ELEVEN

The Item-Based Nature of Children's Early

Syntactic Development
Michael Tomasello
By all accounts. a maj or characteristic di st i ngui shi ng human heings from their nearest
primate relatives is the use o' ianguage. A central question in this regard is how human
beings mai nt ain the conventions of a par t i cul ar Ianguage across generations in a
speech communily. t hat is to say. how children acquire a Ianguage. Of special interest
to many developmental psycholinguists is the question of how children acquire the
syntactic structure o' a Ianguage. because they do not hear an adull speaking in
abstract syntactic categories and schemas bul onl y in concrete and part i cul ar words
and expressions.
The best known answer to this question - Hrst proposed by Chomsky and more
recently popularized by Pinker
1
and others - is t hat children do not have to learn or
construct abstract syntactic structures at all. bul rather they already possess them as
a part of their innat e Ianguage faculty. This so-called cont i nui y assurnption (innate
syntactic competence is fundamentally the same at all points in ontogeny
2
) jusiies
the use-of adult-like formal grammars to describe children' s early Ianguage. In t hi s
view. the 5,000 or more nat ural languages of the world each derive from this same
innate universal grammar. di fferi ng from one another only in the composition of their
lexicons and in a lew parametric variations of synt ax that are pretigured in the human
genome.
Recently. however, a nurnber of emprica! l i ndi ngs t hat chal l enge t hi s maj ori t y
view have emerged. Most i mport ant is the discovery t hat vi r l ual l y all of chi l dren' s
early linguistic competence is item-based. That is to say. chi l dren' s early utterances
are organi/ed around concrete and par t i cul ar words and phrases, not around anv
system-wide syntactic categories or schemas. Abstract and adul t - l i ke synt act i c
categories and schemas are observed to emerge only gr adual l y and in piecemeal
l'ashion duri ng the preschool years. Tliese nevv data are most nat ur al l y accounted
Ibr by a usage-based model in whi ch chi l dr en i mi l a t i ve l v l earn concrete l i ngui sl i c
expressions from t he I anguage t hey hear ar ound ( he , and t he n - usi ng t hei r
general cogni t i ve and social-cognitive s ki l l s - categori/e. schcmat i / e. and creativel y
combine (hese i n d i v d u a l l y learneil e\pressons and s l r ucl ur es l o reach adul l
l i ngui s l i c competence.
/7<9 Introduction to Grammat'tcal Development
Some Recent Findings in Language Acquisition
Most of children' s early language is grammatical l'rom t he adult point of view. and
this fact has been taken by some theorists as support for the hypothesis o an mate
universal grammar. Ikit children can also produce "grammat i cal " l anguage b\
simply reproducing the specitif l i ngui st i c t ems and expressions (e.g.. specilic words
and phrases) o adul t speech. which are, by detinition, grammatical. To differentiate
betvveen (hese Uvo hypotheses. deeper analyses of children' s l i ngui st i c competence are
needed.
Observational studies
iViany researchers believe t hat young children oprate l'rom the beginning \ vi t h abstraer
l i ngui st i c categories and schemas bccause they or only follovv adul t grammatical
conventions i'airly well. bul they also on occasion produce some creativo yet canonical
utterances t hal they could not have heard from adult s - vvhich means rhat rhey must
be generating them via abstraer linguistic categories or schemas. The most famous
examplo is "allgone sfic/a/," as reported by Braine.
!
and indeed such creativity is con-
vincing evidence that the child has some kind of abstrae! linguistic knowledge.
However, recent evidence suggests that, in this example. the only absrract knowledge
this child possesses is vvhat kinds of things can be allgone - not. for example. what kinds
of rhings na}
7
be the subjects or objects of verbs. The general methodological problem
is that vve can never tell from a single uUerance in isolalion what is the child's under-
lying structural knowledge. To determine underlying st ruct ural knowledge we must
look at all of a child' s uses - and most especially non-uses - of a whole set of linguis-
tic tems or structures.
Using this more systematic nierhod, Tomasello found that although most of hs
daughter' s early language during her second year of lile was "grammatical." it was also
very limited. uneven, and item-based.
4
The item-based nature of this child's early lan-
guage was most clearly evidenl n her use of verbs. Thus. during exactly the same devel-
opmental period some semanrically similar verbs were used in only one type of sentence
frame and t hat frame was quite simple (e.g.. Gif ). whereas other verbs were used
in more complex frames of severa I difieren! types (e.g., Dmw . Dniw on .
Draw . /ir . drnw on ). In addi t i on, morphological marking (e.g.. for past
tense) was also very uneven across verbs. Wi t hi n a given verb's development, however,
there was great cont i nui t y. wi t h new uses almost always replicating previous uses with
only one small addition or moditication (e.g.. the marki ng of tense or the adding of a
new participan! role). Overall. by Car the best predictor of this child's use of a given verb
on a given day was not her use of other verbs on t hat same day. bul rat her her use of
that same verb on immediately preceding days: there appeared to be no t ransfer
of st ruct ure across verbs. The hypothesis was t hus t hat chi l dren have an early period
in which each of t hei r verbs forms its ovvn i sl and of organi / al i on in an ot herwi se
unorganixed l anguage syslem I t h e Verb I sl and hypot hesi s). t horeby sorvi ng l o del i nc
l exi eyl l y spccitic synt acl i c categories such as "drawer." " t hi ng dr awn. " and " t h i n g
drawn wi t h" l as opposed l o subj ect . object. and i ns t r ume nt ) .
71
Using a combination of periodic sampling and mat ernal diaries, Lieven el al. " found
some very si mi l ar resul t s in a sample of 12 English-speaking chi l dren from 2-3 year.s
of age. In par t i eul ar . t hey Ibinid t hat chi l dren used virtual!) all of their verbs and pre-
dicative terms in one and onl y one sentence frame early in l anguage development - sug-
gesling t hat Ihei r s ynt ax was bui l t around vari ous par t i cul ar t ems and expressions. In
fact. 92% of these chi l dr en' s earliest mul t i -word ut t er ances emanaled from one of t hei r
lirst 25 l exi cal l y based pat t erns. \ vhi ch were di fferent for di f f er ent chi l dren. Following
along these same l i nes. Pi ne and I. i even
1
' f ound t hat when these same children began
lo use the determiners a and t / i c between 2 and 3 years of age. they did so with almost
completely different seis of nouns (i. e. , there was almost no o\
r
erlap in the seis of nouns
used wi t h the two det ermi ners). This suggested t hat the chi l dr en at t hi s age did not have
any kind of abstrae! category of determiner t hat included both of rhese lexical tems.
