COMPARATIVE EVOLUTION OF CEREALS
JACK R. HARLAN, J. M. J. DE WET, AND E. GLEN PRICE
Department of Agronomy, University of Illinois, Urbana, Illinois 61801
Received May 4, 1972
Our studies on the origin and evolution
of a number of cereal crops have indicated
that all of them have followed essentially
the same pathways from wild grasses to
cultivated plants and that the domestica-
tion of cereals in particular and seed crops
in general have many features in common.
In this paper we attempt to summarize
these features in a comparative way.
TAXONOMY
First, we shall introduce the materials
with which we are dealing, and this re-
quires some understanding of formal names
and botanical classification. The taxonomy
of cultivated plants has long been in a state
of confusion. The same array of variation
is treated in radically different ways by
different taxonomists (see Jirasek, 1966;
Jeffrey, 1968). Classifications are cluttered
with Latin names that have little or no
biological meaning, and some individual
taxa are given ranks ranging from variety
to genus depending on who is doing the
classifying. Inept classifications have prob-
ably caused more difficulty in understand-
ing the origin and evolution of cultivated
plants than any other factor.
We shall use the gene pool classification
suggested by Harlan and de Wet (1971)
in order to treat the several cereals on a
uniform basis. In this system the total
array of variation within maximum genetic
reach is partitioned into primary, second-
ary, and tertiary gene pools. The primary
gene pool includes all those races that can
be crossed with the crop, yielding reason-
ably fertile hybrids in which the chromo-
somes pair well and in whose offspring
genetic segregation is reasonably normal.
The primary gene pool corresponds to the
widely accepted concept of the biological
species.
EVOLUTION 27:311-325. June 1973
The secondary gene pool includes all
those species that can be crossed with the
crop but with restricted gene flow. Genes
can be transferred from the secondary to
the primary gene pool, but one must strug-
gle with those barriers that separate bio-
logical species such as sterility, poor chro-
mosome pairing, lethal or weak hybrids,
or poorly adapted hybrid derivatives and
so on.
The tertiary gene pool includes all those
species that can be crossed with the crop,
but the hybrids lead essentially nowhere.
The hybrids are lethal, completely sterile,
or anomalous. If any gene transfer is pos-
sible at all, it must be through radical ma-
nipulation of some sort such as embryo
culture, tissue culture, use of complex hy-
brid bridges and so on (see Harlan and de
Wet, 1971).
Polyploid series in cultivated plants pose
some special problems. As a general rule,
we have suggested (Harlan and de Wet,
1971) that each level be treated as a sepa-
rate gene pool. The barriers between ploidy
levels are not necessarily strong, however,
and morphological differences are some-
times minimal and difficult to describe.
Each series is different and appropriate
treatments must be worked out crop by
crop. Separate gene pools for ploidy levels
in potato or sugarcane may not be appro-
priate at all. In the cereals considered
here, the only problem of separation by
ploidy level occurs in oats where A vena
strigosa and A. barbata races are difficult
to distinguish morphologically.
Our gene pool classification is not in-
tended to be a formal taxonomic system
but rather a simple device to bring a ge-
netic focus to bear on the taxonomies al-
ready available. The conventional epithets
can be used without undue confusion pro-
311
 <.N
......
"->

TABLE 1. Primary and secondary gene pools of the major cereals.

PRIMARY GENE POOL

Spontaneous Subspecies
Ploidy
Crop Level Cultivated Subspecies Wild Races Weed Races SECONDARY GENE POOL

Einkorn 2X Triticum monococcum T. boeoticum T. boeoticum Triticum, Secale, Aegilops

Emmer 4X T. dicoccum T. dicoccoides None
" " "
Timopheevi 4X T. timopheevi T. araraticum T. timopheeui ';-<
" " "
Bread Wheat 6X T. X aestiuum None None
" " " ?O
Rye 2X Secale cereale S. cereale S. cereale ::q
" " " ;;..
Barley 2X Hordeum vulgare H. spontaneum H. spontaneum None :;>;:I
Oats ?V A 'vena strigosa A. hirtula ; A. strigosa Avena spp. t'"
u" ;;..
A. wiestii Z
Ethiopian Oats* 4X A. abyssinica; A. barbata A. barbata t'l
A. vaviloviana " >-3
;;..
Oats 6X A. sativa A. sterilis A. sterilis ; r-'
A. fatua
"
Rice 2X Oryza sativa O. rujipogon. O. rujipogo« Oryza spp.
African Rice 2X O. glaberrima O. barthii O. siapjii
"
Sorghum 2X Sorghum bicolor S. bicolor S. bicolor S. halepense
Pearl Millet 2X Pennisetum americanum P. violaceum P. americanum P. purpureum
Maize 2X Zea mays Z. mexicana Z. mexicana Tripsacum spp. Z. perennis
* Hardly more than encouraged weeds.
 COMPARATIVE EVOLUTION OF CEREALS
vided one has an understanding of what
they really mean biologically. The primary
and secondary gene pools for the cereals
treated in this paper are shown in Table 1
together with their more common conven-
tional epithets.
The primary gene pool of a crop almost
always includes wild and weedy races as
well as cultivated ones, Table 1. We have
proposed a uniform treatment of this phe-
nomenon in which the cultivated races are
included in one subspecies and the spon-
taneous races in a second subspecies (Har-
lan and de Wet, 1971). We have also
recommended that below the subspecific
rank all formal taxonomy be abandoned
and that the Latin names that clutter the
literature at infraspecific levels be dropped.
Gene pool classification provides some
insight into the structure of germ plasm
for each crop as well as the relationships
between crops. Barley for example, has
no secondary gene pool, sorghum a very
small one, while wheat and rice have rather
large ones. Wheat, barley and rye are re-
lated at tertiary levels as are maize, sor-
ghum and sugarcane.
Evolution under domestication has rarely
led to speciation. The evolution of cereals
has taken place almost entirely within the
primary gene pools. Some contribution
from the secondary gene pools has been
suggested, but not proven, for maize
(Mangelsdorf, 1961, 1968) and rice (Chu
and aka, 1970). Wheat, however, is a
clearly demonstrable case. The D genome
from Aegilops squarrosa was added to
tetraploid wheat after domestication and
recently the Secale genome has been added
to both tetraploid and hexaploid wheats
in the production of Triticale. As Harlan
(1965) pointed out, wheat is a well-
buffered polyploid and can stand the ge-
netic shock of alien germ plasm much
better than weakly buffered diploids.
