You are on page 1of 187

Mate Kapović

with emphasis on accentuation

doktorski rad

Mentor: prof. dr. sc. Ranko Matasović

ZADAR, 2006.


It seems that the topic on the reconstruction and accentuation of personal pronouns
in Balto-Slavic was meant to be – or, to be less pathetic, it was always around. As early as in
the course of my undergraduate study, personal pronouns were a nice topic to deal with – in
that time I wrote a paper titled Personal Pronouns from Indo-European to Croatian. However,
back then I could not even imagine that this would indeed one day be the topic of my PhD
dissertation. In spite of all that, the topic itself did not come about that easily. After a lot of
thinking, the idea came to me one sunny summer day in Göreme (Cappadocia, Turkey) when
my girlfriend Kristina said: "Why not pronouns?" Indeed, I said, why not? So it was settled –
and in the train back to Croatia, I jotted down the concept of the future dissertation. The
major part of the thesis was written during my scholarship in Vienna in the spring of 2006.

And now the acknowledgments. My mentor Ranko Matasović read the manuscript on
more than one occasion. I am grateful to him for numerous useful comments and his help in
general. The first draft of the Slavic part of the dissertation was read by Thomas Olander;
Siniša Habijanec discussed Slovak and Polish personal pronouns with me (shortly but
usefully), and Matej Šekli helped me with a few details concerning Slovene.

Special thanks go to Kristina Marenić and Marko Kapović, who have read the English
version of the dissertation with great patience.

And finally, I would like to thank my mother from whom I learned (pretty archaic)
Croatian accentuation, without which my pursuit of Balto-Slavic accentuation would be much
more difficult than it is.

Zagreb, October 24th 2006.


> yields regularly

⇒ yields irregularly (by analogy etc.)
~ related to
* - reconstructed
** - impossible, not attested; internally reconstructed form

A. – accusative
a. p. – accentual paradigm
G. – genitive
D. - dative
dial. – dialectal
I. - instrumental
L. – locative
N. - nominative
pl. – plural
sg. – singular
p. c. – personal communication

Croat. - Croatian
IE – Indo-European
Lith. – Lithuanian
Latv. - Latvian
OCS – Old Church Slavic
Sln. - Slovene
PIE – Proto-Indo-European

Accentual marks

Common Slavic
* ʀ - place of the stress only
* ɾ - old acute
* ʂ - short circumflex
* ʄ - long circumflex
*` - short neo-acute
* ɶ – long neo-acute

[*eʖ represents *ĕ under the stress]

* ` - short accent
* ɶ – circumflex
*´ - acute

*´ - place of the stress


In his famous monograph “Slavonic accentuation”, Christian Stang opens the very
short chapter dealing with the accentuation of the pronouns with the sentence: “These [i.e.
pronouns, M. K.] present a picture of considerable diversity, and it seems impossible to
arrive at any clear definition of the original state” (Stang 1957: 105). This is a good example
of how the accentuation of the pronouns has been treated in most of the works concerned
with (Balto-)Slavic accentuation. It was either completely left out or, in the best case, covered
unsystematically and not in so many details as it should have been. This is one of the things
we shall try to amend in this work, at least in the case of personal pronouns.
Historical Balto-Slavic linguistics is one of the liveliest fields in Indo-European
linguistics today. In spite of the great progress achieved in the last fifty years, there are still
many unsolved questions to be tackled – starting with the problems of phonology and
prosody to morphology and syntax. As already said, much has also been done in the last
fifty years in the delicate field of Balto-Slavic accentuation – a very complicated subject that
continuously attracts the attention of numerous linguists from all around the world. A
revolution in the field of Balto-Slavic accentology was caused by the work of several noted
linguists – a Norwegian linguist Stang (1957) who suggested a new simplified approach to
Slavic accentuation by introducing the concept of the accentual paradigms, a Russian linguist
Иллич-Свитыч (1963) who derived Balto-Slavic nominal accentual paradigms from the Indo-
European proto-language and a number of other linguists like Дыбо (cf. for instance his
monograph from 1981).
Many questions have been solved in their work, but still many problems remain to be
examined. One such problem is the reconstruction of the Balto-Slavic1 accentual system of
personal pronouns and the reconstruction of Balto-Slavic personal pronouns in general. To
my knowledge, this topic has never been discussed in detail up until now – neither from a
phonological and prosodic, nor from a morphological aspect. Thus, it is a fruitful topic to
write about, even more so because the research on Balto-Slavic personal pronouns can
provide us with important insights useful not only in the reconstruction of Balto-Slavic but in
the reconstruction of the Indo-European proto-language as well. Aside from that, the
application of modern accentology on the pronominal system should yield interesting results
if we have in mind the fact that the pronouns very often differ from other declinable words
not only morphologically, but sometimes even phonotactically (as in Proto-Indo-European)
and, in the case of Balto-Slavic, accentologically.
Simply put, this work will deal with the reconstruction of Balto-Slavic personal
pronouns. The reconstruction itself will be discussed exhaustively, as well as all general
problems of Balto-Slavic and Indo-European linguistics related to this subject. Thus, we shall
cover a time span from the modern Slavic and Baltic languages and dialects to the Proto-

1 Some linguists do not believe in the existence of Proto-Balto-Slavic language even today. I find that
very hard to understand. It is impossible to account for numerous features shared by West and East
Baltic and Slavic if one does not assume the existence of a common proto-language. The accentuation
is very important here. Cf. for instance Croatian dial. N. sg. zīmaʂ, G. sg. zimeɶ, D. sg. zîmi, A. sg. zîmu,
N. pl. zîme and Lithuanian N. sg. žiemà, G. sg. žiemõs, D. sg. žieɶmai, A. sg. žieɶmą, N. pl. žieɶmos. This
kind of agreement in accentuation cannot be coincidental or the result of parallel development.

Indo-European times. The topic of this work are personal pronouns and personal pronouns
alone – or, to put it more precisely, the personal pronouns of the 1st and 2nd person singular,
dual and plural in modern Balto-Slavic languages, in Common Slavic2, Proto-Balto-Slavic
and Proto-Indo-European. We shall not handle the pronouns of the 3rd person. These are not
in fact real personal pronouns in many older Indo-European languages3 (and in Proto-Indo-
European)4, but rather demonstrative or similar pronouns. The reflexive pronoun “self”,
morphologically related to personal pronouns, will also not be dealt with. This pronoun is,
however, in Balto-Slavic and Proto-Indo-European generally identical to the personal
pronoun of the 2nd person singular, except for the difference of the initial consonant, i.e. s-
instead of a t- in Balto-Slavic, or a *t(w)- in Indo-European. Thus, since the reflexive
pronoun generally has, as already said, the same oblique case forms as the 2nd person
singular personal pronoun, we shall sometimes, when necessary, mention and consider this
form as well.
This work is structured as follows: the body of the text is divided into four large parts
– a Slavic part, a Baltic part, a Balto-Slavic part and an Indo-European part. In each of these
sections, the subject of the personal pronouns will be examined within their limits. In the
Slavic section, we shall deal with the reconstruction of the Proto-Slavic system of personal
pronouns. In the Baltic section, we shall discuss the problems concerning the Baltic personal
pronouns. The Proto-Baltic system of personal pronouns will not, however, be reconstructed,
simply because of the fact that it seems that the supposed Proto-Baltic language has never
actually existed5 and that all we can really get by reconstruction starting from Lithuanian,
Latvian and Old Prussian is the Proto-Balto-Slavic system. This fact is especially clear when
looking at the personal pronouns, in which we find, many agreements of Old Prussian and

2 I shall call conventionally the reconstructed Slavic forms Common Slavic, rather than Proto-Slavic (the
terms are sometimes used invariantly if the strict distinction is not necessary in a given situation).The
reconstruction of Common Slavic (or Late Proto-Slavic) is traditional, i.e. only a formula. The reason for
this lies in the fact that Slavic languages were already clearly differentiated in the period for which the
accentual system can be reconstructed directly. However, it is not possible to reconstruct directly the
accentual system of Proto-Slavic, the real proto-language from which all present-day Slavic languages
can be derived with all their distinctive traits. The language here referred to as Proto-Slavic was the
homogeneous language spoken sometime around year 600. The language here referred to as Common
Slavic was a heterogeneous language spoken in the period between the 7th and the 11th century (cf. for
instance Holzer 1995). During this period, some pan-Slavic changes still occurred (together with the
changes typical for separate Slavic languages only). After the 11th century, there were no more pan-
Slavic changes (and thus no more Common Slavic).
3 Cf. Latin ille, used instead of a 3rd person personal pronoun, which can also be used as an attribute to
a noun (ille homo “that man”). This is not possible in the case of real personal pronouns (cf. Croatian
ôn, where **ôn čòvjek is not possible). In Classical Greek for instance, instead of the 3rd person
personal pronoun, in the nominative demonstrative pronouns ουʅτο̋, ǫʆκǫιɶɶνο̋ are used, and in the
oblique cases mostly the pronoun αυʆτó̋ “same”. In Sanskrit, demonstrative pronouns sas, ayam are
used instead of the personal pronouns etc.
4 Since there was no 3rd person personal pronouns in Proto-Indo-European, this means that later 3rd
person personal pronouns in separate IE languages are not etymologically related, cf. Croatian ôn,
Latin ille and English he. This also goes for Balto-Slavic and Slavic. One cannot reconstruct anything on
the basis of completely different forms like Croatian ôn, Macedonian тоj and Lithuanian jìs. All these
forms are unrelated and used only secondarily as 3rd person personal pronouns.
5 Cf. Иванов & Toпоров 1958.

Slavic and many disagreements between Old Prussian and the Eastern Baltic languages. In
the Balto-Slavic section, we shall start with the already reconstructed Proto-Slavic system
and the systemized Baltic data and reconstruct the Balto-Slavic system of personal
pronouns. In the last section, Balto-Slavic personal pronouns will be derived from the Indo-
European system of personal pronouns. There, various problems concerning the Indo-
European reconstruction of personal pronouns will be covered, especially when related to the
Balto-Slavic data. We shall try to use the Balto-Slavic data in the best possible way in order
to reconstruct the Indo-European system of personal pronouns. As we shall see, a careful
examination of Balto-Slavic personal pronouns can be of great help in the reconstruction of
the Indo-European system. Because of the laid out plan of the discussion, it is clear that the
problems concerning specific forms will be discussed in more than one place in the text.
Thus, for instance, the problem of the form of the 1st person nominative singular personal
pronoun will be handled separately in each section respectively – in the Slavic section
concerning the changes which belong to Slavic languages and Slavic period, in the Baltic
section when talking about the forms attested in Baltic and problems related to those forms,
in the Balto-Slavic section when talking about the changes and processes of that period, and
in the Indo-European section when talking generally about the reconstruction of the Indo-
European form. It is quite clear that in this kind of work, a comparative work, it is sometimes
impossible, to stick only to Slavic data when discussing some problem in the Slavic section.
Thus, we shall not try to exclude automatically all the data which does not strictly belong to
the section concerned. We we’ll simply try to limit such data to the level of necessity and to
deal mainly with the problems that do belong to a certain section.


§1 Here is an overview of the personal pronouns paradigms. We have included only

standard (literary) languages. Dialectal data will be presented in the discussion.
Upper Lusatian is cited by Šewc 1968, Lower Lusatian by Šẃela 1952 and Janaš 1976,
Kashubian by Lorentz 1925, 1962 and 1971, Slovincian by Lorentz 1903, and Polabian by
Schleicher 1871 and Ceлищев 1941.



N. azъ ty
G. mene tebe
D. mьnĕ/mъnĕ, mi tebĕ, ti
A. mę tę
L. mьnĕ/mъnĕ tebĕ
I. mъnojNj tobojNj

n. my vy
gl. nasъ vasъ
d. namъ, ny vamъ, vy
a. ny vy
i. nami vami

N. vĕ vy/va
GL. naju vaju
DI. nama vama, va
A. na va


N. àз тиɳ
D. на мèне, ми на тèбе, ти
A/obl. мèне, мe тèбе, те

n. ниɳe виɳe
d. на нàc, ни (arch. наɳм) на вàc, ви (arch. ваɳм)

a/obl. нàc, ни вàc, ви


N. jac ти
D. мeне ми, ми тeбе ти, ти
A/obl. мeне мe, мe тeбе те, те

n. ниe виe
d. нам ни, ни вам ви, ви
a/obl. нac нè, нè вac вe, вe


N. jâ tî
GA. meʂne, me teʂbe, te
D. meʂni, mi teʂbi, ti
L. meʂni teʂbi
I. mnôm, mnóme toʂʂbōm

n. mî vî
ga. nâs, nas vâs, vas
d. naʂma, nam vaʂma, vam
li. naʂma vaʂma


N. jâ тиɵ
GA. мèнe, мe тèбe, тe
D. мèни, ми тèби, ти
L. мèни тèби
I. мнôм, мнóмe тoʂʂбōм

n. мî вî
ga. нâs, нac вâs, вac
d. нaʂмa, нaм вaʂмa, вaм
li. нaʂмa вaʂмa


N. jàz tî
G. méne, me tébe, te
D. méni, mi tébi, ti
A. méne, me tébe, te
L. méni tébi
I. menóʜj, mâno tebóʜj, tâbo

n. mî, mêʜ vî, vêʜ
ga. nàs, nas vàs, vas
d. nàm, nam vàm, vam
l. nàs vàs
i. nâmi vâmi

N. mîdva, mîdve/mêʜdve vîdva, vîdve/vêʜdve
GA. nâju, naju vâju, vaju
D. nâma, nama vâma, vama
L. nâju/nâma vâju/vâma
I. nâma vâma


N. já ty
G. mne tebe
D. mnĕ, mi tobĕ, ti
A. mne, mĕ tebe, tĕ
L. mnĕ tobĕ
I. mnou tebou

n. my vy
gal. nás vás
d. nám vám
i. námi vámi


N. ja ty
GA. mňa, ma teba, ťa
D. mne, mi tebe, ti
L. mne tebe
I. mnou tebou

n. my vy
gal. nás vás
d. nám vám
i. nami vami


N. ja ty
GA. mnje, mje tebje, će
D. mni, mi tebi, ći
L. mni tebi
I. mnu tobu

n. my wy
gal. nas was
d. nam wam
i. nami wami

N. mój wój
GA. naju waju
DLI. namaj wamaj


N. ja ty
G. mnjo, mje tebje, tebjo, śi
D. mnje, mnjo, mě tebje, śi
A. mnjo, mě tebje, tebjo, śi
L. mnje, mnjo tebje
I. mnu tobu

n. my wy
gal. nas was
d. nam wam
i. nami wami

N. mej wej
GA. naju waju
DLI. nama wama


N. ja ty
GA. mnie, mię ciebie, cię
D. mnie, mi tobie, ci
L. mnie tobie
I. mną tobą

n. my wy
gal. nas was
d. nam wam
i. nami wami


N. iʠω tǩ
G. mɴe cebɴe, ce
D. mɴe tobɴe, cǩ, tǩ, cï
A. mɴe, mą cebɴe, cą, ce, cebɴą
L. mɴe cebɴe, tobɴe, cebɴi, ce
I. mnNj, mɴNj tobNj

n. mǩ vǩ
gla. nas, nωs vas, vωs
d. nNjm, nama, namï vNjm, vamï
i. namï, nam÷, nama, namǩ vamï, vam÷, vamǩ


N. ma, maiʠuɺ, naiʠuɺ va
mɴe, mǩ, ma vɴe, vǩ, va
GLA. naiʠuɺ, naiʠï vaiʠuɺ, vaiʠï, vaiʠix
DI. nama vama, vNjma


N. jǛu, jå taɺ, tä
G. mjìeʠ, mjä cìeʠbjä, cä
D. mjìeʠ, mjä tșɳȅʠbjä, cä, tä
A. mjìeʠ, mją, mjä cìeʠbjä, cą, cä
L. mjìeʠ, mjä cìeʠbjä, cä
I. mnoɺųʠ tȅboɺųʠ

n. maɺ, mä vaɺ, vä
gl. nãs vãs
d. noɺuʠm, noɺuʠm  ʠm, voɺuʠm
a. nãs, năs vãs, văs
i. nąɶmï nąɶmï

N. mã, ma vã, va
mjìeʠ, mjä vjìeʠ, vjä
GLA. nãjșɺ vãjșɺ
DI. nąɶmƮ vąɶmƮ


N. joz, jo, jaz, ja toiʠ
G. miné, mané, mánǩ, mán tíbǩ, tibé
D. miné, mané, mánǩ, mán tíbǩ, tibé, tĭ, t
A. mą, miné, mané, mánǩ, tíbǩ, tibé, tą, tƮ
mán, mƮ
I. månNjɺ tåbNjɺ

n. moiʠ jaiʠ
g. nos, nas
d. nom, nam, nǩm vom
i. nómǩ nomǩ, vom


N. яɴ тыɴ
GA. меняɴ тебяɴ
DL. мнé тебé
I. мнóй тобóй

n. мыɴ выɴ
gal. нác вác
d. нáм вáм
i. нáми вáми


N. яɴ тиɴ
GA. менé тебé
DL. мeнí тoбí
I. мнóю тобóю

n. миɴ виɴ
gal. нác вác
d. нáм вáм
i. нáми вáми


N. яɴ тыɴ
GA. мянé цябé
DL. мнé тaбé
I. мнóй , мнóю тaбóй, тaбóю

n. мыɴ выɴ
gal. нác вác
d. нáм вáм
i. нáмi вáмi

Personal pronouns in singular

In this section, we shall talk about the 1st and 2nd person singular personal pronouns.
First, we shall talk about the form *ja(zъ) “I”, which demands special attention and thus gets
a separate section, and then about the oblique cases of the 1st person singular and all the
cases of the 2nd person singular. The general idea is to look at the material of all the
languages first, together with the available historical and dialectological data and to explain
the changes that have occurred in specific languages, and then try to reconstruct Common
Slavic forms and paradigms while discussing specific problems concerned with the
reconstruction and Common Slavic forms.

Slavic *ja(zъ)

Here, after a survey of the material, we shall discuss phonological and prosodic
issues of the Proto-Slavic form *ja(zъ), i.e. the Proto-Slavic nominative singular of the 1st
person personal pronoun. It is quite strange to see that so many problems concerning this
short word exist. The problems that will be dealt with here are the vocalism of Slavic form
(*-a- while other Indo-European languages point to *-e-, cf. Latin ego), the disappearance
of the prothetic *j- in Old Church Slavic azъ, the variants *jazъ and *ja, and the complex
problem of the accentuation of this form.


First we shall adduce the forms for “I” in Slavic literary languages: OCS azъ, Bulgarian
aз, Macedonian jac, Croatian jaʄ, Serbian jaʄ, Slovene jàz, Czech já, Slovak ja, Lusatian ja,
Polabian jo, joz, Polish ja, Kashubian iʠω, Slovincian jǛu, Russian/Ukrainian/Byelorussian яɴ.

Old Church Slavic

§2 OCS shows a form without the initial j- (azъ), unlike all other Slavic languages except
Bulgarian. This is usually the only form mentioned in OCS grammars6; however, jazъ is a
hapax in Codex Marianus7, which could be due to the influence of the dialect area in which
the text was written8.
In Croatian and Russian Church Slavonic, there is the same form as in OCS (CCS
azъ/azь, RCS azъ)9. In Psalterium Sinaiticum (Ps. 38, 13), the OCS form a is found. Diels
(1932-4: 214) and Nandriş (1959: 105) take this form to be a mistake, but it probably
represents the OCS j-less pendant of the form ja in other Slavic languages10.

6 Cf. Vondrák 1912: 459, Leskien 1922: 109, Rosenkranz 1955: 96, Bielefeldt 1961: 146-147,
Горшков 1963: 135, Trubetzkoy 1968: 150, Kurz 1969: 76-77, Hamm 1970: 133-134, Lunt 1974: 65,
Damjanović 2005: 95 etc.
7 SP, Diels 1932-4: 77, Słoński 1950: 84, Nandriş 1959: 104. According to ðorñić (1975: 105), jazъ
occurs number of times after the conjunction i – i ězъ (in glagolitic script), i jazъ (in cyrillic script). In
fact, even the occurrence of jazъ (i.e. ězъ) in Codex Marianus is in this position: vъprošNj i ězъ (Mar.
162, 6-7). Weingart (1937-8: 200) considers this to be a sandhi variant of azъ in a hiatus.
8 The origin of Codex Marianus is disputed. According to Jagić (1883), judging by the sporadic changes
Nj > u, y > i, vъ > u, ę > e, the manuscript could have been written in Croatian or Serbian speaking area
(i.e. Štokavian, judging by vъ > u). Hamm thinks that the vocalization of the yers (ъ > o, ь > e) and the
loss of epenthetic l could point to Macedonian origin (Damjanović 2005: 18). Judging by jazъ it could
be both (Croat. jaʄ, Maced. jac).
9 According to Mihaljević (forthcoming), in Croatian glagolitic fragments from 12th and 13th century
only the forms azъ/azь/az appear. There is no attestation of native ja or even jaz in these texts.
10 This would not be the sole occurrence of an unusual feature in Psalterium Sinaiticum. This text is for
instance the only OCS source which preserves the *s of Proto-Slavic *u(s)tro “morning” in the adverb
zaustra, which had disappeared elsewhere.


§3 The Bulgarian form is also j-less, in accord with OCS. Dialectally, besides aз, one also
finds the forms aзека, aзекана, àзкана, я, яɳзе, яɳзка, язекана (ESSJ), ac, йa11, яз (Vasmer),
язека, яcка, aзи, aзe12 etc. The Bulgarian j-less form aз must be of the same dialectal origin
as OCS azъ. For the attestation of яз, я and aз in Bulgarian dialects, cf. БДA I: 160. The form
aз is attested in southeastern Bulgaria, the form я is attested not only in the west of Bulgaria,
but also in the southeast13. The form я could be attributed to the Serbian influence in some


The Macedonian literary form jac shows an initial j-, thus differing from OCS and
literary Bulgarian. In dialects, one also finds ja (SP) – for instance in North-West Macedonia,
jáska (Małecki 1934), ac, acкa, jaзека (ESSJ) etc.


§4 The Croatian literary form is jaʄ (nì jā, ì jā with a Neo-Štokavian retraction of accent).
Most dialects have a z-less form, except for some of the Northern Čakavian dialects and a
few Kajkavian dialects (in Štokavian, there are no forms with the final –z anywhere). Two
types of accentuation exist – the most common variant with the neo-acute ( ɶ), or with ʄ < ɶ
in the dialects that do not preserve ɶ, and the rarer variant with the short falling accent ( ʂ).
Again, the latter is non-existent in Štokavian dialects; it occurs marginally in Kajkavian and
regularly in North Čakavian. In Štokavian for instance: Posavina jã15, Neo-Štokavian jaʄ

In Čakavian16, as already said – there is a neo-acute in the South and in some Central
Čakavian dialects. Cf. in South Čakavian: Blato (Korčula) jã (personal data), Hvar jõ, Vrgada
j°ã (Jurišić 1973), Murter jã (personal data) etc. In Central Čakavian: Rivanj jaʄʘ (no neo-acute,
Radulić 2002), Pag j°aʄ (no neo-acute, Kustić 2002), Senj jã (Moguš 1966: 78) etc.
In some of the Central Čakavian dialects and in North Čakavian, there is a short
falling accent here (and final –z in some dialects). In Central Čakavian: Novi jaʂ (Белић 1909:
199), Silba jaʂz (Bezlaj) etc. In North Čakavian: Istria jaʂ (but also jaʂs, jaʄ) (Jurišić 1973), Cres
(town) jaʂ (Tentor 1909: 172, 1950: 75), Bejska Tramuntana (Cres) jaʂ (Velčić 2003), Orlec
(Cres) jaʂ (Houtzagers 1985), Tometići (near Kastav) jaʂs (but only when by itself, Skok), Kastav
jaʂz (Bezlaj), In Istria, the form jaz was attested already in 1454 (Bezlaj).

11 Стойков 1993: 90, 250.

12 Mирчев 1963: 163.
13 SP, Xapaламниев 2001: 108.
14 Бернщейн 1948: 325-326.
15 Ivšić 1913, II: 34 or 1971: [373].
16 For the grouping of Čakavian, cf. for instance Brozović & Ivić 1988: 87-88.

In Kajkavian, the form with the neo-acute (or long falling if ʄ< ɶ in the final syllable
or in general) and with no –z is the most frequent by far: Samobor jaʄ, jã sem (no neo-acute
in the final syllable, Šojat 1973c: 53), Ozalj jaʄ (no neo-acute there, Težak 1981), Turopolje jaʄ
(no neo-acute in the final syllable, Šojat 1982), Repušnica, Varaždin jaʄʖ (no neo-acute in the
final syllable, Brlobaš 1999, Lipljin), Cerje jaʄ (no neo-acute in the final syllable, Šojat 1973b),
Brdovec (today Kajkavian, but genetically Čakavian) jã/jaʄ (Šojat 1973a: 42). However, in
Bednja (Jedvaj 1956), we find joʂz (older and more frequent according to Jedvaj), joʂ (younger).
A similar thing is found in Gregurovec Veternički: jaʂs and jaʄ (Jembrih & Lončarić 1982-3:

Until the beginning of the 17th century, jaz was common in Kajkavian texts - for
instance in Pergošić’s language. In his Decretum from 1574, jaz occurs only four times
(pages 5, 6, 147, 187), due to the legal nature of the text. Out of these four times, jaz was
written 1x as <iaz>, 2x as <iaaz> and 1x as <iâz> which strongly suggests that it was
indeed jãz (or jaʄz). According to Šojat (1970: 89), the form ja became the main one from the
beginning of the 17th century under the influence of the spoken dialects. However, already in
Vramec’s Postilla from 1586, apparently only ja appears (for instance, on pages 8, 9, 12 –
3x, 13 – 3x, 14 – 3x, 180 etc.). It is interesting that Vramec always writes <ia> for ja, which
might point to jaʂ rather than jã/jaʄ, considering that Vramec often indicates the length by
doubling the vowel18. However, since the doubling of the vowel is inconsistent, one cannot
be certain. According to Šojat (1970: 89) and Junković (1972: 125), in the 16th century works
of Antun Vramec one finds both ja and jaz. I was not able to verify that on the material of
Postilla, as already said. The form jaz is also found, for instance, in Pavlinski zbornik from
the year 1644 (Šojat 1992: 26).

In Štokavian, the form ja is attested from the earliest times, for instance in Povelja
Kulina bana from the year 1189 (ARj IV: 378) and elsewhere. Petar Budmani in ARj claims
that the form jaz (jazь), as well as az/azь, is not native but taken from OCS. He provides two
Štokavian examples of jazь from 1186 and 1198-9. However, it is not clear if these
examples are undoubtedly OCS loanwords (as azь certainly is)19. Budmani also considers the
Kajkavian example jas (or jaz?) from the year 1587 a Slovene loanword, but that seems an
unnecessary assumption.

17 The authors dub the form jaʂs older and the variant jaʄ younger. The form jaʄ is more frequent,
especially if before a word beginning with a consonant, while jaʂs is used before words beginning with a
vowel (e. g. jaʂs oʂnda veĜĩm, jaʂs ıʂdeʚm). It is interesting to note the long falling accent in jaʄ, in spite of
the fact that this dialect preserves the neo-acute in all positions, cf. bĩk, põt, stõp, strĩc etc. (Jembrih &
Lončarić 1982-3: 32). That could perhaps indicate that we are dealing with a loanword from another
Kajkavian dialect here.
18 For instance: znaam “I know”, kluucz “key”, kraal “king”, daal “gave”, deen “day”, deel “part”, zaam
“alone” etc. However, this is not consistent, cf.glaz “voice”.
19 For ja/jaz in Croatian and Serbian old texts, cf. also Даничић 1874: 215.

Čakavian dialects mentioned above


§5 The Slovene literary form is jàz. Snoj also adduces the younger variant jaʄz20. Cf. iaz
already in the Freising Monuments and the forms Jas, ya, ye in the 16th century texts of the
protestant writer Krelj (Bezlaj). In Središče, both jaz and ja exist21. Pleteršnik adduces jaʄ for
Eastern Styria and Bela krajina22, cf. also jà for Bela Krajina (and elsewhere) in Bezlaj. In
dialects one also finds jèst (< *jàst) in Cerkno (FO 1981: 70), jǫʄst in Hrušica (FO 1981: 114)
etc. Forms with final –st appear already from the 15th century23 (Bezlaj).
According to de Courtenay (1929: 228), the Slovene dialect of Rezija distinguished
forms jaz, used by males, and ja, used by females. However, Steenwijk (1992: 119) claims
that the forms are just free variants and not gender-related, the form jas being more


§6 The literary form in Czech is já. In Old Czech, there was also jáz. The form já occurs
from the beginning of the 14th century25. Since there are practically no earlier attestations of
this pronoun at all, this means that both forms were used from the very beginning of Old
Czech literary tradition. The form jáz was used in Czech as up until the end of the 15th
As for some other attestations of Old Czech jáz and já, cf. for instance Dalimila,
which dates from sometime between 1308/1310 and 1314, where jáz is found only eight
times (each time at the beginning of the sentence) and já 27 times; in the Old Czech text
Katonova dvojverší (found in 6 manuscripts – one from the second half of 14th century and 5
from 15th century), we find both jáz (“jáz pomyslil”) and já (“naučím já tě”) (58b); in Závišova
píseň (Jižť mne vše radost ostává) from the end of 14th century one finds only já five times;
já is found also in the 14th century Czech translation of the Bible (SBDO, e. g. page 293) and
in the 14th century epic Vévoda Arnošt etc.


20 This accent is found in Snoj only. The change jàz > jaʄz is due to the very recent sporadic
lengthening of the final short vowels in Slovene. This change is the part of a tendency to eliminate the
quantitative oppositions in Slovene. Cf. also bàt > baʄt, màk > maʄk, kràs > kraʄs etc. (Šekli 2003: 33).
21 Rigler 2001: 361.
22 The attestation of jaʄ in Bela krajina is not so important since that dialect is in fact genetically
Croatian, not Slovene.
23 [s] yields [st] by fortition. Cf. Ramovš 1924: 222. This development can be typologically compared to
German Karst as opposed to Latin carsus, Čakavian kraʂs/kraʄs, Slovene kraʄs (a substrate word
24 In Bezlaj, the forms jás, jàs and jà are adduced for Rezija.
25 It occurs, for instance, already in Alexandreis, which could in fact stem not from the beginning of the
14th, but from the end of the 13th century (each form, jáz and já, occurs seven times in the text, see §
26 Gebauer 1896: 524, Gebauer (dictionary), Trávníček 1935: 335 etc.

§7 The Slovak literary form is ja. Since the real beginnings of the Slovak literary
language are not older than the end of the 18th century, it is hardly surprising that there is
no attestation of final –z in Slovak27. The fact that the modern Slovak dialects do not show
the final –z anywhere is in accord with the same fact in modern Czech. If Slovak were
attested earlier (not just in traces in Czech texts), we would probably also find the final –z
there, as in Old Czech (cf. however Old Polish with only one doubtful attestation of the z-
form, see § 8). The Slovak literary form ja differs from the Czech form já in length only on
the surface. The length was originally there in Slovak as well (as it still is in dialects), but one
cannot see it in Central Slovak because there já > ja regularly (cf. Czech voják, Croatian
vòjāk but Slovak vojak). Thus, Slovak ja stems from an earlier form já still attested in
dialects. Older form still, also attested in dialects, is jaɾ28. The form jaɾ is attested in Central
and Eastern Gemer (North-West Slovakia) and Lower Orava (South-East Slovakia), in Central
Slovak dialects which had no diphthongization aɾ > iʠa, and in Western Liptova. The form já is
attested in the South-West of Slovakia and in Central Orava (where aɾ > á), cf. also SSN,
Orlovský (Gemer), HSSJ. In Central Slovak, in North West Slovak and in East Slovak, the form
ja occurs29. In Central Slovak, this form is explained via the diphthongization of aɾ > iʠa
according to which the old jaɾ yields jiʠa, which then results with a form ja, due to the change
of jiʠ > j. In East Slovak, the form ja is due to shortening (jaɾ > já > ja). In East Novohrad, aɾ >
eiʠ (thus the forms jeiʠ and eiʠ)30.


§8 In Lusatian, only the form ja is attested, including older texts and dialects.


In Polabian, one finds jo and joz when accented, and ja and jaz when unaccented31.
According to Schleicher (1871: 259), the forms joz, jo are more frequent. The form jo occurs
in front of verbs, for instance in jo jis < *ja jesmь.


The modern Polish literary form is ja. However, in Polish dialects which preserve å <
*ā, one finds the form jå with the attested length (i.e. neo-acute), as in Czech. Cf. also
Kashubian iʠω (ω < *ā) and Slovincian jǛu (SP). In Old Polish, the form jaz is attested only as a
hapax in 15th century in Psałterz floriański 108, 332. However, according to some authors33,
this is in fact a Czech loanword. The shorter form ja is attested in Polish already in 13th

27 Stanislav 1958: 290, Pauliny 1990: 150.

28 In Slovak, *ā after soft consonant yields aɾ which then yields iá > ia.
29 Pauliny 1990, ibid., Stanislav 1958, ibid. For já, cf. for instance Ripka 1975: 146.
30 Pauliny 1990, ibid.
31 Cf. Lehr-Spławiński 1929: 185, Ceлищев 1941: 440, PED.
32 Łoś 1927/III: 159, SS, SP.
33 Klemensiewicz, Lehr-Spławiński & Urbańczyk 1955: 321.

East Slavic

§9 Modern East Slavic languages all have the form яɴ. This form appears already in Old
Russian, as early as the year 1130 in Мьстиславова грамота, together with the longer form
язъ and the OCS form aзъ, without any noticeable difference between these three forms – “ce
aзъ Mьстилавъ”, “a язъ далъ”, “a ce я Bceволод”34. Kiparsky (1967: 130-131) says that this
only supports the conclusion that the forms я and яз were used in Russian for a long period
without any significant stylistic difference. This is confirmed by Гадолина (1963) who gives
clear examples of how both я and яз in Russian occurred in the emphasized and in the non-
emphasized position. According to Черных (1962: 218), яз was still a very usual form in the
first half of the 16th century in Moscow (although it was not the only existing form).
According to Черных, both forms co-existed in Russian until the 17th century. However,
Cocron (1962: 133) and Гадолина (1963: 26) claim that the form я was the only possible
form already in the 17th century. Interestingly, the forms яз and яз-to were found in былина
songs up to the 19th century (Черных, ibid.).
In Ukrainian, the form яɴ is attested from 1322 and язъ from 1341 (CCM).

34 For attestations of these forms in Old Russian, cf. Cpeзнeвcкий, CДЯ and CДЯ 2.


1. Standard languages:

Only z-forms: Bulgarian, Macedonian, Slovene

Only z-less forms: Croatian, Serbian, Slovak, Czech, Polish, Upper Lusatian, Lower Lusatian,
Russian, Ukrainian, Byelorussian

2. Modern dialects:

Only z-forms: -

Only z-less forms: Štokavian, Czech, Slovak, Polish, Kashubian, Slovincian, Upper Lusatian,
Lower Lusatian, Russian, Ukrainian, Byelorussian

Z- and z-less forms both: Bulgarian, Macedonian, Čakavian, Kajkavian, Slovene

3. Historical attestations35:

Only z-forms: -

Only z-less forms: Štokavian (?), Slovak, Upper Lusatian, Lower Lusatian

Z-less and z-forms in the past: Bulgarian, Čakavian, Kajkavian, Slovene, Czech, Polish (?),
Polabian, Russian

35 For some of the languages, like Kashubian and Slovincian, there is almost no historical data. In the
case of East Slavic languages, the histories of Russian, Byelorussian and Ukrainian are not easily
discernable. The same is true of Bulgarian and Macedonian and Czech and Slovak.

*ja- in *ja(zъ) and the dropping of the *j-
*j- in OCS and Bulgarian

§ 10 The question of the vocalism in the 1st person singular personal pronoun has been a
subject of many disputes since the very beginnings of Indo-European linguistics. In order to
explain Slavic *a, instead of *e, found in other Indo-European languages (Latin ego, Greek
ǫʆγώ), various theories have been proposed. For an overview of various explanations, cf. for
instance Tpyбачев, where a number of different explanations, serious and less serious, are
Today however, at least the length problem can be more or less easily explained by
Winter’s Law, which accounts for the long vowel. Thus, we start with the usual Indo-
European form *egɴ-37. In Balto-Slavic, we get a long vowel regularly via Winter’s Law, thus
presumably the form *ēź-38. The problem of the Slavic *a vowel (instead of *ě), however, still
At first glance, the easiest way to solve this problem is to explain this *a by the
known development *ě- > *jě- > *ja- at the beginning of the word in Slavic. This is a known
and mostly unproblematic development39 and this way of deriving the initial ja- in this
pronoun is very frequent in the handbooks40. However, should we assume that Slavic jazъ
stems from the older *ězъ, we should also assume that Old Church Slavic azъ is a secondary
development via the change *ja- > a-. However, this development, although accepted
without a question by many41, is not at all simple.
Соболевский (1908) has stated that the occurrence of a- and ja- in OCS (and in
Russian Church Slavic) is attested only in the words that etymologically have either *ā- or
*jā- and never in the words that etymologically have *ě- or *jě- at the beginning. Thus we
find in OCS the forms aky, amo, ablъko, aviti, agnьcь etc. (where the initial a- stems from
*(j)ā- according to the Russian linguist), but only jadro, jazditi, jaxati, jamь, jadъ, jazva etc.
written with either ja- or ě- (where the initial ja- stems from *(j)ě-). Соболевский (1908:
204) concludes that, according to this, the a- in the OCS pronoun azъ cannot be closely
related to the initial e- of the other Indo-European languages. This means that OCS azъ
cannot be derived from an older *ězъ, which would thus be a great obstacle in the derivation
of Slavic jazъ from an older *ězъ42. However, Соболевский is mistaken in one thing – the

36 For a review of older theories, cf. for instance Knutsson 1935 and for a shorter review of some of the
older explanations cf. for instance Skok, Ivšić 1970: 218 etc.
37 For various IE forms of this pronoun, see § 101.
38 For other Balto-Slavic forms and the exact details of the operation of Winter’s Law in this example,
see § 70.
39 Cf. for instance Vondrák 1912: 292-293, Vondrák 1924: 85-87, Vaillant 1950: 180-182, Arumaa
1964: 103, Ivšić 1971: 95-96 etc. Most of the instances where *ě- does not yield ja- in Slavic are to be
explained analogically (for instance, Croatian jeʂsti instead of **jaʂsti by analogy to ìzjesti, pòjesti etc.),
even though it seems that this development was perhaps not absolutely regular. However, the
undisputed result of *ě- in Slavic is ja- in almost all the cases.
40 Cf. for instance Sławski, Brückner, Boryś, Birnbaum & Schaeken 1997: 73.
41 For instance Brückner, Boryś.
42 Соболевский’s argument is accepted for instance by Vondrák (1924: 87).

examples written in OCS with a- or ja- never go back to the old (Proto-Slavic) *ja-43. All
these examples stem from Proto-Slavic *a-, not from Proto-Slavic *ja-. In fact, the
development *ja- > a- is practically non-existent in Slavic as a whole44. All the discrepancies
between (j)a- in OCS/Bulgarian and ja- in all other languages are found in the words in
which Proto-Slavic had *a-, not *ja-. For instance in: *aɾblo “apple”, *avıɾti “notify” (and the
adverb *avě), *aɾgnę “lamb”, *aʄje (*ājьceʀ) “egg”, *aɾbolnъ/ь “poplar”, *aɾgoda “strawberry”,
*aɾlovъ “barren”, *aɾma “hole”, *ārьmъʀ “yoke”, *aɾsenъ/ь “ash-tree”, *asıɾɾka “asp”, *aɾstrębъ/ь
“hawk”, *aɾvorъ/ь “maple” etc. In these examples, for instance, we find the following picture –
OCS Slavic has only the form with initial a- in four examples: ablъko, agnę, ablanь, aice; it
has double forms in three examples, including later, Church Slavic examples: (j)aviti/(j)avě,
(j)agoda, (j)avorь (CS); and it has forms with initial ja- in two examples: jarьmъ and jalovъ
(CS). In Bulgarian, the same tendency (not to have the initial j-) can be noted, but it is not as
widespread as in OCS. There, we find only one form with a- only (àгне), two forms with both
a- and ja- (àвe/яɳвe but only явяɳ in the verb, яйцè/dial. aйцè) and ja- only in all other
forms45. In the rest of the Slavic languages, we only find initial ja- in these examples46. Thus,
in the words beginning with *a- in Proto-Slavic, we find inconsistent a- or ja- in OCS (much
less in Bulgarian) or both, while in all other Slavic languages only ja- is found. The
inconsistency of OCS and Bulgarian is to be explained probably by the late appearance of the
prothetic *j- in front of *a- in Slavic. While this development, *a- > *ja-, was finished
completely in most Slavic languages, it was only a tendency in OCS and Bulgarian, which
resulted in the inconsistency and double forms (the change had not embraced all the
examples at the same time). However, if a word had an initial *ja- in Proto-Slavic, this *ja-
was preserved everywhere, cf. for instance examples *jakъ “which > strong”, *jaʄlъ “jealousy”,
*jaʄrъ “bitter”, *jaʄrъ “heat; wheat etc.”, *jaɾto “flock” etc. The only possible examples of *ja- >
a- are OCS forms like ako, aky, amo (all from the same root) in which a confusion of two
stems has occurred - original *jako and *ako were probably mixed in OCS (j)ako. Thus, the
only possible example is highly disputable an unreliable. Knowing this, we can conclude that
OCS a-/ja- continues only Common Slavic *a-, not *ja-.
Соболевский also notes that the words with the etymological *(j)ā- are never written
with initial ě- in OCS (thus written forms like **ěky or **ěblъko are not attested), which
probably has to do with the fact that the writing of the initial ě- for later [ja-] in Glagolitic
script was not just an orthographical issue in the beginning (as it clearly was in the later
texts). In these cases, yat, or an allophonic variant of it, was indeed pronounced. That
explains why ja- of a non-yat origin was never written like that. In the case of the pronoun
azъ in OCS, we do know that it was sometimes (although rarely) written as ězъ in hiatus (cf. §
2), which might be indicative.
Соболевский claims that there are no examples in which a Proto-Slavic *ě- would
yield a- in OCS. This may be true, but, as we have shown, the change of *ja- > a- is

43 It is an interesting fact that neither Vondrák (1924: 87), who has accepted Соболевский’s argument,
nor Knutsson (1935), who has modified it, did not notice this.
44 Except for a few disputed examples that we shall discuss here.
45 In Middle Bulgarian, however, one finds also aгода, аремникъ besides яремъ and аблъка/яблъка
(Mladenov 1929: 124). All these words have я- in modern Standard Bulgarian.
46 If we ignore a few anomalies, such as Ukranian dial. aйo or Czech vejce, in the word *aʄje, probably to
do with a dissimilation of the two j-sounds in this word.

generally non-existent in Slavic, regardless of the origin of this *ja- (whether it is an original
*ě- or *ja-)47. There is a possible example of *ě- > *ja- > *a- in Middle Bulgarian however -
aдΧщим and aдΧть (besides яcти and ядΧть), and in Bulgarian dial. - da adéš48, if these
examples are old49. But still a problem remains – how are we to explain the development of
the initial *ě- into a- in the 1st person sg. personal pronoun in Old Church Slavic, Bulgarian
and some Macedonian dialects, when it seems that this is the only sure example of this
change50? Actually, only the initial a- of OCS/Bulgarian/Macedonian is problematic here,
since the initial ja- attested elsewhere in Slavic is the expected result of the older *ě-.
Knutsson (1935) tries to offer a solution for this unique development in OCS. He
claims that the change of *ě- > *ja- > *a- has occurred only in azъ in OCS and Bulgarian
because of the supposed special position (or nature) of this particular word. Knutsson says
that he has examined seven OCS texts, where the form azъ occurs about 700 times. Out of
these 700 times, it appears 280 times (about 40%) at the beginning of the sentence and in
25% of times after i or ь. As he says, this statistic does not prove that the initial *j- was
dropped in OCS, but it clearly makes this supposed dropping much more probable. It is not
difficult to imagine that a *j- would be dropped in an absolute anlaut or after i-sounds.
However, his proposition remains a speculation.

Putting the question why and exactly how *j- has disappeared in OCS and Bulgarian
aside, another question remains – if not from *ězъ, how are we supposed to explain the
initial (j)a- in Slavic? If we disregard all entirely speculative and imaginative explanations51,
and if we exclude the possibility of an *o- (i. e. Balto-Slavic *a-) instead of *e- via
Rozwadowski’s Change of Balto-Slavic *e- > *a- because of the relative chronology52, the
only serious explanation which remains is that we posit an ad hoc Indo-European ablaut

47 One could of course claim that the process might have been *a- > *ja- > a- for certain words in OCS
and Bulgarian, but that approach is highly uneconomical and is in fact improvable since no clear cases
of the original *ja- becoming a- exist.
48 Mladenov 1924: 124.
49 Another example by Mladenov, quoted by Knutsson (1935: 99), aднá, aднó (from *edьnaʀ, *edьnoʀ), is
clearly a result of a later development. The same is of course possible for the Middle Bulgarian
examples aдΧщим, aдΧть and dial. da adéš as well. One should also note that the fluctuation of a- and
ja- in Bulgarian occurs in Greek and Turkish loanwords as well (Mladenov 1929: 124).
50 One could assume that there was a difference in the treatment of the original *ē-, which would yield
*ā- and then regularly (j)a- in Slavic, and *ay-, which would yield *ě- and then ja- in Slavic (Ranko
Matasović, p. c.). However, this is very doubtful since both the results of the old *ē- and of the old *ay-
are written as <ě-> in OCS. Moreover, since the number of the examples with the old *ē- seems to be
limited to the words (j)azъ “I”, jasti “eat” and jazъ “channel” (~ jezero “lake”, Lithuanian ežià, ažià), this
appears to be an artificial solution.
51 Cf. a survey of those in Tpyбачев for instance.
52 The material (cf. Andersen 1996) shows that the change must have been rather recent - there are
many double forms and dialectal differences. Winter’s Law, however, was a rather early change, since it
operated before the Balto-Slavic change of *o > *a (cf. PIE *h2ebōl- “apple” > Lith. obuolyɶs, but *nogw-
“naked” > Lith. núogas), and before the deaspiration of PIE aspirated plosives. Thus, *egɴ- first changes
to *ēź- by Winter’s Law, and at the time of Rozwadowski’s Change, there is no short *e- that could be
changed to *a-. Thus, Rozwadowski’s Change cannot be used to explain the vocalism of Slavic “I”.

variant *ogɴ-53. However, if this supposed o-ablaut is not attested anywhere else in Indo-
European (see § 100, § 101), is it reasonable to reconstruct it based only on the Slavic data?
The problem is in the choice of whether it is more economical to assume an ad hoc Indo-
European ablaut variant *ogɴ- or an irregular secondary allegro development of *jazъ (< *ězъ)
> azъ in Old Church Slavic only? The rest of the Slavic, as said, is not problematic. In spite of
the fact that the development *ě- > *ja- > a- (or just *ja- > a-) is otherwise not found in
OCS, in my opinion it is more elegant and economical to assume a secondary, even if not
regular, development of *ě- > *ja- > a- in OCS and some Bulgarian dialects, than it is to
assume a ghost-form *ogɴ- for Indo-European54. We shall return later to the problem of the
IE form (§ 101).

53 This explanation is found not only in the older literature, but also in some of the newer works - cf.
for instance Meillet 1937: 333, but also Andersen 1996: 148.
54 Another possible solution would be to assume an assimilation of early Slavic *ēzan > *āzan (Ranko
Matasović, p. c.), which would then regularly yield Slavic (j)azъ. Although not impossible, one has to
assume an incomplete assimilation (since *ē- > *ā- are long and *-a- is short) and rely on the very
insecure phonetic status of a supposed early Slavic *-an < PIE *-om.

The final *-
*-zъ in Slavic

§ 11 Traditionally55, it has been taken as obvious by most linguists that the form *(j)azъ
with final –z(ъ) is the original one. The form *ja was always considered somehow secondary,
as a result of some sort of secondary dropping of the final –z(ъ). This supposed dropping of
the final –z(ъ) in Slavic has been explained in many ways. Here, we shall mention a few of
those and discuss them briefly.

One of the explanations is that the –z(ъ) was dropped because of the sandhi
positions like jaz sam or jaz znam56 (Solmsen as quoted in ARj for instance). As already
noted very sharply by Budmani in ARj, the case of jaz sam can hardly prove anything, since
the only two modern Slavic languages that have only the shortened s-initial forms of the 1st
person singular present tense of the verb “to be” are exactly the only two modern Slavic
languages which have preserved the final –z (Slovene and Bulgarian). The collocation jaz
znam is more convincing, but it is hard to imagine that this would be so frequent as to cause
the –z to be dropped completely.

An explanation similar to this is the one which states that the supposed shortening of
jaz(ъ) to ja is simply an allegro-process57. This is possible but still ad hoc.

Another explanation is that jaz(ъ) is shortened by analogy to ty (and my and vy)58

because in this way all the nominatives of the personal pronouns are monosyllabic and end
in a vowel. This explanation seems perfectly possible.

Yet another explanation, similar to the previous one, is that ja was created as an
unaccented monosyllabic variant of bisyllabic jazъ so that ja – jazъ would function in the
same way as mi – mъnĕ, mę – mene etc. (Jakobson, cited from Tpyбачев – accepted by
Gluhak for instance).

Vaillant (1958: 443)59 believed that both *ja and *jazъ originate from an older *jaz in
the same way that the original *ot produced the form otъ (with an unetymological final yor)
instead of the expected ot-/o-. Thus, according to this explanation, ja would be a regular,
expected form and jazъ would have an unetymological yor which has risen to stop the
dropping of the final –z, while being in accord with the law of open syllables. However, since
Slavic *-ъ can be derived directly from the attested IE *-om (see § 101), there is no need to
assume that it is a secondary development. More importantly, the case of otъ (and izъ) is not
quite the same as the case of jazъ. In the first case, we have a preposition (used also as a
verbal prefix) and in the second one we have a pronoun form. In the case of ot > otъ and iz
> izъ (cf. Lithuanian at-, dial. and older iž), the reason for the introduction of the

55 Although it is difficult to speak of tradition when so many different theories exist.

56 We shall take the Croatian form as an example here.
57 Cf. for instance Meillet 1934: 452, SP, Boryś.
58 Cf. for instance Гадолина 1963: 14, Иванов 1983: 295.
59 Accepted for instance in Aitzemüller 1991: 107-108, Bańkowski etc.

unetymological –ъ- was to prevent the making of the variants ot- and o-, which would lead
to semantic confusion (cf. the preposition/prefix ob-/o-). Thus, -ъ was introduced to
prevent verbs like *ot-nesti “take away” becoming *o-nesti, while having *ot-iti “go away” in
the same time. There is also a possibility that the final –ъ here was introduced by analogy to
the prepositions kъ, vъ, sъ (Gluhak). In the case of *jaz, there is no reason whatsoever to
introduce an unetymological –ъ. The proto-form *jaz would simply yield *ja and the final *-z
would be gone forever, just like the final *-s that disappeared in the o-stem nominative
singular ending *-us > -ъ. There was no need to preserve the final *-z in *jaz, unlike in *ot,
where the final *-t was crucial for the communication. The final *-z did not have any special
function, and the form *ja was distinctive enough, so there was no reason for a final *-ъ to
be added60. Moreover, if *ja were the only original form, one would expect the acute here as
the result of Winter’s Law - *jaɾ < PIE *égɴ. However, in that case it would be impossible to
explain the neo-acute accentuation that also appears in this form (see § 13). Thus, Vaillant’s
idea has to be rejected, since the cases of otъ and jazъ are not comparable and since the
accentual data cannot be explained by his hypothesis.

As we have seen, some of these explanations are indeed possible. However, there is
another possibility - to presume that both of these forms, both *ja and *jazъ, are original
and thus derived directly from Indo-European forms *égɴ and *egɴHóm. This idea is not new61,
in spite of the fact that it was never very popular. This may strike us as a little bit odd, since
it was clear for quite some time that there were at least two forms of this pronoun in IE (see
§ 100, § 101), which would in Slavic yield exactly the forms we have attested – ja and jazъ.
Thus, it is strange that so many linguists have gone at such trouble to explain the
mysterious supposed disappearance of the final –z(ъ), when there was no need to assume
any disappearance in the first place. This is probably the result of the fact that in OCS only
the form with the final –zъ is attested62. Because of the false unconscious prejudice that OCS
is for all practical purposes identical to Proto-Slavic, many linguists were probably prone to
reject any possibility of ja to be old, since it did not appear in OCS.
There is more than one explanation as to why there is only one example63 of the
short form in OCS. The simplest explanation would perhaps be that this alternative form was
lost in the dialect that was the basis of OCS. That can hardly be considered impossible since
there are quite a number of such cases – it is not always true that OCS forms are the oldest,
in spite of the fact that OCS is the first attested Slavic language. Thus, for instance, OCS has
the ending –tъ in the 3rd person singular of the present tense, which is, according to the
comparative IE evidence, apparently secondary to the ending –tь, attested in Old Russian and
Old Polish. To name another example, in OCS, the ending of the n-stem dative plural is the
obviously secondary –anьmъ, with the final –ьmъ taken from the i-stems, while in Old

60 The case of Dutch [ik] and [ikǩ] is not really comparable. The exact origin of this schwa is not clear
and, in any case, the schwa was not added here to prevent the loss of –k in Dutch.
61 Cf. for instance Kluge, Якубинский 1953: 200, Schmidt 1978: 27. Others have also accepted the two
forms already in Proto-Slavic, even if they did not attribute them directly to IE heritage, cf. for instance
Lehr Spławiński & Bartula 1954: 51 and Arumaa 1985: 159.
62 Although this is also not completely true, since, as we have seen, the form a is also attested in OCS,
if only as a hapax in Psalterium Sinaiticum, see § 2.
63 Or none, if one does not accept the form a of Psalterium Sinaiticum as relevant.