This general fi ndi ng of the tem-based l earni ng and use of language has novv been repli-
cared in a number of differenr languages of many different types (see Box 1 1 . 1 ) ,
Box 11.1 Cross-linguistic evidence for item-based patterns
A number of systematic studies of children learning languages other rhan
English have also found many item-based patterns in early language develop-
ment. For example, Pizzuto and Caselli (Refs a,b) investigated the grammatical
morphology used by three Halian-speaking children on simple, finile. main verbs.
between the ages of about 18 months to three years. Although there are six forms
possible for each verb root (flrsr-person singular, second-person singular, ere.), the
findings were that:
47% of all verbs used by these children were used in one forra only
an additional 40% were used with two or three forms
of the 1 3% of verbs thar appeared n four or more forms, approximately
half of these were highly frequent. highly irregular forms that could only
be learned by rote.
The clear implication is that Italian children do not master the whole verb
paradigm for all their verbs at once, bur rather they inirially master only some
endings wirh some verbs - and often different ones with different verbs.
In a similar srudy of one child learning to speak Brazilian Portugese at around
3 years of age, Rubino and Pine (Ref. c) found a comparable pattern of results,
ncluding additional evidence that the verb forms this child used most frequently
and consistently corresponded to rhose he had heard mosr frequenrly from adults.
Thar is. this child produced adult-like subject-verb agreement patterns for the
parts of the verb paradigm that appeared with high frequency in adult language
(e.g.. first-person singular), but much less consistent agreemenr patterns in low
frequency parts of rhe paradigm (e.g., third-person pl ural ). Similar!}', in a study
of six Hebrew-speaking children - a language rhat is typologically quite different
from European languages - Berman and Armon-Lotem (Ref. d: see also Ref. e)
17, Introduction to Grsmmatical Devetopment
found that Hebrew children's irst 20 vcrb forms were almost all "rote-learned or
morphologically unanalysed" (Ref. d, p. 37). Other similar results have been
reported for Hungarian (Ref. f). Cataln, Germn and Dutch (Ref. g). Inuktitut
(Ref. h), Spanish (Ref. i), and Russian (Ref. j).
Refcrences
a Pizutto, E., & Caselli, C. (1992). The acquisition of Italian morphology. /. Child Lcmg..
19. 491-S57.
b Pizutto. E., & Caselli, C. (1994). The acquisition of Italian verb morphology in a cross-
linguistic perspeclive. In Y. Levy (Ed.), Other Children, Other IjmgtiagL's. pp. 1 37-188.
Erlbaum.
c Rubino, R., & Fine. J. (1998). Subject-verb agreement in Brazilian Portugese: what low
error rales hide. /. Child Lang.. 25, 35-60.
d Berman. R.A., & Armon-Lotem. S. (1995). How grammatical are early verbs? Anales
Littmires de l'Universit de Franche-Comt, 631. 17-56.
e Berman. R. (1982). Verb-pattern alternation: the interface of morphology, syntax. and
semantics n Hebrew child language. /. Child Lang.. 9, 169-191.
f MacWhinney, B. (1978). The acquisition of morphophonology. Monogr. Soc. Res. Child
Dev., No. 43.
g Behrens, H. (1998). How difflcult are complex verbs? Evidence from Germn, Dutch.
and Engiish. Lingustics, 36, 679-712.
h Alien, S. (1996). Aspects of Argument Structure Acquisition in Inuktitut. John Benjamins.
i Gathecole, V et al. (1999). The early acquisition of Spanish verbal morphology:
across-the-board or piecemeal knowledge? Int. }. BingiiaUsm. 3, 138-182.
j Stoll, S. (1998). The acquisition of Russian aspect. First Lang.. 18. 351-378.
Of special note in children's spontaneous speech are so-called overgeneralization
errors because the child has presumably not heard such errors from adults. The over-
generalizations of most intcrest in the context of a focus on syntax are those involving
basic sentence Iranes, for example. "Shefalled me dowu" or "Don't giggle me." in which
the child uses intransitive verbs transitively (i.e., a verb normally used vvi t h a subject
only is used with both a subject and an object). Bowerman'
N
documented a number
of such overgenerali/ations in the speech of her two English-speaking children. and
Pinker
9
compiled examples from other sources as vvell. The mai n result of intercst vvas
t hat these chi l dren produced very f'ew <i f thcse types of overgeneralixations beforc about
3 years of age. This developrnental pat l ern again provides support for the hypothesis
t hat the const r uct i on of abstract l i ngui st i c categories and schernas is a gr adual process
I hat (akes place over many mont hs, and even years. of ontogeny.
Experimental studles
The other mai n met hod lor s t udyi ng t he na t ur e of chi l dr en' s l ngui s t i c kiKm' ledge
involves l eachi ng I hei n novel l i ngui s t i c t ems and sceing u' hat Ihev do \ \ i t h i hem. The
idea is t hat i l the chi l d uses ( he novel i l ei n n creativo vel canoni cal wavs, vve mav nt e r
173
t ha t she has assi mi l at ed t lo some ki nd of abst ract calegory or schema. I I ' she does not
use it in any creative vvays (despite repeated oppor t uni t i es) . bul only in \vays she has
heard from adul t s. the i nference is t ha t t here is no abst ract system to take up the nevv
element. and the chi l d is simply imitatively l ear ni ng a specilic l i ngui st i c tem or st ruc-
t ure ( assumi ng thal there are no performance l i mi t at i ons . i nvol vi ng l i mi t ed memory or
t he l i ke. t hat preven! t he chi l d from demonst r at i ng her synl ael i c competence i n t he
expcriment).
Kxperiments usi ng novel \
r
erbs have demonstrated t hat by 3 years o age most chi l -
dren can readily assi mi l at e novel verbs to abstract synt act i c calegorics and schemas
t hat t hey bri ng to the experi ment . for example. t a ki ng a verb they have heard only in
a passive sentence frame and using it in an activo sentence frame.
