A CROSS-LINGUISTIC STUDY OF EARLY LEXICAL DEVELOPMENT
Maria Cristina Caselli
National Council of Research, Institute of Psychology,
Rome
Elizabeth Bates
University of California, San Diego
Paola Casadio
National Council of Research, Institute of Psychology,
Rome Clinica e Centro di Ricerca Santa Lucia
Judi Fenson
Children's Hospital and Research Center, San Diego
Larry Fenson
San Diego State University
Lisa Sanderl
National Council of Research, Institute of Psychology,
Rome
Judy Weir
Children's Hospital and Research Center, San Diego
This research was supported by the John D. and Catherine T. MacArthur Foundation
Research Network on Early Childhood Transitions, by NIDCD grants DC01289
"Origins of Communication Disorders", and DC00216 "Cross-linguistic
studies of aphasia". The manuscript was completed while Bates was a
visiting researcher at the Institute of Psychology, National Council of
Research, at the Via Nomentana site directed by Virginia Volterra. We are
grateful to Virginia Volterra, Elena Pizzuto and Alison Gopnik for comments
on an earlier version of the manuscript. Please address all inquiries to
Elizabeth Bates, Center for Research in Language, Dept. 0526, University
of California at San Diego, La Jolla, CA 92093-0526.
ABSTRACT
Cross-linguistic studies have shown that children can vary markedly
in rate, style and sequence of grammatical development, within and across
natural languages. It is less clear whether there are robust cross-linguistic
differences in early lexical development, with particular reference to the
onset and rate of growth in major lexical categories (e.g., nouns, verbs,
adjectives and grammatical function words). In this study, we present parental
report data on the first stages of expressive and receptive lexical development,
for 659 English infants and 195 Italian infants between 8 - 16 months of
age. Although there are powerful structural differences between English
and Italian that could affect the order in which nouns and verbs are acquired,
no differences were observed between these languages in the emergence and
growth of lexical categories. In both languages, children begin with words
that are difficult to classify in adult part-of-speech categories (i.e.,
"routines"). This is followed by a period of sustained growth
in the proportion of vocabulary contributed by common nouns. Verbs, adjectives
and grammatical function words are extremely rare until children have vocabularies
of at least 100 words. The same sequences are observed in production and
comprehension, although verbs are reported earlier for receptive vocabulary.
Our results are compared with other reports in the literature, with special
reference to recent claims regarding the early emergence of verbs in Korean.
It is now clear that patterns of language learning can vary markedly across
different language com-munities. Thanks in large measure to a generation
of pioneering studies by Dan Slobin and his colleagues, striking variations
in the patterning of early grammar have been documented (Slobin, 1985, 1992;
see also Bates & Marchman, 1988; Braine, 1976; MacWhinney & Bates,
1989). Although children do go beyond their data and produce creative errors
in every language studied to date, their first sentences are always tailored
to and shaped by the structural properties of their native language-from
the rich inflectional systems of Turkish and Greenlandic Eskimo to the austere
structure of Chinese, which provides no grammatical inflections of any kind.
There is, for example, no "universal tele-graphic stage", and
no single order in which gram-matical structures of a particular type are
acquired.
Much less is known about cross-linguistic variation in early lexical development,
from first words to the onset of word combinations. However, studies of
children exposed to English lead us to expect sub-stantial variability in
this stage of development as well. There are now many studies documenting
large individual differences in rate of language learning (Bloom, Lightbown,
& Hood, 1975; McCarthy, 1954; Miller & Chapman, 1981). In the largest
study to date, Fenson, Dale, Bates, Reznick and Thal (1994) report the first
stages of lexical development in 1,803 normally developing children between
8 and 30 months of age, demonstrating massive variation in onset time and
rate of growth in all aspects of early language and communication (see also
Fenson et al., 1993; Bates, Marchman et al., 1994). Although many children
in their sample offered no systematic evidence for word comprehension before
10 months of age, the parents of some 10-month-old infants report receptive
vocabu-laries of more than 100 words. In the same vein, some children are
reported to produce little or no meaningful speech before 16 months of age,
while others displayed expressive vocabularies of more than 100 words at
or before 16 months. Such large individual differences in rate of language
development present serious problems for the idea that all normal children
develop on a single maturational timetable.
Another line of evidence against this universal timetable comes from studies
documenting individual differences in "style" of language learning,
qualitative variations that are difficult to explain in terms of linguistic
input and/or rate of maturation (Bloom, Light-bown, & Hood, 1975; Nelson,
1973; for reviews, see Bates, Bretherton, & Snyder, 1988; Bates, Dale,
& Thal, 1994; Nelson 1981; Goldfield & Snow, 1985; Pine & Lieven,
1990; Shore, 1995). One stylistic extreme is characterized by an "analytic"
approach to language, with emphasis on words for common nouns early in the
one-word stage (with little or no use of multiword "formulas"),
and an abrupt transition into combinatorial speech marked by "telegraphic"
combi-nations of nouns and other content words. The other extreme is characterized
by an "holistic" or "formulaic" approach to early language,
with hetero-geneous voca-bularies from the very beginning of meaningful
speech (including production of multiword formulae), and use of pronouns
and other grammatical function words (albeit in relatively "frozen"
forms) from the beginning of multiword speech. Bates et al. (1988) have
argued that these stylistic variations are based upon at least two partially
dissociable mechanisms for language learning: a segmenting "parts before
the whole" mechanism, and a suprasegmental "whole before the parts"
mechanism. Both these learning mechanisms are necessary for normal language,
but for a variety of reasons (matura-tional; environmental; perhaps temperamental)
some children rely more on one than the other in the early stages of language
learning.
In this paper, we will present evidence on the composition of early vocabulary
in 195 Italian children between 8 and 16 months of age, compared with 659
English children in the same age range from the Fenson et al. study described
above. These two languages vary along some critical dimensions, with particular
empha-sis on the relative salience of nouns and verbs. Hence they provide
a fair test of the idea that children can learn nouns or verbs in any order,
depending on their language input (Gopnik & Choi, 1990, in press). Our
results will show that this hypothesis is incorrect, at least for English
and Italian. We will demonstrate a strikingly similar distribution of word
types in young children exposed to these two languages, characterized by
the dominance of nouns and the virtual absence of verbs and adjectives up
to the point where total vocabulary exceeds 100 words. Based on these results,
we will propose as a working hypothesis for future cross-linguistic research
that there may be a universal sequence in development from nouns to verbs,
a sequence that cannot be reversed by variations in the relative salience
of nouns and verbs in the input lan-guage.
There are several theoretical and empirical argu-ments in favor of the proposed
noun-verb sequence in its current form, including the following.
Theoretical arguments for the early appearance of nouns include a proposed
constraint on early word learning that Markman (1989) calls the "Whole
Object Constraint," referring to a (presumably innate) tendency for
young children to assume that new words will refer to whole objects, and
not to subparts of that object (e.g., the bunny's ears), or to the actions
or changes of state in which the objects participates (e.g., hopping, holding
still, or nibbling a carrot). For obvious reasons, the Whole Object constraint
would favor the early acquisition of nouns over other lexical items.
Theoretical arguments for the late appearance of verbs were first outlined
in detail by Gentner (1982), who claimed that the semantic structures underlying
verbs and other predicates are inherently more complex and open-ended than
the structures which define noun meaning (see also Clark, 1983). O'Grady
(1987) pro-vides some related arguments based on the logical underpinnings
of nouns and verbs. He points out that nouns are typically used as arguments
or "primaries", referring to some entity or class of entities.
Verbs and adjectives are most often used as predicates or "secondaries",
to indicate an action, state or relationship involving one or more arguments
or "primaries". If a child is capable of using verbs and adjectives
in a meaningful way, then s/he must have some kind of nominal argument in
mind, i.e., "the thing that is pretty", "the thing that is
gone", "the thing that is possessed". For this reason, acquisition
of verbs and adjectives cannot proceed until the child has mastered enough
nouns to support predication.
These theoretical arguments presuppose a trans-parent mapping from Objects
--> Arguments --> Nouns, complemented by an equally transparent mapping
from Actions/States --> Predicates --> Verbs. In fact, the correlations
presupposed by this analysis are far from perfect (e.g. nouns can be used
to set up a relationship, as in "Mary is an engineer", and verbs
can be used to name an activity or event, as in "Mary loves to dance"),
and may be even weaker in the language produced by young children (e.g.,
"Hot" may be used a name for all hot objects-see Methods). Furthermore,
even if it could be shown that children first use verbs to express predication,
it does not follow that the noun arguments of those verbs have already been
expressed in speech. Such arguments could be indicated through pointing,
or presupposed with no explicit signal of any kind (cf. Goldin-Meadow &
Mylander, 1985; Greenfield & Smith, 1976; Iverson, Capirci, & Caselli,
1994; Volterra & Caselli, 1986).
Although there are flaws in the logical argument for a noun-verb sequence,
empirical support is readily available for children acquiring English as
their native language. For example, Bates, Marchman et al. (1994) have reported
that lexical verbs comprise less than 5% of total vocabulary for English-speaking
children with vocabularies under 50 words, compared with approxi-mately
14% for children with vocabularies of 500 words or more (see also Bates
et al., 1988). In other words, lexical verbs are relatively late, at least
in this language. However, the relative lateness of verbs does not necessarily
imply an early bias towards nouns. As noted earlier, some children do produce
a very high proportion of names for common objects during the first phase
of word learning, but others start out with a heterogeneous set of "non-nominals"
that includes proper names, routine expressions (e.g., "Hi" and
"Night-night"), and functors that serve a specific social purpose
(e.g., the word "up", uttered while extending the arms to be picked
up, or the words "more", "mine!" and "no"
to accompany requests, demands and re-fusals). Because there is so much
variation in the ratio of nouns to non-nominal expressions during the first
stages of lexical development, the evidence to date for English-speaking
children may be best interpreted as "early verb avoidance" rather
than "early noun pre-ference."
This brings us to a consideration of cross-linguistic evidence against the
hypothesized noun-verb sequence. The proposal that verbs are universally
late has come under fire in some interesting and provocative papers by Gopnik
and Choi (1990; in press). The thrust of this work is well illustrated by
the title of their most recent paper: "Names, relational words and
cognitive develop-ment in English- and Korean-speakers: Nouns are not always
learned before verbs." In both papers, Gopnik and Choi present some
detailed logical and empirical arguments against the notion that all children
neces-sarily acquire nouns before verbs. They suggest that this proposed
universal may be an epiphenomenon of two methodological facts: (1) the fact
that most studies have been based on English, and (2) the fact that so many
studies have been based on parental report and/or on data collection in
cultural contexts that emphasize object naming. They present data from free
speech and parental interviews for a small sample of children ac-quiring
Korean as their native language, showing that most of the words that these
children use in the early stages of lexical development are non-nominals
(broad-ly defined), including a higher proportion of verbs (narrowly defined)
than one sees in English-speaking children observed under similar conditions.
This differ-ence between English and Korean is explained by a combination
of linguistic and cultural differences, with particular emphasis on the
fact that Korean is an SOV language (placing the verb in a salient position),
with extensive use of subject and object omission (which means that verbs
are often the only content word in sentences spoken to small children).
In other words, they have proposed that cross-linguistic differences in
the salience of verbs can influence the order in which these categories
emerge during the first stages of language learning (although it is worth
noting that their evidence for early verbs in children is actually based
on results for non-nominals, broadly defined to include lexical verbs together
with other non-nominal cate-gories; we will return to this point later,
comparing verbs with the broader non-nominal category in our own data).
We ourselves are quite sympathetic to cross-lin-guistic research, and to
the argument that linguistic input can have a powerful impact on language
and (perhaps) cognitive development (see MacWhinney & Bates, 1989).
Indeed, in an earlier study we used parental-report instruments to investigate
the growth of nouns, verbs and other categories in 14-month-old Italian-
and English-speaking children, and found what looked like evidence in favor
of a cross-linguistic difference in the age at which verbs begin to appear
(Bates, Caselli, & Casadio, 1990). However, that study had methodological
limitations that led to our present comparative analysis, with rather different
results.