Having put the taxonomy of the major
cereals on a comparative basis we shall
now examine the common features of evo-
lution within the 14 primary gene pools
shown in Table 1.
313
DOMESTICATION PROCESS: AUTOMATIC
SELECTION
Seeds of wild grasses are commonly har-
vested by hunting-gathering (or even agri-
cultural) people. Such harvests are a source
of food that could hardly be overlooked or
neglected. Over 60 species of grass are
recorded as being harvested in Africa in
recent years (Jardin, 1967). Grass seeds
are known to be important in the diet of
the Australian Aborigine, and, perhaps, a
score or more species entered the diet of
the American Indian. Wild grass seeds are
not necessarily crude roughage or scarcity
plants. One of the best known is the Ameri-
can wild-rice (Zizania aquatica) which is
still harvested in the wild by American
Indians and which is so appreciated by
whites that it sells for astonishingly high
prices in the supermarkets and specialty
stores.
Archaeological evidence suggests that
wild grass harvests had a substantial role
in the nutrition of proto-agricultural people.
Setaria was harvested on a large scale in
Mexico (Callen, 1965). Cenchrus was a
major item in the diet of Neolithic Mauri-
tanians (Munson, 1968). Wild emmer, ein-
korn and barley were harvested in the Near
East'before these grasses were domesticated
(Van Ziest, 1967; Helbaek, 1966; Renfrow,
1969). Perrot (1965) has even suggested
that the Natufians did not take up the cul-
tivation of cereals because the wild harvests
were so abundant that it was not necessary
to grow the cultivated sorts. Experimen~s
in harvesting wild wheats suggest that this
could well have been the case (Harlan,
1967) .
There seems to be no evidence that even
intensive harvesting of natural stands has
any appreciable genetic effect on the pop-
ulations. The senior author has spent many
years of his life harvesting wild grasses in
one way or another. Hand stripping, the
sickle, the scythe, mechanical blowers,
strippers, beaters, headers, binders, and
combines have all been used and tons of
seed obtained. It is rare that more than
314 J. R. HARLAN ET AI..

TABLE 2. Adaptation syndromes resulting from automatic selection due to planting harvested seed.

SELECTION PRESSURES ASSOCIATED WITH HARVESTING RESULT IN:

A. Increase in % seed recovered
Adaptations: (1) N onshattering
(2) More determinate growth
(a) Growth Habit I.-Cereals whose wild races have lateral seed-
bearing branches; e.g., maize, coix, sorghum, pearl millet. There
is a trend toward apical dominance resulting in: fewer in-
florescences, larger inflorescences, larger seed, greater daylength
sensitivity and more uniform ripening.
(b) Growth Habit n.-Cereals with unbranched culms, e.g., barley,
emmer, rye, einkorn, rice. There is a trend toward more syn-
chronous tilering and uniform whole plant maturation.
B. Increase in seed production
Adaptations: (1) Increase in % seed set,
(2) Reduced or sterile flowers become fertile.
(3) Increase in inflorescence size, especially in maize, sorghum, pearl millet.
(4) Increase in number of inflorescences especially in wheat, barley, rice, etc.
SELECTION PRESSURES ASSOCIATED WITH SEEDLING COMPETITION RESULT IN:
A. Increase in seedling vigor
Adaptations: (1) Greater seed size
(2) Lower protein; higher carbohydrate
B. More rapid germination
Adaptations: (1) Loss or reduction of germination inhibitors
(2) Reduction in glumes and other appendages
SELECTION PRESSURES ASSOCIATED WITH TILLAGE AND OTHER DISTURBANCE
RESULT IN THE PRODUCTION OF WEED RACES
Adaptations: (1) Plants competitive with cultivated races, but
(2) Retaining the shatter habit of wild races.

half the total potential seed is recovered.
The growth of grasses, especially peren-
nials, is so indeterminate that one must
allow the earliest material to shatter before
starting harvest and must waste the late
maturing material in order to obtain an
optimum amount of harvestable seed. With
the methods available to hunter-gathers,
ample seed is wasted to provide a full stand
the following year.
As long as human activity is confined
to harvesting, the genetic effect on wild
populations is likely to be negligible. It is
the seeds that escape the harvester that
contribute to the next generation, and, if
there is any selection pressure at all, it
would be in favor of such wild-type char-
acters as shattering, indeterminate growth
with maturation over a long period of time,
seed dormancy and so on. As soon as man
starts to plant what he has harvested, how-
ever, the situation changes drastically. Now
there are two populations with selection
pressures in opposite directions. The seeds
that are harvested are those that contribute
to the sown population, and any modifica-
tions that would enhance seed recovery and
competition in the new environment would
be selected favorably. Automatic selection
for interrelated syndromes of characteris-
tics is set up immediately (Table 2).
Selection pressures in these directions
appear as a matter of course whenever man
shows that which he reaps. In addition,
artificial selection pressures were imposed
according to the desires of individual culti-
vators. Man may also have selected for
larger seed, apparent yield, reduction in
glumes and appendages, naked seeds, varia-
tions in color, flavor and storage quality.
The initial establishment of a domesticated
plant can proceed through automatic se-
lection alone. The several adaptations are
amplified and annotated in the following
sections.
N onshattering .-0f all the adaptations
 COMPARATIVE EVOLUTION OF CEREALS
that separate wild from cultivated cereals,
the nonshattering trait of cultivated races
is the most conspicuous. It is, taxonom-
ically, the most diagnostic in separating
domestic subspecies from spontaneous sub-
species and is crucial in establishing the
disruptive selection that effectively main-
tains separation of the two kinds of popu-
lations. Most of the seeds that do not shat-
ter are harvested; most of the seeds that
shatter escape the harvest.
Mutants toward nonshattering probably
occur in all large populations of wild
grasses. Harlan (unpubl.) once obtained a
population of sand bluestem (Andropogon
hallii) with excellent seed retention char-
acteristics by the simple device of harvest-
ing seed after Christmas. Since seed is
usually shattered in November, the selec-
tion pressure was extreme and a single
generation was sufficient to establish a
relatively nonshattering population. Many
other grasses behave in a similar way, and
improvement in seed retention was obtained
in every species seriously studied in this
respect. A nonshattering wild-rice (Zizania)
was found in Minnesota (Johnson pers.
comm.). It disappeared from the natural
population, but if agronomists should wish
to domesticate this valuable wild cereal it
should be easily done insofar as establish-
ing nonshattering types is concerned.
