Content uploaded by Mutsumi Imai
Author content
All content in this area was uploaded by Mutsumi Imai on Oct 20, 2014
Content may be subject to copyright.
17 Revisiting the Noun-Verb Debate:
A Cross-Linguistic Comparison
of Novel Noun and Verb Learning
in English-, Japanese-, and Chinese-
Speaking Children
Mutsumi Imai, Etsuko Haryu, Hiroyuki Okada,
Li Lianjing, and Jun Shigematsu
To understand the nature of lexical development, it is crucial to in-
vestigate how children learn a wide range of word classes, including nouns, verbs,
and adjectives, along with closed class words such as prepositions and classifiers.
An important question is whether a particular type of concept, over others, uni-
versally invites children to name it at early stages of word learning to serve as the
entry point into language, that is, whether there is, in Gentner’s words, “an initial
set of fixed hooks with which children can bootstrap themselves into a position
to learn the less transparent aspects of language” (Gentner, 1982). A number of
researchers have proposed that basic-level object nouns serve such a function and
argue that the basic-level object categories reflect the natural clustering of the
world and are hence conceptually privileged (e.g., Gentner, 1982; Gentner &
Boroditsky, 2001; Rosch, 1978). On the other hand, some researchers disagree
with this view, arguing that the relative dominance of a particular type of words
in the early lexicon is determined by distributional and structural properties of
children’s native language, and hence that the class of words that children learn
earliest should differ across different languages (e.g., Choi & Gopnik, 1995;
Tardif, 1996).
In this chapter we address this issue, reporting results from a series of cross-
linguistic studies that examined how English-, Japanese-, and Chinese-speaking
children generalize newly learned nouns and verbs. Based on the results, we eval-
uate the two competing positions in the noun-verb debate. We then explore
450
Q1
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 450
universal and language-specific factors that affect the ease or difficulty of early
noun and verb learning.
Current Debate Concerning the Universal Advantage in
Noun Learning
In her natural partition hypothesis, Gentner (Gentner, 1982; Gentner & Boroditsky,
2001) has proposed that nouns will predominate over verbs in children’s initial vo-
cabularies because the meanings of concrete nouns are easier for children to dis-
cover than those of verbs and other relational terms, because the concepts which
nouns typically denote (e.g., balls, cups, dogs) are individuated from the environ-
ment in the world and hence are already prepackaged to be named when children
start learning language. In contrast, relational concepts, which are typically denoted
by verbs, are not as obviously accessible as basic-level object categories, as verbs are
cross-linguistically variable in their semantic structure and their meanings cannot be
learned independently of the semantic system of the language. Gentner thus pre-
dicted that noun learning is universally easier than verb learning, independent of the
structural and distributional properties of the language that children are learning.
Some researchers have challenged this view, however, arguing that verbs can
be learned faster and easier than nouns if verbs predominate in the input (Choi &
Gopnik, 1995; Tardif, 1996). For example, in some languages, including Chinese,
Korean, and Japanese, arguments (both subjects and objects) are often dropped
from a sentence. As a result, verbs tend to appear more frequently than nouns
in the maternal input (e.g., Choi & Gopnik, 1995; Kim, McGregor, & Thompson,
2000; Ogura, 2001; Tardif, 1996). Based on this observation, some researchers pre-
dicted that children who are learning these languages should learn verbs earlier,
hence more easily, than nouns (e.g., Choi & Gopnik, 1995;Tardif, 1996).
Mixed results have been reported with respect to whether Chinese-, Korean-,
or Japanese-speaking children learn verbs earlier than nouns, however. Some stud-
ies reported the predominance of verbs in Mandarin-speaking (Tardif, 1996) and
Korean-speaking (Choi, 2000) children’s early productive vocabularies. However,
other studies reported approximately equal proportions of nouns and verbs (Choi
& Gopnik, 1995). Yet some other studies reported that the proportion of nouns
was higher than verbs in Korean (Au, Dapretto, & Song, 1994; Kim et al., 2000)
and in Japanese (Ogura, 2001; Yamashita, 1997) children.