1
"" llowever. the
same is not t r uc for younger chi kl ren. l-'or example. Tomasello and I3rooks
l j
cxposed
2- 3-ycar-oId children to a novel verb used to refer to a hi ghl y t ransi t i vo and novel action
in which an agent was doing something to a pat i cnt . In the key condi t i on the novel verb
vvas used in an i nt ransi t i ve sentence frame such as "Tlic suck is Idinmina" (to refer to a
s i t uat i on in vvhi ch. for example, a bear vvas doing somet hi ng t hat caused a sock to
"tam" - similar to the verb roll or spii. Then. vvith novel characters performing the
targel action. the adul t asked chi l dren the quest i on. "U' /wt is lite thggie doingr" ( \ vhen
the dog vvas causing some nevv character to t am) . Agent quest i ons of t hi s type encour-
age a transitive reply such as "He's tammiiui the car." vvhi c h vvoul d be creative as the
child has previously hoard this verb only in an inlransive sentence frame. The outcome
vvas t hat very f'ew chi l dren at either age produced a transitive ul t erance vvi t h the novel
verb. As a control, children also heard another novel verb introduced in a transitive sen-
tence frame. and in this case vi r t ual l y all of t hem produced a transitive utterance. This
demnstrales that children can use novel verbs n Ihe transitive construction vvhen they
have heard Ihem used in thal vvay (see fi gure 1 1 . 1 ) .
The generality of Ibis flnding is demonslrated by a number of similar studies using
differenl modeled constructions and measurement procedures. These studies have used
children of many different ages and have tested for a variety of different constructions
(see Box 1 1. 2) . Most of Ihe lindings concern chi l dren' s abi l i t y to produce a simple tran-
sitivo utterance (subject-verbobject; SYO). given that they have heard a novel verb
only in some other sentence frame (e.g.. intransitive. passive. imperativo, etc.). VVhen
all of these Hndings are compiled and quant i t at i vel y comparec, vve see a continuous
developrnental progression in vvhich children graduall y become more productivo vvi t h
novel verbs duri ng their t hi rd and l' ourth years of l i fe and beyond (see f i gur e 1 1.2 and
Table 1 1. 1 ). It is cloar t hat this overall pat t ern is not consisten! vvi t h the hypothesis I hat
children possess abstract l i ngui s t i c knowledge early in devel opment , but ral her il is con-
sisten! wi t h a more constructivist or usage-based model in which young children begin
language acqui si t i on by mitalively l ear ni ng l i ngui s t i c t ems di rccl l y from adul t l an-
guage. onl y laler discerning the kinds of pa t erns t hat enable them to const r uct more
abstract l i ngui st i c categories and schemas.
The val i di t y of these l i ndi ngs s f ur t hor corroborated by two cont rol st udi es t ha l deal
vvi t h al t er nat i vo hypolheses. Fi r st , t is possiblo t ha t young chi l dr en are simply reluc-
an! to use nevvl y l oarnod vvords in novel wavs. Il owever. \ vhen oven younger children
(11 mont hs ) are l a ughl novel I I OL I I I S . Ihey LISO I hem q u i t e freely i n novel sent ence
I r an es .
1
' '
1 1
Young cl i l l dr on ar e i hus no! r el i cenl v v i t h al l no\ vl v l earned words. and
174
Introduction to Grsmmatical Development
14 -,
12 -
10 -
o 6-
Age (yrs)
Figure 1 1 . 1 Imilative production o' novel verbs. The numbcr of utterances children produced
vvith the novel verb in one condition of the Tomasello and Brooks study.'- Conservative utter-
ances (liglit gray) were those in which children heard an intransitive use of the novel verb
and thcn reproduced a similar intransitivo utterance. even when they were encouraged to
produce a transitive utterance. Productive utterances ( dar k gray I were those in which children
used the novel verb (heard in an intransitive ut t erance) in a transitive utterance. (Adapted
l'rom Ref. 12 ]
indeed they seem to l'orm something ke a category of "concrete noun" qui t e early in
development (see also Ref. 1 5 and related studies for addi t i onal evidence.) Second, it
might be that children' s lack of productivity in t he novel verb studies does not have lo
do vvith t hci r l i ngui st ic knowledge. bul only with production difculties. Howcver. in
comprehension tests they perform no betlcr. That is. they are lirst laught a novel verb
in a simple sentence trame ("Lwk! Taininmn! 77i/s ix ailled tnnwtinii!). and they are askecl
to act out a transitive construction vvi t h t hat verb I "S/im\' me: f / i c (/<><;' x taniininii tlu- ca").
Perhaps surprisingly. children younger t han 3 years of age do no better in comprehen-
sion t han they do in production. "' (The study of Naigles
1
' is somctimes t aken to be cus-
crepant vvi t h trese i ndi ngs. bul in fac it is not relevan! because the t vvo senlenccs t hat
were comparec in that study were "The dui'k is iilorpinj /ic biiwiu" and "77ic Ininifi mu
the diu'k are glorpiuj" - wi t h one pi ct ure cl epi ct i ng the duck doi ng somet hi ng to t he
bunny and t heot her depi ct i ngt het vvo participantsengaged i n t he same paral le act i on.
The problem is t hal chi l dr en mi ght very uel l have been t i si ng the vvorcl and as an i ndi -
cat or o (he par al l el act i on pi c t ur e .
l > -
Children's hsrly Syntactic Development
175
Box 11.2 Other experimental studies of children's early productivity
A number of studies have used the same basic design as Tomasello and Brooks
(Ref. a), but with diffcrent age children and with the novel verbs presented in
difieren! sentence frames. With specific reference to children's ability to genrate
a novel transitive (subject-verb-object; SVO) utterance:
1 Children were presented with a novel verb in a presentational construction
such as "This is called gorping." and encouraged va questions to produce
a transitive utterance (Refs b.c,d).
2 Children were presented with a novel verb in an imperative construction
such as "Tam, Arma!", and encouraged va questions to produce a transi-
tive utterance (Lewis and Tomasello, unpublished data).
3 Children were presented with a novel verb in a passive construction such
as "Ernie s getting meeked by the doy," and encouraged va questions to
produce a transitive utterance (Ref. e).
In all of these studies the overall finding was that children below 3 years of age
were very poor at using their newly learned verbs in the transitive construction.
with the vast majority of children below this age never producing a single
transitive utterance. In most cases we also had control conditions in which those
very same children did produce a transitive utterance (using different object
ames as subject and object) when they had heard a novel verb modeled for them
in this way.
