Вы находитесь на странице: 1из 57

MARKING, MARKEDNESS, AND

PERSON-GENDER-NUMBER PATTERNING
IN THE ARABIC TENSES AND MOODS*

ROBERT A. FRADKIN

1. PRELIMINARIES

The inflectional system of the verb in Contemporary Stan-


dard Arabic (henceforth: Arabic)^ provides an instructive exam-
ple of the relationship between the meaning of grammatical cat-
egories and their paradigmatic morphological encoding, that is,
between semantic markedness and morphological marking. The
patterning of the person, gender, and number (PGN) categories
in Arabic demonstrates, on the one hand, the need for a binary
analysis of these features and, on the other hand, the possibility
of discovering semantic unity in the formal identity of apparently
distinct affixes (cf. Zwicky 1977). Of special interest in this re-
gard are two phenomena within the imperfect tense of the verb

An early version of this paper was presented at the Fifteenth North Ameri-
can Conference on Afro-Asiatic Linguistics, Los Angeles, 1987. Many thanks
to Rntie Adler, Lloyd B. Anderson, Robert Hoberman, Michael Silverstein,
Yishai Tobin, and Cornelis van Schooneveld for their helpful feedback on the
subsequent revisions of the manuscript. Naturally, I take all responsibility for
any errors or misstatements in the final product.
IVanscription notes: phonemic vowel length and consonant gemination are
indicated by double vowel letters, rather than macrons and raised dots. Hy-
phens reflect morpheme and lexical boundaries which Arabic orthography
does not: the prepositions bi- 'in, at, with' and It- 'to, for'; the conjunctions
WO' 'and', fa- 'and then, but'; the definite article (^a)l-, but not the indef-
inite marker "cue ending-(-n", e.g., 2al-kitdabu 'the-book-Nom.', kitdabun
'book-a-Nom.'; the anaphoric pronominal suffixes for verbal complements,
but not the PGN affixes under investigation (perfect qat^a-haa 'he killed
her', qatdltu-ka 'I killed you'; imperfect yaqtulu-haa 'he kills her', ?aqtulu-ka
'I kUl yon'), po—csrives {kitdabu-haa, kitdabu-ka 'her/your book (Nom.)'),
and prepoaitional objects {min-haa 'from her', md^a-ka 'with you'). Stress is
not phonemic in Arabic, but stress marks are provided throughout for read-
ing convenience. Most of the example sentences and their translations cited
in Section 2 below are from Cantarino (1974-75). IVanscriptions and literal
are mine.

/OK.
610

that have not, as far as I can determine, received a synchronic


treatment elsewhere:
(a) the intra-paradigm identity of the form tdqtuh 'yaam.,g./8iLe
kill(s)' for both 2nd masc. sg.-3rd fem. sg. and its dusd taq-
tuldani for 2nd ("common" gender) du.-3rd fem. du., both
forms with the t- prefix;
(b) the inter-paradigm identity of certain forms of the subjunc-
tive and jussive moods, on the one hand, and of the two
types of energetic mood, on the other (cf. Table 2, p. 653,
and Section 2 below).
The subject pronouns and the two verbal tenses, represented
by the perfect qdtala 'he killed' (Table 1, p. 652) and the imper-
fect ydqtulu 'he (will) kill(s)^ (Table 2, p. 653) mark the cate-
gories traditionally labelled lst-2nd-3rd person, masc.-fem. gen-

The Semitic languages are known for their interlacing of consonantal roots
and (basically) vocalic patterns. The consonants ate traditionally thought to
form the lexicon, usually exemplified by q-t-l with a basic meaning 'loll', and
the vocalic patterns signal grammatical and word formative adjustments of
the basic, abstract lexical meanings. The two "tenses" are the perfect qdtala
'he lolled', which functions as the dictionary citation form, and imperfect
ydqtulu 'he kills', respectively. The perfect signals person, gender, and num-
ber with sufRxes on the stem qatal-. The imperfect uses a set of prefixes for
(basically) person and a set of suffixes that supply gender and number. The
imperfect stem serves as the basis for four moods, which information is syn-
thesized with the gender-number suffixes, on which see Section 2., below. The
-II mood is called the indicative. There is also a subjunctive in -a, ydqtida, a
jussive in - 0 , ydqtvU, and an energetic in 'Onna, yaqtuldnna, which also has a
truncated variant yaqtuldn. The jussive and energetic with a zero prefix give
the imperative, (2u)qtul, (iu)qtulan, with an utterance-initial prothetic syl-
lable before consonant clusters. As for the semantics of the tenses and moods,
the native Arab grammatical tradition prefers the terms mdadi 'past' for the
perfect and muddari^ 'resembling' for the imperfect, the latter because of
the morphological resemblance of the first three mood endings, -u, -a, - 0 ,
to the three case endings -u, (nominative), -a (accusative), -i (genitive). See
Fleisch 1979: 134-135; Fradkin 1985: 254-261. The passive counterparU are
the internally-fleeted perfect qutila 'he was killed' and imperfect yuqtidu 'he
is (being) killed', which use the same suffix set and prefix-suiRx set, respec-
tively, aa the active. See Tables 1 and 2 for details. The semantic character of
the qdt<iUa/y6qtulu opposition as perfective-imperfective aspect or past-non-
pMt tense is a perennial question in Arabic and Semitic linguistics. The use
of the terms perfect and imperfect in this paper is a matter of convention,
but see Fradkin (19ft5, Chapters 3-4) and (forthcoming b.) for a full analysis
of the tense-aspect issue and the semantic analysis of the oppositions within
the mood system.
611

der, and sg.-pl.-du. number in slightly different configurations.^


The purpose of this paper is to isolate the obligatory semantic
relations in these sub-systems and delineate the minimal contri-
bution to a given utterance that each of these PGN morphemes
makes by surveying its range of use. Praguian markedness theory
will then allow us to characterize each of the affixal morphemes
in terms of the binary oppositions of these semantic properties
and to draw a dear picture of the asymmetric interaction of
the referential shifters (person) with the referential non-shifters
(gender and number) (cf. Jakobson 1957; Silverstein 1976a, 1986;
van Schooneveld 1984). Several typolo^cal patterns of PGN ex-
pression emerge in the Arabic inflectional system, overlapping and
operating simultaneously. Further, we can see what limits the sys-
tem sets on the accumulation of these semantic features in any one
pronoun or verbal affix, that is, how informative or uninformative
the system permits any given morpheme to be or insists that it

The array of derived forms in Table 3 (Appendix), serves only to under-


score that all perfect stems, regardless of their voice or word-formative se-
mantics, conform to one type of PGN inflection, and all imperfect stems
conform to another type. IVaditional grammars list a basic form qatala and
nine so-called derived stems employing internal vowel patterns, geminations
and lengthenings, atnd additional consonantal affixes. Native Arab tradition
cites the paradigms of these derivations by the 3rd masc. sg. of the perfect,
e.g., qdtcUa, qdttala, etc. Western Semiticists use a Roman numeral nomen-
clature as a shorthand for each perfect-imperfect-active-passive stem group,
e.g.. Form I {—qdtala, ydqtulu; qutila, yuqtalu), Form II (ssqdtttUa, yuqdttilu;
qittila, yuqdttcUu). Some sources give ten Forms, including XI as a secondary
derivative from IX. Table 3 shows Forms I-XI plus some rarer, unproductive
types for a total of fifteen stems. The important point for present purposes is
that all the stem derivations participate in exactly the same PGN oppositions.
See Schramm (1962) for a dassic Bloomileldian description of the stems and
also Kilani-Schoch and Dressier (1984) for more details. The increasingly in-
fluential studies in McCarthy (1981) and McCarthy and Prince (1989) 'offer a
tiered prosodic approach to the intercalation of consonantal roots with vocalic
patterns. For the semantic relations among these derived verbs, which render
a wide range of lexical and lexico-syntactic types including some notion of
"basic" and semantic additions to it, such as causative, reflexive, conative,
estimative, see Beeston (1970: 72-75) and MacDonald (1978).
612

be.* The fact that Arabic does not code non-refereatial factors
like social deference into its morphology makes it an especially
useful test case. This is not merely a problem of syntactic agree-
ment of the verb with the pronoun because the verb, at least in
the 3rd person, can both agree and disagree with its subject, more
on which below. (See Cantarino 1974 vol. I: 83-95 for details of
agreement, Fassi Fehri 1984 for a lexical-functional analysis, and
Comrie 1984 on the usefulness of studying agreement phenomena
without studying agreement, per se, including some interesting
Arabic and Maltese data.)
The theoretical point of departure here is the idea of se-
mantic markedness (cf. Jakobson 1932, 1957) and the applica-
tions, developments, and arguments for and against it in the
works of several authors on several languages.* Linguists use
the term markedness to cover a wide variety of phenomena,
but they do not always state which of the many uses of the

In this paper I do not deaj with the anaphoric pronominal suffixes, but they
express the same PGN relationships as the subject pronouns. Also excluded
from the discussion are the active vs. passive participles gdatU-, maqtuul-.
They are nominal forms that mark gender, nun:iber and definite/indefinite
in the three cases — nominative qdatilti(n), accusative qdatUa(n), genitive
qdatUi(n). They share many syntactic properties with the verb and some
grammarians consider them part of the verbal paradigm. However, they do
not participate in the person opposition. (Compare the Hebrew situation
mentioned in fn. 19 below.) See (Fradkin forthcoming a.) for a sketch of
markedness typologies across contemporary Semitic, including several repre-
sentative Arabic dialects, Hebrew, and modern South Arabian.
Linguists writing on this brand of markedness theory or appealing to an as-
sumed markedness framework include, for general theory and cross-linguistic
comparsion, Greenberg's classic work of 1966, Kurylowicz (1972a.), Silver-
stein (1976) and (1987), van Schooneveld (1977), Waugh (1982), Viel (1984),
Fradkin (1985) and (forthcoming b.); for Hebrew and Arabic: Rundgren
(1961), Khrakovskii (1965), Kurylowicz (1972b.), Levy (1982), Tobin (1985);
for Russian and Slavic: Jakobson (1932, 1936, 1957, 1958) and others, van
Schooneveld (1978a, 1978b), Sangster (1982) Shapiro (1983), Andrews (1984,
1990), Soudakoff (1986); for French: Waugh (1975, 1979); for Classical Greek:
van Schooneveld (1989). This is not to say that they all interpret markedness
the same way or all agree with each other.
613

term they mean on any given occasion: morpholo^cal muk-


ing as a physical fact (e.g., Tiersma 1982), semantic marked-
ness as a statement of differential interpretation,^ or other no-
tions such as basic/specialized, frequent/infrequent (e.g.. Green-
berg 1966), expected/unexpected use in a certain syntactic en-
vironment or text type (e.g., Fleischman 1985), normal/special
behavior within the speech community (McCawley 1985), natu-
raJ/not natural (Dressier, Mayerthaler and Wurzel, 1987), pri-
mary/derived (Kurylowicz 1972a), distributionaJly asymmetric
(van Valin, p.c), a special case of hyponymy (Kucera 1984: 65,
cf. Lyons 1977: 305-311), and universal/rare correlating with first
learned/last lost (cf. Jakobson's well-known work on acquisition
and aphasia and the whole field of study it spawned). Still oth-
ers use the terms marked/unmarked in the plus-or-minus sense
in which Trubetzkoy originally conceived it for phonological fea-
tures. In this paper I use the marked/unmarked dichotomy only
in the binary semantic sense of presence of a semantic feature
vs. lack of specification of its presence (cf. Jakobson 1932, 1957).
When both types are concerned in the course of this discussion I
state "semantically marked" or "morphologically marked."
My intention is not to argue against the other views or to
claim that markedness is a self-standing and complete theory of
language. It does capture certain insights into language structure
and operation cross-linguistically and can be seen as the intersec-
tion of several parts of several theories. A full discussion of the
generativist contributions to markedness theory in Belletti (1981),
the diverse descriptive approaches in Eckman et al. (1983), and

Unfortunately, some of this confusion between the "iorm" use and the "mean-
ing" use of the term "marking" may be due to Greenberg's own ambiguous
uses (1966), as McCawley notes in his review (1968: 566). Owens, in a specif-
ically Arabic context (1988: 199-225), also focuses on the "basic" reading
of Gteenberg's "unmarked." Such a mixed use of the term may also come
from Kurylowicz's view of primary and secondary forms and primary and
secondary uses of a form. Lyons makes almost the same points about the two
uses of markedness in his introductory chapter (Lyons 1970: 17), but he does
not seem to favor or operate with one or the other. Vinter (Winter) 1982 is one
of the few authors who call for an explicit terminology for marked/unmarked,
on the one hand, and something like normal/non-normal or simply "corre-
sponding to a norm," on the other, to accommodate expectation, syntactic
environment, and text frequency. Mayerthaler's extensive discussion (1987)
explores many implications of a theory of morphological markedness.
614

whether their use of those terms and concepts is appropriate, ef-


fective, or too diluted to be of any value is still necessary' but
hardly possible in the present limited format.
One can examine morphological or lexical forms as pairs that
contribute certain information to a given message, such that one
form states the presence of a given informational property, while
the other member of the pair remains ambiguous. Jakobson's
early formulation "statement of (property) A vs. either statement
of non-A or (mere) non-statement of A" (paraphrase of 1932:1
and 1957: 47) did not provide a methodology for determining any
given "A," how many "A's" one needed to deal with, or what their
interrelationship niight or should be, but it did stress that the so-
called "marked" form has to be "marked for something", that
is, with respect to some property, while the other form does not
necessarily, but could possibly, convey the property. The marked
form is the encoder's way of introducing the property into the
message, that is, of raising the question ofits perceptibility in or
relevance for the given situation. The opposing unmarked form
is so designated, we may now extrapolate, because it leaves it to
the decoder to identify or iail to identify the property from other
signals in the given context.
A frequent diagnostic tool for determining the marked and
unmarked terms of an opposition has been that the unmarked
form could represent neutralization of the opposition. This has,
however, always been difficult to define and less than reliable. I
suggest that this is better articulated as the non-resolution of the
question posed. In the classic example of the use of the "present"
tense to refer to the past or the use of the "singular" to refer
to the plural, the given form is not enough to signal pastness
or plurality independently. This leaves it to other clues in the
context to push the interpretation of the form in the direction of
one or the other tense or number, but the fact remains that a
decoder, upon hearing a present tense form, has to examine the
rest of the message differently from a message that contains a past
form. One must further investigate the ramifications of avoiding
specific pastness. Some of the better studied examples of the
use of the present for clear reference to a past are the tendencies
in English narrative to accord greater authority to the action or
belief in the actor (Johnstone 1987) and in Hebrew narrative to
615

signal an emotional bond to that past action or tag it as close to


one's own experience (Tobin 1989).
In a purely morphological sense markedness — perhaps bet-
ter termed, marking — means logging the presence vs. absence of
a physical element with only secondary concern for its referential
potential. The 3rd sg. of the verb in Turkish or Hungarian has
often been cited as a typical example of the unmarked member
of the paradigm by virtue of its zero-marking vs. the other verb
forms with person endings. The singular is often credited with
being the unmarked number vs. the plural because it can have
a zero ending vs. the plural, unless both numbers are morpho-
logically marked, cf. Greenberg (1966: 28). These assessments of
markedness relations are accurate but not only for those reasons.
Implicit in Jakobsonian, and partly Greenbergian, marked-
ness is a kind of "bottom-up" conceptualization of hierarchical
paradigmatic structures, such that less marked forms — at the
bottom — make the most distinctions, while certain of those dis-
tinctions are suppressed in, or do not "reach" into, the progres-
sively more complex strata of the hierarchy. This is, assumedly,
why the 3rd persons make gender distinctions which the 2nd and
1st persons do not usually make, why the singular tends to make
gender distinctions which the plural does not (e.g., German, Eng-
lish, Russian), why the nominative and accusative cases qiake a
masculine-neuter distinction, while the more marked cases do not
(e.g., Russian), why the singular or internal plural in Arabic dis-
tinguishes three cases (nom./acc./gen.), while the external plural
makes only two distinctions (nom./other), and many more. As
will become clear in the course of this paper, the Arabic verb and
pronoun paradigms afford an excellent example of a "bottom-up"
typology. Of course, apparent counterexamples are not hard to
find, such as the exceptionless distinction in the Russian case
system of dative and locative plural, while the feminine singular
dedension never distinguishes these two cases. In Section 4. be-
low I will return to the question of "bottom-up" vs. "top-down"
markedness schemes.
The presence or absence of a grammatical morpheme is nei-
ther necessary nor sufficient to dub a linguistic form marked or
unmarked. In the Arabic verb paradigm, for example, no form
is morphologically unmarked. All forms consist of a stem plus
616