Croatian (in Povelja Kulina bana from 1189) one finds the ending –am which represents the
original ending. The same goes for the OCS ending of the instrumental singular of i-stems
where we find the ending –ьjNj (taken from ā-stems and there from the pronouns), while in
modern Croatian we find a variant –i which is the reflex of the original IE ending *-ih₁. One
can think of other reasons why there was no short form in OCS texts – for instance, the
longer form *jazъ (i.e. azъ in OCS) could have been regarded as a solemn form in opposition
to ja, which could have been regarded as a colloquial or intimate variant. This is of course
speculation, but it is nonetheless possible.

The question to be asked is – why should we presume the existence of two forms, *ja
and *jazъ, at all? What are the reasons to consider this theory at all? There are a couple of
good answers to this question, some of which have already been mentioned:

Firstly, the simple fact is that both *ja and *jazъ can be derived very easily from Indo-
European and Balto-Slavic. In fact, no one really doubts that the form with the final *–zъ is
old64, and this form is attested outside Balto-Slavic (Vedic ahám), while the form ja (from
*jaz in pre-law-of-the-open-syllables Slavic) has a direct correspondence in Baltic65. If there
is a short form attested in Baltic, why should we assume that this short form had
disappeared in Slavic just to be resurrected again via some irregular allegro process or
analogy? Thus, if one can explain both forms easily by deriving them regularly from IE, there
is no point in struggling to explain ja in a different fashion.

A second reason for the assumption of the two original forms is the fact that, as was
already seen (cf. the section Material), in many Slavic languages, the z-less forms occur
rather early. For instance:

-in Štokavian only ja occurs already from the 12th century

-in Kajkavian texts, ja and jaz seem to occur at the same time
-in Old Czech, the z-less form occurs together with the longer one from the very beginnings
of the literary tradition
-in Polish, the z-less forms is practically the only one since the first texts
-in Russian both forms occur together from the earliest times

Thus one has to take into consideration the option that Proto-Slavic had both the form *ja
and *jazъ, and that only later one of these forms was generalized in various Slavic languages
- the form *ja was usually more successful.

The third reason is that, like the two variants of the same pronoun that exist in Slavic,
two types of accent can also be found. This has up until now, to my knowledge, never been
taken as an archaism. These two types of accentuation must be related to these two different
forms – it can hardly be a coincidence that there is evidence not only for two different forms,

64 Except Vaillant, see above.

65 Except for the fact that the final consonant is devoiced there, which also solves the disagreement in
the vocalism, see § 70.

but for two different accents of the same pronoun. The existence of the two types of
accentuation (see § 13) provides an independent indication that there were two different
forms for “I” in Slavic. If there had been only one form, there would be no reason for it to
have two different accents.

Two forms in Old Czech

§ 12 As we have already said, in this work we shall begin with a presumption that Proto-
Slavic originally had two different forms of the 1st person singular personal pronoun in the
nominative - *ja and *jazъ, both of which were a direct heritage from Indo-European. If
Common Slavic had two variants of the same pronoun, one can assume that they had
different (or slightly different) functions. E.g. one could have belonged to the higher style,
one to the lower, one could have been used more emphatically than the other66 etc. Of
course, this need not be true. A language can have two forms of a same pronoun without
really distinguishing them in any way. This is exactly what Kiparsky (1967: 130-131) claims
for (Old) Russian, as was already said. He says that it is quite possible that the forms я and
яз have co-existed in Russian without problems for quite some time without any significant
stylistic difference whatsoever. In fact, that is seen right from the earliest Old Russian texts
like Мьстиславова грамота, where these two forms, together with OCS loanword azъ, occur
together without any noticeable difference (see § 9). Kiparsky adduces a typological example
of a Swedish dialect in Finland in which the 1st person singular personal pronoun, written
<jag> in Swedish, is pronounced both [ja:] and [ja:g] without any significant difference67.
Kiparsky says that [ja:g] indeed occurs when the pronoun is specially emphasized (like in the
sentence “det är jag” – “that’s me”), but that it also occurs when it is not emphasized at all.
That is exactly what Гадолина (1963), as already mentioned, has shown – that both я and яз
in Russian could have been in the emphasized and in the non-emphasized position68. Thus,
we may conclude that in Russian both forms co-existed for quite some time (up untill the
17th century) without any difference69.

However, as we have already said, it is possible that there was some difference in
Common Slavic, which was perhaps lost in Russian. Thus, we should consider examining the
occurrence of these two forms in other Slavic languages that preserve them70. We have
already excluded Russian. Polabian can be excluded for the lack of sufficient data. In Polish
jaz appears only once (if at all). In Bulgarian, there seems to be no z-less forms in the old
texts. In Štokavian, only ja appears from the earliest times. In Kajkavian, both forms occur,

66 One could say that personal pronouns are emphatic by default in flective languages like Slavic.
However, that is not completely true. For instance, "I" is used always (or almost always) in Russian and
Lithuanian (which are, of course, flective), and in Croatian not every "I" must be used emphatically.
67 Since the forms *ja and *jazъ do indeed look like variants of the same form, the first form looking
like a truncated version of the second one, one could also compare the use of these two forms with the
invariant use of the adjectival-pronominal endings –og/-oga, -om/-omu/-ome in modern Croatian
and other cases when more than one form is used in a language without any significant difference in
meaning, function or style. In Slavic, the forms ja and jaz(ъ) were probably functioning like [ja:] and
[ja:g] in the mentioned Swedish dialect, or like Dutch [ik] and [ikǩ], older English i and ic etc., where
there is also no difference in meaning. The Slavs could not have known the historical origin of these
two forms.
68 Kiparsky 1967: 131 quotes four sentences in order to prove this.
69 Kiparsky criticizes Meyer (1923), who unwarrantedly suggested that the difference in the use of the
two forms depends on the position of the form in the sentence.
70 Or rather, which had preserved them in a certain period in the history.

but, as we have already seen, in the languages of the earliest writers (Pergošić and Vramec)
they apparently do not appear in the same texts.

Thus, we are left with only one possible candidate that could have a functional
opposition of ja and jaz - Old Czech (for Rezija Slovene, cf. § 5). To determin if Proto-Slavic
had a difference in the usage of the two variants of this pronoun, it is necessary to explore
whether such a difference is maintained in the earliest stages of the Old Czech, which
exhibits both forms. Of course, it is possible, as already said, that at that time there was no
difference anymore (if it ever existed).

For this purpose, I have examined three Old Czech texts from the 14th century in
which both forms appear71.
In Alexandreis (from around the year 1300), the form já and jáz both appear exactly
seven times. No pattern can be noticed. Each form appears both with emphasis and without
it, and there is no significant difference concerning their position in the sentence – já
appears six times in the middle of the sentence and once isolated (Kráľu milý, já, člověk
svých přátel zbylý), while jáz appears twice at the beginning of the sentence, and five times
in the middle of it.
In Dalimila, which dates from sometime between 1308/1310 and 1314, jáz appears
only eight times (each time at the beginning of the sentence), and já 27 times in all positions
(approaching the end of the text, jáz appears less and less). Both forms appear with an
emphasis and without it72.
In Mastičkář muzejní, a text from the 2nd half of the 14th century, jáz is attested nine
times in all positions, while the form já is attested nine times as well, but only in the middle
of the sentence73. Again, each form appears both with emphasis and without it.

We can conclude that one cannot find a pattern of use or a difference between Old
Czech já and jáz. Both are used in all positions in a sentence, and both are used when
accented in a sentence and when not. The phonological shape of the next word also plays no
role. A mild tendency may be noted in some texts for jáz to be used at the beginning of a
sentence and já in the middle of it. However, it is only a mild tendency, since either form can
appear at the beginning as well as in the middle of a sentence. If there was a difference of
*ja and *jazъ in Common Slavic, it was not preserved in Old Czech, and it seems that it is to
stay out of our reach for good.

71 Not all texts can be included for this kind of purpose, since in some the pronoun is not attested at
all, in others only one form appears, so there is no variation, and yet in others there is simply not
enough forms to conclude anything. Thus, the corpus of suitable Old Czech texts is not very large.
72 The form jat appears three times. This form was made by the wrong analysis in the sandhi – it
appears twice in jat to (before to) and once before s- (Jat Svatoplukovi…) in this text. It is also attested
in other Old Czech texts.
73 The form ját appears here once as well together with its variant jáť (both in ját/jáť sem).

The accent of *ja(zъ)

§ 13 As we have already seen in the chapter Material, two types of accentuation are
attested in the 1st person nominative singular pronoun in Slavic. Most languages point to the
neo-acute (*jãzъ): Old Štokavian jã, Neo-Štokavian jâ (ì jā), South and part of Central
Čakavian jã, Kajkavian (most dialects) jã, Czech já (Old Czech já(z))74, Slovak dial. já, Polish
dial. jå, Kashubian iʠω and Slovincian jǛu. On the other hand, Slovene jàz (dial. also jà), North
and part of Central Čakavian jaʂ(z) and Kajkavian jaʂ(z) (in some dialects) point to the old acute
(*jaɾzъ). In Old Russian, язъ is an «энклиномен»75 (which points to the old circumflex), but
this must be due to the analogical influence of the other personal pronouns ты, мы, вы in
which this is expected76 (see § 14, § 29). Here is an overview of the reflexes in Slavic:

Neo-acute Štokavian, South and part of Central Čakavian, Kajkavian (most dial.),
Czech/Slovak, (Old and dial.) Polish, Kashubian/Slovincian

Acute Slovene, North and part of Central Čakavian, Kajkavian (marginally)

Circumflex Old Russian77

The problem of two kinds of accents is obviously very interesting. It is a stunning fact
that it was, to my knowledge, never treated in the literature in more than a couple of

Three solutions are possible here – the first one is to assume that the acute is the
original accent; the second one is that the neo-acute is original; and the third one is that
both accents are original, i.e. that neither is secondary.
The first possibility is took up by Kortlandt78 who claims, very shortly, that the acute,
preserved in Slovene jàz (as well as in the neighboring Čakavian dialects), is the original
(Balto-)Slavic accent. Kortlandt believes that Old Štokavian jã and South Čakavian jã are the
result of the later secondary lengthening79. A similar claim is made by Vaillant (1958: 443)

74 Czech já(z) could theoretically be derived from *jaɾzъ as well but in the light of Slovak dial. já, this
seems highly unlikely, and the Czech form should also be derived from *jãzъ.
75 Зализняк 1985: 143.
76 Cf. a typological parallel in Štokavian, where tĩ, mĩ and vĩ have the neo-acute by analogy to the form
jã where this is expected.
77 The innovative Slovene form jâz (cf. § 5f) has no connection to this form.
78 Cf. for instance Kortlandt 1997: 29 and Kortlandt forthcoming b.
79 It must be noted, however, that Kortlandt bases his theory on the subject on his own doctrine of
Slavic accentuation, many postulates of which I do not accept. Thus, his views on the subject are hardly
explainable in the non-Kortlandt doctrine - for instance, in Kortlandt forthcoming b, he says that
Czech and Slovak have preserved the original short reflex of the acute in ty, my, vy. However,
according to my view, and to the view of the most Slavic accentologists, the original reflex of the acute

who says that the Štokavian jâ is the result of a compensatory lengthening and that (North)
Čakavian jaʂ(z) represents the original Proto-Slavic form. Kortlandt and Vaillant aside, there
are several reasons why the theory of the original acute cannot be correct. First of all, the
proposed lengthening is completely ad hoc and there is no real explanation for it (except, of
course, the ever present allegro change, which is not very convincing). Secondly, if we were
to suppose that the acute is the original accent here, we would have to assume a secondary
ad hoc lengthening not only in Štokavian and South Čakavian, but also in Kajkavian, Czech,
Slovak, Polish and Kashubian/Slovincian. Somehow, this possibility does not strike me as
very economical. One could assume compensatory lengthening due to the dropping of
*-zъ, like *jaɾzъ > *jã (or *jaz > *jā), but it is not convincing that this would occur in so many
languages (in Czech *jaɾzъ would yield jáz, the same as *jãzъ), especially taking into account
that there are absolutely no evidence that would point to the form *ja being more recent
than *jazъ. Besides, in North Čakavian the form često jaʂ without final –z, and the supposed
lengthening occurs quite frequently, and the same goes for Slovene dial. jà, also without the
supposed shortening. Old Kajkavian <iaaz> in Pergošić's language, but <ia> in Vramec's (cf.
§ 4) goes directly against a conclusion like the one mentioned. We can conclude that the
neo-acute/length in this pronoun has nothing to do with compensatory lengthening.

The second possibility is to assume that the neo-acute is the original accent here,
and that the short reflexes in Slovene, North Čakavian, and marginally in Kajkavian are
secondary and due to an allegro shortening80. However, this view, in spite of the fact that it
is more economical than the former, because it supposes allegro shortening only in three
neighboring dialects (and not in seven or so widely spread ones), is again troubled by the
fact that yet again one must subside to the help of the mysterious allegro changes and ad
hoc developments. It is unclear why jã(z) would be shortened to jaʂ(z), while there was no
shortening of tî, mî, vî anywhere.

Thus, as already said, the deficiency of both of these approaches, in spite of the fact
that these kinds of irregular secondary lengthening and shortening in short forms like these
ones are theoretically possible, is that they presume ad hoc lengthening and shortening, not
attested elsewhere in these languages. For instance, there is no parallel for the change of jaʂ
> jã elsewhere in Štokavian. Kortlandt’s theory is even more unconvincing since it includes,
among other things, lengthening in Slovene tî, but the absence of it in jàz81.

is long in Czech and short in Slovak. This is one of the reasons why Kortlandt’s theory cannot really be
accepted by accentologists who are not working within the framework of the Leiden doctrine.
80 This is the view of Bulcsú László (personal communication).
81 Kortlandt (1997: 29) believes that the falling tone in Slovene tî, mî, vî, but the rising one in Posavina
and southern Čakavian points to the secondary lengthening of an originally short vowel (i.e. the reflex
of the old acute). However, he provides no reason for this supposed “secondary lengthening”. Besides,
Old Russian and Middle Bulgarian point to the “real” Proto-Slavic cirkumflex as well (cf. § 14),
exhibiting Vasilev-Dolobko’s Law and the shift of the accent to proclitics. It seems highly unlikely that
a “secondary circumflex”, resulting from later (irregular) lengthening would in Old Russian behave
exactly the same as an original circumflex. Cf. Croatian neo-circumflex in examples like G. pl. klaʄdā,
which is different from the original circumflex in A. sg. ruʄku (òd klādā/od klaʄdā, but naʂ rūku). If the
circumflex in Slovene tî (and in Old Russian тыɴ, иɴ ты, ты жè) is due to Meillet’s Law (see § 57), it is

The third approach, however, seems more reasonable. If we have established that
there were probably two forms of the pronoun in question already in Common Slavic, *ja and
*jazъ, and since we already know for sure that there were at least two different forms of this
pronoun in PIE82, why not assume that there were two types of accent in Slavic, if this is what
the material points to? Why not assume that somehow both the old acute and the neo-acute
are original in Slavic? Since we already have two forms - *ja and *jazъ, why should we not
relate these two forms to the two different kinds of accentuation that the material seems to
point at? The logical thing to do would be to assume that the old acute is the original accent
of the shorter form *jaɾ, and that the neo-acute is the original accent of the longer form
*jãzъ. In this way, Common Slavic *jaɾ could be derived from PIE *égɴ with the acute as the
result of Winter’s law83. The absence of the acute in Lithuanian àš is easily explained if we
derive Lithuanian àš and Old Lithuanian eš from a PIE *ekɴ84, a sandhi variant with the
devoiced ending, which was generalized in Baltic and in front of which, logically, there was
no Winter’s law and hence no acute. Common Slavic *jãzъ is to be derived from the older
*jāzъʀ85, with the neo-acute as the result of Ivšić’s Rule86. It is only natural to relate a neo-
acute to a form that has two syllables, because the neo-acute demands an extra syllable (i.e.
a weak yer) in order to be explained convincingly. Slavic *jāzъʀ can also be directly derived
from the PIE form *egɴHóm (Vedic ahám) with the final accent87. The lengthening in Slavic can
be explained here also via Winter’s law88. Here, Winter’s law has produced the acute pretonic
length. In Slavic, the acute has been lost in unaccented positions89, and when the stress was
finally retracted via Ivšić’s Rule, the syllable received the usual neo-acute intonation.

expected that it would act exactly the same as the old circumflex. Cf. Croatian glaʄvu, naʂ glāvu, with the
circumflex due to Meillet’s Law, and ruʄku, naʂ rūku, with the original circumflex.
82 *égɴ as attested in Old Lithuanian eš, and *egɴHóm as attested in Slavic jazъ or in Vedic ahám. See §
83 For the problem concerning this form and Meillet’s Law, see § 57.
84 For this reconstruction, cf. also Bańkowski, who takes it as a dialectal IE form.
85 In the Slavic nominal system, *-ъ in the nominative-accusative sg. (< IE *-os, *-om) could not be
accented before Дыбо’s Law. Slavic had an initial accent in the nominative-accusative sg. of the a. p. c,
having thus lost the *-ós of the IE o-stem oxytones (which were the source of Slavic a. p. c). In *jāzъʀ,
however, the original place of the accent was preserved. This is not surprising, since the pronominal
system often behaves differently from the nominal one – cf. Lithuanian forms like anàs (the nominative
–as cannot be accented in the nominal declension). Moreover, the form *jāzъʀ was isolated and not in a
close relation to the nominal forms, so that the preservation of the final accent need not surprise us.
Cf. also the accented gen. pl. *-ъʀ in a. p. c.
86 The retraction of the accent from a weak yer.
87 Ivšić (1911: 194) first proposed this explanation.
88 Cf. the closed syllable, which is necessary according to Matasović’s interpretation of Winter’s law
(see for instance Matasović 1995).
89 This is the only logical assumption. That is clear from the fact that the old mobile acute paradigm
has remained mobile, which points to the conclusion that the accent was not retracted to the acute
pretonic vowel, as could be presumed, or the old mobile acute nouns would all get a fixed old acute on
the root (thus, Lithuanian a. p. 3 would be related not to Slavic a. p. c, but to Slavic a. p. a). The old
acute pretonic long vowels behave just like the old non-acute pretonic long vowels, which must mean
that the acute has just disappeared if it was not accented.

PIE *égɴ > Slavic *jaɾ
PIE *egɴHóm (Vedic ahám) > Slavic *jāzъʀ > *jãzъ

Thus, as we have concluded, there were two forms originally in Slavic - *jaɾ and *jãzъ
< *jāzъʀ. What has happened after Common Slavic period? These two forms must have mixed,
probably from an early age. Thus, the forms *jaɾ and *jaɾzъ (cf. Slovene dial. jà and standard
jàz) have emerged, as well as *jãzъ and *jã (cf. Old Czech jáz and já). Various
languages/dialects have generalized one of the two accents in both forms, and that is what
we see now in modern Slavic forms. First, there were two forms and two accents, then two
forms and one accent, and finally one form and one accent. The generalization of the accent
had finished before the disappearance of the two different forms. In the small area
consisting of Slovene, part of Kajkavian and North Čakavian, the acute was generalized in
both forms (cf. North Čakavian jaʂ and jaʂz). In the rest of the Slavic languages (excluding the
unclear East Slavic), the neo-acute was generalized in both forms (Old Czech já and jáz).
Thus, the generalizing of the accent, as already said, occurred earlier than the generalization
of one of the two forms and is therefore independent of it.

Slavic *ty

§ 14 The reconstruction of Common Slavic *tŷ is completely unproblematic. All languages

show regular correspondences here.
The circumflex in *tŷ is confirmed by Slovene tî, some Čakavian dialects (for instance
Novi) tî, Kajkavian (Bednja) teʄi, Czech ty, Slovincian taɺ, Old Russian тыɴ, which is an
«энклиномен»90 (cf. иɴ ты, ты жè91) and Middle Bulgarian ты бωɳ, ты жè92.
Old Štokavian, South Čakavian and Kajkavian (most dialects) tĩ is secondary, due to
analogy to the form jã (thus also secondary mĩ, vĩ, and in some dialects even õn, instead of
the older oʂn, which is also attested93), where the neo-acute was a salient feature. In Neo-
Štokavian, the acute original of tî (and mî, vî) is seen in ì tī (and not **ıʂ tī).

90 Зализняк 1985: 143.

91 Дыбo 1981: 51.
92 Дыбo, ibid.
93 Posavian õn (not oʄn) points to analogy with other pronouns as the source, rather than to the
lengthening before a sonant (which yields ʄ, not ɶ, in Posavina, cf. Kapović 2005: 54-55.

Oblique cases of 1st and 2nd person singular

We shall discuss first the material of all Slavic languages separately, together with
historical and dialectological relevant data. In general, only internal developments shall be
discussed. Of course, a minimal amount of comparison with other languages and/or
Common Slavic will also be applied when necessary. In the end, Common Slavic paradigm
will be reconstructed, and special problems, concerning Common Slavic (or earlier) period
will be discussed. Since 1st and 2nd person singular share many traits, it is only logical that
we discuss them consecutively.

Old Church Slavic

§ 15 In the genitive singular in OCS, the form mene is attested. This form is quite
unproblematic. The form mne, which occurs a couple of times already in OCS (and which
reminds one of modern West Slavic forms like Czech mne or Polish mnie), is obviously an
innovation due to the influence of the dative singular mnĕ < mьnĕ and the instrumental
singular mnojNj < mъnojNj. With OCS genitive singular tebe, there are even fewer problems,
or rather – there are no problems whatsoever. This is in fact the only attested form.

The dative singular form is, however, very problematic. In OCS, it occurs as mьnĕ, but
also as mъnĕ (the form mnĕ, also attested, is obviously younger). Since the reconstruction of
the first vowel in this form in Slavic depends practically solely on OCS data (except for Old
Russian), due to the disappearance of the weak yer in other Slavic languages, we shall
discuss the problem of whether this form had ь or ъ here in this section, while leaving the
general discussion about the origin of this form for later.
The usual presumption is that the form mьnĕ is older, although some linguists do not
agree94. The basic reason for this presumption is the obvious parallelism of the OCS dative
and instrumental singular forms of the 1st and the 2nd person singular95. One finds the forms
mьnĕ and tebĕ in the dative singular, which have a front vowel in the root (ь – e), while in the
instrumental singular the forms mъnojNj and tobojNj appear, which have a back vowel (ъ – o)
in the root. It is also interesting to note that the front vowels occur in front of a front vowel
in the next syllable (-ĕ in the dative), while the back vowels occur in front of a back vowel in
the following syllable (-o- in the instrumental). It is, however, strange that while the
connection of mьně and tebě (ь ~ e) is regularly noted (for OCS), no researcher has noticed
that mъně would indeed be in accord with North-West Slavic form tobě in the dative. Thus,
the parallelism of mьně and tebě in OCS is irrelevant.
The supporters of the thesis of the originality of the form mъnĕ are not so numerous.
The usual reason for the presumption that the form mъnĕ is older is a supposed connection
of this form with a dialectal Lithuanian form mùnei96. In fact, for the Baltic dative stem mun-,

94 For instance Ivšić 1971: 219. Cf. the discussion there.

95 Cf. for instance Vondrák 1912: 459, Weingart 1937-8: 201 or Ceлищев 1952: 113.
96 Thus for instance Беседина-Heвзорова 1962: 178. The theory of Hujer (1911, 1912) is not very

Latvian dialectal data must be quoted, since Lithuanian mun- is probably secondary (see §
59, § 62).
A third, we could call them fatalistic (or maybe just realistic), group are those
linguists who do not believe that it is possible to deduct the original state of affairs in the
roots of the dative and instrumental singular in OCS (and Slavic in general), due to the
inconsistencies in OCS spelling97. It must be noted that in the case of the originality of the
form mьnĕ, the secondary form mъnĕ can easily be explained as analogical to the
instrumental form mъnojNj. In the case of the originality of the form mъnĕ, one must
attribute the secondary form mьnĕ:
-to the general mix up of the yers in OCS98,
-to the influence of the final front vowel99,
-or to the reshaping of the ablaut pattern - *en : *ъn (in *mene : *mъnĕ) > *en : *ьn (see §

Another angle could perhaps shed some light on this problem. The form mьnĕ is
attested in OCS in Codex Zographiensis and in Sava’s Book, while the form mъnĕ is attested
in Codex Marianus, and for instance in Ostromirovo Evangelie in Old Russian100. Here might
lie another argument for the originality of the form mьnĕ, since Codex Zographiensis is
considered to be second only to the Kiev Folia when it comes to the preservation of the
original difference of the yers, and Sava’s Book is also considered to be one of the more
archaic OCS texts in that regard101. In return, Codex Marianus, in which the form mъnĕ is
attested, is notorious for its unreliability when it comes to the distribution of the yers. There,
a tendency of a general spread of ъ, instead of the original ь is attested102, and, in general, a
ъ can take the place of any original ь. This makes the words with attested ь true
archaisms103. However, this is not conclusive, since a number of examples in Codex
Zographiensis and Sava’s Book show secondary changes of yers there as well. Thus, in
Codex Zographiensis, a tendency is attested for a yer to appear when there is a front vowel
in the next syllable, and for a yor to appear when there is a back vowel in the next syllable.
Thus, we have, for instance, secondary bьdĕti instead of bъdĕti and secondary vĕrъny
instead of vĕrьny, besides the original vĕrьni when the front vowel is in the following
syllable104. Thus, we are back at the beginning of the problem since the form mьně can be
secondary in that text as well.

Putting aside the (in)consistency of the OCS spelling, the parallelism of the 1st and 2nd
person singular forms in the dative and the instrumental singular is not strong enough an
argument to presume that the form mьnĕ was indeed the original one. Especially since the

97 This is the opinion of Vaillant 1950: 136 and Lunt 1974: 65 for instance.
98 This explanation could of course be possible in the first case, mьnĕ > mъnĕ, as well.
99 Cf. for instance Leskien 1990: 109, who considers the option of a secondary ь, due to the influence
of the ĕ in the next syllable, possible.
100 Vondrák 1912: 459.
101 Vondrák 1912: 268-269.
102 Vondrák 1912: 272.
103 Leskien 1990: 29.
104 Nandriş 1959: 39, Leskien 1990: 28.

parallelism scheme is valid for South Slavic and OCS Slavic only, since in North and West
Slavic, as already mentioned, one finds the form tobě in the dative singular, which would be
parallel to mъně, not mьně. Also, this parallelism could easily be secondary. For a further
discussion on mъně and mьnĕ, see § 46.
In OCS, the dative singular form of the 2nd person singular is tebĕ. This form is
unproblematic in OCS, although apparently not in Common Slavic, as is to be seen.

In the accusative singular, the original OCS forms were mę, tę, which were still
accented in that time. The use of the original genitive forms mene, tebe as the accented
accusative (while using mę, tę as clitics) is an innovation in OCS. This was an early change
that has occurred in all Slavic languages.

The locative singular is the same as the dative singular – mъně/mьnĕ, tebĕ, except
for the absence of the clitic forms mi, ti that do not exist in the locative.

The instrumental singular form of the 1st person is mъnojNj in OCS. The form mnojNj is
obviously younger, showing the disappearance of a weak yer. The 2nd person pronoun is
tobojNj, with an –o- in the root, which, as already said, corresponds to the reduced –ъ- in the
1st person pronoun.


§ 16 In Modern Bulgarian, not much is preserved due to the loss of most cases. The loss,
which was nearly complete in the nominal declension (setting the vocative aside), was a little
bit more merciful in the pronominal declension. There, a larger number of forms was
preserved105. However, the process of analitization continues up to the present day, as we
shall see clearly in the case of the dative forms.

The genitive forms have formally remained the same: мèне, тèбе, while merging with
the reflexes of the old dative forms *mьnĕ (analogically ⇒ *menĕ with the *-e- from the
genitive) and *tebĕ, which have also regularly (*ĕ > e), yielded мèне, тèбе. There are no real
genitive forms in Bulgarian, of course, the old genitives *mene and *tebe have become, as
elsewhere in Slavic, the accented accusatives, which are today accusative/oblique forms in
The forms мèне, тèбе have, up until recently, also functioned as datives on their own
(as the rightful heirs of the old *mьnĕ, *tebĕ), but they are pretty much archaic nowadays in
Modern Bulgarian, which prefers analytical phrases на мèне, на тèбе instead. The short,
unaccented dative forms are preserved: ми, ти, as well as the old accusatives, which have
become (as elsewhere in Slavic) unaccented accusative forms: мe, те.

We can find out a lot more in Middle Bulgarian. Cf. the system of Evangelie 1139
(Hock 1992 I: 119): мeнé, мнΟ, мΧ (мΧɴ), мнóΧ and тебé, тебΟ, тΧ (тΧɴ), тoбóΧ. The accent in

105 This is usually the case, cf. Romance languages or English, which have preserved the remains of the
declension in pronouns only.

the genitive is final, as we can see: мeнé, тебé. This must be due to an early analogy with the
end stress of the dative-locative мнΟ, тебΟ (see § 52). The Modern Bulgarian forms мèне,
тèбе are probably due to the regular stress retraction like *seloʀ > Bulgarian cèлo. In the
instrumental, this Middle Bulgarian dialect has the accent on the first syllable of the ending:
-óΧ. However, since we find the same accent in both a. p. b and a. p. c ā-stems in that
dialect (cf. женóΧ, странóΧ, Hock 1992 I: 90), these forms are not as indicative as Croatian
ones (see § 17, 54). The accent is final in forms like вь мΧɴ, зa мΧɴ (cf. Neo-Štokavian ú me,
zá me).


In Macedonian, the situation is generally not much different from the Bulgarian
situation. However, the dative forms мeне and тeбе (or rather – doubled мeне ми and тeбе
ти) do still function alone, without the preposition на, unlike in Modern Bulgarian. Forms like
на мeне instead of just мeне, do however appear in dialects and in spoken language106.
In Macedonian, the longer forms are generally (unlike Bulgarian) always «doubled»
with the shorter forms (thus мeне мe and not just мeне), if we put aside special cases like
ellipses and oaths. In Bulgarian, the longer forms can appear by themselves, in spite of the
fact that the doubling of the object is quite frequent there as well.


§ 17 In Croatian, the genitive forms mene and tebe have remained unchanged. Today, they
function as the accented accusative as well, and the original accusative forms me and te are
now the clitic genitive-accusative forms.

In the dative-locative form, the ending –i is the same ending one finds in ā-stem
nouns like žèni ⇐ *ženě (the soft-stem ending –i is generalized). In Jekavian Štokavian
dialects, the expected form should be menje. This is indeed attested in Old Croatian, for
instance in Dubrovnik (ARj). In the Ekavian dialects of Štokavian, one finds both the
analogical meni, tebi in some dialects, as well as the original mene, tebe in others107. The
vocalism of the dative-locative is younger, due to analogy to the genitive-accusative form.
The older forms mьně, mni, mani108 are attested in older texts and in dialects (ARj). The
analogical form meně is attested since the 14th century. The original stem mn- was
eliminated in the dative-locative, but not in the instrumental (at least not in all dialects,
including the Standard Croatian). The same thing occurred, for instance, in Ukrainian (see §

106 Koнески 1965: 123.

107 The explanation of Mareš (2001: 136), that meni is a mix of the original mene < *menĕ (obviously,
this would work for Ekavian dialects only) and the unaccented mi is completely off track, since the form
meni occurs in Jekavian dialects as well, and one can hardly separate this development from the overall
tendency of the generalization of soft endings in Štokavian dialects.
108 Cf. here «jaki vokalizam», i.e. «full vocalism» in Čakavian.

In Western (Croatian) Neo-Štokavian dialects, in both the genitive-accusative and in
the dative-locative singular, one finds the short falling accent: meʂne, meʂni and teʂbe, teʂbi (cf.
also zà mene, òd tebe, ò tebi etc.). This is the Standard Croatian accent as well. In Eastern
(Serbian) Neo-Štokavian dialects, one finds the short rising accent – cf. Standard Serbian
мèнe, тèбe, за мèнe, o тèби etc. In Neo-Štokavian, one of these accents is usually
generalized in both forms (for Kajkavian, cf. Bednja maʂna, taʂba, maʂnje, taʂbe). In Posavina109,
according to Ivšić one usually finds meʂne, meʂni, teʂbe, teʂbi, but also sebıʂ, mèni, tèbi, menĩ,
tebĩ in DL. and rarely even mène, tèbe in GA. (some dialects in Posavian have a secondary ɶ
instead of ʂ in DL., like some Čakavian dialects). Cf. my own data from Posavina - Sapci od
meʂne, kod meʂne, kod seʂbe, mèni, k mèni, tèbi, Glogovica kòd mene, mèni, oko seʂbe, sèbi, na
sèbi, o sèbi, Oprisavci meʂne, po teʂbe, sèbi, na sèbi, Bebrina mèni, tèbi, Gundinci na meʂne,
tebıʂ, Orubica mèni, Dubočac mèni, Siče tèbi. This data attests only the pattern meʂne –
mèni/menıʂ in Posavina, while Ivšić claims that meʂne - meʂni is the commonest pattern. I have
an attestation of meʂne, but mèni and teʂbe, but tèbi for Slavonski Brod (Neo-Štokavian) as
well. Ivšić (1971: [374]) adduces also Lumbarda Čakavian menĩ, tebĩ (but also meneɶ), and
cites Kušar's data for the villages on Rab, where meʂne, meʂni/menıʂ and teʂbe, teʂbi/tebıʂ
appear. He concluded that the pattern meʂne, menıʂ is the original one, and that meʂne, meʂni
and meneʂ/mène, menıʂ/mèni are just generalizations of the accent of the genitive or the
dative. For the preservation of the original pattern (attested for instance in North Čakavian),
cf. also (Jurišić 1966: 84-85, Jurišić 1973): Vrgada meʂne, meʂni and teʂbe, teʂbi, but in the
locative po menĩn, o tebĩn etc.110, Susak111 mêne, menıʂ, têbe, tebıʂ etc., but Cres meneʂ, menıʂ.
In Molise Croatian, one finds interesting data: in Mundimitar (Piccoli & Sammartino) meʂne,
but tébe (< *meʂne, *tebeʂ), while in Živavoda Kruč (Breu & Piccoli: 404-405) one finds ména,
téba (< *meneʂ, *tebeʂ)112. Rešetar (1911: 212) adduces the Molise forms mén(e), téb(e) in the
genitive, but meʂn(e), teʂb(e) in the accusative (notice the unusual differentiation of the
genitive and accusative by means of accent!) and meʂn(i), teʂb(i) in the dative113.
One should also note that the accent meʂne, teʂbe was originally the absolute initial
accent (the original circumflex) – thus oʂd mene etc. (cf. Vrgada koʂd mene, preʂs tebe, ıʂza
mene, ıʂspri mene, uʂ tebe, uʂ mene, Susak ôd mene, kêl mene, ôl tebe, kêl tebe – with
secondary lengthening etc.). However, in Western Neo-Štokavian (and Standard Croatian),
this is not retained – there, these forms are òd mene, zà mene etc. There, the generalizated

109 Ivšić 1971: [370, 373, 374].

110 This kind of difference, between the accent (and form!) of the dative and the locative is seldom
found. It is probable that Vrgada menĩn, tebĩn in the locative stem from the older *menĩ, *tebĩ < *menıʂ,
*tebıʂ, while the dative form meʂni took the accent of the genitive. Cf. also in Vrgada D. ńeʂmu, but L.
ńemũ, both from the older *ńemuʂ.
111 In Susak, there is probably no distinctive intonation and that is the reason for leaving out the neo-
acute mark from the examples from Susak. The length in Susak examples like mêne, ôd mene is, of
course, of secondary origin.
112 The transcription of the Živavoda Kruč dialect is somewhat adapted, for simplicity's sake.
113 The accusative forms meʂn(e), teʂb(e) point to the conclusion that both rising and falling accent was
possible in the predecessor of that dialect in the genitive-accusative and dative-locative. The variant
with the falling accent has prevailed in most cases (cf. also the instrumental forms meʂnom, teʂbom), but
the rising accent was generalized in the genitive forms in order to differentiate them secondarily from
the accusative forms. A similar case of secondary development in the accentuation of the genitive and
dative forms is attested in Kašić's grammar (cf. Kapović forthcoming b).

short falling accent in the genitive-dative started functioning as the secondary falling stress
(it was indeed secondary in the dative, but original in the genitive). These forms were
influenced by other secondary falling accents in pronouns - toʂgā, òd togā (instead of the
older tòga), njeʂga, zà njega (instead of the older njèga) etc.

As already said, the old accusative forms me and te are the clitic genitive-accusative
forms in Croatian. However, their old accentedness is seen in the fact that they can stand
with the prepositions alone: zá me, zá te, ná me, ná te, pó te etc. (always with the long rising
accent in Standard Croatian in disyllabic forms). However, in polysyllabic forms one finds the
following pattern: poʂdā te, uʂzā me etc. In Neo-Štokavian, the accentuation of the type zá me
is common. However, in Posavian (Old Štokavian), some Čakavian dialects and Kajkavian, one
finds the pattern zâ me, nâ te etc. Cf. for instance Bednja naʄu ma, naʄo ta, zaʄu ma, praʄd
ma114, Orlec zaʄ te, vaʄ me, poʄ me115 etc. Some Čakavian dialects (like Vrgada) have forms like
n°ã me, z°ã te, where the accent surely originates from the older *zā meʂ (like Neo-Štokavian
zá me). That is attested in Posavina as well (Ivšić 1971: [375])116.

In the instrumental singular, two forms are attested – mnóme (usually when alone),
and the original mnoʄm (usually with a preposition). In the 2nd person, there is toʂbōm
(tobome is quite rare nowadays but attested in ARj)117. The navezak –e is attested in mnome
since the 16th century (ARj). The origin of this –e is not quite clear118. Before the added –e,
only the form mnom existed (attested from the 14th century), which itself stems from the
older mnov (ARj) < *mnou < *mnoju < *mъnojNj (the same development as in the ā-stems).
In Čakavian, one finds the forms mnu (Antun Dalmatinac), maʂnū in North Čakavian (maʂnū <
*mъnǍ < *mъnNjNj ⇐ *mъnNjjNj ⇐ *mъnojNj), cf. Susak z maʄnu; in South Čakavian, the attested
form has no contraction of *-ojNj, as in Štokavian, but has the «strong vocalism»: maʂnōm.
Neo-Štokavian, with mnoʄm, mnóme (and toʂbōm, tobóme – the latter accent according to
Белић 1964: 104, which could be artificial), would at first glance point to the older mnõm
(and mnōmeʂ), but, in fact, in Posavina (Ivšić 1971: 370), only mnoʄm is attested119. The
accent mnóme, thus, must be secondary in Neo-Štokavian - due to the influence of njîm,
njíme, where the first form originally indeed was njĩm (cf. Posavina njĩm, njíme, but mnoʄm).
For the development of njîm, njíme, cf. also the genitive plural glávā (not *glâvā) from the
older form glãv120. This kind of development was generalized in mnoʄm, mnóme, in spite of

114 Jedvaj 1956: 306.

115 Houtzagers 1985: 102.
116 The type ú me also occurs in Posavina, but rarely.
117 Mareš (2001: 138) again has a nice theory that only mnóme exists, and no *tobome, *sobome,
because –e is supposed to have been used in order to produce a disyllabic form (which would be in
accord with the already disyllabic mene, meni etc.), and tobom, sobom was already disyllabic so there
was no need for this -e. That is not entirely wrong, but he disregards the fact that tobome was attested
at the same time as mnome – already in the 16th century (ARj). In addition, one cannot simply separate
the instrumental navezak –e from the locative one. Moreover, the forms tome, lijepome etc. exist in
instrumental, which is to say that –e appears in longer forms as well.
118 Белић (1964: 104) thinks that it stems from the particle –re/-r < *že.
119 In Hvar (Jurišić), however, besides toʂbon/teʂbon and meʂnon, the accent menõn is also attested.
120 When another syllable appears, the neo-acute sort of “goes back” to the final syllable. The basis for
these kinds of processes are, of course, regular patterns like pũt, pūtaʂ/púta etc.

the fact that there was no neo-acute there originally. Cf. also the accents saʂ mnōm, zà
tobōm; dialectally also zaʂ tobōm, cf. Vrgada zaʂ menōn, zaʂ tobōn, Oprisavci (Posavina, my
data) zaʂ sobom, but Prapatnice zà tebōn, Korenica (Jurišić) prè tobōm. In many dialects, the
genitive root has been generalized - meʂnōm instead of mnôm.

The unaccented forms of D. mi, ti have remained unchanged. Standard Croatian also
has the reflexive form si, unlike Standard Serbian.

In Križanić's grammar (1984: 110), one finds the following forms: N. Jâz, Jâ, GA.
Méne, Me: Menęɳ, Męɳ, DL. Méni, Mi: Mně, I. Ménoju: Mnóju and N. Tî, GA. Tébe, Te: Tebęɳ, Tę,
DL. Tébi, Ti: Tebě, I. Tóboju: Tobóju. The genitive-accusative forms Menęɳ, Tebęɳ, the dative
forms Mně, Tebě, and the instrumental form Tobóju are probably taken from Russian. His
native Croatian dialect seemingly had initial stress in the genitive, dative and instrumental.


§ 18 In Slovene, the explanation for the flection of GDAL is the same as in Croatian: the
dative-locative got the analogical men-/teb- from the genitive, the ending –i in the dative-
locative is to be explained in the same way as in the ā-stem nouns, and the old accusative
forms me, te function today as the clitic genitive-accusative forms. The accent is rising in
both of these forms, which is in accord with Serbian мèнe, тèбe, мèни, тèби.

The old accusative forms *mę, *tę are used in the combination with prepositions:
naʄme, naʄte, čeʖʄzme, čeʄʖzte, poʄʖdme, poʄʖdte etc.

The instrumentals menóʜj and tebóʜj (with the analogical root vocalism) point to the
older neo-acute at the ending - menóʜj ⇐ *mъnojNjʀ, which agrees with Old Russian (see § 27).
The other Slovene form, which occurs with the prepositions, preserves the old vocalism of
the root: z mâno < *sъ mъnǍ, but the accent is difficult to explain (see § 54). The form tâbo
is analogical to the form mâno.


§ 19 The genitive form *mene has been replaced with mne, by analogy to the mn- of the
dative-locative and instrumental, from the earliest times, as in all other West Slavic
languages. This new genitive form (as well as the unchanged tebe) has taken over the role of
the accented accusative, as elsewhere. The old accusative forms *mę and *tę > *mä, *tä >
mĕ, tĕ function as the unaccented forms. The dative-locative full form mnĕ is preserved, and
so are the unaccented dative forms mi, ti. However, the difference of the dative-locative mnĕ
and the genitive-accusative mne, mĕ exists only in literary Czech. In spoken Czech, all of
these forms are pronounced exactly the same: [mńe]121. Thus, in spoken Czech, like in
Polish, there is only one form in GDAL.

121 In the dialect of Prague and in Central Czech dialects in general, mĕ is generally pronounced [mńe].

In the dative and locative singular of the 2nd person, Czech has tob-, like Polish and
other West Slavic languages. The form tebĕ is attested in Old Czech dialectal texts122 and in
modern Czech dialects.

In Old Czech, short accusative forms were used with prepositions, for instance na tě.
This is impossible in Modern Czech, except in phrases like dojde na tě "it will come to you
(you will see what you did)".

In the instrumental, Old Czech form mnú (< *mъnǍ < *mъnojNj) has regularly yielded
modern mnou, while the original Old Czech tobú has taken the secondary root vocalism –e-
(instead of the original –o-) from the genitive-accusative form (thus, the modern form is
tebou). The analogical form tebú first appeared in the 14th century, and spread very quickly,
taking the older tobú out of the picture already in the 15th century. The oldest texts, from
around 1300, still had the regular tobú123.


§ 20 The original *men- in the genitive has been replaced, as elsewhere in West Slavic,
with mn- from the other oblique cases (dative-locative and instrumental). The clitic forms mi
and ti in the dative are preserved.

In the dative, mne represents the original form, while in the 2nd person, the root teb-
could have been imported from the genitive(-accusative), cf. Czech tobĕ. The other way to
look at this is that Slovak tebe with an –e- represents yet another isogloss that connects
(Central) Slovak with South Slavic (in South Slavic, only the form tebĕ is attested), and that
both forms existed (at least in some Slovak dialects) from the very beginning124. Pauliny
(1990: 151) believes that the form sobĕ (and thus also tobĕ), must have been more
widespread in Old Slovak. He thinks that personal names which contain this stem prove this.
He adduces the following names attested in Cividalský evanjeliár (9th century) on the
territory of the present day Slovakia: Sobĕmysla, Sobĕgojь, Sobĕslava, Sobina but Sebĕtuxъ,
Sebĕdragъ, Sebĕtĕxъ, Sebĕmyslъ. He also compares contemporary name Sebeslavce with
Hungarian Szoboszló (in 1230/1285 – Scebeslou, 1294 – Zebezlouch, but 1271/1284
Zobozlo and in 1234 zoboslaus, comes nitriensis). The forms sobe, tobe also appear in old
texts, with the attestations ranging from 1479 to 1665, according to Pauliny (1990: 152).
For the DL. forms tebe, sebe, he adduces examples as early as from the year 1380. There is
one more thing which points to the existance of the old tobe, sobe. In East Slovak, in GDAL.
one finds the forms tebe, sebe. However, one would expect, at least in GA. (where the –e-
must be old) the forms **ťebʆe, **sʆebʆe. The hard consonants in East Slovak tebe, sebe are
explained as an original trait of the old DL. *tobʆĕ, *sobʆĕ125.

122 Cf. Gebauer 1896: 528, Vážný 1964: 121.

123 Gebauer 1896: 529.
124 Mareš (2001: 136) thinks that these forms in Slovak can be regarded without any difficulty as one of
the numerous Slovak meridionalisms.
125 Fedák 1937: 74.

The old ending of the genitive-accusative -e in mne, tebe is preserved in East Slovak
dialects: GA. ťebe, sebe was attested in the speech of older speakers in the dialect of
Liptova126, and the GA. forms mne – me, ťebe/tebe – ťe/te are, for instance, attested in
Gemer127. This is due to the fact that in this area, the short nasal *ę yielded e (thus *mę, *tę
> me, ce). In Central and West Slovak dialects (and in Standard Slovak), where *ę yielded a,
the short forms have influenced the long ones and the ending –a has analogically become
the ending of the long forms as well (Slovak DL. mňa, teba). Thus, the same change occurred
in the genitive-accusative as in Russian (see § 27). In these dialects, the dative-locative
ending remained unchanged (i.e. mnĕ, tebĕ > mne, tebe), but in the genitive-accusative, the
original mne, tebe changes to mňa, teba (already from the 15th century, cf. Pauliny 1990:
153), analogically to the short forms ma and ťa (originally the accusative forms and now the
genitive-accusative clitics). Thus, the GA. and DL. forms remain different in these dialects
(otherwise GA. mne, tebe and DL. mně, tebě would yield GDAL. mne, tebe).
In the short forms, the process of change goes like this: mę > mä ⇒ ma and tę > ťä
> ťa. In most Central Slovak dialects, the form mä yields ma by analogy to ťa (this is still a
change in progress). In Central Slovak dialects, in which the form mä still occurs, there is
also ťebä, sebä128. The short accusative forms cannot be combined with prepositions in
In the instrumental, the original mn- has been preserved, since this has become the
stem of all oblique cases, while in the 2nd person, the original tob- was replaced by younger
teb- from the genitive(-accusative), as in Czech. This is hardly surprising since the form teb-
is generalized in all oblique cases in (Central) Slovak129. In West Slovak documents from the
15th century, both the forms sebou and sobu are attested. The forms tebu and sebu are
mostly found in Central and East Slovak, but the forms with to-, so- are attested in the
North and East of Slovakia, as well as in Gemer130.

The (Central) Slovak ending –ou looks identical to the Czech ending, but it is not.
While in Czech the diphthong is due to the diphthongization of the older –ú (which is derived
from the contracted ending *-Ǎ < *-ojNj), in Slovak, -ou is the regular reflex of the non-
contracted ending *-ojNj after the disappearance of the *-j- (*-ojNj > *-oju > -ou). In
dialects, one finds forms mnú and mnu besides mnou131.

Upper Lusatian

§ 21 In Upper Lusatian, the typical West Slavic development of generalizing the stem mn-
in all oblique cases is also present. The genitive and the accusative are merged in the usual
way. The old accusative forms mje, će are now short forms, while the old genitive forms
mnje, tebje are the full forms (in GA. of course). Unlike Czech, Polish and the second-half-
of-the-20th-century Lower Lusatian (see § 19, § 23, § 27), Upper Lusatian still clearly

126 Stanislav 1932: 315.

127 Orlovský 1975: 80.
128 Pauliny 1990: 151.
129 Unlike Czech, which has tob- in the dative-locative, and in Old Czech in the instrumental as well.
130 Pauliny 1990: 153.
131 Pauliny, ibid.

distinguishes the genitive-accusative forms from the dative-locative ones in both 1. and 2.
person: mnje, tebje and mni, tebi, in spite of the fact that one finds teb- and not tob- in the
dative-locative and mn- in all oblique cases. This is due to the fact that the ending –i < *-i
(from the feminine jā-stems, like zemja, DL. sg. zemi), instead of the original –je < *-ě has
been introduced to the dative-locative132. The short dative forms are also preserved.

In the dative form of the 2nd person, as already said, the stem teb- occurs. This could
have been introduced from the genitive-accusative instead of the older tob- attested in Old
Lower Lusatian (see § 22) or it could be old. The old stem tob- was however, in opposition
to Czech, preserved in the instrumental form tobu. Here, we can clearly see how levellings
can function in different ways in different languages – in Lusatian, the dative-locative was
analogically changed (if not old), in Czech – the instrumental, and in Slovak (if the dative
tebe is not considered to be a South Slavic isogloss) both the dative-locative and the
instrumental were changed analogically.

Lower Lusatian

§ 22 In Lower Lusatian, most forms and developments were similar or identical to Upper
Lusatian, except for the innovation which happened in the 20th century. In the older norm of
Lower Lusatian, from the beginning of the 20th century, represented in Šẃela 1952 (: 38-
39)133, the GDAL. full forms were not completely merged, but were still partly distinguished
in the 1st person. Thus, there was GA. mnjo but both mnjo and mnje in the DL. In the 2nd
person, these had already practically merged: GDAL. tebje, tebjo (Šẃela, however, does not
include the form tebjo as possible in the locative)134. In the short forms, there was mje, śi in
the genitive, and mĕ, śi in the DA. Thus, as we can see, these forms were still partly
differentiated, as in Upper Lusatian, but the merger had already started. In the younger
norm, from the second part of the 20th century, represented in Janaš 1976 (: 172-176),
these forms are already completely merged. In the GDAL. full forms (used after
prepositions), there are only mnjo and tebje, while in the short forms (used without
prepositions), mje occurs in the genitive and mĕ in the DA., while in the 2nd person, śi is
used throughout.

In the dative, as in Upper Lusatian, the stem teb- was either analogically introduced
from the genitive-accusative, or it represents a parallel form inherited from Common Slavic.
The form tobje is attested in Jakubica’s Lower Lusatian translation of the New Testament
from the first half of the 16th century135.

In the instrumental, the forms are mnu, tobu as in Upper Lusatian. The analogical
form tebu was attested as a dialectal form at the end of the 18th century136.

132 Vaillant 1958: 458.

133 This book was actually written in 1905.
134 For the sake of simplicity, I write nj and bj instead of ń and bɴ.
135 Muka 1891: 390.
136 Muka, ibid.

One more thing should be mentioned for both Upper and Lower Lusatian – as Mareš
(2001: 137) notes, the forms in teb- should really have the form **ćeb- in Upper Lusatian,
and **śeb- in Lower Lusatian (at least in the genitive-accusative where –e- is old) via usual
sound laws in these two languages. The forms in teb- are obviously analogous to the
nominative, old dative-locative variant with an –o- and instrumental forms, which all have t-
at the beginning.


§ 23 In Polish, the development of personal pronouns is similar in many ways to their

development in other West Slavic languages. The genitive and the accusative are unified in
the usual Slavic way, and via sound laws the genitive-accusative form mnie became identical
to the dative-locative form (the mn- is of course generalized in all oblique cases). The
difference of the dative-locative and genitive-accusative still exists in the short forms and in
the 2nd person singular, where the DL. still preserves the old West Slavic stem tob-, while the
stem *teb- > cieb- (not levelled to **teb- like in Lusatian!) appears in the GA. The
instrumental forms mną, tobą correspond exactly to Old Czech mnú, tobú (< *mъnǍ <
*mъnojNj and *tobǍ < *tobojNj).

Meillet (1934: 453) thought that the short genitive-accusative forms with no nasals –
mie, cie (thus not like like Standard Polish mię, cię, derived from *mę, *tę), which appear in
some Old Polish texts and in some Polish dialects, as well as in Kashubian, Slovincian and
Polabian - are old and that they should be derived from Proto-Slavic *me, *te, which are in
turn derived from the very same Indo-European forms137. These non-nasal forms were used
as clitic forms with verbs, while the nasal forms were used after prepositions. For instance in
Psałterz floriański (Old Polish) – na mię, but czemu mie męczysz, in Slovincian – (na-)mją,
but –mjä, in Polabian (no-)mang, but –me. However, there is probably no reason to assume
that these forms are archaic and that they stem from a Common Slavic variant *me, *te,
rather than *mę, *tę (as attested elsewhere). These non-nasal forms are probably to be
attributed to the influence of the full GA. forms (originally genitive forms), which had no
nasal to begin with – thus mie by analogy to mnie138. These were subsequently confined to
the use with verbs, while the older nasal variants were still used with prepositions.