The inheritance of shattering vs non-
shattering in the major cereals varies, as
one would expect, and has not been ade-
quately studied in all cases. In barley,
there are two semiallelic genes, BtBtBt2Bt2.
A recessive mutation at either locus will
suppress the shattering trait. In sorghum,
at least two systems are known (Ayyangar
et al., 1936; Karper and Quinby, 1947);
one is a single recessive gene for shattering
(sh1 ) and the other a pair of complementary
dominant genes for shattering (Sh2Sh g) .
In rice, more than two genes are involved
(Chang, 1964) although, in some crosses,
a single recessive gene (sh) confers non-
shattering. Other crosses give more com-
plex results and there appear to be modi-
fying genes and semishattering genotypes.
315
In oats, the fatuoid type of shattering (each
seed falling separately) is controlled by a
single gene; the other types (all seeds of
a spikelet falling together) may be more
complex. In wheat, not surprisingly, non-
shattering has been reported as under
monogenic and multigenic control (Porter,
1959). The situation in maize has not been
adequately studied.
It appears that in most of the major
cereals, shattering is under relatively simple
one or two gene control. Appearances may
be somewhat deceiving. Intermediate semi-
shattering forms are known in all cases,
but are relatively uncommon. Such a con-
dition is not well adapted to either cultiva-
tion or spontaneous conditions. Strong
disruptive selection for either one state or
the other will produce at least the appear-
ance of simple either/or inheritance. As
far as domestication is concerned, however,
the establishment of nonshattering traits
is, genetically, one of the easiest and sim-
plest steps in the entire process.
Determinate growth.-The well-known
"sunflower effect" applies to cereals with
lateral seed bearing branches, in this case
maize, sorghum, and pearl millet. Wild
and weedy sunflowers have many branches
bearing a large number of small heads.
The ultimate in domesticated types are the
mammoth "Russian" cultivars with single
unbranched stalks bearing enormous, single
terminal heads. Maize has followed the
same pathway. Spontaneous maize (teo-
sinte)" has a branching system in the axil
of several leaves on each stem and each
branching system includes several small
two-ranked fragile ears each enclosed in a
husk. Early maize, which is well repre-
sented in archaeological sites from south-
western U.S. to southern Mexico, appar-
ently had clusters of very small four-ranked,
mostly nonfragile ears in the axils of several
1 We are well aware of the tripartite theory
of Mangelsdorf and Reeves (1939; 1959) which
denies that teosinte is wild and weed maize. For
reasons published elsewhere, (de Wet et al., 1971;
de Wet and Harlan, 1972) we do not accept the
theory.
316 J. R. HARLAN ET AL.
leaves of each stem. From this condition,
there was a gradual progressive evolution
toward fewer and larger ears at a node
until modern, high yielding cornbelt cul-
tivars were achieved which usually have
one ear at a node and average somewhat
less than two ears per stem.
Evolutionary changes in sorghum and
pearl millet are very similar. Wild sorghum
tillers well and, although seed heads are
terminal, the stems are often branched.
The most derived of modern sorghums are
likely to have a single stem with a large
terminal head. The contrast between the
open, lacy panicle of wild sorghum and
the heavy, compact high-yielding heads of
modern cultivars is striking. The evolution
of pearl millet heads from the numerous,
small (1 dm or less) heads of Pennisetum
violaceum to the most derived cultivars
(over 2m in some cases) is nothing short
of spectacular. The process, however, is
the same in all of these cereals.
Both "multiplication" and "condensa-
tion" may be involved. By multiplication
is meant an increase in rank or row num-
ber or an increase in number of branches
of the inflorescence. This is particularly
notable in maize and pearl millet and to
some extent in sorghum. Condensation re-
fers to a shortening of internodes and
branches, especially notable in sorghum.
The trend from many small inflores-
cences to a few or a single large inflores-
cence is usually accompanied by an increase
in seed size. There are other selection
pressures that favor large seeds, but a part
of the increase may come automatically
with the increase in size of inflorescence.
The end product of the trend is a mon-
strous structure completely unadapted for
survival in the wild. The head of a com-
mercial cultivar of sunflower, an ear of
maize, a head of modern grain sorghum or
grain type pearl millet are each amazingly
different from their wild progenitor forms.
Yet the evolutionary pathway is essentially
the same in all.
An increase in the percent of seed re-
covery is obtained with more uniform ma-
turation. Tropical cultivars in particular
are extremely sensitive to day length. The
difference between the longest and shortest
days of the year near the equator may be
only a matter of minutes, but tropical races
can often detect and respond to differences
of such magnitude. Selection pressures are
very strong for maturation at the end of
the rainy season and the earlier part of
the dry season. A cultivar that matures in
the middle of the rains may be a total
failure. Sensitivity is such that it is not
uncommon for stands established in Feb-
ruary, April and June all to mature at the
same time in October or November (e.g.,
sorghum and millet in West Africa).
The wild races are often less sensitive.
It is common in our collections from trop-
ical Africa for the wild races to flower well
at Urbana, Illinois, while the cultivars from
the same region do not bloom at all or
start to flower about the time of the first
killing frost. Maturation over a long period
of time has selective value for wild plants,
but is detrimental to cultivated races. The
disruptive selection for this trait reinforces
that established by the shattering vs. non-
shattering characteristics.
Einkorn, emmer, rye, barley, oats, and
rice respond somewhat differently to the
same selection forces. They all have un-
branched culms with terminal inflores-
cences. An increase in uniformity of mat-
uration is obtained by tillering over a
shorter period of time. The life cycle of
cultivars is more rigidly controlled in time
than that of the wild races. At a given
stage, tillering essentially ceases and there
is a tendency for the whole plant to mature
at once. This response may also be strongly
conditioned by daylength sensitivity, not
only in rice from the tropics but in cereals
of Mediterranean origin as well. The win-
ter races of the temperate cereals are, in
addition, regulated by cold (vernalization)
requirements. Tillering in wheat, barley,
rye and oats, is strongly influenced by
temperature as well as moisture, soil fer-
tility and day length. Whatever the mech-
anisms involved the selection pressures are
 COMPARATIVE EVOLUTION OF CEREALS
strong in the direction of uniform maturity
and maximum seed recovery at harvest.