We have persisted in this comparison of English and Italian, because we
believe that these languages display striking contrasts in the relative
salience of nouns and verbs (albeit somewhat different contrasts from those
that Gopnik and Choi report for English and Korean). As such, English and
Italian represent a strong test of the hypothesis that cross-linguistic
differ-ences in the input to children can affect the order in which nouns
and verbs are acquired. This claim rests on several interacting contrasts
between English and Italian, briefly summarized as follows.
(1) Word order variation. The order Subject-Verb-Object (SVO) is
preserved quite rigidly in Eng-lish, compared to most of the world's languages.
To be sure, there are some alternative word orders in English, including
object-fronting in Wh-questions (e.g. "Who did you see?"), and
auxiliary fronting in yes-no questions (e.g. "Are you coming?").
Since questions are plentiful in the speech addressed to children, this
means that English children are exposed to some variation in basic SVO.
Nevertheless, deviations from SVO are far more common in Italian. In fact,
all possible orders of subject, verb and object can and do occur in informal
Italian speech, in all sentence types (i.e., declaratives, interrogatives
and imperatives). In conversations among Italian adults, SVO is still the
predominant order (i.e., the most frequent order, and the one used in pragmatically
neutral conditions). How-ever, Bates (1976) has shown that the speech addressed
to young Italian children is evenly divided between subject-initial and
subject-final forms (when subjects are expressed at all-see below). Hence
Italian infants receive even more word order variation than we would expect
based on analyses of adult speech, due perhaps to the fact that word order
variation is used for empha-sis and clarification, two very common functions
in child-directed speech.
(2) Subject omission. English is a language
in which subjects are obligatory in free-standing declara-tive sentences,
including empty subjects like the "it" in "It is raining."
This property (which English shares with some of the other Indo-European
languages, e.g., German and French) contrasts with so-called "pro-drop"
or null-subject languages like Spanish, Chinese or Italian, where subjects
are only mentioned if they are new and/or emphatic information (and not
mentioned at all in constructions like "Piove" - "(It) is
raining"). Bates (1976) reports that subjects are omitted approxi-mately
70% of the time in informal conversation among adults, and in speech to
very young children. This means that verbs are the first element in the
sentence a high proportion of the time. It also means that verbs constitute
a much higher proportion of the content words that Italian children hear,
compared with their English counterparts.1
(3) Clitic pronouns. English uses full pronouns, in all sentence positions.
By contrast, Italian distinguishes between phonologically strong full pro-nouns
(which are always used for contrast or emphasis) and phonologically weak
clitic pronouns (the default way of referring to an element that is already
established in discourse). Clitic object pronouns must occur immediately
before the verb in a declarative or interrogative form (as in "Lo voglio"
- "It (I) want"), forming an SOV or (when subjects are omitted)
an OV structure. Because such sentences are very common in the speech addressed
to young children, this means that the input to Italian children often approximates
the input to children who are learning an SOV language (e.g., Korean).
(4) Agreement marking. Verbs carry a relatively small number of agreement
contrasts in English com-pared with other Indo-European languages. For example,
the English present indicative conjugation contains only contrast (eat vs.
eats), compared with six contrasts in Italian, one for each person and number.
As a result, the English verbs carries very little information about the
identity of the subject (i.e., "Who did what to whom"); by contrast,
Italian listeners can and do rely on agreement contrasts to assign sentence
roles (Devescovi, D'Amico, Smith, & Bates, 1993; MacWhinney, Bates,
& Kliegl, 1984).
Because of the first three properties (i.e., rigid word order, obligatory
subjects and full pronouns), the verb is almost always located somewhere
in the middle of the sentence in English, where it is hardest to hear and
hardest to remember. And because of the fourth property (minimal agreement
marking), the verb also carries less information in English than it does
in many other languages. Putting these facts together, we should not be
surprised to find that verbs are a relatively late acquisition for English-speaking
children. By contrast, verbs are very salient in Italian, and especially
salient in the speech addressed to small Italian children. Hence Italian
ought to provide a good test of the robustness and cross-linguistic generality
of the noun-to-verb shift.
To confirm or disconfirm the proposed noun-verb sequence in English and
Italian, we need to quantify the categories "noun" and "verb".
Can we assume that adult definitions of "noun" and "verb"
are appropriate for use with small children? Bates, Camaioni and Volterra
(1975) have noted that children often start their linguistic careers with
words like "Vroom" that are difficult to classify into adult part-of-speech
categories (see also Ninio, 1993, 1994). Indeed, it could be argued that
none of the adult part-of-speech categories are present in the first stage
of language learning (cf. Tomasello, 1992; Ninio, 1994). One-year-olds often
use words in ways that deviate from adult patterns of usage for the very
same word, e.g., the "Hot" example described above (Volterra,
1979; Volterra & Caselli, 1986). Because there is no necessary one-to-one
relationship between adult part-of-speech categories and the way that words
are used by individual children, it is fair to ask whether anything can
be learned from a study of vocabulary composition that relies on adult classifications.
Our solution to this problem is to analyze the child's reported vocabulary
from the point of view of adult part-of-speech categories. An item is classified
as a "noun", "verb" or "function word" if
that item func-tions as a noun, verb or function word in the adult language.
However, we do not assume that children in this age range necessarily represent
or recognize categories like noun, verb, or grammatical function word. Instead,
we treat adult part-of-speech categories as independent variables. That
is, we treat these categories as a summary of the child's linguistic input,
including differences between Italian and English in the phonological, semantic,
morphological and syntactic properties of nouns and verbs. To the extent
that children treat nouns differently from verbs, content words differently
from grammatical function words, and so forth, we can assume that they have
been affected by these differences. Hence developmental sequences in the
acquisition of part-of-speech types may be taken to reflect changes in the
child's ability to deal with these types. They do not necessarily reflect
the emergence of explicit or implicit categories like noun or verb from
the child's point of view. In the same vein, cross-linguistic differences
in vocabulary compo-sition (with vocabulary size held constant) can be taken
to reflect the child's sensitivity to variations in the nature of the input
language. They do not necessarily reflect cross-linguistic differences in
the "moment" (if there is a moment) at which such categories emerge
in the mind of the child.
METHOD
Subjects
English-speaking subjects were 659 infants be-tween 8-16 months of age whose
parents participated in a large norming study of the Macarthur Communicative
Development Inventories, conducted in San Diego, Seattle and New Haven (Fenson
et al. 1993, 1994). As described in more detail in Fenson et al. (1993),
the sample represents families across the socioeconomic spectrum within
the United States, but is dispro-portionately weighted toward families with
a high-school education or better (i.e., 79% of the mothers report at least
12 years of formal education). The sample was screened (based on a parental
questionnaire about family and medical history) to exclude children with
serious medical problems, neurological deficits and/or mental retardation.
Children with significant exposure to languages other than English were
also excluded from the sample. There were at least 30 boys and 30 girls
in each one-month age group. For a detailed breakdown by sex and age, see
Table 1.
TABLE 1:
AGE BY GENDER BREAKDOWN FOR EACH LANGUAGE GROUP
ENGLISH ITALIAN
Age in months Male Female Male Female
8 32 33 11 8
9 37 32 4 10
10 35 32 10 11
11 38 46 7 13
12 45 41 16 18
13 40 36 7 8
14 42 42 7 10
15 32 33 15 12
16 33 30 13 15
TOTAL 334 325 90 105
There were 195 Italian subjects spread across the 8 - 16-month age range,
with approximately equal distri-bution of males and females at each age
level (i.e., a chi-square test for distribution of males vs. females across
the 9 age levels failed to reach significance-see Table 1 for a breakdown
of the Italian sample by age and sex). With one exception, the children
and their families were all residents of cities in the central to northern
regions of Italy (52% from central regions, including Rome; 48% from the
northern areas of Italy). Parents filled out a basic information sheet based
on the questionnaire used in the MacArthur norming sample for English, including
information on sex and birth order, the child's medical history, exposure
to lan-guages other than Italian, and parental education and occupation.
The same exclusionary criteria were adopted for both the English and the
Italian samples. Similar to the U.S. sample, a full spectrum of socio-economic
groups were represented among the 195 families in the study, but the sample
was dispro-portionately weighted toward families in which the parents had
a high-school education or better.
Materials
The English-language version of the CDI:Infants is composed of two sections.
Part I is a checklist of 396 words that are among the first to appear in
the vocabularies of young English-speaking children. Next to each word,
the parent is asked to indicate if the child (a) understands that word,
and (b) understands and produces that word. The checklist is divided into
19 semantic categories: sound effects (e.g., "moo", "vroom"),
animal names, vehicle names, toys, food items, articles of clothing, body
parts, furniture, house-hold objects, outside things and place to go, people
(including proper nouns), routines and games (e.g., "peekaboo"),
verbs (called "action words" on the CDI form), words for time,
adjectives (called "descriptive words" on the CDI form), pronouns,
question words, prepositions, and quantifiers. All forms are presented in
their "citation form" (e. g., verbs are listed as stems); Part
II is a checklist of 63 communicative and/or symbolic gestures that also
develop in this age range. Data for Part II of the Infant Scale will not
be discussed in this paper.
As described in more detail by Fenson et al. (1994), the English and Italian
versions of the MacArthur Communicative Development Inventories have been
developed in tandem across a twenty-year period, from open-ended interviews
used in both countries (e.g., Bates, Benigni, Bretherton, Camaioni &
Volterra, 1979; Bates, Camaioni & Volterra, 1975) through various intermediate
versions of increasing specificity, with more and more reliance on recognition
memory in a checklist format. The content of the Italian MacArthur CDI reflects
many years of parent report, home obser-vation and laboratory studies on
the acquisition of Italian, but the current version is modelled as closely
as possible on the English version in terms of overall format, number and
type of lexical categories, and number of items. The Italian word checklist
includes 401 words divided into the same 19 semantic categories listed above
for English. For all content word categories (from sound effects through
adjectives), the number of items within each category is identical in English
and Italian. The small difference in number comes entirely from function
word categories that are necessarily longer in Italian (e.g., four forms
of the definite article instead of one).
Procedure
Parents of the children in the U.S. norming study were contacted through
local doctors' offices and day care programs, and were asked to complete
the inven-tory and accompanying family history questionnaire, and return
both in a postage-paid envelope. Approxi-mately 30% of the families that
were contacted completed and returned a questionnaire. (Unfortu-nately,
we have no way of knowing whether there are demographic differences between
this sample and those families who declined to participate in the project.)
For English, where the scale is printed on machine-scan-nable forms, all
returned questionnaires were first subjected to visual analysis (to assure
that parents had followed directions and forms were appropriate for machine
scanning), and then scored mechanically.
In the Italian study, families were contacted through area pediatricians
and staff members within the national health care system for each region.
The questionnaire was accompanied by a letter to the families and the family
and medical history question-naire. Approximately 50% of the families who
were contacted completed and returned the questionnaire (we have no information
on the demographic characteristics of families who declined to participate).
The Italian data were entered into the computer manually, within a scoring
program developed for this purpose.
In both English and Italian, the forms provided an opportunity for parents
to write in comments or add additional words, but these word additions were
not used in any of the counts described below, for either language. This
was also true for the proper noun category. For example, the child would
be given a score of 1 item if the parents checked the word for "grandfather",
but no additional credit was given if the parent wrote in names for two
different grandfathers.