It is also noteworthy that the few novel verb studies on languages other than
English (although using slightly different syntactic constructions) have found
very similar results - a general lack of productivity with novel verbs before 3 years
of age (Ref. F, Hebrew; Childers and Tomasello, unpublished data. Childean
Spanish).
One other study is of speeial importance because it did not only show children
failing to be creative: it actually succeeded in inducing children to produce non-
grammatical English utterances (which should not be possible if certain innate
parameters. such as head direction. were already set). Akhtar modeled novel
verbs for novel transitive events for young children at 2;8, 3:6. and 4:4 years of
age (Ref. g). One verb was modeled in canonical English SVO order, as in "Ernie
meeking the can," whereas two others were in non-canonical orders, either SOV
("Ernie the w tamming") or VSO ("Gopping Ernie the cow"). Children were then
encouraged to use the novel verbs with neutral questions such as "What's hap-
pening?" Almost all of the children at all three ages produced exclusively SVO
utterances with he novel verb when that is what they heard. However. when they
heard one of the non-canonical SOV or VSO forms, children behaved differently
at different ages. In general, the older children used their verb-general knowledge
of English transitivity to "corred" the non-canonical uses of the novel verbs
to canonical SVO l'orm. The younger children. in contras!, much more often
matched the ordcring patterns they liad heard with the novel verb, no matter how
bizarre that pattern sounded to adult ears. Interestingly, many of the younger
children vacillated betvveen imitation of the odd sentence patterns and "correc-
tion" of these patterns to canonical SVO order. This indicated that they knew
enough about English word-order patterns to discern that these were strange
utterances. but not enough to overeme completely their tendency to imitatively
learn and reproduce the basic structure of what the adult was saying with the
novel verb.
References
a Tomasello, M.. & Brooks. P. (1998). Young children's earliest transitive and intransitive
constructions. Cognit. Linguist.. 9, 379-395.
b Olguin. R., & Tomasello, M. (1993). Twenty-five month od children do not have a
grammaticai category of verb. Cognit. Dev., 8, 245-272.
c Akhtar, N.. &Tomasello. M. (1997). Young children's productivity with word order and
verb morphology. Dev. PsychoL, 33, 952-965.
d Dodson. K., &Tomasello. M. (1998). Acquiring ihe transitive construction in English:
The role of animacy and pronouns. /. Child Lang.. 25, 55 5-574.
e Brooks, R, & Tomasello. M. (1999). Young children learn to produce passives with
nonce verbs. Dev. Psijchol, 35, 29-44.
f Berman. R. (1993). Marking verb transitivity in Hebrew-speaking children. /. Child
Lang.. 20. 641-670.
g Akhtar, N. (1999). Acquiring basic word order: evideuce for data-driven learning of
syntactic structure. /. ChildLang.. 26. 261-278.
Implications for Theories of Language Acquisition
Combining the results from naturalistic and experimental studies, it is clear t hat young
children are productivo with their early language in only lirnited ways. They begin by
learning to use specific pieces of language and only gradually crate more abstrae! lin-
guislic categories and schemas. These (indings have importan! implications for curren!
theories of child language acquisition.
Unguistc natvism
Classically. as espoused by Chonisky for example, l i ngui sl i c nat i vi sm has emphasixed
t hat chi l d language acqui si t i on: ( 1 ) takes place qui ckl y and et'fortlessly because chi l -
dren have l' ull l i ngui sl i c competence at bi rt h and need only to learn to express l i l i s com-
petence overtly in performance: ( 2 ) relies only indirectly on the language chi l dren hear
( i . e. , "i nput " only serves to "trigger" innale syntactic structures orto "sel parameters");
and 1 5 ) is creaI i ve from early in onlogeny because i! is generaled by an abslrac!
gramniiir. The dat a j ust rcviewed are cl earl v at vari ance wi t h each of t hese cl ai ms . and
i n addi t i on. t he dat a cal i i nl o quesl i on al t ogel her t he use of a d u l l - l i k e gr ammar s l o
describe cl i i l dr en' s earl y language.
Children's Early Syntactc Development
77
100-
90
70
60-
50 -
40
30-
P1
P3 P4
14
H3
10
S2
11 6
A 2
H2
P2
S1
20-
7
10-
1 2
1
4
,-j j ^__
2;0
9
3
5 H1
13 12
A1
2:6 3;0 3;6
i i /
4;0 4;6 5;0
Age (yrs;mths)
Figure 11. 2 Productive iransitive utleranees in difieren! studies. Percenlage of children
(or responsos in some cases - see Table 1) that produced Iransitive utleranees of a novel verb
that was heard in some other sentence trame. The data points correspond to the studies Usted in
Table 1 1 . 1 .
The classic response of l i ngui st i c nat i vi sm to children' s synl act i e l i mi t at i ons is to
invoke hypohesi/ed (but never measuredl performance limitations that i nhi bi l the l'ull
expression of chi l dren' s innate l i ngui st i c competence (e.g.. l i mi t ed working memory).
19
Many of the control conditions in the ahove experiments. however. pul performance
demands on chi l dren very si mi l ar lo tlise of the experi ment al condilions. but children
experienced no learning diflicullies - for example. in using a newly learned noun in
novel ways and n usi ng a newly learned verb in a t r ans i t i ve ut t er ance when they liad
heard it modcled in t hat way. 11 is also noteworthy I hat children' s performance \vas also
conservalive and item-based in lwo difieren! comprehension experiments. whi ch pi afe
many fe\ver performance demands on young chi l dren.
Recenlly. some l i ngui s t i c nat i vi s t s ha\
r
e also proposed t he idea t ha t chi l dren are nol
born wi t h l' ully adul t - l i ke s ynt act i c compelence. On t hi s vi e\ v. chi l dren' s early l anguage
devel opment mi ghl be i t em-based and pi ecemeal . bul t he genes for many a dul t - l i ke syii-
l acl i c st r uct ur es begin lo "Ui r n on" somet i me hetween 1 and > years of age/" The
problem i n t hi s case i t t h a t . i n t he exper i ment al dal a r evi ewed. t he gr adual and piecc-
jyg Introduction to Grsmmstlcsl Oevelopment
Tablc 1 1 . 1 Research using novel verbs
Relerence
11
Data
: i l !
poi nl Age Productivity - i n
g. 2 l yr s i i nl hs l m
, uistic Ki i dl i ng Scori ng
ode quest i on
Re. 1-1
Re. 1 2
Rd'. 4 ?