This is also a situation for disruptive selec-
tion reinforcing the nonshattering trait.
Increase in percent seed set.-Perennial
wild grasses are notoriously poor or erratic
seed setters. For them, survival does not
depend upon annual seed production and
a full set of seed is an uncommon event.
Wild annual grasses, however, set seed
much better and would not survive without
rather abundant yearly seed production.
This phenomenon has only peripheral in-
terest to the cereals under consideration.
Only rice and rye might have come from
perennial progenitors. In Africa, it is very
clear that African rice was domesticated
from the wild annual species Oryza barthii.
In Asia, the picture is not as clear. Oryza
rufipogon has both annual and perennial
races and the origin of Asian rice may be
more complex. We suggest that it is most
likely that the domesticated strains were
selected primarily from the annual races
but the evidence at present is tenuous. '
A perennial progenitor has been claimed
for rye. If this is true, then a considerable
improvement in seed set has occurred in
the course of domestication. Again, we
suggest that it is more likely, for this and
other reasons, that cultivated rye was se-
lected out of a weed rye which in turn was
derived from a wild annual progenitor all
in the species Secale cereale. ' (Harlan, 1968).
It is of significance, however, to note
that almost all of the cereals considered
have related perennial species within ge-
netic reach in the secondary or tertiary
gene pools. The perennials mayor may not
have had much to do with the evolution
of the crops but can be crossed with them
artificially, except for A vena.
Crop Perennial relative
Oats Avena spp. (Avenastrum
spp.) No hybrids yet
made.
Wheat Agropyron spp. Elymus
spp.
317
Rye Secale montanum
Barley Hordeum spp.
Rice Oryza longistaminata, O.
rufipogon et alia
Sorghum S. biclor in part, S.
halepense
Pearl millet Pennisetum purpureum
Maize Tripsacum spp., Zea
perennis
Fertility of reduced or sterile flowers.-
In the course of evolution in the Gramineae
it is not uncommon for florets of a spikelet
to become sterile and/or reduced and for
spikelets to become male, neuter, reduced
or entirely suppressed. Such reduction
series are to be found in most of the tribes
of the family. It is not uncommon under
domestication for that which has been sup-
pressed to be restored.
In Hordeum there are three spikelets at
a node, the central one fertile, the lateral
ones male or neuter. Throughout the en-
tire genus the wild barleys are "two-rowed."
In the cultivated six-rowed barley the
sterile lateral spikelets are fertile.' The
change is rather simple genetically in that
a single recessive mutation is adequate to
transform a two-rowed barley into a six-
rowed one. Actually there are at least two
loci involved with an allelic series at
each controlling various intermediate forms
In sorghum, there are cultivars with two
seeds per spikelet. A suppressed floret has
become fertile. These "twin-seeded" lines
are seldom very productive and have not
attracted much attention. The inheritance
of the character has not been thoroughly
worked out.
In wheat, the number of fertile florets
varies greatly and is to a considerable ex-
tent under genetic control. Some genes
cause the lower florets to become sterile
and the upper ones fertile; other genes re-
verse the arrangement. Of particular in-
terest is the fact that even the lower
glumes that are "always" sterile can be
made to bear seeds under certain genetic
318 J. R. HARLAN ET AL.
circumstances. Four doses of Q are re-
quired, but it is evident that organs once
suppressed can be restored upon genetic
command (Wright, 1958; Frankel et al.,
1969).
In wild maize (teosinte) spikelets are
borne in pairs, one pedicellate, the other
sessile. In the male inflorescence both
spikelets produce anthers; in the female
inflorescence the pedicellate spikelet is sup-
pressed. An early step in maize evolution
was the restoration of the pedicellate spike-
let to fertility making four rows of a two-
ranked ear and eight rows of a four-ranked
ear. Inheritance of this character is said
to be under single gene control (Collins,
1919; Rogers, 1950; Langham, 1940), but
some stocks give ambiguous ratios and the
matter needs further study. Paired fertile
spikelets is independent of rank number.
In addition, the female spikelets of the
American Maydeae have two florets, the
lower one reduced to small scales and the
upper one fertile. In a few races of maize,
the sterile floret is restored to fertility and
each spikelet has two seeds. The cultivar
'Country Gentleman' is a familiar example.
Increase in inflorescence size.-Increase
in size of the head or ear is closely related
to the reduction in number of inflores-
cences. But, selection in this direction is
reinforced by selection for increased seed
production. The plant that contributes the
most seed to a harvest is likely to con-
tribute more offspring to the next genera-
tion. Selection is automatic in this respect,
but is likely to be still further reinforced
by human selection for apparent yield. All
of these selection pressures go in the same
direction and away from wild-type pro-
genitors. Maize, sorghum, and pearl millet.
are more affected than the other cereals
concerned.
Increase in number of inflorescences.-·
In wheat, barley, rye, oats, and rice this is
achieved through an increase in tillering .
Head size may also increase under selection
pressure for greater seed production, but
not as spectacularly as in maize, sorghum,
and pearl millet.
Greater seed size.-The cultivated field
is a very different environment from a wild
habitat. The seedbed is favorable for ger-
mination and competition with other spe-
cies is reduced. The competition between
seedlings of the same species, however, can
become extremely intense. The first seeds
to sprout and the most vigorous seedlings
are more likely to contribute to the next
generation than the slow or weak seedlings.
Within species, large seeds have more vig-
orous seedlings than small seeds (see Knee-
bone and Cremer, 1955).
Selection for highly competitive seedlings
results automatically in selection for larger
seeds, up to a point. The plant that pro-
duces the greatest number of seeds also has
an advantage and this factor may not be
compatible with the largest seeds. Even-
tually, a balance is reached in which selec-
tion is continuous for a large number of
seeds yielding competitive seedlings. This
balance can be easily changed by human
activities, e.g., larger seeds can emerge
from deeper planting than smaller seeds.
Great variation in seed size may, therefore,
be expected in cultivated cereals, but they
are usually (not always) larger than seeds
of the wild races.