§ 24 The forms are generally in accord with the typical West Slavic development. GDAL.
have the same form in the 1st person as in Polish (with a generalized *mn- > mɴ-). In the 2nd
person, the distinction is maintained due to the preservation of the old *teb- > ceb- and
tob- difference. In Kashubian, the form tobɴe is the only one in the dative, but in the locative
both tobɴe and cebɴe appear. The latter form might be an innovation (due to the influence of
the other cases), or an archaism inherited from Common Slavic. Slovincian is even more
radical in this way (see § 25), there the *to- form is not even attested in the locative. The old

137 Cf. also Schmidt 1978: 50.

138 Vaillant 1958: 444, Aitzemüller 1991: 108.

forms mnNj and tobNj are preserved in the instrumental (cf. Polish mną, tobą). The short
forms are also preserved, and the accusative short forms can still be used with


§ 25 The characteristic of Slovincian is that all forms, except the instrumental, have an
accented and an unaccented variant – including not only the locative (which has no short
form elsewhere in Slavic), but the nominative as well. The forms and development are,
except for this fact, mostly the same as in Kashubian (the same form in the GDAL. accented
form in the 1st person with a generalized *mn- > mj- , the difference of *tob- and *teb- in
the D. and GA. in the 2nd person etc.).

We have seen in Kashubian that besides the form tobɴe in the locative, the form cebɴe
also exists. Slovincian is more consistent here in a way, and the form cìeʠbjä is the only one
that appears in the locative. This is either due to the analogical development in the locative,
or to the generalization of one Common Slavic variant in the locative and the other in the
dative (theoretically, it could be an archaism as well, from Common Slavic D. sg. *tobě, L. sg.
*tebě). Due to this curious development (whatever its origin ultimately is – an analogy or
inheritance), Kashubian and Slovincian are one of the rare Slavic languages/dialects that
distinguish the dative and the locative of the personal pronouns. This kind of secondary
distinction exists for instance in the Croatian dialect of Vrgada as well, see § 17.

In the unaccented dative forms, one finds both cä and tä in the 2nd person. The latter
form has the t- by analogy to tșɳȅʠbjä140. This is proved by the fact that we find this
secondary tä only in the dative, where the accented form begins with a t-, while this
unaccented form does not appear in the GAL., where the accented forms have a c- at the
beginning. This analogical development is either not very old, since the t- does not appear
in the locative (if the *te- form is not old there), or it can be regarded as evidence for the
originality of *te- in the locative, i.e. that the *te- form is inherited from Common Slavic.

It is interesting to note that the unaccented genitive form is really the unaccented
dative form in origin (not the old accusative form as usual elsewhere!). This is the situation
we find in both 1st and 2nd person – mjä and cä in the GDAL. The origin of this must be in the
fact that the form mjä is the regular outcome of both the genitive *mъ/ьne and the dative
*mъ/ьnĕ. Thus, the form mjä functions as both the genitive and the dative form due to
phonological merger. The double function of cä is due to analogy to the 1st person141. The
forms mjä and cä have shown to be very aggressive and have penetrated to the accusative as
well (cf. the old accusative short forms mją, cą).

139 Cf. Lorentz 1925: 156.

140 Cf. also Lorentz 1903: 275.
141 Lorentz 1903: 275.

An interesting point is the final accent in the instrumental case mnoɺųʠ and tȅboɺųʠ
which could indicate the old neo-acute (and thus old *-ojNjʀ), like Slovene (but not
necessarily, cf. § 54).


§ 26 Concerning Polabian, the unreliable orthography and the lack of data are the biggest
problems142. In Polabian, like in Slovincian, unaccented forms exist even in the nominative.
The development of the forms, as far as it can be judged from the attestations, does not
differ much from the average West Slavic state.

In the genitive-dative-accusative (old genitive forms are used for accusative of

course), both the forms with final (miné, mané, tibé) and with initial accent (mánǩ, tíbǩ) are
attested. In the 1st person, man- is the expected result of old *mъ/ьn-, while min- must be
analogous to the 2nd person tib-. One would expect i < *e in front of a labial like in síbe <
*sebe, zímɴa < *zemja, but not in front of a *-n- like in zéna < *žena143. Thus, this is not an
example of the preserved stem *men- in West Slavic. Since the dative and genitive forms
have more or less merged in Polabian144, it goes without saying that the stem *tob- is not
attested in the dative-locative forms – it was either replaced by the analogical *teb-, or
inherited from Common Slavic. Cf. also the cases of Kashubian, Slovincian, Slovak and
Lusatian where the same problem occurs. The original stem *tob- is preserved, though, in
the instrumental: tåbNjɺ (cf. Polabian vådƮ < *voda for the development *o > å)145.

The old accusative mą is still used after prepositions (the 2nd person form is not


§ 27 Old Russian forms were basically the same as OCS ones, except for the fact that in
the dative-locative the form mъnĕ was the only one and that besides tebĕ, there was also the
form tobĕ (as in West Slavic). The form mъně, unlike in OCS, is in accord with the form tobě
(both ъ and o being back vowels).

In Ukrainian and Byelorussian, we find in the 2nd person dative-locative only to-, as in
Czech or Polish. However, in Modern Russian only the form тебé exists. In Old Russian,
besides тебΟ, the form тoбΟ is also attested. Both тeбΟ and тoбΟ appear from the earliest
records - for instance in the same text from 1073146.

142 For instance, cf. Polabian forms adduced by Schleicher (1871: 259) and by Ceлищев (1941: 440),
which are not completely alike.
143 Cf. Schleicher 1871: 47, 49.
144 Ceлищев 1941: 440, but cf. Schleicher 1871: 260.
145 Schleicher (1871: 262) attributes the form tåbNjɺ to the analogy with the form månNjɺ, although it is
not clear why, since both *ъ and *o would yield the same result here.
146 Meyer 1923: 147.

According to Дypново (1924: 258), the forms тебΟ and cебΟ in Old Russian are due
to the OCS influence, which he tries to prove by pointing out that in some of the Old Russian
texts, like in Aрхангельско евангелие (1092), where Ο is normally clearly distinguished from
e, -Ο is written consistently in тoбΟ, coбΟ, but not in тебΟ, cебΟ, which also appear as тебe,
cебe. This is indeed a very good argument and this proves that the forms тебΟ, cебΟ were
really OCS forms in some Old Russian texts. However, this does not mean that it was so in all
texts/dialects, and the question is not very important anyway.
Since the forms with –e- appear in most of the modern Russian dialects, we know for
a fact that this must be a native Russian form, not an OCS loanword. The thing that is not so
clear is the question of whether this тебΟ in Russian is just a product of a later analogy to
the genitive-accusative forms (which would mean that only the form *tobĕ existed in Russian
originally), or whether both forms are equally original in Russian (which would mean that Old
Russian has inherited both *tobĕ, like West Slavic languages, and *tebĕ, like South Slavic
languages and probably some West Slavic languages as well, from Common Slavic147). This is
a difficult question to answer since this kind of innovation, the analogical spread of –e-
instead of –o- to the dative-locative singular, might have occurred very early, like the West
Slavic innovation of *mъ/ьne instead of *mene148. If *tebĕ is a Russian innovation (and not an
inheritance from Common Slavic), that must have been a very early change indeed, since, as
already said, both of these forms are attested in Old Russian from the earliest records. It
would seem more economical to assume that the form with the *-e- in the dative-locative
was inherited rather than recreated anew by analogy149. However, according to Mareš (2001:
135-136)150, the first reliable attestation of Old Russian (and not Church Slavic) тебΟ, cебΟ
stems from the 14th century. If this view is to be trusted, the –e- forms might be regarded as
an internal Russian innovation, although this still remains controversial151.

According to Kiparsky (1967: 133), the form тoбΟ was more frequent than тебΟ in
Old Russian. However, since the 14th/15th century, the dative-locative form with an –e- starts
gaining power, which results in a complete disappearance of the –o- dative-locative form by
the end of the 17th century in most dialects.
Черных (1962: 218) thinks that the forms with –e- suppressed the forms with –o-
due to the literary tradition of OCS. This may be true in explaining the victory of the –e-
forms in the literary language, but since the same thing also happened in most dialects, the
influence of the literary tradition of OCS was hardly very important.
The –o- forms have, however, managed to survive in certain dialects – cf. the forms
тoбé, coбé in North Russian, and тaбé, caбé in South Russian152. The tendency to unite the
stems of the DL. and the GA. was strong and it resulted in the victory of the тe- forms in

147 That is the opinion of Eckert, Crome & Fleckenstein (1983: 140), for instance.
148 *mene with an *-e- appears nowhere in West Slavic, not even in the earliest records.
149 There is another angle to this story, Kiparsky (1967: 133) believes that the form *tobĕ was originally
a locative, while *tebĕ is originally a dative. For the discussion, see § 47, § 72.
150 Cf. also Филин 1972: 430-431 and Бopковский & Кузнецов 1965: 232-233.
151 Mareš (ibid.) himself says: “It is rather difficult to say, if Sorbian and Russian forms teb-/seb- DL sg.
can be interpreted as a belated branch of a tendency whose root was in the South Slavic territory, or if
we are dealing in this case with a quite independent and typologically parallel phenomenon (…)“.
152 Бopковский & Кузнецов 1965: 229.

Russian. However, before the final winner was clear, the opposite scenario has occurred as
well. The тo- forms were attested in the GA. since the 14th century, but not too often (this is
attested even today in dialects). The first occurrence of тoбe instead of тeбe in the GA. is
from the year 1382153.

In the merged genitive-accusative154, an unusual ending is found: меняɴ, тебяɴ.

However, these forms are really not that difficult to account for. As explained already by
Vatroslav Jagić, these forms are due to the merging of the original менe, тебe with the short
forms мя, тя < *mę, *tę. Mareš (2001: 136) explained why this has happened: «The
phonemes /e/ and /ĕ/ were merging in Russian, and in the XVth century this coalescence was
already accomplished in most of the Great Russian dialects. Accordingly, the genitive-
accusative and the dative-locative forms of the pronouns ty and sę became homophonous:
tebe, sebe. This undesired lack of communicative clearness led to the assumption of the
ending of the shorter forms tę > tja, sę > sja (A sg.): tebé/sebé GA sg. X tja/sja GA sg. >
tebjá, sebjá. All other endings having been the same with all three pronouns, the pronoun ja
was adapted in the framework of the whole system: mené GA sg. X (mę >) mja A sg. >
menjá, like tebjá, sebjá (though there was no homophonous fusion in genitive-accusative
and in dative-locative: mene vs. mne)»155. The present day form меняɴ was first recorded in
1383 and тебяɴ in 1461-1464. In the 16th century the new forms grew strong, and by the 17th
century other competitive forms were completely and utterly gone156.

The short forms have disappeared altogether from Modern Russian (and from most
modern East Slavic dialects in general) in 16th/17th century. Old accusative forms, мя and тя,
were still used both in accented and in unaccented position in 13th/14th century (Kiparsky
1967: 134). According to Eckert, Crome & Fleckenstein (1983: 141), the forms ми and мя
were used in Moscow up until 17th century157, but according to Бopковский & Кузнецов
(1965: 232), these unaccented forms were used only in traditional formulas already in the
15th century. Before they disappeared, these forms were reduced, cf. Modern Russian
reflexive –cь and –cя with verbs. The short forms are attested in some Russian dialects, but
not all of them represent a continuation of the old short forms.

In instrumental, the forms мнóй and тобóй represent an innovation, attested in

Russian (and in Byelorussian) from the 13th/14th century158. The older forms with the final –
ю, мнóю and тобóю, are today used only as poetic forms.

Concerning the accent in Old Russian, the situation is a little bit different from that in
Modern Russian. Modern Russian forms like меняɴ and тобóй (тобóю) are not really relevant
by themselves, since the first form is innovative and the accent could be secondary as well,
and in the second one the ending is shortened and thus the accent would have to be

153 According to Kyзнецов 1959: 122.

154 The original accusative forms have disappeared altogether in Modern Russian.
155 Cf. also Vaillant 1958: 456.
156 Kiparsky 1967: 131-132.
157 Cf. also Черных 1962: 219.
158 Mareš 2001: 138.

retracted even if it was not originally on the first part of –oю. In Old Russian159, the forms of
the genitive and dative, мéне, тéбe, мъɴнΟ, теɴбΟ/тóбΟ, were enclinomena, i.e. stem stressed,
although they had variants with the end-stress (which corresponds to modern меняɴ, тебяɴ). In
instrumental, the accent was on the final syllable – мнoюɴ and тобoюɴ.


§ 28 In Ukrainian, the historical development of the personal pronouns oblique singular

cases was much simpler than in Russian. Most of the forms represent just a regular outcome
of the original North Slavic forms. The only exception is the dative-locative of the 1st person
pronoun, where the old мн- was analogically changed to мeн- by the influence of the
genitive-accusative form (cf. Russian мнé). The old form мнí is preserved in some West
Ukrainian dialects160.
In the dative-locative of the 2nd person, the old North-West Slavic –o- is preserved.
Unlike Russian, there was never any danger of the coalescence of the genitive-accusative
and dative-locative form in the 2nd person, not only because of the теб-/тoб- difference (the
мeн-/мн- difference is gone in the 1st person though), but also because *ě and *e did not
merge in Ukrainian, unlike Russian.
Mareš (2001:137-8) notes a couple of interesting points concerning the Ukrainian
dative forms. The stem мeн- in the dative-locative in Ukrainian is the only such example in
North Slavic (Russian and Byelorussian have the old мн-), and, more interestingly, that is the
only example in Slavic of the stem men- in the dative-locative, while there is no parallel teb-
(and seb-) in the dative-locative. Mareš interprets this development in the following way: the
Ukrainian form мнí was the only case of an isolated monosyllabic non-nominative form. It
was the only monosyllabic form, and thus it was analogically made bisyllabic. In Russian and
Byelorussian, the 1st person dative-locative form was not an isolated monosyllabic non-
nominative form, since there the instrumental form was either mono- or bisyllabic161, so
there was no ground for this kind of change.

In both the genitive-accusative and dative-locative, the accent is desinential in

Modern Ukrainian, like in Russian. In the instrumental, the old final accent was replaced by
the accent on the first syllable of the ending, in correspondence to the ending –óю in ā-stem
nouns. The original dative and accusative short forms have, like in Russian, disappeared
from Ukrainian. They are however preserved in some Ukrainian dialects (Meyer 1923: 146).


Byelorussian is the most archaic East Slavic language when it comes to the
preservation of personal pronouns. In Byelorussian, only the regular sound change has
occurred - there were no analogical changes whatsoever, the only minor exception being the
shortening of the instrumental forms, мнóй and тaбóй, like in Russian (the old long forms
мнóю, тaбóю are poetical variants). The old *ě and *e have merged in Byelorussian just like

159 Зализняк 1985: 143.

160 Mareš 2001: 137.
161 Cf. Ukrainian мнóю but Russian, Byelorussian мнóй/мнóю.

in Russian, but there was no merger of the genitive-accusative and dative-locative forms
since these are clearly differentiated by the stems (GA. мянé - DL. мнé, GA. цябé – DL. тaбé).
The accent of the genitive-accusative and the dative-locative is on the second syllable, like
in Modern Russian and Ukrainian (but unlike Old Russian). The short forms have
disappeared, as in Russian and Ukrainian.

Personal pronouns in plural

Slavic *my and *vy

§ 29 The reconstruction of Common Slavic *myʄ and *vyʄ, like that of *tŷ, is not
problematic. The correspondences are quite clear.

The circumflex in *mŷ and *vyʄ is confirmed by Slovene mî, vî, Čakavian (for instance
Novi, Orlec) mî, vî, Kajkavian (Bednja) meʄi, veʄi, Czech my, vy, Slovincian maɺ, vaɺ, Old Russian
мыɴ, выɴ - which are «enclinomena»162 (cf. иɴ мы, иɴ вы, мы жè, вы бò163), and Middle Bulgarian
мы жè, вы жè164.
Old Štokavian, South Čakavian and Kajkavian (most dialects) mĩ and vĩ are secondary,
the result of the secondary spread of the neo-acute from jã (thus also secondary tĩ, see §
14). In Neo-Štokavian, the neo-acute original of mî, vî is seen in ì vī, ì mī (and not **ıʂ mī, **ıʂ

The reflexes of the original my and vy have been preserved in most languages. The
only exceptions are Bulgarian ниɳe and Macedonian ниe, which have the n- from oblique
cases and a secondary ending by analogy to the demonstrative forms like тие (< *ti-ję with a
*ję originally from the accusative)165.
The form ny instead of my is also a feature of the dialect of the Kiev Folia, where it is
the only form attested (7x). The form ny instead of my appears also once in the OCS text
Glagolita Clozianus, but there it can be explained as writer's error166.
The form ниɳe is attested in Bulgarian since the 13th century167. The form мие with the
old м- appears in the Macedonian dialect of Mijaci.
In Slovene, the new separate feminine plural forms were made by analogy to the
nominal o- and ā-stem endings: mî, mêʜ, vî, vêʜ168.
In Polabian, the form jai instead of the expected *våi is attested. This is probably a
loanword from Lower German jî169.

162 Зализняк 1985: 143.

163 Дыбo 1981: 35, 51, Дыбo 2000: 62.
164 Дыбo 1981: 51.
165 Георгиев’s sandhi explanation *my esmъ > *mye smъ and *vy este > *vye ste (Xapaламниев 2001:
108-9) is not very convincing.
166 Cf. Weingart 1937-8: 202.
167 Xapaламниев 2001: 109.
168 The same thing happened in Slovincian and Kashubian in dual nominative forms.
169 Cf. Schleicher 1871: 262, Vaillant 1958: 452. Polabian jai could also be derived from older Slavic *ji.
Curiously, this is the very form we would expect to get from the IE *yūs (the inital *y- is otherwise lost
in Slavic, cf. § 48). However, it is highly unlikely that this kind of archaism would be preserved in
Polabian only (Vaillant, ibid.).

Oblique cases of 1st and 2nd person plural

The forms of the 1st and 2nd person plural are always the same, except for the initial
consonant (m-/n- : v-). The development of the plural oblique forms in Slavic languages is
simpler than the development of the singular oblique forms.

Old Church Slavic

§ 30
30 The genitive and locative forms are identical in OCS (and in other Slavic languages).
This is due simply to the phonetic merger of the old endings *-som and *-su in Slavic (see §
The old accusative forms were ny, vy. In modern Slavic languages, the genitive forms
nasъ and vasъ have taken over their place completely (due to the general tendency in Slavic
to merge pronominal genitive and accusative forms). This process is seen already in OCS,
where nasъ and vasъ also appear instead of the expected ny and vy in the accusative. This is
a clear later innovation in OCS texts.
In some OCS texts (Euchologium Sinaiticum, Codex Suprasaliensis, Glagolita
Clozianus), the forms ny and vy were also used as the unaccented dative forms. This feature
also appears in other Slavic languages – like Macedonian and Bulgarian, Old Russian, Old
Croatian, some Montenegrin and Serbian dialects etc.


§ 31 As already said, Bulgarian has lost most of its declension. Thus, not many original
forms are preserved. The following forms are preserved: old genitive-locative forms нàc,
вàc, which are now general oblique forms; the old accented accusative/unaccented dative
forms ни and ви, which are now used as the unaccented dative-accusative forms (the old
genitives нàc, вàc are used as the accented accusative, like in most other Slavic languages);
and the old dative forms наɳм and ваɳм, which are today archaic (unlike in Macedonian) and
instead of which the new analytic forms на нàc and на вàc are used.
We can find out more from Middle Bulgarian (Hock 1992 I: 119-120), where we find
the forms practically identical to those of OCS. The form corresponding to Croatian dial.
naʂmi, vaʂmi and Czech námi, vámi is нáми, вáми. In Evangelie 1139 (Hock 1992), one finds
only secondary на ныɴ, на выɴ in the accusative. The old forms like наɴ ны are facultatively
preserved in other Middle Bulgarian monuments (Дыбo 1981: 50).


In Macedonian, the situation is similar to Bulgarian but not the same. The accented
forms, the dative нам and вам (preserved in Modern Macedonian, unlike Bulgarian, where
they are archaic) and the accusative/oblique нac, вac (originally genitive) are usually doubled
with the unaccented forms. In the dative, the unaccented forms are ни and ви, but in the
accusative, new forms нè and вe have been formed analogically (like in some Montenegrin


§ 32 In Neo-Štokavian, the original declension was changed in many ways. The older
declension pattern was preserved in many Čakavian and Kajkavian dialects.

The old genitive-locative forms *nasъ, *vasъ are now the genitive-accusative forms,
which can be both accented and unaccented. These forms are attested as the accusative
from the 15th century. When accented, some northern Čakavian dialects (for instance Cres,
Susak), Kajkavian (for instance Bednja GAL. noʂs, voʂs) and some South-West Štokavian
dialects have the original forms naʂs, vaʂs, while South Čakavian and Posavina Old Štokavian
have the secondary neo-acute here - cf. Posavina nãs, Hvar nõs. This form is also the source
of the Neo-Štokavian naʄs, vaʄs, nà vās, zà vās170. The original accent is still seen in Štokavian
possessive pronouns naʂš and vaʂš. The source of the secondary neo-acute in Štokavian and
South Čakavian nãs, vãs must be sought in the pronominal forms like njĩh, teɶʖh (> tîh, tıʂjeh
etc.), and in the frequency of the neo-acute in the genitive plural in general – cf. nominal
genitive plural forms like kostĩ, rukũ, and in Štokavian also glāvã171. This change is probably
not very old in Neo-Štokavian – according to Rešetar (1911: 212), in Molise these forms have
the old accent: naʂs, vaʂs. The Molise Croats came to Italy in 15th/16th century, which means
that in that time the forms naʂs, vaʂs were still used in Neo-Štokavian - at least in the territory
between the rivers Cetina and Neretva, where they came from. Today, only the forms naʄs,
vaʄs are attested there. Thus, the innovation naʂs, vaʂs ⇒ nãs, vãs (naʄs, vaʄs) was probably a
later isogloss, which has encompassed most of Štokavian and South Čakavian.

In the standard language, and in many Štokavian dialects, the original dual forms
naʂma and vaʂma are used for the dative-locative-instrumental plural. The form vama in the
dative, for instance, is attested from the 15th century (ARj). In the locative, nama, vama
appear from the 17th century, and in the instrumental from the 14th century172. The forms
nami, vami in the dative appear from the 14th century, and in the locative from the 17th
It is clear that the system was not stable from the very beginning. In the instrumental,
for instance, the forms nam and vam instead of nami, vami are attested as early as the 13th
century174. The use of nas, vas in the locative is preserved mostly in Kajkavian and Čakavian
dialects. In Štokavian, the use of nas in the locative has been declining from the 17th century
(ARj). The old dative form nam is still used as the unaccented dative form175. Old
instrumental forms naʂmi and vaʂmi are also used as dative-locative-instrumental forms in

170 In the light of Posavina nãs, vãs, it is clear that Stang’s (1957: 106) and Дыбо’s (1981: 35, 2000:
63) claims that Neo-Štokavian naʄs, vaʄs is the result of the old neo-circumflex are blatantly wrong.
Kajkavian Prigorje genitive vaʄs, but locative vaʂs (Rožić 1889-94/116: 130) is an example of a real neo-
circumflex (see § 55 for the discussion). Cf. Bednja GAL. noʂs, voʂs.
171 Белић’s (1964: 106) argument that Štokavian naʄs, vaʄs is older than Čakavian naʂs, vaʂs makes no
172 Jurišić 1992: 152.
173 Белић 1964: 105.
174 Jurišić, ibid.
175 The variation –ma : -m is common in the plural of pronouns and adjectives in Croatian.

many Štokavian dialects (for instance in Vrgorska krajina). In Posavina176, the forms naʂmī and
vaʂmī (analogically also naʂmā) are attested, which have the long –ī by analogy to the –ī in the
locative-instrumental of the o-stems (zubĩ, voʂlī with the –ī originally from a. p. b *-īxъ, a. p.
c *-ĩxъ). This is in accord with Slovene naʄmi, vaʄmi and Slovak dialectal namí, vamí (see § 33,
§ 34).
According to ARj, the form ni was never used for the accusative, except due to the
influence of OCS177. The form ni as the dative enclitic is attested in Boka kotorska and in
some Montenegrin and East Serbian dialects. The dialects which have enclitic dative ni
usually have the new accusative ne (like in Macedonian), due to analogy with nominal
declension (ne like kònje) and other pronouns (accusatives teʄ, njeʄ, sveʂ etc.).
Unlike the form ni, attested only in modern dialects, and not in old texts, the
unaccented dative vi is well-attested in Old Croatian, as well as the accusative vi (ARj). From
the 15th century, the old genitive form vas is used instead.


§ 33 In Slovene, the development of the plural oblique forms was rather simple, unlike
Croatian. The old genitive forms are used for the accusative, as usual, and from the accented
forms, the new clitic forms in the genitive-accusative and the dative were made.

Slovene nàs, nàm is in accord with Čakavian naʂs and Czech nás, nám. The neo-
circumflex in the instrumental naʄmi, vaʄmi can be explained in a similar fashion as Croatian
(Posavina) naʂmī, vaʂmī. The reason for the neo-circumflex (thus a former long *-ī) is to be
found in Slovene forms like mîšmi, obraʄzi, leʄʖti, meʖʄsti, goraʄmi, možmí, kostmí etc. The
reason for the long *–ī in these examples is in the compensatory lengthening of *-Cь/ъmi >
*-Cmī, and in the regularly preserved length in the locative forms like mîših, leʖʄtih, možéʖh
(instead of **možíh) etc. Under the influence of the long *–ī in some instrumental forms (in
i- and u-stems), and the length of the ending *–īxъ in the locative, the long *–ī was
generalized in Slovene.


§ 34 In Czech, the forms remained practically unchanged except for the usual Slavic nás,
vás in the accusative. The long –á- in all forms is in accord with Slovene and northern
Čakavian dialects.


In Slovak, the original genitives nás and vás function also as the accusatives.
However, older accusative forms may have been attested in Slovak dialects178.

176 Ivšić 1971: [374].

177 Križanić (1984: 110) has the forms Nás: Nì and Vás: Vì in the accusative plural. However, the short
falling accent in these forms (cf. the long accent in the nominative forms Mî, Vî) points to the
conclusion that these forms were artificial in his language and not a part of his native dialect.
178 Pauliny 1990: 155.

In the instrumental, the forms nami, vami are expected, compared to Czech námi,
vámi and Croatian dialectal naʂmi, vaʂmi. However, the long vowel in the forms nás, vás, nám,
vám cannot be explained so easily (the acute yields short syllable in Slovak). In dialects,
regular nas, vas also appears179. An ad hoc solution would be some kind of sporadic
lengthening, maybe originally only in nám, vám in front of a sonant (cf. sám, sama - where
sám is secondary for *sam). Another possible solution would be an analogy to the
possessives náš, naša, naše and váš, vaša, vaše, which have the length only in the
monosyllabic forms (where it is due to analogy with môj, tvôj), and a short syllable in the
disyllabic forms (just like nás, nám, nami)180. The borrowing of these forms from the Czech
literary language is excluded, since the long forms appear in dialects as well (the same thing
is attested in Old Polish, see § 35).
In the instrumental, further variation is attested in dialects – in the east of Slovakia,
besides or instead of the form nami, the older dual form, nama, is used (like in Croatian). In
Central Slovakia, the form namí is used181, which is to be explained in a similar fashion as
Croatian (Posavina) and Slovene forms.


In Lusatian, one finds no special developments except for the usual use of the
genitive-locative nas, was in the accusative as well.


§ 35 The same that goes for Lusatian, goes for Polish as well - Lusatian and Polish plural
forms of the personal pronouns are exactly the same. The old accusative ny is attested, as
usual in Polish, as a hapax182.
In Old Polish Bible of Queen Sophia (Biblia królowej Zofii) from the 15th century,
before the elimination of the distinctive length, one can see forms <nas>, <nam> but
<náma> (with <á> being [a], and <a> being [å]), which is in accord with Slovak nás, nám,
but nama. These long forms are difficult to account for, but they must be a younger


The old genitive-locative is, as usual, used in the accusative as well. In the dative,
besides the original nNjm and vNjm, the original instrumental forms namï and vamï are used
in some dialects, and in the case of the 1st person, the older dative-instrumental dual form
nama is also attested for D. pl. (like in Croatian and Slovak dialects). In the instrumental,
variation is also present, and the form nama is used instead of the original namï in some

179 Pauliny 1990: 154.

180 Siniša Habijanec (p.c.).
181 Pauliny, ibid.
182 Klemensiewicz, Lehr-Spławiński & Urbańczyk 1955: 322.


As usual, the original genitive-locative forms nãs and vãs are used in the accusative
as well. The new enclitic forms năs, văs have developed, and the same thing happened in the
dative. Unlike Kashubian, the old plural forms are preserved in all cases without secondary


No special development here, except for the regular sound laws.

East Slavic

§ 36 Russian, Ukrainian and Byelorussian have the same plural oblique forms. Except for
the usual use of the old genitive-locative forms in the accusative, nothing else changed. The
original accusative forms ны and вы were used in Russian texts untill the 17th century, but
the use of the genitive form instead of it is attested from the 13th for васъ, and from the 14th
century for нacъ183.
According to the statistical analysis of Caмойленкo (1962), the forms ны and вы were
more common in Old Russian than the forms нacъ and васъ in the accusative, but in the
dative they were far less common than the forms нaмъ, вамъ.

The GDLI. had an initial stress in Old Russian, as in all modern East Slavic languages.
This points to the old acute, cf. Old Russian над нáми, оɺ вáсъ184.
In the old accusative forms, the accent “jumps” to the preposition which points to the
old circumflex (*nyʄ, *vyʄ): Old Russian нá ны, нá вы185.

183 Kiparsky 1967: 137.

184 Дыбо 1981: 35, Дыбo 2000: 62-63.
185 Дыбо, ibid.

Personal pronouns in dual

The reconstruction of the dual forms is a little bit more difficult, because the dual
forms were less well attested than the singular and plural forms, and because of the non-
existance of this category in most modern Slavic languages. However, the dual forms do
appear in older phases of all Slavic languages attested sufficiently early. The only modern
languages that preserve the dual are Slovene, Lusatian and Kashubian (as well as now extinct

Old Church Slavic

§ 37 Already in OCS, the dual shows quite a fluctuation of forms. In 1st person nominative
dual, besides the form vě (very well attested), a secondary na, originally an accusative form,
appears, though very rarely. In the 2nd person nominative dual, forms va and vy appear. The
second form is identical to the 2nd person nominative plural form – that, however, does not
mean that these forms are of the same origin (see § 48, § 50). The form vy in the nominative
dual is much more frequent than the form va186, and it is consistently used in older glagolitic
texts Codex Zographiensis, Codex Marianus and Codex Assemanius187. The nominative dual
form va, identical to the accusative dual va, is attested in cyrillic texts Codex Suprasaliensis
and Sava’s Book188.

In the 1st person accusative dual, besides the original na, younger naju (originally a
genitive) appears due to the overall tendency of the pronominal genitives to become the
accusatives189. Originally plural forms ny (accusative) and nasъ (originally genitive plural and
secondary also accusative plural) also appear as secondary A. dual190.
In the 2nd person, instead of va, the original genitive dual form vaju appears, as well
as vy which is either the nominative dual form or the accusative plural form.
Use of ny, vy for the accusative dual was consistent in the older glagolitic texts
Codex Zographiensis, Codex Marianus and Codex Assemanius; the forms na, va are attested
in younger, cyrillic texts Codex Suprasaliensis and Sava’s Book191.

In Codex Suprasaliensis, in the 2nd person dative dual, the unaccented form va is
used besides the accented vama, like vy and vamъ in plural192. The form na in this function is
not attested in OCS.

186 Cf. Mareš 2001: 177.

187 Nandriş 1959: 104.
188 Diels 1932: 214.
189 Cf. also in the singular mene instead of mę, in the plural nasъ instead of ny etc.
190 Mareš (2001: 178) believes that the form na was in the process of elimination because of the
similarity with the preposition na. However, since the pronoun form was stressed, this could hardly be
a major factor in the elimination of the dual accusative na.
191 Diels, ibid., Weingart 1937-8: 202, Kurz 1969: 77.
192 Diels, ibid.

Middle Bulgarian

In Evangelie 1139 in Middle Bulgarian (Hock 1992 I: 119), one finds the forms: вΟ,
наɴю, нáма, нá and ваɴю, вáма, вá (the 2nd person nominative dual is not attested in this


§ 38 The dual does not exist in modern Croatian (in any dialect), but the dual forms have
influenced the plural of the literary language. The original dative-instrumental dual forms
naʂma, vaʂma are used as the dative-locative-instrumental plural forms.

The forms *vě, *na and *va are unattested even in Old Croatian according to ARj (but
see below for Bednja)193. The forms naju, vaju are found in the texts up untill the 17th
century, not only in the original function of the genitive-locative dual, but also as the
accusative dual and instead of the plural forms (ARj). There is no attestation of the
accentuation of these forms, except for Križanić’s Náju, Váju (1984: 110), i.e. naʂju, vaʂju.

In the archaic Bednja Kajkavian dialect194, a trace of the old dual va > Bednja –vo had
been preserved in the speech of very old people (in the middle of the 20th century) as a
particle added to the plural pronouns meʄi and veʄi when talking about two people: mıʂvo dvaʄo
“us two”, vıʂvo dvaʄo “you two”195. The preservation of this –vo here is remarkable. It is not
clear if this –vo < *va is originally the old N. dual of the 1st person (like va in Old Czech and
Old Polish for instance, cf. § 38, § 40) or the old A. dual of the 2nd person.


§ 39 Modern Slovene still preserves the old dual forms, but with certain changes. In the
nominative, the old forms were replaced by the new agglutinative forms mîdva,
mîdve/mêʜdve, vîdva, vîdve/vêʜdve (like in Lithuanian), which differentiate gender, like plural
forms. The old genitive-locative forms naʄju and vaʄju have spread to the accusative (as in
plural). In the locative, both the old forms naʄju and vaʄju and the new naʄma and vaʄma
(originally dative-instrumental) can be used. In the genitive-accusative and dative, new
atonic forms naju, vaju, nama, vama have developed.

Both naʄju, vaʄju and naʄma, vaʄma have a neo-circumflex in Slovene. In naʄma, vaʄma,
the neo-circumflex is to be explained as in the plural – the length of the final *-ma is due to
the compensatory lengthening in old i- and u-stems (*-Cь/ъma > *-Cmā), like mîšma, and
due to analogy to the plural forms naʄmi, vaʄmi (see § 33). The forms naʄju and vaʄju probably
have the neo-circumlfex analogically to naʄma, vaʄma. One could also imagine an analogical

193 Vaillant (1958: [252-253]) says that “le vieux serbo-croate” had these forms, but it is not clear what
he means by that, since the existence of these forms is clearly rejected in ARj.
194 Jedvaj 1956: 306.
195 It is difficult to explain the accent ʂ here. It is hard to believe that this could be a pre-Meillet old

development similar to the development of the long –ú in najú, vajú in Old Czech (see § 40),
and of the long -ū in Croatian196, but, strictly speaking, that is improvable, since the old dual
ending –u is not preserved in Slovene except in naʄju, vaʄju.

Old Czech

§ 40 In Old Czech, the form vě is attested in the 14th/15th century. In the 15th century, a
younger form va appears. This form is probably due to analogy with the new verbal ending –
va (instead of the older ending –vě, both of which are attested in Old Czech) in the first
person dual. Another likely influence was the number dva. One could also imagine the
influence of the 1st person accusative dual form *na (cf. va : va in 2nd person dual in OCS),
but the accusative na is not attested in Old Czech. In the verbal ending, –vě > -va was due to
the influence of the 2nd person ending –ta. The ending –va in verbs is attested from the 14th
century, and the nominative dual va is attested from the 15th century197. That also points to
the conclusion that the new form was transferred from the verbal ending to the pronominal
form198. In the year 1502 and 1672, an even newer form ma, with the analogical plural m- is
attested199. In the 2nd person nominative dual, there is no form *va, unlike in OCS. Only the
form vy is attested there, which is an archaism (see § 50).

Old accusative forms *na and *va are not attested at all in Old Czech. The original
genitive dual forms najú and vajú are used instead (due to the general tendency of using the
genitive forms instead of the accusative ones in Slavic). Later, the plural forms ny and vy are

The old forms have been preserved in the genitive-locative dual: najú, vajú (> nají,
vají). The length of the final –ú is secondary200 and due to analogy with the forms which
developed the long –ú by contraction: jú < *jejuʀ, tú < *tojuʀ, dvú < *dvojuʀ, obú < *obojuʀ, mú
< *mojuʀ, tvú < *tvojuʀ, svú < *svojuʀ. This long –ú has been generalized not only in the forms
najú, vajú, but also in all genitive-locative dual nominal forms in Old Czech: očú, ušú, hosťú,
zubú, letú etc. The length of the Slavic old acute in the first syllable in Czech najú, vajú has
disappeared because of the secondary length of the ending201. For the reconstruction of
Common Slavic *naɾju, *vaɾju, cf. Slovene naʄju, vaʄju and Old Russian в нáюɺ, в вáюɺ202.

196 In Croatian, the long -ū (cf. forms like noʂgū, rùkū, slúgū), which is today one of the genitive plural
endings, is due to the influence of the i-stem genitive plural –ī, in which the length is the result of the
contraction of the old ending *-ijь < *-ьjь. The accent of the Old Croatian naju, vaju is not known (for
Križanić's forms, see § 38).
197 Gebauer 1898: 15-16.
198 In Old Czech from the end of the 14th century, the following types of sentence are attested: vě na
každý den obět vzdáváva (Mareš 2001: 180).
199 Mareš 2001: 180.
200 All the long vowels in final open syllable in Slavic are secondary, cf. Kapović 2003: 57f.
201 Cf. Kapović 2003, Kapović 2005.
202 Дыбо 1981: 35, 2000: 63.

The dative-locative forms náma, váma are preserved in Old Czech as well. These
forms are used for plural in some Czech dialects203.


§ 41 Lusatian has, as already said, maintained dual as a category, unlike most other Slavic
languages. The old pronoun *vě has been preserved in Upper Lusatian dialectal wi204, with a
secondary –i by analogy to the o-stem plural –i205. In Lower Lusatian, this has been replaced
by younger forms mej (1st person), wej (2nd person), and in Upper Lusatian by mój, wój.
According to Mareš (2001: 183), these forms are due to analogy with, also younger,
dual verbal 1st person endings – cf. LL njasomej “carry” : mej (analogically also wej), UL smój
“are” : mój (analogically also wój). The unifying of the verbal ending and the personal
pronoun exists in Lusatian in the plural as well: cf. LL njasomy : my (and wy), UL smy : my
(and wy). In the plural, the verbal ending –my is a secondary development due to analogy
with the personal pronoun my (cf. also –my in Polish and in Old Czech).
Vaillant (1958: [252]) offers a different explanation of the Lusatian nominative dual
forms. He thinks that the forms mej, wej have the ending –ej from the pronouns tej, wonej.
This seems a little bit strange and Mareš (2001: 183f) rejects this idea vehemently. For the
UL mój, wój, Vaillant (ibid.) takes mój as analogous to wój, this form itself being the old *vy
with the dual –j added. The development of *y > ó is, according to him, due to the preceding
labial (cf. the form móš besides the form myš “mouse”). The explanation via the influence of
the verbal endings seems to be more economical than the latter one.

The old genitive-locative forms, naju and waju, are used in Lusatian as the genitive-
accusative forms. For the locative, the old dative-instrumental form is used (cf. the situation
in Slovene, § 39).

The old dative-instrumental forms are used for the locative as well. In Lower
Lusatian, these forms are unchanged: nama, wama. In Upper Lusatian, these forms got a
secondary –j at the end, namaj, wamaj, due to the influence of the –j in the nominative: mój,

Old Polish

§ 42 Unlike Old Czech, where younger va was attested besides (actually after) older vě, in
Old Polish only the newer form wa is attested in the 15th/16th century (due to the same
analogy as in Czech, see § 40). In 2nd person nominative, only the form wy is attested, as
found in Old Czech and Old Russian.

The old accusative dual forms were lost before the earliest records – old genitive
plural forms nas, was are used instead in Old Polish.

203 Mareš 2001: 181.

204 Muka 1891: 393-394.
205 Vaillant 1958: 458.

In the 15th/16th century, in the genitive-locative and dative-instrumental, the old
forms, naju, waju and nama, wama, were preserved in Old Polish. These forms are still
widely used in Polish dialects, mostly functioning as plural206.


§ 43 In Kashubian, the dual has been preserved as a category, but not without a lot of

In the nominative dual forms, special secondary feminine forms have developed (like
in Slovene) in some Kashubian dialects. In the 1st person dual, the forms ma, maiʠuɺ, naiʠuɺ are
attested. The form ma is the original one here, explained via *vě > *va > ma, like in Old
Czech (-a due to the verbal ending, and m- due to the influence of the plural form). The
form naiʠuɺ is the original genitive-locative (in modern Kashubian also accusative) form. The
same goes for the form maiʠuɺ, which has a secondary m- due to the from ma (and plural mǩ).
The use of the accusative for the nominative is hardly an unusual development in Slavic (cf.
the ā-stem nominative plural in –y, which is originally the accusative plural ending). In the
feminine gender, either the same form ma is used, or the secondary forms mɴe, mǩ (with
feminine endings *-ě, *-y). In the 2nd person dual, the form va is used in the masculine
gender. This va is probably due to analogy with the 1st person dual form ma, and it is
unlikely to be connected to the OCS nominative form va. The development of the secondary
feminine forms is the same as in the 1st person.

The old genitive-locative forms are, as usual, used for the accusative as well (and the
nominative in the 1st person). A couple of secondary forms, besides the original naiʠuɺ, vaiʠuɺ
have developed (cf. for instance vaiʠix with a genitive plural ending –ix).

The old forms nama, vama are preserved in the dative-instrumental.


§ 44 In Slovincian, the dual has also been preserved. The development is quite similar to
Kashubian, of course.

The nominative forms mã and vã are to be explained in the same way as in Kashubian
(see § 43). Lorentz (1903: 274) thought that the dual nominative form mã is due to analogy
with the old accusative *va (sic! not *na), which first became the 2nd person nominative, and
then influenced the 1st person nominative form. This is not impossible, but the old
accusative forms *na, *va are not actually attested anywhere in West Slavic. 2nd person vã
could, thus, be either analogical to mã, which got its –a from the verbal ending. Even if the
2nd person nominative vã is due to analogy with the accusative *va, this does not necessarily
mean that this vã is the same as OCS va. These could be unrelated innovations (using the
accusative va as both the nominative and accusative) in Slovincian (and Kashubian) and OCS.

206 Cf. Klemensiewicz, Lehr-Spławiński & Urbańczyk 1955: 322.

Besides these forms, new clitic forms ma, va exist, as well as the secondary feminine forms
with the feminine ending *-ě: mjìeʠ, vjìeʠ (here the numeral forms dvã, dvjìeʠ probably played a
crucial role), also with new clitic forms mjä, vjä.

The original genitive-locative forms nãjșɺ, vãjșɺ are of course used for the accusative
as well. In the dative-instrumental, the original forms nąɶmƮ, vąɶmƮ were preserved.

Old Russian

§ 45 In Old Russian, the situation with dual personal pronoun forms is very similar to OCS.
The form вa is attested only as the accusative dual and only in the OCS influenced texts207.
The form вы in the nominative dual is attested already in the year 1056208. According to
Mareš (2001: 179), the forms на, ва are not attested for the dative dual in Old Russian.
According to Vondrák (1912: 460, citing Соболевский), these forms are attested very well in
Old Russian209 – this probably refers to Church Slavic texts. For the accentuation of на, ва,
which points to Slavic *naʄ, *vaʄ, cf. Зализняк 1985: 143.

207 Гадолина 1963: 112-113.

208 Соболевский 1907: 206.
209 Cf. also Birnbaum & Schaeken 1997: 75 for this.

Reconstruction of personal pronouns in Common Slavic

Up until now, we have reconstructed Common Slavic forms *jaɾ and *jãzъ < *jāzъʀ and
*tyʄ, *myʄ, *vyʄ. The first two forms were discussed exhaustively, while the reconstruction of
the last three is fairly simple. We have also laid out all available material for the
reconstruction of other forms of personal pronouns in Slavic.
Now, we shall first discuss some specific problems concerning the reconstruction of
the Common Slavic personal pronoun system, and then we shall reconstruct all the
paradigms, 1st and 2nd person singular/plural/dual, with the accompanying accents of
course. Up until now, we have mainly discussed the problems of the development in the
historical times in the separate Slavic languages. Here, we shall discuss the problems of the
development in Common Slavic and Proto-Slavic times210. The problems of the derivation of
the Slavic forms from Balto-Slavic will be discussed mainly in the section dealing with the
reconstruction of the Balto-Slavic personal pronoun system (see § 69, § 70 and passim).

210 Of course, in many cases it is not easy to distinguish these periods.

Dative-locative singular *mъ/ьně

§ 46 As already said, the dative-locative singular form of the 1st person personal pronoun
presents a very difficult problem, the problem being in the practical impossibility of
ascertaining the nature of the first syllable (at least within Slavic), i.e. whether we are dealing
with *ь or *ъ here. OCS is ambiguous concerning this problem, since both ь and ъ appear
there and both could be secondary (see § 15), and the rest of the Slavic languages are quite
useless in this respect, since the yers are simply lost everywhere in that position.
The argument for the originality of mьně usually adduced (see § 15), is that the yer
here is parallel with e in tebě. However, this argument is valid for OCS (and South Slavic)
only. In West and East Slavic, the original form in the dative singular of the 2nd person was
not only tebě, but tobě as well211. This tobě would be parallel not with mьně, but with mъně.
Even if the –o- in tobě is not very old (see § 47, § 72 for its origin), that must not necessarily
be an argument against the originality of the form mъně212. It is even possible that both
variants (*mъně and *mьně) existed in Common Slavic, which would then be parallel to
*tebě/tobě. Anyhow, the possible parallelism of mьně with tebě and mъně with tobě does
not really shed any light on the origin of this form.

There are two basic possibilities for the origin of this mъn-/mьn-. If the original form
was mьně, the yer could be explained via reduction. This is also possible for mъně, but not
so likely, since it seems that the cases of a reduced *ъ are practically non-existant. This
would be the same kind of sporadic reduction seen in Slavic words like *čeloveʖɾkъ - *čьloveʖɾkъ
"man", *česoʀ - *čьsoʀ "of what", *četyɾre - *čьtyɾre "four", *veʂčerъ "evening" - *vьčeraʀ
"yesterday" and *mèčь/mьɳčь "sword" (< Gothic mēkeis)213 and *žeravь - *žьravjь "crane".
However, this kind of reduction is difficult to explain. The adduced examples all have
a *č before or after the yer (except for the "crane" word), which is not very similar to *mьně.
The sonant *-n- after the vowel, however, would favor reduction. That would also explain
the possible reduction in *mьně, but not in *tebě. The following *-b- would not really favor
such a reduction. An important thing in this reduction is that the reduced vowel is always
unaccented, which is also the case in DL. *mьneʖʀ (cf. G. *meʂne without reduction). However,
since other examples of the reduction of *e to *ь in front of *n are not attested, at least to
my knowledge, this approach does not seem to be very fruitful. Anyhow, explaining the form
*mьně via reduction is a clear case of an obscurum per obscurius explanation.

If the form mъně is the original one, then it is simply an archaism connected to
Latvian (and maybe Lithuanian, see § 59, § 62) dialectal stem mun-. The form mьně would
then be secondary. Since the Baltic (i.e. Latvian) data does indeed point to the stem *mun- in
the dative, the easiest solution is to assume that the form mъně is the original one. For the
discussion of the origin of this form see § 73.

211 It is not clear whether all of North Slavic had both forms from the beginning, or if some of the
dialects had just one of them (*tobě), see § 20-28, § 47.
212 The –o- in tobě could have already been there in the time when the form with mъn- was made.
213 Cf. also a different case of the loanword *cěsarь - *cьsarь.

We have concluded that the form mъně is probably older. How are we then to explain
the form mьně in OCS? One possible explanation was already mentioned - the assimilation
of the back yor to the front yat in the following syllable. This is especially possible if we bear
in mind that OCS had only the variant tebě in the 2nd person. This assimilation would yield
the parallelism of –e- and –ь- here. There is another possible explanation. The original form
being mъně, there was an ablaut –en- : -ъn- here214. This pattern could have been changed
to –en- : -ьn-, which was a very productive ablaut pattern215.

214 Or at least an ablaut-looking vowel change in the root.

215 For this kind of development, cf. Matasović 2004.

Dative-locative singular *tebě/tobě

§ 47 As we have seen, in the dative-locative singular of the 2nd person personal pronoun,
both the forms with teb- and with tob- are attested. In South Slavic, only the form *tebě is
attested. In East Slavic, both forms are attested, but it is not certain if the form *tebě is
indeed old (inherited from Common Slavic, together with *tobě), or a younger independent
innovation in Russian (see § 27). In West Slavic, both *tebě and *tobě appear – again, it is
impossible to tell when (or if) this *teb- is inherited from Common Slavic, and when it is
analogous to other cases.
The change of the dative-locative tobě to tebě due to the influence of the genitive-
accusative tebe is quite imaginable, due to the importance and frequency of genitive-
accusative forms. However, the to- forms can hardly be secondary, since there is no reason
why the dative-locative tebě should be, for instance, influenced by the instrumental form
tobojNj216. Instrumental is neither very frequent, nor especially close to dative-locative in any
way217. Thus, every to- form can generally be regarded as an archaism, and every te- form in
the dative-locative can be an innovation as well as an archaism.
In Czech and Slovak, both te- and to- are attested; in Polish only to- appears; in
Lusatian both stems appear; Kashubian has to- in the dative and te-/to- in the locative;
Slovincian has to- in the dative and te- in the locative; in Polabian only te- is attested. Here
is an overview:

Only te - OCS, Macedonian, Bulgarian, Croatian, Slovene, Polabian

Only to - Polish, Ukrainian, Byelorussian218

Both te to-
- and to - Czech, Slovak, Upper Lusatian, Lower Lusatian, Kashubian,
Slovincian, Russian

216 It is hard to think of a different source for the supposed secondary –o- in tob-.
217 Exactly that kind of ad hoc explanation is provided by Schenker 1995: 128.
218 Ukrainian and Byelorussian presence in this group must be taken as somewhat arbitrary. To my
knowledge, the forms with te- are not attested in modern Ukrainian and Byelorussian dialects.
However, since there is no practical difference between Old Ukrainian, Old Byelorussian and Old
Russian, the presence of the te- forms in Old Russian points to their presence in the earlier stages of
Ukrainian and Byelorussian as well.

One could claim that only to- is the original stem in the dative-locative in West
Slavic, while all te- forms are secondary and analogical219. This is possible in theory, but not
very likely, since practically every (sic!) West Slavic language has the te- form attested (at
least as the dialectal variant or historically). It is not very likely that this was an independent
parallel change in almost all West Slavic languages. If we were to assume that this supposed
analogical change occurred early in some Proto-West-Slavic, we would either have to assume
that the form *tobě was the only existing form in Common Slavic (which is not likely, see §
47), or that the variant *tebě first disappeared in West Slavic and that it was then
reintroduced by analogy. The last approach is not economical.
Thus, we may conclude that in all probability, at least some of the West Slavic
languages have inherited Common Slavic form *tebě. Of course, it is quite possible that
*tebě is inherited in some West Slavic languages, while it is analogical in others.

A question arises – how to explain the variants *tebě and *tobě in the dative-locative
in Common Slavic? If we were to reconstruct the Common Slavic paradigm based solely on
Slavic evidence, the simplest solution would be to reconstruct only *tobě in the dative-
locative singular. If we were to start with an original paradigm *tebe - *tobě – *tobojNj (as
attested, for instance, in Polish or Ukrainian), we could easily get to *tebe - *tebě - *tobojNj
(as attested, for instance, in Croatian), or even further to *tebe - *tebě - *tebojNj (as attested,
for instance, in Slovak).
Another easy way to explain the variants *tobě/tebě in the dative-locative would be
to assume that one was the original dative form, and the other the original locative form.
Thus, if we were to believe that Slovincian is the most archaic here, we should reconstruct
*tebe - *tobě - *tebě - *tobojNj, which would yield all attested Slavic patterns even more

However, even though we are concentrating mostly on Slavic internal evidence in this
section, in historical linguistics external evidence should also be considered whenever
possible. In the light of the comparative evidence (and the dative-locative forms like Old
Prussian tebbei and Avestan taibya, see § 72, § 98), there is no evidence indicating that the
form *tobě could have been the original one in the dative220. Thus, this theory is not very
likely and Slovincian must have innovated here, generalizing one of the inherited forms in
the dative and the other one in the locative. The secondary introduction of the stem teb- in
the locative only is also a possibility.

A different kind of solution is offered by Vaillant (1958: [249]) and Kiparsky (1967:
133). They claim that the original pattern was *tebě in the dative and *tobě in the locative.
They derive this from the fact that in the parallel reflexive pronoun the stem *sob- is found
in some derivatives that seem to be in connection with the locative form. Thus, in OCS there
is the adverb osobě «especially, separately» (~ *o sobě), posobije «help» (~ *po sobě), Czech
násobný «diverse, manifold» (~ *na sobě). This interesting solution is downplayed by the fact

219 One could claim that for South Slavic as well!

220 For the locative form, such a possibility exists, see § 72.

that no Slavic language preserves this kind of pattern (i. e. D. *tebě, L. *tobě). What is even
worse, the only Slavic language that has the stem distinction of the dative and the locative
form by the stems *teb-/tob-, Slovincian (and not so strictly Kashubian), has the exact
opposite distribution – *tobě in the dative, and *tebě in the locative! These derivatives
indeed point to the conclusion that the form *sobě (and thus also *tobě) existed even in the
earlier phases of those languages, like OCS, in which it was never attested. However, these
derivatives do not prove that these *-o- forms were the only possibility in L. or that they
were not possible alternants in the dative as well. This theory could be correct but the
distribution of the stem *teb- in the dative and *tob- in the locative would have to be pre-
Common Slavic.