Data Reduction
Following the procedures outlined by Bates, Marchman et al. (1994), we calculated
a series of proportion scores to examine the composition of production and
comprehension vocabulary, as follows:
(1) Percent nominals, broadly defined to include
common nouns, proper nouns, places to go, and sound effects that are often
used by young children to refer to animals and objects;2
(2)Percent common nouns, more stringently de-fined
to include only words from the adult language that stand for concrete objects
(animal names, vehicles, toys, food and drink, clothing, body parts, furniture
and rooms, small household objects); the nominal categories excluded from
this count are names for people, sound effects, and "places to go";3
(3)Percent proper nouns and terms for people;4
(4)Percent sound effects (e.g., "vroom", "meow");
(5)Percent routines, a heterogeneous category that includes social words
like "hi" and "bye", daily routines or events like "breakfast"
or "nap", and familiar com-mands like "don't";
(6)Percent verbs;
(7)Percent adjectives;
(8)Percent predicates (verbs and adjectives com-bined);5
(9)Percent grammatical function words, which includes pronouns, question
words, prepositions and quantifiers.6
In calculating these scores (and in the original
design of the Italian and English scales), every effort was made to use
comparable criteria for word classi-fication in each language. We also tried
to stick closely to adult-based classifications, consistent with our claim
that adult part-of-speech categories should be treated as independent variables
in the analysis of child language. Nevertheless, because the adult categories
themselves are idealizations, we were forced to make some arbi-trary (and
arguable) classifications in many cases, especially for routines, sound
effects and other marginal categories (i.e., items that one cannot simply
look up in a dictionary). There are a few items that ended up in different
categories for English vs. Italian, even though they have a similar function
in each language. For example, the word "uh-oh" is classified
under sound effects in English, while its nearest equivalent "bum"
is classified as a routine in Italian. In the same vein, the word "allgone"
was classified as an adjective in English (because of the sentence contexts
in which it is typically used by adults), while its equivalent in Italian
(a frozen phrase "Non c'è più"-"There is no
more") was classified among the routines.7
All of these classi-fications are probably best viewed as approximations,
so that our quantitative results (based on category proportion scores) must
be considered together with the qualitative item analyses presented later.
RESULTS AND DISCUSSION
We will begin with a global comparison of age-related changes in vocabulary
comprehension and production in the two language groups. Then we will examine
similarities and differences between English and Italian in the composition
of vocabulary, with children grouped according to vocabulary size rather
than age. This is the section that will provide our most important test
of the hypothesis that verbs can develop before nouns in a language that
renders verbs particu-larly salient. Finally, we will undertake a qualitative
comparison of the first 100 words acquired by English vs. Italian children,
in production and in compre-hension. This analysis will not change our conclusions
about the late onset of verbs, but it will clarify and illustrate some item-specific
effects that can emerge as a consequence of linguistic and cultural differences.
Age-related changes in vocabulary size
To determine whether the two groups differ in rate of vocabulary development
from 8 - 16 months, we carried out two separate 2 (language) by 9 (age in
months) between-group analyses of variance, for total comprehension and
total production, respectively.
In the analysis of word comprehension, there was a significant main effect
of age (F(8,853) = 65.05, p .0001), but no main effect of language group
(F(1,853) = 0.485, n.s.) and no interaction between language group and age
(F(8,853) = 1.86, n.s.). There is indeed substantial growth in receptive
vocabulary size across this age range, from a mean of 36 words at 8 months,
to 86 words at 12 months, to 191 at 16 months. This developmental increase
in word comprehension is illus-trated in Figure 1 (although there was no
significant effect of language, we have graphed data for each group separately,
for comparison with the data on production).
Figure 1
There were also large individual differences in rate of growth for word
comprehension, in both English and Italian. Across the two groups, the mean
number of words comprehended was 105, but the standard devia-tion was 86.5,
and we observed a range from 0 to 396-in other words, from no words at all
to all of the words on the list! The mean was quite similar for both language
groups (105 for English vs. 107 for Italian), as was the standard deviation
(86.5 for English, 83.3 for Italian). Hence these two populations are quite
comparable in size, rate of growth, and amount of variation observed for
word comprehension.
By contrast, the analysis of word production did reveal a consistent difference
in rate of growth for English vs. Italian. There was a significant main
effect of age (F(8,853) = 33.12, p .0001), but also a signi-ficant effect
of group (F(1,853) = 15.75, p .0001), and a significant age by group interaction
(F(8,853) = 3.37, p .001). Figure 1 shows that Italian children lag behind
the English group in total vocabulary size at most ages, although the difference
is most evident at the highest levels. At 8 months of age, the average score
for English children is 1.8 words, compared with 1.1 for Italian. By 12
months, the English average is 10 words, compared with an Italian mean of
9. The groups move further apart at 15 months, when the mean for English
is 48 words compared with 22 for Italian, and further still at 16 months,
where the corresponding numbers are 64 and 27. There are substantial individual
differences in rate of growth within each language group, but the variance
is larger in the English sample (English standard deviation = 39.39; Italian
standard deviation = 18.12).
This apparent cross-linguistic difference in rate of growth and range of
variation for vocabulary produc-tion may be an artifact of differences in
sample size. As Figure 1 clearly indicates, the data for vocabulary production
are characterized by floor effects across the period from 8 - 16 months,
with many children who produce almost no speech at all. In a situation like
this, where the most of the sample is still "hugging the floor",
there is nowhere for the variance to go but up. As sample size increases,
we increase the probability of picking up children with exceptionally high
scores, but we cannot pick up additional children with exception-ally low
scores because one cannot go any lower than zero (and zero is not an unusual
score). Our sample includes 195 Italians-a very large number within the
literature on child language, but a much smaller number than the 659 subjects
that we have for English. Thus, our probability of picking up exceptionally
high scores is approximately 3.4 times greater in English than it is in
Italian; our probability of picking up exceptionally low scores is also
3.4 greater in English, but 3.4 times zero is still zero. As a result, there
is an inevitable upward skewing of the mean as sample size goes up. In view
of this statistical confound, we suspect that our data do not support a
true "English advantage" in productive vocabulary. However, because
there is a consistent group difference in vocabulary size, all the analyses
that follow will be based on language level rather than age, so that English
and Italian children are matched for total vocabulary.
Cross-linguistic and developmental differences in vocabulary composition
For analyses of word production, children were divided into six groups based
on their total vocabulary size: (1) zero words, (2) 1 - 5 words, (3) 6 -
10 words, (4) 11 - 20 words, (5) 21 - 50 words, and (6) more than 50 words.
A chi square test confirmed that the two language groups are evenly balanced
across these six vocabulary levels. For analyses of word compre-hension,
children were also divided into six groups based on vocabulary size, as
follows: (1) 0 - 20 words, (2) 21 - 50 words, (3) 51 - 100 words, (4) 101
- 150 words, (5) 151 - 200 words, and (6) more than 200 words. Again, a
chi square test confirmed that the English and Italian groups are evenly
balanced across vocabulary levels.
Table 2 reports mean proportion scores in produc-tion for each language
and vocabulary group; Table 3 contains the corresponding subgroup means
in comprehension. For each modality (production and comprehension), each
of these proportion scores was analyzed within a Language by Vocabulary
Level analysis of variance (2 ¥ 6 for analyses of compre-hension; 2
¥ 5 for analyses of production, excluding children with zero words
for obvious reasons). We will start below with analyses of word production
(following Bates, Marchman et al., 1994, for English), and then move on
to analyses of word comprehension (which have not been presented for English
or Italian in any of our previous work).
TABLE 2: COMPOSITION OF PRODUCTION VOCABULARY
AS A FUNCTION OF LANGUAGE AND VOCABULARY SIZE
Variable Language Number of Words in Production Vocabulary:
1-5 6-10 11-20 21-50 >50 Total
% Nominals English 80.4 75.4 70.7 72.8 73.6 75.5
Italian 91.0 84.5 74.7 70.4 72.6 82.0
% Common Nouns English 16.4 22.9 32.9 44.5 54.1 30.2
Italian 20.6 28.1 31.4 36.5 46.2 28.8
% People English 32.3 24.8 16.5 10.7 6.5 21.1
Italian 35.1 35.4 19.6 16.9 11.1 27.6
% Sound Effects English 30.7 27.1 20.4 15.5 7.7 22.7
Italian 35.3 21.0 23.6 16.5 11.8 25.3
% Routines English 14.2 19.5 18.7 15.0 9.1 15.5
Italian 7.8 13.3 20.7 20.1 16.4 13.8
% Verbs English 0.5 1.1 2.1 2.7 6.8 2.1
Italian 1.3 0.7 2.8 4.1 4.5 2.3
% Adjectives English 1.7 0.5 2.7 4.3 4.8 2.5
Italian 0.0 0.4 0.5 1.0 2.1 0.5
% Predicates English 2.2 1.5 4.8 7.0 11.6 4.6
(Verb + Adj) Italian 1.3 1.0 3.3 6.0 6.6 2.8
% Function Words English 3.2 3.6 5.8 5.2 5.4 4.4
Italian 0.0 1.1 1.3 3.5 4.0 1.4
Changes in the composition of expressive vocabulary
As Table 2 shows, the category Nominals (broadly defined) accounts for 75.5%
of total production voca-bulary in the English sample, compared with 82%
in Italian. Under the broad definition of non-nominals discussed by Gopnik
and Choi, this means that 24.5% of vocabulary falls into the non-nominal
category for English children from 8 - 16 months, compared with 18% for
Italian. So, contrary to predictions based on verb salience, there is actually
a slight advantage for nominals in the Italian group, reflected in a significant
main effect of language (F(1,674) = 10.49, p .001). There was also a main
effect of vocabulary size (F(4,674) = 9.54, p .0001), with no significant
inter-action (F(4,674) = 1.97, p .10). The main effect of vocabulary size
actually reflects a slight drop in percent nominals, as other classes start
to grow (e.g., verbs, adjectives, and closed-class words-see below). The
source of the small Italian advantage on nominals is made clear by analyses
looking at subtypes within the Nominal category.
When analyses were carried out over the more restricted category of Common
Nouns, we found a large main effect of vocabulary level (F(5,674) = 54.20,
p .0001), but no main effect of language (F(1,674) = 0.02, n.s.), and no
interaction (F(5,674) = 1.80, n.s.). Note first that common nouns occupy
a much smaller pro-portion of total production vocabulary than nominals
more broadly defined: an average of 30.2% for English (compared with 75.5%
for all nominals), and 28.8% for Italian (compared with 82% for all nominals).
How-ever, in line with results described by Bates, Marchman et al. (1994),
common nouns also occupy an increasing proportion of total vocabulary as
the lexicon expands, from 17.4% in children with only 1 - 5 words, to 52.7%
in children with more than 50 words. In this respect, our results for Italian
exactly mirror previous reports for English on the disproportionate representation
and growth of object names in the first phases of lexical development. Figure
2 illustrates these developmental changes in the common-noun category, for
each lan-guage. The flat dotted line around 45% is included to illustrate
the proportion of the entire checklist composed of common nouns. As discussed
by Bates, Marchman et al.(1994), this line indicates the "checklist
baseline". If acquisition of nouns were a constant or random process
across this period of development, with no change in noun bias, then common-noun
proportion scores should hover around 45% at every level of development,
for both languages. Obviously this is not the case. In both groups, common
nouns are under-represented in the early stages but increase in importance
as vocabulary expands.
Figure 2
We also looked at individual variation around the mean for common-noun
scores, to see whether Italian children display the same "referential-to-expressive"
dimension of variation that has been described pre-viously for English (Bates
et al., 1988; Bates, March-man et al., 1994; Nelson, 1973). Figure 3a presents
the mean percent common-noun scores at each vocabulary level for English,
together with scores for children who are 1.28 standard deviations above
and below the mean (parametric values that approximate the top and bottom
10th percentiles-Bates, Marchman et al., 1994); Figure 3b presents the corresponding
values for Italian. It is clear from these figures that the same develop-mental
and stylistic factors are at work in each lan-guage. That is, nouns increase
in importance across this developmental range for both groups, but there
is also considerable variation in "nouniness" within each group.