Lower protein: higher carbohydrate.-A
trend in this direction is automatic in that
most of the increase in seed size in cereals
is due to an increase in endosperm. The
embryo is richer in protein and oil, but
does not increase in the same proportion
as the endosperm. In addition to this, there
is a general inverse relationship between
yield and protein content. As yield in-
creases the percent protein decreases. For-
tunately the relationship is only general
and exceptions are possible, again up to a
point.
Loss or reduction of dormancy.-Most
wild grasses have some sort of seed dor-
mancy, which breaks down with time. The
dormancy prevents premature germination
and, when it lasts for several years helps
to maintain a seed supply in the soil only
some of which can germinate in a given
season. There is an obvious selective ad-
 COMPARATIVE EVOLUTION OF CEREALS
vantage for this adaptation in wild plants.
Wild oats, einkorn and emmer in the Near
East each have an elegant adaptation to
the erratic rainfall of the region. In all
three wild grasses there are two seeds in
each spikelet one without appreciable dor-
mancy and the other sufficiently dormant
that it will not normally sprout for a year
or more after shedding. The nondormant
seed of the pair is usually about twice as
large as the dormant one. The nondormant
seeds germinate with the first rains in the
fall and must compete with dense popula-
tions of other annual plants. Large seed
has a selective advantage under these con-
ditions and some races of wild barley,
emmer, and oats have seeds larger than
some cultivated races of these crops. If the
rains fail after this first emergence and the
plants die without producing seed, there is
still a reserve of dormant seeds which can
sprout the following year.
While dormancy has adaptive value for
wild and weed races, it is nonadaptive in
cultivated races unless it is of short dura-
tion. Some cultivars lose dormancy alto-
gether, but in regions where rainfall at
harvest time is likely, seeds may sprout in
the ear and so be lost. A dormancy that
breaks down between harvest and planting
time may have selective value under some
conditions. Automatic selection pressures
are very strong for seeds that come up
when planted; therefore dormancy is much
reduced in cultivated races. Human ma-
nipulation is important in ways other than
mere reaping and sowing. Many barley
cultivars of the Near East have some dor-
mancy; Ethiopian cultivars do not. It has
been suggested (Harlan, 1970) that Ethi-
opian barleys have been strongly selected
for prompt germination because the house-
wife brews beer almost weekly and freshly
harvested seed must be used when the new
crop comes in.
Reduction of glumes and other append-
ages.-Man has, no doubt, selected inten-
sively for less chaff, but there are certain
automatic selection pressures set up in the
same direction. Dormancy is often to a
319
considerable degree controlled by inhibitors
in the enclosing glumes, lemmas and paleas,
Selection for reduced dormancy may act to
reduce these structures. An increase in seed
size also has the feature of reducing the
relative amount of chaff in the material
harvested. The various selection pressures
are all interlocking and tending in the same
directions.
Production of weed races.-When man
tills the soil and prepares a seedbed for
his crops, he also provides an environment
favorable for wild species with adaptations
that can take advantage of the new situa-
tion. We have defined weeds as organisms
adapted to human disturbance (Harlan and
de Wet, 1965). One of the most widespread
of human disturbances has been the agri-
cultural practice of tillage. The cereals
have responded with weed races adapted
to the conditions of cultivation. As a mat-
ter of fact, almost all field crops and many
horticultural crops have weed races as well
(Harlan, 1965).
In most cases, the weed races can be
clearly distinguished from wild races mor-
phologically. The small, grassy wadi race
of wild barley can be readily separated
from the robust, thick-stemmed roadside
weeds of southeastern Turkey (Harlan and
Zohary, 1966). There is no difficulty in
recognizing the shattercanes of sorghum,
the weed rices of Asia and Africa, or the
weedy Chalco and Guerrero races of maize
(teosinte). The morphology of the weed
races is usually intermediate between the
wild and cultivated. The adaptations are
intermediate as well. The weeds are
adapted to disturbed environments, but
retain the shattering habit and, frequently,
the dormancy and seed appendages of the
wild races.
The weed and cultivated races commonly
interact genetically. This can be shown
experimentally, but is also clearly evident
in the tendency of weeds to mimic the
cultivars with which they are associated.
In both lowland Ethiopia and Eastern
Sudan where the sorghums grown are dur-
ras with compact heads, the shattercanes
320 J. R. HARLAN ET AL.
(weed sorghums) that infest the fields have
semicompact to compact heads. In high-
land Ethiopia, where the sorghums grown
are durra-bicolor with open panicles, the
shattercanes also have open panicles. In
Niger where the Fulani tribe grows a race
of pearl millet with heads one-two meters
long, the shatter types also have extra-
ordinarily long heads. Since the wild races
have heads one decimeter or less in length,
there can hardly be any confusion between
wild and weed races in that area. In the
Valley of Mexico the weed maize (teosinte)
takes on the characteristics of the partic-
ular cultivars with which it is growing
(Wilkes, 1967). If the maize is purple-
stemmed and hairy, the weed maize is
purple-stemmed and hairy and so on.
In eastern India, weed rice is a serious
pest of paddy fields. Indian plant breeders
once developed some distinctive purple-
leaved cultivars so that farmers could dis-
tinguish weeds from cultivated rice at an
early stage and could clean their fields by
pulling up the weeds. Within a few years,
however, the weed race had picked up the
purple-leaved character. There are a num-
ber of other demonstrable examples of gene
flow from a crop to its companion weed.
In some cases, on the other hand, it is
very difficult, if not impossible, to separate
the wild from the weed races morpholog-
ically and the only difference appears to
be in habitat. We see no consistent differ-
ence between the two-seeded wild and weed
einkorns. The one-seeded race appears to
be almost entirely or entirely a weed. The
situation in rye needs more careful inves-
tigation, but the weedy Secale cereale is
thought by some not to occur in the wild
and that, therefore, the truly wild S. mon-
tanum must be the progenitor of S. cereale.
Some of the races of spontaneous barley
occur in reasonably primary habitats but
appear identical to forms occupying segetal
habitats nearby. Wild and weed Avena
sterilis races appear indistinguishable.
Actually, wild races can be weedy with-
out change in morphology or interaction
with cultivated races. An example is the
behavior of wild sorghum in eastern Sudan.