Judging by the reflexes in Slavic languages, it seems only reasonable to reconstruct

both *tebě and *tobě in the dative-locative singular for Common Slavic. The form *tebě must
be reconstructed because this is what is attested in South Slavic as the only possibility, and
in the rest of the Slavic together with *tobě as one of the variants (when looking at West and
East Slavic as a whole). The form *tobě must also be reconstructed for Common Slavic,
because there is no possibility of this form emerging secondarily in North Slavic. As already
said, the instrumental is not so frequent as to be the source of this kind of analogy.

How are we then to explain these two variants in the dative-locative singular? Within
Slavic itself, there seems to be no obvious reason for the stems *teb-/tob- to be attested in
the dative-locative, in the same manner as there seems to be no reason for the stem *teb-
to be attested as the only one in the genitive, or for *tob- to be the only stem attested in the
instrumental221. The same goes for the ablaut alternations in the 1st person singular oblique
stems. Thus, this problem must be dealt with in the Balto-Slavic section (see § 72).

221 There is a possibility of the preference of the stem *tob-, instead of *teb-, because of the *-o- in
the ending *-ojNj, but that can hardly be called a solution.

Nominative plural *my and *vy

§ 48 In the nominative plural of personal pronouns, we find the forms my and vy. The
apparent characteristic of these forms is that they have the same ending as the accusatives
ny and vy (and as the nominal o- and ā-stem accusatives). Thus, these endings have
generally been explained in the literature as the original accusative endings that have been
taken analogically to the nominative222. However, there is simply no reason to take the
supposed accusative vy for the nominative as well. This kind of development would be a little
strange typologically, since, in general, languages tend to preserve the accusative-
nominative difference in pronouns223, and do not eliminate it without any apparent reason.
However, cf. Kashubian maiʠuɺ, naiʠuɺ in N. dual (see § 43). It is difficult to presume that a
highly flective language like Slavic would eliminate the difference of N. and A. pl. of the 2nd
person personal pronoun for no reason (although it is possible to look for the source of
analogy in the ā-stems which have the same ending –y in both cases).

Another, more fruitful approach has also been presented224. According to Lithuanian
jũs (and the forms in other IE languages), we would expect *jūs > *jū > *jī > *ji in Slavic. But
before the change of *jū(s) > *jī(s)225, the initial *j- in Slavic was replaced with the *w- from
the oblique cases. Thus, the form *wūs was made, which regularly yielded attested vy. The
form my is then made analogically to vy (cf. Lithuanian meɶs).
This kind of development explains Slavic nominative-accusative vy via a simple
sound merger of the original *wūs and *wāns (and my via convincing analogy to vy), and not
by assuming an unmotivated change by which the accusative form has become both the
nominative and the accusative form (or that the accusative ending was taken to the

222 Cf. for instance Meillet 1934: 454, Vaillant 1958: 452, Arumaa 1985: 168.
223 Cf. English we : us, a pattern inherited from PIE, in spite of the almost complete lack of cases
outside pronouns.
224 Cf. for instance Беседина-Heвзорова 1962: 178, Aitzemüller 1991: 111.
225 Cf. also Schmidt 1978: 210.

Unaccented forms *ny and *vy in dative plural

§ 49 In Slavic, besides the regular dative plural forms *namъ and *vamъ, the forms *ny and
*vy, identical to the original accusative plural forms, appear as clitics in the dative. This was
often regarded as a secondary development226. However, there are two basic problems with
this explanation. Firstly, it is very difficult, if not impossible, to explain exactly how the
accented accusative forms became the unaccented dative forms. Secondly, ny and vy in the
dative function are attested very well in the whole variety of Slavic languages – in OCS (and
Bulgarian & Macedonian), Old Russian, Old Croatian, and in Montenegrin and Serbian
dialects. Thus it must have been already a Common Slavic feature.
Since the explanation of this phenomenon as an innovation is not very convincing,
one should consider the other possibility – that the use of ny, vy in the dative is not an
innovation, but an archaism. Slavic accusatives ny and vy are derived from IE unaccented
GDA. forms *nos/nōs, *wos/wōs (see § 94), cf. GDA. nas, vas in Vedic. In (Balto-)Slavic, the
*-n- was added in the accusative, thus *nōs > *nōns > *nāns > ny. Since this *nōns with a
secondary accusative *-n- has taken the place of the old *nōs in the accusative, the old *nōs
was replaced with *nōns in the dative as well. Thus, we got the unaccented ny, vy in the
dative, identical to the accusative forms (which have become accented forms already in
Balto-Slavic, see § 79). Here, the polyfunctionality of Vedic nas, vas is still seen in Slavic227.
This is the only way to explain the appearance of the same form in the accusative and in the
dative plural.

226 Cf. for instance Hamm 1970: 134.

227 Cf. Meillet 1934: 455, Vaillant 1958: 453.

Nominative dual of 2nd person

§ 50 Traditionally, many linguists have been hasty to accept OCS 2nd person nominative
dual va as the oldest Slavic form228. The reason for this was the fact that this va was a special
form, different from the 2nd person nominative plural form vy. The logic is clear – if the dual
is a separate category in Slavic, it must also have separate pronoun forms to differentiate it
completely from the singular and the plural. However, after a careful inspection of the
material, one gets a different picture.
The form va is attested in OCS, but only in the younger, cyrillic texts (see § 37).
Except for that, it is attested only in Kashubian and Slovincian, where it is probably a
younger form created due to the influence of the 1st person forms (Kashubian ma, see § 43).
The exact origin of Bednja –vo is not clear (cf. § 38). Everywhere else, including OCS older,
glagolitic texts, Old Russian, Old Czech and Old Polish, only the form vy appears in the
function of the 2nd person nominative dual. That this can hardly be explained as an
innovation, i.e. taking the plural form for the dual, is clear from the fact that the 1st person
nominative dual form vě (or younger va) is very well attested in Old Russian, Old Czech and
Old Polish. It is not very likely that in all these languages, the 1st person form would be
preserved, but that the separate 2nd person form would have vanished without a trace.
Thus, another approach is necessary – and that is to assume that the best attested
2nd person form, vy, is indeed the oldest form229, in spite of the fact that it is identical to the
plural form. The reason why the dual and the plural form are identical is simple – they were
once different, but later, due to sound change, they became identical. We have already
explained the plural form vy from the older *jūs where *j- was substituted with *w- (see §
48). According to Lithuanian dialectal jù-du for the 2nd person nominative dual, we can
assume a pre-Proto-Slavic form *jū (see § 69, § 70). Here, *j- was also substituted with *w-
by analogy to the oblique forms (GL. vaju, DI. vama, A. va), and the new dual form *wū
merged with the new plural form *wūs when all final consonants were dropped in Slavic by
the law of the open syllables.

In this way, as we have shown, we have come to the point where both the plural and
the dual 2nd person nominative form was vy. This was, of course, a little bit problematic
because it seemed that the plural form was used instead of the dual form. This functioned in
Old Czech, Old Polish and Old Russian for instance (only for some time though, since the
dual eventually disappeared in every one of those languages), but in the OCS of the cyrillic
texts, the unfortunate nominative form vy was already replaced with the original accusative
dual form va, which was distinct from the plural form. In this way it was possible to
differentiate the dual as a category from the plural more clearly. In OCS glagolitic texts,
however, the older dual form vy was preserved, which led to the general instability of the
difference of the dual and plural in the nominative-accusative personal pronoun forms. Since
the nominatives were already the same in the plural and dual (vy = vy), this led to the
transfer of the accusative plural vy into the dual – thus we got N. vy – A. vy in the dual, just

228 For instance Mareš 2001: 176.

229 For this approach, cf. for instance Vaillant 1958: 455.

like in the plural. Since the accusative plural vy appeared in the dual as well, this was taken
even further and the accusative plural ny was also used instead of the original na. Thus, the
original accusative dual forms na and va are not even attested in OCS glagolitic texts (nor in
Old Czech and Old Polish) where the original nominative dual vy was preserved. That the
original accusative forms na and va are attested in cyrillic texts is explained by the fact that
there the older nominative vy was replaced with a new va, which helped to preserve the
general difference of the plural and dual better. The preservation of the form vě (unlike va
and na) in OCS glagolitic texts, Old Czech and Old Polish (and not its replacement with the
plural form my, which is exactly what happened later in Czech and Polish though), is easily
explained – the form vě (or younger va) was strongly supported by the verbal dual 1st person
ending –vě. Thus, vě was preserved, while na and va were eliminated.
It is highly probable that the accidental merger of the pronouns of the 1st person
nominative plural and dual (the older *jūs and *jū both becoming vy) has facilitated the
overall disappearance of the dual as a category in most of the Slavic languages. This was one
of the ways in which the system was made unstable. One could even claim that this merger
was the trigger that eventually led to the disappearance of the dual as a category in Slavic.
That this is not an overstatement is proved by the fact that in all Slavic languages which have
preserved the dual as a category up to today (Slovene, Lusatian and Kashubian – all fringe
dialects, by the way), the nominative plural and dual of the 2nd person personal pronoun
forms are different – cf. Slovene vî : vîdva, Lower Lusatian wy : wej, Kashubian vǩ : va.

Dative dual na, va

§ 51 The normal dual dative forms are nama and vama. However, in Codex Suprasaliensis,
the form va appears as the unaccented dative form as well230. The form na is not attested in
the dative in the canonical OCS texts, but both na and va appear frequently in this function
in Church Slavic texts231. Again, this function of na and va is considered secondary and
innovative by many232, but Meillet (1934: 455) and Vaillant (1958: 455) think of it as an
archaism233. As is the case with the clitic ny, vy in the dative plural, it is hard to think of a
reason why the accented accusative forms would become the unaccented dative forms. This
can only be explained if one takes both of these forms to stem originally from an oblique
clitic form234. Thus, the forms na, va were originally the GDA. clitics that are preserved as the
secondary stressed A. dual and D. dual clitics. (cf. § 80).

230 Diels 1932: 214.

231 Vaillant 1958: 454.
232 Mareš 2001: 176.
233 Cf. the Vedic GDA. dual clitic forms nau and vām.
234 The implausibility of the transference of function from the accented accusative to the unaccented
dative is not to be mistaken with the IE general oblique unaccented forms. That one clitic form can
serve for more than one oblique case is not at all unusual.

Accentuation of 1st and 2nd person genitive/dative singular

§ 52 As we have seen in the material, in the reflexes of the genitives *mene, *tebe and the
datives *mъně, *tebě/tobě, in Slavic languages one mostly finds the same kind of
accentuation - on the first or on the second syllable, in both cases.
For the initial accent, cf. Croatian meʂne, teʂbe, meʂni, teʂbi, Old Russian мéне, тéбe,
мъɴнΟ, теɴбΟ/тóбΟ, and for the desinential accent, cf. Serbian мèнe, тèбe, мèни, тèби, Slovene
méne, tébe, méni, tébi and Middle Bulgarian235: мeнé, мнΟ, тебé, тебΟ236  .
In Polabian one finds both the initial and the desinential accent: miné, mané, mánǩ,
tíbǩ, tibé.
The accentuation of Slovincian cìeʠbjä, tșɳȅʠbjä is not relevant, since the final accent
would be retracted from the short final syllable anyway.
The answer to this distribution is, as already mentioned, attested in archaic Croatian
dialects (Štokavian and Čakavian), which preserve the initial accent in the genitive: meʂne,
teʂbe, oʂd mene, and the desinential accent in the dative: menıʂ, tebıʂ (see § 17 for the dialectal
attestations of this). Thus, one should reconstruct Common Slavic genitive *meʂne, *teʂbe, and
the dative *mьneʖʀ/mъneʖʀ, *tebeʖʀ/tobeʖʀ.

235 Evangelie 1139 (Hock 1992 I: 119).

236 Modern Bulgarian forms мèнe, теɳбe are the result of stress retraction.

Accent of *mę, *tę

§ 53 The problem of the accent of the original accusative singular forms *mę and *tę is
difficult to solve. The reason for this is that these original accusatives were preserved as
such only in OCS, and there, there are no accentual marks. In the present day Slavic
languages, these forms are always, if preserved, unaccented, which does not help us in
determining the original accent. The only useful material for the reconstruction of the accent
of *mę and *tę is the accentuation of the combination of these pronouns with a preposition,
like *za mę, *na tę etc. Here, only Croatian (Štokavian, Kajkavian, Čakavian) and Slovene
material are useful, since there are no combinations like this in Russian, and in West Slavic
the shortness of the preposition (cf. Old Czech na tě) is not very useful, because it is hard to
expect that prepositions could, for instance, be long in these combinations, when they are
otherwise always short.

In Slovene, there is ʄ on the first syllable in these combinations: naʄme, zaʄte. One also
finds this in Kajkavian (for instance Bednja), some Čakavian and some Old Štokavian (like
Posavina) dialects: naʄ me, zaʄ te.
In Neo-Štokavian, the accentuation of these forms is different: ná me, zá te (from the
older *nā meʂ, *zā teʂ)237. This kind of accentuation is also attested in some Čakavian dialects,
but with a secondary stress retraction, cf. Vrgada: n°ã me (from the older *nā meʂ)238.
However, if the yer at the end of the preposition is vocalized in Neo-Štokavian, another
pattern is observed: poʂdā me, naʂdā te, uʂzā me (from *podъ mę etc. with a vocalization of the
yer, or a secondary prop vowel, in order to break the cluster *-d m-). In Vrgada, one finds
the same pattern here as in the combinations with no vocalized yer: uz°ã me.

The reconstruction of the Common Slavic accentuation of these combinations is not

easy because of the differences shown in the material we just saw. Kajkavian, Čakavian and
Old Štokavian accentuation would point to Common Slavic *naʄ mę, i.e. to *męʄ. This *męʄ
would than be in accord with the rest of the personal pronouns accusatives (*nyʄ, *vyʄ, *naʄ,
*vaʄʄ, see § 56, § 57). In Old Russian, мя,тя do not function like ны, вы, на, ва – first two
being “плюсовые энклитики” (like ми, ти), which looks like an innovation, and the latter
forms being “минусовые энклитики”239. Thus, мя and тя did not participate in the
combinations like нá ны, нá вы240. Middle Bulgarian вь мΧɴ, зa мΧɴ (Hock 1992 I: 120) must
also be secondary (cf. also нa ныɴ there, as opposed to нá ны, preserved in some other
Middle Bulgarian texts).
Kajkavian, Čakavian and Old Štokavian forms like naʄ me can easily be derived from
Slavic *naʄ mę. These would then be the sole examples of the preserved (i.e. not shortened)
long circumflex on the preposition with the absolute initial accent in Croatian. In all other
examples, like naʂ svijēt, naʂ strānu, zaʂ mēso etc., the circumflex was shortened due to the

237 The length is generalized in cases like pó me, pó te where we would expect **pò me, **pò te.
238 This kind of retraction is known in Čakavian, in spite of the fact that there is no general retraction
of stress, cf. Vrgada tũći for older *tūć ıʂ (this is typical for all of Čakavian), pũta instead of *pūtaʂ etc.
239 Зализняк 1985: 143.
240 Дыбо 1981: 35, Дыбо 2000: 62.

number of syllables, while *naʄ mę would preserve the length like *męʄso > meʄso241.
However, the Slovene forms are not so simple as to be explained in such a manner. Slovene
naʄme cannot be derived regularly from the Common Slavic *naʄ mę, since this should yield
**na meʖʄ in Slovene, with a progressive shift of the circumflex. It is not clear how Slovene
forms like naʄme should be explained. An ad hoc solution would be to presume the analogy
with the cases like *nî za mę > ni zaʄ me, but it is unlikely that combinations with two
prepositions were common enough to be strong enough an influence in this case.

Neo-Štokavian forms like ná me < *nā meʂ, would, on the other hand point to the
possible Common Slavic *nā męɾ. However, forms like poʂdā me could be in accord with
dialectal Croatian naʄ me (and Vrgada uz°ã me would be taken one step further in the
innovation). It is possible that in these cases the old accentuation was preserved, while naʄ
me changed to secondary ná me. Considering the practical isolation of Neo-Štokavian in this
accentuation (together with Middle Bulgarian) and structural reasons242, it would be safest to
assume that the accentuation naʄ me is the archaic one243. Old Czech na tě also confirms
It is not difficult to explain the secondary accentuation in Neo-Štokavian. The forms
like naʄ me, zaʄ te were isolated in the system – this was the only case where the long falling
accent was placed on the preposition; in all other cases, the accent on the preposition (if the
accent was there) was either ʂ (like in zaʂ strāne, naʂ rūku, zaʂ brōd) or ɳ/´ (like nà ulicu, zà
kuću, and in combinations like trí dāna < trî daʄna, stó godīnā < stoʄ goʂdīnā, dvá trī < dvaʄ
trî). Thus, the unusual forms like naʄ me were eliminated and adjusted to the type of the
newer Neo-Štokavian accentual type like dvá trī. On the other hand, dialectal Croatian naʄ me
is hard to explain via any analogy or secondary development, because, as already said, ʄ on
the prepositions appears in this case only. This is also one of the reasons why one should
assume that the accentuation like naʄ me is more archaic than the accentuation like ná me -
it is the usual approach in historical linguistics to assume that the form that looks odd and
cannot be interpreted in any way as innovative must be regarded as archaic.

241 For a thorough discussion of these problems, cf. Kapović 2003: 58-61, Kapović 2005: 77-81.
242 One would expect a circumflex in the accusative, not an acute, because of Meillet’s Law, see § 57,
and because one finds a circumflex in the other personal pronoun accusatives.
243 In spite of the fact that Slovene does not provide real support for that solution.

Accentuation of 1st and 2nd person instrumental singular

§ 54 As is known, the ending of the pronominal instrumental singular, *-ojNj, like in

*mъnojNj and *tobojNj, is found also in the ā-stems, like ženojNj. This ending was transferred
from the pronouns to the ā-stems (for the origin of this ending, see § 77). In the ā-stems,
on the structural grounds, one expects the ending *-òjNj in a. p. b and *-ojNjʀ in a. p. c244. In
ā-stems, one finds the reflexes of the old a. p. c *-ojNjʀ generalized (in both a. p. c and in a.
p. b) in Čakavian nogõn, ženõn (< nogõm < nogõv ⇐ *nogouʂ ⇐ *nogojuʂ < *nogojNjʀ)245 and
in Slovene nogóʖ, ženóʖ. On the other hand, Russian ногóю, женóю (thus also in Old Russian),
would point to the generalization of the old *-òjNj. However, Russian –óю might also be
derived from the older –oюɴ by analogy to other cases, in which the accent is always on the
first syllable after the stem. The accent –oюɴ would be the only exception in the paradigm. In
Old Russian, one finds мнoюɴ and тобoюɴ, where the –oюɴ is indeed attested – there, the
declension had different forms, including ones like мя, тя, which were totally different from
the other accented forms, and the initial accent in the dative/genitive forms, so there was no
real pressure for –oюɴ to become –óю, which might have happened in the ā-stems (if the –óю
there is not to be attributed to the old a. p. b *-òjNj, which is also, of course, possible). In the
Middle Bulgarian dialect of Evangelie 1139 (Hock 1992), one finds the forms мнóΧ, тобóΧ,
with the same accent as in the nominal forms женóΧ (a. p. b), странóΧ (a. p. c).
There is one dialect, though, that seems to have preserved the opposition of the old
*-òjNj in a. p. b and *-ojNjʀ in a. p. c - Bednja Kajkavian. There, one finds the instrumental
storešĩnu from the noun storešēīnoʂ (a. p. b), but the instrumental rukũ from the noun rōūkoʂ
(a. p. c)246. As noted by Vermeer (1978: 376), rukũ is to be derived from *rNjkNjɶ < *rNjkojNjʀ and
storešĩnu < *starěšīnNjʄ < *starěšīnòjNj, with a typical Kajkavian development of ɷ ʄ > ɶˇ as
defined by Ivšić (cf. Kajkavian *zābaʄva > zãbava, but *otaʄva > otaʄva).

The accentual paradigms of both 1st and 2nd person personal pronoun belonged to a
special case of mobile a. p. c, which is clear from the forms like *meʂne, *oʂtъ mene, but
*mьneʖʀ etc. (see § 56 for more details). Since the paradigm was mobile, one would expect on
structural grounds the instrumental forms *mъnojNjʀ and *tobojNjʀ with final accent. This is
exactly what is found in Old Russian мнoюɴ, тобoюɴ, Slovene menóʖj, tebóʖj, and possibly also
in Slovincian mnoɺųʠ, tȅboɺųʠ. However, Slovincian forms could be from *-òjNj as well, if the
contraction occurred after Ivšić's Law247, as it did in Bednja. In the Bednja form, there was no
Ivšić's Law in the form *starěšīnNjʄ - the accent was retracted only later by a specific Kajkavian
development. There is another possible example of the final accent in Čakavian (Hvar)
menõn. This could, however, easily be a younger form due to analogy with nouns like ženõn.
Cf. the variant meʂnon in the same dialect.

244 Cf. Stang 1957: 62.

245 The ending –õv is long due to –u becoming asyllabic (-u > -v), which caused compensatory
lengthening of the preceding syllable ( ɶ is expected since the accent was desinential: -ouʂ > -õv).
Vaillant's (1931: 13-14) theory that –ovь comes from pronominal *-ojejNj is not valid.
246 Cf. Jedvaj 1956: 300.
247 I.e. the retraction of internal * ʄ.

However, there is some evidence which seems to point otherwise. In Slovene, after
the prepositions one finds forms maʄno and taʄbo (the second is obviously analogous to the
first one), which are very difficult to explain. Vaillant (1958: 460) derives this from the form
that had the accent on the preposition (like Croatian saʂ mnōm), where the accent was shifted
to the next syllable yielding the usual ʄ, and with the usual development of a lengthened yer
into a. However, Slovene z maʄno cannot be compared to Croatian saʂ mnōm, because ʂ in the
latter form represents an original short neo-acute * ɳ, and not an original circumflex248 – see
below for the development of the Croatian form. An ad hoc solution would be to reconstruct
a prepositional proto-form *sъʂ mъnojNj, which would yield Slovene z maʄno (as suggested by
Vaillant), besides the usual form *mъnojNjʀ, without a preposition, which yielded Slovene
menóʖj. But how is one to explain such a strange form *sъʂ mъnojNj? Again, an ad hoc solution
would be to suggest an analogy with the forms like *oʂtъ mene (the trouble is that even
*meʂne is not attested in Slovene), but that is not even remotely convincing. For now, the
problem remains unsolved.

In Croatian, these forms have the accentuation mnoʄm, saʂ mnōm/zaʂ mnōm (dial.
Čakavian also maʂnōm with a vocalized yer) and toʂbōm (zà tobōm, but also dial. zaʂ tobōm).
Posavina mnoʄm confirms the originality of the ʄ (thus not mnoʄm < **mnõm, which would
also be possible as a development in Neo-Štokavian). Knowing that the circumflex in mnoʄm
is original, the forms saʂ mnōm and toʂbōm are easily explained by Ivšić's Law – the neo-
circumflex made by the contraction of *-ojNj was retracted to a preceding syllable (if there
was one). In mnoʄm, it did not move because the yer disappeared in the first syllable (but cf.
dial. maʂnōm, where the yer was vocalized, and where, expectedly, a retraction has occurred
because of that). In saʂ mnōm, toʂbōm, as already said, the accent was retracted to the
preceding syllable as a short neo-acute by Ivšić's Law: *sъ m(ъ)noʄv > *sъɳ m(ъ)nōv ⇒ saʂ
mnōm, *toboʄv > *tòbōv > *toʂbōv ⇒ toʂbōm. The accent zà tobōm is the original one, while
the accent zaʂ tobōm is analogous to saʂ mnōm. Cf. this kind of secondary retraction of the
original short neo-acute in Croatian proʂlijēm - neʂ prolijēm, but the original lack of it in lıʂjēm
- nè lijēm. The situation in Croatian (toʂbōm ⇐ *toʂbōv < *toboʄv < *toboʂju) is in accord with
Middle Bulgarian мнóΧ, тобóΧ249.

From these forms we can see that Croatian (Štokavian and Čakavian) and Middle
Bulgarian (Evangelie 1139) point to the old *mъnòjNj and *tobòjNj, unlike Slovene and Old
Russian. Since the nominal accent has generalized the variant *-ojNjʀ (Old Štokavian/Čakavian
nogõm, ženõm), it is difficult to understand this *-òjNj in personal pronouns, since other
languages point to *-ojNjʀ, which is also the expected accent on structural grounds, as already
said. The usual way to explain such discrepancies would be to assume *-òjNj in one pronoun
and *-ojNjʀ in the other, but we have no other clues that would point to that direction. One
also has to take into consideration that the mobile paradigm in pronouns is not quite the
same as in nouns (see § 56), and that it is not impossible that *mъnòjNj/mъnojNjʀ represents
an old dialectal isogloss in Slavic. Here, I shall reconstruct both forms in Common Slavic.
See § 78 for further discussion.

248 Which would be shifted progressively in Slovene.

249 Hock 1992 I: 119.

Accent of genitive/locative plural

§ 55 The accent of the genitive and locative plural of personal pronouns is the same in
most Slovene, Kajkavian and Čakavian dialects, cf. Slovene GL. nàs, vàs, Kajkavian (Bednja250)
GL. noʂs, voʂs, Čakavian GL. naʂs, vaʂs (for instance Cres). However, one finds different forms in
the Kajkavian dialect of Prigorje251 – G. naʄs, vaʄs with a neo-circumflex, but L. naʂs, vaʂs.
A question appears whether this could be archaic – that is, has there been a neo-
circumflex in the genitive plural already in Common Slavic252? This is not very likely. Firstly –
the pattern of Prigorje is isolated. Secondly – it could be explained as secondary very easily,
due to the neo-circumflex appearing regularly in the genitive plural, like in nouns like kraʄv,
njîv, sît etc. And thirdly – it is not even certain that the neo-circumflex is of pan-Slavic origin
in the nouns, since there is no trace of it in Russian, cf. Russian G. pl. корóв but Croat. dial.,
Sln. kraʄv, Czech krav.
Thus, it seems safest to assume that in the case of Prigorje Kajkavian forms one
deals with an innovation, not an archaism when it comes to Common Slavic itself (in
Kajkavian-Slovene area it is an archaism – Slovene nàs and Kajkavian naʂs are L. pl., not G. pl.
in origin). The original accent in the genitive-locative plural is the acute – as attested in
Čakavian naʂs, vaʂs, Czech nás, vás etc. For the explanation of Štokavian and Slovak innovative
forms, cf. § 32 and § 34.

Jedvaj 1956: 306.
251 Rožić 1889-1894/116: 130.
252 Stang 1957: 106.

Mobile paradigm of Slavic personal pronouns

§ 56 As is widely accepted, all Slavic morphological paradigms are divided into three
accentual paradigms – a. p. a, a. p. b and a. p. c. These paradigms are not the same
everywhere – differences exist because of the differences of the different parts of speech,
different declension and conjugation types etc. However, accentual paradigms always have
basic things in common – the a. p. a always has an acute on the stem, a. p. b always has the
accent on the first syllable after the stem (or a neo-acute on the stem, if the accent is
retracted), and in a. p. c the accent is mobile – it can be initial (then the accent «jumps» to
the preceding clitic), final, or an acute on the first syllable of the ending.
The mobility can be regarded as a prototype of a paradigm which has to have the
three mentioned characteristics (the initial, final or the acute accent in the first syllable of
the ending), even though not all three must be included in a single paradigm. The way in
which this abstract mobility will appear on a surface in a particular paradigm depends
basically on the type of endings that appear in that paradigm (for instance, some endings
attract the accent, other do not).
One kind of mobility is exactly what we find in personal pronouns (cf. Зализняк 1985:
143 who puts all personal pronouns into Old Russian a. p. c). We find all the traits of a. p. c
in personal pronouns: absolute initial accent - *meʂne, *męʄ, *tyʄ, *teʂbe, *tęʄ, *myʄ, *nyʄ, *vyʄ, *veʖʄ,
*naʄ, *vaʄ253 (like *goʄrdъ «town», *goʄlvNj «head»), final accent - *mь/ъneʖʀ, *mъnojNjʀ, *te/obeʖʀ,
*tobojNjʀ254 (like L. *golveʖʀ, *golvojNjʀ, *gordomъʀ), and old acute on the first syllable of the
"ending" (see below) *naɾsъ, *naɾmъ, *naɾmi, *vaɾsъ, *vaɾmъ, *vaɾmi, *naɾju, *naɾma, *vaɾju, *vaɾma
(like *golvaɾmъ, *golvaɾxъ, *golvaɾmi, *golvaɾma). Of course, the mobile paradigm of personal
pronouns has its own special peculiarities, just like verbs have theirs, adjectives theirs, and
just like the i-stems are different from the o- and u-stems etc. One can compare
pronominal forms like *tyʄ with nouns like *kryʄ «blood», *syʄnъ «son», *teʂbe with *goʄrda,
*te/obeʖʀ with the locative *golveʖʀ, *męʄ with *goʄlvNj, *tobojNjʀ with *golvojNjʀ, *myʄ with *goʄlvy,
*naɾsъ with *golvaɾxъ, *naɾmъ with *golvaɾmъ, *naɾmi with *golvaɾmi, *nyʄ with *goʄlvy, *naɾma with
*golvaɾma etc. One can see that many features are similar to those found in other
morphological types, but nowhere is the exact same pattern found.
The forms *naɾsъ, *naɾmъ, *naɾmi, *vaɾsъ, *vaɾmъ, *vaɾmi, *naɾju, *naɾma, *vaɾju, *vaɾma
have to be explained separately. As clearly seen, these forms did not really have an internal
old acute like *golvaɾmъ, *golvaɾxъ, *golvaɾmi, *golvaɾma. However, this acute behaved
synchronically and was regarded as an internal acute, as well as the accent on the first
syllable of the ending, in connection to the operation of Meillet's Law (cf. § 57).

253 Cf. Зализняк 1985: 143 for Old Russian вΟ which is most likely an “enclinomenon“. The
reconstruction of *naʄ, *vaʄ is based on structural grounds, since the direct evidence does not exist.
254 The instrumental accentual variants *mъnòjNj and *tobòjNj (see § 54) can be regarded as a special
trait of the pronominal mobile paradigm.

Meillet's Law in nominative/accusative forms of personal pronouns

§ 57 In the nominative-accusative of the personal pronouns, we find the forms with the
circumflex in Slavic255, i.e. unaccented words or the words with the absolute initial accent:
*męʄ, *tyʄ, *tęʄ, *myʄ, *nyʄ, *vyʄ, *veʖʄ, *naʄ, *vaʄ. Forms like Old Prussian toū, ioūs or Latvian jũs
point to the Balto-Slavic acute in these forms (see § 81). This discrepancy of Baltic and Slavic
can be explained by Meillet’s Law – the elimination of acute in forms with initial accent in
mobile accentual paradigms256. The original acute in NA. plural/dual is proven by oblique
cases as well – all oblique cases have the acute (*naɾsъ, *naɾmъ, *naɾmi, *vaɾsъ, *vaɾmъ, *vaɾmi,
*naɾju, *naɾma, *vaɾju, *vaɾma). Working out the exact details of the operation of Meillet’s Law
in personal pronouns is not an easy task, as we shall see.
The exact conditions of the operation of Meillet's Law are not certain. Stang (1957:
10) assumes an analogy – the acute has disappeared first in the pretonic syllable in *golvaʀ,
and then by analogy in the accented position as well (*gılvNj > *goʄlvNj). This connects
Meillet’s Law with the loss of acute in pretonic syllables (see below). Rasmussen (1999b:
474-475) explains it by absolute polarization of accent in mobile stems in Slavic, thus
connecting it to the “jumping” of the accent to proclitics. Perhaps one should connect these
two mentioned hypotheses. The loss of acute in *goʄlvNj might have been analogous to
*golvaʀʀ; the form *goʄlvNj had the accent on the first mora (unlike earlier *gılvNj with the
accent on the second mora) and this polarization of accent was carried a step further and
resulted in *naʄ golvNj etc. However, that remains a speculation.
The exact place of Meillet’s Law in the relative chronology is uncertain as well.
However, it seems that it was a rather recent development. According to Holzer 2005 (: 43-
44), Meillet’s Law operated after the metathesis of liquids. For instance, Holzer adduces the
examples of Russian/Ukrainian ланиɴтa, Czech lanitva “cheek” and Russian рамяɴнo “very” (~
Slovene rámeno), where the ra-/la- (and not ro-/lo-) point to the acute in the first pretonic
syllable. If the examples are reliable, they would actually prove that the elimination of acute
intonation in pretonic syllables occurred after the metathesis of liquids. This hypothesis
presupposes that the elimination of pretonic acute and the loss of acute in initial syllables of
mobile paradigms were simultaneous, i.e. a part of the same process (see above).

There are two main problems to be addressed concerning Meillet’s Law in personal
pronouns. The first one is the absence of Meillet’s Law in the form *jaɾ, and the second one is
the question of the operation of the Law in plural – how to explain the forms *myʄ, *nyʄ and
The operation of Meillet’s Law is clear in the 2nd person singular paradigm. The
paradigm is obviously mobile – *teʂbe, *te/obeʖʀ, *tobòjNj/tobojNjʀ (cf. § 56). Thus, the forms
with the initial acute accent obtain the circumflex by Meillet's Law - *tyɾ > *tyʄ, *tęɾ > *tęʄ. A

255 Kortlandt’s doctrine that all personal pronouns in Slavic had an unchanged original acute (cf. for
instance Kortlandt 1997: 29) is unsubstantiated in the material (see § 4, § 6-8, § 13, § 14, § 29 for the
examples). All the forms point to the original circumflex in *ty, *my, *vy – Kortlandt’s hypothesis is
thus completely ad hoc.
256 Cf. Дыбо 1981: 37 for the explanation of *tyʄ, *myʄ, *vyʄ via Meillet's Law.

question appears - if the new circumflex forms came into existence via Meillet’s Law in the
nominative and accusative forms of the 2nd person personal pronoun, how are we then to
explain the preserved acute in one of the nominative 1st person singular forms - *jaɾ (see §
13 for this form)? There is a simple answer to this question. In order for Meillet's Law to
work, one has to have a paradigm – a mobile paradigm. Now, the 1st person singular is a
clear case of mobility, cf. *meʂne, *mъneʖʀ (cf. § 56). Yes, the mobility is indeed there, but
where is the paradigm? That the forms *naʄgъ “naked”, G. *naʄga or *golvaʀ, A. *goʄlvNj belong
to the same paradigm is quite clear. That the forms *tyʄ, *teʂbe, *te/obeʖʀ, *tęʄ, *tobòjNj/tobojNjʀ
belong to the same paradigm is not quite as clear because this paradigm is irregular, but
still, all these forms are similar enough to be considered a paradigm, although a far less
prototypical one than the first two mentioned. Thus, the operation of Meillet’s Law in *tyʄ and
*tęʄ is not surprising, since these two forms belong to a mobile paradigm. But in the case of
the 1st person singular, it was not so. In this case, we had a form which was completely
isolated in a number of ways. Firstly, its nominative forms (*jaɾ and *jãzъ < *jāzъʀ) were
completely different than the rest of its “paradigm”. Secondly, the 1st person nominative
singular forms are also special because there are two of them – which is not the case
anywhere else. Thirdly, the form *jaɾ was not similar to any of the other pronominal
nominative or accusative forms in any visible way (neither in the initial consonant, nor in the
final vowel). Thus this form was simply not a part of the same paradigm as the forms *meʂne,
*mъneʖʀ, *męʄ, *mъnòjNj/mъnojNjʀ. The form *jaɾ was an isolated form (*jāzъʀ as well), not a part
of a mobile paradigm257, and thus, logically, Meillet’s Law did not operate in it and the form
managed to preserve its acute258.

It is easy to understand that *tyɾ > *tyʄ259 because of the *teʂbe, *te/obeʖʀ (a mobile
paradigm), but why did this occur also in the old *myɾ, *naɾsъ, *naɾmъ, *naɾmi? If all the forms
had an acute originally, why was the paradigm *myɾ, *naɾmъ identified with *gılvy, *golvaɾmъ
(the acute mobile paradigm), and not with *ryɾby, *ryɾbamъ (the acute immobile paradigm, i.e.
a. p. a)? The basic problem is in the forms *gılvy and *ryɾby – what was the difference
between them? One can assume that there was no phonetic difference, i.e. that they were
pronounced in the same way. The difference was probably only in the fact that the accent of
*gılvy was not on the same syllable in other cases, while the acute of *ryɾby always remained
on that syllable260. The fact that *myɾ, *naɾsъ (and *vyɾ, *vaɾsъ, and dual forms in line with

257 A suppletivity of completely unrelated (or very different) forms is thus not regarded as the same
paradigm, but as two (or more) different forms. One can set a provisional hierarchy of the prototypical
paradigms – the most prototypical ones would be those that make a clear paradigm, the so-called
regular words like Croatian koʂnj, kònji or English horse, horses. The less prototypical paradigm would
be that of the irregular words, like Croatian tî, teʂbe or English man, men. The least prototypical, i.e. not
really a paradigm, is a case of synchronic suppletivity like Croatian čòvjek - ljuʄdi, doʂbar - boʂljī or
English good – better. Traditional grammars may teach us that the form better is a comparative of
good, but one should know better – especially when talking about what really is in a native speaker’s
mind and not in grammar-books.
258 In the form *jāzъʀ, the acute was, as already said, eliminated in the pretonic syllable.
259 The operation of Meillet’s Law in monosyllables is seen in the 2nd and 3rd person of aorist, cf.
Croatian bıʂh, bî, bî (dòbih, doʂbi, doʂbi) in a. p. c, but šıʂh, šıʂ, šıʂ (ùših, ùši, ùši) in a.p. a.
260 A typological parallel is found in Croatian – the forms noʂgu and kuʂću are phonetically pronounced
exactly the same (in most dialects), the only difference being the way these forms function in other

them) were identified as the acute mobile paradigm (and not as the acute immobile
paradigm) probably has to do with Meillet's Law operating in *tyɾ > *tyʄ (where the mobility
was obvious). By analogy to *tyɾ > *tyʄ, the forms *myɾ, *nyɾ and *vyɾ, which rhymed with *tyɾ
(and the NA. dual forms after them) were also interpreted as having the «mobile» rather than
the «immobile» acute. This points to the conclusion that in personal pronouns, Meillet's Law
was not a completely phonetic change. It was partly analogical – as seen by the operation of
it in *myʄ, *nyʄ and *vyʄ.
This explains the absence of Meillet's Law in oblique plural/dual forms (kao *naɾsъ,
*naɾmъ, *naɾmi itd.) – there was no Meillet's Law because the acutes in these forms were
regarded as a part of the ending, thus word internal, rather than word initial (like in
*golvaɾmъ etc.). If these acutes had been regarded in the system as initial, they would have all
been changed to the circumflex by Meillet's Law (however, that would be very strange since
elsewhere there was no circumflex in those cases – thus, the change never occurred there).
Of course, this behavior was hardly etymological, since originally these forms were to be
analyzed like *naɾ-sъ, *naɾ-mъ, *naɾ-mi (< *nō-su, *nō-mos, *nō-mih1s, see § 69), unlike
*golv-aɾmъ, *golv-aɾxъ, *golv-aɾmi, *golv-aɾma (< *gholHw-eh2mos, *gholHw-eh2su, *gholHw-
eh2mih1s, *gholHw-eh2moh1). The *-aɾ- in the pronouns was originally a part of the root,
while the *-aɾ- in the ā-stems was a part of the ending (or a thematic vowel that makes a unit
with the ending). Thus, one has to suppose a misanalysis in Slavic, which is not problematic
– these endings were synchronically identical.
It can be argued that the operation of Meillet’s Law in *tyɾ > *tyʄ was perhaps what
triggered off its operation in plural (and by analogy in dual). The nominative-accusative
circumflex in *tyʄ, *tęʄ was the ground for the establishment of the circumflex in the
nominative-accusative plural and dual. The analogy was simple, as already said - *tyʄ
influenced *myʄ and *vyʄ in the nominative. The 2nd person accusative plural *vyʄ was the same
as the nominative plural, the 1st person accusative *nyʄ rhymed with the other forms. Thus,
the circumflex was easily established in plural. In dual, the 2nd person nominative was *vyʄ as
well (see § 50). Thus, by analogy to the plural forms, the circumflex was introduced to the
nominative and accusative dual as well. However, there was one more crucial factor that
made this kind of change possible – the already mentioned analogy with the ā-stem acute
mobile plural paradigm. Since there was no Meillet's Law in the DLI. plural (and DI. dual) of
the ā-stems (where the accent was in the middle of the word - *golvaɾmъ, *golvaɾxъ,
*golvaɾmi, *golvaɾma), it is understandable that there was no Meillet's Law in the
morphologically similar cases of personal pronouns in plural/dual261. The parallelism of
personal pronoun plural/dual paradigms and ā-stem plural/dual paradigms is obvious. The
distribution of the acute on the ending and the initial circumflex is the same. The only
difference is that the ā-stems are at least disyllabic in NA. plural/dual and that the oblique
cases are at least trisyllabic, while NA. is monosyllabic in the pronouns and the oblique cases
are disyllabic.

cases (nòga, but kuʂća), or with a preposition (naʂ nogu, ù kuću). These forms, like Common Slavic
*gılvy and *ryɾby are the same on the surface, but different in the way they behave in various
261 It was convenient that N. pl. *my – A. pl. *ny, N. pl. *vy – A. pl. *vy had the same endings as N. pl.
*golvy – A. pl. *golvy (although only accidentally in the nominative, see § 48).

The parallelism of personal pronouns and ā-stems:

N. *vyʄ *goʄlvy < pre-Meillet's Law Slavic *vyɾ, *gılvy

D. *vaɾmъ *golvaɾmъ
A. *vyʄ *goʄlvy < pre-Meillet's Law Slavic *vyɾ, *gılvy
L. *vaɾsъ *golvaɾxъ
I. *vaɾmi *golvaɾmi

D/I. *vaɾma *golvaɾma

Thus, we have concluded that the operation of Meillet's Law in the nominative-
accusative plural/dual of personal pronouns was due to analogy with the regular *tyʄ, *tęʄ in
singular and to analogy with the plural/dual pattern of the acute mobile ā-stems. One might
argue against such an analysis claiming that it is somewhat far-fetched, i.e. not very likely.
However, the Slavic nominative-accusative forms of personal pronouns very clearly attest the
circumflex accent, while the same forms in Baltic point to the acute (as well as the oblique
cases in Slavic). The change of acute to circumflex can be explained via Meillet's Law only 262.
The question is only – why did Meillet's Law operate there? The proposed scheme, although
complicated at first glance in some regards, provides us with the answer to that question.

262 Rasmussen (1999b: 481) explains Slovene tî, mî, vî by his rule that only circumflex is possible in
monosyllables in Balto-Slavic. I cannot accept that rule, not only because of the Slavic aorist forms like
*bıɾ, *šıɾ (Croatian bıʂ, šıʂ), which do not look analogical, but also because of the existence of Slavic *jaɾ
(North Čakavian jaʂ).

Reconstructed Common Slavic paradigms

§ 58 Here, we shall just adduce all Common Slavic personal pronoun paradigms. Most of
the forms and their accents are reflected very well in the daughter languages, but the
accentuation of some of the forms (like *naʄ, *vaʄ) is based on structural grounds263.

1st person singular 2nd person singular

N. *jaɾ, *jãzъ < *jāzъʀ N. *tyʄ

G. *meʂne G. *teʂbe
DL. *mъneʖʀ, *mi DL. *tebeʖʀ, *tobeʖʀ, *ti
A. *męʄ A. *tęʄ
I. *mъnòjNj/*mъnojNjʀ I. *tobòjNj/*tobojNjʀ

1st person plural 2nd person plural

n. *myʄ n. *vyʄ
gl. *naɾsъ gl. *vaɾsъ
d. *naɾmъ, *ny d. *vaɾmъ, *vy
a. *nyʄ a. *vyʄ
i. *naɾmi i. *vaɾmi

1st person dual 2nd person dual

N. *veʖʄ N. *vyʄ
GL. *naɾju GL. *vaɾju
D. *naɾma, *na D. *vaɾma, *va
A. *naʄ A. *vaʄ
I. *naɾma I. *vaɾma

Дыбо 1981: 35, for instance, does not reconstruct the nominative and accusative dual because of
the lack of evidence.


Standard languages are cited. In Old Prussian, the most representative forms are
adduced (other attested variants are found in the discussion).


N. àš tù
G. manęɶs tavęɶs
D. mán táv
A. manè tavè
L. manyjè tavyjè
I. manimì tavimì

n. meɶs jũs
g. mūsų jūsų
d. mùms jùms
a. mùs jùs
l. mūsyjè jūsyjè
i. mumìs jumìs

NA. mùdu jùdu
GL. mù-dviejų jù-dviejų
D. mù-dviem jù-dviem


N. es tu
G. manis tevis
D. man tev
A. mani tevi
L. manĩ tevĩ

n. meɶs jũs
g. mũsu jũsu
d. mumʄs jumʄs
a. mũs jũs
l. mūsuoʄs jūsuoʄs


N. as, es toū, tu
G. maisei twaise
D. mennei tebbe(i)
A. mien tien
I. māim

n. mes ioūs
g. nuson/noūsan iouson/ioūsan
d. nūma(n)s/noūma(n)s iūma(n)s/ioūma(n)s
a. mans wans

Since we have to deal with only three languages here264, we shall not divide the
presentation of the material into separate sections for the discussion of the personal
pronouns in singular/plural/dual. All the forms belonging to one language will be discussed
under the same heading.

264 Conveniently labeled Baltic, although the author does not believe in the existence of the supposed
Proto-Baltic, as already pointed out (see Introduction).


§ 59 In dialects and in Old Lithuanian texts, various forms, different from the Modern
Standard Lithuanian ones, are attested265. The Lithuanian forms are closely related to the
Latvian forms. This is normal, since those two languages form the East Baltic branch of
Balto-Slavic. They are, however, very different from the Old Prussian forms, which are as
close to Slavic as they are to East Baltic (still, there are some developments that are shared
by Lithuanian, Latvian and Old Prussian). This is also one of the facts that points to the
conclusion that West and East Baltic do not represent a unity inside Balto-Slavic. In Old
Lithuanian and in Lithuanian dialects, secondary cases, like allative, inessive etc., have
developed. Since the purpose of this work is strictly comparative, we shall not tackle these
forms, unless they bear some importance in the discussion of cases inherited from Proto-
Balto-Slavic and Indo-European (i.e. the «original» cases). The general characteristic of the
Lithuanian system of personal pronouns is that it is not very archaic. The East Baltic system
of personal pronouns is in general much more innovative than that of Old Prussian and

The nominative singular of the 1st person personal pronoun is àš in Modern
Lithuanian. In some Old Lithuanian texts and in dialects, the form eš is attested, which
seems closer to other IE languages (cf. Latvian es, Latin ego etc.). This is due to the complete
confusion of the original Balto-Slavic *a- (< IE *h2e-, *Ho-) and *e- (< IE *(h1)e-) in
Lithuanian, called Rozwadowski's Change. Old *a- and *e- yield e-/a- indiscriminately in
Lithuanian266. Dialectal (Anykščiai) õš (with õ- < a-) is problematic, since it cannot really be
derived from the older *eš, cf. esmù «am» in the same dialect, with the unchanged *e-267. In
dialects, the adding of various suffixes in both 1st and 2nd person is usual: cf. dialectal ašaĩ,
àšai, àšen268, àšenai; tùje, tùjei, tùjeinas, tùjeino(s), tùjen, tùjenai etc.269
Standard Lithuanian 2nd person nominative singular is tù. This form is ambiguous. It
can be derived from the old *tū, with the length (as in Old Prussian tū, toū and OCS ty) and
with the acute shortened by Leskien's Law, as well as from the old short *tù (as in Latvian tu,
Greek Doric τύ etc.). However, dialectal North-West Žemaitian toʜɳ unambiguously points to
the older short *tù. Thus, the Standard Lithuanian tù is also to be derived from the older *tù.
The possibility that both the form *tū and *tù are attested in Lithuanian dialects is not very
likely. Since the short form is attested in Latvian too, it points to the East Baltic
generalization of the short form.
The development of Lithuanian genitive and accusative forms must be dealt with
together. Numerous different forms exist in dialects and in older texts. The oldest forms

265 For a good overview of various attested complete systems of Lithuanian dialects and Old Lithuanian
texts, see Rosinas 1995: 12-21.
266 Kortlandt forthcoming a.
267 Endzelīns 1971: 186.
268 This –en could very speculatively be related to PIE *-om (> Slavic –ъ), as a reflex of a variant *-em
(Steven Young, p. c.).
269 Senn 1966: 190.

must have been *mene (cf. OCS mene, Avestan mana) and *tewe (Vedic táva) in the genitive,
and *mēn, *tēn270 (Old Prussian mien, tien, OCS mę, tę) in the accusative. The modern
Lithuanian genitive and accusative forms are obviously innovative (as is the whole paradigm).
The old stems *men- and *tew- were changed to man- and tav- in Lithuanian (dialectally
also mun- and tev-) in all oblique forms, thus also in the genitive (and accusative). The old
*mene, *tewe are reshaped to *mane, *te/awe in Lithuanian271 (see § 74-75). The whole
development of the modern forms is to be explained in the following fashion272: the old
genitive form *mane (we shall take the 1st person form as the example, but the same goes
for the 2nd person as well) starts being used for the accusative as well (cf. the secondary use
of mene in the accusative in Slavic). Then, this *mane is contaminated with the old accusative
form *mēn, which probably had an acute - *mēn. Thus, one gets *manēn in both the genitive
and accusative273. Instead of a contamination of *mane and *mēn, one could also imagine
that the form *mēn was just analogically transformed into *m-an-ēn, with the stem *man-
that is found in all other oblique cases. This new form *manęɴɴ looked like an accusative,
because of the nasal. Thus, a new genitive form *manęɴɴs was created with the typical genitive
*-s. The acute was then changed to the circumflex, *manęɴɴs ⇒ *manęɶs, by analogy to the
circumflex in the nominal genitive forms (žiemõs, avieɶs, sunaũs). In the accusative *manęɴɴ,
the final syllable is shortened to manè by Leskien's Law (dialectally also manì for instance)274.
Here is a schematic presentation of these developments:

East Baltic G. *mane, A. *mēn

G. *manēn, A. *manēn

G. *manęɴɴs, A. *manęɴɴ

Standard Lithuanian G. manęɶs, A. manè

Stang (1966: 249) says that it is possible that manè could also be due to the irregular
shortening of the form manęɶ. That is not entirely impossible, but not very likely. Dialectal
accusative form manęɶ probably ows its accent to the genitive. In some dialects, the form
manęɶ appears in the genitive (no *-s was added), while in the accusative, the form manè is

270 The long *ē here seems necessary because of the Slavic form at least.
271 Stang (1966: 250) reconstructs the accent *manè, and Rosinas (1995: 37) *menè and *teuʠè here.
These forms are completely unsubstantiated and obviously biased by the accent of the innovative
Modern Lithuanian accusative forms manè, tavè.
272 Cf. Stang 1966: 249-250.
273 Cf. exactly the same process in Russian, where the new genitive-accusative form меняɴ has
developed through the contamination of the original genitive менe and the original accusative мя, see
§ 27.
274 The reason for the reconstruction of *manęɴɴ for the modern manè is that this is the only way to
explain the nasal in the genitive manęɶs (Stang 1966: 250). If there is a nasal in the genitive, then it
must have been there in the accusative as well. If not for this, one could theoretically reconstruct
*manè > manè.

attested. Arumaa's claims275 that the –ęɶs in the genitive is older than the –ęɶ, because of the
allative –ęɶsp(i) together with the genitive –ęɶ as the only form in some dialects, is not very
convincing. In the dialect of Garliava276, three versions of the genitive form are attested:
manè, manès and manęɶs. The first one could be a regular reflex of the old *manęɴɴ, and the
second one of the old *manęɴɴs without an analogical circumflex (which is also attested in this
dialect). In North-West Žemaitian, forms mõʜneʜs and tãveʜs/teɶveʜs are attested, which could
point to the original accent on the first syllable (as in Slavic *meʂne, Vedic máma). In dialects,
the ending –ies, borrowed from the i-stems, also appears in the genitive singular personal
pronoun forms - manieɶs. In the accusative, Bezzenberger (1877: 162) records a possible,
but uncertain, example of Lithuanian tuwe.

Concerning the confusion of the genitive and the accusative, Stang (1966: 250)
makes two opposite claims on the same page of his book. First, he explains it as obvious –
the merger being due to the fact that the genitive and the accusative of the singular personal
pronouns were syntactically closer to each other than the same cases in the nominal
declension. The reason for this was that the genitives of the singular personal pronouns
were not used in the possessive constructions277. Therefore, the genitive forms had a limited
use - only after some prepositions and in negative sentences. Thus, they were syntactically
more alike to the accusative than is usually the case. A little bit further in the text, however,
Stang says that it is not easy to explain this confusion of the genitive and the accusative. He
claims that the situation in Slavic is different, because there the genitive is used instead of
the accusative for living beings278. In spite of this claim, the use of the genitive form in the
accusative does not appear very strange. The genitive and the accusative were syntactically
close – cf. the use of the accusative in the positive and the use of the genitive in the negative
sentences for the direct object. To avoid the problem of explaining the confusion of the
genitive and the accusative, one can simply take the new accusative form *m-an-ęɴɴ as the
analogically transformed old *m-ęɴɴ, as already suggested. The confusion of the forms *manęɴɴ
and *mane would then be more likely to occur because of the similarity of these two forms.