Figure 3a
Figure 3b
So where does the slight Italian advantage on nominals come from? Table
2 includes all language by vocabulary level scores for two further subsets
from the global nominal category, proper nouns and sound effects. Analysis
of variance revealed a significant decrease in proper noun proportion scores
as a function of vocabulary level (F(4,674) = 26.49, p .0001), from an average
of 33% in children with 1 - 5 words, to 7% in children with more than 50
words. The interaction was not reliable (F(4,674) = 0.45, n.s.), but there
was a small main effect of language (F(1,674) = 5.31, p .022), reflecting
slightly more production of names for people in the Italian sample. As we
shall see later in the item analyses, this finding may reflect cultural
differ-ences in the number and proximity of grandparents and other family
members.
By contrast, there was no group difference in the analysis of sound effects
(F(1,674) = 0.50, n.s.), nor was there a significant interaction (F(4,674)
= 0.79, n.s.), but there was a significant main effect of voca-bulary level
(F(4,674) = 17.17). As we just saw for proper nouns, sound effects are another
category that decreases in relative size as vocabulary grows. For children
with 1 - 5 words, the sound effect category constitutes 32% of total vocabulary
on average, compared with 8.5% in children with more than 50 words. This
confirms the oft-reported finding that sound effects are (like proper names)
among the first nominals produced by one-year-old children. However, these
items decrease in importance as "true" common nouns take off-a
finding that holds for both language groups.
Routines like "bye" and "pattycake" form another category
that decreases in relative size as vocabulary expands. Indeed, just as proper
nouns and sound effects constitute the "starter set" within the
nominal category, these routines can be viewed as a "starter set"
in the category of non-nominals. Analysis of variance re-vealed a significant
main effect of vocabulary level (F(4,674) = 5.58, p .0001), no main effect
of language (F(1,674) = 1.28, n.s.), but a reliable albeit small interaction
between language and vocabulary size (F(4,674) = 3.13, p .02). In both language
groups, the development of routines is characterized by an inverted U-shaped
function, with an initial increase in relative size followed by a decline.
However, the shape of this function is slightly different for English and
Italian children. The English children reach a peak proportion score of
19.5% for children with 6 - 10 words, falling to 9.1% for children with
more than 50 words. The Italian children reach a peak of 20.7% for children
with 11 - 20 words, falling to 16.4% for children with more than 50 words.
We may speculate that the protracted emphasis on verbal routines in Italian
is one more reflection of cultural differences in adult-child interaction-a
tend-ency to "show off" to the relatively large number of aunts,
uncles and cousins that fill the lives of small Italian children. But it
may also reflect decisions about where to put early acquired items like
"allgone" and "uh-oh". Cross-linguistic differences
as small as the ones reported here can be caused by a very small number
of items, a point that will be clear in the qualitative analysis of first
words presented later on.
This brings us to the critical analysis for competing claims regarding the
emergence of nouns and verbs. As Table 2 shows, verbs occupy a very small
proportion of total expressive vocabulary in both the English and the Italian
sample. The overall mean is 2.06% for English and 2.27% for Italian. Analysis
of variance revealed no significant main effect of language (F(1,674) =
1.28, n.s.), and no significant interaction (F(4,674) = 1.91, n.s.). But
there was a large main effect of vocabulary size (F(4,674) = 24.08, p .0001),
reflecting a gradual increase in the contribution of verbs to total vocabulary
size, from less than 1% in children with 1 - 5 words to 6.4% in children
with more than 50 words. This developmental change is illustrated in Figure
2 for both languages, on the same figure with the scores described above
for common nouns (to facilitate comparison). Note also that all the verb
proportion scores fall well short of the percentage that verbs contribute
to the checklist as a whole. That is, because verbs represent approximately
14% of the whole checklist, a hypo-thetical child who develops by selecting
randomly across categories would average somewhere around 14% across the
first stages of lexical development. Instead, verbs constitute far less
than 14% of voca-bulary throughout this period.
Before we accept the conclusion that nouns de-velop before verbs in both
English and Italian, we need to consider a possible confound: The list as
a whole contains far fewer verbs than nouns. Perhaps children are developing
equally fast in both categories, but verb development is undersampled by
the checklist. As Bates, Marchman et al. (1994) point out, the dispro-portionate
numbers of nouns and verbs on this list is no accident. The final form of
the MacArthur Communicative Development Inventory was developed over a 20-year
period, with several revisions of the word checklist based on parental feedback
and laboratory observations. There are approximately three times as many
nouns as verbs on the final checklist because (1) this ratio is representative
of the adult language, and (2) this ratio is also reflected in diary studies,
laboratory observations, and in previous studies using versions of the word
checklist. Nevertheless, following Bates, Marchman et al. (1994), we also
carried out an analysis of the percentage of all noun and verb opportunities
that were checked by parents of children at each vocabulary level. That
is, we cal-culated a "noun opportunity score" (dividing the com-mon
nouns reported for each child by 182, the total number of common nouns provided
on the list) and a "verb opportunity score" (dividing the verbs
reported for each child by 55, the total number of verbs provided on the
list). These scores were analyzed in a multi-variate analysis of variance
comparing noun vs. verb opportunity scores (hereafter Word Type) as a function
of language and vocabulary level. Word type was treated as a within-subjects
variable; all other factors were between groups.
Results of this analysis are illustrated in Figure 4. All main effects and
interactions were statistically reliable: Language (F(4,665) = 21.80, p
.0001), Vocabulary Size (F(4,665) = 82.03, p .0001), Word Type (F(4,665)
= 381.95), Language by Vocabulary Size (F(1,665) = 13.65, p .0001), Word
Type by Vocabulary Size (F(4,665) = 89.54, p .0001), and a three-way interaction
(F(4,665) = 4.15, p .003). Several conclusions can be drawn from these complex
results. First, we should note that children produce a very small percentage
of the available items across this range of development, for both nouns
and verbs. Hence there is little danger that ceiling effects on the checklist
have skewed our results in some way. Second, some English children are filling
up the various noun and verb opportunities more quickly than their Italian
counterparts, although this is only true for children with more than 50
words. This finding reflects the upward skewing due to larger sample size
that we discussed in our analyses of raw scores for word production. That
is, the sample of 659 English-speaking children contains more children with
extremely high scores than the sample of 195 Italians. Despite these differences,
it is clear that the common-noun category is "filling up" faster
than verb category in both language groups, even though this analysis has
equated the two word types for the number of items available on the checklist.
Figure 4
We submit that the underrepresentation of verbs in very young children
reflects a fundamental develop-mental fact: Lexical verbs do not develop
until common nouns are a well-established component of the emerg-ing lexicon.
Our data suggest that this sequence is not affected by crucial cross-linguistic
differences in the salience and availability of verbs.
A similar conclusion holds for adjectives, although we do find some small
cross-linguistic differences here. As Table 2 shows, adjectives are very
rare in the expressive vocabulary of 8- to 16-month-old children, constituting
2.5% of total vocabulary for the whole English sample, and only 0.5% of
total vocabulary for the Italian group. Analysis of variance revealed a
main effect of language (F(1,674) = 11.82, p .001), a main effect of vocabulary
size (F(4,674) = 7.85, p .001), with no significant interaction (F(4,674)
= 0.92, n.s.). As we shall see later in our qualitative analysis of first
words, the American advantage on adjectives is due almost entirely to 2
- 3 specific words.
Following Bates, Marchman et al. (1994), we also combined verbs and adjectives
into a single predicate category. As Table 2 shows, this category constitutes
4.6% of total vocabulary for the full English sample, compared with 2.8%
for Italian. Analysis of variance revealed a large and reliable main effect
of vocabulary size (F(4,674) = 27.71, p .0001), a small but reliable main
effect of language (F(1,674) = 3.96, p .05), with no significant interaction.
It is already clear from the analyses above that this small cross-linguistic
difference comes from a few adjectives that are acquired early by English-speaking
children.
We end these analyses of word production with the heterogeneous category
of grammatical function words, known to be relatively rare in the first
stages of lan-guage learning. As Table 2 shows, grammatical function words
constitute 4.4% of total vocabulary for 8- to 16-month-old children acquiring
English, compared with 1.37% for children acquiring Italian. Since function
words constitute approximately 9% of the items on each checklist, this means
that function words are indeed underrepresented in the vocabularies of 8
- 16-month-old English and Italian children (i.e., this result cannot be
viewed as an artifact of the list itself). Analysis of variance revealed
a significant main effect of language (F(1,674) = 11.17, p .001), a significant
main effect of vocabulary size (F(4,674) = 2.47, p .044), with no significant
interaction (F(4,674) = 0.33, n.s.). In fact, there is very little growth
in the function word category across this period of develop-ment. On average,
scores increase slightly from 2.4% in children with only 1 - 5 words, up
to 5.2% in children with more than 50 words. As Bates, Marchman et al. (1994)
have pointed out for English, the early use of function words bears no predictive
relationship to the subsequent emergence of grammar. And as we shall point
out here for Italian and English children, the first function words are
restricted almost entirely to words that can be used alone in single-word
commands and routines (e.g., "Mine!"). The small cross-linguistic
difference favoring English comes from a subset of these items.
To summarize so far, there are far more similarities than differences in
the composition of early productive vocabulary in these English- and Italian-speaking
children. Above all, there is no evidence whatsoever for an Italian advantage
in verbs (narrowly defined) or non-nominals (broadly defined). Instead,
both groups provide evidence for a developmental sequence from reference
to predication, reflected in the early onset and growth of nouns and the
paucity of verbs and other predicates. Let us now turn to the composition
of receptive vocabulary, to see whether the Gopnik and Choi hypothesis of
cross-linguistic differences in the sequencing of nouns and verbs receives
more support from what these children know about language, as opposed to
what they actively do with the words they know.
Changes in the composition of receptive vocabulary
Previous studies comparing the early stages of comprehension and production
have concluded that verbs play a much larger role in early receptive language
(Bates et al., 1979; Benedict, 1979; Goldin-Meadow, Seligman, & Gelman,
1976). However, as Margaret Harris (1993) has also pointed out, this conclusion
is subject to some methodological criti-cisms. In particular, adults tend
to collect evidence for early word comprehension from the child's response
to commands (e.g., "Spit it out!" or "Get the book!'"),
a fact which may weight their estimates heavily toward the acquisition of
action verbs. It is also clear from Harris' review that the proportion of
verbs reported for early receptive vocabulary tends to be greater in studies
that rely on the adult's recall memory (i.e., informal diaries or open-ended
questions), and smaller in studies that rely on adult recognition memory
(e.g., checklists like the one employed here). In the present study, parents
used the same checklist to report word comprehension and production. This
means that they can rely heavily on recognition memory. The side-by-side
comparison of comprehension and production may bias them to use similar
criteria in the assessment of each-a fact which may be an advantage or a
dis-advantage, depending on one's view. In any case, if there is a verb
advantage in reports of word com-prehension, we may be in a position to
see cross-linguistic differences that did not emerge in our data on word
production.
Table 3 presents the nine proportion scores for word comprehension, for
each of the language and vocabulary-level groups in this 2 ¥ 6 design.
Starting with the class of Nominals (broadly defined), Table 3 reports an
overall mean of 61.8% for English and 63.8% for Italian. Although this is
smaller than the overall means of 75.5% and 82% described above for word
production, it is clear that nominals constitute the largest category of
word types in both modalities. However, the developmental patterns uncovered
for comprehension and production do differ in some respects. Analysis of
variance revealed a small but significant main effect of language (F(1,843)
= 4.08, p .05), but no main effect of vocabulary size (F(5,843) = 0.94,
n.s.), and no interaction (F(5,943) = 1.30, n.s.). In other words, we don't
find the same developmental decrease in the nominal category that we uncovered
in our analyses of word production.