There truly wild sorghum is enormously
abundant. Large tracts had never been
farmed for lack of surface water in the dry
season. With the construction of the Aswan
Dam on the Nile, people from Wadi HaIfa
and vicinity were moved to a new irriga-
tion project in eastern Sudan. Many thou-
sands of hectares of wild sorghum stands
were plowed down, leveled, and irrigated.
The wild sorghum has done well as a weed
and since the people from the north are
accustomed to growing wheat instead of
sorghum there has been essentially no in-
teraction between wild and cultivated sor-
ghum. The weed sorghum there retains the
wild morphology and does not look like a
shattercane. In cases where wild and weed
races are difficult to distinguish, it may
be that the wild forms have simply ad-
justed to artificial disturbance.
According to Vavilov (1951), both rye
and oats are secondary crops, i.e., they were
first weeds and then insinuated themselves
into domestication by being more aggres-
sive than the crops they infested. This may
well be true but the Ethiopian oats fol-
lowed a somewhat different pathway. Oats
is almost never grown as a crop in Ethiopia,
but rather as a mixture (weed) in barley
or emmer. Nonshattering and semishatter-
ing races evolved that are harvested along
with the barley and emmer. The seeds are
all threshed together and the mixture
seeded as harvested. The farmers do not
object to the contamination and make no
attempt to get rid of the oats. Neither do
they try to grow pure oats. The non-
shattering and semishattering races are
clearly related to the tetraploid A vena bar-
bata which was probably introduced into
Ethiopia as a weed of the primary cereals.
It may not be clear whether Ethiopian oats
is a domesticated plant, but it is clear that
nonshattering can evolve without delib-
erate selection by man.
There are a few other adaptive trends
that can take place without the necessity
of deliberate selection. Very hard vitreous
seeds store better and suffer less insect
 COMPARATIVE EVOLUTION OF CEREALS
damage in the wet tropics than soft chalky
seeds. Tendencies in this direction are
notable in maize and sorghum. Lax, open
heads have selective values in high rain-
fall areas in permitting more rapid drying
and less insect damage. This adaptation
is seen in sorghum especially but also in
wheat, barley, and rye.
DOMESTICATION PROCESS: DELIBERATE
HUMAN SELECTION
Throughout the process of domestication,
deliberate human selections have been su-
perimposed on automatic selection pres-
sures. In many cases, they are in the same
direction and reinforce each other. In se-
lecting for apparent yield, man will also
select for larger heads, larger seeds, more
seeds, better seed set, more determinate
growth, daylength sensitivity, easier thresh-
ing and so on. But deliberate selection adds
new dimensions to the process. Human se-
lection may be more intense and absolute
and is often biologically capricious or even
whimsical.
Without deliberate selection a given geno-
type may have a certain statistical chance
of contributing offspring to the next gen-
eration. It is a component of a population
and even if it is not among the best fitted
to .the environment, elimination may be
relatively slow (see Harlan and Martini,
1938, for natural selection experiments with
barley). But most cultivators in what we
call "primitive agriculture'? are very par-
ticular about the seed they sow. Each year
at harvest time, they carefully choose cer-
tain heads of sorghum or ears or maize
and seed from these only will be planted
for the next generation. To be sure this
procedure is less universal among the small
grains, but even in these individual heads
selection is common.
This practice provides a new order of
selection pressure. The population becomes
an array of deliberately chosen components.
2 The term shows a certain prejudice on our
part.
321
It may still be rich in variation because
cultivators of traditional agriculture have
an appreciation for mixtures, but the mix-
tures will conform to whatever an indi-
vidual selector chooses. The total potential
range of variation will be fragmented into
landrace populations or primitive cultivars.
Different cultivars will be grown for dif-
ferent purposes or to fit different ecological
niches of the agricultural system.
Man selects for color, flavor, texture,
and storage quality. He selects maize for
popping, boiling, eating off the cob, flour
quality, for making hominy, and ceremonial
cultivars for religious rites. He selects
sweet sorghum for chewing, white-seeded
types for bread, small, dark red-seeded
types for beer and strong-stemmed, fibrous
types for house construction and basketry.
He selects glutinous rice and nonglutinous
rice, long-grained rice and short-grained
rice, red rice, white rice and aromatic rice.
He selects barley for food, barley for beer,
and barley for livestock feed. He selects
grains that grind well in his apparatus for
grinding or grains that process well in a
mortar. He delights in bright colors and
curious and unusual variants. High yield
is seldom a factor in traditional agriculture,
but consistent and reliable yield is abso-
lutely essential. He knows his materials
well because survival depends on it.
Man in traditional agriculture has an
intuitive feeling for nutritive value. Cer-
tain cultivars are said to be good for
pregnant women, others good for nursing
mothers. Some cultivars are prized as food
for young children and others are said to
be "strong" and reserved for periods of
heavy work in the field. Without chemical
analyses or laboratory rats for testing, the
intuitive feeling usually has considerable
nutritional merit. Under automatic selec-
tion, the fact that a line of sorghum makes
good dumplings confers no particular fit-
ness for survival, but under human selec-
tion, fitness may be total. The line sur-
vives only if man plants it.
Examples of fitting cultivars into eco-
322 J. R. HARLAN ET AL.
logical niches are too numerous to cata-
logue here, but we shall describe one case
to illustrate a point. In many parts of
West Africa, sorghum is transplanted. like
rice in Asia. Seedlings are grown III a
sandy seed bed, pulled up by the roots,
and planted in deep dibble holes in the
field. This is done as waters recede after
the annual flood of a river or as wet areas
dry up at the end of the rains. These cul-
tivars must mature seed on residual mois-
ture only and must be ephemeral, short-
season types. The same cultivator may
grow full-season types during the rainy
season. The transplant sorts may mature
in 90 days, the rain fed sorts may take 180
days. Such practices set up separate pop-
ulations that have little chance of inter-
acting with each other. One population
matures while the other is in the seedling
stage.
In West Africa, the most common rain
fed race is guinea and the most common
transplant race is durra. We have, how-
.
ever, detected some quinea-durra interac-
tion (Harlan and de Wet, 1972). ThIS can
come about by late summer or fall planting
of a mixture of the two races when the
shortening days may bring them into bloom
together. Mixtures of this kind may occur
through careless handling by the cultivator,
but are more likely through seed purchase
at the local market.