In the dative, the forms mán and táu are attested in Standard Lithuanian. In Old
Lithuanian, the form mani is also attested (but rarely), from which mán stems (with the
truncation of the final –i). Cf. also dialectal tãvi. The form mán points to the older *màni279.
This is the same ending –i that appears in Old Lithuanian nominal datives like šuni "dog" or
dukteri "daughter". This ending clearly stems from the IE ending *-i, which was the locative
ending in the nominal declension. Instead of the stems man- and tav-, the stems mun- and
tev- also appear in dialects. Žemaitian mun- is secondary according to Stang (1966: 252),
due to the phonetic change of –an- > -un-. The stem tev- appears in some North-West
Žemaitian dialects. Dialectally, the dative forms with the final –ie and –ei are also attested –
mánie/mãnie, mãnei. This ending must stem from the IE dative ending *-ey (cf. Old Prussian

275 As quoted by Stang 1966: 250.

276 Stang 1966: 251.
277 Cf. the Lithuanian possessive forms màno, tàvo – the genitive form is not used in this function.
278 That is, however, secondary in Slavic.
279 Stang 1966: 168.

In Old Lithuanian and in some modern dialects in the genitive, dative and accusative
singular, the short unaccented forms mi and ti appear (also truncated –m, -t). They can be
affixes or independent enclitics. According to Rosinas (1988: 160, 1995: 10), these are used
in the genitive, dative and accusative function. Stang (1966: 253) claims that these forms
were used in the function of the dative and accusative only. The reason for these different
claims must lay in the fact that Stang did not consider the appearance of these forms in
negative sentences important enough to regard them as the genitive unaccented forms as
well. Indeed, all except one of the Rosinas’s examples (1995: 10) of the genitival use of mi
and ti are in negative sentences, the exception being the sentence “pa mi gaylies”280.

The forms of the locative – manyjè, tavyjè, and of the instrumental - manimì, tavimì
are obviously secondary, due to the influence of the i-stems (cf. also the dialectal genitive -
ies forms). Colloquially, these forms appear as manyɶ(j) and manimɶ.

§ 60 The developments in the plural were also far-reaching. Here, the analogical influence
of the nominal u-stems was notable. This was due to the analogical spread of the *-ū- from
the stem of the 2nd person plural nominative form, which was later associated with the *-u-
from the u-stems.

The nominative plural forms are meɶs and jũs. In dialects, the form m÷s is attested,
which is either due to analogy with the 2nd person pronoun or to the original variant *mēs281.
The accent of the form meɶs is analogical to jũs and the circumflex in the 2nd person form is
likewise secondary – the original acute is not only attested in Latvian jũs, Old Prussian ioūs,
but also in the Lithuanian genitive form jūsų. The circumflex of Lithuanian jũs is due to
metatony, the form without it being attested in dialects282. In Standard Lithuanian, the acute
does not appear in the final or the only syllable. Another reason for this change is the
imitation of the pattern of the u-stems, cf. the nominative sūnūs with –ūs and the accusative
sūnus with –us. Thus also jũs : jùs.

The genitive forms are mūsų, jūsų, with the long –ū- originally from the 2nd person
nominative form. This –ū- has also spread into the 1st person oblique cases.

In the dative and instrumental forms, the –u- of the stem is short, which must be due
to analogy with the u-stems. The long –ū-, which is a part of the stem, has been
reinterpreted as part of the ending and thus shortened, like the –u- in the u-stem endings.
The endings are the usual plural endings –ms and –mis. Theoretically, the dative forms
mùms, jùms could be derived from the old *mūms < *mūmus, and not only from *mùm(a)s,

280 For the material and examples of Old Lithuanian mi, ti, cf. Hermann 1926: 15-88.
281 Cf. also the analogical influence of the 2nd person concerning the plural oblique stem.
282 Cf. for instance Otrębsky 1958: 239. In Žemaitian dialects that preserve the original rising
pronunciation of the acute, the form jūs also has the original rising accent.

but the instrumental forms mumìs, jumìs attest the short –u-, which is then to be assumed
in the dative as well283.

In the accusative, one finds the forms mùs, jùs. These forms are in accord with
Latvian mũs, jũs and must be due to the operation of Leskien’s Law. The preforms were
*mūns, *yūns, the first form being analogous to the second one. These forms are usually
explained through the influence of the u-stems as well. Another way to explain them is
through the insertion of the accusative *-n- in the original (nominative) form *yūs. This new
made form, *yū-n-s, replaced the old accusative form, attested in Old Prussian wans and
OCS vy. That the proto-forms were indeed *mūns, *yūns is seen by the fact that dialectally,
nominative forms męɶs and jųɶs are attested. The nasal in the nominative forms is obviously
analogous to the old accusative forms284. The old *jųɴs is also confirmed by the form iunspi
from a Lithuanian text from the year 1711285. It must be noted that mùs, jùs could be
derived directly from *m/nōns, *wōns, as Old Prussian mans, wans and OCS ny, vy. However,
this is not certain, since the *-ū-, originally stemming from the 2nd person nominative plural,
was generalized in all oblique cases in East Baltic.

Unlike other oblique cases (GDAI. pl.), which have the usual plural nominal endings,
the locative has an analogical singular i-stem ending in Standard Lithuanian. This is certainly
due to the influence of the singular forms manyjè, tavyjè286. In dialects, the forms with the
plural locative endings are also widely attested – mūsuosè, jūsuosè (with the o- and u-stem
ending), mumysè, jumysè (with the i-stem ending).

In Old Lithuanian, the nominative form juos and the genitive form juosų are attested.
Stang (1966: 256) also adduces the forms mousu (together with junsu) from the 17th
century, the form mousu probably being an error in Stang’s book, instead of muosu287.
Stang explains this –uo- as a remnant of the old stems *nō-/wō-, attested in OCS na-, va-.
If these forms are reliable, this would mean that the generalization of the *-ū- in the plural
is post-Proto-East-Baltic (see § 67).

§ 61 The Standard Lithuanian (Aukštaitian) dual forms are all innovative, made by the
agglutination of the stems mu-, ju- and the number «two». The 1st person form is obviously,
as usual, analogically based on the 2nd person form (the –u- in the stem), and the initial m-
is due to the influence of the plural form meɶs288. The original *we- (cf. OCS vĕ) is seen
dialectally, in Žemaitian: veɶdoʖ, and in Old Lithuanian289. The form jùdu is ambiguous. It can

283 Endzelīns 1923: 379.

284 Stang 1966: 255.
285 Arumaa 1933: 22.
286 In the singular forms, the i-stems were a major influence, while in the plural, the u-stems had that
287 However, if it was a mistake, it was not recorded in the addenda and corrigenda to his book (1975).
288 Cf. the same development in the Old Czech and Kashubian/Slovincian dual forms, see § 40, § 43, §
289 Cf. Rosinas 1995: 21 for vedu and muddu in the same text.

be derived from the form *yù-du, from the form *yu-dù with a secondary stress retraction,
or from *yū > *yù > *yù-du with Leskien’s Law shortening. Leskien’s Law would have to
operate before the adding of the –du, of course. The same goes for veɶdoʖ as well, it seems.
The form veɶdoʖ is derived from vèdu, in which the short accent can be derived in the same
way as in jùdu.

In Old Lithuanian, the dative-instrumental dual forms mùma, jùma are attested290
(also mùmu, jùmu in dialects). These are probably to be derived from *mumō, *yumō (cf.
OCS nama, vama). Rosinas (1995: 30) also adduces dialectal mumì, jumì as archaic
instrumental dual forms, which are to be derived from *mumī, *yumī. These forms must be
an East Baltic (or Lithuanian) innovation, since the dative and instrumental dual are the same
in OCS and Vedic.
The dialectal accusative form nuodu has often been cited291 as the archaic form and
the only remnant of the old accusative dual form *nō, attested in OCS na for instance. In
Lithuanian, the form nuodu is attested only in the book of folk poetry Lietuvių tauta III
(1920-1922: 417). However, apart from that source, this form is not attested anywhere else.
By checking the manuscript of the poem in which it supposedly appeared, Sabaliuskas (1976)
concluded that the form does not really exist. In the manuscript, there is just the usual
Lithuanian form mudu. The supposed old dual form nuodu is just an error made in print,
when the form mudu from the manuscript was wrongly copied as nuodu. Thus, there is no
evidence for the old dual form *nō in Lithuanian292.

290 Vaillant 1958: 455f.

291 Otrębsky 1956: 139, Stang 1966: 257, Zinkevičius 1966: 306 etc.
292 It is unfortunate that this inexistent form is still found in the handbooks, cf. Fortson 2004: 127.


§ 62 Due to its many changes, Latvian is morphologically less archaic than Lithuanian.
Many of the developments show similarities to Lithuanian. Some of these are to be attributed
to the Proto-East-Baltic, and some are just independant parallel developments293. It is not
always easy to separate the two.

In the nominative singular, the form es and tu are attested. The form eʔs appears
dialectally. The High Latvian form as stems from this. Thus, the a- here has no connection to
the Lithuanian form àš or Old Prussian as.

In the oblique stems, Latvian, just like Lithuanian, has generalized the secondary
stem man-. Unlike Lithuanian, it has generalized the stem tev-, not tav-. In dialects,
however, tav- also appears. In High Latvian, instead of the stem man-, the stem mun-
appears. In some dialects, it appears in all oblique cases and in some only in the dative (cf.
OCS mъnĕ). The stem mun- is also found in the possessive adjective muns in all High Latvian
dialects, even in the ones that only have man (and not mun) in the dative294. Unlike the
Žemaitian mun- (see § 59), the Latvian stem mun- cannot be explained phonetically as a
secondary development.

The Standard Latvian genitive form manis295 can be compared to both the Lithuanian
genitive manęɶs and to the Lithuanian dialectal manieɶs. Both endings would yield the same
result in Latvian. Since the form manęɶs is much more common in Lithuanian, it seems as a
likelier source of Latvian manis as well296. Latvian dialectal man, tev is of the same source as
Lithuanian dial. manè in the genitive. Latvian dialectal manes, teves is due to analogy with
the ē-stems.

The dative forms man, tev (dialectally also tav) are to be derived from the older
*mani, *tevi, as in Lithuanian (see § 59).

For the accusative forms mani, tavi, again more proto-forms are possible -
*manēn297 or *manin, due to analogy with the i-stems298. The first possibility seems more

The instrumental singular has coincided with the accusative in Latvian. In dialects,
the forms manim, tevim are found, with the original i-stem ending, as in Lithuanian manimì,

293 For an overview of Latvian dialectal systems, see Rosinas 1995: 32-33.
294 Endzelīns 1923: 375.
295 The same goes for tavis, of course.
296 Endzelīns 1923: 373.
297 As in Lithuanian dialectal manęɶ, with the same origin, see § 59.
298 Which, like in Lithuanian, have influenced the declension of the singular personal pronouns.

The locative forms manĩ, tevĩ also have the originally i-stem ending, and correspond
exactly to Lithuanian manyjè, tavyjè. The forms manē, tevē, corresponding to Lithuanian
dialectal man÷, tav÷, appear dialectally.

§ 63 The development of the plural forms is almost identical to the Lithuanian one. In the
nominative, the old acute is still attested – meɶs, jũs (dialectally also meʔɶs). The length of the
first form is either due to the second one, or it is a reflex of the original *mēs, and is to be
compared to Lithuanian dialectal m÷s. The short version can be found in Latvian dialects –
mes, meʔs (> mas). This corresponds to Lithuanian meɶs.

The genitive plural forms mūsu, jūsu also correspond to Lithuanian forms. These
forms are attested with all three possible accents – mũsu/jũsu, muʄsu/juʄsu and mūsu/jūsu.
The first one is the original one. The second one is due to the spread of mobility in the East
Baltic (cf. Lithuanian dial. mūsuosè), which must be secondary (see § 81). The third one is
due to analogy with the dative mùms/jùms299.

The dative forms mums/jums are either the original dative or the original
instrumental forms300. Again, all three accents appear here. In mùms, jùms, the short –u- is
due to the influence of the u-stems. The forms mumɶs/jumɶs are to be derived either from the
original *mūms, *yūms, or due to the influence of the nominative and genitive forms with

The accusative forms are to be derived in the same manner as in Lithuanian (see §
59), with the old acute preserved. The dialectal forms muʄs/juʄs owe the accent to the analogy
with the other cases. The forms mùs, jùs are due to analogy with the dative form.

The locative forms mūsuoʄs, jūsuoʄs correspond to Lithuanian dialectal mūsuosè,

jūsuosè. Cf. also Latvian dialectal mumīs, corresponding to Lithuanian dialectal mumysè,
with the original i-stem ending.

The dual is, of course, lost in Latvian. The dialectal Rucava forms mūduiʠ, jūduiʠ are
due to the Lithuanian influence.

299 Endzelīns 1923: 379.

300 There is no separate instrumental form in Latvian.
301 Endzelīns, ibid.

Old Prussian

§ 64 Old Prussian system of personal pronouns is much more conservative than that of
East Baltic. Old Prussian, for instance, preserves the suppletive pattern of the plural forms,
which is completely missing in East Baltic. The problem with Old Prussian is the inconsistent
orthography and the disputable phonetic realization of many forms.

Old Prussian form as is attested 46 times, and form es only twice (in the Second
Catechism)302. At first glance, this resembles the state in Old Lithuanian, where aš/eš are
attested. Kortlandt (2000) believes that the old *e- always and without exception yields a- in
Old Prussian and, accordingly, that only as is the real Old Prussian form for «I». He claims
that the form es is due to a «secondary fronting» and that it has no connection with the
original *es. According to his theory, this had regularly yielded as in Old Prussian. However,
he proposes no explanation for this «secondary fronting», which makes this an ad hoc
solution. It just may well be that the old *e- always yields Old Prussian a-303, but we cannot
just classify Old Prussian by-form es as secondary and disregard it. There is a good chance
that it just might be a remnant of the older form with initial e-. In my view, that seems likely.

In the 2nd person nominative, one finds the following variants: tu (often), tou (often),
toū (6x), tū (once)304. The long vowel corresponds to Slavic ty and not to East Baltic *tu. The
writing toū, with a length on the second part of the diphthong, points to the old acute. The
short tu frequently occurring in the texts probably represents a short variant of this
pronoun, corresponding to East Baltic forms305. The writing <ū> and <oū> is a result of the
development of diphthongization in Old Prussian306. The older pronunciation was [ū] and the
younger [oū]. This can also be seen in the plural personal pronoun stems (see § 65)307.

In the genitive, the forms maisei and twaise appear, clearly originally the genitive
forms of possessive pronouns308. Palmaitis (1976), tries to explain the use of the possessive
form here in a complicated, although possible way. According to him, the unaccented form
*mey (used for the genitive and dative) analogically got the genitival *-s309. This *meys was
remodeled to *meyse, by analogy with the demonstrative pronoun stese. Then this *meyse

302 Stang 1966: 247. For attestations of Old Prussian pronouns, cf. Toпоров.
303 Or in most cases, cf. other examples in Kortlandt 2000.
304 Stang, ibid. Other writing variants were also: tau, du, thou, thu, thuo, ton (obviously a lapsus)
(Schmalstieg 1974: 138).
305 Considering the Indo-European evidence (see § 90), Schmalstieg’s (1974: 139) opinion that the
short forms in East Baltic are “in all probability a later development” is completely unfounded.
306 If <oū> was not really a diphthong, it would be very difficult to explain exactly why the German
scribes would have written down Old Prussian /ū/ with a rising accent as <oū> with a macron on the
second letter. That seems impossible, and, thus, diphthongal pronunciation must have been real.
307 For a discussion on the diphthongization in Old Prussian, cf. for instance Kortlandt 2001.
308 Cf. the Latin genitive meī, also originally the genitive of the possessive pronoun.
309 Cf. Old Latin mīs < *mey-s.

again coalesced with the original (and still used) *mey forming the new form *meysey. This
form was subsequently easily confused with the original genitive maisei of the possessive

In the dative, the pronouns mennei (9x) and tebbei (often), tebbe (4x) appear (cf. also
sebbei attested 3x)310. The same ending *-ey is found dialectally in Lithuanian (see § 59).

In the accusative singular, the archaic forms mien and tien are attested (cf. OCS mę,
tę). How these forms are supposed to be read is, as always in Old Prussian, disputed. Stang
(1966: 248), for instance, thinks that this should be read as [miʠen], while Schmalstieg (1971:
137, 1974: 137-139) thinks that this was pronounced either [men] or [min]. Besides the
forms tien, sien (the reflexive pronoun), there are also enclitics –tin, attested once, and –sin
(9x), -si (4x).

In the instrumental case, the form māim is attested (twice, maim once). This form is
not easy to explain. Berneker (1896: 209) and Trautmann (1910: 269) derived it from the
possessive pronoun stem, like the genitive form, which seems possible311. Stang (1966: 249)
considers a possibility of deriving this form from the older clitic *moy + instrumental *-mī.
For an idea that māim actually represents the form /manim/, corresponding to Lithuanian
manimì, which does not sound too convincing, cf. Schmalstieg 1974: 138. The biggest
problem with this proposition is that one would have to have the stem man- in the
instrumental (like in East Baltic), while Old Prussian knows only men- in the dative. The form
māim is attested one time in the dative function as well. This must be due to a slave
translation from German (cf. German dative mir and mit mir)312.

§ 65 In the nominative plural, the attested forms are mes and ioūs. First one corresponds
to Lithuanian meɶs (the vowel is short), while the other, like Latvian jũs, points to the acute.
The form mas is also attested once in Old Prussian. According to Schmalstieg (1976: 195)
this is either a “scribal error or an indication of the very open pronunciation”. However, this
is neither possible to know, nor all that important.
In the genitive plural, many different written variants are attested: nuson, nusun,
nusan, nusen, noūson, noūson, noūsan, noūsen, nouson. In the 2nd person, the variants
ioūsan, iouson, iousan are attested313. The writing <oū>, as elsewhere, points to the acute.
Both the older monophthongal and the younger diphthongal pronunciation are reflected
here, as in the 2nd person nominative singular pronoun. The vocalism of the stem of the
genitive and dative is clearly due to analogy with the nominative plural of the 2nd person.
There, the *ū is original314. The pronunciation of the genitive plural ending is disputable, as

310 According to a disputed theory of Kortlandt’s (1974), the double consonants in these forms point to
the accent on the syllable after it.
311 In spite of Schmalstieg’s rejection of this possibility (1974: 138).
312 Endzelīns 1944: 132.
313 Schmalstieg 1974: 138-139.
314 Cf. the same independant development in East Baltic.

well as whether it is supposed to be derived from the older *-ōn (like the ending in East
Baltic), or from the older *-on (like the ending in Slavic).

In the dative plural, again many forms are attested315: nūmas, noūmas, noūmans,
noūmons, noumans, naūmans. In the 2nd person: ioūmas, ioūmans, ioūmans, ioumas,
ioumus. For the vacillation of the monophthongs and diphthongs in the stem, see § 64. The
vocalism of the dative plural stem is, as in the genitive plural, due to the influence of the
nominative plural form ioūs. The new vocalism was generalized first in the 2nd person forms,
and then, by analogy, in the 1st person as well. As for the ending, here we find both the
secondary ending –mans, and the original ending –mas. The latter was preserved in personal
pronouns only316.

In the accusative plural, the forms mans and wans are found. The latter corresponds
completely to OCS vy317, and the former has the initial m- by analogy to the nominative
form318. Endzelīns’s (1944: 134) suggestion that the form mans is due to the dissimilation of
the old *nans (cf. OCS ny) could also be correct, in the sense that the occurrence of n-n in
this form could have been a factor in the analogical change of n- to m-.

315 Schmalstieg, ibid.

316 For a discussion on the dative plural ending in Balto-Slavic, cf. Olander 2005.
317 Cf. the suppletion of Old Prussian ioūs : wans, gone from East Baltic and Slavic.
318 The suppletion is preserved in OCS my : ny.

Reconstruction of personal pronouns in East Baltic

§ 66 Here, we shall try to reconstruct the Proto-East-Baltic system of personal pronouns,

i.e. the system out of which Lithuanian and Latvian systems have developed. First, we shall
discuss some of the problematic issues, and at the end, a provisional reconstruction will be

In the case of the nominative singular of the 1st person, there is no doubt that the
form *eś must be reconstructed. Lithuanian vaccilation of e- and a- (Rozwadowski’s Change)
is of secondary nature. In contrast to the Slavic final –zъ, both East and West Baltic show the
variant *-ś, which is due to the generalization of the unvoiced sandhi variant319. Cf. also
Lithuanian iš to Slavic iz < *jьz for the same kind of development. The generalization of the
unvoiced sandhi variant also explains the short vowels in East and West Baltic – if the original
voiced consonant had been preserved, we would expect the operation of Winter’s Law, as in
Slavic. Stang’s opinion (1966: 247) that a dropping of the final *-an can be justified in this
word by its special psychological position is unnecessary. The easiest solution is that there
was no *-an from the beginning320.

Lithuanian clitics mi, ti are usually derived from the older *mie, *tie, derived from the
even earlier *mey, *tey (Otrębsky 1956: 137, Zinkevičius 1980: 50)321, cf. OCS mi, ti.
However, Rosinas (1995: 12) compares these Lithuanian forms to Hittite mi, ti322. Since we
already need to reconstruct *mey, *tey for Slavic (which Rosinas does not reject), and since
the supposed *mi, *ti do not appear anywhere else, except maybe in Hittite, which is highly
questionable, this proposition seems unlikely323. In the 16th century, the dative function was
more frequent, but in the 17th century, it was the accusative that was more frequent, which
according to Stang (1966: 253) proves the original dative function (unlike Rosinas). Hermann
(1926: 82) proposed that –m(i), -t(i) could also have had another source – old unaccented
accusative *me, *te. This could probably also explain numerous examples of Old Lithuanian
–se instead of –si. However, this remains highly controversial (although not impossible).

The most difficult problem in the reconstruction of the East Baltic324 system of
personal pronouns is the ablaut (or quasiablaut) pattern in the singular. When it comes to
the changes of the stem vowel in personal pronouns, it is clear that we are not just dealing
with “regular” ablaut patterns. The different “ablaut-grades” of the stem are mostly due to
soundchange and analogy, and not to PIE or Balto-Slavic ablaut patterns. We shall try to
reconstruct the East Baltic system here (or one of the possible reconstructible systems).

319 Cf. for instance Endzelīns 1971: 186.

320 Thus also Schmalstieg 1976: 195.
321 The development *-ey > *-ie > -i is usually expected under the acute accent. Here however, it
seems to be due to the unaccentedness of these forms (Stang 1966: 254).
322 Cf. for instance Иванов 1965: 230.
323 This would also demand a reconstruction of not only of the IE *moy and *mey (see § 99), but also of
the form *mi.
324 And Balto-Slavic in general.

However, the major part of the discussion should be left for the Balto-Slavic section (see §
71, § 73-76). In East Baltic, stems man- and mun- appear in the 1st person. The stem mun-
is found in some Latvian dialects325. The stem mun- appears in all oblique cases in some
dialects, while in others, it appears only in the dative. Since one has to explain the dialectal
patterns with man- and mun- generalized in all oblique cases, the easiest solution is to
assume that the stem mun- is original in the dative only (man- being the stem of the other
cases). This seems to be confirmed by external evidence (OCS mъnĕ).
In the 2nd person, stems tav- and tev- appear326. The easiest way to explain this is by
assuming the original tev- in the dative and original tav- elsewhere, with dialects
generalizing one or the other later on. The reasons for reconstructing tev- in the dative lie
outside East Baltic. Since we know from other languages (Old Prussian tebbei, OCS tebĕ) that
the original dative stem was really teb-, we can explain the stem tev- in the dative in the
following way - the stem *tew- was regularly changed to *taw- in front of a vowel in Balto-
Slavic, while it remained unchanged in *teb-. Later, this *teb- was analogically changed to
tev- in East Baltic. Of course, one could also reconstruct *teb- here for East Baltic. However,
since there is no concrete evidence for the stem *teb- in East Baltic, it is methodologically
more correct to assume that it was already *tev-. Naturally, we also assume that by that time
the change *ewV > *awV was no more active. One also has to note that in order to assume
this kind of development, one has to take the change *ewV > *awV as unconditional, i.e. as
operational in front of both back and front vowels. Some linguists assume that this change
was carried out only in front of back vowels327. That makes the explanation of East Baltic
forms much more difficult. This change, which obviously has tremendous importance for the
reconstruction of personal pronouns, will be discussed in the Balto-Slavic part (see § 71).
One also has to note that the question of the East Baltic stems in 1st and 2nd person singular
is related to the problem of the stems of possessive pronouns. In Lithuanian, tãvas is
attested, while in Latvian, there are both mans and muns, and tavs and tuvs. See more in §

§ 67 As we have seen, in the plural stems, both the 1st and the 2nd person have a younger
*-ū- in the stem328. It is not certain whether this change was already carried out in Proto-
East-Baltic. Here, we shall assume it did. Although this kind of analogical change can hardly
be problematic, Rosinas (1995: 29) suggests a cause for this change. He relates the
analogical introduction of *yū- in the oblique cases with a regular sound change *wō- > *ō-
in East Baltic329. Thus, the old *wō- in the 2nd person oblique stems (cf. OCS vasъ, vamъ) was
changed regularly to *ō-. This *ō- in the oblique cases was awkward, so *y- from the
nominative *yūs was introduced. Then this new *yō- was finally changed to *yū-, making it,
thus, identical to the nominative in the stem. Soon, the 1st person forms followed, and the
old *nō- was replaced by *nū- and, finally, *mū-330. This kind of scenario seems plausable.

325 It also occurs in some Lithuanian dialects, but it is irrelevant there, see § 59.
326 The possible Lithuanian stem tuv- is very doubtful, see § 59.
327 For instance Stang 1966: 32-33.
328 And, of course, a younger initial *y- and *m- from the nominative as well.
329 Cf. for instance Mažiulis 1970b for this change.
330 It might have been replaced with *mū- at once, of course.

The provisional reconstruction of East Baltic system of personal pronouns:

1st person singular 2nd person singular

N. *èś *tù
G. *màne, *mey *tàwe, *tey
D. *mùni, *mey *tèwi, *tey
A. *mēn, *mey *tēn, *tey

1st person plural 2nd person plural

n. *mès/mēs *yūs
g. *mūsōn *yūsōn
d. *mūmas *yūmas
a. *mūns *yūns
i. *mūmīs *yūmīs

1st person dual 2nd person dual

N. *wè/wē *yù/yū
D. *mūmō *yūmō
I. *mūmō/mūmī *yūmō/yūmī

A few comments concerning the proposed reconstruction are necessary. In the

genitive and accusative I reconstruct *mane and *mēn respectively. It is fairly possible that
the forms like *manēn(s) were already Proto-East-Baltic331, but here, due to various attested
genitive and accusative forms in East Baltic, I have chosen to reconstruct the older phase of
those cases.

Since the stem men- is not attested in East Baltic, I reconstruct *man- in the genitive.
The *mun- in the dative is reconstructed on the basis of Latvian dialectal evidence (see § 62)
and on the basis of the external evidence in Slavic *mъnĕ.

The accent of the original genitive form *mane is not certain. Lithuanian evidence
points to the initial accent in the genitive and dative (see § 59). Thus *màne etc., which is
apparently in accord with Slavic *meʂne and Vedic máma.

The acute accent of *mēn, *tēn is also a bit speculative, but I believe it explains the
final short accent in Modern Lithuanian manè, tavè (cf. Russian меняɴ)332.

The ending *-i in the dative is attested in dialects and in Old Lithuanian. I take it to
be the oldest one, because it seems that it was also the oldest one in PIE (see § 101). The
variant ending *-ey (seen also in Old Prussian mennei) could easily be secondary, but it

331 Cf. Rosinas (1995: 34), who reconstructs *mene/ē, *menes, *menen(s) and *menā (for Modern
Lithuanian possessive màno) for the genitive, and *mene, *menen for the accusative. The stem *men-
here is based on external evidence, of course.
332 Stang (1966: 249) considers the accentuation of *mēn, *tēn unknown.

could also have been Proto-East-Baltic or even Proto-Balto-Slavic. However, it is impossible
to determine that. The locative was perhaps identical to the dative originally, the modern
forms being younger.

I have chosen not to reconstruct instrumental *manimī for the proto-language,

despite the fact that it is possible that it was already there.

As for the clitic forms, only *mey and *tey are reconstructed (confirmed by external
evidence). The forms *mi, *ti are not likely and the evidence for *me, *te is inconclusive (see
§ 66).

In plural, the short *mès is reconstructed, as well as the variant *mēs. It is possible,
however, that the latter is a secondary development due to the influence of *yūs.

The stems *mū- and *yū- are reconstructed throughout, although it is possible that
this analogous development was not entirely finished back in the East Baltic proto-
language333. However, in order to reconstruct the older stage of these stems, we would have
to reach out for external evidence, which we shall try to avoid whenever possible334. For the
ending of genitive plural, cf. § 70.
In all oblique plural cases, the acute has been reconstructed on the stem. This must
be due to analogy with the nominative – the accent was transferred together with the
stem335. The mobile accent in modern dialects is certainly secondary.

In the dual, only three cases are reconstructed. In the case of *yù/yū, the evidence is
not clear and both forms can be reconstructed. It is also possible that both variants existed.

The dative and instrumental dual are reconstructed on the basis of Lithuanian
dialectal evidence, the instrumental being East Baltic (if not Lithuanian) innovation – cf. the
ending *-mī without the plural *-s at the end.

The oblique stems *mū-/yū- are reconstructed in the dual as well, although it is
possible that they were indeed *mu-/yu-, corresponding to the modern attestations of these
forms (mùmu, mùma, mumì), and corresponding to the possible nominative short *yù. If
there was no nominative *yū, it would be imaginable for the oblique dual stems to have a
short *-u-336.

333 For possible traces of the old *-ō- from the Balto-Slavic *nō-/wō- in Lithuanian, cf. Stang 1966:
256 and § 60.
334 This is one of the general problems of reconstruction. Forms can, and do, simply disappear from a
language, and one cannot reconstruct something that has disappeared from the daughter languages
altogether (although, in reality, that form was once indeed there in the system). If East Baltic were an
isolated language group, in all probability there would have been no possibility of reconstructing
anything but *mū- and *jū- in the plural oblique cases.
335 Although the acute was already there in all oblique cases in Balto-Slavic, see § 81.
336 But the long *-ū- could have been introduced from the plural of course.

Reconstruction of personal pronouns in West Baltic

§ 68 Here, we shall attempt to reconstruct the West Baltic, i.e. pre-Old-Prussian, system of
personal pronouns. This is what can be reconstructed based on Old Prussian material
(together with the minimal use of external evidence):

1st person singular 2nd person singular

N. *eś *tū, *tu
D. *meney *tebey
A. *mēn *tēn

1st person plural 2nd person plural

n. *mes *yūs
g. *nūso/ōn *yūso/ōn
d. *nūmas *yūmas
a. *ma/ōns *wa/ōns

The nominative *eś, with initial *e-, is reconstructed on the basis of the Old Prussian
es. If this is really secondary (according to Kortlandt), *a- should be reconstructed. *-ś is
reconstructed on the basis of external evidence of course.

In the 2nd person, besides a long acuted *tū, a short *tu should be reconstructed. It is
not likely that every Old Prussian <tu> was /tū/, and not /tu/.

The older genitive forms cannot be reconstructed on the basis of internal Old
Prussian evidence. One could claim that Old Prussian dative mennei (cf. OCS mъně) is due to
the older genitive *mene, but it could just as easily be analogous to the 2nd person tebbei.

The accentuation of the dative forms *meney and *tebey is not known. If the –nn-
and –bb- really point to the final accent in Old Prussian (Kortlandt 1974), we could
reconstruct *menèy, *tebèy. However, Kortlandt also assumes a forward stress shift in Old
Prussian337. In that case, Old Prussian /mennéi/ could be derived both from the older
*menèy and from the older *mèney, which would render those forms irrelevant.

The pronunciation of Old Prussian mien, tien is disputable, as already said. In spite of
Schmalstieg (1976: 195), who says that he cannot see the evidence for the reconstruction of
*ē in these Old Prussian forms anymore than he can see them anywhere else in Baltic, it
seems far more probable that *mēn, *tēn should be reconstructed here. Original *men, *ten
would be expected to yield **men, **ten, cf. *mes > mes.

337 The hypothesis that the geminates indicate the stress on the following syllable and the hypothesis
of a progressive stress shift are allegedly independent, i.e. one hypothesis does not depend on the
other according to Kortlandt.

The instrumental form is not reconstructed, since the Old Prussian māim is obviously
not archaic.

The analogical *nū- and *yū- are reconstructed in the genitive/dative plural. The
older ending *-mas is reconstructed in the dative, while in the genitive one cannot say
whether one has to reconstruct *-ōn (as in East Baltic) or *-on (as in Slavic), cf. § 70. The
same goes for the accusative, where both *-ōns and *-ans can be reconstructed. However, a
structural analysis points to the pre-forms with *-ōns (see § 79).
The spread of the *-ū-stems in the oblique plural is a pan-Baltic tendency, although
it is not sure whether this is a common development or an independent innovation. It must
be noted that Old Prussian did not introduce new forms into the A. pl. and that it does not
have the stem *mū- but *nū- in the 1. person, unlike Lithuanian and Latvian.


§ 69 Here, we shall reconstruct the Balto-Slavic system of personal pronouns on the basis
of our previous reconstructions of Slavic, East Baltic and West Baltic systems. The proposed
reconstruction will be based mostly on internal Balto-Slavic material, with the use of external
evidence when necessary. Here, the changes occurring in Balto-Slavic and from Balto-Slavic
to Slavic, East and West Baltic shall be considered. The changes that have occurred from
Proto-Indo-European to Balto-Slavic time shall be discussed in the Indo-European section.
Once more, we shall show the reconstructed systems of Slavic (or rather pre-Proto-
Slavic this time, reconstructed with the help of external, Baltic and Indo-European evidence),
East and West Baltic, followed by the Balto-Slavic reconstruction and a discussion:


1st person singular 2nd person singular

N. *ēź, *ēźàn *tū
G. *mène *tèbe
D. *munà/āʱy, *mey *te/abà/āʱy, *tey
A. *mēn *mēn
L. *munà/āʱy *te/abà/āʱy
I. *munayān *tabayān

1st person plural 2nd person plural

n. *mūs *wūs
g. *nōsan *wōsan
d. *nōmas, *nōns *wōmas, *wōns
a. *nōns *wōns
l. *nōsu *wōsu
i. *nōmīs *wōmīs

1st person dual 2nd person dual

N. *wē *wū
GL. *nōyaw *wōyaw
DI. *nōmō, *nō *wōmō, *wō
A. *nō *wō

In the dative-locative singular, one cannot tell whether Slavic *-ĕ represents an older *-ay or
The original accent of the instrumental singular is not certain (see § 54, § 78).
These are pre-Meillet's Law forms.

East Baltic

1st person singular 2nd person singular

N. *èś *tù
G. *màne, *mey *tàwe, *tey
D. *mùni, *mey *tèwi, *tey
A. *mēn, *mey *tēn, *tey

1st person plural 2nd person plural

n. *mès/mēs *yūs
g. *mūsōn *yūsōn
d. *mūmas *yūmas
a. *mūns *yūns
i. *mūmīs *yūmīs

1st person
person dual 2nd person dual
N. *wè/wē *yù/yū
D. *mūmō *yūmō
I. *mūmō/mūmī *yūmō/yūmī

West Baltic

1st person singular 2nd person singular

N. *eś *tū, *tu
D. *meney *tebey
A. *mēn *tēn

1st person plural 2nd person plural

n. *mes *yūs
g. *nūso/ōn *yūso/ōn
d. *nūmas *yūmas
a. *ma/ōns *wa/ōns


1st person singular 2nd person singular

N. *ēź, *èś, *ēźàn *tū, *tù
G. *mène, *mey *tàwe, *tey
D. *mùni, *mèni, *mey *tèbi, *tùbi, *tey
A. *mēn, *mey (?) *tēn, *tey (?)
L. *tawày (?)
I. *menàyān (?) *tawàyān (?)

1st person plural 2nd person plural

n. *mès, *mēs *yūs
g. *nōso/ōn, *nōns *wōso/ōn, *wōns
d. *nōmas, *nōns *wōmas, *wōns
a. *nōns, *nōns *wōns, *wōns
l. *nōsu *wōsu
i. *nōmīs *wōmīs

1st person dual 2nd person plural

N. *wē, *wè (?) *yū, *yù (?)
G. *nōyaw, *nō *wōyaw, *wō
D. *nōmō, *nō *wōmō, *wō
A. *nō, *nō *wō, *wō
L. *nōyaw *wōyaw
I. *nōmō *wōmō

General notes on reconstruction of Balto-
Balto-Slavic personal pronouns

§ 70 In the 1st person nominative singular, three forms must be reconstructed. The form
*èś was just a sandhi variant of the primary form *ēź in front of voiceless consonants. The
form *èś is generalized in East and West Baltic, while the other two forms are reflected in
Slavic. The difference in length is explained via Winter's Law, which operated in front of
voiced *ź < *gɴ, but not in front of voiceless *ś < *kɴ. In the form *ēźàn, Winter's Law operated
in the closed syllable338, cf. PIE *egɴHóm (see § 101).
One could presume that positing three forms for “I”, *ēź, *èś and *ēźàn, is
inconceivable. However, *ēź and *èś were not really different forms, as already said. At first,
the alternation was automatic and then simply inherited – the form *ēź appearing in front of
voiced and *èś in front of unvoiced consonants. It is also possible that one of the forms (*ēź
or *èś) was generalized in various dialects already in Proto-Balto-Slavic. This kind of
alternation, especially in the light of the presence of yet another form, *ēźàn, must have
been obliterated by generalization very early.

In the 2nd person nominative singular, two forms are reconstructed - *tū and *tù. The
first one is reflected in Slavic and Old Prussian, and the second one in East Baltic and
probably in Old Prussian.

The existence of Balto-Slavic *yù is not certain, since Lithuanian jù-du points to
either *yū or *yù. The form *yū is confirmed by Slavic vy, so it must be reconstructed in
Balto-Slavic. The form *yù might also have existed in Balto-Slavic, but there is no way of
proving it.

The accent in the genitives *mène and *tàwe is reconstructed on the basis of Slavic
*meʂne, *teʂbe and Vedic máma, táva. Lithuanian also seems to point to the initial accent (see
§ 59), but since the original form is not preserved, the data is inconclusive.

In the dative, Lithuanian mán, táu point to the initial accent, while Slavic points to the
final accent. Slavic has probably innovated here, this innovation perhaps being related to the
new ending. The exact origin of the Slavic final accent is not clear. In Slavic, the ending *-ĕ is
stressed in the locative singular of a. p. c ā-stems, and it is unstressed in the dative singular
of a. p. c ā-stems and locative singular of a. p. c o-stems. The ending *-i is reconstructed
because the IE data seems to point to the conclusion that this ending is the oldest one (see §
101). This idiosyncratic ending was changed in different ways, perhaps already in Balto-
Slavic. In East Baltic, the ending *-ey (from the consonant and i-stems) also appears, and the
same ending is the only one attested in Old Prussian. Slavic ending *-ĕ is often interpreted
as DL. ending of ā-stems339. However, it could also be a reflex of PIE *-oy, the locative
ending of o-stems (see § 72 for that theory). The ending *-ĕ in the DL. must have been one
of the reasons for the introduction of the instrumental ending *-ojNj to ā-stems in Slavic.

338 Matasović 1995.

339 Cf. for instance Birnbaum & Schaeken 1997: 74.

Since the dative-locative singular ending was the same in the ā-stems and personal
pronouns340, this was the basis for the introduction of the pronominal instrumental ending
*-ojNj to ā-stems341.

Slavic *męʄ, *tęʄ, Lithuanian manè, tavè and Old Prussian mien, tien seem to point to

In the nominative plural of the 1st person personal pronoun, we have reconstructed
the form *mès together with the variant *mēs. The latter is reflected probably in Latvian meɶs
and Lithuanian dialectal m÷s. Latvian and Lithuanian dialectal forms could also be explained
via the influence of the 2nd person plural forms or the 1st person plural verbal ending *-mēs
(whatever its origin). The variant *mēs, if archaic, is a case of monosyllabic lengthening342.

The genitive plural ending had two variants: *-on and *-ōn343. The first one is
reflected in Slavic, and the second one in East Baltic344. The languages have obviously
generalized the same ending as in nouns.

The ending *-yaw in the genitive-locative dual has been taken over from
demonstrative pronouns (cf. OCS toju), where the *-y- was part of the stem345.

The DALI. plural and DI. dual have the endings identical to the nominal ones. This is
mostly due to Balto-Slavic innovation. The accusative dual endings in *nō, *wō were identical
to the accusative dual of the o-stem ending in Balto-Slavic. Both endings stem from PIE *-
oh1 (see § 95).

340 Whatever the ultimate origin of the pronominal dative-locative *-ĕ is.
341 The reason for the elimination of the old IE ending of the ā-stem instrumental singular was that it
was the same as the nominative singular endings, *-ā (PIE nominative *-eh2 and instrumental *-eh2eh1
have both yielded *-ā in Slavic regularly). The Lithuanians, however, do not seem to mind the fact that
both the nominative and the instrumental ending is –à in Lithuanian.
342 Cf. Stang 1966: 254.
343 We shall consistently write *-n for the Balto-Slavic reflex of PIE *-n and *-m.
344 The reconstruction of Lithuanian –ų < PIE *-ōm presupposes that PIE *-om regularly yields
Lithuanian –ą in A. sg. of o-stems. If one wants to explain this accusative form as secondary to the
nominative –as, it is possible to derive Lithuanian –ų from PIE *-om, like Slavic –ъ. In that case, one
should reconstruct Proto-Balto-Slavic *-on only. However, we shall provisionally reconstruct both *-ōn
and *-on in the genitive plural.
345 The PIE dual ending was actually *-ow.

Problem of *ew > *ow in Balto-

§ 71 Before we discuss the problem of the singular stems in Balto-Slavic personal

pronouns, we must discuss the problem of the change of heterosyllabic *ew > *ow in Balto-
Slavic. Obviously, the question of whether PIE *tewe would regularly remain *tewe in Balto-
Slavic, or whether it would yield *tawe, is of great importance in the matter of the
development of the singular stems of personal pronouns.
There are two basic approaches here. One is that the change of heterosyllabic *ew
was unconditioned346, and the other one is that is occurs only in front of back vowels347. The
relative chronology of this development is also disputed, i.e. whether this is a Balto-Slavic
development or an independent innovation in Baltic and Slavic348. There seems to be no
reason to question Balto-Slavic origin of this change349.

Mikkola (1913: 44) takes the change as regular in front of all vowels and explains
apparent exceptions as secondary. Thus OCS devętь and Lith. devynì (instead of **novętь
and **navynì) are due to analogy to desętь and deɶšimt. This is quite acceptable, since we
need the analogy to explain the initial d- anyway. He explains the lack of the change in OCS
nevĕsta and nevodъ by assuming that these are compounds, which in turn explains why ne-
was not changed here. This is also a satisfactory explanation. Even if the quoted words are
not real compounds (the etymology of nevĕsta is disputable), it is certain that they could
have been analyzed as such, which makes the question irrelevant.
Vaillant (1950: 110) thinks the change was regular only in front of back vowels and
he gives an example of OCS revNj/rovNj «roar», where the alternation of –ev-/-ov- is to be
explained as the different generalization of the older *rovNj - *revetъ. This example seems
convincing at first glance, but the forms reveši, revetъ, revemo etc. could also be explained
by the assimilation of *o-e > e-e (*rovetъ > revetъ etc.) or by analogical reintroduction of
the e-grade into the root.
Shevelov (1964: 357-359) takes the forms novъ : nevĕsta as his Paradebeispiel, but,
as already said, even if these two words are of the same origin, the other could have easily
been misanalyzed as a compound. He also adduces OCS adverb drevlje «once» (obviously
secondary, since one would expect *-u- here, and not –ev-) and various Slavic reflexes of
the adjective *drevьnъ «ancient». The latter, however, is surely not archaic and can be
explained as a younger derivative350. Lithuanian dreɶv÷ «hollow tree trunk» and Latvian dreve
are deemed irrelevant by Arumaa (1964: 69), since –÷ here is derived from the older *-yā
(Old Indic dravya- «belonging to a tree»)351. Dialectal Russian adjective нéвенний «thin» that

346 For instance Mikkola 1913: 44, Arumaa 1964: 68-69.

347 For instance Vaillant 1950: 110, Shevelov 1964: 357-359, Stang 1966: 32-33, Endzelīns 1974: 32.
348 The same change also occurs in Celtic and Italic languages.
349 Shevelov’s (1964: 359) argument against the Balto-Slavic origin of this change makes no sense.
350 The etymology of this word is also not completely clear. It is traditionally related to the PIE stem
*d(e)r(e)w- “tree”, but the semantic development “tree” > “ancient, old” is hardly straightforward.
351 Cf. also Lithuanian dravìs, Latvian drava “Waldbienenstock”, Old Prussian drawine “Beute, Bienenfaß”
and Latvian drava “Höhlung im Bienenstock” (Pokorny). These are not necessarily the reflex of PIE *-

Shevelov quotes (making a possible connection with нaв «dead») does not look very relevant.
Shevelov takes the trouble to explain some of the supposed paradigmatic generalizations.
He explains Slavic slovo, slovese as generalized to the nominative-accusative singular. That
is very possible, in spite of the fact that in most cases, according to him, *slev- would be
expected. However, he also explains the nominative plural synove as secondary to the
genitive plural synovъ. He does not even mention the dative singular form synovi. It is not
very clear why –ev- in the dative singular and nominative plural would be changed to –ov-,
due to the influence of the genitive plural. That kind of analogy seems very improbable -
why and how would genitive plural influence dative singular? In any case, the possibility of
analogical generalization in the root, like in slovo, is understandable. However, we are
dealing with endings here. Why would the supposed endings *-evi, *-eve and *-ovъ, which
are not very close to begin with, be remodeled to –ovi, -ove and -ovъ? And more so, why
would they be remodeled by analogy to genitive plural352, which is itself only one form, in
comparison to two forms with the supposed regular *-ev-? This kind of analogical change
seems very strange and highly unlikely. Shevelov also adduces Russian можжевельник
«juniper» as supposed evidence of the old *-ew-, but he fails to notice the -жж- in front of
it. Shevelov also explains the present forms like plovNj, slovNj, trovNj etc. by the generalization
of –ov- in the paradigm. This is possible, in spite of the fact that one has to assume that the
stem of only two forms (1st singular and 3rd plural) was victorious in the end.
Arumaa (1964: 68-69) takes another approach. He believes that the change is regular
in all cases, but that there are a few cases in which the change was stopped due to the
influence of the following front vowel. He ascribes this to a tendency, rather then to a strict
rule. This looks like an ad hoc solution, but it cannot be ruled out, since this kind of sporadic
tendencies have been known to occur.
Stang's (1966: 32-33) main examples are Latvian personal pronouns like tevis, tevim
etc. However, he does not mention that the forms with tav- are attested in dialects and in
Lithuanian, nor does he mention that this tev- is easily explained as analogical due to the
influence of the stem *teb- (cf. Old Prussian tebbei). Stang also mentions two Lithuanian
examples that need to be explained as secondary in order for his claims to be correct:
Žemaitian šlavù (which he claims can also be derived from PIE *-ow-) and Daukša's
sráwanczio (IE *srew-), which he does not explain.

If one assumes that *ew > *ow occurred only in front of back vowels, the pronominal
forms in *taw- are very difficult to explain (see § 74). The evidence adduced in order to
support the theory that the change occurs only in front of back vowels is not very strong and
is definitely inconclusive. All the examples can be explained differently (see above), while
some of the supposed analogical changes (like the one in the u-stems) seem highly unlikely.
What is also suspicious is that it seems that it is always –ov- that is generalized in the
paradigm. Why would that be so, if both *-ov- and *-ev- were there earlier, and if the forms
in *-ev- were even more numerous353? The only possible answer seems to be that the
change of heterosyllabic *ew to *ow was regular in all conditions. This is clearly seen in the
paradigms, where only –ov- is ever attested. In some of the other examples, the change was

352 Not nominative-accusative singular as in slovo, which are likely to be a role model for other cases.
353 Like in the present tense of words like plovNj and in the u-stems.

reversed due to various additional processes. The advantage of this interpretation also lies in
the fact that in this way, as already mentioned, the development of the 2nd person singular
personal pronoun oblique stems is much more easily explained.

Balto-Slavic singular stems of personal pronouns

§ 71 Concerning the stems in the singular oblique cases in Balto-Slavic, there are many
problems to be dealt with. These problems involve some forms in Slavic, but East Baltic is
notorious in this respect. The problem of the East Baltic singular stems of personal pronouns
is that they make no sense in the wider scope of IE, and are, of course, very difficult to
account for. For instance, one can note that Stang (1966: 253) says that he will not even try
to explain the origin of East Baltic man-. Since this work deals with the reconstruction of
Balto-Slavic personal pronouns per se, cannot afford that kind of luxury.

What is there to be found in the genitive and dative-locative singular354 stems in

Balto-Slavic? In Slavic, there is men- in the genitive, mъn- in the dative-locative355, teb- in
the genitive-dative-locative and tob- in the dative-locative. In Old Prussian, there are menn-
and tebb- in the dative. In East Baltic, it is a little bit more complicated. In Lithuanian, one
finds man-356 and tav-, tev-, depending on the dialect357. In Latvian, one finds man-, mun-
(in all oblique cases or in the dative only) and tav-, tev-, depending on the dialect. Here is an

Slavic G. men-, teb-, DL. mъn-, teb-/tob-

Old Prussian D. menn-, tebb-

Lithuanian GD. man-, (mun-), tav-, tev-

Latvian GD. man-, mun-, tav-, tev-

If we start from PIE and Balto-Slavic, we can expect *men- (like in Slavic and Old
Prussian), *taw- (like in Lithuanian and Latvian) and *teb- (like in Old Prussian and Slavic). By
assuming a simple analogy of *teb- > *tew- (via the influence of *taw-), one also gets *tew-
(attested in Lithuanian and Latvian). This also goes for *tab- (in Slavic), due to analogy with
*teb-, instead of older *taw-. The Balto-Slavic stem *men- is derived from the PIE G. sg.
*men-, *taw- from the PIE G. sg. *tew- and *teb- from the PIE D. sg. *tebh-.

In Old Prussian, the dative stems menn- and tebb- need no special explaining. The
same goes for the Slavic men- in the genitive.

354 The Slavic instrumental form has been explained elsewhere (see § 77). The East and West Baltic
instrumental forms are secondary, as well as the East Baltic locative forms.
355 This seems to be the oldest form, according to the comparison with Latvian.
356 And also mun-, but probably always secondary (see § 59).
357 Perhaps also tuv-, see § 59.

Slavic *tebě/*tobě

§ 72 Slavic teb- (instead of **tov-) in the genitive and teb-/tob- (instead of only *teb-) in
the dative need some explaining. The *-b- instead of *-v- in the genitive is explained by a
simple analogy to the dative form, and the same goes for the *-e- instead of *-o-. Thus,
analogically to *tebě, we get *tebe instead of *tove358. Nevertheless, how does that explain
the two variants in the dative? This can be interpreted in two ways.
When the genitive form *tove changed its *-v- to *-b-, a form *tobe was made. This
genitive form *tobe gained a secondary variant *tebe not only by analogy to the dative *tebě,
but also by analogy to *mene. Thus, we got two forms in the genitive - *tebe and *tobe. By
analogy to this, in the dative singular, a secondary *tobě was also created - in addition to
the original *tebě. Thus, for a while there were *tebe, *tobe in the genitive and *tebě, *tobě
in the dative. Then the form *tebe completely pushed out the other form in the genitive. This
was due to the powerful influence of the form *mene in the 1st person pronoun. However, in
the dative, there was *mъně359 in the 1st person pronoun. This form could not influence the
variants *tebě and *tobě, and thus, both were preserved in Common Slavic and reflected in
Slavic languages360. Since there is no reflex whatsoever of the reconstructed Slavic *tobe in
the genitive, we can safely assume that this form had disappeared from Slavic before its
Here is a schematic presentation of these changes (it is not possible to ascertain in
details at exactly which stages particular changes occurred):

Indo-European G. *mene, *tewe, D. *tebhi

Balto-Slavic G. *mene, *tawe, D. *tebi

Proto-Slavic G. *mene, *tabe, *tebe, D. *tebay

Proto-Slavic G. *mene, *tabe, *tebe, D. *tebay, *tabay

Common Slavic G. *mene, *tebe, D. *tebĕ, *tobĕ

Another way to explain Slavic *tebě/tobě in the dative is to start with a locative form
*twóy in pre-Balto-Slavic (cf. Old Indic tvé)361. In Balto-Slavic, this *twóy would by analogy
become *tewóy, which would later yield *tawáy regularly. This form would yield *tabáy in
Slavic, since the *-w- of the 2nd person singular was always replaced by analogy to the *-b-
there. This *tabáy would yield *tobeʙʀ in the locative, and the ending *-ě would be taken over
to the dative *tebě.

358 Here, for the sake of simplicity, we shall use the traditional Slavic notation of the reconstructions
(i.e. *v instead of *w, *o instead of *a, *ъ instead of *u etc.).
359 And secondary *mьně.
360 Cf. the same explanation for the preservation of the dative *tobě and the disappearance of the
genitive *tobe in the frame of a different theory in Hujer 1911.
361 For *twoy, cf. for instance Stang 1966: 253, Schmidt 1978: 130-131.

Vaillant (1958: [249]) and Kiparsky (1967: 133) assume that the stem *tob- was
originally found only in the locative. They, however, came to this conclusion through an
analysis of various derivatives of the stem *sob- in Slavic (see § 47), and not by deriving the
forms from PIE like here. However, since the only Slavic language that has a difference of the
stem in the dative and locative, Slovincian, shows exactly the opposite distribution of these
stems362, it seems probable, that if this scenario is true, the forms *tebě and *tobě
functioned both as the dative and as the locative in Common Slavic.
The advantage of this theory is: not only does it explain the variation of *teb- and
*tob- in the Slavic dative-locative, but it also explains the ending –ě363 and the dative-
locative final accent *t(e)wóy > *tobeʙʀ364. Thus, one should perhaps reconstruct *tawáy in the
locative. Also, one cannot rule out the possibility that both of our solutions for the
development of the Slavic dative-locative *teb-/tob- are correct. It is impossible to
determine which of the solutions is more useful, but the explanation involving the proto-
form *twóy has an advantage that it explains everything and not just the teb-/tob-

362 *tob- in the dative and *teb- in the locative, see § 25.
363 The explanation that –ĕ was taken from the ā-stems lacks a convincing reason for this sort of
364 The accent of *mъneʖʀ would then be analogous to the 2nd person form.