TABLE 3: COMPOSITION OF RECEPTIVE VOCABULARY
AS A FUNCTION OF LANGUAGE AND VOCABULARY SIZE
Variable Language Number of Words in Receptive Vocabulary:
1-20 21-50 51-100 101-150 151-200 >200 Total
% Nominals English 60.4 62.6 61.2 63.2 61.4 61.8 61.8
Italian 66.8 67.0 63.7 64.2 60.4 60.7 63.8
% Common Nouns English 22.0 37.2 44.2 48.6 47.5 48.4 41.5
Italian 31.7 44.5 47.0 49.2 47.8 48.2 45.2
% People English 31.5 14.0 8.2 5.4 4.7 4.1 11.1
Italian 29.4 14.8 8.6 6.4 4.8 4.0 10.9
% Sound Effects English 6.2 8.8 5.2 4.8 4.3 3.7 5.6
Italian 5.1 7.0 5.4 4.5 3.7 3.5 4.9
% Routines English 27.3 18.6 13.1 9.8 8.6 6.6 14.1
Italian 25.2 18.7 12.5 9.9 8.4 6.4 13.3
% Verbs English 6.8 10.0 14.6 15.0 16.6 16.0 13.2
Italian 6.9 10.8 17.6 17.7 20.0 17.5 15.3
% Adjectives English 2.8 5.2 6.7 6.9 7.2 7.8 6.1
Italian 0.5 1.6 3.0 4.4 5.5 6.5 3.6
% Predicates English 9.7 15.2 21.3 21.9 23.9 23.8 19.3
(Verb + Adj) Italian 7.4 12.4 20.6 22.1 25.5 24.1 19.0
% Function Words English 2.5 3.6 4.1 4.9 5.8 7.1 4.5
Italian 0.6 1.9 3.2 3.5 5.4 8.0 3.7
Turning to the more restricted category of common nouns, scores were actually
larger for comprehension than they were for production, averaging 41.5%
for English (compared with 30.2% in production) and 45.2% for Italian (compared
with 28.2% in production). Analysis of variance on the common-noun scores
in comprehension revealed a significant main effect of language (F(1,843)
= 10.23, p .001), reflecting slightly larger common-noun proportion scores
for Italians (again, the opposite of predictions based on verb salience).
There was also a significant main effect of vocabulary size (F(5,843) =
70.86, p .0001), and a small but reliable interaction (F(5,843) = 2.43,
p .04). Figure 5 shows that common nouns make up an increas-ing proportion
of total receptive vocabulary in this period of development, from 24.2%
in children with fewer than 20 words to 48.3% in children with between 100
- 150 words (common-noun proportion scores hold steady after this point).
The initial rise in common nouns starts somewhat earlier for children acquiring
Italian, but the basic pattern is the same. Note also that the proportion
of common nouns in comprehension actually "overshoots" the checklist
ceiling (i.e., the dotted line indicated the 45% contribution of common
nouns to the checklist as a whole). In other words, common nouns are overrepresented
in children with receptive vocabularies between 50 and 200 words.
Figure 5
We also examined growth within two small subsets of the broad nominal
category, proper nouns and sound effects. In contrast with our analysis
of word pro-duction, we found no evidence for a cross-linguistic difference
in comprehension of proper nouns (F(1,843) = 0.04, n.s.), nor did language
interact with vocabulary size (F(5,843) = 0.39, n.s.). However, we did find
the expected developmental decrease in proper-noun proportion scores (F(5,843)
= 186.04, p .0001), with scores dropping from an average of 31% in children
with under 20 words to 4% in children with more than 200 words. In line
with our analyses of word produc-tion, we also found no cross-linguistic
difference in the proportional representation of sound effects in word comprehension
(F1,843) = 2.07, n.s.), and no significant interaction (F(5,843) = 0.56,
n.s.), but there was (as expected) a significant effect of vocabulary size
(F(5,843) = 15.99, p .0001), reflecting an overall drop in the contribution
of sound effects from around 26% in children with fewer than 20 words to
3.6% in children with more than 200 words.
In our earlier analyses of word production, routines constituted another
category that decreases sharply with development, leading us to propose
that routines are the "starter set" for non-nominals, while proper
names and sound effects serve as the "starter set" for nominals.
The same finding appeared in our analyses of word comprehension. There was
a significant main effect of vocabulary size (F(5,843) = 114.2, p .0001),
reflect-ing the fact that routines comprise around 27% of receptive vocabularies
in children with 20 words or less, compared with an average of 6.5% in children
with more than 200 words. In contrast with the small Italian advantage that
we found for word production, there was no main effect here for language
(F(1,843) = 0.45, n.s.), and no interaction (F(5,843) = 0.26, n.s.).
The main question here revolves around the pro-portional representation
of verbs in early word compre-hension. Table 3 shows that verbs are still
a relatively small category, compared with nominals (broadly defined) or
common nouns (narrowly defined). Over-all, verbs constitute 13.2% of the
items reported for English, compared with 15.3% for Italian. Analysis of
variance revealed a significant main effect of language (F(1,843) = 11.73,
p .001), and a significant main effect of vocabulary size (F(5,843) = 44.36,
p .0001), with no interaction (F(5,843) = 0.78, n.s.). The developmental
data for verbs in comprehension are illustrated in Figure 5 (plotted separately
for each language), where we also plotted the data for comprehension of
common nouns. It is clear that verbs do increase markedly across this range
of development, from approximately 6.8% of total receptive vocabulary in
children with 20 words or less, to 16.3% in children with more than 200
words. In fact (in contrast with our findings for production), verb proportion
scores rise above the checklist baseline for children with more than 50
words in their receptive vocabulary. It is also clear from Figure 5 that
Italians are ahead by 1 - 2 percentage points in verb comprehension across
this period of development. Hence we do find a small version of the predicted
Italian advantage on verbs in our data for comprehension, although a difference
this small could be explained by two or three salient words (see item analyses
below). It is also clear that the verb class is always much smaller than
the common-noun class, for both English and Italian children, even in comprehension.
Similar results are obtained for the adjective category, except that this
time the advantage goes to English-speaking children. Overall, adjectives
account for 6.1% of total vocabulary in English, vs. 3.6% in Italian. This
is higher than the averages reported for production, but it is still very
low. Analysis of variance revealed a significant main effect of language
favoring the Americans (F(1,843) = 73.05, p .0001), a signi-ficant main
effect of vocabulary size (F(5,843) = 33.07, p .0001), with no significant
interaction (F(5,843) = 1.64, n.s.). The vocabulary size result reflects
an increase from 2.8% in children with under 20 words to 7.8% in children
with more than 200 words. The American advantage of approximately 2 percentage
points at each development level is contributed (as we shall see below)
by a handful of specific words.
When the verb and adjective categories are added together, we find that
the combined predicate category accounts for 19.2% of comprehension vocabulary,
on average, compared with 43% for common nouns and 62% for all nominals.
Analysis of variance on predicate scores revealed no significant main effect
of language (F(1,843) = 0.84, n.s.), and no significant interaction (F(5,843)
= 0.96, n.s.). But there was a significant main effect of vocabulary size
(F(5,843) = 68.18), reflecting a clear increase in the relative size of
the predicate class as the receptive lexicon expands from under 20 words
(mean = 9.1%) to over 200 words (mean = 23.8%).
Finally, Table 3 summarizes results for the category of grammatical function
words in compre-hension. In line with our findings for production, these
words constitute only 4.3% of receptive vocabulary across this age range,
from a low of 2% in children with fewer than 20 words to a high of 7.3%
in children with more than 200 words. The overall average was 4.5% for English,
3.7% for Italian. Analysis of variance revealed a small but reliable effect
of language (F(1,843) = 5.55, p .02), a significant main effect of vocabulary
size (F(5,843) = 17.53, p .0001), with no significant interaction. Hence
there is a small advan-tage in this category for children acquiring English,
and a small but still reliable increase in the proportional representation
of function words as a function of receptive vocabulary size.
We conclude that the data for comprehension and production both provide
evidence for a universal se-quence from nouns to verbs and adjectives. This
pattern holds up in both languages, and it holds up across modalities (even
though verbs and adjectives do emerge earlier and grow faster in comprehension
than production). The few cross-linguistic differences that we do find in
vocabulary composition always involve differences of 1 - 3 percentage points-a
difference that could be contributed by a few salient items. This brings
us to a qualitative analysis of the first 100 words acquired within each
language, in production and in comprehension.
Item analyses
Table 4 presents the first 50 words in the expressive vocabularies of American
and Italian children, where "first acquired" is defined in terms
of the percentage of children in the sample who were reported to produce
that word.8 These percentages are collapsed over age levels, but it is worth
pointing out that Fenson et al. (1994) report extremely high correlations
between this criterion for the assessment of age of acquisition, and a different
criterion based on the age at which 50% of the sample is reported to use
the word. In Table 4, the words are listed in order of acquisition within
each language group. Within each list, words that would qualify as non-nominatives
according to the broad criteria used by Gopnik and Choi (in press) are marked
with an asterisk (regardless of how those words were classified on the MacArthur
list). Those non-nominatives that qualify as lexical verbs in the adult
language are also marked with an asterisk, and presented in capital letters.
TABLE 4: FIRST 50 WORDS IN PRODUCTION
FOR ENGLISH VS. ITALIAN INFANTS
ENGLISH ITALIAN
Rank Word % sample Word Translation % sample
1. Daddy 54.9 mamma mommy 49.7
2. mommy 52.9 papa daddy 46.7
3. * bye 43.1 bau-bau (dog sound) 41.5
4. * hi 39.3 * pappa (food/mealtime) 36.9
5. * uh-oh 35.5 nonna grandma 32.8
_____________________________________________________________________________
6. Baa-baa 31.9 brum-brum (vehicle sound) 28.7
7. Ball 30.9 acqua water 27.2
8. dog 30.6 nonno grandpa 23.1
9. * no 28.5 * nanna (sleep/bedtime) 21.5
10. bottle 25.2 * no no 21.5
_____________________________________________________________________________
11. woof 24.9 miao (cat sound) 21.0
12. baby 24.6 * grazie thanks 20.5
13. * yum-yum 24.1 * ciao hi/bye 17.9
14. grr 23.5 * cuccu-settete (hiding game) 16.9
15. kitty 21.8 palla ball 16.4
16. vroom 20.2 muuh (cow sound) 15.9
17. book 19.9 * non c'è più (is no more) 14.9
18. bird 19.6 scarpe shoes 14.4
19. duck 18.8 coccode (rooster sound) 13.8
20. balloon 18.4 beh-beh (sheep sound) 12.8
21. cat 18.2 (child's own name) -- 12.8
22. * night-night 17.1 ih-oh (donkey sound) 12.8
23. quack 17.0 bimbo child 12.3
24. shoe 17.0 * pronto (hello on phone) 11.8
25. meow 16.6 * bum boom 11.3
26. banana 16.3 grr (lion sound) 10.8
27. * hot 16.0 qua-qua (duck sound) 10.8
28. Juice 15.4 (babysitter's name) -- 10.3
29. Eye 14.8 cip-cip (bird sound) 10.3
30. Grandma 14.3 * si yes 10.3
31. moo 14.2 tuttu (train sound) 10.3
32. * thank-you 14.0 * zitto hush/quiet 10.3
33. * up 14.0 * (fare) popo/pipi (make) pee/poo 9.7
34. cookie 13.5 clop-clop (horse sound) 9.2
35. Nose 13.5 * bua hurt/owie 8.7
36. * ouch 13.4 * (dare) toto (give) spanking 8.2
37. Cracker 12.3 * mio my/mine 8.2
38. grandpa 12.3 pane bread 8.2
39. * shh 12.0 biscotto cookie 7.7
40. bath 11.8 cane dog 7.2
41. keys 11.8 ciuccio pacifier 7.2
42. Bubbles 11.4 zio uncle 7.2
43. * down 11.4 latte milk 6.7
44. car 11.2 orologio watch/clock 6.7
45. * yes 11.0 zia aunt 6.2
46. cheese 10.9 banana banana 5.6
47. bear 10.7 * basta enough/stop 5.6
48. * hello 10.6 bambola doll 5.1
49. Fish 10.4 * DARE to give 5.1
50. * allgone 10.3 gatto cat 5.1
51. Hat 10.3 mela appole 5.1
For English, the first five words reported in produc-tion are "daddy"
(54.9%), "mommy" (52.9%), "*bye" (43.1%), "*hi"
(39.3%) and "*uh-oh" (a sound effect that is difficult to classify
but typically occurs when things fall down or break - 35.5%). None of these
are common nouns by our criteria, and none of them are verbs, although three
(especially "*uh-oh") would qualify as non-nominals with a "verb-like"
status using the criteria described by Gopnik and Choi (in press). For Italian,
the first five words are "mamma" (49.7%), "papa" (46.7%),
"bau-bau" (a sound effect for dogs - 41.5%), "*pappa"
(an all-purpose word for food or mealtime - 36.9%) and "nonna"
(for grandmother - 32.8%). None of these are common nouns by our criteria,
and none are verbs. All five of these words were classified as nominals
on the MacArthur list (and in the analyses over children), because "*pappa"
was listed with food words. If we use the Gopnik and Choi criteria and reclassify
"*pappa" as a non-nominal, it is still the case that nominals
dominate among the first five words in Italian. Note also that the first
five words already indicate a greater use of words for people in the Italian
sample, since "grandma" ranks up at the top with "mommy"
and "daddy" on the Italian list.