some recordable affix. The 3rd ma£c. sg. perfect qdtala (that is,
qatal-{-a) is marked, as is the 1st pi. qatdlnaa. (On the case for
the iconic simplicity of the 3rd person suffix — that is, mono-
phonematic -a vs. all others, which are polyphonematic — see
Fradkin forthcoming a.) In the Arabic imperfect all forms are
equally prefixed, with 3rd masc. sg. ydqtulu carrying as much
morphological material as 1st sg. idqtulu,, 1st pi. ndqtulu or 2nd
masc/3rd fem. sg. taqtulu. Thus, even the diagrammatic iconicity
factor ceases to be a barometer of internal semantic composition.
The semantic PGN oppositions are no less present, however.^ I
assume, further, that the morphological opposition in question
carries a consistent general meaning, an approach that begins
with Jakobson's Gesamtbedeutung (1936, 1958) and has much in
common with the work of Garcia (1975: 38-58), van Schooneveld
(e.g., 1977a., 1978a., 1986), Tobin (1985, 1988), Andrews (1990),
and Ruhl's explicitly "monosemic bias" (1989: 3-5).
Extending the markedness notion to clause-level morphosyn-
tax and semantics requires access to the referential side of mor-
phology and underscores the essential asymmetry of the form-
meaning correlation. Reanalyzing the tripartite categories of per-
son and number into binary semantic features along the lines
of Jakobson (1932, 1957), Kurylowicz (1972, 1974), Silverstein
(1976b., 1987), van Schooneveld (1978b., 1988a, b.), Foley and
van Valin (1985), and to some extent, Benveniste (1946) and
Greenberg (1966, 1988), among others, permits a much finer in-
terpretation of the minimal obligatory reference of the inflected
forms and the interaction of these semantic features as a sys-
tem. The semantic oppositions of person under investigation
here, therefore, break down into the generally accepted univer-
sals participant in the speech situation (1st and 2nd person) vs.
non-participant (3rd person) and plural (plural and dual) vs. non-
plural (singular). The already binary feminine vs. non-femiiiiiie
does not readily submit to a single definition, but it is still a useful
formal-functional designation for present purposes. The notation

Benveniste's remark that "(i)n Semitic the 3rd sing, of the perfect does not
have an ending" (1946: 198) is accurate for Arabic dialects, Hebrew, and
South Arabian, but not for Standard Arabic. Of course, this does not explain
why precisely the marked tense (perfect) should have the form with the lero
ending, while the unmarked tense (imperfect) has no lero forms.
617

"+" vs. " 0 " in the following discussion signals "marked" vs. "un-
marked" in the sense of "explicit presence of feature" vs. "feature
possibly relevant and not specifically absent."
The features [-f-/0participant] and (-)-/0plural] break down
further. The 1st and 2nd persons are both marked [-l-partici-
pant], and the 1st person is, I contend, marked additionally as
[-l-participant, -l-central] vs. the 2nd person, namely, [^participant
0central], because some of the uses of the 2nd person form can in-
clude the speaker in a generalized reference but cannot specify the
speaker. (The term central is a provisional mnemonic here,® and
in note 25 below I return to the problem of markedness relations
between the 1st and 2nd persons.) The 3rd persons are unmarked
for these features, namely, [Oparticipant Ocentral]. Some uses of
the 3rd person can include reference to speaker and addressee,
even if this is not the most usual use.^
Within the plural group, the dual is marked as [-|-plural
-l-minimal] vs. the plain plural, which is [-|-plural Ominimal]. This
latter form simply signals the fact of plurality but need not specify
quantity. The dual signals not just a restricted or limited plural
Q

In the hierarchical system of recursive conceptual features that van Schoon-


eveld presents in a number of publications, this [+central] feature is roughly
a version of his third feature, "distinctness," e.g , (1982: 452), also termed
"pre-identity," (1988a.). This means that the referent so marked has a quality
that makes it identifiable only in the given speech situation, hence its shifter
quality.
The "displaced" or indirect uses of person forms that Zwicky excludes from
consideration (1977: 715-716) are actually more telling than it appears at
first glance, particularly the "phoney inclusive" we. When the nurse says to
the patient, "Are we ready for dinner?" (Zwicky's example, 1977: 716), the
chamberlain says to the king, "And how are we today, Sire?", or the parent
confronts the child with the report, "We were naughty in school today, weren't
we?", these speakers are using a form that signals not only the obvious traits
(1) that the group consists of more than one member (plural) and (2) that
the speaker is among them and can be idenitfied only by those observing
the speech situation (typical deictic center), but also (3) since the speaker
controls assignment of reference, that the speaker at the linguistic center has
the option of choosing the actual physical focus of the message. That such
uses tend to occur between parties of unequal social status is also interesting.
618

(cf. Silverstein 1976:169) but the specifically minimal quantity.^"


The singular is unmarked for these features, namely, [0plural
0minimal]. Within gender I consider [-J-feminine] semanticaily
more complex than the masculine, whidi is better characterized
[0feminine]. I take up the syntactic repercussions of this relation-
ship in Section 2. The five features of this Arabic system, then,
are (designated [-I-/0]):

participant [+/0prt]

within that, central [+/0 cnt];

plural [+/0P1]
within that, minimal [+/0 nrin];

feminine [+/0 fem].

Figure 1: Semantic Features in Arabic Conjugation

To designate the forms called 1st person as [-^participant


-f-central], the forms called 2nd person as [-|-participant 0cen-
tral], and the forms called 3rd person as [0participant 0central]
is not just a translation into new symbols of the familiar tri<ui

Referring again to van Schooneveld's features, the feature I am calling [-|-min-


imal] is a literal interpretation of the feature he calls [extension], which in-
dicates that the referent so marked is minimally affected by its relationship
with the given category (1982). Consistent with his approach, the plural is
marked for [transitivity] (cf. van Schooneveld 1980), also termed [plurality]
(cf. van Schooneveld 1986), and the dual is marked in addition with exten-
sion, cf. van Schooneveld forthcoming. I prefer the term minimal to, e.g.,
Silverstein's limited (1976b.) because there is no reason in principle why the
form could not limit the plural group to six or seventeen.
Arabic has no forni which is marked, e.g., *[0participant -|-central]. If such
a feature combination existed in a language, it might have to do with the
inclusive-exclusive distinction, but even that is suspect. Centrality in Arabic
seems to co-occur only with participation but not vice-versa. Below we will see
the significance of the fact that 0participant, by definition of being unmarked,
does not exclude reference to a participant.
Similarly to the person feature [central], the minimality feature co-occurs
with plurality. There is no Arabic form marked *[0plural-(-minimaI]. On the
other hand, the form called 1st pers. pi., as I will show below, is marked
[-(-plural 0minimal], since it covers the semantic space of both 1st pi. and the
otherwise not formalized 1st du.
619

[ispeaker, ihearer, iother] encountered in earlier work, such as


Buchler and Freeze (1966) or Ingram (1978). These authors use
binarity in the plus-or-minus sense akin to phonological distinc-
tive features. However, since "speaker" and "hearer" are equipol-
lent specific functions, those terms do not provide a basis for
claiming that one person form is marked over another. "Hearer"
is just one — albeit, the statistically preponderant one — of the
contextual applications of the form called 2nd person, as I will
elaborate on below. The asymmetric interpretation of binarity as
[+] vs. [0] affords a much clearer explanation of why one form is
considered semantically more complex or less complex than an-
other and of which minimal combinations of features occur in all
contextual applications of a form and which do not. As will be-
come clear below, the status of [-|-/0central] and [-|-/0minimal]
as features separate from [-|-/0participant] and [-|-/0plural] has
important implications for Arabic. The gender and number fea-
tures (non-shifters) combine in different ways with the participant
features (shifters). Both [-|-feminine] and [-[-minimal] may com-
bine with [-f participant], yielding in each tense five 2nd person
forms, including gender:

Perf. Impeif. (Ind.)

2 m.s. OQtuttCl tdijiulu = [+pn 0cnc 0fein 0pI 0min]

2f.s. qadM tadtuUina = [+pn 0cnt+feni 0 p l 0inin]

2 m.p. qatdltum taQtulliuna = [+pit 0cnt 0fem-t-pl 0iTun]

2f.p. qatdtiima taqtHlna = [+prt0cfit+fcm+pl0min]

2du. QauUtumaa UiQtuldani =• [-t-pit 0cnt 0fem-t-pl-Hnin]

Figure 2: Arabic 2nd Person System

but the same two features cannot combine with [-i-central], that is,
there is no form for 1st fem. or 1st du. (If there is a 1st plural group
consistingg of two women,, an adjective
j the speaker
p uses to describe
ffiam iiriii f\f /•/Mirco Ko lorn Hiial a ir n/iCnti mn^iAntAnrti *nrA
them will, of course, be fem. dual, e.g., ndfinn mariidatdani *we
620

(pi.) sick-fem. (Nom. du.)' or kunnaa mariidatdyni'^were-we (•ph)


sick-fem. (Ace. du.)', cf. the Hebrew case mentioned in note 19).

Perf. Imperf. (Ind.)

Isg. Qatdlat ^d^nJu = [+pn +cnt 0pl ]

Ipl. qadlnaa ruiqtulu = [+prt+cm+pl ]

Figure 3: 1st Person System

All feature combinations occur in the 3rd person, however:

Perf. Imperf. (Indicative)

3 msc. sg. qdtala ydqmlu = [0prt 0cm 0fem 0pl 0min]

3 fem. ss. qdtakit tdqtulu = [0prt 0cm+fem 0pl 0min]

3 msc. pi. qdtcduu yaqtuluuna = [0prt 0cnt 0fem -t-pl 0min]

3 fem. pi. qcadlna yaqiulna = [0prt 0cnt+fem+pl 0niin]

3 msc. du. qdtaloa yaqmldani = [0pn 0cnt 0fem+pl+min]

3 fem. du. qatdlataa taqtiddani = [0prt 0cnt +fem +pl +min]

Figure 4: Arabic 3rd Person System

In the number categories in Arabic conjugation the singu-


lar is the unmarked, namely, [0plural Ominimal], and the plural
is the marked number, namely, [-|-plural Ominimal]. The dual,
namely, [-f-plural -}-minimal], is called the marked plural or dou-
bly marked number. Greenberg is correct that where there is a
dual in the system the meaning of the plural is adjusted to "three
or more" (1966: 34), but this is so only in the event that counting
is at issue.^^ Whereas plural signals "more than one," the dual
13
Arabic nouns are noted for their great variety of word-internal plural for-
mations, termed "broken plurals," some of which seem to have a specifi-
cally semantic underpinning. Wright gives an exhaustive list of the possibili-
ties (1896: 199-234). Many singular nouns have several corresponding plural
forms. Where there is a contrast, certain plural forms are reserved for count-
ing with the numerals 3-10 while others are reserved for groups of ten or
621

says just how much more, in this case, the minimal candidate for
designation as plural.
At this point I will make reference to the [-l-human] feature.
All the Arabic pronouns are assumed to refer to humans, as is
generally the case in languages, and the 3rd sg. pronouns may also
refer to the entire nominal lexicon.^* The presence of [-|-plural]
in the Arabic verb, pronoun, and adjective, however, is reserved
for human plurals. Non-human plurals have fem. sg. agreement,
as shown by these examples:
(1) ?uqdbbilu l-safaf)dati wa-2idaa hiya bayddalu xaaliyatun
I-turn the-pages-[fem. pi. Ace] and-lo she white-[fem. sg.]
empty-[fem. sg.]
'I turn the pages and lo, they are white and empty.'
(2) wa-maa hiya tilka l-xayaldatu wa-l-2af;ldamu
and-what she this-[fem. sg.] the-images-[fem. pi.] and the-
dreams-[masc. pi.]
'And what were those images and dreams?'
(3) 2dmmaa bd^du-haa fa-kdanat ?aswdata diikati f»dqqan
as for some-her well-was-she voices-[masc. pi.] roosters-
[Gen.] truth-[Ace]
'Some of them were really the voices of the roosters.'
Arabic syntax requires that verbs preceding their subjects be only
singular, regardless of the number of the following subject, as in
(3) just above, but gender agreement is variable. Examples and
discussion are in Section 2 below. Since plurality of non-humans
is coded syntactically as fem. sg. agreement in the pronoun, verb

more, termed paucal and abundant plurals. (See also, for example, Fleisch
1961: 495-6.). These, along with the cooperative duals like 2abawaani 'two
fathers' = 'parents', qamardani 'two moons' = 'sun and moon' (cf. Green-
berg's reference to the Arabic term tagliib 'dominance', 1966: 30) have no
special agreement morphology in the verb beyond the regular plural and
dual, respectively.
Kuryiowicz's interesting position on this matter is that presence of animacy
is the unmarked state of affairs, and 3rd persons — in this Arabic case, the
3rd sg. — are marked, since they occupy a privileged position by virtue of
their unique ability to refer to either animates or non-animates (1972: 123).
However, that makes the referential potential of the 3rd sg. pronouns, vis.,
Arabic huwa-hiya, much broader than other pronouns, which is one of the
classic criteria of the unmarked.
622

and adjective, there is no overt animacy agreement other than


that implicit in the 3rd pi. of a following verb. Therefore, I do
not include animacy in the features under discussion here. Fur-
ther, morphological gender in the non-participant categories can
refer to either natural or grammatical gender, while in the partic-
ipant category, specifically the 2nd persons, gender refers to sex
of humans.
As in all linguistic work, it is important to keep the tradi-
tional label of the form separate from its total range of use in
the language. The unmarked forms, which I denote, e.g., [0plu-
ral] are often read "non-plural" and are written [-pi], meaning
"having no plurality" However, it is crucial to bear in mind that
[0plural] here indicates "a form not necessarily referring to plu-
ral but not excluding the possibility of reference to a plural." The
forms so indicated may certainly take on the implied reference of
"absence or exclusion of the feature" in many contexts, but to
say, for example, that the category singular is "non-plural" in the
sense of "excluding reference to a plural," is not the broadest, and
certainly not the most useful, characterization of the form called
singular. Both the terminology "non-plural" and the notation
[0plural] are meant here as "unmarked for plurality"
Similarly, the traditional 3rd persons in the strict sense are
not participants in the speech exchange (cf. the native Arabic
term 2al-faa7ib 'the absent one' and the Hebrew ha-nistdr 'the
hidden one'), but the linguistic form characterized as [Opartid-
pant] or "non-participant" may, indeed, designate a physical par-
ticipant in the speech situation — and not just in Arabic. Typical
examples of the 3rd person pronoun (and corresponding 3rd per-
son verb form) are the forms of polite address or social distance
in German (Sie) Polish {Pan, etc.), or Italian (Let), all of which
forms can refer to either 2nd or 3rd persons, though which one
is usually clear in any given context. Arabic, similar to Russian
and Hebrew, can use the 3rd masc. pi. verb form with no pronoun
for the sa called impersonal — perhaps better, omnipersonal^' —
type, as in

The term omnipenonal first appe&rs for this construction, as f v as I know,


in Fradkin 1991: 273.
623

(4) yaquluuna 7inna l-ftubba 2tfimda


say-[3 masc. pi.] that the-love-[Acc.] blind
'People say that love is blind.'
One of the pragmatic reasons for a speaker to use such a construc-
tion is to mask his or her own opinion of the popular wisdom. The
real-world referent of such an action may be any person, possibly
including but not specifiying the given speaker or addressee.^^
The linguistic category singular can be just as specific about
its reference to "oneness" as plural can be about "more-than-
oneness," but the paradigmatic linguistic form labelled singular
in Arabic, as in many languages, has a range of reference much
broader than its use as referring to "one item." Within the plu-
ral opposition, by the same token, the non-minimal (plain) plural
may imply any potential two of a generic plurality but not spec-
ify any given two, as the dual must. It can also refer to a dual
where there is only morphological plural as in the 1st. pi. in Ara-
bic, where the [-I-/0 minimal] opposition is not overtly realized.
Similarly, a masculine form can make a positive reference to a
masculine referent, while the form called masculine may include
reference to, but not specify, a feminine.