Limitations of Checklist or Production Data as an Index
for Relative Noun-Verb Advantage
Why are there such discrepancies in the literature concerning the noun-verb de-
bate? It seems that the discrepancies have in large part arisen from the fact that
REVISITING THE NOUN-VERB DEBATE 451
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 451
the studies reviewed above relied on either maternal reports using checklist inven-
tories or production data. While these methodologies no doubt are very useful and
provide us with invaluable data for understanding how children build up their vo-
cabulary, there are some limitations inherent in these methodologies when we rely
on them as a sole index for relative ease or difficulty of noun and verb learning.
Relative Use of Nouns and Verbs Differ Across Different Contexts
In the studies using either checklist or production data, the relative proportion of
nouns and verbs in children’s early vocabulary is taken to be an index of the rela-
tive ease or difficulty of noun learning and verb learning. However, the relative
proportion of nouns and verbs may vary depending on the context in which the
production data are collected even within a single language (Choi, 2000; Ogura,
2001; Tardif, Gelman, & Xu, 1999). Thus, samples collected in different contexts
may yield different results. Given this, it is difficult to draw a definitive conclusion
about whether nouns are learned more easily than verbs (or vice versa) in a partic-
ular language based on production data alone.
Using a Word Does Not Guarantee the Full Mastery of the
Meaning of the Word
A further limitation in using the proportion of nouns and verbs in children’s vo-
cabulary as the index of relative ease of noun versus verb learning is that the fact
that a word learner “knows” a word does not necessarily mean that she has ac-
quired the full meaning of the word. In other words, children may use a particular
word in a particular situation appropriately, but the total number of situations in
which they could use the word appropriately may be much more limited that sit-
uations in which the word is used by adults (Bowerman, 1980; Dromi, 1987). For
example, a child may use the verb throw when she sees someone throwing a ball,
but she may not fully understand that when one throws something, one can throw
not only balls but any object that can be held with one’s hand. She may also not
realize that one can throw things in many different ways (e.g., with two hands or
one hand, underhand or overhand, etc.), but one cannot throw things using legs
(Forbes & Poulin-Dubois, 1997; see also Huttenlocher, Smiley, & Charney, 1983).
This problem applies to verbs more severely than nouns because verb meanings
are often made up of a combination of abstract semantic features.
Here, we report an interesting anecdote that clearly shows that using a verb in
certain context does not mean that the child understand the full, adult-like mean-
ing of it. In Japanese, there are two verbs, ageru and kureru,corresponding to
the English verb give.However, the two must be clearly distinguished and cannot
be used interchangeably. Ageru is used when the giver gives something to someone
other than the speaker, while kureru is used when the giver gives something to the
speaker. Thus, if a mother gives a candy to a child, the child must say, “Okaasan
452 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 452
(Mother) ga (nominal particle) ame (candy) wo (accusative particle) kureta (give
me),” but when the mother gives a candy to the child’s sister, the child must say,
“Okaasan ga ame wo ageta (give someone other than myself ).” Mika, a 4-year-old
girl, had been using ageru since she was 2-1/2 years old, and the adults around her
had assumed that she knew the meaning of this verb. However, at 42 months old,
she said “Ojiichan (grandfather) ga (nominal particle) ame (candy) wo (accusative
particle) age-ta (ageru-past, “give to other than self ”)” when noting that her grandfa-
ther gave her a candy, where she should have said “ojiichan ga kureta,” because the
receiver of the candy was herself, the speaker. Her aunt (the first author) asked her
to repeat what she had just said to confirm if it was just a slip of the tongue. She
used ageru again with confidence. This shows that even though she used ageru cor-
rectly in many situations, her representation of the meaning was not quite adult-like
in the sense that it did not include the crucial semantic component of whether or
not the recipient of giving is the speaker. Nonetheless, if her mother had been asked
to fill out a vocabulary inventory when Mika was only 2 years old, she would have
definitely checked off ageru,thinking that Mika had already learned the word.