Moving down to the next five words, the English favorites are "baa-baa"
(a sound effect for sheep, although some families reportedly use this sound
for dogs - 31.9%), "ball" (30.9%), "dog" (30.6%), "*no"
(28.5%) and "bottle" (25.2%). If we follow Gopnik and Choi and
take "*uh-oh" out of the sound effect category (where it was counted
broadly as a nominal) and put it into the category of non-nominal routines,
then we have an expressive lexicon that breaks down into 50% nominals (20%
names for people, 10% sound effects and 30% common nouns), and 50% non-nominals
(with no lexical verbs). Moving down the corresponding list for Italian,
the second set of five words includes "brum-brum" (a sound effect
for vehicles - 28.7%), "acqua" (the word for water - 27.2%), "nonno"
(for grandfather - 23.1%), "*nanna" (a general term for sleeping
and for naps - 21.5%) and "*no" (which means exactly what it means
in English - 21.5%). Hence the first ten words in Italian break down into
70% nominals (broadly defined using the Gopnik and Choi criteria), including
40% words for people, 20% sound effects and 10% common nouns. There are
only three non-nominals ("*no", "*pappa" and "*nanna"),
constituting only 30% of the list, and no lexical verbs.
To summarize so far, it is clear lexical verbs do not appear among the first
ten words produced by English or Italian children. And if we adopt the very
broad criteria proposed by Gopnik and Choi, treating all non-nominals as
verbs, then we actually find a small disadvantage in the non-nominal category
for Italians-although this difference probably has more to do with the salience
of words for people in the Italian culture (i.e., the first 10 words for
Italian include "mommy", "daddy", "grandma"
and "grandpa", while the first 10 words for English are restricted
to "mommy" and "daddy"). However, this does not mean
that children begin their lexical careers by naming objects. Indeed, most
of the items in the Top Ten for English and/or Italian are not common nouns.
We will return to this point in the final discussion.
We can obtain a more stable picture of vocabulary composition by considering
the first 50 words in each language (Table 4).8 In English, this list includes
no lexical verbs whatsoever. In Italian, the corresponding list contains
exactly one lexical verb "*DARE" ("to give"). This word
was used by only 5.1% of the sample, which means a total of 10 children
out of 195. Hence we may conclude that lexical verbs are sorely lacking
in the first stages of language development, in both English and Italian.
If we follow Gopnik and Choi (1990) and enlarge our definition to include
all non-nominals, then we find a total of 15 non-nominal words for the English
sample (29.4% of the list). This includes two adjectives ("*hot"
- 16%; "*allgone" - 10.3%), and two grammatical function words
("*up" - 14%; "*down" - 11.4%). On the Italian side,
we find a total of 17 non-nominals among the first 50 words (33% of the
list). This includes the one verb mentioned above, no adjectives, 13 items
that fall under the heterogeneous category of routines, and one gram-matical
function word ("*mio", or "mine", used by 8.2% of the
sample).
We went on to examine the next 50 words on the list for English and Italian,
to see whether conclusions based on the first 50 words would change markedly
in any direction. In English, the first lexical verbs finally do appear
when the list is expanded to 100 items, including "*GO" (7.5%
of the sample), "*SEE" (5.8%), "*EAT" (5.6%) and "*BITE"
(5.3%). Still, this means that verbs constitute only 4% of the first 100
words acquired by English-speaking children. In Italian, four additional
verbs appear when the list is expanded to 100, including "*PIANGERE"
("to cry" - 2.6%), "*DONDOLARE" ("to rock"
- 2.1%), "*MANGIARE" ("to eat" - 2.1%) and "*APRIRE"
("to open" - 2.1%). This brings the total number of verbs in Italian
to five, constituted 5% of the first 100 words. In other words, the emergence
of lexical verbs continues to follow a parallel course in English and Italian.
It should be clear from these qualitative analyses that the very small cross-linguistic
differences reported earlier for adjectives and function words must be treated
with caution. In every case, they are due to fewer than three items, a result
that could shift in either direction if items were moved from one category
to another. The main results to emerge from these item analyses come from
the larger and more robust categories (i.e., common nouns and lexical verbs).
In this regard, the item analyses support our conclusions in the earlier
analyses over children: Verbs are relatively late in both English and Italian,
common nouns outnumber verbs and other predicates at every point in this
age range, and common nouns increase in importance at the expense of those
hard-to-classify items that dominate in the very first stages of lexical
development (e.g., routines, sound effects, names for people).
In the same spirit, Table 5 presents the first
50 words reported for comprehension, in English and Italian.8
The numbers are considerably larger and more stable in this case, since
comprehension gets off the ground much earlier than production (another
universal that withstands a cross-linguistic test.....). But the item analysis
complements and confirms what we have already learned from our analyses
over subjects, for both comprehension and production.
TABLE 5: FIRST 50 WORDS IN COMPREHENSION
FOR ENGLISH VS. ITALIAN INFANTS
ENGLISH ITALIAN
Rank Word % sample Word Translation % sample
1. mommy 95.0 mamma mommy 91.3
2. daddy 93.5 papa daddy 88.2
3. * bye 88.6 (child's own name) --- 82.6
4. * no 86.3 * ciao hi/bye 82.6
5. * peekaboo 84.3 * pappa (food/mealtime) 81.1
_____________________________________________________________________________
6. bath 76.2 * cuccu-settete (hiding game) 81.0
7. ball 75.0 acqua water 79.0
8. bottle 75.0 * no no 77.9
9. * hi 74.0 palla ball 75.9
10. * allgone 71.9 bau-bau (dog sound) 75.4
_____________________________________________________________________________
11. dog 70.8 nonna grandma 75.4
12. book 68.7 cane dog 74.9
13. * night-night 68.5 biberon bottle 71.8
14. diaper 67.4 telefono telephone 70.3
15. * KISS 66.2 * bravo good 67.7
16. * uh-oh 65.1 nonno grandpa 66.7
17. * pattycake 62.6 scarpe shoes 66.2
18. juice 61.9 biscotto cookie 65.6
19. shoe 61.9 * BERE to drink 65.1
20. baby 61.6 miao (cat sound) 64.6
21. grandma 61.3 latte milk 64.1
22. outside 61.0 * nanna (sleep/bedtime) 63.6
23. car 60.1 mano hand 63.1
24. * EAT 59.7 * basta (enough/stop) 62.6
25. kitty 58.8 * pane bread 62.6
26. * DRINK 58.1 * (fare) bagno (have/do) bath 62.1
27. keys 56.3 * gatto cat 62.1
28. * DON'T 55.8 bimbo child 60.5
29. comb 55.4 * DARE to give 59.5
30. nose 55.4 * MANGIARE to eat 59.5
31. * HUG 54.9 piede foot 59.0
32. banana 54.4 * BACIARE to kiss 59.0
33. cookie 54.2 * BALLARE to dance 59.0
34. bathtub 53.2 automobile car 57.9
35. balloon 52.9 * non c'è più (is no more) 56.9
36. milk 52.9 panolino diaper 56.9
37. cat 52.7 * si yes 56.9
38. Cracker 52.7 bavaglino bib 56.4
39. telephone 52.6 capelli hair 56.4
40. * yes 52.6 bocca mouth 55.9
41. cheerios 51.4 bicchiere glass 54.9
42. bird 50.4 uccellino bird 54.4
43. * yum-yum 50.4 passegino stroller 53.8
44. grandpa 50.1 * pronto (hello on phone) 53.3
45. woof 49.5 ciuccio pacifier 51.3
46. * DANCE 49.3 letto bed 51.3
47. baa-baa 49.0 naso nose 50.3
48. meow 48.3 televisione television 49.7
49. * LOOK 48.2 * ANDARE to go 49.2
50. mouth 48.2 cucchiaio spoon 49.2
51. * PETTINARE to comb 49.2
52. * SALUTARE to greet 49.2
Starting with the Top Ten Hits in comprehension for each language group,
the breakdown for English is as follows: "mommy" (95%), "daddy"
(93.5%), "*bye" (88.6%), "*peekaboo" (84.3%), "*no"
(86.3%), "bath" (76.2%), "bottle" (75%), "ball"
(75%), "*hi" (74%) and "*allgone" (71.9%). Following
the criteria proposed by Gopnik and Choi, this corresponds to 60% non-nominals,
although none of these items are lexical verbs. The corresponding list for
Italian is "mamma" (91.3%), "papa" (88.2%), the child's
own name (82.6%), "*ciao" (which means both "hi" and
"goodbye" - 82.6%), "*pappa" (a general word for food
or mealtime - 81.1%), "*cuccu-settete" (a word used in hide-and-seek
- 81%), "acqua" ("water" - 79%), "*no" (77.9%),
"palla" ("ball" - 75.9%) and "bau-bau" (a
sound effect for dogs - 75.4%). According to the Gopnik and Choi criteria,
this corresponds to 40% non-nominals. Again, none of these items are lexical
verbs. Hence the data for comprehension in English and Italian are consonant
with our findings for production: no lexical verbs, very few common nouns,
with a preponderance of routines, sound effects and names for people.
Surveying the top 50 words in comprehension for each language, we find a
total of six lexical verbs for English ("*KISS" - 66.2%; "*EAT"
- 59.7%; "*DRINK" - 58.1%; "*HUG" - 54.9%; "*DANCE"
- 49.3%; "*LOOK" - 48.2%). If we also include the word "*DON'T"
(which was classified among routine expressions on the MacArthur list),
this brings the total to seven verbs, or 14% of the list. Hence lexical
verbs are still a minority in compre-hension, but (in line with results
by other investigators) there are more verbs in comprehension than we found
in production. Applying the criteria proposed by Gopnik and Choi, we can
classify 17 of the first 50 words in comprehension as non-nominals (i.e.,
34%, compared with 29.4% in production).
Looking at the corresponding list for Italian, we find a total of eight
lexical verbs in Italian (15.4% of the list). These include "*BERE"
("to drink" - 65.1%), "*DARE" ("to give" -
59.5%), "*MANGIARE" ("to eat" - 59.5%), "*BACIARE"
("to kiss" - 59%), "*BALLARE" ("to dance"
- 59%), "*ANDARE" ("to go" - 49.2%), "*PETTINARE"
("to comb" - 49.2%) and "*SALUTARE" ("to greet"
- 49.2%). Using broad criteria to define the class of non-nominals, the
total for Italian is 20 items, or 38.5% of the list (compared with 33% for
production).
If we move on to examine the next 50 words in comprehension, the verb category
does show some proportional increase. Briefly, we find a total of 18 verbs
among the first 100 items for English (18% of the list), compared with 24
verbs among the first 100 items for Italian (24% of the list). Hence, as
we saw in our analyses over children, there are indeed more verbs in early
comprehension than we find in early production (cf. Benedict, 1979; Harris,
1993). But nominals are still the dominant class for both comprehension
and production as soon as children move beyond their first 5 - 10 words,
and common nouns expand more rapidly than verbs as the list grows. There
does appear to be an Italian advantage for verbs in comprehension when we
consider the first 100 words, but the difference is still very small.