2 . MARKING AND MARKEDNESS IN ARABIC

The verbal inflectional affixes across Semitic, both the suf-


fixes of the perfect qdtala and the prefixes of the imperfect
ydqtulu, have long been thought to derive historically from per-
sonal pronouns (cf. Moscati 1964: 137, Givon 1976), and this

One syntactic construction in Arabic shows that the 3rd masc. sg. is truly
unmarked for person. In the so-called exceptive construction, e.g.,
lam ydpdar 2Ulaa 2dnal2dntum
not come [3 msc. sg. juss.) except I [nom.]/you-pl. [nom.]
'No one came but me/you.' or 'Only I/you came.'
the semantically less specific 3rd person form merely sets up the expectation
that this, strictly speaking, subjectless construction can have a verbal referent
of potentially greater semantic complexity. That information comes only later
in the utterance. The nominative after the particle 2Maa may refer to any
person, both physically and linguistically. Here it is possible to say that 2anaa
is the semantic referent of yafidar and not the syntactic subject.
624

creates the deceptive impression that the affixes are straightfor-


ward agreement copies for those pronouns. In the attested an-
cient and modern Semitic languages, however, that resemblance
is only partial, at best: probable for the 2nd persons, plausible
for the 1st persons, but fanciful for the 3rd persons (cf. Gesenius
1909: 126, also Russell 1984). On closer inspection it becomes
clear that the pronoun, perfect, and imperfect signal the PGN
features each in its own way.
As Table 1 (Appendix) shows there are twelve independent
(basically, subject or topicalizing) pronouns and thirteen PGN
suffixes for the perfect tense stem qatal-. The stem -qtul-, the
basis for the imperfect tense including the moods, distributes the
signalization of the PGN features over a set of four consonan-
tal prefixes for (basically) person, with some inherent gender or
number:

= 1st sg.

n- •• 1 s t p i .
= 3rd persons only, but not all
= all 2nd persons, also 3rcl fern, sg.

Figure S: Imperfect Prefix Set

Eighteen suffixes give gender, number, and mood:

[0pl0fem] [0pl +feinl [•t-plOfem] [+pl +imn] [+pl +feinl


Ind. -u -iin^ -uun/a •aan/i
Sbjn. -a
•ii -uu -aa -na
Jss. -0
Enrg. -an(na) -in(na) •un(na) •aanni -naanni

Figure 6: Imperfect Suffixes, Including Moods

(Of the Energetic forms the full form, including the material
in parentheses, is Energetic I. With the parenthesed material
deleted, the form is Energetic II, and I count it separately in
the eighteen suffixes. The indicative suffixes with a slash indicate
625

that the auslaut short vowel is deleted before pause, but it is not
counted as a separate item.) Together they form forty-two prefix-
suffix groupings, cf. Figure 16 below, and each mood paradigm
consists of eleven of them (Table 2, p. 653). Table 4, p. 656,
compares the maximum potential PGN oppositions that such a
feature system can make with those actually morphologically en-
coded in Arabic. This is especially relevant for the present study
because the pronoun and the verb in Arabic syntax can lead very
separate lives. Pronouns can be matched by zero-verb in so-called
equational sentences, and a single inflected verb form can form the
entire core*' of a clause. (See Fassi Fehri 1982 for the syntax of
pro-deletion, zero-anaphora, and control in Arabic.) The object
of inquiry here is not just the individual morphemes, themselves,
but the ways in which these readily identifiable morphemes do
or do not support the opposition of the semantic features in the
system. There is only partial, but nonetheless telling, feature-to-
form isomorphism.
As is clear from Tables 1 and 2 (p. 652-653), nearly every
pronoun has a distinct perfect suffix and imperfect prefix/suffix
cooperation to match it. There are, however, three classic exam-
ples of neutralization in the agreement system:
(a) The 3rd du. of the verb distinguishes gender — per-
fect masc./fem. qdtalaa-qatdlataa, imperfect masc./fem. yaq-
tiildani-taqtuldani — while the corresponding 3rd du. pro-
noun hiimaa is unmarked for gender. (The entire 2nd du.,
by comparison — pronoun 2dntumaa and perfect/imperfect
qatdltumaa-taqtuldani — is unmarked for gender, for which
phenomenon one sometimes encounters the term "common
gender.")
(b) Verbs that precede their subjects — which implies only
3rd persons — may agree in gender but must not agree in
plural number. Thus, the verb in such a clause can be only
3rd masc. sg. or 3rd fem. sg.
(c) Three of the four imperfect prefixes are unambiguous as to
person: 3- and n- for 1st, y- for 3rd. Only the prefix <-, when

Foley and v u Valin (1984) and van Valin (1989) work with this unit in their
Role and Reference Grammar and not with separate NP and VP. The core
connsta of the bare predicate plus its obligatory arguments.
626

it combines with a mood-only suffix, e.g., indicative tdqtulu,


can be either 2nd masc. sg. (pronoun 2anta) or 3rd fem. sg.
(pronoun htya). The same holds for the du. taqtuldani and
its moods for 2nd du. (pronoun 2antumaa) and 3rd fem. du.
(pronoun humaa). That is, this t- is the only form in Arabic
inflection that blurs the line between participants and non-
participants, but we are not justified in saying that person
or participant is neutralized in the form tdqtulu.
Cases (a) and (b) are fairly straightforward examples of the gen-
erally agreed upon unmarked character of the masculine vs. the
feminine and of the singular vs. the plural, hence one dual pro-
noun sufficing for both genders and the lack of explicit plural
agreement in verbs preceding their subjects. In many languages
the verb makes exactly as many PGN distinctions as the pronoun
(as in Turkish, with PN but no G), discounts one of the pronoun
features (as in Romance or Germanic, with gender in pronouns
but not in verbs, and most of Slavic, where the verb has either
person and number or gender and number) or discounts two of
the three (as in English, with minimal person marking in the
present and no PGN in the past). In (a) above the Arabic verb
makes a distinction (gendered duals) not paralleled in the pro-
noun. In other words, the verb carries morphological expression
of a particular and systemically plausible combination of seman-
tic features which the pronoun system does not utilize.^® The

Indo-European languages with gendered 3rd person pronouns or special


accommodations for social distance (e.g., Dutch U, Polish Pan, Pant,
Patlttwo), or indefinite (human) subject (e.g. French on, German man) do
not mark those categories on the verb. If the verbal form is historically a
participle, as in the Slavic past tenses in -( or the Hebrew present qotil, cog-
nate with Arabic qaatil-, then gender is part of the verb system but usually
incompatible with simultaneous person marking, cf. Jakobson (1957: 53) on
Russian. (This is why it is somewhat misleading for Anderson and Keenan to
say that such a gendered predicate as Hebrew ant medabtr 'I speak-masc.'
vs an{ medahiret 'I speak-fem.' forces a gender distinction in the otherwise
non-gendered 1st person, cf. 1985: 269-270. The 1st person in Hebrew is
unmarked for gender and not non-gendered.) However, it is worth noting
that languages with a 1st person inclusive-exclusive or a 3rd person obviate
category, as Moravcsik (1978: 357) points out, mark it on the verb but not
the pronoun. (Ideally, studies of pronoun or person systems, such as Ingram
1978, Wiesemann 1986, or Greenberg 1988, should mention the corresponding
conjugation system more consistently.)
627

morphemic non-differentiation of 2nd masc. sg. and 3rd fem. sg.


in (c), however, is puzzling, indeed.
The prefix t- of the imperfect in case (c), viz., indicative
taqtulu 'youm.«j./she kill(s)' (either 2nd masc. sg. or 3rd fem. sg.)
is not only ambiguous for person but at the same time involves
gender, but not number. Note, however, that in the plain plurals
[-|-pl 0min] the prefixes t- and y- are the sole markers for 2nd and
3rd persons, while the suffixes separate the genders:

2nd msc. pi. taqttiluuna 3rd msc. pi. yaqaduuno

2nd fem. pi. laqnilna 3rd fem. pi. yaqtulna

Figure 7: Imperfect, Non-Central Plurals

Rather than merely attribute the formal identity of the 2nd masc.
sg. prefix t- and the 3rd fem. sg. prefix t- to two historical sources,
as traditional Semiticists do (on which, see below) or dismiss t-
as simply a homophonous or polysemous prefix, as contemporary
authors implicitly do, I contend that in synchronic perspective
this t- is a single prefix and that there is a unifying semantic ba-
sis for the two readings in terms of the five component features
that combine to appear as person, gender, and number. In sim-
ilar fashion, though much more schematically, I will propose a
reason why certain forms of the subjunctive and jussive moods
are identical:

taqadii 2 fem. sg.


taeitulmt 2 msc. pi. yaqtuluu 3 msc. pi.
3 fem.-2 du. yOQtuloci 3 msc. du.

Figure 8: Imperfect, Subjunctive-Jussive Identity

and the two forms that serve for indicative, subjunctive, and jus-
sive:
628

saqtMna 2 fern, pi.


yaQtiUna 3 fem. pi.

Figure 9: Imperfect, Indicative-Subjuncdve-Jussive Identity

(I choose the neutral term identity as inescapable observation


over the more technically loaded syncretism). The last four forms
cited are the same ones that in the energetic mood permit only
one form, while all other energetics have a long and short form:

wqnddanrd 2nd du.-3rd fem. du.

yaqtuldanm 3rd msc. du.

tcufmlfidofvu 2nd fem. pi.

yoQtidnddtvu 3rd fem. pi.

Figure 10: Imperfect, Single-Fonn Energetics

Semitic philological tradition connects the t- prefix with two


other Semitic Vs: the fem. suffix -/ of the noun, the 3rd sg. per-
fect -at, or the demonstrative, e.g., tilka, on the one hand, and
the 2nd masc. pronoun idnta, on the other.^^ I have seen no
explanation in the literature, in terms of either synchronic Arabic
or some notion of synchronic Proto-Semitic, why only these two

One of the clearest statements comes from the encyclopedic and historical-
comparative-philological Gesenius-Kautzsch-Cowley (1909: 126):
...The preformative t- of the second persons is, without doubt, connected
with the t- of 2atta, 2attem, &c... The preformatives of the third persons (y-,
(-) have not yet met with any satisfactory explanation. With (- might most
obviously be compared the original feminine ending •( of nouns and of the
3rd sing. fem. perfect.
See also Fleisch (1979: 120). This situation holds throughout all Semitic,
and I have found no explantion for it in the literature. Gray mentioned the
view that this four-prefix system indicates that Proto-Semitic had only four
personal pronouns (1934: 65), though he does not attempt to pout the forms
of these pronouns or suggest the referential range of whatever proto-form
yielded the ambiguous prefix (•. Givdn (1976) proposes the syntactic state of
Proto-Semitic that gave rise to the prefix- vs. suffix-conjug»tion, but it is not
his aim to dissect the paradigm-internal category relations.
629

feature bundles (2 masc. sg., 3 fem. sg.) map into the same form
tdqtulu. Even in a diachronic framework it is not dear why a femi-
nine demonstrative and a masculine personal pronoun should have
coalesced, in the first place, and survived for so many millenia in
Semitic speech territory, in the second place. Indeed, no Semitic
speech community has, as far as I know, regularized gender ex-
pression by extending the suffixal 2nd masc./fem. sg. proportion
tdqtulu/taqtuliina to the 3rd masc./fem. sg. ydqtulu/tdqtulu, that
is, there is no ydqtulu/*yaqtuliina. Even in Maltese and the other
North African dialects where gender is a strictly 3rd person fea-
ture, the same proportion cognate with yaqtulu/taqtulu still ex-
presses 3rd masc./fem. sg., and the prefix t- still crosses the
participant boundary.
Formal ties between the 2nd and 3rd person verb forms are
familiar from the pronoun and conjugation systems of some Paleo-
Siberian and Amerindian languages (cf. Zwicky 1977: 727-8), and
several branches of Indo-European exhibit the same tendency.^"
Similarly, alongside other universal tendencies to divide, for ex-
ample, demonstratives and spatial or temporal deictics according
to speaker-oriented, addressee-oriented and other-oriented, there
are also systems with speaker-orientation vs. all else, cf., e.g.,
Kurylowicz (1972b, 1974), Anderson and Keenan (1985: 280-
^^ What is unusual about this Arabic verb phenomenon is

French unites the 2nd-3rd sg. of all non-periphrastic tenses of the indicative
(outside of environments of elision). The Dutch present has 2nd-3rd sg. in
-( vs. 1st sg. in - 0 , e.g., jij-hxj slaapt vs. ik ilaap 'you-he vs. I sleep'. One
might also consider the submorphemic umlaut in the corresponding German
forms, e.g., du schlSfst, er ichlaft vs. ich schlafe 'you, he vs. I sleep', a similar
phenomenon. The same is true of the Old Slavic past tenses, e.g., 'pray': 2nd-
3rd sg. aorist moli vs. 1st sg. molix'h and 2nd-3rd sg. imperfect moljaaie
vs. ist sg. moljaax'i. Modern Serbo-Croatian, Macedonian, Bulgarian, and
Lusatian have the same situation in these tenses, and Lusatian has it in the
dual of all tenses. Slovenian has identical 2nd-3rd dual in ite only conjugated
tense, the present.
^^ Compare well-known three-term demonstrative and deictic adverb systems
like Spanish and 7\irkish with the two-term systems like the (basically)
speaker vs. non-speaker in this vs. that. An interesting case is Hebrew, which
in normal speech has only a single demonstrative masc. sg. xe, fem. sg. toft),
pi. ^le or ^{u. In order to narrow down speech roles Hebrew resorte to the
personal pronoun-based masc. sg. ha-hu, fem. sg. ha-hi, masc. pi. ha-Mm,
fem. pi. ha-him, which gloss only 'that' but with little concern for physical
proximity to addressee-orientation. It is, rather, indicative of a contrast of at
630

that it splits up part of the 2nd person with part of the 3rd per-
son and accentuates the different roles that gender and nuniber
play in Arabic syntax and morphology, as the examples below
illustrate. What has this to do with markedness relations?
Greenberg (1966) exemplifies several of his well-known
markedness criteria with data from Arabic. (See also the dis-
cussion of these in Owens 1988: 199-226.) As a case of the
marked form having less morphological irregularity (perhaps bet-
ter termed, more morphological diversity), he cites the perfect
tense qdtala as marked vs. the imperfect ydqtulu because of the
greater range of denotations of the latter — a semantic/syntactic
criterion — and the formal fact that the imperfect has suffixal
mood variations, namely subjunctive ydqtula, jussive ydqtul, and
energetic yaqtuldn(na), while the perfect does not (1966:48-49).^^
As for the Arabic derivational system of verb stem types,
given here in Table 3, (p. 655), Greenberg (1966: 29) invokes the
criterion of greater morphological variation to point out that Form
I qdtala is unmarked vs. the other nine Forms, e.g., II qdttala, IV
2dqtala, etc., again, because Form I has three varieties of medial
stem vowel (between the 2nd and 3rd root consonants) in the
perfect, namely, a, i, u in Form la. qdtala, Ib. qdtila, Ic. qdtula,
while the other perfects have only the medial stem vowel a, e.g..
Form n qdttala, IV 2dqtala, X (li)stdqtala. The corresponding
imperfects of Form I also have three medial vowellings u, a, t,
namely, ydqtulu for la., c. ydqtalu for Ia.,b. and ydqtilu for la..