Thus, populations are fragmented by
human activity and often kept apart by
agricultural practice. Such barriers leak,
and differentiation-hybridization cycles are
set up that enhance variability and broaden
the base for plant selection (Harlan, 1965).
The process is repeated on a geographic
scale. Cultivars and landraces evolve,
adapted to particular ecological zones or
geographic regions. These are brought to-
gether periodically through shipment of
seed or migration of people. The genetic
system of differentiation-hybridization
cycles is especially effective in exploiting
stored variability and producing new re-
combinations.
DISRUPTIVE SELECTION, FITNESS, AND
LINKAGE
Despite the spectacular arrays of varia-
tion in cultivated cereals, the genetic dif-
ference between wild and cultivated races
does not appear to be enormous. In no
case has speciation occurred except in hexa-
ploid wheat. Genetic barriers have not
developed. In crosses between wild and
tame races either one or the other mor-
phological type can be quickly recovered
in backcrosses. The small, two-ranked,
fragile ear of wild maize (teosinte) is
strikingly different in appearance from the
large, multirowed, nonfragile ear of culti-
vated maize, yet both morphological types
can be recovered in a single F 2 population
(Beadle and Galinat, unpubl.; see Galinat,
1971), either, only a few genes are in-
volved or the genes that are involved are
tightly linked on only a few chromosomes.
Similar crosses in other cereals give about
the same results.
Considering the observed interaction be-
tween spontaneous and cultivated races in
the field, this is not unexpected. Gene flow
can be easily detected as we have shown
by the mimicry of weed races and the in-
fusion of cultivated traits into them. Yet,
intermediate morphologies are rare. Fragile
six-rowed barleys are encountered as
a byproduct of spontaneous X cultivated
crosses in the Near East (Zohary, 1959,
1963, 1971), but are nonadapted and
quickly disappear from hybrid swarm pop-
ulations. Harlan has collected both shat-
tering caudatums and shattering guinea
sorghums in Africa, but plants of this mor-
phology are rare and probably ephemeral
as well. Even in interacting populations,
the morphologies are either spontaneous
or cultivated, one or the other.
A few genetical principles are applicable
to the situation and are presented below
in condensed form:
1) Disruptive selection encourages poly-
morphisms (Mather, 1955; Doggett and
Majisu, 1968).
2) Polymorphisms are maintained even
 COMPARATIVE EVOLUTION OF CEREALS
in self-pollinating populations despite high
inbreeding coefficients (Allard et al., 1968;
Jain and Allard, 1960).
3) Maintenance of polymorphisms is
associated with chromosome segments and
linkage (Jain and Allard, 1960, 1966).
4) There is a distinct advantage in the
tight linkage of all genes concerned with
balanced polymorphism (Fisher, 1930).
5) Linkage between morphology and fit-
ness genes is a general feature of the ge-
netic architecture of plant species (Grant,
1967).
6) Under strong selection pressure a
tight linkage of fitness genes results in a
rapid initial increase in fitness but a slower
rate of fitness increase in later generations
(Lewontin, 1964a, b).
7) The highest fitness is reached when
there are combinations of loosely linked
genes and tightly linked blocks of genes
for fitness (Lewontin and Hull, 1967).
Fitness for primary habitats is one thing;
fitness for the cultivated field is another.
We would expect a considerable linkage of
the genes that control these differences as
well as those genes controlling wild-type
vs. cultivated morphologies.
Viewing the plant domestication process
as a whole, it is evident that similar se-
lection pressures produce similar results in
whatever species the selection is applied.
Genetic responses and evolutionary path-
ways are fundamentally alike whether
plants are diploid or polyploid, autogamous
or allogamous. Such broad similarities in
variation patterns were pointed out many
years ago by Vavilov (1951) in his Law
of Homologous Series in Variation. We
now have a better understanding Of .the
genetic basis for homologous variation.
SUMMARY AND CONCLUSIONS
1) When the taxonomies of the major
cereals are adjusted to a comparative basis,
the common features of their evolution be-
come apparent.
2) When man started to plant the seeds
323
he had harvested, automatic selection pres-
sures on the population were put into effect.
3) Automatic selection pressures are
multiple and interlocking, all leading in
the same direction and mutually reinforc-
ing, i.e., toward better seed retention, more
determinate growth, larger inflorescences,
larger seed, loss of dormancy, and so on.
4) Deliberate human selection pressures
are more absolute and capricious, involve
color, flavor, texture, uses, culinary and
nutritive values, curios and freaks, and
vastly increase the diversity of cultivated
races.
5) The wild and cultivated populations
interact; genes from cultivated races flow
into spontaneous races, and genes from
spontaneous races flow into cultivated
races.
6) The gene flow is always across bar-
riers of some sort, but the barriers are not
of the kind that separate species.
7) Barriers of gene flow between spon-
taneous and cultivated races are e.g.
self-fertilization, time of flowering, game-
tophytic factors, nonadaptation of inter-
mediates, human manipulation, disruptive
selection and so on.
8) The differences between wild and
cultivated races are likely to be linked.
9) The differences between wild and
cultivated races are relatively simple ge-
netically; domestication has led to great
variation but not speciation.
LITERATURE CITED
ALLARD, R. W., S. K. JAIN, AND P. L. WORKMAN.
1968. The genetics of inbreeding populations.
Advan. Genet. 14:55-131.
AYYANGAR, G. N. R., U P. RAo, AND T. V. REDDY.
1936. The inheritance of deciduousness of the
pedicelled spikelets of sorghum. Curro Sci. 5:
538.
CALLEN, E. O. 1965. Food habits of some pre-
Columbian Mexican Indians. Econ. Bot. 19:
335-343.
CHANG, T. T. 1964. Present knowledge of rice
genetics and cytogenetics. LR.R.L Tech. Bull.
No. 1.
CHU, Y. E., AND H. 1. OKA. 1970. Introgression
across isolating barriers in wild and cultivated
Oryza species. Evolution 24:344-355.
COLLINS, G. N. 1919. Structure of the maize
324 J. R. HARLAN ET AL.
ear as indicated in Zea-Euchlaena hybrids. J.
Agr. Res. 17:127-135.
DE WET, J. M. J., AND J. R. HARLAN. 1972.
Origin of Maize: The Tripatrite Hypothesis.
Euphytica 21:271-279.
DE WET, J. M. J., J. R. HARLAN, AND C. A. GRANT.