Our Approach: Learning New Nouns and New Verbs Introduced
During a Dynamic Action Event
Given the limitations in using the relative dominance of nouns and verbs in chil-
dren’s earliest vocabulary as the index of the relative ease or difficulty of noun and
verb learning in young children, we decided to approach the question by asking
how easily (and correctly in the adult sense) children learning different languages
extend newly learned nouns and verbs to new instances in experimental settings.
A merit of this paradigm is that it allows us to assess the general knowledge chil-
dren possess about the given word class rather than the knowledge of the mean-
ings of specific words, which must vary across different individuals depending on
their specific experiences with the words.
The principles governing the extension of nouns and verbs are very different.
An object can appear in many different actions. For example, a ball can be rolled,
thrown, kicked, and so on. Thus, when a noun is introduced in a scene in which
the referent object is used in a particular action, in extending it, children must
know that the noun should be extended on the basis of the sameness of the object
per se but not on the basis of the sameness of the action with which the object is
used. Likewise, an action can be done with many different objects. For example,
we can throw a ball, a Frisbee, a stone, a disk, or almost anything we can lift up
with our hands.Thus, in extending a verb that has been mapped onto an action in-
volving an object, the object must be separated from the action and be treated as a
variable that can be changed.1
In this chapter, we report the results of a cross-linguistic study that investi-
gated how young children learning English, Japanese, or Chinese map novel nouns
REVISITING THE NOUN-VERB DEBATE 453
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 453
and verbs onto ongoing action event scenes and how they extend the novel
words. Specifically, we asked whether children understand two basic principles for
noun extension and verb extension: (a) a noun gets extended on the basis of the
sameness of the referred object, and the particular action in which the object is
used is not relevant for noun extension; (b) a verb gets extended on the basis of
the sameness of the action, and the objects (both the agent and theme object) that
appear in a particular action event are variables that can be replaced across differ-
ent instances of the event referred to by the verb. Before describing the studies,
however, we briefly discuss some linguistic properties of the three languages and
discuss predicted patterns based on the linguistic comparisons.
Distributional and Structural Properties of English, Japanese,
and Mandarin Chinese
Comparing children learning English, Japanese, and Chinese is extremely interesting
because the three languages are different from one another along the dimensions
that have been assumed to affect the relative ease or difficulty of verb learning by
children. Argument dropping is allowed in Japanese and Chinese but not in English.
This means that in Japanese and Chinese, when the arguments can be inferred from
the context, it is possible that a verb is the only word in the sentence. As a conse-
quence of this linguistic property, children learning Japanese or Chinese tend to
hear verbs more frequently than children learning English do.As mentioned earlier,
because of this distributional property, some researchers predict that children learn-
ing Japanese or Chinese will learn verbs earlier (and hence more easily) than nouns
(Choi & Gopnik, 1995;Tardif, 1996). However, at the same time, this property may
lead to the opposite prediction. It has been proposed that inferring the meaning
of a verb is very difficult even for adults without cues from the argument struc-
ture (Gillette, Gleitman, Gleitman, & Lederer, 1999), and that children do utilize
the structural cues in inferring verb meanings (e.g., Fisher, 1996). Thus, one could
make the prediction that verb learning should be more difficult for children who
are learning a language that occasionally allows argument dropping. (In fact, in
Japanese, argument dropping occurs more than occasionally—it is usually dropped
when the speaker believes that the arguments can be inferred from the context.)