SUMMARY AND CONCLUSION
In this cross-linguistic study of early lexical devel-opment, the following
conclusions emerge.
(1) If we restrict our attention entirely to the first 5 - 10 words in each
language, we find a preponderance of sound effects, routines, and names
for people, with a few common nouns scattered in. These are the very items
that are hardest to classify in adult part-of-speech categories. In other
words, the very first words produced by English and Italian children are
neither nouns nor verbs-a point to which we shall return below.
(2) When we expand our developmental window to include the first 50 - 100
words, the overwhelming majority are nominals (broadly defined); even when
we restrict the definition of nominals to common nouns (i.e., names for
classes of concrete objects), we cannot escape the conclusion that nouns
predominate and grow sharply (in proportion to other items) across the first
stages of lexical development, for both English and Italian.
(3) Verbs are very rare among the first 50 - 100 words produced by young
children, even in a language like Italian in which verbs are particularly
salient and informative.
(4) Verbs emerge earlier and grow faster in comprehension than they do in
production, but they are still outnumbered by common nouns at every point
across this period, in both English and Italian.
(5) Adjectives and grammatical function words are also very rare among the
first 50 - 100 words acquired by English and Italian children, in both comprehension
and production.
As we pointed out in the introduction, all of these analyses have been carried
out from the point of view of adult part-of-speech categories. We do not
assume (and should not assume) that these categories are somewhere in the
mind of the child. Rather, we treat adult cate-gories like noun and verb
as independent variables, a short-hand way of referring to semantic, grammatical
and/or phonological differences in the input to which small children are
exposed. Verbs differ from nouns in all languages, along many different
semantic, gramma-tical and phonological dimensions (including prosody).
To the extent that children treat nouns differently from verbs in the first
stages of lexical development, we may infer that they have been influenced
by these differ-ences in their linguistic input. That is, there is some-thing
"in the child" that responds differently to nouns and verbs, something
which must be due to the characteristics that distinguish nouns from verbs
(and from adjectives and function words) in the adult language.
At the same time, verbs ought to be far more salient for Italian children.
Had we seen any differences between English and Italian children in the
emergence and growth of major lexical categories, we would be justified
in concluding that children are influenced by differences in the relative
salience of these categories in their respective input languages. We did
find differ-ences in order of emergence (suggesting that verbs are somehow
more difficult than nouns), but we did not find any differences between
the two languages in the onset and growth of major lexical categories. The
robustness of this noun-verb sequence despite striking cross-linguistic
differences in the input lead us to propose that this sequence may be a
developmental universal.
However, we should also stress that common nouns (i.e., object names) are
relatively rare in the first 5 - 10 words produced by English and Italian
children. It would, therefore, be incorrect to argue that all children begin
their lexical careers with object naming. We agree with Gopnik and Choi
(and many other investigators in the field-e.g., Bloom, Hafitz, & Lifter,
1980; Nelson, 1981; Tomasello, 1992) that non-nominals play an important
role in early lexical development. But how should that role be defined?
We think it is a mistake to confuse the broad category of non-nominals used
by 8 - 16-month-old children with true predication, i.e., with the explicit
postulation of a state that exists in one entity (i.e., single-argument
predicates) and/or a relation that exists between two or more entities (i.e.,
n-place predicates). The hetero-geneous set of non-nominals that appear
in Tables 4 and 5 are hard to place in any clear-cut part-of-speech category-which
is why we have been forced to use non-technical categories like "sound
effects" and "routines" that do not appear in any grammar
or dictionary of the adult language. In earlier work by members of our research
team (e.g., Bates, Camaioni, & Volterra, 1975; Bates, Benigni, Bretherton,
Camaioni, & Volterra, 1979), we proposed that the referential function
of words may emerge gradually from verbal routines that are best viewed
as speech acts or performatives, vocal conventions that children use in
particular situations to achieve some social function. Some of these vocal
routines happen to be nouns in the adult language, others (rarely) might
be verbs or adjectives, while others (like "hi" and "uh-oh")
defy classification. The important point is that these words are best construed
as actions, language games, things that children do with speech in a familiar
and well-structured context (see also Barrett, 1982; Dromi, 1987; Lieven
& Pine, 1990; Nelson & Lucariello, 1985; Ninio, 1993 & 1994;
Tomasello, 1992).
From this point of view, we should not be at all surprised to find that
children pick up the highest-frequency games and routines that are proposed
to them by adults in this period of development. Indeed, some of the item-specific
differences between English and Italian that we have uncovered here seem
to fall into this category. For example, the first 50 words produced by
English-speaking children include an adjective "hot", reported
for 16% of the sample. The adjective "caldo" (which means "hot")
does not appear among the first 100 words for Italian. American mothers
often use the single word utterance "HOT!" to warn infant away
from stoves, lightbulbs and other dangerous items (and, as we have already
noted, the word "hot" is often used by English-speaking children
as a name for bright and/or hot objects). In the same context, Italian mothers
are more likely to respond with "SCOTTA!" ("Burns!"),
or simply "NO NO!". Hence these two "equivalent" items
are differentially available to children learning English vs. Italian. In
the same vein, it is interesting that "nonna" ("grandmother")
and "nonno" ("grandfather") are among the first 10 words
for Italian children, while the corresponding items rank 30th and 38th,
respectively, for children acquiring English. This "linguistic"
phenomenon finds a ready explanation in the fact that Italian families are
more likely to live in the same city with grandparents and other relatives,
resulting in regular visits that foster the early acquisition of names for
these family members. Without dwelling further on details of this sort,
we will simply conclude that cultural differences do influence the content
of infant vocabulary. In this regard, we agree with Gopnik and Choi. Nevertheless,
these cultural differences are apparently not sufficient to overcome the
developmental progression from nouns to verbs, a shift which is (we assume)
motivated by critical differences in the input characteristics of nouns
and verbs, and developmental differences in children's sensitivity to these
characteristics.
The proposed noun-verb shift must be amended in two respects. First, as
we have just noted, children do not necessarily begin with common nouns,
but with a heterogeneous class of items that are perhaps best viewed as
routines or language games. Second, verbs are relatively late, but so are
adjectives and function words. Following O'Grady (1987) and Bates, Marchman
et al. (1994), we would like to suggest that verbs and other relational
terms are rare within the first 100 words because it is difficult for children
to understand the meaning and purpose of "secondaries" (i.e.,
terms that express predication) before they have a firm grasp of "primaries".
(i.e., terms that set up the arguments of a predicate). Hence the proposed
universal sequence in the acquisition of lexical cate-gories may reflect
changes at the level of underlying meanings, a developmental sequence that
may be summarized as Routines -> Reference -> Predication.
This proposed universal is a hypothesis, not a con-clusion, for two reasons:
(1) the limited utility of parental report data for the study of semantic
change, and (2) other cross-linguistic findings that lead to a different
conclusion.
With regard to the first point, a test of the proposed referent-to-predication
shift will require detailed infor-mation on the way that children use their
words in context. Previous studies of this sort have yielded mixed conclusions
regarding semantic change in the second year of life. Some investigators
have concluded that there is an important shift late in the one-word stage,
from naming and language games to the use of single words to express relational
meanings (e.g., DeLaguna, 1927; Howe, 1976). Others have argued that predication
is present early in one-word speech, in the way non-nominals and some nominal
expressions are used (e.g., Gopnik, 1981; Rescorla, 1981). Based on our
findings here, we would predict that further observational evidence will
come down on the side of a referent-to-predication shift during the second
year. But our own data are silent on this point.
With regard to the second problem, our findings for English and Italian
differ from those of Gopnik and Choi for English and Korean (see also Cheng,
1994, whose findings for Chinese parallel those of Gopnik and Choi for Korean).
However, these authors also used a different methodology (i.e., free-speech
samples complemented by a parental interview). What would happen if our
checklist methodology were applied to large samples of Korean children in
the same age range? Relevant findings come from an unpublished dissertation
by Pae (1993), who adapted the MacArthur Infant Scale for Korean, and administered
it to the mothers of 90 Korean infants between 12 and 23 years of age, all
residents in middle-class apartment com-plexes in Seoul, Korea. Pae's results
for both com-prehension and production were remarkably similar to ours,
suggesting that there are few differences among Korean, Italian and English
children in the onset and growth of nouns vs. verbs when comparable parental
report methods are employed. She concludes that "Contrary to the verb
bias argument, the Korean-speaking children in this sample did not show
any sign of relative ease in learning verbs" (Pae, 1993, p. ii).
The contrast between Gopnik and Choi's findings based on free speech and
those of Pae using parental report lead to the suggestion that cross-linguistic
differ-ences in the emergence of non-nominal expressions (especially verbs)
may reflect differences in the forms that children prefer to use, and not
to differences in the words they are able to use (see also Bates et al.,
1988). If this is the case, then we might expect to find differences between
English and Italian children in early verb use in a free play situation-a
possibility that we are currently pursuing.
Another possible explanation for these conflicting findings revolves around
the issue of homophony, i.e., the use of a single sound for two different
meanings or functions. The English language is notorious for the flexibility
with which words can move back and forth across the noun-verb boundary (Clark
& Clark, 1979). In particular, virtually any noun can be used as a verb
if the context is set up just right (e.g., "He Richard-Nixoned the
tapes before appearing at the Senate hearing...."). Because so many
English nouns have corresponding verb homophones (and vice-versa), the form
class assignments that we adopted in the MacArthur CDI are perhaps best
viewed as a classification into "most typical or frequent word class",
as opposed to an absolute classification based on inherent properties of
each word. Such denominalized verbs are far less common in Italian, and
when they do occur, they involve so much morphological alteration in the
shape of the word that the respective noun and verb forms are not homophones
at all. Indeed, the line between nouns and verbs on the Italian MacArthur
CDI is clearly indicated not only by class membership (i.e., verbs are listed
under "action words"), but also by the inflectional form in which
the verbs are listed (i.e., in the infinitive, the citation form for Italian
verbs). Despite these differences between English and Italian, we have observed
the same sequence of acquisition for nouns and verbs. However, it is possible
that the results observed to date for Korean and Chinese may revolve (at
least in part) around the homophony issue. Like Italian, Korean and Chinese
are languages that permit extensive noun omissions and word order variation.
However, like English, Korean and Chinese are lan-guages with a high degree
of noun-verb homophony. This leads to some interesting possibilities, including
the following:
(1) Korean and Chinese children do not acquire
verbs earlier than English or Italian children. Rather, they appear to have
acquired verbs earlier because they have misclassified adult verbs as nouns.
Hence they are using more verbs as nouns (to name objects, or perhaps to
name events), and/or they are using more verbs in rote routines.9
(2) Korean and Chinese children really do acquire verbs earlier than English
children do. Italian children would do the same if they could, but they
are "blocked" from picking up those enticing and salient verbs
due to the complex morphology that distinguishes verbs from nouns in their
language.
These issues can only be settled by converging methodologies, complementing
maternal diary and parental report studies with detailed analyses of the
way that infants use their early words in context.
FOOTNOTES
1 Of course the pro-drop
parameter is an idealization. Subject omission is obligatory in English
imperatives, and subjects are frequently omitted in response to questions
(e.g., "What did you do?" -- "Went to the movies"),
in some informal conversations (e.g., "Really blew it this time!),
and in most of the interviews granted by former President Bush. Never-theless,
subject omission in English never comes close to the 70% ratio observed
in informal Italian, in either adult- or child-directed speech.
2 The formula [1
- percent nominals] yields a score corre-sponding approximately to the proportion
of total vocabulary composed of non-nominals as defined by Gopnik and Choi
(in press).