Kef. I 8
Re'. 59
Ref. 40
' n children
"',, children
% ehildren
";, chi l dr en
Ref. 38
Ref. 1 1
Ref. 10
Ref. 4 1
Ref. 42
Kel. 44
12
13
14
P3
P4
P5
VI
Al
A2
A 3
I I I
1-12
H ',
4:6
3:10
5: 1
6:1
7:1 1
5:0
2:8
3:6
4:4
2:9
3:9
0. 2T
0. 38
l' resentitional Neut r al
I nt r ans i t i ve Agent
Iknv freq.
Hnglish verbsi
Passive Agent
InranMive Agent
SOV and \'S() Neut ral
I nt r ansi l i ve Sentence
(Hebreivl completion
1 st or 3rd Neut ral
persnn verb
Spanish)
' > children
iresponses
, responses
( act i on verbs)
% children
' i , chi l dren
% responsos
"u children
meal developmenlal process \viis al l \ \ i i h i n t he same svnt acl i c sl r uet ur e. namely. i he
linglish t r ans i t i vo const r uct i on. Chi l dr en who can use i he si mpl e' t r ansi l i ve const ruc-
t i on Ibr f a mi l i a r verbs prcsumabl y al r eadv havc t he required genel i e bases i n place, and
so it beeomes a mvstery xvhy they eannol use these same genelc bases lo use novel
verbs i n t r a n s i t i v o H t enmeos i n experimental cont ext s .
179
Finally, it is also possible lo posit Ihat ehildren' s early language is itom-based. bul
thal aft er "sulieient" l i ngui st i e t r i gger i ng experionces. i t beeomes l i nked \ vi t l i t he
i nnat o uni \ ' ersal grammar.
21
The problem in t hi s ease is t hat thore is only one serious
theory o' hovv t hi s l i nki ng mi ght tako plaee - l' inker' s t heory of i nnat o l i nki ng rules"
- and t hi s theory does not l i t \ vi t h tho emprica! dat a"
2
' (seo Ref. 24 on problems of
hypothosi/ing linguistie universals). In general, it is very difticult to envision ho\v ai i
i nnat o uni ver sal gr ammar eould be biologieally propared ahoad of t i mo to l i n k up its
speeilic oatogorios and sehemas lo the part i cul ar syntactic conventions of the many dif-
lerent languages of the \vorld (o.g.. ergalive-ahsolutive vorsus nominative-accusative
svstems).
Usage-based accounts
Usage-based approaches to language acquisition attempt to characteri/e ehildren' s
language nol in lerrns of i nnat o, adul t -l i ke, formal grammars. bul ralher in lerms of
the cognitive and communicative processes involved. VVith rcspecl to the data revievved
above. tho hypothesis would be t hat ehildren' s earliesl language is based on tho speeilic
l i ngui st i e tems and expressions they comprehend and produce. Children begin to form
an abstracl category of "concreto noun" qui te early. and this allows them t o uso any
symbol eategori/ed in t hi s way productvely in a wide rango of linguistie contexts. It
takes some time Cor children to categrico or schematize the relational-syntactic
structure of thoir various item-based (verb island) constructions. hoivever. and theroby
to becomo productivo \vith their language in more adult-like vvays. The adult endpoint
of this developmental process is nol an abstrae! formal grammar. but rather an "inven-
tory of symbolic resources" including everything from vvords and morphemos to
whole grammatical construclions as kind of linguistie gestalts
21
(seo papers in Ref. 26).
This developmental trajoctory depends on cognitive and social-cognitive processes
common to all human beings (and. of course. on experiential human universals like
growing up in the midst of language usors). Three of these processes are especially
important.
First in importance is cul t ural learning or. more specifically. imitative learning in the
specific sense used by Tomasello ot al.
2
On this view, imitative learning is not simply
repeating or mimicking the surface form of adult utlerances. Rather. it is the at t empt
by children to reproduce the language adul t s produce and for the samo comni uni cal i vo
t ' unct i on - the reproduction is of bolh (he linguistie form and its conventional com-
muni cat i ve f unct i on. Al one levol of analysis. t hi s absolutoly mus be truc because all
children l earn the language to \ vhi ch they are exposed. and for all non-canonical
aspeis of language sl ruet ure - all i di oms. l exi cal tems, qui r ky const ruct i ons. and the
like - nobody has ever proposed any mechanism other t han some form of i mi t at i ve
l ear ni ng. (For oxample, only by observing and reproducing par t i cul ar l i ngui st i e
symbols can one l earn I hat . in F. nglish, "Tlnit won't jo dinvu well witli um" moans t hat
he u' on' t l i ke t h a l . I The curren! proposal s si mpl y t h a t . i ni t i a l l y. i mi t a t i v o l ear ni ng is nll
I hat chi l dren do f or al l l i ngui s t i e const r ucl i ons. canoni cal and qui r ky al i ke. Tl i i s
approach t hus hi ghl i ght s t he rolo of (he l anguage t ha l chi l dr en hear a round l l i em. and
il also takes seriously the pos s i hi l i t y of i n d i \ ' i d u a l dilerences haseil bol h on ehi l dren' s
pot enl i al l y differenl perceptual and cognilive ski l l s and on t hoi r pol ent i al l \ ' differont
l anguai e l e a r n n <> cm i r omnent s.
2 N
]QO Introduction to Grammsical Development
Secondly. children go beyotid thesc early tem-bascd constructions n el u e course. The
only way they can do t hi s is to iind pat t cr ns in the l anguagc they are hear i ng. and
thereby to orn some ki nds of abstrae! categories and schemas. Chi l dren do this in the
case of the category of concrete nonn qui t e early. Bul in addi t i on they abstrae! across
more complcx relational structures as well. for example. ivhole constructions such as
the simple t r ansi t i ve consl ruct i on. Al t hough t here are no good dat a on how they do
t hi s . t he vrork of l i ent ncr on analogy and "st ruct ure mappi ng" provides some i nt er-
esting hypothescs.'"The idea is t hat children must see both t hest ruct ural and the func-
tional similarities in utterancessuch as "lilmwtmi." "Shekissednie." "I liit ]effrei," "Yon
uit] Monimii," "iiiiiic kickincj hill." in terms of t hei r r el at i onal st ruct ure. independent of
the speclic \vords involved. A reasonable assumption s t hat there must be some "crit-
ieal mass" of exemplars of particular utterance types necessary for the human cogni-
tive apparatus to be able to make the requisite analogies and subsequent categories and
schemas." It may be t hat the cri t i ca! factor is the number of di fferent verbs heard in
the construction - because verbs are the central organizing element in utteranee-level
constructions and because many exemplars vvith only one or a lew verbs vvould seem
to be a very inadequate basis for generali/ing the construction.