1971. Origin and evolution of teosinte (Zea
mexicana). Euphytica 20:255-265.
DOGGETT, H., AND B. N. MAJISU. 1968. Dis-
ruptive selection in crop development. Heredity
23: 1-23.
FISHER, R. A. 1930. The genetical theory of
natural selection. Oxford Univ. Press, Oxford.
FRANKEL, O. H., B. SHINEBERG, AND A. MUNDAY.
1969. The genetic basis of an invariant char-
acter in wheat. Heredity 24:571-591.
GALINAT, W. C. 1971. The origin of maize.
Ann. Rev. Genet. 5:447-478.
GRANT, V. 1967. Linkage between morphology
and viability in plant species. Amer. Natur.
101:125-139.
HARLAN, H. V., AND M. L. MARTINI. 1938. The
effect of natural selection in a mixture of bar-
ley varieties. J. Agr. Res. 57:189-200.
HARLAN, J. R. 1965. The possible role of weed
races in the evolution of cultivated plants.
Euphytica 14:173-176.
1966. Plant introduction and biosyste-
matics. In Frey (ed.), Plant Breeding. Iowa
State Univ. Press.
- - . 1967. A wild wheat harvest in Turkey.
Archaeology 20: 197-201.
- - . 1968. On the origin of barley. In Barley:
Origin, Botany, Culture, Winterhardiness, Ge-
netics, Utilization. USDA Handbook 338,
Washington, D.C.
- - . 1970. On the origin of barley: A second
look. Barley Genet. II:45-50.
HARLAN, J. R., AND J. M. J. DE WET. 1965.
Some thoughts about weeds. Econ. Bot. 19:
16-24.
- - AND - - . 1971. Toward a rational classi-
fication of cultivated plants. Taxon 20:509-
517.
- - AND - - . 1972. A simplified classification
of cultivated sorghum. Crop. Sci. 12: 172-176.
HARLAN, J. R., AND D. ZOHARY. 1966. Distribu-
tion of wild wheats and barley. Science 153:
1074-1080.
HELBAEK, H. 1966. Commentary on the phyto-
genesis of Triticum and Hordeum. Econ. Bot.
20:350-360.
JAIN, S. K., AND R. W. ALLARD. 1960. Popula-
tion studies in predominantly self-pollinated
species. I. Evidence for heterozygote advan-
tage in a closed population of barley. Proc.
Natl, Acad. Sci. 46: 1371-1377.
JARDIN, C. 1967. List of foods used in Africa.
FAO. Publication, Rome, 320 p.
JEFFREY, C. 1968. Systematic categories for cul-
tivated plants. Taxon 17: 109-114.
JIRASEK, C. 1966. The systematics of cultivated
plants and their taxonomic categories. Preslia
38:267-284.
KARPER, R. E., AND J. R. QUINBY. 1947. The
inheritance of callus formation and seed shed-
ding in sorghum. J. Hered. 38:211-219.
KNEEBONE, W. R., AND C. L. CREMER. 1955.
The relationship of seed size to seedling vigor
in some native grass species. Agron. J. 47:
472-477.
LANGHAM, D. C. 1940. The inheritance of in-
tergeneric differences in Zea-Euchlaena hy-
brids. Genetics 25:88-107.
LEWONTIN, R. C. 1964a. The interaction of
selection and linkage. I. General consideration;
heterosis models. Genetics 49:49-67.
- - . 1964b. The interaction of selection and
linkage. II. Optimum models. Genetics 50:
757-782.
LEWONTIN, R. C., AND P. HULL. 1967. The in-
teraction of selection and linkage. III. Syner-
gistic effect of blocks of genes. Der Ziichter
37:93-98.
MANGELSDORF, P. C. 1961. Introgression in maize.
Eyphytica 10: 157-168.
- - . 1968. Cryptic genes for "Tripsacoid"
characteristics in Maiz Amargo of Argentina
and other Latin-American varieties. Bol. Soc.
Argent. Bot. 12: 180-187.
MANGELSDORF, P. C., AND R. G. REEVES. 1939.
The origin of Indian corn and its relatives.
Tex. Agr. Exp. Sta. Bull. 574.
- - AND - - . 1959. The origin of corn. I.
Pod corn, the ancestral form. Bot. Mus. Leaf-
lets Harvard Univ. 18:329-356.
MATHER, K. 1955. Polymorphism as an out-
come of disruptive selection. Evolution 9:
52-61.
MUNSON, P. G. 1968. Recent archaeological re-
search in the Dhar Tichitt region of south
central Mauritania. West African Newsletter
10:6-13.
PERROT, J. 1966. Le Gisement Natufien de
Mallaha (Enynan), Israel. L'Anthropologie
70:437-484.
PORTER, K. B. 1959. The inheritance of shatter-
ing in wheat. Agron. J. 51:173-177.
RENFREW, J. M. 1969. The archaeological evi-
dence for the domestication of plants: methods
and problems, p. 149-172. In Ucko and
Dimbleby (eds.), The Domestication and Ex-
ploitation of Plants and Animals. Aldine Publ.
Co., Chicago.
ROGERS, J. S. 1950. The inheritance of inflores-
cence in maize-teosinte hybrids. Genetics 35:
541-558.
VAN ZEIST, W. 1967. Late quaternary vegeta-
tion history of western Iran. Rev. Paleobot.
Paolynol, 2: 301-311.
 COMPARATIVE EVOLUTION OF CEREALS
VAVILOV, N. I. 1951. The origin, variation, im-
munity, and breeding of cultivated plants
(Trans. K. S. Chester). Chron. Bot. 13: 1-364.
WILKES, H. G. 1967. Teosinte: the closest rela-
tive in maize. Ph.D. thesis, Harvard Univ.
Bussey Institute.
WRIGHT, G. M. 1958. Grain in the glume of
wheat. Nature 181:1812-1813.
ZOHARY, D. 1959. Is Hordeum agriocrithon the
325
ancestor of six-rowed cultivated barley? Evo-
lution 13:279-280.
- - . 1963. Spontaneous brittle six-row barleys,
their nature and origin. Barley Genet. 1:27-
31.
- - . 1971. Origin of South-West Asiatic Ce-
reals: Wheats, barley, oats, and rye. In Davis
et al. (eds.) , Plant Life of South-West Asia.
Botanical Society of Edinburgh.