The second dimension is the presence of morphological inflection in verbs. On
this dimension, Chinese contrasts not only to English but also to Japanese. While
verbs are inflected in both English and Japanese, they are not in Chinese. In other
words, nouns and verbs are not morphologically distinguished in Chinese (Erbaugh,
1992). Remember that in Chinese and Japanese, verb arguments are often dropped,
and the verb alone can constitute a sentence in the language. In the case of Japanese,
even when this occurs, verbs can be identified by inflectional morphology. That
is,when a verb is produced without the arguments, as in “Mite (Look), X-teiru
(X-ing),” one can tell that the word Xis a verb. However, in Chinese, when a
word is produced on its own (and this can happen in a conversational discourse), it
454 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 454
is difficult to tell whether it is a noun or a verb. In other words, one can identify a
novel word as a verb only when it is embedded in an argument structure (see Li,
Bates, & MacWhinney, 1993, for a discussion of how Chinese-speaking adults deter-
mine grammatical classes of words and their thematic roles in sentence processing).
It is of great theoretical interest to see whether the morphological simplicity of Chi-
nese makes verb learning even easier when compared to Japanese (Tardif, 1996).
Given these distributional and structural properties of English, Japanese, and
Chinese, comparing children of these three language groups should provide us
with invaluable insights onto the question of what linguistic factors might influ-
ence early verb learning.
How Japanese-, Chinese-, and English-Speaking Children Map
Novel Nouns and Verbs Onto Dynamic Action Events:
Cross-Linguistic Comparisons
The Task and Procedure
In this study, 3- and 5-year-old children from three language groups—Japanese,
Mandarin Chinese, and English—were tested (Imai, Haryu, & Okada, in press;
Mayer et al., 2003; Haryu et al., 2004). The children were all from monolingual
families living in Japan (a suburban Tokyo metropolitan area), China (Beijing),
and the United States (Philadelphia), respectively.
Six sets of video action events served as stimulus materials. Each set consisted
of a standard event and two test events. In each standard event, a young woman
was doing a novel repetitive action with a novel object. The two test events were
variants of the standard event. In one, the same person was doing the same action
with a different object (action-same-object-change, henceforth AS) from the stan-
dard event. In the other, the person was doing a different action with the same ob-
ject (action-change-object-same, henceforth OS; see figure 17.1 for a sample set;
also see table 1 in Imai et al., in press, for a complete description of the actions and
the objects used in the stimuli).
The standard event was presented on a computer monitor for about 30 sec-
onds. While watching the standard event, a child heard either a novel noun or a
novel verb, depending on the condition. The child was then shown the two test
videos and was asked to which event the target word should be extended.
Conditions and Instructions
Our major interest was to examine whether Japanese-, Chinese-, and English-
speaking children understand the basic principles governing noun generalization
and verb generalization, so in all three language groups, children learned either six
novel nouns or six novel verbs. In addition, we wished to see whether dropping of
the verb arguments affects children’s performance in learning novel verbs.Thus, in
REVISITING THE NOUN-VERB DEBATE 455
Q2-3
Q4
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 455
English and Japanese, we presented the verbs in two different forms: one with full
arguments (full argument verb condition) and the other with no arguments (bare
verb condition). In providing the arguments, in English, the pronoun she served as
the subject, and it as the object of the sentence (e.g., “Look, she is X-ing it”). In
Japanese, the word oneesan (girl) is used for the subject, and nanika (something)
was used in referring to the novel object.
As we noted earlier, in Chinese, when both arguments are dropped, one can-
not tell whether the word is a verb or a noun. We thus conducted only the noun
and the full argument verb conditions. In the noun and the full argument verb
conditions, special care was taken so that there was absolutely no ambiguity over
whether the target word was a noun or a verb, respectively. In presenting the verb
in the full argument verb condition, an aspect marker zai,which marks the imper-
fective aspect and is usually used in expressing an ongoing action, accompanied
the verb along with the subject ayi (the girl) and the theme object “yi-(one) ge
(generic classifier) dongxi (thing)” (i.e., something).2The conditions carried out in
each language and the corresponding instructions are given in table 17.1.