3 The category "places
to go" was excluded from the count of common nouns because many of
its items function more like adverbials in the adult language.
4 Words like "aunt"
that could describe more than one individual were included in "nominals"
but excluded from "common nouns"
5 We are using the
term "predicate" in as a theory-neutral descriptor at this point,
and will return later to the more controversial issue of the meanings that
children intend with such words
6 The ambiguous class
of "words about time" was excluded from the function word count,
although it includes some highly frequent terms like "now".
7 Another cultural
difference is reflected in the composition of the two word lists. Words
for eliminatory functions (i.e., "pee" and "poo") and
words for genitalia were excluded from the English list even though these
words are reported for some children in the 8 - 16-month age range. The
decision was made because so many parents (and some investigators) expressed
embarrassment or discomfort in reporting these items. The Italian team reached
a different decision for these particular items, a fact which may be related
to cultural differences in comfort level regarding sexuality and bodily
functions.
8 The list of words
in Table 4 actually includes 51 items for each language, instead of 50,
due to the fact that the last two items on each list were tied in rank ("allgone"
and "hat" were both produced by 10.3% of the Americans; "dare",
"gatto" and "mela" were all produced by 5.1% of the
Italians). In the same vein, Table 5 contains 50 words for English but 52
words for Italian, due to a tie in rank for the last four words ("andare",
"cucchiaio", "pettinare" and "salutare", all
compre-hended by 49.2% of the Italian sample).
9 Some investigators
have argued that lexical verbs do predominate in the first signs produced
by children exposed to American Sign Language (Launer, 1982), and in the
signing systems that emerge among deaf children of hearing parents (Goldin-Meadow
& Mylander, 1985). However, this apparent difference between signed
and spoken languages may have more to do with the ambiguity of nouns and
verbs within the adult language, where names for common objects are some-times
derived from the actions that are carried out on those objects. From a slightly
different point of view, Volterra (personal communication) has proposed
that verbs may be rare among the first words produced by hearing children
because they are able to express verb-like meanings in the gestural modality
(e.g., drinking from cups, putting a tele-phone receiver to the ear). However,
this proposal is also controversial, since many investigators view such
gestures as a form of gestural recognition or naming (e.g., Acredolo &
Goodwyn, 1988). This is a matter that merits further investigation.
REFERENCES
Acredolo, L.P., & Goodwyn, S.W. (1988). Symbolic gesturing in normal
infants. Child Development, 59, 450-456.
Barrett, M. (1982). The holophrastic hypothesis: conceptual and empirical
issues. Cognition, 11, 47 - 76.
Bates, E. (1976). Language and context: Studies in the acquisition of pragmatics.
New York: Academic Press.
Bates, E., Benigni, L., Bretherton, I., Camaioni, L., & Volterra, V.
(1979). The emergence of symbols: Cognition and communication in infancy.
New York: Academic Press.
Bates, E., Bretherton, I., & Snyder, L. (1988). From first words to
grammar: Individual differences and dissociable mechanisms. New York: Cambridge
University Press.
Bates, E., Camaioni, L., & Volterra, V. (1975). The acquisition of performatives
prior to speech. Merrill-Palmer Quarterly, 21:3, 205-226. (Reprinted in
E. Ochs & B. Schieffelin, Eds., Developmental pragmatics (pp. 111-128).
New York: Academic Press, 1979.
Bates, E., Caselli, M.C., & Casadio, P. (1990). A cross-linguistic study
of early lexical development using parental report. In C. Rovee-Collier
(Ed.), Abstracts of the Seventh International Conference on Infant Studies.
Special issue of Infant Behavior and Development , 13, 258. Norwood, NJ:
Ablex.
Bates, E., Dale, P.S., & Thal, D. (1994). Individual differences and
their implications for theories of language development. In Paul Fletcher
& Brian MacWhinney (Eds.), Handbook of child language (pp. 96-151).
Oxford: Basil Blackwell.
Bates, E., & Marchman, V. (1988). What is and is not universal in language
acquisition. In F. Plum (Ed.), Language, communication and the brain. New
York: Raven Press, 19-38.
Bates, E., Marchman, V., Thal, D., Fenson, L., Dale, P., Reznick, S., Reilly,
J., & Hartung, J. (1994). Developmental and stylistic variation in the
composition of early vocabulary. Journal of Child Language, 21:1, 85-124.
(Reprinted in K. Perera, Ed., Growing points in child language. Cambridge:
Cambridge University Press, in press).
Benedict, H. (1979). Early lexical development: Comprehension and production.
Journal of Child Language, 6, 183-200.
Bloom, L., Hafitz, E., & Lifter, K. (1980). Schematic organization of
verbs in child language and the acquisition of grammatical morphemes. Language,
6, 386-412.
Bloom, L., Lightbown, P., & Hood, L. (1975). Structure and variation
in child language. Mono-graphs of the Society for Research in Child Development.
Braine, M.D.S. (1976). Children's first word combinations. With commentary
by Melissa Bowerman. Monographs of the Society for Research in Child Development,
41, Serial # 164.
Cheng, S.-W. (1994). Beginning words of three children acquiring Mandarin-Chinese.
Manuscript in preparation, National Chung Cheng University, Taiwan.
Clark, E. (1983). Meanings and concepts. In J. Flavell and E. Markman (Eds.),
Cognitive Development, Volume III of P. Mussen (Ed.), Handbook of Child
Psychology. New York: John Wiley, 787 - 840.
Clark, E., & Clark, H. (1979). When nouns surface as verbs. Language,
55, 767 - 811.
DeLaguna, G. (1927). Speech: Its function and development. Bloomington:
Indiana University Press.
Devescovi, A., D'Amico, S., Smith, S., Mimica, I., & Bates, E. (1993).
The development of sentence comprehension in Italian and Serbo-Croatian:
Local versus distributed cues. Manuscript, University of Rome "La Sapienza".
(submitted for review).
Dromi, E. (1987). Early lexical development. New York: Cambridge University
Press.
Fenson, L., Dale, P.A., Bates, E., Reznick, J.S., & Thal, D. (1994).
Variability in early communicative development. Monographs of the Society
for Research in Child Development, 59, Serial No. 242, No. 5.
Fenson, L., Dale, P., Reznick, J.S., Thal, D., Bates, E., Hartung, J., Pethick,
S., & Reilly, J. (1993). The MacArthur Communicative Development Inven-tories:
User's guide and technical manual. San Diego: Singular Publishing Group.
Gentner, D. (1982). Why are nouns learned before verbs: Linguistic relativity
versus natural partitioning. In S.A. Kuczaj II (Ed.), Language development:
Vol. 2. Language, thought and culture. Hillsdale, NJ: Erlbaum.
Goldfield, B.A., & Snow, C. (1985). Individual differences in language
acquisition. In J. Gleason (Ed.), Language development. Columbus: Merrill
Publishing Co.
Goldin-Meadow, S., & Mylander, C. (1985). Gestural communication in
deaf children: The effects and non-effects of parental input on early language
development. Monographs of the Society for Research in Child Development.
Goldin-Meadow, S., Seligman, M.E.P., & Gelman, R. (1976). Language in
the two-year-old. Cognition, 4, 189-202.
Gopnik, A. (1981). Development of non-nominal expressions in 1-2 year olds:
Why the first words aren't about things. In P. S. Dale & D. Ingram (Eds).,
Child language: An international perspective. Baltimore, Maryland: University
Park Press.
Gopnik, A., & Choi, S. (1990). Do linguistic differences lead to cognitive
differences? A cross-linguistic study of semantic and cognitive develop-ment.
First Language, 10, 199-215.
Gopnik, A., & Choi, S. (in press). Names, relational words, and cognitive
development in English- and Korean-speakers: Nouns are not always learned
before verbs. In M. Tomasello & W. Merriman (Eds.), Beyond names for
things: Young children's acquisition of verbs. Hillsdale, NJ: Lawrence Erlbaum.
Greenfield, P.M., & Smith, J.H. (1976). The structure of communication
in early language development. New York: Academic Press.
Harris, M. (1993, July). The origins of early lexical comprehension. Paper
presented at the Sixth International Congress for the Study of Child Language,
Trieste, Italy.
Howe, C. (1976). The meanings of two-word utterances in the speech of young
children. Journal of Child Language, 13, 29-47.
Iverson, J., Capirci, O., & Caselli, M.C. (1994). From communication
to language in two modalities. Cognitive Development, 9(1), 23-43.
Launer, P.B. (1982). "A plane" is not "to fly": Acquiring
the distinction between related nouns and verbs in American Sign Language.
Unpublished doctoral dissertation, City University of New York.
Lieven, E., & Pine, J.M. (1990). [Review of From first words to grammar:
Individual differences and dissociable mechanisms - Bates, E., Bretherton,
I., & Snyder, L.]. Journal of Child Language ,17, 495-501.
MacWhinney, B., & Bates, E. (Eds.) (1989). The crosslinguistic study
of sentence processing. New York: Cambridge University Press.
MacWhinney, B., Bates, E., & Kliegl, R. (1984). Cue validity and sentence
interpretation in English, German and Italian. Journal of Verbal Learning
and Verbal Behavior, 23, 127-150.
Markman, E.M. (1989). Categorization and naming in children: Problems of
induction. Cambridge, MA: MIT Press/Bradford Books.
McCarthy, D. (1954). Language development in children. In L. Carmichael
(Ed.), Manual of Child Psychology. New York: John Wiley.
Miller, J.F., & Chapman, R. (1981). The relation between age and mean
length of utterances in morphemes. Journal of Speech and Hearing Research,
24, 154-161.
Nelson, K. (1973). Structure and strategy in learning to talk. Monographs
of the Society for Research in Child Development, 38, (1 & 2, Serial
#149).
Nelson, K. (1981). Individual differences in language development: Implications
for development and language. Developmental Psychology, 17, 170-187.
Nelson, K., & Lucariello, J. (1985). The development of meaning in first
words. In M. D. Barrett (Ed.), Children's single-word speech. New York:
Wiley.
Ninio, A. (1993). On the fringes of the system: children's acquisition of
syntactically isolated forms at the onset of speech. First Language, 13,
291-313.
Ninio, A. (1994). Words with holes: the acquisition of the predicateness
of predicates. Emory Conference on Cognitive and Functional Approaches to
Grammatical Development, Emory University, Atlanta, Georgia, January 1994.
Pae, Soyeong (1993). Early vocabulary in Korean: Are nouns easier to learn
than verbs? Unpublished doctoral dissertation, University of Kansas.
Pine, J.M., & Lieven, E. (1990). Referential style at 13 months: Why
age-defined cross-sectional measures are inappropriate for the study of
strategy differ-ences in early language development. Journal of Child Language,
17, 625-631.
Rescorla, L. (1981). Category development in early child language. Journal
of Child Language, 8, 225 - 238.
Scollon, R. (1976). Conversations with a one-year-old: A case study of the
developmental foundation of syntax. Honolulu: University Press of Hawaii.
Shore, C. (1995). Pathways into language: Individual differences in language
development. Thousand Oaks, CA: Sage Publications.
Slobin, D. (Ed.). (1992). The crosslinguistic study of language acquisition
(Vol. 3). Hillsdale, NJ: Erlbaum.
Slobin, D. (Ed.). (1985). The crosslinguistic study of language acquisition
(Vols. 1 & 2). Hillsdale, NJ: Erlbaum.
Tomasello, M. (1992). First verbs: A case study of early grammatical development.
Cambridge [England] ; New York: Cambridge University Press.
Volterra, V. (1979). First words in language and action: a qualitative look.
In Bates, E., Benigni, L., Bretherton, I., Camaioni, L., & Volterra,
V. (1979). The emergence of symbols: Cognition and communication in infancy.
New York: Academic Press, 141 - 182.
Volterra, V., & Caselli, M.C. (1986). First stage of language acquisition
through two modalities in deaf and hearing children. Italian Journal of
Neurological Science, Supplement 5, Maturation and Learning, 109-115.