12
Third and tinally. children also combine various ki nds of l i ngui st i c constructions
creatively. involving both concrete and abstract constructions of varying levis o com-
plexity. They combine much more complex structures than ust words or word classes.
As one example, one child' s earliest utterances vvith three or more words were t hi ngs
like ".Vt' Daddy's car." But previously this chi l d had said things like ".SV? hall" and "See
Momimi." on the one hand, and also things like "Daddy's shirt" and "Daddy'spen," on
the other. So, the likelihood is that she creatively combined something like a "See "
schema \vith a "Daddy's " schema. Note t hat to do t hi s she had to understand that
"Daddy's car" as a complex expression was n some sense equivalen! to the other things
she previously had been talking about seeing (Ball and Mommy), and so this combina-
tion indcales some knowledge of the funct i onal equivalence of these different referring
expressions. It should be noted (hat many different procedures may be used to combine
established constructions in these ways. For example. a child might combine an item-
based construction with a more abstract construction. or she might combine two item-
based or two abstract constructions with one another. Diessel and Tomasello "report a
further i l l ust rat i on of these processes in more complex constructions (.e.. those wi t h
sententia! complements: see Box 1 1 . 3 ) .
Conclusin
11 gr ammat i cal st ruct ures do not come di recl l y (rom the human genome. as the above-
reported dat a suggesl they do not. and i f chi l dren do nol i n\
r
ent t hem de novo, as they
clearly can not. then it is l egi t mat e to ask, Where do grammat i cal structures come
from? The answer is t hat . in the lirst instance, thev come l'rom processes of grammat i -
cal i xat i on in l anguage history. That is to say. at some poi nt in h u ma n evol ul i on. Homo
suirii'jia e\' olvecl t he a b i l i l y t o commun cal e wi t h one anot her symboiicallv.
H
\ Vhe n
huma n beings comni uni cat e symbol i cal l y ui t h one anot her i n extended discourse
nteractions. the stringing together o symbols begins to become grammal i cal xcd: for
Children's Early Syntsctic Developtnent j,
Box 11. 3 A more complex example of structure combining
As a more complex example of structure combining, Diessel and Tomasello (Ref.
a) looked at the earliest complex sentences with sentential complements of six
children. They found that virtually all early complement sentences are composed
of a simple sentence schema that the child has already mastered, combined with
one of a handful of matrix verbs (see also Ref. b).
These matrix verbs are of two types. First are epistemic verbs such as think and
know. fn almost all cases children and / think to indcate their own uncertainty
about something, and they basically never used the verb think in anything but
this first-person form (i.e., no examples of "He thinks . . . ," "She thinks . . . ," etc.).
This form was also virtually never negated (no examples of "I don't think .. ."),
virtually never used in anything other than the present tense (no examples of
"Z thought. . ."), and never with a complementizer (no examples of "J think that
. . ."). It thus appears that / think is a relatively fixed phrase meaning something
like may he. The child pieces together this fixed phrase with a full sentence, but this
piecing together does not amount to "sentence embedding" as it is typically por-
trayed in more formal analyses - it is more like simple concatenation because the
main verb (think) is not really acting as a verb. Second. children also use
attention-getting verbs like look and see in conjunction with full sentences. In this
case, they use them almost exclusively in imperative form (again no negations,
no non-present tenses, no complementizers). Therefore, these early complex
sentences do not appear to be abstract sentence embeddings, but rather concate-
nations of a formlale expression and a full sentence.
Examples from Sarah:
/ think he's gone
think it's in here
/ think my daddy took it
I think I saw one
It's a crazy bone. I think
I think dis is de bowl
References
Examples from Nina:
See that monkey crying
See Becca sleepng
See that go
See my hands are washed
See he bite me
See him lie down
a Diessel. H.. & Tomasello. M. Why complement clauses do not have a that comple-
mentizer in early child language. Berkcley Linyuistics Society f i n press).
b Bloom, L. (1992). Language Development from Two to Three. Cambridge University Press.
example. cont en words such as nouns and verbs become funct i on words such as prepo-
s i t i ons and aux l i ar i es . and loosely concat enal ed symbols acqui r e synt acl i c r el at i on-
ships i n' ol vi ng const i t uency and dependency (see Box I 1. 4) . Thesc ( r ansf or mat i ons of
linguistic structure occur as a rcsult of social-interaclive processes n \vhich ( 1 ) speak-
ers t ry to abbreviate l i ngui st i c expression as much as t hev can. and 11 \ lisleners t r v to
}&Z
Introduction to Grammatical Development
Box 11.4 Grammaticalization
Each of the 5,000 or more languages of the world has its own inventory of
linguistic conventions, including syntactic conventions. which aliow its users
to share experience with one another symbolically. This inventory of symboiic
conventions is grounded in universal structures of human cognition, human
communication, and the mechanics of the vocal-auditory apparatus. The pecu-
liarities of particular languages come from differences in the kinds of things
that different speech communities think it important to talk about and the
ways they think it useful to talk about them - along with various histrica! "acci-
dents." All of the conventions and constructions of a given language are
not invented at one time, of course. and once invented they often do not stay
the same for very long, but rather they evolve, change, and accumulate over
time as humans use them with one another. This set of processes is called gram-
maticalization. and it involves such well-attested phenomena as free-standing
words evolving into grammatical markers. and lose and redundantly organized
discourse structures congealing into tight and less redundantly organized syn-
tactic constructions (see Refs a.b for some recent research). Some examples
are as follows:
1 The future tense marker in many languages is grammaticized Irom free-
standing words for such things as volition or movement to a goal. So in
English the original verb was wl, as in "I wiU to happen," and this became
grammaticized into " wl happen" (with the volitional componen!
"bleached" out). Similarly, the original use of go was for movement ("J'm
going to the store") and this was gramrnaticizied into "I'm going to die some
day" (with the movement bleached out).