Predicted Patterns
What patterns are predicted? If children understand that a noun refers to an ob-
ject and that the particular action in which it is used is irrelevant to the meaning
456 HOW LANGUAGE INFLUENCES VERB LEARNING
Figure 17.1. Sample stimulus set used in the novel noun/verb learning study.
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 456
REVISITING THE NOUN-VERB DEBATE 457
of the noun, we would expect them to select the OS event when asked to extend
a novel noun. If they understand that a verb maps to an action, and that the agent
and the object of the action event are variables that can be changed across differ-
ent instances of the event referred by the verb, they would select the AS event in
extending a novel verb.
The question of most importance is whether children would perform equally
well in learning novel nouns and verbs. If there is any asymmetry between noun
learning and verb learning, it is extremely interesting to see whether there are any
cross-linguistic differences in the pattern of novel noun and verb learning. If the
universal noun advantage view is correct, we may expect that children in all three
languages perform better in learning new nouns than new verbs. On the other
hand, if the relative ease of noun and verb learning is determined by distributional
properties of the input language, we may expect that Japanese- and Chinese-
speaking children do better in extending new verbs than English-speaking chil-
dren. If structural properties of language such as morphological simplicity affect
the ease of word learning (Tardif, 1996), we might expect that Chinese children
perform even better than Japanese children, as Chinese verbs are morphologically
simpler than Japanese verbs.
Children’s Performance in Novel Noun Learning
Children in all three languages in both age groups succeeded in the novel noun
extension task. They extended a novel noun to the same object/different action
event, and there was no cross-linguistic or developmental difference. Thus, 3-year-
olds, regardless of the language they are learning, have a clear understanding that
nouns refer to objects and that the actions in which the referent object is used are
irrelevant to the noun meaning.
Children’s Performance in Novel Verb Learning
In contrast to the success in the novel noun learning task, in none of the language
groups were 3-year-olds able to successfully extended novel verbs (see table 17.2).
It is not until they are 5 years old that children reliably can extend a novel verb to
an event involving the same action but a different object. In this sense, the results
suggest that learning a new verb is more difficult than learning a new noun. With
this overall pattern in mind, however, we should also note that the performance
of Japanese-, Chinese-, and English-speaking children was not totally uniform. In
fact, we found intriguing cross-linguistic differences in the pattern of novel verb
learning. Specifically, the condition in which 5-year-olds successfully extended
newly learned verbs varied across the three languages, which in turn suggests that
children speaking different languages rely on different cues in learning verbs. Be-
low, we describe how children of the three language groups generalized novel
verbs in our task, starting with Japanese children.
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 457
Table 17.1 Conditions carried out in the three languages and the corresponding instruction.
Language Condition Instruction during verb presentation Instruction for test
English Noun “Look! This is an X!” “Where is the X? Can you point to the X?”
Bare verb “Look! X-ing” “Where is X-ing?”
Full argument verb “Look! She is X-ing it” “Where is she X-ing it?”
Japanese Noun “Mite (Look)! X-ga (nominal particle) “X-ga aru (exist)-no (nominal particle)-wa
aru (exist)” (Look! There is (an) X)(topic particle) docchi (which movie)?”
(In which (movie) is there (an) XK?)
Bare verb “Mite (Look)! X-teiru (X-progressive)” X-teiru-no (genetive particle)-wa (topic
(Look, X-ing) particle) docchi (which movie)?” (In which
(movie) is (she) X-ing?)
Full argument verb “Mite (Look)! Oneesan (girl) ga (nominal “Oneesan (girl) ga (nomial particle) nanika
particle) nanika (something)-wo (accusative (something)-wo (accusative particle)
particle) X-teiru (X-progressive)” (Look, she X-teiru (X-progressive) no (genetive
is X-ing something) particle)-wa (topic particle) docchi (which
movie)?” (In which (movie) is she X-ing
something?)