least two items, and not surprisingly, the second member of two pointed out.
Therefore, it covers only part of the semantic space of, e.g., English that. As
Adler (1986) has shown the use of ha-hu in Hebrew discourse also serves to
create a speaker-addressee in-group and give deliberately opaque anaphora
of an entity known to speaker and addressee and specifically excluding all
other hearers.
The perfect is, indeed, semantically marked vs. the imperfect, cf. Kurylowici
(1972a: 91), Fradkin (1985: 242), but the number of imperfect forms is not
in itself sufRcient to claim this. In many tense systems, e.g., Romance or
Germanic, there is only one present tense opposed to at least tvro past tenses,
but that does not change the fact that the past tense is semantically marked
and the present tense, unmarked.
631

only." The imperfects of Forms II-IV and VII-XI all have an


i-vowel, e.g., II yuqdttilu, IV yuqtilu, X yastdqtilu, and only V
and VI have an o-vowel, viz., V yataqdttalu, VI yataqdatalu. In
terms of number Greenberg gives the dual as the marked number
for two reasons: first, because the gender distinction maintained
in the singular and plural of the pronoun does not appear in the
dual (2nd 2dntumaa, 3rd hiimaa) and second, because it signals
a cooperative pair where one member dominates, cf. 2abawdani,
the dual of 2ab 'father', for 'parents = father and mother'.
Several other observations about markedness in Arabic fol-
low quite naturally. In the case system the singular and internal
plural distinguish three: nominative in -u, accusative in -a, gen-
itive in -t. External plurals reduce to two cases, where syntactic
accusative and genitive in -Una, whether definite or indefinite, op-
pose nominative in -uuna. (The secondary two-case subsystem,
where indefinite accusative and genitive in -a oppose nominative
in -u, deserves its own presentation on another occasion.)
Returning to the verb system, formally, the four forms of the
Arabic imperfect with (person) prefix and minimal (mood-only)
suffixal information — 2dqtulu, ndqtulu, tdqtulu, ydqtulu 'I, we,
youma»c.«j./she, he kill(s)' — all carry equal amounts of mor-
phological material, but semantically, the 1st and 2nd persons

Semitic philological tradition seems to see this as an implicit hierarchy of


syntactic restriction or conceptual complexity from la. to Ib. to Ic. Form
la. emcompasses — to use the inadequate traditional terminology — both
transitive and intransitive verbs; Ib. is intransitive or temporary state, such
as marida-yamradu 'be sick', although some are undeniably transitive, such
as idriha-ydirahu 'drink'); Ic. is always intransitive, but of an adjectival verb
or permanent state, such as kohun-yakhuru 'grow, get big'. See, e.g., Fleisch
(1979: 229-262). (In terms of the sound structure of these forms, Chomsky
and Halle discuss them as polarity phenomena, with a tonality reversal of
high/non-high or low/non-low in the perfect vs. imperfect stem vowels, cf.
Arabic a-ti in ftftala-ydgtuiu and i'O in qdtUa-ydqtalu. The corresponding
Hebrew ntuation is a-o, as in qatdl-yiqtol 'kill'. Unfortunately, the Hebrew
example Chomsky and Halle cite, lamdd-yilmod 'learn' (1968:356) was poorly
choaea: it belongs to a small class in Hebrew that regularly has a prefix-
form stem vowel a, namely, lamdd-yilmdd, which disproves their point about
tonality. The imperfect *yilm6d does not exist. A more typical verb like
qatdl-yiqtdl 'kill' or katdv-yixtdv 'write' would have served them better.)
632

are marked vs. the 3rd persons.^^ The short-vowel suffixes and
zero, viz., -u, -a, - 0 , -an(na) carry mood information only, leav-
ing person, gender, and number indication to the meager forces
of the prefix. The main issue in this paper is the person-gender
relation, as raised by the existence of tdqtulu. The long-vowel suf-
fixes -uun(a), -iin(a), -aan(i) and the one consonant-initial suffix
-na combine only with the prefixes t-, y- and signal either marked
number [+pl. 0fem.] (2 m. pi. taqtuluuna, 3m. pi. yaqtxduuna),
marked gender [0pl. -|-fem.] (2 f. sg. taqtuliina), or in the case of
-na, both at once [-|-pl. -|-fem.] (2 f. pi. taqtulna, 3 f. pi. yaqiulna).
These problems have their solution only in such a componential
feature approach as I employ here.
The so-called subordinate moods of the imperfect, the sub-
junctive ydqtula and the jussive ydqtul, are generally mentioned in
grammar books only insofar as they are governed by various syn-
tactic particles, cf. Gantarino, vol. I: 77-8. However, the moods
present a semantic hierarchy of their own, built up of two seman-
tic features. To explain the nature of those features here would

The question of the markedness relations of Ist vs. 2nd person or 2nd vs.
1st is far from settled for the same reason as many problems of marked-
ness: it is not always stated from the outset whether form, general meaning,
specific function or a combination of these is at issue. Greenberg sees the
2nd as marked vs. 1st on the basis of the formal individuality in, e.g., the
German past tense du gingst vs. the 1st and 3rd ich/er ging 'you vs. I/he
went' (1966: 44). Jakobson asserts that 1st is marked on the semantic basis
of its unique identification of the referent (1932, 1957), and van Schooneveld
corroborates this view (1982: 452-54, 1988a, b). Silverstein suggests a func-
tional basis for the markedness of the 2nd person: whereas the 1st person
assumes the existence of a speaker, the 2nd person is actually the more infor-
mative category (hence, marked) because it selects and, thereby, creates the
addressee (1976b: 171). Moravcsik (1978: 355) and Tiersma (1982: 845-846)
summarize other arguments in support of the marked status of the 2nd over
the 1st, including text frequency and the direction of diachronic paradigm
levelling. Kudera (1984: 65) suggests that the 2nd person is marked in view
of Czech because in the past tense the 2nd person requires an auxiliary verb,
while the 1st and 3rd do not. It is arguable that the function of reference to
an addressee is a "marked" function in some sense, but the point remains that
the linguistic form called 2nd person has as part of its range of application
the ability to include reference to a 1st person, while the reverse is not so. I
maintain, then, that the 1st is marked, cf. the following omniperaonal usage
of the 2nd, typical of Arabic and many other languages:
2d0manu l-hibdati hibatun laa tdjhalu waahiha-haa
precious-[superl.] the-gifts-[Gen.] gift-(Nom.] no you-know giver- [Acc.]-her
The most precious gift is the gift of which you are ignorant of the donor.
633

take us well beyond the scope of the present work, but suffice
it to say for now that the subjunctive ydqtula is marked for one
of these features vs. the (unmarked) indicative ydqtulu, and the
jussive ydqtul is marked for the other feature vs. the (unmarked)
indicative.'* The indicative is the unmarked mood semantically,
although it is just as morphologically marked as the subjunctive
and more marked than the zero-suffixed jussive. (This recalls the
situation of the English present tense where the semantically un-
marked 3rd sg. is morphologically marked in -s and opposes the
form in -0.) In the suffixed forms for marked gender [-l-fem.] or
marked number [-1-plural], the two mood oppositions [subjunc-
tive vs. indicative] and pussive vs. indicative] reduce to a single
opposition of one semantically marked mood (in the sense that
it carries more information about the perception of the verbal
process) vs. (unmarked) plain indicative,'^^ thus:

The two features involved characterize the perceptibility of the verbal process
in the given narrated situation, as I explored in Fradkin (1985: 228-242). In
terms of van Schooneveld's system, the subjunctive is marked for extension,
meaning that the verbal process may or may not be perceivable in the nar-
rated situation. The jussive is marked for the feature objectiveness, which
means that the stated verbal process has no guaranteed referent in the nar-
rated situation. The energetic, as I discuss below, is marked simultaneously
for both features. See van Schooneveld (1989a) for the notion of semantic
levels in tense vs. mood.
The weakest version of the form-meaning correlation — that form signals
meaning — prevents the claim that the long-vowel suffixes signal one mood
or the other. If each of the two moods has its own feature vs. the indicative,
how can their shared suffixes -ii, -uu, -aa be marked for either? A stronger
version is that the moods function in opposition. All that is clear is that the
suffixes without final -n are identifiable as "not the indicative" and specify
more about the verbal process than the indicative does. The abstract feature
itself does not correspond to any single morpheme, and only other clues in the
syntax will indicate the import of the form in question. The same problem
has been evident since the beginning of this kind of inquiry, viz. Jakobson
(1932, 1936): it is hard to claim that the accusative case in Russian is marked
for "directionality" (Jakobson 1958: 109) or "extension" (van Schooneveld
1986: 383) vs. the unmarked nominative, given that in three of the four sin-
gular noun declensions in Russian the nominative and accusative are identical
and all four in the plural have the nominative the same as the accusative.
This does not invalidate the feature notion. It is obvious that the form alone
does not uniquely announce the feature: only in syntax can there be a nomi-
native or accusative, a subjunctive or Jussive, a 2nd masc sg. or a 3rd fem sg.
What is crucial is that the form, which syntax can assign to nominative or
accusative positions, signals to the receiver the minimum information neces-
sary for the interpretation of the form. This is why morphology is allied more
634

2 f. sg. 2 m. pi. 3 m. pi. 3 f.-2 d. 3 m. pi.


Sbjn-Jss taqtulli taqmliiu yaqtu!uu taqndda y(U}ndda
Indicative taqndilna Uu/tulilna yaqtulmna uujtuldani yaqtuliani

Figure 11: Imperfect, Fonns with One-Mood Opposition

In all these pairs the semantically marked form is anti-dia-


grammatically (cf. Kilani-Schoch and Dressier 1984) or counter-
iconically (Mayerthaler 1987: 49) shorter by one, and optionally
two, phonemes -n(a). In the forms that are simultaneously [-f-plu-
ral -f-fem.], namely 2nd/3rd fem. pi. all three moods merge as
one, namely, taqtulna/yaqtulna. The general observation that the
marked forms have less morphological variation is now roughly
quantifiable for Arabic: the marked moods oppose each other in
the relatively unmarked categories (with neither gender nor num-
ber signalled in the suffix), but merge in the categories marked for
the non-shifters, either gender [-(-fem.] or number [-1-pl.]. (This in-
cludes the dual [-1-plural -|-minimal], since gender in the dual of the
imperfect is not given in the suffix.) When both gender and num-
ber information come into play those two moods — "indicative-
plus" vs. plain indicative — reduce even further to one, gener-
alized mood, namely, 2nd/3rd fem. pi. taqtulna/yaqtulna for in-
dicative, subjunctive, and jussive.
The energetic yaqtuldn(na) is always distinct in form from
the other moods. Formally, it is at lejist bi-phonematic {-an) and
optionally bi-syllabic (-anna), making it also the most complex
of the mood forms. Semantically, it is marked for the features
of both the subjunctive and the jussive (Fradkin 1985: 239-240;
note 25 above), making it the semantically most complex mood,
as well. Based on the coincidence of the jussive and subjunctive
paradigms as just discussed above, one might expect this doubly
complex mood to merge with the others, as well, but this is not the
case. Instead, the pattern replicates itself in another way. In its

with syntax in current work, such as Anderson 1989, than phonology, as in


the early days of transformational grammar. The important point is still the
fact of formal-functional oppositions, and the task of the present paper is to
establish the significant oppositions in Arabic morphology as a case in point.
Later they need to be integrated into a theory of syntax.
635

semantically unmarked suffix categories, namely, masculine sin-


gulars [0pl. Ofem.], there is both a longer form, termed Energetic
I, yaqtuldnna and a shorter form, termed Energetic II, yaqtuldn.^''
This is also true when the suffix indicates either marked number
or marked gender:

[0pl. 0fem.] mood only: P*-aqnddn(no)

r+pi. 0fem.] 2 msc. pi. taqttdun(na)

3 msc. pi. yaqtidunfna)

f0pi. +fem.] 2 fem. sg- tcuitid(n(na).

Figure 12: Imperfect, Forms with Energetics I and D


(P4 indicates all four prefixes can combine with this suffix.)

The energetic forms with both marks [-|-pl. -l-fem.] or with double
plural marking [-t-pl. -|-min.] admit only one type (cf. Figure 10,
above):

2-3 fem. pi. taqadndanm -yaqmlndanni

2 du., 3 fem. du.-3rd msc. du. tcujmliianni •yaqaddanni

Figure 13: Imperfect, Doubly Mariced Fonns with One Energetic

Note, too, that these semantically complex forms have an un-


usually complex syllable structure for Arabic, namely, long vowel
in closed syllable: (G)VVCG. (See also Fleisch 1979: 119 for dis-
cussion of shortened forms for these, as well.) Thus, again, the
marked moods do not simply all collapse with increased seman-
tic complexity. This energetic mood, with its two-feature load, is
unique in the system and reflects that semantic uniqueness in its
formal distinctness.