2 The English past perfective, using have, is very likely derived from sentences
such as " have a brokenfinger" or " have theprsoners bound" (in which have
is a verb of possession) into something like "/ have broken afinger" on which
the possession meaning of have is bleached out and it only now indicates
perfective aspect).
3 English phrases such as "on the top of" and "in the side of" evolved into "on
top of" and "inside of" and eventually into "atop" and "inside." In some lan-
guages relator words such as these spatial prepositions have also become
attached to nouns as case markers - in this instance as possible locative
markers.
4 Lose discourse sequences such as "He piiiled the door and it opened"
may become syntacticized into "le pulled the door open" (a resultative
construction).
5 Lose discourse sequences such as "Mi/ boyfriend. , . he plays piano . . . he
plays in a band" may become "Mi/ boyfriend plays piano in a batid." Or, simi-
l arl y. "My boyfriend . . . he rdes horses. . . he bets on them" may become "My
boyfriend. wlw rides horses. bets on them."
Chlldren's Early Syntactic Development )f
6 Similarly. i someone expresses the belief that Mary will wed John, another
person might respond with an assent "I believe that." followed by a repeti-
tion of the expressed belief that "Mary will wed John" - which become
syntacticized into the single statement " believe that Mary will wed John."
7 Complex sentences may also derive from discourse sequences of initially
seprate utterances, as in " want it... 1 buy it" evolving into "/ want to
bny it."
References
a Traugott, E.. & Heine, B. (1 991). Approaches to Grammaticalization (Vals 1 and 2). John
Benjamins.
b Hopper. R, & Traugott, E. (1993). Grammaticalization. Cambridge University Press.
make sure thal speakers do not go so far in this direction that the message becomes
incomprehensible. Grammaticalization processes are well attested in the written
records of numerous languages in t hei r relatively recent pasts, and it is a reasonable
assumption that the same processes were at work in the origin and early evolution of
language, t urni ng loosely organized sequences of single symbols i nt o grammaticized
linguistic constructions.'"
Even so. grammaticalization by itself s not enough because it does not account for
the abstractness of linguistic structures. Abstrae!ness. as Chomsky recognized in even
bis earliest writings. must be contributed by the mi nds of individual children as they
acquire the use of par t i cul ar pieces of par t i cul ar languages. It s possible - albeit very
difficult - to imagine that children make this contribution by simply linking an innate
universal grammar with the par t i cul ar st ruct ures of the part i cul ar language they are
learning. However. it is also possible. and more in accord with recent data, to imagine
t hat children make t hi s contribution in more extended developmental processes in
which they apply their general cognitive. social-eognitive, and vocal-auditory process-
ing skills to the histrica! producs of grammaticalization.'
7
Ovcrall. then, we may
hypothesize t hat human language originated ul t i mut el y from a species-unique biol-
gica! adapt at i on for symboiic communi cat i on. but the act ual grammatical st ruct ures
of modern languages were humanl y created t hrough processes of grammaticalization
duri ng particular cul t ural histories, and t hrough processes of cul t ur al learning.
schema formation. and structure combining during particular i ndi vi dual ontogenies.
Outstanding Questions
When chi l dren imitativcly l earn some compicx l i ngui s t i c expression. how do f hey
come l o underst and t he communi cat i ve limclions of t l i e di l ' l crenl const i t uent s
involved-
On \ \ ha l basis do chi l dr en make analogies or form schemas as lliey abst r acl
across l l i ei r verb i s l and and ol her r el at i onal l i ngui s l i c schemas?
J4 introduction to Grammatical Development
' VVhat principies govern the ways in \vhich children combine established l i ngui s-
t i c const ruct i ons xvi t h one anot her crcatively?
Ho\v do chi l dren select \ \ hat they nred (rom all t he l anguage they liear around
thcm?
VVhat is the nat ur e of the cross-l i ngui st i c and i ndi vi dual differences t hat can be
observed i n chi l dren acqui r i ng nat ur al languages?
Acknowledgements
The aut hor \vould like to t hank Hiena Lieven. Heike Bchrcns. Holger Dicsscl, Nameera
Akhtar. and Patty Brooks for helping to devclop the ideas and the studies reportcd in
this papen
References
1 Pinker. S. ( 1 994). The Language Instincl: Hmv the \Iind C'rcntcs Language. Morrow.
2 Pinker, S. ( 1984) . Language Learnahilij and Language Dm'lopment. Harvard Universily
Press.
5 Braine. M.D.S ( 1 9 7 1 ) . On t\vo types of models of the nternalization of grammars. In
D.I. Slobn ( hd. ) . The Ontognesis of Gramimir, pp. 1 S 3-186. Academic Press.
4 Tomasello. M. ( 1 992). First Verbs: A Case Studi inEarh/ Grammatical Development. Cambridge
I.'niversity rress.
5 Licvcn. E. et al. ( 1997) . Lexically-based learning and early grammatical development.
/. CliillLuifi.. 24. 187-220.
6 Pine. J.. & Lieven. E. ( 1 997). Slot and trame patterns in the development of the determiner
category. Appl. Psyilwlmgiiist.. I K. 123-1 38.
7 Bowerman, M. (1982). Reorganizationai processes in lexical and syntactic development. In
L. Cileitman. & E. Wanner (Eds. ) , Langiuige Acqdsitlon: The State of tlie rt. pp. 319-34IS.
Cambridge Kniversity Press.
8 Bowerman. M. (1988). The "no negative evidence" problem: how do childrcn avoid con-
strucling an overgeneral grammar? In ).A. Hawkins ( Ed. ) . Explaining Language Universals.
pp. 73- l Ol . Basi l Bl ackwel l .
9 Pinker. S. ( 1989) . Learnabit and Cognition: The Acquisition of Verb-argumenl Structwe.
Harvard University Press.
10 Maratsos. M. et al. ( 1 987). A study in novel vvord learning: theproductivity of thecausative.
In B. MaeWhinney (Ed.). Medmnisms of Language Acquisition. pp. 89-1 14. Erlbaum.
11 Pinker. S. et al. ( 1987) . Productivity and constraints in the acqui si t i on of the passive.
Cognition. 26. 195-267.
12 Tomasello. M.. & Brooks. P. ( 1 9 9 8 ) . Young ehildren' s earl i est t r ansi t i ve and i nt ransi l i ve
constructions. Cognit. Litigis!..
l
>. 379-595.