Balto-Slavic *muni

§ 73 In the dative singular, we have reconstructed Balto-Slavic *muni, on the basis of

Latvian dialectal mun-, attested only in the dative in some dialects, and on the basis of the
OCS mъn- in the dative365. Since these two forms cannot be explained as secondary in any
convincing way, they must be an archaism going back to Balto Slavic. However, this form is
very difficult to explain. From the PIE *megɴh- (see § 101), one would expect Balto-Slavic
*men- here (since no *meź- is attested).
The unusual stem *mun- can perhaps be explained in the following way – besides the
form *tebi in the dative366 there was also a variant *tubi in Balto-Slavic, coexisting with *tebi
for some time. This *tubi probably has the same origin as Old Indic túbhyam367. This Balto-
Slavic-Indic isogloss could be regarded either as a PIE dialectalism368, or as an independent
innovation (see § 98 for the discussion on PIE reconstruction). However, there is no tub-
attested anywhere in Balto-Slavic369, which makes this supposition speculative. What
happened next is that this variation *tebi/tubi also created new forms *meni/muni in the 1st
person. The dative stem *men- is attested in Old Prussian mennei, if that is not analogous to
tebbei. The dative stem *mun- is attested in Latvian and OCS370, and *teb- is attested in
Slavic and in East Baltic371. The stem *tub- is probably not attested anywhere, as already
said, which is the biggest problem of this hypothesis372.
There are a couple of other solutions by which we can get Balto-Slavic *tu- in the
oblique cases, which would in turn cause the making of *mun-. One is by supposing an
analogy to the possessive pronoun *tuwos (cf. Latvian dialectal tuvs), a by-form of *twos (cf.
Old Indic tvás and Greek σó̋)373. However, it is not clear why the possessive would influence
the dative.
Another possibility is the existence of old locative *twoy (see § 72), which would have
a by-form *tuwoy. This could then be mixed with the dative *tebi and thus new forms with
*tub-/tuw- would appear in the dative. However, the same problem remains in all three
hypotheses – *tub- or *tuw- are attested nowhere in Balto-Slavic (except for a slight
possibility of the attestation of tuv- in Lithuanian, as mentioned).

365 For mъn- in the instrumental, see § 77.

366 Attested in Old Prussian tebbei and Slavic tebĕ.
367 Cf. Endzelīns 1971: 187.
368 Kortlandt forthcoming c, reconstructs PIE dative *h1mighi, *tubhi and Balto-Slavic *minoy, *tubhoy (I
have normalized the reconstructions) based on (within Balto-Slavic) Polish mnie, tobie, sobie (and the
corresponding forms in Czech and Old Russian), East Baltic dial. mu- from tu- and Old Prussian subs
“self”. Outside Balto-Slavic he adduces Sanskrit túbhyam and Oscan sífeí. However, he does not note
that Polish mnie does not tell us whether there was *-u- or *-i- in Balto-Slavic. He also does not
explain exactly how *tubhoy (= Common Slavic *tъbĕ) would yield Polish tobie, not to mention Czech
tobĕ (in Russian, at least strong *ъ > o). For the reconstruction of PIE forms see § 98, § 101.
369 Except for the one possible occurrence of Lithuanian tuv- in the accusative, see § 59.

370 Perhaps also in Lithuanian, if some of the examples of mun- are old.

371 Changed analogically to tev- there.

372 Cf. also Stang 1966: 252.

373 Cf. also Stang 1966: 252 for *tuwos.

Balto-Slavic *mun- would be a regular reflex of the PIE *mnʘ-, but it would be very
difficult to explain the appearance of the zero grade in front of a vowel. This would
necessitate the introduction of a laryngeal, thus *mnʘH-. Ranko Matasović (p. c.) suggests an
analogy to the 1st person singular middle present ending *-h₂ey – thus *bȹeroh₂ey mnʘh₂ey “I
carry to myself”, parallel to the introduction of the 1st person singular active present ending
*-oh₂ into *egɴoh₂ from *egɴoh₂ bȹeroh₂ “I carry”. This idea is interesting but the development
would be rather unusual.
Matasović (p. c.) offers another possible solution. One should start form the PIE 1st
person dative singular clitic *-mu, preserved in Hittite –mu (dative-accusative clitic) and Old
Irish –m- (a form of infixed pronoun and 1st sg. pronominal ending of pronominal
prepositions). In Old Irish, the fact that this –m- is never palatalized would point to the
proto-form *-mu, cf. duit "to you" < *to-tī < ?*to-toy, but dom < *to-mu (and not **duim).
In Balto-Slavic, this *mu- would be trivially extended with the *-n- from the genitive *mene.
In this case, the form *mu would be an archaism and the form *moy (attested in Greek and
Old Indic) an innovation, perhaps by analogy to *toy.
It is not easy to determine which of these theories is the most probable. Here, I shall
tentatively support the idea of PIE *túbȹi as a possible source of Balto-Slavic *muni (see §
98), but the question remains open.

East Baltic tav /tev-
-/tev -

§ 74 If one assumes that *ew > *ow is regular in all conditions in Balto-Slavic, the East
Baltic distribution of tav- (from Balto-Slavic *taw- < IE *tew-) and tev- (from Balto-Slavic
*teb-, with an analogical *-w-) is easily explained. However, if one assumes the change of
*ew > *ow only in front of back vowels, the stem tav- in East Baltic is not so easy to account
for. Stang (1966: 251) assumes the influence of the possessive *tavas < *tewos, or of the
possible instrumental *taway- < *teway- (1966: 253). It is also possible to assume the
locative *twoy ⇒ *tewoy > *taway (see § 72). However, these solutions are not very elegant.
It is hard to imagine why the stem tav-, originally supposedly found only in the instrumental,
locative and the possessive form, would become the only stem in many East Baltic dialects.
Not to mention the fact that it has influenced the original *men- to become man- (see § 75).
Since the reconstruction of the pre-forms of these supposed forms is not very secure in any
case, one can only conclude that such explanations cannot really explain the facts.
Therefore, our assumption on the overall regularity of *ew > *ow in Balto-Slavic (see § 71)
must be correct.

East Baltic man -

§ 75 The next problem to be dealt with is the mysterious East Baltic stem man-. As is
known, besides the stem mun-, which we have already tried to explain, East Baltic only has
the stem man-. The stem **men- is not attested anywhere. This seems strange at first
glance, since both tav- (from Balto-Slavic *taw- < IE *tew-) and tev- (from Balto-Slavic *teb-
, with an analogical *-w-) are widely attested. The –a- in man- cannot be original (cf. Slavic
G. mene, Old Prussian D. mennei) - it must be secondary. The obvious source of analogy is
tav-. The lack of the stem *men- in East Baltic does not perhaps need to worry us. If we are
to assume that the analogical change of *men- ⇒ *man- occurred already in the days when
there was no separate locative and instrumental, we would simply get *mane in the genitive
and *muni in the dative (with a variant *mani, instead of *meni). It was enough to change the
stem analogically in one case only (in the genitive), and the stem *men- was lost for good.
When the separate locative and instrumental emerged, the stem man- was the only one
Although there seems to be no other choice than to regard the stem man- as
analogous to tav-, Stang (1966: 253) notes that the problem lies in the fact that the stem
man- in Latvian is most scarcely attested375 exactly in those dialects where only tav- and
sav- is attested, not only in the dative, but also in the possessive. However, this could be the
result of a later development.

374 Together with mun-, of course.

375 I.e. the stem mun- is attested there.

Possessive pronouns in East Baltic

§ 76 The possessive pronouns in East Baltic have often been connected to the oblique
singular personal pronoun stems. Thus, mun- in the oblique cases of the 1st person personal
pronoun in Latvian was explained as a result of the influence of possessive pronouns.
However, the reason for this possible analogy is not at all clear. The origin of the mun- in
the possessive muns, attested in almost all High Latvian dialects, is also uncertain. That is a
clear case of obscurum per obscurius.
The explanation of muns by analogy with the possessive tuvs, suvs (from *tuwos)
seems highly unlikely, since exactly those Latvian dialects that have muns, very often have
only tavs, savs376.
One would be more prone to believe that the possessive forms like muns were built
by analogy to the oblique cases with the stem mun-. Such a premise could also lead us to
believe that tuvs came from the oblique tuv- (otherwise not attested, see § 71). This is,
however, very speculative. The –u- in tuvs could be due to analogy with the nominative tu
just as well, which seems like a good solution.

376 Stang 1966: 252.

Slavic instrumental
instrumental singular

§ 77 In the instrumental singular, the East and West Baltic forms seem very recent (see §
64, § 67). Slavic forms, however, could be somewhat older, although it is not sure that they
should be reconstructed already in Proto-Balto-Slavic - perhaps only as Proto-Balto-Slavic
dialectal variants?
How does one explain the Slavic forms mъnojNj and tobojNj?377 A start should be made
with with pronominal stems *moy- and *twoy-, which get the ending *-oh1. This ending
looks strange, since it must be a misanalysis from the o-stem ending *-o-h1. It is not clear
why exactly this ending would be attached to *moy-/twoy-, but it is the ending one needs in
order to explain Slavic *-Nj. Now, the forms *moy-oh1, *twoy-oh1 are analogically remodeled
to *menoyoh1 and *tewoyoh1 due to the influence of *mene, *tewe. The form *tewoyoh1
regularly yields *tawayō. The usual (Balto-)Slavic singular instrumental ending *-mi is now
added at the end. Thus we get *menayōmi and *tawayōmi. These four-syllable long forms
were too long compared to other pronominal forms. Therefore, they were shortened – the
final *-i disappeared378. The new forms were *menayōn and *tawayōn. The latter received an
analogical *-b- instead of *-w- in Slavic, as in the genitive singular, and is regularly
reflected as tobojNj.
The other possibe explanation is that the ending –ojNj originated in the demonstrative
instrumental tojNj < *toy- oh1-m(i)379, and that it was analogically transferred to personal
pronouns. However, that would perhaps deprive us of the possibility that *-mi was
shortened to *-m, because the forms *menayōmi and *tawayōmi were too long. Besides, it is
not clear why personal pronouns would take the feminine ending from demonstrative
pronouns. If the ending –ojNj is indeed secondary in personal pronouns, it must have been
introduced there before it was introduced into the ā-stems (see § 70). It is also possible that
the ending –ojNj appeared in personal and demonstrative pronouns at the same time.
As for the 1st person form, one mysteriously finds an old *-u- instead of *-e- there.
Thus mъnojNj, and not **menojNj. The obvious solution is to try to find a source for that –ъ-
in the dative form mъnĕ, but the motivation for this influence is quite unclear. A possible
explanation would be that tobojNj had a stem tob-, just like one of the dative forms (tobĕ).
That could have been the reason for mъn-380 to be imported to the instrumental by analogy.
Another solution would be to assume that the new forms of instrumental (which was non-
existant in PIE) were simply made from the dative stems *mun- and *tob-, whatever their
origin. However, this is only a formal solution. It is not clear why the instrumental would be

377 Old Indic máyā, tváyā have no direct relation to the Slavic forms. They are younger forms (cf. Vedic
tvā, *mā is not attested), due to analogy to the ā-stem instrumental singular –áyā, with the usual
secondary –ay-, and to the demonstrative táyā, where the –ay- is old (IE demonstrative stem *toy-; cf.
also the abscence of Brugmann's Law in táyā).
378 Of course, this was not a regular process, but such irregular shortenings in pronominal forms are
hardly unusual. For example, contemporary Lithuanian exhibits such changes (like manimì > manimɶ).
Cf. also dialectal Croatian meʂn < meʂni etc.
379 For the stem *toy-, cf. *toysom > OCS tĕxъ, *toymos > tĕmъ.
This was also perhaps only one of the dative forms, the younger mьnĕ might have already been

built from the dative form (although there were probably only two different stems – the
genitive and the dative one). In the 2nd person, this derivation is somewhat problematic since
the stem *tob- in the dative is not very archaic. In case of a derivation like this, Slavic
instrumental forms would have to be rather young, i.e. they would have to be younger than
the dative in *tob-. If we were to accept this, the ending *-ojNj would be unaccounted for,
since the scheme proposed above (which presupposes an older origin of these forms) would
not be possible.

Accent of Slavic instrumental singular

§ 78 The accent of the instrumental singular forms in Slavic is not easy to explain. As we
have shown (see § 54), Croatian (Štokavian and Čakavian) and Middle Bulgarian of Evangelie
1139 point to the old *mъnòjNj and *tobòjNj, while Slovene and Old Russian point to the old
*mъnojNjʀ, *tobojNjʀ. These could have been Common Slavic dialectal variants. But how are we
to explain them?
If we start from the original *moy-, *twoy-, we can assume that the new
instrumentals *móy-oh1, *twóy-oh1 had the initial accent like all other oblique forms
originally had. The result are the transformed *menàyōn and *tawàyōn with the unchanged
place of the accent. Now, these supposed *menàyōn and *tawàyōn could be the direct source
of the Common Slavic *mъnòjNj and *tobòjNj. In this case, the final accent of Old Russian and
Slovene (a dialectal variant in Common Slavic), could be explained in the following way:
The original forms were *mъnòjNj and *tobòjNj. When this *-ojNj was taken over
analogically to the ā-stems381, new accentual variants appeared in nouns to incorporate
them into the nominal accentual system: in a. p. b *ɪ-ojNj (*-òjNj after Dybo’s Law), and in a.
p. c *-ojNjʀ382. With time, the new a. p. c *-ojNjʀ could have been transferred to *mъnojNjʀ and
*tobojNjʀ in some Proto-Slavic dialects383, since there the desinential accent was also nicely
incorporated in that specific pronominal mobile paradigm. Thus, the forms *mъnòjNj and
*mъnojNjʀ could be Proto-Slavic dialectal isoglosses.
However, if Dybo’s Law operated in the case of medial syllables of a. p. c (in the
interpretation of Rasmussen384), and not only in a. p. b, the forms *mъnojNjʀ, *tobojNjʀ, attested
in Old Russian and Slovene, would be regular reflexes of the old *menàyōn and *tawàyōn.
However, in this case, the variants *mъnòjNj and *tobòjNj would be almost impossible to
interpret. Štokavian and Čakavian have generalized the variant *-ojNjʀ in the instrumental of
the ā-stems in both a. p. c (where it is expected) and in a. p. b (where it is not expected) -
cf. dialectal Croatian –õv in both paradigms. Thus, it is completely unclear in which manner
the innovative *mъnòjNj and *tobòjNj could have developed. The question of the operation of
Dybo’s Law in the medial syllables of a. p. c is much more complicated, of course, and
cannot be dealt with here.

381 Because the original instrumental ending *-ā (< PIE *-eh₂eh₁) was identical to that of the nominative
382 In a. p. b, the accent was in pre-Dybo Slavic on the last syllable of the stem, like in all other cases in
a. p. b, and in the a. p. c, the instrumental became the “strong” case, with the stressed ending. The
new ending *-ojNj was apparently interpreted as having its own accent (thus being +, not – in the
valence theory).
383 Old Russian –oюɴ in personal pronouns, but –óю in nouns perhaps contradicts this possibility (see §
384 Cf. Rasmussen 1999b: 478. Dybo’s Law would operate here in the same way as in *pečeʀte >

Balto-Slavic plural and dual oblique stems

§ 79 In Balto-Slavic oblique cases of plural and dual, all the forms were rebuilt on the
basis of the stems *nō- and *wō-. These stems are derived from the PIE GDA. unaccented
plural forms *nōs and *wōs385. In the dual, the basis were the dual unaccented NA. forms *nō
and *wō < PIE *noh₁, *woh₁386 (cf. § 100, 102).
In the plural, even the accusative was remodeled: *nōs > *nōns and *wōs > *wōns,
cf. OCS ny, vy and Old Prussian mans, wans. In the dual, the original *nō and *wō were
preserved (OCS na, va), but they became accented - at least in the accusative function. These
newly modeled plural cases were often explained as *nōs, *wōs + endings387. This functions
for the explanation of *nōs-o/ōn, *wōs-o/ōn (or *nōs-so/ōn, *wōs-so/ōn, with the
pronominal ending *-so/ōn, like in *toyso/ōn) and *nōs-su, *wōs-su. In the case of the D.
pl. *nōmas, *wōmas and I. pl. *nōmīs, *wōmīs, one could assume a development *-sm- > *-
m- like in Slavic tomu < *to-sm- (cf. Sanskrit tasm-). However, in the case of A. pl. *nōns,
*wōns, no such explanation would work. Thus, it is easier to assume that the basis were the
bare stems *nō-, *wō- from the start. When the *-n- was inserted in the accusative plural,
*nōns and *wōns were analyzed as *nō-ns, *wō-ns, and thus new stems were made. The
acute from the original *nōs and *wōs remained, and thus all these new forms now had the
acute on the first syllable (for the origin of this acute, see § 106).
According to what has already been said here, the stems *nō- and *wō- in the
oblique cases, attested in Slavic, are the oldest. In East Baltic, the oblique stems were
completely reshaped by analogy to the nominative forms. In West Baltic, only the accusative
form of the 2nd person has managed to preserve the original wa-388. In the other oblique
cases, the development was the same as in East Baltic (except that 1st person GD. pl. have n-
instead of m-).
However, one could claim that there were no Balto-Slavic *nō- and *wō- stems in the
oblique plural cases and that that is just a Slavic innovation. Balto-Slavic could have had
different forms in these cases, perhaps inherited directly from PIE. These would then be
changed with the new *nō-, *wō- stem forms in Slavic, with the new *mū-, *yū- in East
Baltic and with the new *nū-, *yū- in West Baltic independently389. Nevertheless, this
possibility seems improbable. The *n- in the 1st person genitive-dative plural in West Baltic
seems to point to the original *nō- here. Otherwise, we would expect the stem *mū- here,
like in East Baltic, with an *m- as in the nominative mes and accusative mans. If there were
no Balto-Slavic oblique *nō- to begin with, Old Prussian n- in the genitive-dative would be
hard to explain. Likewise, the Old Prussian accusative wans, stemming from the PIE *wo/ōs
transformed to a «regular» accusative, points to the earlier presence of this w- in the

385 Cf. Vedic nas, vas < *nos, *wos and Latin nōs and uōs.
386 Cf. unaccented Gatha Avestan nā and Vedic vā-m.
387 Cf. for instance Birnbaum & Schaeken 1997: 74-75.
388 In the 1st person, the m- was introduced from the nominative plural or by dissimilation (cf. § 65).
389 Since West Baltic had the *n-, and East Baltic the *m- here, a common Baltic innovation is not very
likely (cf. § 68).

Since the stems *nō-, *wō- were generalized in all oblique cases, this points to the
conclusion that the Old Prussian A. pl. forms mans, wans were originally *nōns, *wōns and
not *nons, *wons.
The development of the oblique plural stems in Balto-Slavic can be summarized as

1. Pre-
Balto-Slavic unknown stems, probably *ins-, *us- < PIE *nʢʢs-, *us-390

2. Balto-
Balto-Slavic *nō-, *wō- (from the PIE oblique clitics *nōs and *wōs)

3A. Slavic preserved *nō-, *wō-

3B. West Baltic 1 preserved *nō-, *wō-

West Baltic 2 A. pl. *nō-, *wō-; analogical GD. pl. *yū- and *nū-

Old Prussian A. pl. *wō-; analogical GD. pl. *yū- and *nū-, analogical A. pl.

3C. East Baltic 1 preserved *nō-, *wō-

East Baltic 2 analogical *mū-, *yū-391

390 For these forms, see § 102.

391 It is very possible that the change in East Baltic was also carried out in a couple of phases. However,
we have no evidence of that.


§ 80 The clitics have gone completely different ways in East Baltic and in Slavic392. In Old
Lithuanian one finds only the GDA. singular clitics –mi, -ti < *mey, *tey, which have
preserved their original polyfunctionality (and pushed it a little bit further, see below). In
Slavic however, *mey, *tey are preserved as well as mi, ti, but only in the dative function.
In general, Slavic has preserved only the dative function of the clitics, not only in
singular, but also in plural and dual. In dual, the PIE polyfunctional *nō, *wō have become the
accented accusatives, but have also remained unaccented dative forms (thus in OCS). In
singular, the polyfunctionality of *mey and *tey is proven by Old Lithuanian. There are no
such direct evidence for the polyfunctional use of clitics in plural and dual, since these are
dative only in Slavic and non-existant in Baltic. However, considering the polyfunctional
forms in singular and the existance of polyfunctionality in PIE (see § 100), the reconstruction
of Balto-Slavic GDA. plural/dual clitics seems rather certain.
One has to note that the pronouns *mey/moy, *tey/toy were used only for GD. sg. in
PIE (see § 99). The use of these pronouns in A. sg. is either BSl or East Baltic/Lithuanian
innovation. Here we have tentatively reconstructed *mey/tey for BSl as well, although it is not
sure whether they were accusative clitics already then.
In plural, the old PIE polyfunctional unaccented oblique forms *nōs, *wōs have been
remodeled. They were used as a base for all new (accented) oblique cases, but they were also
preserved as the unaccented dative forms (thus in OCS, see § 30, § 49). However, since old
*nōs, *wōs were remodeled to *nōns, *wōns in the accusative393, the dative unaccented forms
*nōs, *wōs were also changed to *nōns, *wōns. Since these are attested as such only in
Slavic, one cannot be sure whether this occurred already in the Balto-Slavic proto-language.
In dual, both the accented accusative *nō, *wō and the unaccented dative *nō, *wō
remained unchanged, since their ending *-ō was the ending of the NA. dual of o-stems as

392 Old Prussian accusative clitics –tin, -si(n) look secondary, i.e. made by cliticization of the accented
forms tien, sien.
393 Cf. OCS ny, vy and Old Prussian mans, wans, which probably proves that this change was already

Balto-Slavic accentual paradigm of personal pronouns

§ 81 As far as can be seen, the accentual pattern of Balto-Slavic paradigms of personal

pronouns was not mobile, i.e. it was always on the first syllable394. Many forms were
monosyllabic (like *ēź, *tū, *mès, *nō etc.), and in some of them the initial accent was
inherited from PIE (like in *mène, *tàwe, *tèbi etc.). The initial acute of the oblique plural and
dual stems was derived, as mentioned, from the original clitic forms that received the acute
in Balto-Slavic - *nōs, *nō etc. All Balto-Slavic personal pronoun forms with the long vowel
in the first syllable had an acute. For the origin of this acute, see § 106.

In Slavic, the singular pattern became mobile due to the desinential accent of the
dative and instrumental. The accent of the instrumental is, as the form itself, not very
conclusive (see § 77, § 78). In the dative-locative, the desinential accent can perhaps be
explained by assuming that the dative-locative ending of ā-stems has been introduced in
Slavic instead of the original *-i. As already said, the dative-locative ending of ā-stems was
unaccented in the dative and accented in the locative of the a. p. c. To explain Slavic
desinential accent in *mъneʖʀ, *tebeʖʀ, one would have to assume original dative *mъʀnĕ and
locative *mъneʖ395
ʀ , which was later generalized to DL. sg. *mъneʖʀ. Although possible, this
explanation seems only formally satisfactory. . See also § 72 for a possible development
*t(e)wóy > *tobeʖʀ etc.

The Balto-Slavic forms *mène and *tàwe were interpreted as unaccented in Slavic396,
when the whole paradigm was interpreted as mobile due to the end-stress in the dative-
locative and instrumental. For the developments concerning Meillet's Law in Slavic see § 56,
§ 57.

In East Baltic, as already said, the accentual pattern was originally immobile, with the
acute in all the forms that had a long vowel stem. The mobile paradigm, seen in present day
Lithuanian and Latvian, is of a later date. The rise of the mobility in personal pronouns
(Lithuanian manimì, manyjè, mumìs etc.) is obviously directly related to the introduction of
the new nominal endings, the making of the new forms (like the new locative singular) and
analogical changes like the shortening of the *-ū- in the dative-instrumental plural etc.

394 Of course, in some of the cases, the accent cannot be reconstructed reliably. One form that had a
desinential accent was *ēźàn (but the pretonic length had the acute as well, due to Winter's
395 Like D. sg. *voʂdĕ – L. sg. *vodeʖʀ, cf. Croatian D. sg. voʂdi, L. sg. vòdi.
396 I.e. enclinomena, words with the absolute initial accent.


§ 82 Here, we shall reconstruct the system of personal pronouns in Proto-Indo-European.

Since the main subject of this work is the reconstruction of Balto-Slavic personal pronouns,
we shall mostly deal with Indo-European problems related to Balto-Slavic in order to derive
Balto-Slavic forms from their Proto-Indo-European origin. Of course, it is necessary to deal
with the problems concerning other Indo-European languages as well. However, many of the
innovations concerning specific languages only cannot be dealt with here. The glottogonic
problems, very popular in the research of the Indo-European personal pronouns, will also be
left out of consideration397. Needless to say, the literature on the problems of the
reconstruction of IE personal pronouns is vast. It would be impossible to list or cite even a
small part of it. Therefore, we shall not make any claim regarding the comprehensiveness of
the literature quoted. For an overview of the literature and references, cf. Schimdt 1978 and
Szemerényi 1989.


Here, we shall list the paradigms of the personal pronouns in various Indo-European
languages (we shall leave out Balto-Slavic of course)398. After that, we shall list the PIE
paradigms of personal pronouns as reconstructed in some of the works dealing with the
problem. Again, the list is representative rather than comprehensive. We shall then briefly
criticize some of the proposed reconstructions and then propose our own reconstruction
together with the discussion. As already said, the emphasis will be on the derivation of
Balto-Slavic data from the Indo-European source.


N. uk, ugga zik
G. ammēl, -miš tuēl
D. ammuk, -mu tuk, -ta, -du
A. ammuk, -mu tuk, -ta, -du
Ab. ammēdaz tuēdaz

397 Thus for instance, we shall reconstruct PIE *egɴHóm, disregarding the possible analyses of this form
(*egɴ-Hóm or *egɴH-óm), the origin of the initial *egɴ- and the final *-(H)om etc. These questions are
undoubtedly very interesting, but they are outside the scope of this work and are immanently very
398 As a rule, we shall list only the oldest and the most archaic forms, e.g. Vedic and not Classic
Sanskrit, Old and not Middle Hittite, Aeolic or Doric instead of Attic Greek (regarding the plural
personal pronoun forms) etc. Younger and obvious innovations in specific languages will not be dealt
with except when of particular interest.

n. wēš šumēš
g. anzēl šumēl, šumenzan, -šmiš
d. anzāš, -naš šumāš, -šmaš
a. anzāš, -naš šumāš, -šmaš
ab. anzēdaz šumēdaz


N. ahám tuvám
G. máma, me táva, te
D. máhya(m), me túbhya(m), te
A. mām, mā tvām, tvā
L. máyi tvé, tváyi
I. máyā tvā, tváyā
Ab. mát tvát

n. vayám yūyám
g. asmākam, nas yušmākam, vas
d. asmé, nas yušmé, vas
a. asmān, nas yušmān, vas
ab. asmát yušmát

N. vām yuvám
G. āváyos yuvós, yuváyos
A. āvām, nāu yuvām, vām
Ab. āvát yuvát


N. azǩm, azǩɷm tū, tuvǩɷm
G. mana, mōi tauuā, tōi
D. maibiiā, mōi taibiiā399, tōi
A. mąm, mā θβąm, θβā
I. θβā
Ab. matʣ θβatʣ


399 The forms maibiiā, taibiiā etc. are written also as maibyā, taibyā and maibyā, taibyā.

n. vaēm yūš, yūžǩɷm, yūžǩm
g. ahmākǩm, nǩɷ, nō yūšmākǩm, vǩɷ, vō
d. ahmaibiiā, nǩɷ, nō yūšmaibiiā, vǩɷ, vō
a. ahma, nå, nō -, vå, vō
i. ǩɷhmā xšmā
ab. ahmatʣ yūšmatʣ

N. vā
G. yauuākǩm
A. ǩɷǩāuuā, nā


N. ǫʆγώ, ǫʆγών τύ, τυɷνη, σύ
G. ǫʆµǫɴο, µǫο σǫɴο, τǫɴο, σǫο
D. ǫʆµoί, µoι σoί, τoί, σoι
A. ǫʆµǫɴ, µǫ σǫɴ, τǫɴ, σǫ

plural (Aeolic)
n. αʆɴµµǫ̋ υʆɴµµǫ̋
g. αʆµµǫɴων υʆµµǫɴων
d. αʆɴµµι(ν) υʆɴµµι(ν)
a. αʆɴµµǫ υʆɴµµǫ

NA. νώ, νωɶǫ σϕώ, σϕωɶ̓
GD. νωɶʛν, νωɶ̓ν σϕωɶʛν, σϕωɶ̓ν


N. egō tū
G. meī, mīs tuī, tīs
D. mihī tibī
A. mēd tēd
Ab. mēd tēd

n. nōs uōs
g. nostrum, nostrī uostrum, uostrī
d. nōbīs uōbīs
a. nōs uōs
ab. nōbīs uōbīs


N. es du
G. im kɈo
D. inj kɈez
A. is kɈez
I. inew kɈew
Ab. inēn kɈēn

n. mekɈ dukɈ
g. mer jer
d. mez jez
a. mez jez
i. mewkɈ jewkɈ
ab. mēnǐ jēnǐ


N. ik þu
G. meina þeina
D. mis þis
A. mik þik

n. weis jus
g. unsara izwara
d. uns, unsis izwis
a. uns, unsis izwis

N. wit
G. igqara
D. ugkis igqis
A. ugkis igqis


NA. mé tú
GDA. -mɈ, -m -thɈ, -t

n. sní sí
gda. –nn -b


N. ñäś, ñiś twe, tuwe
G. ñi tañ
obl. ñäś, ñiś ci

n/obl. wes yes
g. wesi, wesäñ, wesämʖ yesi, yesäñ, yesämʖ
gda. –me -me

N/obl. wene yene


N. unë ti
GDA. mua, më ty, tyɷj, të
Ab. meje(t) teje(t)

n. ne ju
gda. nē, neve, na juve, u
ab. nesh jush

Previous reconstructions

Here, as already said, we shall list some of the previous reconstructions of the Proto-
Indo-European system of personal pronouns. The survey is not comprehensive and the
emphasis is on recent works. We shall briefly comment on each of the proposed
reconstructions. The comments on the previous reconstructions will also find their place in
the main discussion. The list comprises the standard recent Indo-Europeanist general
handbooks - Szemerényi 1989, Beekes 1995, Sihler 1995 and Meier-Brügger 2002, the
classic compendium on the IE personal pronouns - Schmidt 1978, and a very good recent
dissertation on the topic of IE personal pronouns - Katz 1998.

§ 83 Schmidt 1978400

N. *egʄ-H > *egʄ *egʄ-H-o/ōm
G. < GAb. *(e)me-ne/ē
DL. *(e)m-oy
D. < L. *(e)me-gʄ-bhi > *(e)megʄhi
L. *(e)me-gʄe
A. *(e)me, *mē *(e)me-gʄe
Ab. < I. *(e)med (*(e)m-ed)

N. *tu-H *tu-H-om
DG. < L. *tew
D. < L. *tu-bhey
A. *tu *te (enclitic)
L. *tew *tw-i(-n), *tow-oy
Ab. < GAb. *tus
I. *tw-ē/ō

n. *mes *we-y (younger)
g. *mʘser-om (*moser-om)
d. < l. *mʘs-bhey
a. *mōs *mos (enclitic)
l. *mʘser (*mos-er)
l. > dg. *mos (enclitic)
i. –
ab. < gab. *mʘs- e/os

400 Schmidt’s reconstructed paradigms are cited by his Zusammenfassung sections in the book (1978:
110, 144, 205-206, 245-246). I have omitted the possessive forms, changed the order of the cases
and translated German notes to English.

n. *yuH-s
g.(ab.) *wos *wos-om (only g.)
dg. (enclit.) < l. *wos
d. < l. *wos-bhi

a. *us-me
l. *us-m-ey
i. *wos?

N. *nʘH-we *we (elliptic) *we-de (extended)
G. *noH-Hu
D. < L. *noH-bhi (orthotonic) *noH (enclitic)
L. *noH-Hu

N. *yuH
GAb. -
G(L.?) *yuH-eH2qw-Hu
D. < L. *yuH-eH2qw-bhi? *woH
A. *sgwhes *woH
L. *yuH-eH2qw
I. -

Schmidt’s book is unavoidable in any serious study of Indo-European pronouns

because it provides a great survey of the pre-1978 material and theories. However,
Schmidt’s reconstruction of the Proto-Indo-European system of personal pronouns is not at
all satisfactory. As is clearly seen, many of his reconstructions are arbitrary, fanciful and
insufficiently substantiated in the attested material.
He insists on the reconstruction of seven case forms, in spite of the fact that there is
no real evidence for it. PIE personal pronoun declension seems to have had fewer cases than
nominal declension. He also seems too prone to glottogonic solutions, reconstructing, for
instance, enclitic *mos (instead of *nos) with a *m- that is not attestad in any Indo-European
As an example of unreasonable reconstructions, we can mention his 2nd person
singular DGL. *tew, not attested anywhere, his GAb. *tus, attested only in Germanic (Gothic
þus), his 1st person plural L. *mʘser (*mos-er), his only 2nd person plural G. *wos-om (in spite
of the fact that Slavic vasъ is obviously not archaic) etc. The most monstrous reconstruction
is his 2nd person dual form *yuH-eH2qw-, based practically solely on Indo-Iranian (Vedic

§ 84 Szemerényi 1989

N. *eg(h)om, *egō *tū, *tu
G. *mene, *mei/moi *tewe/tewo, *t(w)ei, *t(w)oi

D. *mei/moi, *mebhi *t(w)ei/t(w)oi, *tebhi
A. *(e)me, *mē, *mēm *twe/te, *twē/tē, *twēm/tēm
Ab. *med *twed

n. *wei, *nʘsmés *yūs, *usmés (uswes?)
g. *nosom/nosōm *wosom/wosōm
d. *nʘsmei *usmei
a. *nes/nos, *nēs/nōs, *nʘsme *wes/wos, *wēs/wōs, *usme, *uswes
ab. *nʘsed/nʘsmed (*used?)/usmed

Szemerényi’s approach to personal pronouns is unusual in many ways and it cannot

be accepted. His glottogonic attempts to get rid of the suppletion do not really represent a
sound methodology401.
The accusative *mēm is unnecessary in PIE, since the final *-m is an obvious
innovation in certain languages. The same goes for the ablative forms. The form *mebhi is
mostly a product of Szemerényi’s strong dislike of suppletion.
The supposed nominative forms *nʘsmés and *usmés obviously cannot be PIE. Not to
mention the fact that forms like these are attested only in Greek. They can hardly be archaic
and Szemerényi’s reconstruction of them is guided only by his glottogonic aspirations. The
genitive plural forms are also unattested in the material, except for Slavic, where they are
clearly secondary.
The main flaw of Szemerényi’s approach is that he values theory (which itself is not
very convincing) above the factual attestations. To get rid of the suppletion, he even
reconstructs such blatantly inexistent forms as *nʘsmés, whose reconstruction goes against
any logic and linguistic methodology402.

§ 85 Beekes 1995

N. *h₁egɴ (-oH, -Hom) *tuH
G. *h₁méne, *h₁moi *téwe, *toy
D. *h₁mégɴȹyo, *h₁moy *tébȹyo, *toy
A. *h₁mé, *h₁me *twé
L/I. *h₁móy (h₁moí) *toy
Ab. *h₁méd *twéd

n. *wey *yuH
g. *nʘs(er)o-, *nos *yus(er)o-, *wos
d. *nʘsmey, *nʘs *usmey

401 For a reasonable critique of Szemerényi, cf. Katz 1998: 28-37.

402 It is completely impossible that the majority of Indo-European languages would rather prefer to
generalize the nominative form *wey-, which is completely different from the oblique cases, than the
form *nʘsmés, which would make a regular pattern with the oblique cases.

a. *nʘsmé, *nōs *usmé, *wos
l. *nʘsmí *usmi
ab. *nʘsmed *usmed

Beekes’s approach is not very radical. The characteristic points of his reconstructions
are the unnecessary initial *h₁- in the 1st person singular forms, and the assumption that Old
Indic –ya < *-yo might be the oldest dative ending, since it is otherwise unexplainable. He
disregards the fact that it is not attested outside Indo-Iranian. The reconstruction of the
separate ablative forms ending in *-ed also seems somewhat far-fetched – these could have
easily been formed later due to analogy with the nominal declension. His reconstruction of
the genitive plural forms *nʘs(er)o- and *yus(ero)- (cf. Gothic unsara, izwara) is highly

§ 86 Sihler 1995

N. *egɴoH *ti/ī (tu/ū)
G. *mé-me, *mos (adj.) *té-we, *tos (adj.)
D. *mébȹi, *mey, ? moy *tebȹi, *tey/toy
A.*mé, *me *t-wé, *te
Ab. **mm-ét (> *mét) *tw-ét

n. *we-i *yūs (? yuHs)
g. *nʘsóm, *nos/nōs *usóm
d. *nʘsm-éy, *nos/nōs *usm-éy, *wos/wōs
a.*nʘsmé, *nos/nōs *usmé, *wos/wōs
ab. *nʘsm-ét *usm-ét

N. *weh₁ *yuh₁ (? yūh₁)
A. *nʘh₁-wé, *no/ōh₁ uh₁-wé, *wo/ōh₁

Sihler gives an extensive survey of the reconstruction of personal pronouns in his

comparative Greek and Latin grammar. His approach is also characterized by a few unusual
For instance, he reconstructs PIE *mébȹi on the grounds of Avestan maibiiā,
disregarding Old Indic máhyam and Latin mihī. However, the whole idea of *m-bh being
dissimilated into *m-gɴh is not very convincing. It is not at all clear why *bh would yield a
palatalized (sic!) aspirated velar by dissimilation. Thus, it is much more reasonable to
assume that Avestan maibiiā is analogous to taibiiā (see § 101). Also, the reconstruction of
*méme in the genitive (Old Indic máma) seems unnecessary and based on glottogonic
considerations. On these problems, Sihler has the same opinion as Szemerény.

Sihler also reconstructs PIE nominative *ti/ī based on Hittite zik. That does not seem
very reasonable, in the light of the possibility that Hittite -i is actually a regular reflex of PIE
*-ū here (see § 90).
Reconstruction of the ablative forms is highly unlikely, as already said, since it is
quite clear that PIE system of personal pronouns did not have all the cases as in the nominal
declension403. In addition, since the ablative has a separate ending only in the o-stems, one
would hardly expect to find it in the declension of personal pronouns.
The reconstruction of the genitive plural forms *nʘsóm, *usóm is quite ad hoc.
Structurally, it works just fine, the usual plural oblique *nʘs-/us- + *-(s)om, but there are no
real attestations of these forms. Either PIE did not have the accented genitive plural forms, or
the original forms were lost everywhere (or transformed beyond recognition). It seems safer
not to reconstruct genitive plural forms, than to resort to reconstructing completely
speculative forms, however may they fit into a proposed system.

§ 87 Katz 1998404

N. *egʄ(H)-(o/ō(H?)m)405 *tū/tú
G. *méne, *moy *téuʢe, *toy
A. *(m-)mé *t(uʢ)é

n. *uʢéy(e)s, *més *iʢūs
a. *nʘsmé, *no/ōs *usuʢé, *uʢo/ōs

N. *uʢē *iʢū
A. **nʘh3mé, *noh3 *uh3uʢé, *uʢoh3

Katz reconstructs the suffix in the 2nd person plural accented A. as *-wé, rather than
as *-mé as is usually reconstructed406. The usual reconstruction *usmé is based on the clear
correspondence of Greek Aeolic αʆɴµµǫ, υʆɴµµǫ407 and Old Indic asmá-, yušmá-. However, Katz

403 For this, cf. for instance Petersen 1930: 164.

404 Katz does not really deal with the reconstruction of the whole system in this work. He just discusses
the accented plural (and dual) accusative forms in various IE language families, while reconstructing in
passing some of the other forms. Since his view of these forms is different from the classic approach, I
believe it is justifiable to list his reconstructions here as well. Since Katz does not adduce the complete
PIE paradigms, some of the adduced forms perhaps need to be taken conditionally.
405 This is, of course, rather a formula, than an actual reconstruction. Katz does not present his ideas
on the reconstruction of the dative singular forms.
406 It is interesting to note that Szemerényi, whose views Katz vehemently disapproves, also
reconstructs *usw- in the 2nd person plural. However, while Katz proposes that reconstruction on the
basis of a thorough research on the material of the “non-classic” IE languages, Szemerényi reconstructs
his *usw- out of glottogonic reasons mostly (his only real trace of *usw- being Gothic izw-).
407 Cf. Doric ʅαɷµµǫɴ and ʅυɷµµǫɴ for the older accent.

points to various traces of older *-wé in other IE branches – Celtic, Germanic, Anatolian and
Armenian (he also deals extensively with Tocharian forms). The evidence from each branch
itself is perhaps not too strong, but put together, it makes quite a strong case.
In Celtic, Katz derives Old Irish sní, sí from Indo-European *nʘsmé, *uswé with the
aphaeresis of the initial *nʘ- and *u- and a (monosyllabic or emphatic) lengthening of the
forms408. This explanation is not very problematic409 and is by far superior to other
explanations of these forms410.
In Germanic, Katz derives Gothic izw- in izwis from PIE *uswé. This form yields
strong support for the reconstruction of *-wé. The –z- in Germanic is due to Verner’s Law
and the initial i- is due to the dissimilation *u- -w- > *i- -w-411.
The evidence in Anatolian412 is less conclusive. Katz’s derivation of Hittite anzāš,
Cuneiform Luwian 1st person ānza and 2nd person ū(n)za from PIE *nʘswé, *uswé413 is
possible, but not certain. On the other hand, the derivation of Hieroglyphic Luvian á-zu-za
/antsunts/, u-zu-za /untsunts/ from the same forms seems far more likely. The –u- in the
second syllable of these forms is supposed to be a reflex of PIE *-w-. However, I cannot
accept Katz’s derivation of Hittite šumāš from aphaeretized *swé > *suwé. The old
assumption that šum- is to be derived by methatesis of old *usw- (or *usm-)414 seems much
more convincing.
In the case of Armenian, the evidence for *-mé and *-wé is indirect415, but in the
light of other evidence, Katz’s analysis seems satisfactory.
We can conlude that it seems probable that the suffix *-wé in the accented 2nd
person accusative plural form in Indo-European indeed existed. This suffix was perhaps only
one of the possible variants (see § 100, § 102 for the reconstruction).
In dual, Katz reconstructs oblique forms with the *-h3-. Even though this would
enable us to derive Homeric νωɶǫ directly from *nʘh3wé, I cannot accept Katz’s416 claim that
this *-h3- does not have to be related to the nominal dual ending *-h1. Therefore, I think *-
h1- should be reconstructed here as well, despite the problems (see § 102 for further
discussion on this).

408 Cf. Katz 1998: 77-106. In Celtic, *ē > *ī regularly, *-w- drops regularly in Old Irish (cf. Old Irish sí
and Middle Welsh chwi). The whole process is: *nʘs-mé ⇒ *nʘs-né (assimilation) ⇒ *sné (aphaeresis) ⇒
*snē (lengthening) > *snī > Old Irish sní, Gaulish SNÍ, Middle Welsh ni. Similar in: *us-wé ⇒ *swé
(aphaeresis) ⇒ *swē (lengthening) > *swī >Old Irish sí, Middle Welsh chwi.
409 The notion of aphaeresis is not strange, as this is typologically very frequent in pronouns. Cf.
Croatian dialectal nàkō “like that”, vàkō “like this (1. sg.)” from onàkō, ovàkō (though analogy with tàkō
“like this (2. sg.)” is also possible).
410 Usually deriving the forms from the ad hoc forms *s-nēs and *s-wēs (in relation to the usual
*no/ōs, *wo/ōs). Cf. for instance Schmidt 1978: 218.
411 Katz’s (1998: 111) solution to the problem of the initial i- is unacceptable. For the final –is, the 1st
person form etc., cf. Katz 1998: 107-131. Katz’s explanation of the ending –is is not completely
convincing, but *usw- > izw- seems quite certain.
412 Katz 1998: 134-151.
413 The ending *-wé was apparently generalized in Anatolian, unlike Indo-Iranian and Greek.
414 In Hittite, one cannot tell whether it was *-w- or *-m-, since both yield –m- after –u-. If it was *-m,
this would mean that *-wé was not generalized in Anatolian, but that seems less likely.
415 Cf. Katz 1998: 186-193.
416 1998: 209f.

§ 88 Meier-
Meier-Brügger 2002

singular plural
N. *egɴóh₂, *egɴh₂óm *túh₂ *wéys *yúHs
G. *méne *téwe
D. *me-gɴȹey, *me-gɴȹy-om *te-bȹey, *te-bȹy-om *nʘs-mé-i *us-mé-i
G/D. unacc. *moy *toy
A. *mē, *me * tē, *t(w)e *nʘs-mé, *nos *us-mé, *wos

Meier-Brügger’s reconstruction is rather conservative. In the dative singular, he

reconstructs the ending *-ey, based on such forms as Old Prussian tebbei or Latin tibī. This
seems reasonable, but one must be aware that a common dative ending *-ey can easily be
secondary. The other ending *-y-om that he reconstructs is based on the authority of
Sanskrit, the “favorite” Indo-Europeanists’ language of all times. Still, since it is widely
known that Sankrit is fond of the *-om particle in personal pronouns, it is unlikely that this
*-om should be attributed to PIE here.
A good side of Meier-Brüggers conservativeness is that he does not venture into the
reconstruction of cases like the ablative and instrumental singular or genitive plural, where
there is no material available for a plausible reconstruction of these PIE forms.

Reconstruction of personal pronouns in Indo-

As pointed out many times before, this is mostly a research on Balto-Slavic personal
pronouns and their Indo-European origins. Accordingly, we shall pay most of our attention
to the problems of Indo-European reconstruction concerning Balto-Slavic. In fact, the main
concern will be the problem of monosyllabic lengthening in Indo-European, which is of great
interest for the questions of Balto-Slavic accentuation. Another important question is the
Balto-Slavic form *mès and its Indo-European origin. The question of possible Balto-Slavic
dative singular form *tubi and its connection to Old Indic túbhya(m) is also important, but
inevitably inconclusive and not really solvable. The rest is practically uncontroversial, though
one should mention the unaccented forms *moy/mey, *toy/tey. We shall avoid detailed
discussions on the problems about which one is unlikely to say anything original.

Personal pronoun forms with possible monosyllabic leng thening

§ 89 It is widely known that many PIE personal pronoun forms show reflexes with both a
short and with a long vowel, cf. for instance Greek σύ, but Latin tū. These reflexes have been
explained in two basic ways: either by a laryngeal – thus for instance *tuH (or *tuh2), or by
monosyllabic lengthening – thus, *tú/tū417. The disadvantage of reconstructing a laryngeal
here is that we cannot convincingly explain the short reflexes418. Thus, we shall assume that
whenever both the short and the long reflexes exist, we must be dealing with the
monosyllabic lengthening419. If we assume that the long and short reflexes are due to this
process, it is easy to understand the origin of the double forms. The original short vowel was
lengthened (already in PIE or later, in separate languages) and, thus, two forms were made.
Phonetic lengthening in monosyllabic word is an extremely common typological
phenomenon, so there is no reason whatsoever not to assume that it could have been
present in Proto-Indo-European as well.
We shall first take a look at all forms with a possible case of monosyllabic

2nd person nominative singular (accented)

§ 90
*tu – Greek Doric τύ, Latvian tu, Lithuanian Žemaitian tòʖ, Old English þu (German du), Old
Irish tu-ssu420, tú421, Old Indic t(u)vám422, tú (particle), Gatha-Avestan tuvǩɷm, Hittite A.
tuk423, Old Prussian tu, Tocharian B t(u)we (?)

*tū – Latin tū, Greek Homeric τυɷνη, Avestan tū, OCS ty, Old Prussian tū, toū, Old English þū
(English thou)424, Albanian ti, Old Indic tū (particle), Hittite zik (?)425, Breton te426

417 The idea of the facultative lengthening in monosyllables stems already from Hirt (1921-37/II: 227).
418 Ranko Matasović (p. c.) has suggested the following scenario to me. One starts with the original
short *tu and a variant *tuh2om (like *egɴ and *egɴh2om). This would make Old Indic tuvám archaic (cf.
also Oskic tiium). From *tuh2om, the form *tuh2 might have been made by wrong segmentation. I find
this explanation formally satisfactory, but in the light of, for instance, Greek τύ, τυɷɴνη, this hypothesis
just does not seem convincing enough. Besides, monosyllabic lengthening occurs in other examples as
419 Cf. Rasmussen 1999a: “Given the possibility of emphatic forms, I see no reason to read laryngeals
into the forms *uʢē ‘we two’, *iʢū ‘you two’ or even *iʢūs ‘you’, secondary lengthening of old short vowels
will do”.
420 Cf. Pedersen & Lewis 1937: 215. Schmidt 1978: 116 believes that Old Irish tu-ssu is analogous to
me-sse (and that Breton te is analogous to me). That does not seem too convincing.
421 Probably with the Old Irish monosyllabic lengthening, like in *me > mé.

422 The pre-form *tū-om or *tuH-om would have probably yielded Old Indic **tūvám, and the same
goes for Avestan. This is disputable - examples like Old Indic ba-bhūvān from the root bhū- (PIE
*bhuH-) seem to confirm that, but cf. also Old Indic bhúvat.
423 Melchert 1994: 132, 188.
424 Old English situation could be secondary (the same goes for the rest of the older Germanic
languages). There, the long variant was accented and the short one unaccented. Thus, this variation

In some languages, forms can be derived from both variants - for instance Armenian
du427. Old English seems to have preserved both PIE variants (if it is not secondary), just like
Balto-Slavic (most probably also Old Prussian), Indo-Iranian and Greek.
Having considered the evidence, we have to reconstruct PIE *tú, *tū, with the second
form being the result of the monosyllabic lengthening. The reconstruction **tuH is
impossible, since it cannot account for the numerous short reflexes.

2nd person nominative plural (accented)

§ 91
ūs – Avestan yūš, Old Indic yūyám428, Lithuanian jūs, Latvian jũs, Old Prussian ioūs, OCS vy

Armenian duk ʅ is ambiguous, as is Gothic jus429. All forms point to the long vowel430.
Thus, we are left with a choice to either reconstruct the laryngealistic proto-form *yuHs or
the secondary lengthened *yūs, in spite of the fact that the short variant seems to be
unattested. One could suppose that the long variant was generalized already in Proto-Indo-
European, or that the short **yús is accidentally not attested. On the grounds of other forms
(especially the 2nd person dual form, see § 92), I shall tentatively reconstruct *yūs, to be
derived from the older, unattested **yús by monosyllabic lengthening431.

1st person nominative dual (accented)

§ 92
*we – Gothic wit

*wē – Avestan vā, OCS vĕ


Vedic vām and Žemaitian veɶdoʖ are inconclusive. In spite of only three clear
correspondences, PIE *wé, *wē can probably be reconstructed432.

2nd person nominative

nominative dual (accented)

could be the outcome of later lengthenings or shortenings (cf. also secondary Old English īc "I"). It
seems that the length in PIE was not dependent on the accent (see § 96).
425 Melchert 1994: 84, 132.
426 In Brythonic, *ū > *ī > *i regularly, and in Breton *i > e regularly.
427 PIE *u and *ū > Armenian u.
428 With the –y- by analogy to vayám.
429 In North and West Germanic, the form *jiz has a vowel analogous to the 1st person plural form *wiz,
and, therefore, tells us nothing about the original vowel.
430 Albanian ju could be derived from PIE *yus with a short vowel. However, the initial *y- should rather
yield gj- or z- in Albanian, so this derivation is not very reliable.
431 However, it is not impossible that Indo-European had, for instance, *tú/*tū but *yúHs.
432 The final –t in Germanic is usually attributed to PIE *dwoh1 “two”, but the details are not clear.

*yu – Old Indic yuvám433

*yū – OCS vy

Again, Lithuanian jùdu is inconclusive434 (see § 61), but we can reconstruct PIE *yú,
*yū. As in the 1st person nominative dual, there seems to be no need to reconstruct the
laryngealistic *weh1, *yuh1, despite the insufficient evidence in the case of the 2nd person

1st person accusative singular (accented/unaccented)

§ 93
*me – Greek ǫʆµǫɴ, µǫ, Old Irish mé435, Gothic mi-k

*mē – Old Indic mā, Avestan mā, Latin mēd (?)436, Welsh mi (?)437, Gaulish mi (< *mē)

Balto-Slavic also seems to point to *mē indirectly (see § 70). The reconstruction of
PIE *mé, *mē, *me, *mē seems quite certain. The long and short forms were present in both
accented and unaccented form.

2nd person accusative singular

singular (accented/unaccented)

*t(w)e – Greek σǫɴ, τǫɴ (Doric), σǫ

*t(w)ē - Old Indic tvā, Avestan θβā, Latin tēd (?)


The same that goes for the 1st person form, goes for the 2nd person as well. We can
reconstruct PIE *twé, *twē, *te, *tē. For the reconstruction of *tw-/t-, see § 101.