These two energetic forms are largely in free variation, although there has
been no special study of that variation, to my knowledge. Cantarino reminds
us (1974: 82) that since the orthography masks the difference between them,
since printed texts indicate neither consonant gemination nor auslaut short
vowels, it is not always possible to tell which one an author means in any
given context.
636

Semitic morpholo^cal studies in the American Structuralist


tradition used to describe such a form as 2nd fem. sg. taqtultina as
having a discontinuous morpheme, t—Una opposed to 2nd masc.
sg. tdqtulu with discontinuous t—u. More recent morphological
terminology might call this a circumfix, since both parts envelop-
ing the stem are necessary for the correct interpetation of the
form. However, it is evident from my presentation so far that
I consider each part of these Arabic affixes to have its own se-
mantic characterization. Each element contributes its separate
component of meaning to the given verb form, as Table 5 (Ap-
pendix) shows.^®
The most significant aspect of the syntax of number in Ara-
bic which demonstrates the unmarked character of the singular
is that in VSO clauses with a nominal subject the 3rd person
verb can be only singular regardless of the number of the fol-
lowing subject. Gender agreement is possible but not obligatory,
especially if that feminine subject is separated from the verb, as
by an adverbial in the less usual VXS order (see, e.g., Cantarino
1974: 83-95 for details). A clause-initial masc. singular verb form
leaves open the possibility that a subject of either gender or either
number will follow. An initial fem. sg. verb, however, excludes one
possibility, namely, a following masc. sg. subject.^^ By implica-
tion, then, a clause-initial verb with plural marking is anaphoric.
Following are some examples of frequent patterns of verb-subject
"disagreement":

At present, I cannot explain why the prefix (- is read only as 2nd person
given either gender or number information in the suffix — cf. 2nd fem. sg.
taqtultina, 2nd masc. pi. taqtultiuna, 2nd fem. pi. taqtulna — except for
the dual taqtuldani, which is still ambiguous for person. In addition, one
could well argue that the fem. sg. imperfect suffix -iina, since it combines
only with the prefix t-, should carry a person marking [-hprt 0cnt-|-fem &pi]
in its feature specification. This would make it the only suffix with specific
person information and undermine the complementarity of person and gender
between prefix and suffix. For now I will leave person marking out of its
specification.
One can argue that this does not prove the unmarked character of the singular
at all because the choice of singular is obligatory in this position and not a
mere "lack of specification" of number. Though there can be no syntactic
opposition in such circumstances, the singular form nonetheless continues to
raise the question of the [-i-/0plural] feature.
637

(5) masc. sg. verb with:


(a) fem. sg. subject
kdana lit mrdJatun
was-[3 masc. sg.] to-me woman-[Nom. fem. sg.]
I had a wife.
(b) masc. non-hum, pi. subject
lawlda-hu la-4dcfia 2asmda2u ka$iirin min-a l-kutubi I-
nafAsati
if-not-him then-got lost-[3 masc. sg.] names-[Nom. masc.
pi.] many-[Gen. many] from the-books-[Gen. masc] the-
precious-[fem. sg.]
If it had not been for him the titles of many precious books
would have been lost. (sg. 2ism 'name', kitaab are masc.)
(c) masc. hum. pi. subject
7ixtdlafa l-mu^arrixtiuna
disagreed [3 masc. sg.] the-historians [Nom. masc.]
Historians disagree (have disagreed).
(d) masc. du. subject
mddaa 2usbuufdani fdlaa tilka l-ldylati
past [3masc. sg] week-2 [Nom. masc] on that the-night
Two weeks had passed since that night.
(e) fem. pi. hum. subject
jdaZa fit 8uhbdti-haa ntswatun 2drbcfi un
came [3mascsg.] in accompaniment-[Gen.]-her women-[Nom.]
Four women arrived accompanying her. four
(f) fem. du. subject
yakjii-hi min-a l-ndwmi saa*atdani
[3 masc. sg.]-suffice-him from the-sleep hour-2-[Nom.]
Two hours of sleep are enough for him.
Consistent with Russell (1984), in 5 a, e, and f, the feminine sub-
ject is either indefinite or "delayed," that is, separated from the
verb in this assumedly V(X)S(O) construction. Such subjects are
le8B topiicalized and tend toward less anticipatory gender agree-
ment.
(6) fem. sg. verb with:
(a) masc. non-hum, pi. subject
tdfmuru kaa2indatun 'yariibatun min-haa /-tamoasuKu l-ldtii
tazddridu l-naasa zdirdadan...
638

live-[3 fem. sg.] beings-[Nom. masc] strange-[fem. sg.]


from-her the-crocodiles-[Nom.] the-which-[fem. sg.]
swallow-3[fem.sg.] the-people-[Acc.] a swallow-[Acc.]
Some strange beings live there ... among them are
crocodiles which can easily swallow people ...
(sg. kaa2inun 'being', timsdafihun 'crocodile' are masc.)
(b) masc. hum. pi. subject
qdalat l-hukamda2u
said-[3 fem. sg.] the-wise men-[Nom. masc]
The wise men have said ...
(c) fem. non-hum, pi. subject
sdwfa tamujTu nasiimdatun ladiifatun wa-tdf^milu budiiura-
kum 2ilaa l-samsi
will pass-[3 fem. sg.] breezes-[Nom. fem. pi.] gentle-
[Nom. fem.sg.] and-carry-[3 fem sg.] seeds-[Acc.]-your-
[masc.pl.] to the-sun-[Gen.]
Gentle breezes will blow and carry your seeds to the sun.
(d) fem. du. subject
2intdlaqat-i l-sayyaaratdani
left-[3 fem.sg.] the-cars-2 [Nom.]
The two cars drove off.
Agreement in both gender and number is, however, obligatory
when the verb follows the subject. Thus, in conjoined predicates
(7) or compound verb phr<ises (8) there is first disagreement, then
agreement:
(7) masc. sg. verb-|-masc. pi. subject-|-masc. pi. verb
jda2a l-2awldadu wa-qdaluu...
came [3 masc. sg.] the-boys and-said [3 masc. pi.]
The boys came and said....
(8) masc./fem. sg. verb-f-fem. pi. subject-|-fem. pi. verb
(a) compound past tense: kaana 'be'-f imperfect
kaana l-nisda2u yataf>addd9na 2ildy-him
was-[3 masc. sg.] the-women-[Nom.] speak-[3 fem. pi.) to-them
The women used to talk to them.
(b) Aktionsart-iyYte: auxiliary verb-|-imperfect
jdfalat nisda2u qurdyiin yamSiina xildala sufuufi-haa
started-[3fem.sg.] women-[Nom.] Quraush-[Gen.] waIk-[3
639

fem. pi.] among ranks-[Gen.]-her


The women of Quraish started to walk through its ranks.
All the foregoing applies, of course, to [Oparticipant] subjects
but not to [+participant] subjects: a 2nd pi. subject will not take
a 2nd sg. verb, nor will a 2nd fem. sg. subject take a 2nd masc. sg.
verb or a 1st pi. subject take a 1st sg. verb. This is no doubt tied
to the general non-obligatory occurrence of these pronouns with
verbs and to the fact that as the hierarchical markings accumulate
they have narrower and narrower fields of reference. Thus, 3
masc. sg. forms can be impersonal, and 3 masc. pi. forms can be
omnipersonal, as in example (4), p. 623 above. 2nd masc. sg.
can also be a, perhaps more intimate, kind of omnipersonal, as
in note 24. However, neither 2 nor 3 fem. sg. or pi. can be so
general.
Gender in Arabic, as is now apparent, presents a single
morpho-syntactic opposition of feminine vs. non-feminine which
serves several related purposes. Feminine nominals vs. masculine,
for example, often signal more concrete or individualizable items,
at least in derivational pairs distinguished by gender:

Masc. coll: idjarun 'trees' verbal noun: iarbun 'hitting'


Fem. item: idjaratun 'a tree' instance: iarbatun 'ahif

Figure 14: Gender in some noun types

Feminine marking in morphology, then, can indicate, vis-a-vis


the masculine, some extra information about its referent.^ The
agreeing preposed verb can use only this marked category to antic-
ipate either of the marked non-shifters, gender or number. Arabic
syntax either uses this gender property as a surrogate for plurality

This lends more than incidental support to Russell's astute obaervations


(1984) on the higher correlation of feminine agreement of a clause-initial
verb to a following definite feminine subject, as opposed to an indefinite one:
stricter feature agreement is concomitant with greater topicality. That is,
the fenunine verb form already announces that its upcoming subject is more
thui "just unmarked." It would also be interesting on another occasion to see
whether the agreement hierarchy that Corbett worked out for Slavic (1983)
is u y put to Ara,bic.
640

within the [0participant] category or recognizes that both non-


shifter categories signal additional information about their refer-
ents. In the increasing complexity of the participant hierardiy,
gender combines with [-f-participant] to signal the sex of human
referents (cf. p. 622 above), but crucially, it cannot do so with the
more complex [-|-participant -l-central]. The gap in the otherwise
symmetrical pattern — namely, the lack of 1st. fem. — has a
basis in the binary feature system.

3 . FORM AND MEANING IN THE PRONOUN AND VERB^^

Taking a closer look at Table 1, we see that the Arabic pro-


nouns represent the features just discussed in a straightforward
manner: [-^participant] is always formally separate from [0par-
ticipant], and they seem to bear out many of the generally ac-
cepted statements on markedness and the tendency of forms more
marked in one respect to make fewer distinctions in other respects.
There are clearly two pronominal stems, one built on (2a)n-,
which serves to unite all and only the [-{-participant] and another
built on /)-, which signals all and only [Oparticipant]. Participant
is, thus, the crucial opposition of the pronouns, with gender and
number not represented at the submorphemic level of the stem.
As for number signalization, it is well known that, given the het-
erogeneous character of the members of a "we" group, languages
signal plurality in the 1st person most often by lexical suppletion
or near-suppletion, while second and third persons display either

In this section I will have occasion to refer to the Arabic sound system.
The two phonological distinctive features of the three-vowel system will be
relevant for the following discussion:
fdiffusei fflat

Length is phonemic. For discussion of the Arabic sound system see Beeston
(l»70: 16-19). The choice of [-(-/-flat] over [acute/grave] is connected with
the role of flatness in the system of velarized consonants, on which see FVad-
Idn (1985: 41-45). [Acute/grave] does, nonetheless, figure in the consonant
system, e.g., / n / vs. /m/.
641

lexical or inflectional plurality (cf. Moravcsik 1978: 234, recalling


Forchheimer 1953). Arabic is no exception. The Arabic 1st per-
son pronouns ianaa, nafjnu are partially suppletive and partially
inflectional. They do have in common with each other and the
whole [-{-participant] pronoun set the phoneme / n / . The phoneme
/fi/ in the plural pronoun, however, could suggest a different lexi-
cal stem or compound stem, since the pharyngeal consonant series
occurs only in root morphemes and does not play a role in the
affixal repertoire of Arabic.^^
All and only the 2nd persons add t- to the base 2an. Within
the singulars, the [-|-/0fem.] opposition correlates with the [ + / -
high] tonality of the final segment. The vowels employ the phono-
logical feature [+/- diffuse]: masc. 2anta ending in a low-tone
vowel; fem. 2anti ending in a high-tone vowel. The plurals con-
tinue this tonality opposition but with the consonantal feature
[-|-/-acute]: 2nd masc. pi. 2antum in non-acute (low-tone) /m/ vs.
2nd fem. pL 2antunna with acute (high-tone) / n / , cf. fn. 33 below.
The 3rd person stem h- also reflects gender opposition in tonal-
ity. Without going into the exact morphological processes, the
singulars use a skeleton h—a and insert a high-vowel-plus-glide
distinguished by the feature [-{-/- flat]: 3rd masc. sg. huwa with
low-tonality (rounded vowel and labial glide) vs. 3rd fem. sg. hiya
(unrounded vowel and palatal glide) with high-tonality.^^ The

^^ The lack of the initial 2a- syllable is not by itself ground for dismissing
from the inflectional system. Hebrew avoids this ambiguity with the 1st sg./pl.
pair 2an(/2and(inu. Thus Hebrew unites all its [-(-participant] pronouns under
the stem 7a- (better a- for the modern language) with nasal / n / in [-f-central]
lsi ant, andfinu opposing non-nasal / t / in [0central] 2nd atd, at, atim, atin.
I am not claiming that the phonological features of high and low tone are
themsdves somehow "marked" and "unmarked" respectively although just
such evidence might fuel an argument over the comparative complexity of
sonnd segments. It is, nonetheless, worthy of note that the affixes of a co-
herent semantic group so consistently present the same kind of phonological
compodtion. I stress that this is simple diagrammatic iconidty of semanti-
cally similar forms bearing similar phonological stamps. Any hint of organic
overtonea of 'high tone/low tone' having even the remotest connection with
female/male, light/heavy, or weak/strong, as a colleague of mine once half
•erioiuly mused, would be totally irresponsible.
642

plurals employ the same tonality-opposed nasals as the 2nd per-


sons: 3rd masc. pi. hum vs. 3rd fem. pi. hunna.^ The dual
pronouns are unmarked for gender and are formed with the ag-
glutinating suffix -aa on the plural pronoun also unmarked for
gender, namely, the [-^plural 0fem] forms 2antum, hum, giving
2antumaa, human, despite the fact that in the rest of the nominal
system the dual is formed from the sg. of the noun, regardless of
whether it has a "broken" (=internal) plural, e.g., sg. kitaab, du.
kitaabaani, but pi. kutub. (See Fontinoy 1969 for the evolution of
the dual from both plural and singular stems.)
We have, thus, a split paradigm with interesting seman-
tic implications. The pronominal stems correlate with a parti-
cipant/non-participant distinction, while the inflectional remain-
ders of the forms correlate with the central/non-central distinc-
tion. The form marked [-{-participant], 2anta, can add up to two
other features, be they both non-shifters — [-|-feminine] for 2anti;
[-l-plural] for 2antum; both [+fem. -|-pl.] for 2antunna, and [-|-pl.
-|-min.] for iantumaa — or one shifter plus one non-shifter: [-t-cen-
tral] for 2anaa, which can add only [-l-pl.] for nahnu.
U the central participant is considered to bear two semantic
marks [-{-participant -{-central], and the dual number also bears
two marks [-{-plural -{-miminal], then the lack of a 1st du. sug-
gests that in the Arabic system of features no one form can bear
more than one such "double mark" in addition to a single mark,
whether [-{-participation] or [-{-feminine]. If gender were to enter
into the dual at all, according to this hierarchy, it would be in the

Rice and IVager suggested (1954) that feminine -tunna and hunna are an-
alyzable as *tum-(-na, *hum-hna with regressive assimilation of the nasals.
This analysis does, indeed, simplify the description of the afRxal morphol-
ogy by not positing an abstract '-tun, -hun + na, but there is no other
synchronic Arabic evidence for a regressive assimilation of /mn/ to /nn/.
Even historically and comparatively the opposition -m/-n signalled the gen-
der opposition in the plural, as it does in Hebrew -ttm, -ten, where the nasal
consonants carry the entire burden of distinction. In South Arabian and those
Arabic dialects that have gender in the plural, e.g.. Gulf Arabic, this tonality
oppontion in the consonant is matched in the vowel, as well, cf. South Arar
biu/Gulf perfect: 2nd masc. pi. .kumf-tu, 2nd fem. pi. -il;en/-(in. (Note also
Gulf Arabic 2nd fem. sg.-pl. -ti, -tin, where the -n can be seen as « pluraliier
of the feminine. The cognates of Arabic 1st and 2nd masc. tg. perfecta •(«, -to
in Gulf are both •(, that is, [-f-participant 0central]. See Fradkin forthcoming
a. for elaboration.)
643

3rd person, namely, [0partidpant]. The 3rd person of the verb


does, indeed, distinguish gender, and the 3rd du. pronoun does
not. I have no concrete explanation for this, but it merely points
up the different role of the minimality feature in the verbal and
nominal subsystems. Thus the 2nd and 3rd person pronouns show
the greatest number of feature combinations and the 1st persons,
thfi smallest.
The suffix system of the perfect tense opposes, basically,
consonant-initial suffixes — specifically a dental — to vowel-initial
ones, correlating with the opposition [+/Q participant]. (The ob-
vious exception is the consonant-initial 3rd fem. pi. -na, which
I treat just below.) The prefixes fall into the following feature
scheme:

0min +min
r0pii r+Pii [+pl +niin]

[+participant+central] 1st. -tu •naa

[+participant 0centra]] Im.-i. -ta,-ti •tum,-iunna -tumaa

[0participanl 0central] 3m.-f. -a,-at -uu,-no


Figure 15: Perfect Suffix System

Even if some of these suffixes were of pronominal origin in Proto-


Semitic, it is still interesting that synchronic Arabic formal struc-
ture lines up by semantic features. The 1st sg. -tu is clearly not
from the pronoun 2anaa, but rather, displays its affinity with the
other [-^participant] pronouns by means of the initial oral den-
tal. The traditional philological view is that the 1st sg. -t is by
analogy with the 2nd persons. This feature scheme now gives this
intuition a principled basis. It is also worthy of note that the oral
dental opposes a nasal dental in the suffixes -naa, namely, [-Hpar-
ticipant +central -l-plviral] (the most marked of the participants)
and -na, namely, [0prt -)-feminine -fplural] (the most marked of
the non-participants). These are the forms with complex mark-
ings, one of which is plurality.
The two suffixes -aa and -na are perhaps better termed post-
fixes. They are singled out in this system as being agglutinative
on forms not marked for centrality. The dual -aa, itself, car-
ries no gender specification. In the [-^-participant] category it
644

affixes to the unmarked gender of the marked number, namely,


[-{-participant -l-plural 0fem] -tum+aa —» -tumaa. In the bon-
participants it affixes to the unmarked number of the gender-
opposed 3rd masc. -a and 3rd fem. -at, giving -aa, -ataa. The
feminine plural -na always signals both non-shifters and affixes
directly to the stem of the perfect, namely, qatdlna, and to a
theoretical 2nd fem. pi. *-tun, cf. note 33 above. Furthermore,
these two suffixes carry no person marking of their own and ap-
pear to "travel" between the perfect and imperfect. (It is even
questionable whether the 3rd masc. pi. perfect -uu in qdttduu is
the same as in 3rd masc. pi. imperfect yaqtuluu/n(a)?^) What
is important for the present investigation is that -na and -aa
are both marked [-|-plural] and, in addition, [-|-fem.] in the first
case and [-|-min.] in the second and that they have in common
the reinforcement of the [-|-/0central] opposition: a form with
gender or dual reference is automatically not speaker reference.
This strengthens the assertion made above, p. 642, that the two
pronominal stems and the structure of the perfect suffixes signal
[-}-/0 participant], while the inflectional system concentrates on
[-H/Ocentral].
The prefix system of the imperfect is not so clear-cut. Of
the four prefixes 2-, n-, t-, y-, only the first two are unambiguous
for person and number. Both are marked [-|-participant -l-central]
and distinguish further [-|-/0plural] only. They combine only with
the four mood suffixes, -u, -a, - 0 , -an(na). The other two pre-
fixes signal, by themselves, neither gender nor number, neither do
they neatly fit 2nd vs. 3rd person, as the pronouns and perfect
suffixes do. We can say with certainty that y- is always [0partic-
ipant 0centra]], but leaves gender specification for the suffix, cf.
3rd masc. pi. yaqtuluuna, 3rd fem. pi. yaqtulna, cf. Fig. 4 above.
The only referential constants of the t- prefix are [0central] to-
gether with either [-{-participant] or [-|-fem.]. Any suffix that adds
gender or one number feature fixes the prefix as [-(-participant],
namely, 2 fem. sg. taqtuliina, 2 masc. pi. iaqtuluuna, 2 fem. pi.