1 3 Tomasello. M., i Ol gui n. R. (199 3). Twenty-three-month-old children liavc a gr ammat i cal
category o I' noun. Cot/nil. I)e\'., N. 451164.
14 Tomasello. M. el al. ( 19971. Di f f e r e nt i a l productivity in young ehi l dr en' s use of nouns and
verts. /. Cliill Laiig.. 24. 573-387.
1 5 Taylor. M.. & Ci cl nKi n. S. I 19881. Adj ecl i u' s and nouns: chi l dr en' s st ral egi es for l ear ni ng
ncu- vvords. CliildDcv.. 5 9 . 4 1 1-419.
Children's Early Syntactic Development ]&$
1 6 Akht ar . ,Y. & Tomasello, M. ( 1 9 9 7 ) . Young children's produetivi.y with word order and verb
morphology. Oev. Psuchol. i ? . 952-965.
Naigles. L (1 990). Children usesynt ax lo lean, verb meanings. /. Child Lang.. I 7. 557~ 5-4
Olguin. R.. & Tomasello. M. ( 1 9 9 3 ) . Twenty-five-month-old children do not have a grilm-
mat i cal category of verb. Cognit. On
1
.. X. 245-272.
19 Val an. V. ( 1 9 9 1 i . Syntactic subj ect s in the early speeeh of American and I t a l i a n chi l dr en
C ocinilMii. 40. 2 1 - 8 1 .
20 Boerer. H.. & Wexler. K. ( 1 9 9 2 ) . Bi -uni quc relalions and the mat ur al i on of gr ammat i cal
pri nci pi es. .Vn. hing. l.inijitixl Tlieori/. I). 147-187.
21 Hyams. \. ( 1 9 9 4 ) . Non-discreteness and vari at i on in child l anguage: i mpl i cat i ons for
Principie and parameter models of l anguage acqui si l i on. In V. I.evv ( Kd . l . Ollicr Clnhtn-n.
Other Languages. pp. 1 1-40. Erl baum.
Bowerman. M. ( 1 9 9 0 ) . Mapping ( hemat e roles onto syntactic functions: are chi l dren
clped by mnat e l i nki ng rules? Lingitistics. 2<S'. 125 3-1 289.
2 3 Slobin. I). 11997) . On the origin of grammat i eal i zable noons: beyond the i ndi vi dual mi nd.
1. Slobin ( Ed. ) . The Cross-litigiiistic Stud/ of Languagc AcquisitwiHVl. 5), pp. 1 1 9-15 3.
Erl baum.
ryer. M. ( 1 9 9 7 ) . Are grammat i cal relalions uni versal ? In J. Bybee et al.. ( Eds. ) . /fssii/.v on
Language Fwiclion tmtt Langiutge Ti/pe, pp. 11 5-144. John Benjarains.
2^ Langacker. R. ( I 991) . Fmmilutioiis of C ognit ive Grammar (\'ols I and 2) Stanlord Tri i versi t v
Press.
26 Tomasello. M. ( Ed. ) ( 19981. The .Vu> rwliolog, of Language Cognitive and Fimclional
Approaches. Erl baum.
Tomasello. M. et al. ( 1 9 9 3 ) . Cul t ur al learning. Bchm: Bram ScL. / 6. 495-511.
28 Lieven. E. ( 1 9 9 7 ) . Vari at i on in i cross-linguislic context. In D. Slobin ( Ed. ) . The Cross-
linguistic Stud/ of Lanauage Acquisition (Vol. 5). pp. 2 3 3-287 Erlbaum
29 Tomasello. M.. & Brooks. P 11999). Early syntactic development. In M. Barrett (Ed.). Tlw
Development oj Language. pp. 1 61-190. l'CL Press.
30 Gentner. D.. & Markman. A. ( 1 9 9 7 ) . St ruct ure mapping in analogv and similaritv. Am
Psi/chol. 52. 4-1-56.
31 Marchman. V.. & Bates. E. ( 1994) . Cont i nui t y in lexical and morphological development:
a test o the cri t i cal mass hypothesis. /. Child Lana.. 21. 339-366.
32 Bybee. |. ( 1985) . Morplwlogu. John Benjamins.
3 3 Diessel. H.. & Tomasello. M. VVhy complement clauses do not have a /m-complementizer in
early child language. Berkeleii Limmislics Socielii (in press).
Deacon. T. 11 998). The Sumbolic Species. Harvard l' niversity Press.
3-5 Traugott. F... & Ileine. 15. ( 1991) . Approaches to Grammatimli:ation (Vols 1 and 2). John
Benjamins.
36 Civn. T. 1 1 995) . Fimctitiniilixm imd C.nunina. John Benj ami ns
Tomasello. M. , 1999) . The Cultural Origins of Uwnan Cogniton. Harvard Lni versi l v
3H I ngham. R. ( I 99 5 1. Cr i t i cal i nl l uences on the acqui si t i on of verb t r ans i t i vi l y. I n I). Messer,
& t. Turner ( Eds . l . Critical Infhences o, Child Langnage Acquisition and Developinenl. pp.
19-38. Maximillian Press,
-iy Dodson. K. . & Tomasello. M. ( 1 9 9 8 ) . Acqui ri ng the Iransitive construction in Hngl i sh:
Ul e role of ani macv and pronouns. /. ChihlLang.. 25. 555-574.
40 Brooks. P, t Tomasello. M. ( 1 999). Young children learn lo produce passix'es l l h nonce
\ erbs. )ev. l'si/cliol., J 5 . 29 44.
Akht ar . \. ( 1 9 9 9 ) . Acc| u r i ng basic vvord order: evidence for dal a- dr i ven l ear ni ng o
syntactic st ruct ure. /. Child .mu/.. 26. 261-378.
\S>6 Introduccin to Gratnmatical Developmetit
42 Hermn. R. ( 1 99 5). Marki ng verb I r ansi l i vi l y in Hebrew-speaking chi l dren. /. C' /n' W Lniiij..
20. Ci41-670.
4? l. ewis. I... & Tomasello. M. C' hi ki ren' s i hi l i t y lo gfi i erai i xe no\'el verbs useci in impeailives.
l ' npubl i sher manuscri pt .
44 Cliildt-rs. ].. & Tomiisello. M. Spanish-speaking childrcn' s ahi l i l y lo generalice nonce \-erhs
lo nevv constructions. Submitled.