1st person genitive-

dative-accusative plural (unaccented)

§ 94
*nos – Old Indic nas, Hittite –naš, Young Avestan nō, Gatha-Avestan nǩɷ, Albanian na, Greek
νόσ-ϕι(ν) “apart (from)”

*nōs – Latin nōs, OCS na-, Gatha-Avestan nå (?)438


433 See the note for t(u)vám though.

434 In Gothic, this form is not attested, and Old English git has taken the vocalism of the 1st person
form *wit.
435 The length in mé is a later Old Irish development. PIE *mē would have yielded Old Irish **mí.
436 The reliability of the Latin form is questionable, since the origin of the final –d is not clear. The
most common explanations are the adding of a particle or confusion with the ablative form.
437 Welsh mi is unreliable since it could have the –i by analogy to ti < *tū. Cf. Pedersen & Lewis 1937:

The evidence is quite strong and we can posit PIE *nos, *nōs without any doubt439.

2nd person genitive-

dative-accusative plural (unaccented)

*wos – Old Indic vas, Young Avestan vō, Gatha-Avestan vǩɷ

*wōs – Latin uōs, OCS va-, Gatha-Avestan vå (?)


Like in the 1st person form, PIE *wos, *wōs can also be reconstructed without any

1st person genitive-

dative-accusative dual (unaccented)

§ 95
*noh₁ – Greek νώ, Old Indic nau, Avestan nā, OCS na, Old Irish ná-(tha)r440

All the reflexes point to the originally long vowel. Because of that, but also because
of the presence of a laryngeal in the accusative dual accented forms (cf. Old Indic āv- <
*nʢh1-), it seems that, unlike elsewhere, laryngealistic *noh1 must be reconstructed.

2nd person genitive-

dative-accusative dual (unaccented)

*woh₁ – OCS va, Old Indic vām441


Despite only two attestations, it is clear that the form *woh1 is to be reconstructed
(cf. the well-attested length in *noh1).
Structural reasons also validate our claim that *noh1, *woh1 should be reconstructed.
Compare the parallel of unaccented GDA. dual *noh1, *woh1 and the accented accusative
dual *nʢh1wé, *uh1wé with the relation of the unaccented GDA. plural *nos, *wos and the
accented accusative plural *nʢsmé, *usm/wé.
It should be noted that the lengthened variants *nōh1, *wōh1 can also be
reconstructed, but are unprovable since they would yield the same result as the short *noh1,

438 If not analogous to the accusative plural. In Young Avestan, GDA. pl. are nō, in Gatha-Avestan GD.
pl. are nǩɷ and A. pl. is nå. That is not in accord with Old Indic GDA. pl. nas.
439 Old Indic nas could theoretically be derived from *nes as well, but Latin nōs and OCS na- point to
the o-vocalism.
440 Probably from *noh1-tero-, analyzable as the old unaccented dual form plus the contrastive suffix
441Old Indic vām is probably due to the dissimilation from the older *vau. In this way, one can explain
nau and vām in the same way. If one were to assume that vām is made by the usual adding of the
particle –am in Old Indic, it would be difficult to explain the reason why these two parallel forms get
such different treatment. The form vām is a Rgvedic hapax, the normal nominative form āvám being
due to analogy with the accusative form (āvám : āvām like tvám : tvām, cf. Katz 1998: 202f).

Monosyllabic lengthening

§ 96 We have already seen that the presumption of monosyllabic lengthening is very

useful in the reconstruction of personal pronouns. Here, we shall try to handle this
phenomenon in more detail.

The basic characteristics of monosyllabic lengthening are:

1) Monosyllabic lengthening is not regular in the neo-grammarian sense. Thus, it cannot be

described as a regular sound law.

2) Monosyllabic lengthening operates in clitics as well (cf. clitic PIE *nōs, *mē).

3) The forms made by monosyllabic lengthening were probably just phonetic variants (thus,
*tū was just one possible phonetic realization of the phonological *tu).

The first point is obvious from the fact that there are numerous short monosyllables
in PIE, cf. for instance *égɴ “I” or *nókwts “night”. This reduces monosyllabic lengthening to a
tendency, i.e. a facultative change – PIE monosyllables could have been phonetically
lengthened, but it was only a possibility.
The second point seems to invalidate the claim that monosyllabic lengthening was
originally an emphatic lengthening. It is hard to envisage emphatic use of clitics. Some of PIE
lengthened clitics, like *mē, could be interpreted as analogous to accented forms like *mē,
but in cases like *nōs for instance, there is no source of analogy. One could suppose that the
length of *nōs and *wōs is post-PIE, but that does not seem likely442. As we have already
mentioned, monosyllabic lengthening is typologically very well attested. It occurs not only in
PIE, but also in the separate histories of some IE language families and languages – for
instance in Germanic, Latin, Old Irish etc. The typological confirmation for long pronominal
clitics is found in Croatian, for instance: cf. Croatian long clitic forms jē “of her” or jōj “to
The third point is closely related to the first point. The results of this sporadic
monosyllabic lengthening were not phonemic. They were probably only phonetic variants.
Thus, long *ū and *ī in the monosyllables were just phonetic realizations of inherently short
phonemic *u and *i. Later, after *uH > *ū and *iH > *ī, these allophones became separate
phonemes. Some languages generalized short (Latvian tu), some long variants (OCS ty), and
sometimes both variants were preserved (possibly Old English þū, þu). It is quite possible
that in some cases, the short or the long variant was not preserved by mere chance (like in

442 The argument of Sihler (1995: 38) that *tu was originally unaccented and *tū accented is not
possible, since *nōs was always unaccented and still one gets both the short and the long variant.
443 In Croatian, all unstressed variants of the long accented personal pronoun forms are long as well
(when posttonic, of course). Standard language grammars do not mark this, but it is certain that this is
a feature of at least some Neo-Štokavian dialects. Cf. for instance zóvnijo bi nās "he would call us" in
Imotski/Bekija (Šimundić 1971: 120).

*yūs, where only the long variant is attested). Of course, it is possible that in some cases
there was never any lengthening.

Here, we shall see what one gains with the assumption of monosyllabic lengthening
and which are this theory’s faults.
The advantages of monosyllabic lengthening are:

1) A simple answer to the problem of the double pronominal forms like *tú, *tū, which
cannot be solved by laryngeal theory.

2) The assumption of monosyllabic lengthening explains the reconstruction of many IE

particles and adverbs (accented and unaccented):

PIE *nu “now” > Old Indic nú, Greek νυ, Latin nu-nc, Old Irish nu, no, Latvian nu, OCS nъ
“but” (?)
*nū444 > Old Indic nū, Avestan nū, Greek νυɶν, OCS ny-nĕ, Lithuanian nūnaĩ

PIE *ud “on high” > Old Indic úd

*ūd > Gothic ut [ūt], the length is seen in English out, German aus445

PIE *ne “not” > Old Indic ná, OCS ne, Latin ne-que, Gothic ni-h
*nē > Latin nē, Old Irish ní, Gothic nē, OCS nĕ-

PIE *gɴhi (particle) > Old Indic hí, Greek –χι

*gɴhī > OCS –zi

PIE *u, *ū (particle) > Old Indic u, ū

PIE *gwhe/o (particle) > Old Indic gha, ha, OCS že, -go446, Greek -ϑǫ (?)
? *gwhō > Croatian ga

PIE *ni “downward” > Old Indic ní

*nī > OCS ni-zъ

The possible problems concerning monosyllabic lengthening:

1) The reconstruction of long *ī and *ū in PIE, which were otherwise not present in PIE. This
is not problematic, since these were probably mere allophones in PIE (see above).

444 The secondarity of this length is also seen from the fact that this *nu “now” is probably just a zero
grade of the stem *new-o- “new”.
445 OCS vy- is due to Winter’s Law.
446 In pronominal G. sg. like togo "of this".

2) The occurrence of monosyllabic lengthening in clitics – one could be skeptical about a
process that involves lengthening of clitics. This seems counterintuitive at first glance, but
long clitics are attested typologically (see above), so this does not pose any real problem.

3) The irregularity of monosyllabic lengthening. This is also not a real problem, since it could
be the result of the generalization of individual variants, and since the regularity is not really
expected in a process such as this one to begin with.

We still have to take a look at the length appearing in root nouns. Could this also be
a case of monosyllabic lengthening? Could this be a possible explanation of the otherwise
strange lengthening in root nouns like *pōds “foot”447? Many of these nouns, like the already
mentioned *nókwts "night" do not have a long vowel in the nominative. It is likely that the
lengthening in the words like *pōds is not completely the same phenomenon as the
lengthening in *tú, *tū448. The nonexistence of dual (long and short) forms in root nouns
points to that conclusion (root nouns are either long or short in N. sg.). The details might be
impossible to reconstruct, but monosyllabic lengthening could have occurred in many
phases in PIE. It seems that the long vowels in PIE root nouns are treated the same way in
Balto-Slavic as all other long vowels in PIE. However, the length made by monosyllabic
lengthening in pronouns appears to have a different outcome in Balto-Slavic concerning the
accentual development (see § 106). This would point to the conclusion that the stable length
in root nouns is ultimately not of the same origin as the unstable length in personal
pronouns (and monosyllabic particles and adverbs). For our purposes, we should probably
leave the root nouns out of the consideration for monosyllabic lengthening. Thus, the
monosyllabic lengthening, which we are dealing with here, was probably constricted to
pronouns and particles/adverbs. The reason why monosyllabic nouns were left outside of
this sporadic process lies probably in the fact that the length in the nominal forms was
strictly governed by the ablaut processes. Thus, if any phonetic length appeared there, it
could have been quickly eliminated by analogy, back-formation etc. Also, the occasional
length that could have appeared in a root noun like *pōds could have quickly been stabilized
(thus phonemized) by the analogy to other lengthened grades, like the long vowels in sonant
stems, which were not the result of monosyllabic lengthening (like *kɴwōn “dog” etc.) 449. This
would explain the different accentological treatment of the length in pronouns and the
length in root nouns in Balto-Slavic (see § 106).

In the end, putting aside the controversial lengthening in nominal forms, the
secondary lengthening in monosyllabic personal pronouns and particles/adverbs seems
obvious. Thus, we must conclude that our reconstruction of PIE personal pronouns is correct
and that one has to take the monosyllabic lengthening as a real PIE phenomenon.

447 Not every length in the nominative of a root-noun has to be due to monosyllabic lengthening. For
instance, the length in the form *kēr “heart”, might be due to the dropping of the final *-d, if this
occurred already in PIE (cf. Greek κηɶρ, Hittite kīr).
448 If one insists on explaining the length in the root nouns via some kind of monosyllabic lengthening,
one could assume the existence of the first and second PIE monosyllabic lengthening for instance.
449 In the sonant stems, the length is due to the dropping of the nominative ending *-s: **kɴwons >

Indo-European *més?

§ 97 The Balto-Slavic 1st person nominative plural form *mes and the according Armenian
form mek ʅ have always been troublesome. The reason for that lies in the fact that these
forms are completely different from the forms found in other IE languages – Old Indic vayám,
Gothic weis, Hittite wēš, which are all reconstructible as the older *wey-((e)s).
There are two basic ways how to deal with this problem. The first is to assume that
these *m-forms are archaic. This is, of course, just what the glottogonists want, since this
*mes is in perfect accord with the singular 1st person oblique stem *me-, with the oblique
plural stem *nʢs- < **mʢs-450, and with the verbal 1st person singular ending *-m and 1st
person plural ending *-me. The problem with the reconstruction of this *mes in PIE is that it
is attested only in two languages451, and that is not enough for the reconstruction of PIE if
one of the languages in question is not Sanscrit.
The other explanation is that these forms are somehow innovative. The basic
approach is to try to reduce these forms to some sort of the old *wey-((e)s)452. The initial
*m- is attributed either to the influence of the *m- in the singular, or to the influence of the
*-m- in the 1st person plural verbal endings. The analogy of the nominative plural with
oblique singular stem does not seem very likely453, but the idea that a verbal ending could
influence pronouns is not impossible.
Cf. for instance German dialectal (Bavarian-Austrian) 2nd person plural forms like dir,
des454, instead of Standard German ihr, with the d- due to the misanalysis of the verbal
ending –d and the consonantless initial of the pronoun. The Swedish 2nd person plural
pronoun ni and (Old) Norwegian 1st and 2nd person dual and plural pronouns in m- and þ-
can also be explained in a similar fashion455. The famous Bavarian form mir (instead of wir),
very frequently mentioned in this context, is due to the assimilation of the verbal ending –n
and the initial pronominal w- in the inversion during fast speech456. It must be noted that all
these forms are due to sandhi misanalysis in inversion and that they do not really
correspond to the Balto-Slavic-Armenian situation. This kind of development is not possible
in Balto-Slavic, since the verbal ending was certainly not just *-m, and the inversion of the
verb and personal pronouns was hardly so frequent there (if it existed at all).
Another example of the mutual influence of personal pronouns and verbal endings is
the Polish 1st person plural verb ending –my, made by analogy to the pronoun my. A further
parallel is the influence of the 1st person dual verb ending –va on the dual pronoun vĕ, due

Cf. the probable *-ms > *-ns in the accusative plural of nominal declensions (*-m being the
singular accusative ending and *-s the general plural ending, of course).
451 The idea that the old *mes is seen in Greek forms like αʆɴµ-µǫ̋ is not very convincing.
452 The idea that this *mes is a pluralization of the singular stem *m(e)- (Szemerényi 1989: 229) looks
highly unlikely.
453 In a case of this type of suppletion, one would expect the plural oblique stem initial consonant *n-
to be taken into the nominative plural analogically, rather than the singular oblique stem initial
consonant (sic!). Thus, it is hard to accept such an analogy.
454 Howe 1996: 280-281.
455 Howe 1996: 300, 316-318.
456 Howe 1996: 280-281.

to which this pronoun was changed to va in Old Czech and Old Polish (see § 40, § 42). The
second case seems typologically closest to the supposed influence of *-m- 1st person plural
verbal forms on *wey-((e)s). However, it must be noted that in the case of Balto-Slavic, the
initial consonant was also changed, unlike the second example in which the final vowel was
changed. In the first example, the pronoun has actually influenced the verbal ending, and
not vice versa.
It may well be accepted that the initial *m- of Balto-Slavic and Armenian could be
due to some sort of analogy like the ones mentioned above, despite the fact that the real
typological parallel is missing. However, it is not the initial *m- that is problematic. While it
is possible to derive Armenian -ek ʅ from *-eyes457, there is no obvious way of deriving
Balto-Slavic *-es from the older *-eys, *-eyes or even *-ey. The easiest solution would be to
assume that the original *wey (nowhere attested as such, by the way) first changed to *mey,
and then to *mes458. However, this also seems very questionable459.
Because of the uncertain derivation of Balto-Slavic *mes from the PIE *wey-((e)s), I
shall tentatively conclude that the form *mes might indeed be old, together with the
elsewhere attested *wey- based forms460.

457 Matzinger 1997: 88f.

458 The *-s being from the verbal ending or by analogy to *yūs.
459 A possible suggestion that *wey((e)s) was transformed to *mes with the final *-es also from the
verbal ending *-mes seems very unconvincing.
460 The easiest way to explain the two 1st person plural personal pronouns would be to suppose that
one was inclusive and the other exclusive. Since *mes has a common stem with the 1st person oblique
*m-, this could be the exclusive form, while *wey- could be the inclusive form since it apparently
shares a common stem with the *w- from the 2nd person plural pronoun (cf. for instance Prokosch
1939: 282 or Watkins 1969: 47).

Accented 2nd person dative singular

§ 98 Old Indic dative form túbhya(m) has always been controversial. Usually, it was
regarded as secondary to the older **tábhya(m), attested in Avestan taibiiā, Old Prussian
tebbei, OCS tebĕ and Umbrian tefe. However, the origin of this analogy was very difficult to
explain. Why would t(u)vám influence only the dative form? An old rule of historical
linguistics says – if you cannot interpret it as an innovation, it must be archaic. Thus, there
are some who reconstruct *tubh- in the dative singular461. The parallel to Old Indic túbh- is
found in Balto-Slavic, where the supposed *tub- is found in the influence it had made on the
Latvian dialectal dative stem mun- and maybe in the Old Prussian subs “self”.
Can we indeed reconstruct Proto-Indo-European *tubh- in the dative singular based
on such meager evidence? The rest of the forms, which point to *tebh- are easily explained
as analogous to *téwe (and *méne, *mégɴhi). What would we get by supposing that this is an
archaism? We would explain the Old Indic form, which is otherwise solved with an ad hoc
analogy with the nominative. The otherwise strange occurrence of the u-vocalism in Latvian
dialects, also in the dative singular, would be explained as well. It seems too much of a
coincidence that the unusual u-vocalism forms appeared independently in Balto-Slavic and
in Indic. Therefore, we shall tentatively reconstruct Proto-Indo-European dative form *túbhi.
The more regular variant *tébȹi might have existed already in PIE.

461 For instance Hujer 1912, Schmidt 1978 (cf. :127-129 for the discussion), Kortlandt forthcoming c

Unaccented 1st and 2nd person genitive-
genitive-dative singular

§ 99 We have reconstructed Balto-Slavic 1st and 2nd person genitive-dative (and

accusative?) singular unaccented pronouns *mey and *tey. This is the easiest reconstruction
that yields OCS mi, ti. It was long ago noted that OCS mi and ti are not in accord with Greek
µοι and τοι, which clearly point to the o-vocalism. However, it seems more reasonable to
assume PIE variants *mey/moy, *tey/toy462, than to claim that only *moy and *toy are
original (on the authority of Greek alone), and that OCS mi, ti have to be explained via some
special sound law463.
Old Indic me and te do not tell us whether their pre-form was *mey/tey or *moy/toy.
However, the support for the reconstructions *mey, *tey could perhaps be found in Old
Latin, where the form mīs is most easily explained from the old *mey with the addition of the
usual genitive ending *-s464. Perhaps one could derive Old Latin mīs from *moy-s as well465,
but that is questionable. The same goes for Latin mī (< Old Latin mei < PIE *mey) in the
vocatives like pater mei!, mī fīlī!. One objection might be that it is not certain that Old Latin
mīs was a clitic, since clitics were usually written together with orthotonic words, for
instance –que, -met, -pte etc. But the derivation of mīs < *mey-s does not need Old Latin
mīs to be a clitic, since the development of clitics to accented forms is very well attested in
all Indo-European languages.
One might suppose that Old Indic locative forms máyi and tváyi could point to the old
stems *mey- and *twey-, since *moy- and *twoy- would here yield a long –ā- due to
Brugmann’s Law. However, these forms do not seem to be relevant, since they are probably
far too recent to be reliable (cf. Old Indic older locative tvé)466.
The support of Latin mī (and mīs) to the Balto-Slavic data allows us to reconstruct PIE
*mey/moy, *tey/toy.

462 For the possibility of PIE *mu instead of *mey/moy, see § 73.
463 For instance, that *-oy yields –i in clitic forms in Slavic. One could also compare it to the nominal –i
< PIE *-oy etc.
464 Cf. for instance Sihler 1995: 377.
465 Cf. the PIE o-stem instrumental plural *-ōys > Latin -īs. However, in this case, we would probably
have to deal with a short ending *-oys in a probably unaccented monosyllable, so the development
might not be the same.
466 Some researchers believe that tvé is not old, for this cf. Schmidt 1978: 79-84. Cf. also Old Indic
demonstrative feminine instrumental táyā, which also does not have a long –ā- in spite of the fact that
the PIE stem was *toy-. This seems to prove that this kind of forms are not reliable enough.

Reconstruction of Proto-
Indo-European paradigms

§ 100 Here, after discussing all major problems concerning Balto-Slavic, we shall
reconstruct the whole paradigms of the Proto-Indo-European personal pronouns. The
reconstructions will be explained after the given paradigms, except for the cases already
discussed (see § 90-95, § 97-99). The basis of our approach is that the Proto-Indo-
European declension of personal pronouns did not have as much different cases as the
nominal paradigm467. In the cases where the attested forms in the daughter-languages point
to completely different forms, we shall assume that PIE just did not have those forms (or, if it
had them, that they were irreversibly lost). This approach seems much more logical than
reconstructing full paradigms, where many of the forms are based on the data from one
language only (usually Old Indic), or just on theoretical considerations (see § 83, § 84, § 86
for that kind of reconstructions). Thus, in the singular, we shall reconstruct N., G., D. and A.
and in the plural and dual the accented N. and A. only, with the unaccented GDA. forms.

1st person singular 2nd person singular

N. *égɴ, *egɴHóm, *egɴóh2 *tú/tū
G. *méne, *mey/moy *téwe, *tey/toy
D. *mégɴhi, *mey/moy *túbhi/tébhi, *tey/toy
A. *mé/mē, *me/mē *twé/twē, *te/tē

1st person plural 2nd person plural

n. *wéy, *més *yūs
g. *nos/nōs *wos/wōs
d. *nos/nōs *wos/wōs
a. *nʢsmé, *nos/nōs *uswé/usmé, *wos/wōs

1st person dual 2nd person dual

N. *wé/wē *yú/yū
G. *noh1 *woh1
D. *noh1 *woh1
A. *nʢh1wé, *noh1 *uh1wé, *woh1

467 Although it is perhaps not completely sure that it is justified to reconstruct the classic full
paradigms in the plural of PIE nouns for instance, since many of the supposed PIE endings behave in
many ways as postpositions.

Notes on the
the reconstruction:

§ 101 In the 1st person nominative singular, we reconstruct the forms *égɴ (Hittite ūg468, Old
Lithuanian eš, Slavic ja, ?Avestan azǩ469), *egɴHóm470 (Old Indic ahám, Avestan azǩɷm, OCS
azъ, Greek ǫʆγών471) and *egɴóh2 (Latin egō, Greek ǫʆγώ, Venetic ego)472. The third variant
could be an innovation, due to the pattern *-oh2 : *-om in verbs, but since it occurs in at
least three IE languages, it was probably there already in the PIE times. There seems to be no
grounds for assuming an initial *h1- in PIE (although it is not impossible) or a final *-H in the
form *égɴ473. The form *égɴ is obviously the most archaic one, with *egɴHóm and *egɴóh2 being
younger, dialectal variants. The sandhi form of *égɴ - *ékɴ, with final devoiced plosive, is
attested in Baltic (Old Prussian as, es) and apparently in the Hittite variant ugga474.
The evidence for the reconstruction of PIE *egɴHóm with final *-óm is not as firm as is
usually presumed. The attestation of *-om in Indo-Iranian is not very relevant since this *-
om is practically omnipresent in personal pronouns. However, the agreement of the accent
of Vedic ahám and Slavic *jāzъʀ (see § 13) points to the PIE *-óm and is highly indicative. It is
possible that the proliferation of *-óm in Indo-Iranian started from this form. Greek ǫʆγών
almost certainly reflects this *-óm indirectly and at least some of the Germanic forms can
probably be derived from *egɴHóm. The strongest case for the reconstruction of PIE *-óm
comes from Slavic –ъ, since this cannot be explained as secondary as Vaillant wants to (see §
11). Thus, in spite of the difficulties, it seems that the reconstruction of PIE *egɴHóm is

468 Hittite ūg has the vocalism by analogy to the old *tū. Cf. for instance Melchert 1994: 7, 84.
469 Once in Gatha-Avestan, according to Beekes 1995: 207. However, Beekes 1988a (: 137) says that
the interpretation of the form is doubtful.
470 Instead of an unknown laryngeal, *-h2- is often reconstructed here, but there is no secure evidence
that would prove the assumption. The presence of the laryngeal is confirmed by the –h- in Old Indic.
The closed syllable also permits us to explain the operation of Winter’s Law in this example (cf.
Matasović 1995).
471 -ω- is analogous to ǫʆγώ. Latin egomet is to be analyzed synchronically as ego-met, cf. also nōs-
met, uōs-met. However, it is perhaps possible that egom-et was analyzed secondarily as ego-met, and
that the forms like nōs-met were made by analogy to egomet after this reanalyzation. Oskic tiium and
Umbrian tiom probably point to analogy to the lost *egom as well.
472 Germanic is problematic. One could easily derive Gothic ik, Old English ic, Old High German ih from

PIE *egɴHóm. However, Rhunic Germanic (on the golden horn of Gallehus) shows the form ek, but horna
“horn” < PIE *kɴrʢnom. Thus, Rhunic ek cannot be derived from *egɴHóm with *-om, but from *egɴV with a
mysterious final short vowel (Sihler 1995: 369). Even stranger is the Old Nordic unaccented –eka
besides the accented ek. The form –eka could be derived from PIE *egɴHóm, but it seems unusual that
the final vowel would be preserved in an unstressed form, and lost in a stressed one (cf. Schmidt 1978:
24-25). Armenian es (instead of expected *ec, probably due to a sandhi phenomenon), is inconclusive
regarding the ending. It could be derived from both *egɴ and *egɴHóm.
473 Schmidt’s (1978: 27f) claim that the pre-form *égɴh2 cannot be reconstructed, because *-h2 would
yield *-o in Slavic, is false. In the case of the eh2-stem vocative, Slavic –o is to be reconstructed from
PIE *-e(h2) > *-a(h2) > *-a > -o, with the laryngeal dropping out before a pause, and not from a simple
*-h2. Schmidt’s derivation of Russian devjano-sto “90” directly from PIE *-kɴmʠth2 is hardly worth
474 Ranko Matasović (p. c.).

The genitive *méne475 is reconstructed based on Avestan mana and OCS mene. Old
Indic máma shows an assimilation of *m-n > *m-m and is not archaic. Welsh fyn “my” can be
derived from *men’476, a syncopated form of *méne.

The stem *mégɴh- in the dative singular is reconstructed on the basis of Old Indic
mah- and Umbrian meh-477. The secondarity of Avestan maibiiā (by analogy to taibiiā) is
proven by the Wakhī form maž “me” < Iranian *mazya478. The *men- in Balto-Slavic (besides
*mun-) is due to analogy with the genitive *méne.
The final *-i is attested in Old Lithuanian (see § 59) and in Indo-Iranian with the
added particle –a: Vedic máhya. Cf. the same particle in the o-stem dative –āy-a < PIE *-ōy.
The final –m in Vedic máhyam is due to analogy with –am in the other forms479. This unusual
final *-i, on the origin of which we shall not speculate here, was replaced by *-ey from the
nominal declensions in Italic and Old Prussian, for instance. The final *-i must be the oldest,
because it is otherwise impossible to explain, while the ending *-ey is explained by a simple
analogy. For the possibility of PIE 1st person singular dative clitic *mu instead of *mey/moy,
see § 73.

There was probably no ablative singular in PIE. Old Indic mát and Old Latin mēd are
most probably independent innovations480. There is no reason to reconstruct either the
locative or the instrumental.

2nd person nominative singular forms with *-om, like Old Indic t(u)vám, Tocharian
t(u)we, Oskic tiium and Umbrian tiom481 probably all represent independent innovations and
point indirectly to the existence of the old *-om in 1st person as well.

2nd person genitive singular *téwe is reconstructed based on Old Indic táva, OCS tebe
(with the –eb- from the dative) and East Baltic tav-.

In the 2nd person accusative singular, the distribution of *tw- in the accented forms
*twé/twē and *t- in the unaccented forms *te/tē is in fact nowhere attested. Some languages
generalized in oblique cases *t- everywhere (like OCS, Gothic, Latin), some *tw- (like

475 Oblique cases of the 1st person singular had an initial *m-. The initial a- in Hittite, ǫʆ- in Greek and
i- in Armenian does not stem from an original *h1-, but rather from the analogical *e-, taken from the
nominative form.
476 Of course, *mem’ is also theoretically possible.
477 Kortlandt’s (forthcoming c) argument that Latin mih- in mihī is archaic is not convincing. Latin mihī
and tibī have a secondary weakened min- and tib- (cf. Sihler 1995: 40 for a possible explanation). Cf.
Umbrian mehe, tefe.
478 Morgenstierne 1945: 233.
479 Cf. for instance Schmidt 1978: 65.
480 This is especially true for Latin, which introduces a secondary *-d into the ablative of all the
nominal declensions.
481 Cf. Schmidt 1978: 113-115, for a discussion on the development of these forms, which clearly
seem to be innovative.

Armenian), and some have both but without any visible difference (Greek dialects). Only
Indo-Iranian preserves the difference. There, *t- appears in the unaccented *te/oy (Old Indic
te), and *tw- appears in the accented accusative *twe- (Old Indic tvām). The Old Indic
unaccented accusative tvā is analogical. Since the unaccented *te/oy only had *t-, we
presume that the A. *te/tē also had *t- by parallel, and this makes the reconstruction of the
accented *twé/twē and the unaccented *te/tē probable482.

The same that goes for the 1st person instrumental and ablative singular goes for the
2nd person as well. Also, there was probably no locative in PIE. Vedic tvé < *twóy and
possible pre-form *twóy of Slavic tobĕ (see § 72) are probably just independent innovations,
if these forms are related at all.

§ 102 In the 1st person nominative plural, the form *wéy (not attested as such anywhere) is
reconstructed on the basis of Old Indic vay-ám (Avestan vaēm), Gothic weis (from *weys or
*weyes)483 and Hittite wēš (from *weys or *weyes). The forms *wey-s and *wey-es, with the
additional plural *-(e)s, might have existed already in PIE.

Except for the accented accusative plural form, no other accented oblique case can
be reconstructed for PIE. All other forms attested in the IE languages look like clear
innovations. This is especially obvious in the genitive plural. In the dative plural, the
accented *nʠsméy is usually reconstructed on the authority of Vedic asmé. However, this
seems to be an apparent innovation of the basis of the accusative *nʠsmé (*-ey being a
normal dative singular ending). Greek (Aeolic) αʆɴµµι cannot really be derived from the same
pre-form as Vedic asmé.

In the 2nd person accented accusative plural, I reconstruct both *uswé and *usmé (see
§ 87). The secondary *usmé seems to have been a dialectal variant already in PIE, replacing
the older *uswé.

In the accented accusative dual forms, the ending *-wé seems to have been
generalized already in PIE, although one should probably reconstruct *-mé in the 1st person
dual in pre-PIE. The original distribution of *-mé and *-wé according to the persons (1st
person: *-mé, 2nd person: *-wé) is confirmed by the singular forms *mé and *twé. This
system, arranged according to the persons, was being transformed to the system according
to number probably already in PIE484. Thus, *-wé was generalized in the dual accusative
forms, while *-mé was generalized in some languages in the plural (Indo-Iranian, Greek,

482 Cf. for instance Rasmussen 1999: 259.

483 In North and West Germanic, the reflexes like Old High German wir and Old Islandic vér point to
short *wiz. This is probably due to the influence of the dual *wit. Cf. also Schmidt 1978: 168-169.
484 Since PIE was just a normal language, like any other, I see no problem in reconstruction of an
“imperfect” system. Pre-PIE might have had a “perfect”, symmetric distribution of *-mé and *-wé, but
that certainly does not seem to be valid for the last phase of PIE we are reconstructing.

The evidence for the accusative dual forms is slim, but the reconstruction seems
accurate485. The form *nʢh1wé is reconstructed based on Avestan ǩɷǩāuuā /āwa/ and Old Indic
āvá-486, while the form *uh1wé is based on Old Indic yuvá- with the secondary y- and short –
u- from the nominative. The reconstruction of *uh1wé is, thus, based more on the supposed
structure, than on actual evidence. Greek (Boeotian and Ionic) νωɶǫ487 seems to point to
*nóh1we, a form transformed from the original *nʢh1wé, by analogy to the unaccented form
*noh1 (which became the accented νώ in Greek). The accentuation of Greek νωɶǫ points to its
secondary nature. If the form was to be derived directly from PIE *nʢh3wé, as the supporters
of the *-h₃ in the dual want to, the expected form would be **νωǫɴ.

485 For the reconstruction, cf. Cowgill 1965: 169. Even though the accusative dual forms are
reconstructed on the material of Indo-Iranian and Greek alone, structural reasons (i.e. the parallelism
of these forms with the plural ones) indicate that they are indeed archaic.
486 Vedic accusative āvām is due to the adding of the usual particle *-om of course.
487 According to Cowgill (1965: 170), the Attic form νώ is a contraction of this form, rather than a
reflex of the old unaccented *noh1, as is usually presumed.

Reflex of PIE lengthened grade in Balto-

§ 103 We have seen that all long vowels in the Balto-Slavic paradigm of personal pronouns
have the acute accent (cf. § 69 and § 81). The PIE origin of these long vowels (except for the
unaccented dual forms and perhaps the 2nd person nominative plural form) are phonetically
long vowels made by monosyllabic lengthening (cf. § 89-95). In order to account for the
origin of the acute in personal pronoun forms in Balto-Slavic, we shall first have to tackle a
great problem in the area of Balto-Slavic accentology – the reflex of PIE lengthened grade.
The traditional approach (of which Jens Elmegård Rasmussen is a modern advocate, for
instance)488 claims that PIE lengthened grade yielded the acute intonation in Balto-Slavic, just
like the long vowels made by dropping of laryngeals (which yielded the acute intonation as
well – that is accepted by everybody). In another approach, advocated by Frederik Kortlandt,
PIE lengthened grade yielded circumflex intonation in Balto-Slavic, in opposition to long
vowels made by dropping of laryngeals, which yielded the acute.
Before we start, a couple of preliminary remarks are necessary. The first question that
appears is the relevance of the problem. Simply, there seems to be a rather limited number
of the inherited PIE lengthened grade forms in Balto-Slavic. If we are talking about
approximately 10 reliable words or even less, the question of the intonation of these forms
is not very important. The second point is that one does not need to search for a simple
solution. Thus, the answer is perhaps not simple – that the regular reflex is either acute or
circumflex. The reflex might be different according to specific conditions, the reflex of the
new Balto-Slavic lengthened grade might be different than that of the old, PIE lengthened
grade, the reflex of the normal PIE lengthened grade may be different from the reflex of the
monosyllabic lengthening length etc. The third point is that one should look for examples in
which the PIE origin of the lengthened grade is certain, i.e. for the examples in which there is
evidence from other PIE languages for the reconstruction of the lengthened grade. If the
lengthened grade is found only in Balto-Slavic, that can hardly be taken as the evidence of
PIE lengthened grade, since vrddhi-formations were very productive in Balto-Slavic (see
An interesting suggestion is that the original PIE lengthened grade yields the Balto-
Slavic circumflex, while the new, Balto-Slavic lengthened grade yields the acute489. This
supposition seems to be confirmed by Slavic iteratives with a new Slavic lengthened *-i-/-y-
from the original *-ь-/-ъ-/-u-. Cf. Croatian màći : mıʂcati, dotàći : dòticati, (pro)drijéti :
pròdirati, (u)prijéti : ùpirati, (u)mrijéti : ùmirati, braʂti : ùbirati, -súti : sıʂpati, uzéti : ùzimati,
prodrijéti : pròdirati, sàhnuti : sıʂsati, svànuti : svıʂtati etc. Additional evidence of the acute in
this kind of iteratives is found in Baltic490: cf. Lithuanian mýnioti, Latvian mĩnât, meɶtât etc.
The quoted examples probably go to the same Balto-Slavic phase as the famous examples of
the acute Balto-Slavic vrddhi: Croatian vraʂna, slaʂʂva ~ Lithuanian várna491, Žemaitian šlóv÷.
The example of Croatian sjeʂći and Lithuanian s÷kti probably stems from this phase as well.

488 Cf. also for instance Panzer 1999: 327.

489 Matasović 2006.
490 Cf. for instance Endzelīns 1971: 227.
491 The other solution is to reconstruct a suffix *-h₃n- here (Kortlandt 1985: 121), cf. Greek κóρ-αξ
"raven" but κoρ-ών-η "crow".

However, I am inclined to believe that this kind of behavior is correct only for the (Balto-
)Slavic lengthened grades in particular phases. In other phases of (Balto-)Slavic, the new
lengthened grade could have behaved differently. Because of this possibility, the only cases
that may be taken as relevant in our cause are the cases where we can convincingly
reconstruct a PIE lengthened grade. It is very dangerous to lump all lengthened grades
together, PIE, Baltic and Slavic, since it is well known that this was a very productive
morphonological process in Balto-Slavic. As already mentioned, the lengthened grade could
have been prosodically treated in different ways according to the time of its origin. Cf. also
Croatian iteratives pomòći : pomágati, slòžiti : slágati, prelòmiti : prelámati, umòriti :
umárati, opròstiti : opráštati, skòčiti : skákati, prèsjeći : presijécati, izvúći : izvláčiti etc.,
which seem to stem from another, probably Slavic, phase492. However, cf. otvòriti : òtvarati,
hroʂm : hraʂmati (but also hrámati), glaʂbati (besides glábati) with an exceptional acute in the
a-iterative, and okrénuti : òkretati. In Baltic, words like Lithuanian žol÷ «grass» may be an
example of the younger circumflex-vrddhi formations. The circumflex examples could be
either Slavic, Baltic or Balto-Slavic.
Possible reflexes of PIE causatives like Croatian kániti "intend", sáditi "plant", gásiti
"extinguish fire" are not reliable. The PIE origin of the long vowel is not certain – the
formations might be younger, and sáditi and gásiti are a. p. c originally (cf. dial. Croatian
gasĩm, sadĩm). There, the acute would be eliminated either by Meillet’s Law or in pretonic
position (see below). The verb kániti could also belong to a. p. c originally, since Neo-
Štokavian kaʄnīm is inconclusive (cf. Kapović 2003: 67-68) and the word is attested in South
Slavic only. Cf. also Croatian dáviti (a. p. b, cf. Slovene dávim) from PIE *dȹow- with the
causative length. However, there are no IE cognates and this could be a younger derivation.
Another thing that must be cleared out is the question of relevance, since a lot of
irrelevant or inconclusive evidence seems to be adduced when trying to prove a case in these
discussions. In Slavic, the circumflex (in a. p. c) can replace the original acute in many cases.
The most common way is Meillet's Law493. Except for Meillet's Law, the secondary circumflex
in Slavic originates in deverbative suffixless o-stem nouns (cf. Croatian straʂšiti > strâh), in
the late spread of mobility in the i-stems (cf. Sandžak dialectal o-stem zveʂr, but Standard
Croatian i-stem zvijêr)494, and in some cases due to the secondary spread of mobility in early
Slavic (Common Slavic *dêʖverь, but Lithuanian dialectal díeveris)495. Cases like these cannot
be cited for the originality of the circumflex.
If we assume that PIE lengthened grade yielded the Balto-Slavic acute tone, these
examples would be a. p. a or a. p. c (via Meillet's Law or the other changes) in Slavic. If we

492 This type was productive in Croatian, cf. the younger stávljati instead of the older staʂvljati (the acute
is in the root here - staʂviti), poprávljati instead of pòpravljati (~ praʂviti), zaborávljati instead of
zabòravljati (~ zabòraviti), otvárati instead of òtvarati (~ otvòriti, see above), okrétati instead of òkretati
(~ okrénuti, see above) etc. Cf. also secondary Croatian bírati, dírati, tícati but original ùbirati, nàdirati,
dòticati, mıʂcati, sıʂpati. The original acute was often preserved in the verbs with a prefix. Cf. however
nazívati (~ zvaʂti).
493 By which all previous mobile acute stems (preserved in Lithuanian) become mobile circumflex
stems. Slavic mobile accentual paradigm always has the circumflex in the case of initial accent.
494 The accent was originally immobile in Balto-Slavic (PIE *gɴhweh1r). Lithuanian žv÷rìs (3) is secondary
to the unattested (1), and the Slavic mobile accent is due to the rising mobility in the i-stems.
495 For the last three developments, cf. Kapović forthcoming a.

assume that PIE lengthened grade had yielded the Balto-Slavic circumflex tone, these
examples would belong to a. p. b or a. p. c in Slavic.

if PIE *ō
*ō > BSl *ō
*ō > Slavic a. p. a/c

if PIE *ō
*ō > BSl *ō
*ō > Slavic a. p b/c

This means that Slavic a. p. c is, for all practical purposes, completely inconclusive by
itself, since it does not tell us whether the original accent in it was acute or circumflex496.
Thus, if one wants to prove that PIE lengthened grade yielded the acute tone in Balto-Slavic,
one must search for examples that belong to a. p. a to validate one's claims. A. p. c
examples are inconclusive unless the word is not attested in Lithuanian or Latvian, and a. p.
b forms are counterexamples. On the other hand, if one wants to prove that PIE lengthened
grade yielded the circumflex in Balto-Slavic, one must look for a. p. b examples in Slavic. A.
p. c examples are inconclusive, while a. p. a forms are direct counterexamples.

Relevant examples in Slavic:

for BSl *ō
* ō > Slavic a. p. a

for BSl. *ō
* ō > Slavic a. p. b

Everything is easier if a Lithuanian or Latvian cognate is attested, since there the

original acute is preserved497.

§ 104 Here, we shall examine the evidence for the reflex of PIE lengthened grade in Balto-
Slavic. We shall start with the examples adduced by Kortlandt 1997498 and Rasmussen
1999b. The latter tries to prove that the acute is the regular reflex, while the former
maintains that the circumflex is the regular reflex. Since Kortlandt 1997 is a criticism of
Rasmussen 1999b (originally 1992), we shall start from that article.
For the circumflex reflex of PIE lengthened grade, Kortlandt first adduces the
following examples: Lithuanian akmuõ «stone», dukt÷ «daughter», Latvian âbuõls «apple»
and Croatian žeʂrāv and Czech žeráv «crane». The first three examples are indeed relevant,
and these reflexes of PIE *-ōn, *-ēr and *-ōl indeed show the circumflex in Baltic. To these, I
would add PIE *-ōm > Lithuanian –ųɶ in the genitive plural of the o-stems499 (this was

496 Thus for instance, *h2ōwyóm “egg” > Croatian (Dubrovnik) jâje is just not reliable, since the word in
question is mobile and there is no Baltic cognate.
497 The cases of secondary metatony in Baltic are usually obvious.
498 Kortlandt has also presented his theory elsewhere (like Kortlandt 1985), but the article in question
is the one that deals directly with our problem.
499 Cf. the acute Lithuanian –mìs, -mì in the instrumental from *-mih1s and *-mih1, due to confusion of
the instrumental endings/postpositions *-mi(s) (cf. the Slavic u- and i-stem instrumental –mь) and *-
h1 (cf. *-oh1 in the Lithuanian o-stem instrumental –ù). The acute of the Lithuanian o-stem A. pl. –ùs
(PIE *-ons) is probably due to analogy with the ā-stems form –às, where it has a laryngeal origin (PIE *-

probably uncontracted in PIE: *-o-om). However, it must be noted that these examples only
prove that the circumflex was the regular reflex in the final syllable in front of a sonant. The
Slavic example is however somewhat difficult and Kortlandt ignores a lot of data. Croat.
žeʂrāv (a. p. c) has a variant žèrav (a. p. a), which is in accord with Slovene žerjàv500. Czech
žeráv is not in accord with Croatian žeʂrāv as Kortlandt seems to suggest501 (one would
expect Czech **žerav in a. p. c). Czech žeráv is most easily derived from the old *žerãv(j)ь (a.
p. b), which is in accord with Russian жураɴɴвль, журaвляɴ. It is difficult to say whether one
should reconstruct *žeraɾv(j)ь here (which would be a counterexample to Kortlandt's
theory502, based on Croatian žèrav and Slovene žerjàv), *žeʂrāvь (based on Croatian žeʂrāv) or
*žerãv(j)ь (based on Czech žeráv and Russian жураɴɴвль). Finally, one can conclude that this
example is not reliable (and clear) enough for us to use it in discussing the reflex of PIE
lengthened grade.
Kortlandt further adduces the Croatian aorist forms dònijēh «I brought», ùmrijēh «I
died», zàklēh «I swore» and the infinitive rıʂjet «say» (Dubrovnik). Instead of the last example,
which is an innovation that is due to the influence of the old s-aorist rıʂjeh «I said», it is
better to take the aorist form itself as an example. These examples are indeed relevant,
being the original a. p. b paradigms (cf. Croat. dial. zaklésmo, zakléste, zakléše) in a. p. c
type words (cf. the present tense donèsem, mrémo, kùnem). These forms are reflexes of the
PIE s-aorists, which had the regular lengthened grade.
The next group of examples are the Croatian 2nd and 3rd person singular aorists dâ
«you/he gave» and lî «you/he poured» compared to the 1st person singular forms daʂh, lıʂh.
Kortlandt derives the first two forms from *dōs, *lēys with a supposed disappearance of the
laryngeal after the long vowel (which then yields a circumflex), and the second two from
*doHs- and *leHis-. It is hard to understand why the 2nd and 3rd person would have a long
vowel and the 1st person would not. Kortlandt's derivation is not convincing. In fact, the
forms dâ and lî owe its circumflex to Meillet's Law and are irrelevant for our discussion503.
The next piece of the supposed evidence is the metatony in the Lithuanian future.
Again, Kortlandt assumes the loss of a laryngeal after a long vowel. I do not think that this is
the right explanation, but we cannot get into that problem here. Anyhow, this is irrelevant
since we are not dealing with PIE lengthened grade here. This kind of examples only gets in
our way of finding the correct answer.
Next, Kortlandt mentions «original root nouns», Lithuanian g÷là «pain», žol÷ «grass»
and m÷sà «meat». The first example is perhaps not very old, so I do not see the relevance.
The vocalism of the second example points to its secondarity (cf. Lithuanianžélti «grow»).
The problem with m÷sà is that it could very well be a loanword from Slavic. According to

500 Kortlandt’s (1985: 113) explanation that this is due to analogy with the homonym meaning
“burning” is strange.
501 The last syllable of the word, if long, is shortened in a. p. c in West Slavic, cf. Kapović 2003: 76-77
and 2005: 104-105.
502 This form was *gérh₂ōws with long grade in PIE!
503These verbs belong to a. p. c, cf. Croatian l-participles dála, dâlo and líla, lîlo. The relation of daʂh :
dâ is to be understood as the relation of polòmih : poʂlomi “broke”. The acute in the 2nd and 3rd person
was interpreted as initial (like in poʂlomi) and was subjected to Meillet’s Law, while the acute in the 1st
person was treated as medial accent (like in polòmih, pròdah), so there was no Meillet’s Law. Cf. the
same situation in the plural personal pronouns (see § 56, § 57).

Fraenkel, the real Lithuanian form was *mensā, reflected in Žemaitian miesa. Thus, this
example is also not conclusive504.
Kortlandt also adduces the following Slavic examples: Croatian rijeʄč «word», čaʄr
«enchantment», saʄm «alone» and Czech čár, čára and sám. Croatian examples are all mobile,
and rijeʄč is moreover an i-stem, so they are not really relevant. Czech čár (< *čãrъ, a. p. b) is
not in accord with Croatian čaʄr (< *čaʄrъ, a. p. c). Czech čár is in accord with čára (< *čāraʀ, a.
p. b), which is in accord with Ukrainian чapá and points to the original non-acute lengthened
grade (a. p. b), the antiquity of which seems to be proven by Avestan čārā. Kortlandt (1985:
118), following Vaillant, adduces more examples of the original PIE root nouns that point to
circumflex in Balto-Slavic. Many of his examples are not relevant (for instance all a. p. c
ones). However, those a. p. b examples that have a reliable attestation of lengthened grade
outside of Balto-Slavic, may very well prove that the circumflex is a regular reflex of PIE
lengthened grade.
In the case of Latvian sāls «salt» (Lithuanian sólymas «brine»), which Kortlandt
mentions, the circumflex is hard to explain, but since I believe this word ultimately stems
from PIE *seh2l-, with a laryngeal, this is of no concern to us here. As for Latvian gùovs
«cow», Kortlandt again reconstructs a lost laryngeal after the lengthened grade. I find it
much easier to reconstruct a simple PIE *gwōws. This is also one of the compelling pieces of
evidence for the circumflex reflex of PIE lengthened grade in Balto-Slavic.
Kortlandt also supposes the connection of Lithuanian ending -÷ with Vedic –dhā and
Latin –dēs. This would be another example of the loss of a laryngeal after a lengthened
grade. However, the example is not compelling.

Kortlandt also criticizes Rasmussen 1999b (= 1992). Many of his objections are valid
and I shall not repeat them. Rasmussen's connection of Lithuanian súolas «bench», Latvian
suôls with Greek (Hesychius) ǫʆɴλµατα «bedstead», attempting to prove that lengthened grade
yields acute accent, is not very persuasive. If the roots are related at all (the etymology is not
very convincing), the lengthened grade is hardly of PIE date. Kortlandt's objections to the
following examples of Rasmussen are also justifiable: Latvian vâts «wound» (I cannot accept
Rasmussen's *H3wātH2wy- as a reconstruction), Latvian smiêtiês «laugh» and Latvian lâcis
«bear» (Rasmussen's etymology is obviously not correct). As far as Rasmussen's example of
Croatian sjeʂći with the acute (cf. Lithuanian s÷kti) is concerned, it is better than the other
examples, but it is surely not of PIE date505, and thus it is irrelevant. This example probably
belongs to the phase of the Balto-Slavic acute vrddhi with the rest of the examples (see §
Kortlandt's critique of Rasmussen's circumflex in the monosyllables theory is also
well-founded. Lithuanian tieɶ "those" is obviously not old, compared to Latvian tieɶ, and
Kortlandt is right in criticizing Rasmussen's ideas about the 2nd and 3rd singular of the aorist

504 Slavic cognates, like Croatian mêso, are inconclusive due to the possible operation of Meillet’s Law.
505 The lengthened grade is attested only in Balto-Slavic.
506 The athematic formation of the iterative is a little bit strange. However, it is easy to presume that
the iterative could have won over the original form in the verb with the meaning “cut” (Ranko
Matasović, p. c.). The displacement of the normal form of the verb with an iterative is quite usual. Cf.
secondary Croatian iterative hódati, which has mostly displaced the original hòditi “walk, go”, šívati
instead of šıʂti “sew”, spávati instead of spaʂti “sleep” etc.

in Slavic. For the explanation of the falling accent in personal pronouns, which I consider to
be due to Meillet's Law, unlike Kortlandt, see § 57.

The example of Croatian vaʂditi, Slovene váditi, if it were to be related directly to

Greek ωʆϑǫɴω (LIV) would be a good example for the acute reflex of the PIE lengthened grade.
However, this might be an independent parallel development in Balto-Slavic and Greek.

§ 105 In the end, we can conclude that it seems that the regular reflex of PIE lengthened
grade in Balto-Slavic is the circumflex tone, in spite of the fact there are few certain
examples. This is confirmed by the following evidence:

PIE *-ōn > Lith. –uõ (akmuõ)

PIE *-ēr > Lith. –÷ (dukt÷)
PIE *-ōl > Latv. –õl(s) (âbuõls)
PIE *-ōm > Lith. –ųɶ (G. pl.)
PIE *-ōts > Lith. -uo (m÷nuo)
PIE *gwōws > Latv. gùovs

Besides these examples, Slavic reflexes of the original PIE root nouns, belonging to a.
p. b, also prove a non-acute reflex of PIE lengthened grade. One such example is:

PIE *kwēr- (Avestan čārā) > Czech čára, Ukrainian чapá

Moreover, the long s-aorists in Croatian, like rijeʄh < *wrēk-, definitely prove that the
circumflex is the true reflex of PIE lengthened grade in Balto-Slavic.

One could claim that the reflexes in final syllables (or in final syllables before a
resonant or just before a resonant) are a special case, e.g. that the circumflex is regular in
these positions only. However, that cannot explain the long vowel in Croatian s-aorists.
One should bear in mind that the circumflex was a regular reflex of Proto-Indo-
European lengthened grade only. In the case of the new Balto-Slavic, East Baltic and Slavic
lengthened grade, the result in some cases was obviously the acute tone (see § 103), and in
some it could have been the circumflex as well. The question of the intonation of the new
vrddhi formations in Balto-Slavic (like the Slavic iteratives) should be discussed with more
scrutiny in the future.

Acute in personal pronouns in Balto-

§ 106 Here, we shall deal with the origin of the acute in Balto-Slavic personal pronouns. As
we have already shown, the regular reflex of PIE lengthened grade in Balto-Slavic was the
circumflex intonation. However, all long-vowel personal pronouns in Balto-Slavic show the
acute and not the circumflex, in spite of the fact that in most cases the origin of their length
is not laryngeal. Thus, we must ask the question – why do the long vowels (made by
monosyllabic lengthening) of these PIE forms have the acute and not the circumflex in Balto-

There are more possible answers to this question. The first one is obvious – analogy.
How can we get the acute in all forms of Balto-Slavic long-vowel personal pronouns if we
assume that only the PIE laryngeals result in the acute intonation? Here, we shall see whether
the acute in Balto-Slavic could be the result of analogy in these cases.

According to our reconstruction (see § 90-95), the only monosyllabic pronominal

forms with the laryngeal in PIE were the dual forms *noh1 and *woh1. These would regularly
yield Balto-Slavic *nō and *wō with the acute. We could suggest that the acute of the
nominative *wē and *yū is analogous to these forms. From *yū, the acute could have spread
to *yūs. However, this could also be a regular reflex of the PIE *yúHs. Since there is no short
**yús attested, we cannot be sure that the correct reconstruction was not indeed *yúHs (see
§ 91). The plural clitics *nōs and *wōs could have gotten the acute from the dual *nō and
*wō, and the form *tū could have gotten its acute from the form *ēź, where it was due to
Winter's Law. In spite of the fact that each of these individual analogical changes is possible,
the whole explanation does not seem very convincing. One would have to assume too much
analogies, and the initial reason for the spread of the acute from only two or three forms
(*nō, *wō and perhaps *yūs) would remain completely obscure. Therefore, explanation that
involves analogy must be discarded.

Let us see some other possible explanations. We could assume that the circumflex in
Balto-Slavic was, like in Slavic, not really a special intonation, but rather a lack of an acute.
Thus, the circumflex would be regarded as an unspecified tone. Since the pronouns were
often used emphatically507, one could suggest that they have got a secondary acute in this
kind of emphatic usage. Since the acute was the only «real» tone, personal pronouns, often
used emphatically, have also acquired this «non-neutral» tone. However, this explanation
also seems a little far-fetched.