The interrelations of the nominal and verbal paradigms have long been a
question for historians of Semitic. The imperfect appears to be nominal not
only in the resemblance of the mood endings to the case endings, but also in
the matter of these gender-number suffixes.
645

taqtulna, cf. Figure 2 above. The dual taqtuldani behaves like the
ungular tdqtulu with respect to [Ocentral], using the same prefix
for 3rd fem. sg. as for 3rd fem. du., although the change of prefix
8ignals the change from 3 fem. sg. to 3 fem. pi. idqtulu/yaqtulna.
The only reason I can offer for this at present is that the dual is
also quintessentiaUy [Ocentral], as discussed just above.^^
These prefixes, then, slight though their current resemblance
may be to the pronouns, nonetheless replicate the [-{-/0central]
distinction suggested by the pronominal structure of lexical 1st
person vs. inflectional 2nd-3rd:
Imperfect Affix System:
Prefix-SufHx Combinations and Complementarity of <jender-Number Features

Prefixes Mood Suffixei


All Prefixes + Mood Only Sbjn Juss Ind Enrg
J0pl]
[-•prt -t-cnt 0fem])—-— Mood-only
l+P'J suffixes =
-a -0 -u -an(na)

i
[-»-prt0cnt0fem] [0pl 0fem]
[01>rt 0cnt -t-fem]
[0prt0cnt0fem]
[0Cnt] Prefixes Only y./ Mood plus other
feature-
[-•prt 0cnt 0fem] — ^ -ii -iina -in(na) [0pl-t-fem] - 2 fsg.
t-
[0prt 0cnt -t-fem] ^^uu -uuna -un(na) [-t-pl Ofem] - 2-3mpl.
[0prt 0cnt 0fem] - 2du-3fdu.
^ ^ -aa -aani -aaiuii [-t-pl -t-min] y- - jmdu.
y- ^-na -naanm [-•-pl -•fem] » 2-3fpl.

Figure 16: Imperfect Prefix-SufTix System

This Is precisely one of the points of typological divergence across Semitic.


Hebrew and some Arabic dialects do generalize (- for the fem. pi., as well as
the sg., cf. Biblical Hebrew, surviving in the very highest registers of the con-
temporary literary language, 2nd masc. sg./pl. tiqtdl/tiqtelu, 2nd fem. sg./pl.
tigUli/tigtdlna, also 3rd fem. sg./pl. tiqtdl/tiqtHna. See Fradkin forthcoming
ft. for details.
646

This prefix pattern, except for the gender problem, recalls the
minimal four-person system Ingram (1978: 227) notes for Ko-
rean, cf. also Gray's remark (note 19 above) about the hypothe-
sized four-pronoun system of Proto-Semitic, The spoken dialects
of Arabic from Egypt to Iraq have the same four-prefix system.
Many of the dialects west of Egypt (the bulk of North Africa,
including Maltese) have only a three-prefix system n-, U, y-, with
plurality in the 1st person signalled infiectionally exactly as in the
other persons (and the same morphophonenuc alternation of the
stem kteb-^-zero-siiffix/ketb+voca^c suffix). Compare the stan-
dard Arabic and Moroccan imperfects:

3 m. sg.-pl. 2 m. sg.-pl. 1st sg.-pl. 2 m./3 f.-3 m.


Arabic ydktubu tdkaibu Idktubu tdknAu
yaktubHuna takmbiuna ndktubu ydktubu
Monxxan yikteb tikteb nikteb Ukteb
(no fem. pi.) ykitbu tkitbu nkitbu yikteb

Figure 17: Arabic vs. Moroccan Imperfect

Czapkiewicz (1971) calls this the accumulation of functions (that


is, of both singular and plural) in the n- prefix. In terms of the
feature system I am using here the features in the dialects are
polarized into prefixal and suffixal territories. The prefix system
handles [-|-/0participant], with the same participant-gender split
in the t- prefix, but the [+/0plural] feature is strictly suffixal.
(Many dialects maintain the imperfect -« suffix in, e.g., 2 fem. sg.
tketbi, thus gender still vascillates between prefix and suffix.)

4 . CONCLUSIONS

The Arabic data I have examined here bear out the pre-
dictions of a bottom-up type markedness scheme. The perfect
and imperfect use different affixal means to arrive at nearly the
same sets of PGN feature expressions: both verb sets have gen-
dered 3rd du., non-gendered 2nd du., and no 1st du. or fem.
The main point of divergence between the perfect and imper-
fect systems is that the formal side of the perfect suffixes rein-
forces the [-|-/0participant] opposition. The formal side of the
647

imperfect prefixes reinforces [-|-/0central]. The [-j-central] pre-


fixes are the only ones that carry inherent number, but then only
singly marked number (plural) and not doubly marked number
(dual). The t- prefix, then, emerges as the [0central-plus-other-
information] affix par excellence. The other information is either
a shifter [-f participant] or a non-shifter [+feminine]. The rest of
the forms that mark person in Arabic do not otherwise violate
the participant/non-participant line, but the additional features
of [-{-feminine] and [-{-minimal] do. Here in this t- prefix is the one
ostensibly person-marked form that signals, without further suf-
fixal elaboration, "lack of specification of centrality." It is either
2nd masc. sg., namely, added participation [-{-participant 0central
0feminine] or 3rd fem. sg., namely, added gender [0participant
0central -{-feminine].
The mood system of the imperfect also shows a kind of
bottom-up typology. The unmarked mood makes the maximum
distinctions. The two singly-marked moods make each the same
number but merge part of their paradigms, by comparison with
the indicative, and it is no accident that the forms in common
are, themselves, the semantically marked ones. All three mood
paradigms overlap just where the forms are otherwise heavily
laden with other information. The energetic recapitulates the
process, beginning over again with complex forms and limiting
their propagation to forms with less gender and number informa-
tion.
But is this the norm for verbal systems cross-linguistically?
Other studies, including Greenberg (1988), remind us that in
many languages the 1st person is the favored category for duals
(and especially indusives, even if in the plural only, in a system
with no dual, cf. Greenberg 1988: 1). The 2nd person is disfa-
vored for duals cross-linguistically. Why should the most marked
1st person — assuming that the 1st, and not the 2nd, is the most
marked — make the greatest number of distinctions?
It is generally agreed, as Ingram's resurvey of Forchheimer's
1953 data (1978: 242-246) and Greenberg's (1988) study show, if
a pronoun system has one plural form it is the 1st, and if it has
only one dual form it is also the 1st, as in Carrier. If all three
persons have a plural, then the 1st and 2nd can have a dual,
but the 3rd need not, as in West Greenlandic. (Old Slavic has
648

a symmetrical number pattern, where all persons have a plural


and dual, with the bottom-up exception that the three genders
in the 3rd person reduce to masc. vs. fem./neut. in the dual.)
Where there is no dual, the 1st pi. can be split into inclusive-
exclusive, as in Tagalog, Tamil, or Algonquian. If in addition to
such a system there is a dual, it is 1st inclusive, as in Southern
Paiute. Of course, the inclusive/exclusive can split both the plural
and the dual without involving the 3rd person in either category,
as in Maya (cf. also Silverstein 1986: 342).^^ This is a sort of
"top-down" pattern where the most marked forms also have the
most formal distinctions, and the less marked forms are sparser.
The same pattern characterizes, for example, French conjugation,
where the forms for nous and vous are the most distinct in the
paradigm and remain distinct, even when all the rest merge.'*

'^ Though Ingram's study is concerned with person and number categories and
not gender, it is also interesting that in Slovenian the dual pronoun, as-
sumedly the semantically most complex, allows a masculine/feminine gender
distinction that the [-f-participant] categories in Slavic cannot otherwise ac-
commodate, viz., lst/2nd pi. mi/vi, 1st masc./fem. du. midva, midve, 2nd
masc./fem. du. vtdtia, vtdve. Of course, this has to do with the compound na-
ture of that dual pronoun: mi+masc. numeral dva 'two' and mi-|-fem.-neut.
numeral dve 'two'. There is no gender agreement in the conjugated verb.
Upper Lusatian brings into its 2nd-3rd dual verb conjugation a masculine
personal distinction, -taj, vs. all else, -tej. Lower Lusatian has only -tej.
In fYench intra-paradigm conjugation the person relations show precisely that
the semantically most-marked forms — the 1st and 2nd plural forms, nous,
vout, marked [-(-participant -t-plural] — never merge with the other forms of
the paradigm. These two forms are distinct from each other and from every
other form of the paradigm in every tense, even when two, three, or all of
the other forms fall together, at least in contexts where no elision is involved
(as I mentioned in Section 1 above): nous regardant, vou$ regardez vs. je-tv-
il/elle-ilt/elle» regard-e-ea-ent, that is, -6, -e vs. - 0 . In fact, it is possible
to demonstrate an implicational hierarchy for French conjugation based on
semantic features. (I assume this has been noticed in the Romance linguistic
literature, but I have not specifically sought out any discussion of it.) If any
other form of the paradigm is distinct, number is distinguished over person,
thus, it will be the 3rd pi. against a single form for the entire singular, as in
the finir type verbs: 1-2-3 pi. finiso, fini»e, finis vs. sg. fini, that is, VC stem
+ -0, -e, - 0 vs. V-final stem. (No doubt other schools of markedness term
this a "marked" (=longer, more informative) stem finis+ vs. an "unmarked"
stem fini+0.) Then the person hierarchy reasserts itself within the singular,
such that if any other form is distinct it is the 1st sg., as in the future and
passe simple, while the 2nd and 3rd sg. are always identical. (I leave aside the
forais of the future that seem to overturn this analysis, namely, homophonic
nou$ regardenns and Us regarderont and those dialects where je finirtu is
649

What this really shows, however, is not that classical marked-


ness theory falls apart in the realm of conjugation categories.
Since the singular/plural split is already in place universally in the
1st person, languages can choose it as the jumping-off point for
further semantic elaborations. They merely elaborate an existing
distinction. That this Arabic "bottom-up" system of conjugation
suppresses this tendency, then, seems to be the minority strategy
among PGN systems but not other grammatical paradigms .^^
Semantic markedness, then, means comparative conceptucd
complexity of the signatum. The physical signans can be a par-
tial indicator of semantic similarity. In the present context, all
the Arabic pronouns and verbal affixes call into question simul-
taneously the presence of person, gender, and number, and the
form that is unmarked for one referential parameter allows the co-
expression of some other parameters, while a form that is more
highly marked tends not to accumulate too many other marks.

homophonouB with vous finirez.)


Calling on French again, however, we see that the case-marking system for
pronouns is also a sort of "bottom-up" type in terms of number of case
distinction relative to semantic complexity of the pronoun. (Again this is not
the same situation as Arabic, because among other factors, gender in the
French system is strictly a [0participant] feature.) As the chart below shows,
the plural participant forms are invariable for case marking, although every
other pronoun has at least three forms that divide up the traditional four
syntactic functions of nominative (N), accusative (A), dative (D), and the
non-verb-governed "disjunctive" (J). Specifically, the non-plural participants
always distinguish N formally from the rest, but A and D are always the same
form and may also be the s&me as J: je (N) - me (A, D) - mot (A, D, J); tu (N)
- te (A, D) - tot (A, D, J), where me - (e occur in (mostly) pre-verbal position
and moi-toi occur exclusively in post-verbal position, coinciding roughly with
unstressed and stressed positions, respectively. The non-participants are like
the participants in keeping N and A formally distinct: sg. i7, elle (N) - le, la
(A); pi. ih, ellti (N) - le$ (A). In addition they also keep A and D distinct but
at the expense of the gender distinction otherwise maintained in the singular:
sg. le, la (A) - lui (D); pi. le$ (A) - leur (D). The two genders then behave
differently. The marked gender merges N and J in both singular and plural:
elU, ellti (N, J). The unmarked gender i7 merges D and J in the singular lui
but keeps them apart in the plural leur (D) - eux (J). Thus, N and J are
always gender-marked; A is gender-marked only in the singular (le, la vs. lei),
and D ia never gender-marked (lux, (eur). The plural participants, however,
are noui, vout across the board. That is, it seems to be that the greatest
differentiation must be between N and A. If those two are not distinct, the
re>t of the paradigm is, by implication, also the same.
650

Questions of statistical frequency, defectivation, or morphologi-


cal variation are secondary and not necessarily diagnostic of the
marked or unmarked character of the members of an opposition.
This binary semantic analysis brings to the fore the semantic
incompatibility in Arabic of dual (marked plural) and gender dis-
tinction in combination with centrality, that is, 1st person. Con-
versely, a form expressing gender or dual must be non-central, just
as forms marked as central cannot distinguish gender or dual. The
system-internal motivation for this is to be found in the feature
matrix itself. Interpreting [-{-central] as doubly marked person
and dual [-{-pHmin] as doubly marked number allows the fur-
ther clarification of that generalizaton: Arabic verbal affixation
exhibits a morpho-semantic prohibition on the co-occurrence of
double marks. A form can be marked for a maximum of three
of these five semantic features. Greater cumulations of feature
marks do not occur, namely *[+ central + participant + fem +
minimal + plural] viz., 1st fem. du. or even without minimality,
1st. fem. pi. In the mood system of the imperfect, given that the
energetic is always a distinct form, the two- and three-featured
prefixes 2- [-{-participant -{-central] and n- [-{-participant -{-central
-{-plural] permit no further gender or number features in the suffix.
The one-featured and unmarked prefixes t- [0central -{-"other"]
and y- [0participant 0central] with no suffixal gender specifica-
tion distinguishes the other three moods: indicative, subjunctive.

Non-Paiticipant Participant
f.s. m.s. f.pl. m. pi. 1 s. 2 s. lpl. 2 pi.

Ace. la le les me te

Dat

disJ.
lui leur

eux
1'
moi
i

toi
nous vous

eUe elles
Nom il ils je tu

Suggested French Case Marking Scheme


651

ydqtul. If the suffix contains gender or number, or no gender but


dual number, the same suffix permits only two moods: indicative
taqtuliina, taqtuluuna, taqtuldani, subjunctive/jussive taqtulii,
taqtuluu, taqtulaa. The suffix with both gender and number per-
mits one mood, e.g., taqtulna. Of the energetic we may also recall
that the same highly-marked suffixes that reduce the other three
moods to two also reduce the two energetic varieties to one, e.g.,
taqtuldanni, taqtulndanni.
I have shown that there is semantic subtlety behind mor-
phological systems thought to be defective. The forms point the
way to an intricate set of conceptual features which interact in
consistent, if not always predictable, ways.

Address of the author: Robert A. Fradkin


Dept. of Foreign Languages
Old Dominion University
Norfolk, Virginia 23529
USA
652

APPENDIX

Table 1: Arabic Pronouns and Perfect Stem QATAL+ suffixes

SINO. PLUR. DU.


3MSC. hiiwa qdtal-a hum qdtal-uu qitd-aa
hiimaa
3FEM. htya qdwl-at hunna qcudl-na qauS-aiaa

2MSC. idn-ta qdtal-ta Mn-nim qatdl-tum


idn-tumaa qadl-tumaa
2FEM. qeM-a 2an-Binna qatal-tuma

1 Tdn-aa qaidl-tu ndlinu qatdl-naa

Remarks on sound structure and semantic content of suffixes:


a. Semantic feature [-f-participant] vs. [0participant] correlates with
(1) consonant-initial vs. vocalic-initial verb sufRx
(except -na, marked gender, marked number)
(2) the two pronoun stems {2a)n- vs. h-.
b. 3rd Persons: Vocalic-initial suffixes. Long vs. short vowel correlates with
[+pl] vS' [0pl]- Unmarked suffix (masc. sg.) is mono-phonematic. Addi-
tion of any other semantic feature matched by at least bi-phonematic or
prosodically marked (bi-moraic) vowel: pi. -uu, dual -aa. Only [-l-fem]
suffixes contain consonant: -at, -na.
c. 2nd Persons: All begin in (•. [0pl] add a short vowel: -ta, -ti. Addition of
phonological feature to vowel correlates with additional semantic fea-
ture: [0fem] is [-diff -flat]; [+fem] is [-l-diff -flat]. Non-central plurals
are single closed syllable in non-fem. -turn, closed syllable plus open in
fem. -tunna, short open plus long open in dual -tumaa.
d. 1st Persons: Also consonant-initial, opposing nasal for [-l-pl] to non-nasal
for [0pl], as well as long vowel vs. short: -naa vs. -tu. Compared to
2nd persons, short vowel in [-(-cnt] -tu is both [-(-diff -(-flat].
653

Table 2: Imperfect Active and the Moods:


Prefix 4€onnector aJu-t- stem Q T U L + Suffixes
Indicative Subjunctive Jussive Energetic

Short -V or Closed SyU. Suffixes

1 sg. 2-d-qtul-u J-d-qtut-a l-d-qtid 2-d-qtul-an(na)


1 pi. n-d-qtul-u n-d-qtul-a n-d-qtul n-d-qnd-an(na)
2 m. sg.

^ t-d-qtul-u t-d-qtul-a t-d-qtul t-d-qtul-at^na)


3f. sg.^

3 m. sg. y-d-qtul-u y-d-qtul-a y-d-qtul y-d-qiul-an(na)

Long-V or Closed Syll. Suffixes


2 f. sg. t-a-qmt-(in(a) t-a-qtul-ii t-a-qtul-(n(nn)
2 m. pi. t-a-qtul-liunfa) t-a-qtul-uu t-a-qtul-iin(na)
3 m. pi. y-a-qtul-liun(a) y-a-qtul-uu y-a-qtul-iinlna)
2du.

t-a-qtul-aa t-a-qad-danni

3 m. du. y-a-qtul-dan(i) y-a-qtul-aa y-a-qad-danni

Cons.-lnit. Suffixes
2 f. pi. t-a-qttil-na t-a-qtul-ndanni
3 f. pi. y-a-qt&l-na y-a-qtul-ndanni

Imperative = Jussive and Energetic with zero-prefix: (2u)qtul, (2u)qtulan(na)


etc. (Prothetic syllable with diffuse vowel in utterance-initial context, with
anticipatory [+H»,i] before [-»-flat] stem vowel.)
Remarks on semantic content and affix structure:
1. Voice is marked in stem, active -qtul- vs. passive 'qtal-, and does not affect
external inflection. Connector vowel provides partial information on voice:
active either -a- (I, V-X) or -u- (II-IV); passive, only -u-. Same inflection.
654

e.g., 3rd sg., pi. indie, ydqtalu, yuqtaluuna; subj. yuqtala, yuqtaluu, etc. Short
auslaut vowel of indicative -uun(a), -iin(a), •aan(i) deleted before pause.
2. Close resemblance between pronouns and some prefixes, but much less
than in perfect: 2- in 1st sg., n- in 1st pi., (- in 2nd pers. and also 3rd fem.
sg., but y- in other 3rd pers. not otherwise present in inflection system.
3. Both subjunctive and jussive are semantically marked vs. indicative.
Semantically complex forms are shared across these two paradigms and merge
gender/number with mood: long-vowel suffixes -ti (2nd fem. sg.), -uu (masc.
pi.), and -aa (dual) are common between these two moods. AU three moods
share the highly marked fem. plur. suffix -na.
4. The energetic, semantically the most complex mood, is always formally
distinct from the other moods. Further, its two varieties do not differ syntac-
tically but might stylistically or possibly semantically: "heavy" with final -na,
'lightened" without it (suggesting that the truncation is a secondary process)
reduce to one variant with unusual syllable structure, (C)VVCCV, in the se-
mantically more complex forms, namely dual -aanni, fem. plur. -naanni.
655

Table 3: Outline of Arabic Verb Stem Patterns

Native Arab traditional citation form is 3 m. sg. perfect. Semiticists in the west

number whole paradigms by Roman numerals, though that is not the most

instructive ordering. All perfects use suffix set in Table 1. All imperfects use the

prefix-suffix material in Table 2. Derivation is by prosodic lengthening (consonant

gemination: n, K ; vowel lengthening: HI; both: XI), affixation (prefix: IV ?a;

V n n-, X sta-i infix -r-: V m ) . or both length and affix (V. VI. XH-XV).

Perfect Stems Imperfect, 3rd masc. sg.

Active Passive Active Passive

[-diffuse vowels] [-Kliffuse vowels] -a/u- connector -u- connector

I. a., b. qauil; gatil- qutil- yaqtul-,yaqtal-,yaqtU- yuqtal

c. qaml- * ** yaqtul- * »»

II qaaal- qimil- yuqanil- yuqaxtol-

m qaatal- quiail- yuqaatil- yuquutal-

IV iaqtal- ?uqtil- yuqtil- ytiqtal-

V taqaltal- tuqunil- yntn/jnnnl. yutaqottdl-

VI fCKjfifitnl- tuquutil- yataqaotal- yuiooaoud'

vn (?i)nqaul- (iu)nqutil- yanqarU- yunqeual-

vm (fi)q-t-atal- (fu)q-t-unl- yaq-t-aul- yuq-t-mcd-

IX (ii)qtdl- * ** yaqtall- *• *

X (fi)staqtal- (Ju)staqtH- yasiaqtil- yustaqtal-

XI (fi)qiaall- ** • yaqiaall- • * •

Rare types. No complete paradigm attested. Usually cited in perfect active only.

XII (fi)qtinvtal- XIII (fi)qtawwal- XIV. (fi)qiardal- XV. (ti)quudaa-


656
Table 4: Feanire Combinations Given Formal Expresnon

A. 18 Potential B. 12 Personal Pronouns

Sing. Plur. Dual Sing. Plur. Dual

3 msc. sg. 3 msc. pi. 3 msc. du. 3 msc. sg. 3 msc. pi.
3du
3 fem. sg. 3 fem. pi. 3 msc. du. 3 fem sg. 3 fem. pi.

2 msc. sg. 2 RISC, pi. 2 msc. du. 2 msc. sg. 2 msc. pi.
2du
2 fem. sg. 2 fem, pi. *2 fem. du. 2 fem. se. 2 fem. pi.

1 msc, sg. 1 msc, pi. *1 msc. du.


lsg. lpl.
•1 fem. sg. •1 fem, pi. • 1 fem. du.

Hypothetical system with all features Gender excluded from dual. Gender and dual
combining symmetrically, (See B.) incompatible with 1st person.

Sing. Plur. Dual Sing. Dual Plur.

3 msc, sg. 3 msc Pl- 3 msc, du. 3 msc. sg. 3 msc. du. 3 msc, pl,

3 fem, sg. 3 fem. Pl- 3 fem, du. 3 fem, sg.= 3 fem. du.= 3 fem. pl.

2 msc, sg. 2 msc, Pl 2 msc. sg. 2du, 2 msc, pl,


2du,
2 fem, sg. 2 fem. Pl- 2 fem. sg. 2 fem, pl.

lsg. lpl. ... lpl.

C. 13 Perfect Tense Suffixes D. 11 Imperfect Tense Prefix-Suffix

Sg-PI-Du order show receding dual and gender Sg-Du-PI order shows formal affinities:
the higher the person marking.
Formal coincidence of 2nd masc-3id fem,
in sing, and dual.
Both tenses have in con:unon:

3rd person: gender in plural and dual of verb but not dual pronoun

2nd person: gender in plural but not dual *

1 st person: no gender and no dual


657

Table S: Semantic Composition of Affixes

I. Perfect
Person Gender Number
Central Panicipant Feniinine Minimal Plural

-a 0 0 3 0 0
-at 0 0 + 0 0
•uu 0 0 3 0 +
-na 0 0 t- 0 +
0 0 3 + +
0 0 + + +
-ta 0 + 3 0 0
-ti 0 + t- 0 0
-tiun 0 + 3 0 +
•tunna 0 + t- 0 +
-tumaa 0 + 3 + +
-tu + + 3 0 0

-naa + + Zt 0 +

II. Imperfect
Person Gender Number
(Indicative only)
Central Panicipant Feminine Minimal Plural

y- 0 0 0 0 0 [unmarked prefix]

t- 0 0 0
:;>
7. + + 0 0 0
n- + + 0 0 +
-u 0 0 0 0 0 [unmarked suffix]
-uuna 0 0 0 0 +
-iina 0 (+) + 0 0
-aani 0 0 0 + +
-na 0 0 + 0 +

No affix can carry more than 3 marks. No co-ofturence of doubly marked


categories. If prcflx carries centrality, it also carries number opposition and takes
only unmarked mood suffix. PreHxes without centrtUty require gender and number
in suffix.
658

REFERENCES

Adler, Rutie. 1986. Deixis and Anaphora in Hebrew. Unpublished MA The-


sis. Linguistics Dept. University of California, Berkeley.
Anderson, Stephen. 1985. "Inflectional Morphology." In Shopen, ed.. III.,
150-200.
Anderson, Stephen. 1989. A-Morphous Morphology, mss.
Anderson, Stephen and Edward Keenan. 1985. "Deixis.'' In Shopen, ed.,
III., 259-308.
Andrews, Edna. 1984. A Theoretical Foundation for Markedness: Asymme-
try in Language from a Mathematical Perspective. Indiana University
Ph.D. Dissertation, Ann Arbor: University Microfilms #8417151.
Andrews, Exlna. 1989. Markedness Theory: A Union of Asymmetry and Se-
meiosis in Language. Duke University Press.
Aoun, Youssef. 1979. "Parts of Speech: A Case of Reconsideration." In: Bel-
letti, ed., 3-24.
Barlow, Michael and Charles Ferguson, eds. 1988. Agreement in Natural Lan-
guage. Stanford: Center for the Study of Language and Information.
Beeston, A.F.L. The Arabic Language Today. London: Hutchinson. 1970.
Belleti, A., et al. eds. 1981. Theory of Markedness in Generative Grammar.
Proceedings of the 1979 GLOW Conference. Pisa: Scuola Normale
Superiore di Pisa.
Benveniste, Emile. 1946. 'The Relations of Person in the Verb." In:
Problemes de Linguistique Generate. Reprinted in: Mary Elizabeth
Meek, trans., 1971. Problems In General Linguistics. Miami: Univer-
sity of Miami Press. (Cited pagination according to 1971.)
Buchler, Ira and R. Freeze. 1966. "The Distinctive Features of Pronominal
Systems." Anthropological Linguistics 8-8, 78-105.
Bybee, Joan, L. 1985. Morphology: A Study of the Relation Between Meaning
and Form. Amsterdam: John Benjamins.
Cantarino, Vicente. 1974-1975. The Syntax of Modern Arabic Prose, I (1974);
Il-Ili (1975). Bloomington: Indiana University Press.
Cantineau, Jean. 1952. "Les oppositions significatives.'' Cahiers Ferdinand de
Saussure 10, 11-40.
Carstairs, Andrew. 1987. Allmorphy in Inflection. London: Croom Helm.
Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New
York: Harcourt-Brace.
Comrie, Bernard. 1984. "Agreement as a Research Tool." ESCOL 1, 13-26.
Corbett, Greville. 1983. Hierarchies, Targets, and Controllers: Agreement
Patterns in Slavic. Pennsylvania State University Press.
Czapkiewicz, Andrzej. 1971. "The Accumulation of Functions in the Catego-
rial Morphemes of the Verb in Modern Spoken Arabic." Folia Orientalia
XIII, 15-32.
659

Dressier, Wolfgang, ed., 1987. Leitmotifs in Natural Morphology. Amster-


dam: John Benjamins.
Eckman, Fred, Edith A. Moravcsik, and Jessica R. Wirth, eds. 1983. Marked-
ness. New York and London: Plenum Press.
Fassi Fehri, Abdelkader. 1982. Linguistique arabe: forme et interpretation.
Rabat: Universite Mohammed V, Faculte de Lettres et Sciences Hu-
maines.
Fassi Fehri, Abdelkader. 1984. "Agreement in Arabic, Binding and Coher-
ence." In Barlow and Ferguson, eds,, 1988, 107-158,
Fleisch, Henri. 1961. Traite de philologie arabe, I. Beirut: Imprimerie
Catholique.
Fleisch, Henri. 1979, TYaite de philologie arabe, II, Beirut: Imprimerie
Catholique.
Fleischman, Suzanne, 1985, "Discourse Functions of Tense-Aspect Opposi-
tions in Narrative: Toward a Theory of Grounding." Linguistics 23-
6, 851-882.
Foley, William and Robert van Valin. 1984. Functional Syntax and Universal
Grammar. Cambridge: Cambridge University Press.
Foley, William and Robert van Valin, 1985. "Information Packaging in the
Clause," In Shopen, ed., I, 282-364,
Fontinoy, Charles, 1969, Le Duel dans les langues semitiques, Paris: Belles
Lettres.
Forchheimer, Paul. 1953. The Category of Person in Language, Berlin,
Fradkin, Robert, 1985, Markedness Theory and the Verb Systems of Rus-
sian and Arabic: Aspect, Tense, and Mood, Indiana University Ph,D,
Dissertation, Ann Arbor: University Microfilms #8527003,
Fradkin, Robert, 1990, Jerzy Kurylowicz's Contribution to the Study of Tense
and Aspect in Semitic, Paper presented at NACAL, Atlanta.
Fradkin, Robert, 1991, Stalking the Wild Verb Phrase: An Adventure into
English Grammar for Speakers of English Learning Other Languages.
University Press of America,
Fradkin, Robert, forthcoming a. "Typologies of Person Categories in Slavic,
Semitic, and Elsewhere." In: Andrews, E. and Y. Tobin, eds. Towards a
Calculus of Meaning: Studies in Markedness, Distinctive Features, and
Deixis,
Fratdkin, Robert, forthcoming b. Verbal Categories and the Linguistic Sign
in Slavic and Semitic,
Garcia, Erica. 1975, The Role of Theory in Linguistic Analysis: The Spanish
Pronoun System. Amsterdam: North Holland Publishing Company,
Gesenius, W.-E. Kautzsch, (trans, A. E, Cowley). 1909, Gesenius' Hebrew
Grammar, Oxford: Clarendon Press. (Engl. tr. of 28th edition of
W. Gesenius, Hebraische Grammatik. Halle, 1813.)
Givon, Talmy. 1975. "Topic, Pronoun, and Grammatical Agreement." In Li,
Charles, ed.. Subject and Topic. New York: Academic Press. 151-188.
660