The third assumption is very simple – let us just assume that the Balto-Slavic reflex
of the normal PIE lengthened grade was different than the reflex of the PIE long vowels made
by monosyllabic lengthening. The first would regularly yield the circumflex in Balto-Slavic,

507 One could even say pronouns are used emphatically by default.

while the second one would yield the acute508. The assumption is rather simple, although
trying to think of an explanation for this kind of development might be a bit more difficult.
The assumption that the length made by monosyllabic lengthening was somehow
phonetically different than the normal length does not seem like a good suggestion. We can
perhaps offer a more convincing one. One could assume that the length in the personal
pronouns remained only phonetic in the earliest stage of Balto-Slavic. Thus, as the form *tū
was only a phonetic variant of the phonological *tú in Proto-Indo-European, the same state
of affaires was preserved in the earliest Proto-Balto-Slavic. In the meantime, when the
vowels preceding a laryngeal (and later vowels preceding voiced plosives) developed a acute
accent, the original PIE long vowels developed a circumflex accent. The long vowels made by
monosyllabic lengthening, which was perhaps still a productive rule in Balto-Slavic, were still
treated as mere phonetic variants, and the long forms acquired no special intonation. Only
later, when the short and long forms became real lexical variants (and not just phonetic
variants)509, did these long variants acquire an intonation. But in that phase of Balto-Slavic,
they acquired not the circumflex, but the acute. Thus, they developed in a different way than
the original (non-monosyllabic-length) PIE long vowels. This conclusion is supported by the
accentuation of the particles/adverbs (see § 107).

508 In this case, one would have to claim that because of PIE *gwōws > Latv. gùovs, the length in such
PIE nouns was of different origin than the length caused by monosyllabic lengthening, or that
monosyllabic lengthening in nouns was somehow different than the one in pronouns (and in particles).
The length in root nouns could be a product of an earlier phase of monosyllabic lengthening. That this
might be the case is indicated by the fact that in nouns, there is no fluctuation of length we find in
pronouns and particles (see § 90-96).
509 Probably at the time when monosyllabic lengthening stopped being a productive rule of Balto-

Acute in Balto-
Balto-Slavic particles and adverbs

§ 107 To confirm the theory layed out in § 106, one can look at the accent of the PIE
particles and adverbs that also show monosyllabic lengthening (see § 96).
As for the reflexes of PIE *nu «now», in East Baltic, Latvian nu shows the short variant
of the root, while Lithuanian nù can point to the older *nù, as well as to *nū, and is thus
inconclusive. Lithuanian nūn is probably due to the metatony and stems from an old acute.
The final accent of Lithuanian nūnaĩ is probably secondary due to the adding of the suffix or
to the metatonized form nūn. However, Old Czech form nýnie (attested for instance in
Alexandreida)510, together with Russian ныɴне points unambiguously to Common Slavic
*nyɾnĕ, which shows the acute reflex of PIE *nū.
The acute intonation is also seen in Croatian neʂ- < older njeʂ- in pronominal forms
like njeʂtko, njeʂšto, njeʂkī, njeʂkakav, njeʂkoliko, as well as in Middle Bulgarian нΌкто, нΌчто,
нΌкыи (Hock 1992, I: 121, II: 398-400) etc. This acute *neʖɾ- in Slavic stems from PIE *nē (cf.
§ 96) and confirms, together with *nyɾnĕ, my theory that the monosyllabic length yields the
acute in Balto-Slavic.

510 Modern Czech nyní is ambiguous, since the acute length disappears in front of long syllable (which
is secondary here), cf. Kapović 2003: 57 and 2005: 77.


In the dissertation, I have reconstructed the Common Slavic, East and West Baltic,
Balto-Slavic and Proto-Indo-European system of personal pronouns. Here, I shall list the
most original and innovative points of this work:

1) The detailed examination of the Slavic variants *ja and *jazъ and the evidence that points
to the direct PIE source of both forms (see § 11).

2) The convincing explanation of the two accentual variants in mentioned forms, the relation
of two different accents with two different forms (*jaɾ and *jãzъ), and the derivation of both
the accentual variants from PIE (see § 13).

3) A comprehensive treatment of the problem of the OCS form azъ and its relation to the
problem of the development of Common Slavic *ĕ-, *a- and *ja- in Slavic languages (see §

4) A comprehensive treatment of the distribution of the stems teb- and tob- in Slavic
languages (see § 47).

5) A comprehensive treatment of the problem of Slavic *mъnĕ and *mьnĕ (see § 15, § 46).

6) The clarification of some minor issues concerning the accentuation of personal pronouns
in Slavic – for instance, the origin of the accent of the Neo-Štokavian forms naʄs, vaʄs (see §

7) The explanation of the accent of the Slavic forms *mę, *tę, and the origin of the secondary
Neo-Štokavian zá me, ná te (see § 53).

8) The explanation of the conditions of the operation of Meillet's Law in personal pronouns
and elsewhere (see § 56, § 57 for the pronouns, and § 104f for the aorist).

9) A comprehensive treatment of the problem of mobility in personal pronouns paradigms,

and the position of this specific mobile paradigm in the wider frame of the accentual
mobility phenomenon in Slavic (see § 56).

10) A comprehensive treatment of the development of East Baltic paradigms of personal

pronouns, especially concerning the problems of the different stems (man-, mun-, tav-, tev-
) and accentuation (see § 66, § 71, § 73-76).

11) A comprehensive treatment of the *ew > *ow problem in Balto-Slavic and its relation to
the problems concerning the reconstruction of personal pronouns (see § 71, § 74).

12) The reconstruction of the Proto-Balto-Slavic paradigm of personal pronouns, including
the accentuation (see § 69, § 70).

13) A comprehensive treatment of the origin of the Common Slavic forms *tebeʖʀ/tobeʖʀ,
including the accentuation (see § 47, § 72).

14) The explanation of the origin of the Slavic instrumental forms mъnojNj and tobojNj,
including the accentuation (see § 54, § 77, § 78).

15) The reconstruction of the Proto-Indo-European personal pronouns' paradigms, with the
assumption these paradigms did not have 7 or 8 different cases like the nominal ones, but
less (especially in the plural) (see § 100-102).

16) A comprehensive treatment of the phenomenon of monosyllabic lengthening in Proto-

Indo-European personal pronouns (and particles/adverbs) (see § 89-95, § 96).

17) The tentative reconstruction of the Proto-Indo-European forms *túbhi (see § 98) and
*més (see § 97), otherwise not usually a part of the communis opinio.

18) A comprehensive treatment of the problem of the reflex of Proto-Indo-European

lengthened grade in Balto-Slavic (with a careful approach to Balto-Slavic data) – the
conclusion being that it yielded the circumflex. However, the idea that one has to count with
more the one phase in the development of the new Balto-Slavic vrddhi is also presented (see
§ 103-105).

19) The explanation of the acute accent in the long vowels of Balto-Slavic personal pronouns
and monosyllabic particles/adverbs, the conclusion being that the reflex of Proto-Indo-
European normal long vowels and PIE long vowels made by monosyllabic lengthening is not
the same. The former obtained the circumflex and the latter the acute accent in Balto-Slavic
(see § 106, § 107).


Aitzemüller, Rudolf 19912, Altbulgarische Grammatik als Einführung in die Slavische

Sprachwissenschaft, U. W. Weilar, Freiburg
Andersen, Henning 1996, Reconstructing prehistorical dialects: Initial vowels in Slavic and
Baltic (Trends in Linguistics. Studies and Monographs, 91), Mouton de Gruyter, Berlin
Arumaa, Peter 1933, Untersuchungen zur Geschichte der litauischen Personalpronomina,
Arumaa, Peter 1964, Urslavische Grammatik. Einführung in das vergleichende Studiun der
slavischen Sprachen, I. Band, Einleitung. Lautlehre, Carl Winter – Universitäatsverlag,
Arumaa, Peter 1985, Urslavische Grammatik. Einführung in das vergleichende Studiun der
slavischen Sprachen, III. Band, Formenlehre, Carl Winter – Universitäatsverlag, Heidelberg
Bader, Françoise 1990 «Les pronoms dans les langues indo-européennes», Mémoires de la
Société de Linguistique de Paris 1990: 23-35
Beekes, Robert S. 1988a, A Grammar of Gatha-Avestan, E. J. Brill, Leiden
Beekes, Robert S. P. 1988b, «The origin of Indo-European pronominal inflection», Festschrift
Polomé: 73-87
Beekes, Robert Stephen Paul 1995, Comparative Indo-European Linguistics: an Introduction,
John Benjamins Publishing Company, Amsterdam/Philadelphia
Белић, Александар 1909, “ЗамΟтки по чакавскимъ говорамъ”, ИзвΟстия отдΟления
русского языка и словесности Академии Наукъ 14: 181-266
Белић, Александар 19644, Историја српскохрватског језика, II/1 Peчи са деклинацијом,
Hayчна књига, Беoград
Berneker, Erich 1896, Die preussische Sprache, Karl J. Trübner, Strassburg
Бернщейн, C. Б. 1948, Paзискования в области болгарской исторической диалектологии,
Toм I, Язык валашских грамот XIV-XV веков, Издательство Aкадемии нaук CCCP, Москва -
Беседина-Heвзорова, B. П. 1962, Старославянский язык, Харьков
Bezzenberger, Adalbert 1877, Beiträge zur Geschichte der litauischen Sprache auf Grund
litauischer Texte des XVI. und des XVII. Jahrhunderts, Gottingen
Bielefeldt, Hans Holm 1961, Altslawische Grammatik, Einführung in die slawischen Sprachen,
Veb Max Niemeyer Verlag, Halle (Saale)
Birnbaum, Henrik & Schaeken, Jos 1997, Das altkirchenslavische Wort. Bildung-Bedeutung-
Herleitung. Altkirchenslavische Studien I, Verlag Otto Sagner, München
Бopковский, B. И. & Кузнецов, П. С. 19652, Историческая грамматика русского языка,
Издательство «Наука», Москва
Brlobaš, Željka 1999, «Kajkavski govor Repušnice», Rasprave Instituta za hrvatski jezik i
jezikoslovlje 25: 77-87
Brozović, Dalibor & Ivić, Pavle 1988, Jezik, srpskohrvatski/hrvatskosrpski, hrvatski ili srpski,
Jugoslavenski leksikografski zavod «Miroslav Krleža», Zagreb
Brugmann, Karl & Delbrück, Berthold 1897-1916, Grundriss der vergleichenden Grammatik
der indogermanischen Sprachen I/1-2, II/1-3, Strassburg – Berlin

Cocron, Friedrich 1962, La langue russe dans la seconde moitié du XVIIe siècle
(morphologie), Paris
Cowgill, Warren 1965, «Evidence in Greek», in: Werner Winter (ed.), Evidence for Laryngeals,
The Hague – London – Paris: 142-180
Черных, П. Я. 19623, Историческая граматика русского языка, краткий очерк, Москва
Damjanović, Stjepan 20055, Staroslavenski jezik, Hrvatska sveučilišna naklada, Zagreb
Даничић, Ђурa 1874, Историja oблика српскога или хрватскога jeзика до cвршетка XVII
виjeка, Биоград
de Bray, R. G. A. 1980, Guide to the Slavonic Languages, 1-3 (Guide to the South Slavic
languages, Guide to the West Slavic languages, Guide to the East Slavic languages), Slavica,
de Courtenay, Jan Baudouin 1929, «Einfluss der Sprache auf Weltanschauung und
Stimmung», Prace filologiczne 14: 185-256
Derksen, Rick 2002, “’Rozwadowski’s change’ in Baltic”, Baltu Filoloăija 11/1: 5-12
Diels, P. 1932-4, Altkirchenslavische Grammatik I-II, Heidelberg
Дypново, Никоай 1924, Очерк истории русского языка, Государственное издательство,
Москва - Ленинград
Дыбо, Владимир А. 1981, Славяанская акцентология. Опыт реконструкции системы
акцентных парадигм в праславянском, Издательство «Наука», Москва
Дыбо, Владимир А. 2000, Mopфологизированные парадигматические акцентные
системы. Tипология и генезис. Toм I, Языки русской культуры, Москва
ðorñić, Petar 1975, Staroslovenski jezik, Matica srpska, Novi Sad
Eckert, Reiner; Crome, Emilia & Fleckenstein, Christa 1983, Geschichte der russischen
Sprache, Veb Verlag Enzyklopädie, Leipzig
Endzelīns, Jānis 1923, Lettische Grammatik, Carl Winter’s Universitätsbuchhandlung,
Endzelīns, Jānis 1944, Altpreussische Grammatik, Latvju grāmata, Rīga
Endzelīns, Jānis 1971, Comparative phonology and morphology of the Baltic languages,
Mouton, The Hague [translation of: Baltų kalbų garsai ir formos, Valstybin÷ politin÷s ir
mokslin÷s literatūros leidykla, Vilnius 1957]
Fedák, Ján 1937, “Zo šarišského skloňovania zámenného”, Sborník Matice slovenskej 15: 72-
Филин, Ф. П. 1972, Происхождение русского, украинского и белорусского языков,
FO 1981, Fonološki opisi srpskohrvatskih/hrvatskosrpskih, slovenačkih i makedonskih
govora obuhvaćenih opšteslovenskim lingvističkim atlasom, ANUBiH, Sarajevo 1981
Fortson, Benjamin W. IV 2004, Indo-European Language and Culture, an Introduction,
Blackwell Publishing, Padstow
Fraenkel, Ernst 1950, Die baltischen Sprachen, Carl Winter, Heidelberg
Гадолина, M. A. 1963, История форм личных и возвратного местоимений в славянских
языках, AH CCCP, Институт славяноведения, Москва
Gebauer, Jan 1896, Historická mluvnice jazyka českého, III-1, Skloňování, Praha – Vídeň
Gebauer, Jan 1898, Historická mluvnice jazyka českého, Tvarosloví - Časování, Praha – Vídeň
Горшков, A. И. 1963, Старославенский язык, Государственное издательcтво «Bысшая
школа», Москва

Hamm, Josip 19703, Staroslavenska gramatika, Školska knjiga, Zagreb
Xapaламниев, Иван 2001, Историческа граматика нa българския eзик, Фадер, B. Tърново
Hermann, Eduard 1926, Litauische Studien, eine historische Untersuchung schwachbetonter
Wörter im litauischen, Weidmannsche Buchhandlung, Berlin
Hirt, Hermann 1921-37, Indogermanische Grammatik, I-VII, Heidelberg
Hock, Wolfgang 1992, Der Flexionsakzent im mittelbulgarischen Evangelie 1139 (NBKM), I:
Akzentgrammatik, II: Akzentwörterbuch, Verlag Otto Sagner, München
Holzer, Georg 1995, «Die Einlichkeit des Slavischen um 600 n. Chr. und ihr Zerfall», Wiener
Slavistisches Jahrbuch 41: 55-89
Holzer, Georg 2005 (2006), «Zur relativen Datierung prosodischer Prozesse im
Gemeinslavischen und frühen Kroatischen», Wiener Slavistisches Jahrbuch 51: 31-71
Houtzagers, H. P. 1985, The Čakavian dialect of Orlec on the island of Cres, Studies in Slavic
and general linguistics 5, Rodopi, Amsterdam
Howe, S. 1996, The Personal Pronouns in the Germanic Languages, A study of personal
pronoun morphology and change in the Germanic languages from the first records to the
present day, Berlin/NY
Hujer, O. 1911, «K slovanské deklinaci zájmenné», Sborník filologický II: 188-207
Hujer, O. 1912, «Zur Deklination der Personalpronomina», Indogermanische Forschungen
XXX: 49-54
Иллич-Свитыч, Bладислав M. 1963, Именная акцентуация в балтийском и славянском,
Издательство Академии наук СССР, Москва [translation: Nominal Accentuation in Baltic and
Slavic, The MIT Press Cambridge & London 1979]
Иванов, В. B. 1965, Общеиндоевропская, праславянская и анатолийская языковые
системы, Москва
Иванов, В. B. 19832, Историческая грамматика русского языка, «Просвещение», Москва
Иванов, В. B. & Toпоров, B. H. 1958, "On the Relations between Slavic and Baltic Languages",
4th International Congress of Slavic Studies, Moscow
Ivšić, Stjepan 1911, “Prilog za slavenski akcenat”, Rad JAZU 187: 133-208
Ivšić, Stjepan 1913, “Današńi posavski govor”, Rad JAZU 196 (I): 124-254, 197 (II): 9-138
Ivšić, Stjepan 1971, Izabrana djela iz slavenske akcentuacije (Gesammelte Schriften zum
slavischen Akzent), mit einer Einleitung sowie Berichtigungen und Ergänzungen des
Verfassers herausgegeben von Christian Alphonsus van der Berk, Slavische Propyläen, Texte
in Neu- und Nachdrucken, Band 96, Wilhelm Fink Verlag, München
Jaksche, Harald 1965, Slavische Akzentuation, II Slovenisch, Otto Harrassowitz, Wiesbaden
Якубинский, Л. П. 1953, История древнерусского языка, Москва
Janaš, Pětr 1976, Niedersorbische Grammatik, VEB Domowina – Verlag, Bautzen
Явнис, K. [= Jaunis, K.] 1908-1916, Граматика литовскаго языка, Петроградъ
Jedvaj, Josip 1956, “Bednjanski govor”, Hrvatski dijalektološki zbornik 1: 279-330
Jembrih, Alojz & Lončarić, Mijo 1982-3, “Govor Gregurovca Veterničkoga”, Rasprave Zavoda
za jezik Instituta za filologiju i folkloristiku 8-9: 5-62
Junković, Zvonimir 1972, “Jezik Antuna Vramca i podrijetlo kajkavskoga dijalekta”, Rad JAZU,
knj. 363: 1-229
Jurišić, Blaž 1966, Rječnik govora otoka Vrgade, I dio, Uvod, JAZU, Zagreb
Jurišić, Blaž 1973, Rječnik govora otoka Vrgade, usporeñen s nekim čakavskim i
zapadnoštokavskim govorima, II dio: Rječnik, JAZU, Zagreb

Jurišić, Blaž 1992, Nacrt hrvatske slovnice. Glasovi i oblici u povijesnom razvoju, Matica
hrvatska, Zagreb [reprint]
Kalsbeek, Janneke 1998, The Čakavian Dialect of Orbanići near Žminj in Istria, Amsterdam –
Kapović, Mate 2003 (2005), «Razvoj starih dugih samoglasa u hrvatskom i ostalim
slavenskim jezicima», Filologija 41: 51-82
Kapović, Mate 2005 (2006), «The Development of Proto-Slavic Quantity (from Proto-Slavic to
Modern Slavic Languages)», Wiener slavistisches Jahrbuch 51: 73-111
Kapović, Mate (forthcoming a), «Reexamining Meillet's Law»
Kapović, Mate (forthcoming b), «Bilješke o naglasku Kašićeve gramatike», Croatica et Slavica
Katz, Joshua 1998, Topics in Indo-European Personal Pronouns (PhD dissertation, Harvard)
Kazlauskas, J. 1968, Lietuvių kalbos istorin÷ gramatika, Mintis, Vilnius
Kiparsky, Valentin 1967, Russische historische Grammatik, Band II, Die Entwicklung des
Formensystems, Carl Winter – Universitätsverlag, Heidelberg 1967
Klemensiewicz, Z., Lehr-Spławiński, Tadeusz & Urbańczyk, S. 1955, Gramatyka historyczna
języka polskiego, Państwowe wydawnictwo naukowe, Warszawa
Knutsson, Knut 1935, “Altbulgarisch azъ”, Zeitschrift für slavische Philologie, XII: 94-102
Kortlandt, Frederik 1974, “Old Prussian accentuation”, Zeitschrift für vergleichende
Sprachforschung 88/2: 299-306
Kortlandt, Frederik 1985, “Long vowels in Balto-Slavic”, Baltistica XXI(2): 112-124
Kortlandt, Frederik 1997, “PIE. lengthened grade in Balto-Slavic”, Festschrift for Eric P. Hamp,
vol. II: 26-31
Kortlandt, Frederik 2000, “Initial a- and e- in Old Prussian”, Linguistica Baltica 8: 125-127
Kortlandt, Frederik 2001, “Diphthongization and monophthongization in Old Prussian”, Res
Balticae 7: 57-65
Kortlandt, Frederik (forthcoming a), “Indo-European e-, a-, o- in Slavic”
Kortlandt, Frederik (forthcoming b), “Miscellaneous remarks on Balto-Slavic accentuation”,
in: M. Kapović & R. Matasović (ed.), Tones and Theories. Proceedings of IWoBA 2005
Kortlandt, Frederik (forthcoming c), “Hittite ammuk 'me'”
Koнески, Блаже 1965, Историја на македонски јазик, «Koчо Paцин» Cкопје/«Просвета»
Križanić, Juraj 1984, Gramatično izkazanje ob ruskom jeziku, JAZU, Zagreb
Kurz, Josef 1969, Učebnice jazyka staroslověnského, Státní pedagogické nakladatelství,
Kustić, Nikola 2002, Cakavski govor grada Paga s rječnikom, Zagreb
Kyзнецов, П. С. 1959, Очерк историческoй морфологии русского языка, Москва
Lehr-Spławiński, Tadeusz 1929, Gramatyka połabska, Lwów
Lehr Spławiński, Tadeusz & Bartula, Czesław 19544, Zarys gramatyki języka staro-
cerkwieno-słowiańskiego na tle porównowczym, Zakład narodowy im. Ossolińskich
wydawnictwo, Wrocław - Kraków
Leskien, August 19226, Handbuch der Altbulgarischen (Altkirchenslavischen) Sprache, Carl
Winters Universitätsbuchhandlung, Heidelberg
Lorentz, Friedrich 1903, Slovinzische Grammatik, St. Petersburg

Lorentz, Friedrich 1925, Geschichte der pomoranischen (kaschubischen) Sprache, Walter de
Gruyter & Co, Berlin - Leipzig
Lorentz, Frydryk 1962, Gramatyka pomorska, Tom III (Fleksja, Składnia), Zakład narodowy
im. Ossolińskich wydawnictwo, Wrocław – Warszawa – Kraków
Lorentz, Friedrich 1971, Kaschubische Grammatik, Verlag Dr. H. A. Gerstenberg, Hildesheim
Łoś, J. 1922-7, Gramatyka polska I-III, Kraków
Lunt, Horace G. 19745, Old Church Slavonic grammar, Mouton, The Hague
Małecki, Mieczysław 1934, Dwie gwary macedońskie (Suche i Wysoka w Soluńskiem). I.
Teksty, II. Słownik, Kraków
Mareš, František Václav 2001, Diachronische Morphologie des Ur- und Frühslavischen, Peter
Lang, Frankfurt am Main
Matasović, Ranko 1995 (1996), "A Reexamination of Winter’s Law in Baltic and Slavic", Lingua
Posnaniensis 37: 57-70
Matasović, Ranko 2004, "The Proto-Indo-European syllabic resonants in Balto-Slavic",
Indogermanische Forschungen 109: 337-354
Matasović, Ranko 2006, Poredbenopovijesna gramatika hrvatskoga jezika (manuscript)
Matzinger, Joachim 1997, “Zu armenischen mek ʅ ‘wir’”, Historische Sprachforschung 110:
Mažiulis, Vytautas 1966, Prūsų kalbos paminklai, Mintis, Vilnius
Mažiulis, Vytautas 1970a, Baltų ir kitų indoeuropiečių kalbų santykiai, Mintis, Vilnius
Mažiulis, Vytautas 1970b, “D÷l balt. *uʠ nykimo”, in: V. Rūėe-DraviĦa (ed.), Donum Balticum
(Festschrift Ch. Stang): 334-337
Mažiulis, Vytautas 2004, Prūsų kalbos istorin÷ gramatika, Vilniaus universiteto leidykla,
Meier-Brügger, Michael 20028, Indogermanische Sprachwissenschaft. (8., überarbeitete und
ergänzte Auflage der früheren Darstellung von Hans Krahe. Unter Mitarbeit von Matthias
Fritz und Manfred Mayrhofer), Walter de Gruyter, Berlin/New York
Meillet, Antoine 19342, Le slave commun, Librairie ancienne Honoré Champion, Paris
Meillet, Antoine 19378, Introduction a l’étude comparative des langues indo-européennes,
Librairire Hachette, Paris
Melchert, H. Craig 1994, Anatolian Historical Phonology, Rodopi, Amsterdam/Atlanta
Meyer, Karl. H. 1923, Historische Grammatik der russischen Sprache I, Bonn
Mihaljević, Milan (forthcoming), “Zamjenice u hrvatskoglagoljskim fragmentima 12. i 13. st.”
Mikkola, Jooseppi Julius 1913, Urslavische Grammatik. Einführung in das vergleichende
Studium der slavischen Sprachen, I. Teil, Lautlehre, Vokalismus, Betonung, Carl Winter –
Universitätsverlag, Heidelberg
Mikkola, Jooseppi Julius 1950, Urslavische Grammatik. Einführung in das vergleichende
Studiun der slavischen Sprachen, III. Teil, Formenlehre, Carl Winter – Universitätsverlag,
Mирчев, Kирил 19632, Историческа граматика нa български eзик, Државно издателство
«Наука и изкуство», Coфия
Mladenov, Stefan 1929, Geschichte der bulgarischen Sprache, Walter de Gruyter & Co., Berlin
- Leipzig
Moguš, Milan 1966, “Današnji senjski govor”, Senjski zbornik 2: 5-152

Morgenstierne, Georg 1945, "Indo-European k' in Kafiri", New Testament Studies 13: 225-
Muka, K. A. [= Mucke, K. E.] 1891, Historische und vergleichende Laut- und Formenlehre der
niedersorbischen (niederlautsitzisch-wendischen) Sprache. Mit besonderer Berücksichtigung
der Grenzdialecte und des Obersorbischen, Leipzig
Nandriş, Grigore 1959, Old Church Slavonic Grammar, University of London, The Athlone
Press, London
Oбнорский, C. & Бapaxyдoв, C. 1938-1949, Xpecтoматия по истории pyccкoгo языка 1-2,
Olander, Thomas 2005, “The dative plural in Old Latvian and Proto-Indo-European”,
Indogermanische Forschungen 110: 273–281
Orlovský, Jozef 1975, Stredogemerské nárečia, Vydavateľstvo Osveta, Martin
Otrębsky, Jan 1956, Gramatyka języka litewskiego, Tom III, Nauka o budowie wyrazów,
Państwowe wydawnictwo naukowe, Warszawa
Otrębsky, Jan 1958, Gramatyka języka litewskiego, Tom I, Wiadności wstępne, Nauka o
głoskach, Państwowe wydawnictwo naukowe, Warszawa
Palmaitis, Mykolas Letas 1975, Baltų kalbų negimininių įvardžių istorija, Vilnius (dissertation)
Palmaitis, Mykolas Letas 1976, Prūsų kalbos negimininių įvardžių formų kilm÷, Baltistica
12(2): 156-165
Panzer, Baldur 19993, Die slavischen Sprachen in Gegenwart und Geschichte, Lang, Frankfurt
am Main
Pauliny, Eugen 1990, Vývin slovenskej deklinácie, Veda - Vydavateľstvo Slovenskej akadémie
vied, Bratislava
Pedersen, Holger & Lewis, Henry 1937, A Concise Comparative Celtic Grammar, Göttingen
Petersen, Walter 1930, «The Inflection of Indo-European Personal Pronouns», Language 6:
Pisani, Vittore 1974, “Zum Personalpronomen”, Zeitschrift für vergleichende Sprachforschung
88: 113-166
Prokosch, E. 1939, A Comparative Germanic Grammar, Philadelphia
Radulić, Ladislav 2002, Rječnik rivanjskoga govora, Matica hrvatska, Zadar
Rasmussen, Jens Elmegård 1999a, «The Constituent Elements of the Indo-European Personal
Pronouns», Selected Papers on Indo-European Linguistics. With a Section on Comparative
Eskimo Linguistics, 1-2, Museum Tusculanum Press, University of Copenhagen: 256-275
[first published in: Arbejdspapirer udsendt af Institut for Lingvistik, Københavns Universitet,
vol. 6/1987: 89-112]
Rasmussen, Jens Elmegård 1999b, «Die Vorgeschichte der baltoslavischen Akzentuierung –
Beiträge zu einer vereinfachten Lösung», Selected Papers on Indo-European Linguistics. With
a Section on Comparative Eskimo Linguistics, 1-2, Museum Tusculanum Press, University of
Copenhagen: 468-489 [first published in: B. Barschel, M. Kozianka & K. Weber (Hrsgg.):
Indogermanisch, Slawisch und Baltisch. Materialien des vom 21.-22. September 1989 in Jena
in Zusammenarbeit mit der Indogermanischen Gesellschaft durchgeführten Kolloquiums (=
Slavistische Beitraege, Bd. 285), Otto Sagner, München 1992: 173-200]
Rešetar, Milan 1911, Die serbokroatischen Kolonier Süditaliens, Alfred Hölder, Wien
Rigler, Jakob 2001, Zbrani spisi 1. Jezikozgodovinske in dialektološke razprave, Založba
ZRC, Ljubljana

Ripka, Ivor 1975, Dolnotrenčianske nárečia, Veda - Vydavateľstvo Slovenskej akadémie vied,
Rix, Helmut 19922, Historische Grammatik des Griechischen: Laut- und Formenlehre,
Wissenschaftliche Buchgesellschaft, Darmstadt
Rosenkranz, Bernhard 1955, Historische Laut- und Formenlehre des Altbulgarischen
(Altkirchenslavischen), Mouton & Co. N. V./’s-Gravenhage – Carl Winter/Universitätsverlag,
Rosinas, Albertas 1988, Baltų kalbų įvardžiai, Mokslas, Vilnius
Rosinas, Albertas 1995, Baltų kalbų įvardžiai: morfologijos raida, Vilnius
Rožić, Vatroslav 1889-1894, “Kajkavski dijalekat u Prigorju“, Rad JAZU 115: 68-136, 116:
113-174, 118: 55-115
Rudzīte, Marta 1964, Latviešu dialektoloăija, Latvijas valsts, Rīgā
Rudzīte, Marta 1993, Latviešu valodas vēsturiskā fonētika, Zvaizgne, Rīgā
Sabaliauskas, A. 1976, “D÷l lie. formos nuodu ‘MUDU’”, Baltistica 12(2): 167
Caмойленкo, C. Ф. 1962, “Из истории основ граматических форм личных местоимений в
славянских языках“, Научные доклады высшей школы, Филологические науки 2 (18): 3-15
Caвченкo, A. H. 1962, “Славянские и балтийские местоимения в соотношении к
местоимениям других индоевропейских языков“, Филологические науки 3(19): 73-81
Schenker, Alexander M. 1995, The Dawn of Slavic, An Introduction to Slavic Philology, Yale
University Press, New Heaven and London
Schleicher, August 1871, Laut- und Formenlehre der polabischen Sprache, St.-Petersburg
Schmalstieg, William R. 1971, “A new look at the Old Prussian pronoun“, Baltistica 7: 129-
Schmalstieg, William R. 1974, An Old Prussian Grammar: the phonology and morphology of
the three catechisms, The Pennsylvania State University Press, University Park – London
Schmalstieg, William R. 1976, Studies in Old Prussian, A Critical Review of the Relevant
Literature in the Field since 1945, The Pennsylvania State University Press, University Park –
Schmidt, Gernot 1968, “Zu den singularischen Genetiven der idg. Personalpronomina“,
Zeitschrift für vergleichende Sprachforschung 82/2: 225-250
Schmidt, Gernot 1978, Stammbildung und Flexion der indogermanischen
Personalpronomina, Harrassowitz, Wiesbaden
Seebold, E. 1984, Das System der Personal-pronomina in den frühgermanischen Sprachen,
Ceлищев, A. M. 1941, Cлавянское языковедение 1, Зaпаднославянские языки, Москва
Ceлищев, A. M. 1952, Cтарославянский язык II, Teксты. Cловарь. Очерки морфологии,
Senn, Alfred 1966, Handbuch der litauischen Sprache, Band I: Grammatik, Carl Winters-
Universitätsverlag, Heidelberg
Shevelov, George Y. 1964, A Prehistory of Slavic. The Historical Phonology of Common Slavic,
Winter-Universitätsverlag, Heidelberg
Shields, Kenneth 1985, “Some Remarks about the personal Pronouns”, Zeitschrift für
vergleichende Sprachforschung 98/1: 10-21
Słoński, Stanisław 1950, Gramatyka języka starosłowiańskiego (starobułgarskiego),
Państwowe zakłady wydawnictw szkołnych, Warszawa

Соболевский, A. H. 19074, Лекціи по исторіи русскаго языка, Москва
Соболевский, A. H. 1908, “Одинъ изъ законовъ церк.-слaвянскаго языка”, Jagić-
Festschrift/Zbornik u slavu Vatroslava Jagića, Berlin 1908: 204-205
Stang, Christian S. 1957, Slavonic accentuation, I kommisjon hos H. Aschehoug & Co. (W.
Nygaard), Oslo
Stang, Christian S. 1996, Vergleichende Grammatik der baltischen Sprachen,
Universitetsforlaget, Oslo - Bergen – Tromsö [with: Ergänzungsband. Register, Addenda und
Corrigenda zur VGbS, Universitetsforlaget, Oslo-Bergen-Tromsø]
Stanislav, Ján 1932, Liptovské nárečia, Turčiansky Svätý Martin
Stanislav, Ján 1958, Dejiny slovenského jazyka. II. Tvaroslovie, Veda - Vydavateľstvo
Slovenskej akadémie vied, Bratislava
Steenwijk, Han 1992, The Slovene Dialect of Resia San Giorgio, Amsterdam
Стойков, Стойко 19933, Българска диалектология, Издателство на българска академия на
науките, София
Svane, Gunnar Olaf 1958, Grammatik der slowenischen Schriftsprache, Rosenkilde und
Bagger, Kopenhagen
Szemerényi, Oswald 19893, Einführung in die Vergleichende Sprachwissenschaft,
Wissenschaftliche Buchgesellschaft, Darmstadt
Šekli, Matej 2003, “Odrazi praslovanskih oblikotvornih naglasnih tipov samostalnikov moške
o-jevske sklanjatve v (knjižni) slovenščini”, Jezikoslovni zapiski 9/2: 29-50
Šewc, Hinc 1968, Gramatika hornoserbskie rĕče, Ludowe nakładnistwo Domowina, Budyšin
Šimundić, Mate 1971, Govor Imotske krajine i Bekije, Sarajevo
Šojat, Antun 1970, “Kratak navuk jezičnice horvatske (Jezik stare kajkavske književnosti)”,
Kaj III/2: 81-96
Šojat, Antun 1973a, “Kajkavski ikavci kraj Sutle”, Rasprave Instituta za jezik 2: 37-44
Šojat, Antun 1973b, “O govoru Cerja kod Sesveta”, Rasprave Instituta za jezik 2: 45-49
Šojat, Antun 1973c, “Govor u Samoboru i njegovoj okolici”, Rasprave Instituta za jezik 2: 51-
Šojat, Antun 1982, “Turopoljski govori”, Hrvatski dijalektološki zbornik 6: 317-493
Šojat, Antun 1992, “Kulturološko-jezične značajke Pavlinskog zbornika (1644)”, Kaj XXV/4:
Šẃela, Bogumił 1952, Grammatik der Niedersorbischen Sprache, Domowina – Verlag
Bautzen, Bautzen
Tentor, Mate 1909, “Der čakavische Dialekt der Stadt Cres”, Archiv für slavische Philologie 30
Tentor, Mate 1950, Leksička slaganja creskoga narječja i slovenskoga jezika protiv Vukova
jezika, Razprave/Dissertationes I, Ljubljana
Težak, Stjepko 1981, “Ozaljski govor”, Hrvatski dijalektološki zbornik 5: 203-432
Townsend, Charles E. & Janda, Laura 2002, Gemeinslavisch und Slavisch im Vergleich:
Einfürung in die Entwicklung von Phonologie und Flexion, Otto Sagner, München [translation
of: Common and comparative Slavic: Phonology and inflection, with special attention to
Russian, Polish, Czech, Serbo-Croatian, and Bulgarian, Slavica, Columbus 1996]
Trautmann, Reinhold 1910, Die altpreußischen Sprachdenkmäler. Einleitung, Texte,
Grammatik, Wörterbuch, Vandenhoeck & Ruprecht, Göttingen
Trávníček, František 1935, Historická mluvnice československá, Praha

Trubetzkoy, Nikolaus S. 19682, Altkirchenslavische Grammatik. Schrift-, Laut- und
Formensystem, Herausgegeben von Rudolf Jagoditsch, Hermann Böhlaus Nachlf., Graz –
Wien – Köln
Užpuris, J. 1990, Trys kalbin÷s studijos, Fundation of Lithuania Minor, Chicago
Vaillant, André 1931, La langue de Dominko Zlatarić, Imprimerie nationale, Paris
Vaillant, André 1950, Grammaire comparée des langues slaves, Tome I: Phonétique, IAC,
Vaillant, André 1958, Grammaire comparée des langues slaves, Tome II. Morphologie.
Deuxième Partie: Flexion pronominale, IAC Lyon
Valeckien÷, Adel÷ 1966, “D÷l dabartin÷s lietuvių kalbos negimininių įvardžių linksniavimo
sistemų susidarymo ir raidos”, Baltistica 1(2): 127-141
Vážný, V. 1964, Historická mluvnice česká II/1, Tvarosloví: skloňovaní, Praha
Velčić, Nikola 2003, Besedar Bejske Tramuntane, Adamić, Mali Lošinj/Beli/Rijeka
Vermeer, Willem 1978, “Innovations in the Kajkavian Dialect of Bednja”, in: Dutch
Contributions to the 8th International Congress of Slavists, Lisse: 347-381
Vondrák, Wenzel 19122, Altkirchenslavische Grammatik, Weidmannsche Buchhandlung,
Vondrák, Wenzel 19242, Vergleichende slavische Grammatik, I Band. Lautlehre und
Stammbildungslehre, Vandenhoeck & Ruprecht, Göttingen
Watkins, Calvert 1969, Indogermanische Grammatik III. 1. Geschichte der indogermanischen
Verbalflexion, Heidelberg
Weingart, Miloš 1937-8, Rukovĕť jazyka staroslovĕnského 1-2, Praha
Зализняк, A. A. 1985, От праславянской акцентуации к русской, Издательство «Наука»,
Zinkevičius, Zigmas 1957, Lietuvių kalbos įvardžinotinių būdvardžių istorijos bruožai, Vilnius
Zinkevičius, Zigmas 1966, Lietuvių dialektologija, Mintis, Vilnius
Zinkevičius, Zigmas 1981, Lietuvių kalbos istorin÷ gramatika, II, Mokslas, Vilnius

Dictionaries and dialectal atlases

ARj = Rječnik hrvatskoga ili srpskoga jezika, t. 1-97 [vol. I-XXIII], JAZU, Zagreb 1881-1976
Bańkowski = Bańkowski, Andrzej 2000, Etymologiczny słownik języka polskiego, 1-2,
Wydawnictwo naukowe PWN, Warszawa
БДA = Български диалектен атлас. Съставен под ръководство иа Ст. Стойков и С. Б.
Бернщейн (I), Ст. Стойков (II), София 1964
Bezlaj = Bezlaj, France 1977-1982, Etimološki slovar slovenskega jezika 1-3, Institut za
slovenski jezik/Mladinska knjiga, Ljubljana
Boryś = Boryś, Wiesław 2005, Słownik etymologiczny języka polskiego, Wydawnictwo
literackie, Kraków
Breu & Piccoli = Breu, Walter & Piccoli, Giovanni 2000 (con la collaborazione di Snježana
Marčec), Dizionario croato molisano di Acquaviva Collecroce. Dizionario plurilingue della
lingua slava della minoranza di provenienza dalmata di Acquaviva Collecroce in Provincia di
Campobasso (dizionario, registri, grammatica, testi), Campobasso
Brückner = Brückner, Aleksander 1993, Słownik etymologiczny języka polskiego, Wiedza
Powszechna, Warszawa [reprint]

Derksen = Derksen, Rick, Slavic inherited lexicon, Indo-European Etymological Dictionary
Project (Leiden University). Available at:
ESSJ = Etymologický slovník slovanských jaziků, Slova gramatická a zájmena 2, Academia,
Praha 1980
Fraenkel = Fraenkel, Ernst 1955-1965, Litauisches Etymologisches Wörterbuch, Göttingen -
Gebauer = Gebauer, Jan 1903-1916, Slovník staročeský I-II, Praha
Gluhak = Gluhak, Alemko 1993, Hrvatski etimološki rječnik, August Cesarec, Zagreb
HSSJ = Historický slovník slovenského jazyka I, Veda/Vydavateľstvo Slovenskej akadémie
vied, Bratislava 1991
Jurišić 1973 = Jurišić, Blaž 1973, Rječnik govora otoka Vrgade, usporeñen s nekim
čakavskim i zapadnoštokavskim govorima, II dio, Rječnik, JAZU, Zagreb
Kluge = Etymologische Wörterbuch der deutschen Sprache, bearbeitet von Elmar Seebold,
Walter de Gruyter, Berlin – New York 200224
Lipljin = Lipljin, Tomislav 2002, Rječnik varaždinskoga kajkavskog govora, Garestin d.o.o.,
LIV 20012 = Rix, Helmut & Kümmel, Martin (ed.), Lexikon der Indogermanischen Verben,
Mažiulis = Mažiulis, Vytautas 1988-1998, Prūsų kalbos etimologijos žodynas I-IV, Mokslas,
Orlovský = Orlovský, Jozef 1982, Gemerský nárečový slovník, Vydavateľstvo Osveta,
Rimavská Sobota
PED = Polański, K. & Sehnert 1967, G. A., Polabian-English dictionary, The Hague-Paris
Piccoli & Sammartino = Piccoli, Agostina & Sammartino, Antonio 2000, Dizionario dell'
idioma croato-molisano di Montemitro / Rječnik moliškohrvatskoga govora Mundimitra.
[Redazione della parte fonematica e croata / Sastavljanje i prireñivanje fonološkoga i
hrvatsko-ga dijela: Snježana Marčec i Mira Menac-Mihalić], Fondazione «Agostina
Piccoli»/Matica hrvatska, Montemitro- Zagreb
Pleteršnik = Maks Pleteršnik 1894-5, Slovensko-nemški slowar, I-II, Ljubljana
Pokorny = Julius Pokorny, Indogermanisches etymologisches Wörterbuch, Francke, Bern
CДЯ = Словар русского языка (XI ~ XVII вв.), 1-26, Издательство «Наука», Москва 1975-
CДЯ 2 = Словар древнерусского языка (XI ~ XIV вв.), I-VII, red. A. Aванесов, «Pyccкий
язык», Москва 1988-2004
Skok = Petar Skok, Etimologijski rječnik hrvatskoga ili srpskoga jezika, Knjiga prva: A-J.
Knjiga druga: K-poni1. Knjiga treća: poni2-Ž. Knjiga četvrta: Kazala. Uredili akademici Mirko
Deanović i Ljudevit Jonke. Surañivao u predradnjama i priredio za tisak Valentin Putanec.
JAZU, Zagreb 1971-4
Sławski = Sławski, Franciszek 1952/6-1975/82, Słownik etymologiczny języka polskiego, I-
V, Kraków
Snoj = Snoj, Marko 20032, Slovenski etimološki slovar, Modrijan, Ljubljana
SP = Słownik prasłowiański (red. F. Sławski), 1-8, Polska Akademia Nauk, Wrocław 1961-

Cpeзнeвcкий = Cpeзнeвcкий, И. И. 1893-1903, Maтepиaлы для cлoвapя дpeвнepyccкoгo
языкa пo пиcьмeнным пaмятникaм I-III, Caнктпeтepбypг
SS = Słownik staropolski, I-X, Polska academia nauk, Warszawa 1953/5-1988/93
CCM = Cловник староукраїнскої мови XIV-XV cт., 1-2, «Наукoва книга», Kиїв 1977-1978
SSN = Slovník slovenských nárečí I, Veda/Vydavateľstvo Slovenskej akadémie vied, Bratislava
Toпоров = Toпоров, B. H., Прусский язык, I-V, Издательство «Наука», Москва 1975-1990
Tpyбачев = Этимологический словарь славянских языков (red. O. H. Tpyбачев), Aкадемия
Hayк CCCP 1974-
Vasmer = Vasmer, Max 1950-8, Russisches etymologisches Wörterbuch, Heidelberg


Alexandreida = Važný, Vaclav 19632, Staročeská Alexandreida, Nakl. Československé

Akademie věd, Praha
Codex Marianus = Jagić, Vatroslav (ed.) 1883, Quattuor evangeliorum versionis
palaeoslovenicae codex Marianus glagoliticus. Characteribus cyrillicis transcriptum V. Jagić
(mit Anhang über Schreibweise und Sprache und einem erschöpfenden Index), Berlin – Sankt
Dalimila = Bláhová, Marie & alii (ed.) 1977, Kronika tak řečeného Dalmila, Praha
Decretum = Ivanuš Pergošić, Decretum 1574, Hrvatski kajkavski editio princeps, za tisak
priredio i priloge napisao Zvonimir Bartolić, Matica hrvatska/Zrinski, Čakovec 2003
Katonova dvojverší = in: Petrů, Eduard (ed.) 1999, Ezopovy bajky, Katonova dvojverší, Rada
otce synovi, Praha
Postilla = Antun Vramec, Postilla, JAZU/Zavod za znanstveni rad Varaždin i Kršćanska
sadašnjost, Zagreb – Varaždin 1990
Psalterium Sinaiticum = Psalterium Sinaiticum, Akademischer Druck- und Verlagsanstalt
[reprint], Graz 1954
SBDO = Staročeská Bible Drážďanská a Olomoucká. Kritické vydání nejstaršího českého
překladu bible ze 14. století. I. Evangelia, Vladimír Kyas (ed.), Praha 1981
Vévoda Arnošt = in: Petrů, Eduard (ed.) 1984, Rytířské srdce majíce. Česká rytířská epika 14.
století, Praha
Závišova píseň (Jižť mne vše radost ostává) = in: Lehár, Jan (ed.) 1990 (: 244-246), Česká
středověká lyrika, Praha


This dissertation provides an overview of personal pronouns and their history in all
Balto-Slavic languages. Common Slavic, East Baltic, West Baltic, Balto-Slavic and Indo-
European system of personal pronouns (including accentuation) is reconstructed, and the
Balto-Slavic system of personal pronouns is derived from Proto-Indo-European. In the
reconstruction of Proto-Indo-European personal pronouns, emphasis is placed on Balto-
Slavic data and its use in the reconstruction. We also deal with several problems of historical
phonology, accentology and morphology of Slavic and Balto-Slavic which are related to the
reconstruction of personal pronouns.
In Slavic, there were two different forms of the 1st person nominative singular
pronoun with two different accents - *jaɾ i *jãzъ < *jāzъʀ. Both forms were inherited from
Proto-Indo-European: *égɴ (>*jaɾ) and *egɴHóm (>*jāzъʀ). Later in Slavic, one of these two
forms and accents was generalized differently in different languages. OCS azъ is due to an
irregular allegro-change of the initial *ě- > *ja- > a-. The problem of the dative/locative
singular variants *mъně and *mьně is also tackled - on the basis of the correspondence to
the Latvian dative mun, the form *mъně is interpreted as older. The dative-locative singular
variants *tebě and *tobě are closely investigated (South Slavic has only *tebě, North Slavic
has both), and possible historical explanations are suggested. The reconstruction of the
accent in the Slavic dative singular *mъneʀʀʖ, *tebeʀʀʖ/tobeʀʀʖ (cf. genitive singular *meʂne, *teʂbe),
accusative singular *męʄ and *tęʄ, and instrumental singular *mъnòjNj/mъnojNjʀ, *tobòjNj/tobojNjʀ
is discussed as well. In Slavic personal pronouns, the circumflex in the nominative-
accusative of all numbers is explained via Meillet's Law. Meillet's Law was partly analogical in
personal pronouns and not a strictly phonetic change. The special type of mobile paradigm
in personal pronouns is tackled, and some secondary Slavic forms, like Štokavian naʄs, vaʄs, zá
me etc. are explained as well. Concerning East Baltic, an overview and historical
interpretation of different singular stems is given: man- (by analogy to tav-), mun-
(originally only in the dative singular), tav- (regullarly from PIE *tew-), tev- (from older *teb-
by analogy to tav-). The Balto-Slavic change of *ew > *ow is important in the reconstruction
of personal pronouns - it is concluded that this change was regular in front of all vowels, not
only back ones. As for PIE personal pronouns, they seem to have had less cases than nouns.
In singular, there were no special forms (or no reconstructable forms) in the locative,
instrumental and ablative. In plural and dual, there were orthotonic forms in the nominative-
accusative and clitics in the genitive-dative-accusative. In Indo-European, I tentatively
reconstruct the 1st person nominative plural form *més (besides *wey-) and the 2nd person
dative singular *túbȹi (besides *tébȹi). Concerning PIE personal pronouns and monosyllabic
particles and adverbs (cf. PIE *tú/tūɴ "thou", *nú/nūɴ "now"), the problem of monosyllabic
lengthening is thoroughly dealt with. The cases of monosyllabic lengthening are interpreted
as having phonetic (not phonologic) length. The length arising from monosyllabic
lengthening yields acute in Balto-Slavic, while regular PIE lengthened grade yields circumflex
accent. As for the reflex of PIE lengthened grade in Balto-Slavic, it is concluded that a more
careful approach to the material is needed. It is likely that there were a couple of phases in
Balto-Slavic during which, depending on the phase, new vrddhi-derivations got either acute
or circumflex accent.


U radu se daje iscrpan pregled sustavâ osobnih zamjenica svih baltoslavenskih jezika
i njihove povijesti te se rekonstruira općeslavenski, istočnobaltijski, zapadnobaltijski,
baltoslavenski i indoeuropski sustav osobnih zamjenica, kao i njihov naglasni sustav.
Baltoslavenski se sustav osobnih zamjenica izvodi iz indoeuropskoga te se posebna pažnja
posvećuje upotrebi baltoslavenske jezične grañe pri rekonstrukciji indoeuropskih osobnih
zamjenica. U radu su takoñer obrañeni mnogi problemi slavenske i baltoslavenske povijesne
fonologije, akcentologije i morfologije vezani uz rekonstrukciju osobnih zamjenica.
U slavenskom su izvorno postojala dva oblika zamjenice 1. lica jednine u nominativu
koja su imala različitu akcentuaciju - *jaɾ i *jãzъ < *jāzъʀ. Oba su ta oblika naslijeñena iz
indoeuropskoga prajezika i izvode se iz praoblika *égɴ (>*jaɾ) i *egɴHóm (>*jāzъʀ). U
slavenskom je poslije došlo do poopćenja jednoga od oblikâ i jednoga od naglasaka kako u
kojem jeziku. Staroslavenski je oblik azъ postao nepravilnom allegro-promjenom početnoga
*ě- > *ja- > a-. Govori se o problemu varijanata dativa/lokativa jednine *mъně i *mьně te se
na osnovi podudarnosti s latvijskim dativom mun zaključuje da je stariji lik bio *mъně.
Detaljno se razmatraju i slavenske varijante dativa/lokativa jednine *tebě i *tobě (u
južnoslavenskom je samo *tebě, a u ostalim je slavenskim jezicima uglavnom oboje) te se
predlažu različite mogućnosti kako je do tih varijanata došlo. Objašnjava se rekonstrukcija
naglaska u slavenskom dativu jednine *mъneʀʀʖ, *tebeʀʀʖ/tobeʀʀʖ (ali u genitivu jednine *meʂne,
*teʂbe), akuzativu jednine *męʄ i *tęʄ, te u instrumentalu jednine *mъnòjNj/mъnojNjʀ,
*tobòjNj/tobojNjʀ. Cirkumfleks se u nominativu i akuzativu svih brojeva u slavenskim osobnim
zamjenicama objašnjava djelovanjem Meilletova zakona, koje je kod osobnih zamjenica bilo
djelomično i analoško, a ne strogo fonetsko pravilo. Osim toga, objašnjava se poseban tip
pomične naglasne paradigme koja se nalazi u osobnih zamjenica te se objašnjava postanak
nekih sekundarnih slavenskih oblika, kao što je štokavsko naʄs, vaʄs, zá me itd. U
istočnobaltijskom se pomno sagledavaju različite jedninske osnove koje se ondje javljaju te
se objašnjava njihov postanak: man- (analogijom prema tav-), mun- (izvorno samo u dativu
jednine), tav- (pravilno od indoeuropskoga *tew-), tev- (od starijega *teb- analogijom prema
tav-). S rekonstrukcijom je osobnih zamjenica povezana i glasovna promjena *ew > *ow u
baltoslavenskom za koju zaključujemo da je bila je djelovala ispred svih samoglasa, a ne
samo ispred stražnjih. Što se tiče indoeuropskih osobnih zamjenica, zaključuje se da su
imale manje padeža od imenica. U jednini nije bilo posebnih oblika (ili se ne mogu
rekonstruirati) za lokativ, instrumental i ablativ, a u množini je i dvojini moguće
rekonstruirati samo nominativ i akuzativ od naglašenih oblika, te nenaglasnica u genitivu,
dativu i akuzativu. U indoeuropskom provizorno rekonstruiramo i oblike *més u nominativu
množine 1. lica (uz *wey-) te *túbȹi u dativu jednine 2. lica (uz *tébȹi). Pomno se razmatra
problem duljenja u jednosložicama u indoeuropskim osobnim zamjenicama te jednosložnim
česticama i prilozima (npr. ie. *tú/tūɴ "ti", *nú/nūɴ "sada"), te se zaključuje da je riječ o
fonetskom duljenju (a ne o fonološkoj duljini). Zaključuje se da duljina nastala duljenjem u
jednosložicama daje akut u baltoslavenskom, dočim obična indoeuropska prijevojna duljina
daje cirkumfleks. Vezano uz odraz indoeuropske prijevojne duljine u baltoslavenskom,
zaključuje se da je potreban pomniji pristup grañi te da je u baltoslavenskom po svemu
sudeći bilo više faza u kojima su vrddhi-tvorbe dobivale katkad akut, katkad cirkumfleks,
ovisno kada je tvorba nastala.