Givon, Talmy. 1976. "On the SOV Origins of the Suffixal Agreement Conjugsr
tion in Indo-European and Semitic." In Juilland, Alphonse, ed. Stfldies
Presented to Joseph Greenberg on the Occasion of his 60th Birthday.
Stanford University Press. 481-503.
Gray, Lewis. 1934. An Introduction to Semitic Comparative Linguistics.
Columbia University Press.
Greenberg, Joseph H. 1963. "Some Universals of Grammar with Particular
Reference to the Order of Meaningful Elements." Universals of Lan-
guage. Cambridge: M.I.T. Press. 73-113.
Greenberg, Joseph H. 1966. Language Universals. The Hague: Mouton.
Greenberg, Joseph H. 1988. "The First Person Inclusive Dual as an Ambigu-
ous Category." Studies in Language 12-1, 1-18.
Greenberg, Joseph H, 1989. "On a Metalanguage for Pronominal Systems: a
reply to McGregor." Studies in Language 13-2, 452-458.
Greenberg, Joseph H. ed, 1978. Universals of Human Language, 3: Word
Structure; 4: Syntax. Standford University Press.
Ingram, David. 1978, "Typology and Universals of Personal Pronouns." In
Greenberg, ed,, 3, 213-248.
Jakobson, Roman O. (1932) "Zur Struktur des russischen Verbums." In:
Jakobson, 1971, 3-15. Also: "On the Structure of the Russian Verb."
In: Jakobson 1984, 1-14. (Cited pagination according to the latter.)
Jakobson, Roman O. (1936) "Beitrag zur allgemeiner Kasuslehre: Gesamtbe-
deutungen der russischen Kasus." In: Jakobson 1971. Also: "General
Meanings of the Russian Cases." In: Jakobson 1984, 59-103. (Cited
pagination according to the latter.)
Jakobson, Roman O, (1939) "Signe Zero," In: Jakobson 1971. Also: "Zero
Sign." In: Jakobson 1984, 151-159. (Cited pagination according to the
latter.)
Jakobson, Roman O. (1957) "Shifters, Verbal Categories, and the Russian
Verb," In: Jakobson 1971, 130-147. Also in: Jakobson 1984, 41-58.
(Cited pagination according to the latter.)
Jakobson, Roman O. (1958) "Morfologiceskie nabljudenia nad slavjanskim
skloneniem." In: Jakobson 1971, 154-183. In English ("Morphological
Observations on Slavic Declension") in: Jakobson 1984: 105-134. (Cited
pagination according to the latter.)
Jakobson, Roman O. (1966) "Relationship between Russian Stem Suffixes
and Verbal Aspects." In: Jakobson 1984, 27-31.
Jakobson, Roman O, 1971, Selected Writirtgs II: Word and Language. The
Hague: Mouton.
Jakobson, Roman O. 1984. Russian and Slavic Grammar (M. Halle and L.
Waugh eds.). Berlin: Moutoir.
Johnstone, Barbara. 1987, "'He Says..So I said.' Verb Tense Alternation and
Narrative Depictions of Authority in American English." Linguistics
25.1 (287), 33-52.
Joseph, John Earl, 1987. Review of Viel, 1984. Language 63-3, 665-668.
661

Khrakovskii, Viktor. 196S. "K Xaraktere oppozicii form kataba/yaktubu


V arabskom ja2yke.'' Kratkie Soob6£enija Instituta Narodov Azii, S6.
ANSSSR. 155-163.
Kilani-Schoch, Marianne and Wolfgang Dressier. 1984. "Natural Morphology
and Tunisian vs. Classical Arabic." Wiener linguistische Gazette 33.
Kueera, Henry. 1984. "The Logical Basis of the Mukedness Hypothesis." In:
Stolz et al. eds., 61-75.
Kurylowicz, Jerzy. 1936. Derivation lexicale et derivation syntaxique.
BSL 37, 79-92.
Kurylowicz, Jerzy. 1964. The Infiectional Categories of Indo-European. Hei-
delberg: Carl Winter.
Kurylowicz, Jerzy. 1972a. Studies in Semitic Grammar and Metrics. Wro-
claw: Wydawnictwo PAN.
Kurylowicz, Jerzy. 1972b. "The Role of Deictic Elements in Linguistic Evo-
lution." In: Esquisses Linguistique II. Munchen: Wilhelm Fink Ver-
lag. 121-130.
Kuryiowicz, Jerzy. 1973. "Verbal Aspect in Semitic.'' Orientalia, Series Nova,
42, 114-120.
Kurylowicz, Jerzy. 1974. "Universaux linguistiques." XI International Congress
of Linguists I, 39-46.
Levy, Yonatha. 1982. "Some Notes on the Feminine Plural Inflection in Mod-
ern Israeli Hebrew." Journal of Psycholinguistic Research 11, 265-273.
Lyons, John. 1977. Semantics, I. Cambridge University Press.
Lyons, John, ed., 1970. New Horizons in Linguistics. Baltimore: Penguin
Books.
MacDonald, John. 1978. "The Arabic Derived Verb Themes: A Study in Form
and Meaning." In: Al-Ani, Salman, ed.. Readings in Arabic Linguistics.
Bloomington: Indiana University Linguistics Club.
Mayerthaler, Willi. 1987. "System-Independent Morphological Naturalness."
In Dressier, ed., 59-98.
McCarthy, John. 1981. "A Prosodic Theory of Non-Concatenative Morphol-
ogy." Linguistic Inquiry 12-3, 377-418.
McCarthy, John and Allen Prince. 1989. Prosodic Morphology, mss.
McCawley, James D. 1968. Review of Current Trends in Linguistics III: The-
oretical Foundations (containing Greenberg 1966). Language 44, 556-
593.
McCawley, James D. 1985. "Kuhnian Paradigms as Systems of Markedness
Conventions." In: Makkai, Adam, ed.. Linguistics and Philosophy: Es-
says in Honor of Rulon S. Wells. Amsterdam: John Benjamins.
McGregor, William. 1989. "Greenberg on the First Person Inclusive Dual:
Evidence From Some Australian Languages." Studies in Language 13-
2, 437-452.
Melduk, Igor. 1982. "Studies of the Russian Language." In Halle, Morris, ed.,
Roman Jakobson: What He Taught Us. IJSLP XXVII: Supplement, 57-
71.
662

Monteil, Vincent. 1960. L'Arabe moderne. Paris: Klincksieck.


Moravcsik, Edith. 1978. "Agreement.' In Greenberg ed., 4, 351-374.
Moravcsik, Edith. 1984. "Agreement and Markedness." In Barlow and Fer-
guson, eds., 89-106.
Moscati, Sabatino. 1964. An Introduction to the Comparative Grammar of
the Semitic Languages. Wiesbaden: Harrasowitz.
Owens, Jonathan. 1988. The Foundations of Grammar: An Introduction to
Medieval Arabic Grammatical Theory. (Studies in the History of the
Language Sciences 45). Amsterdam: John Benjamins.
Rice, Frank and George Trager. 1954. "The Pronoun System of Classical
Arabic." Language 30-2, 224-229.
Ruhl, Charles. 1989. On Monosemy: A Study in Linguistic Semantics. SUNY
Press.
Rundgren, Frithiof. 1961. Das Althebraische Verbum: Abriss der Aspekt-
lehre. Uppsala: Almqvist and Wiksell.
Russell, Robert. 1984. "Historical Aspects of Subject-Verb Agreement in Ara-
bic." ESCOL 1, 116-127.
Sangster, Rodney. 1982. Roman Jakobson and Beyond: Language as a system
of signs. Berlin: Mouton.
Schramm, Gene. 1962. "An Outline of Classical Arabic Verb Structure." Lan-
guage 38, 360-375.
Schramm, Gene. 1967. "The Correspondance of Distinctive Features in Dis-
tantly Related Languages." In: To Honor Roman Jakobson, III. The
Hague: Mouton. 1769-1774.
Shapiro, Michael. 1983. The Sense of Grammar. Bloomington: Indiana Uni-
versity Press.
Shopen, Timothy, ed., 1985. Language Typology and Syntactic Description.
I: Clause Structure; III: Grammatical Categories and the Lexicon. Cam-
bridge University Press.
Silverstein, Michael. 1976a. "Shifters, Linguistic Categories, and Cultural
Description." In: Basso, K. and H. Selby,eds. Meaning in Anthropiology.
Albuquerque: University of New Mexico Press. 11-55.
Silverstein, Michael. 1976b. "Hierarchy of Features and Ergativity." In:
Dixon, R, ed.. Grammatical Categories in Australian Languages.
Reprinted in: van Riemsdijk, H. and P. Muysken, eds., 1987. Features
and Projections. Foris Publications. 163-232. (Cited as 1976b. but
pagination according to 1987.)
Silverstein, Michael. 1979. "Language Structure and Linguistic Ideology.^' In:
Clyne, Paul, ed.. The Elements: A Parasession on Linguistic Units and
Levels. CLS, 193-247.
Silverstein, Michael. 1985. "Language and the Culture of Gender: At the in-
tersection of structure, usage, and ideology." In: Mertz, Elizabeth and
R. Parmentier, eds., Semiotic Mediation: Sociocultural and Psycholog-
ical Perspectives. New York: Academic Press. 219-259.
663

Silvetsteia, Michael. 1986. "Noun Phrase Categorial Markedness and Syntac-


tic Parametridzation." ESCOL 2, 337-361.
Silveretein, Michael. 1987. "Cognitive Implications of a Referential Hierar-
chy." In: Hickinann, Maya, ed., Social and Functional Approaches to
Language and Thought. New York: Academic Press. 125-165.
Stolz, Benjamin. I. Titunik, L. Dolejel, eds. Language and Literary Theory.
(In Honor of Ladislav Matejka). Ann Arbor: University of Michigan,
Department of Slavic Languages. Papers in Slavic Philology 5.
Tieisma, Peter. 1982. "Local and General Markedness.' Language 58-4, 832-
849.
Tobin, Yishai. 1985. "Invariant Meaning: Alternative Variations on an In-
variant Theme." In: Waugh and Rudy, eds.
Tobin, Yishai. 1988. "A Semantic Analysis of the Modern Hebrew Interro-
gratives Eix and Keitsad = 'How'" General Linguistics 28, 183-219.
Tobin, Yishai. 1989. "Space, Time, and Point of View in the Modern Hebrew
VeTb." In Tobin, ed., 61-92.
Tobin, Yishai. ed., 1989. From Sign To Text: A Semiotic View of Communi-
cation. Amsterdam: John Benjamins.
van Schooneveld, Cornelis H. 1977a. "By Way of Introduction: Roman Jakob-
son's Tenets and Their Potential." In: van Schooneveld, C.H. and
J. Daniel Armstrong, eds., Roman Jakobson: Echoes of His Scholar-
ship. Lisse: Peter de Ridder Press. 1-11.
van Schooneveld, Cornelis H. 1977b. "The Place of Gender in the Grammat-
ical Structure of Russian." Scandoslavica 23, 129-138.
van Schooneveld, Cornelis H. 1978a. Semantic Transmutations: Prolegomena
to a Calculus of Meaning. Bloomington: Physsardt.
van Schooneveld, Cornelis H. 1978b. "Contribution al'etude comparative des
cas, des prepositions, et des categories grammaticales du verbe en russe
moderne." In: Raskin, V. and D. Segal, eds. Slavica Hierosolymitana
II, 41-50.
van Schooneveld, Cornelis H. 1980. "A Semantic Proteus: The Transitivity
Feature in Russian." In: Ezikovedski proucvanija v cest na akademik
V.I. Georgiev. Bulgarian Academy of Sciences. 377-385.
van Schooneveld, Cornelis H. 1982. "The Extension Feature in Russian.'
IJSLP 5-26. Studies for Edward Stankiewicz on His 60th Birthday. 445-
457.
van Schooneveld, Cornelis H. 1983. "Contribution to the Systematic Com-
parison of Morphological and Lexical Semantic Structures in the Slavic
Languages." In: Flier, M., ed., American Contributions to the Ninth
International Congress of Slavists: Kiev, 1983. I. Columbus: Slavica
Publishers. 321-347.
van Schooneveld, Cornelis H. 1984. "Agreement in Russian.'' In B. Stolz et
al. eds, 189-214.
van Schooneveld, Cornelis H. 1986. "Jakobson's Case System and Syntax.''
In Brecht, Richard and Jules Levin, eds.. Case in Slavic. Columbus:
Slavica. 373-385.
664

van Schooneveld, Cornelis H. 1988a. "The Semantics of Russian Pronomi-


nal Structure." In Schenker, A., ed., American Contributions to the X
International Congress of SlavisU. Sofia. 401-414.
van Schooneveld, Cornelis H. 1988b. "The Semantics and Syntax of Russian
Pronominal Structure." COLING XII.
van Schooneveld, Cornelis H. 1989a. "Syntagmatic Relations and Paradigms:
Tenses and Moods in Ancient Greek Verbal Structure. A Semantic
Analysis of the Ancient Greek Verb System." In Tobin, ed., 99-121.
van Schooneveld, Cornelis H. 1989b. "Baudouin de Courtenay's Methodolog-
ical Premises for the Investigation of Language and Their Relation to
Present-Day Linguistics." In J. Baudouin de Courtenay a lingwistyka
swiatowa. Wroclaw. 1-16.
van Schooneveld, Cornelis H. forthcoming. "The Dual and Linguistic Struc-
ture: Singulative Identificational Deixis."
van Valin, Robert. 1990. "A Synopsis of Role and Reference Grammar.'' To
appear in: van Valin, R., ed.. Advances in Role and Reference Grammar.
Amsterdam: John Benjamins.
Viel, Michel. 1984. La Notion de marque chez TVubetzkoy et Jakobson. Lille:
Atelier Nationale.
Vinter, V. (=Winter, Werner) 1982. "O markirovannosti, sootvetsvii norme
i jestestvennosti." Voprosy Jazykoznanija 4, 72-77.
Waugh, Linda. 1975. "A Semantic Analysis of the French Tense System."
Orbis 24, 436-485.
Waugh, Linda. 1976. Roman Jakobson's Science of Language. Lisse: Peter
de Ridder Press.
Waugh, Linda. 1979. "The Context-Sensitive Meaning of the French Sub-
junctive." In: Waugh, Linda and Frans Coetsem, eds.. Contributions
to Grammatical Studies. (Cornell Linguistic Contributions, Vol. 2.)
Leiden: Brill. 179-238.
Waugh, Linda. 1982. "Marked and Unmarked: A Choice Between Unequals
in Semiotic Structure." Semiotica 38. 3-4: 299-318.
Waugh, Linda and S. Rudy, eds., 1990. New Vistas in Grammar: Invariance
and Variation. (Proceedings of First International Roman Jakobson
Conference). Amsterdam: John Benjamins.
Wiesemann, Ursula, ed., 1986. Pronominal Systems. Tubingen: Gunter Narr
Verlag.
Wright, William. 1896-98. A Grammar of the Arabic Language, 3rd Edi-
tion. (Both volumes reprinted in one, 1975.) Oxford: Oxford University
Press.
Wurzel, Wolfgang. 1987. "Syslem-Dependenl Morphological Naturalness." In
Dressier, ed., 59-95.
Zwicky, Arnold. 1977. "Hierarchies of Person.' CLS 13, 714-733.

Вам также может понравиться