Chinese Noun “Ni (you) kan (look)! Nali (there) you (exist) “Na (which) zhang (quantifier) tu (picture)
ge (classifier) X”(Look! There is (an) X)li (within) you (exist) ge (classifier) X?” (In
which picture is there (an) X?)
458
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 458
Full argument verb zai only “Ni (you ) kan (look)! Ayi (girl) zai “Na (which) zhang (classifier) tu (picture)
(progressive) Xyi (one) ge (classifier) li (within) ayi (aunt) zai (progressive)
dongxi (thing) ne (mode marking particle)” X yi-(one) ge (classifier) dongxi (thing)?” (In
(Look, (a) girl is X-ing something) which picture is she X-ing something?)
Full argument verb 3 sentences “Ni (you) kan (look)! Ayi (girl) zai (progressive) “Na (which) zhang (classifier) tu (picture)
with different auxiliaries Xyi (one) ge (classifier) dongxi (thing) ne li (within) ayi (aunt) zai (progressive)
(mode marking particle)” (Look, (a) girl is X yi-(one) ge (classifier) dongxi (thing)?”
X-ing something)
“Ni (you) kan (look)! Ayi (girl) zhengzai “Ayi (aunt) zai (progressive) Xyi-(one)
(progressive) Xyi (one) ge (classifier) dongxi ge (classifier) dongxi (thing) de
(thing) ne (mode marking particle)” (Look, (progressive) tu (picture) shi (is) na
(a) girl is X-ing something) (which) yi-(one) ge(classifier)” (In which
picture is she X-ing something?)
“Ni (you) kan (look)! Ayi (girl) yizhizai
(progressive) Xyi (one) ge (classifier) dongxi
(thing) ne (mode marking particle)” (Look,
(a) girl is always X-ing something)
459
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 459
Japanese-Speaking Children
Five-year-olds, but not 3-year-olds, showed understanding of the principle that
verbs get extended on the basis of the sameness of actions, and that the objects that
appear in a particular action event are variables that can be replaced across differ-
ent instances. While the 5-year-olds extended a novel verb to the action-same-
object-change test at reliably above chance level, the 3-year-olds showed only
chance-level performance (Imai et al., in press). To our surprise, Japanese children
performed better when the verb was presented without the arguments than when
it was presented with an explicit mention of the arguments (Haryu et al., 2004).
A series of follow-up studies were conducted to specify the nature of the
Japanese 3-year-olds’ chance level performance in novel verb generalization (Imai
et al., in press, Study 2). Chance-level performance in a forced-choice task must
be interpreted with special caution because there are multiple possibilities to
account for this phenomenon. One possibility is that some children may have
mapped the verb to the object rather than the action ignoring the verb morphol-
ogy (i.e., -teiru). A second possibility is that they might have thought that a novel
verb could refer not only to the action but also to the object used in the action. If
the 3-year-olds in this study had indeed made this assumption, they should have
found both test events plausible. A third possibility, in contrast to the second, is
that 3-year-olds may have thought that the verb referred to the action only with
that particular object. In other words, the 3-year-olds were labeling the action-
object interaction. If this was the case, they should have found neither test event
plausible. Finally, the task may have involved a greater information-processing
load than 3-year-olds could handle. To make a choice in this paradigm, children
had to mentally process three ongoing action events simultaneously, holding the
standard event in their working memory while watching the two test events.
460 HOW LANGUAGE INFLUENCES VERB LEARNING
Table 17.2 Proportion of action-same-object-change responses in each of the noun,
bare verb, full argument verb conditions in Japanese- and English-speaking 3- and
5-year-olds.
Language Age Noun Bare verb Full argument verb
Japanese 3-year-olds 0.27
b
0.64 0.41
5-year-olds 0.08
b
0.85
a
0.69
English 3-year-olds 0.14
b
0.49 0.42
5-year-olds 0.09
b
0.56 0.70
a
Chinese 3-year-olds 0.25
b
—0.06
b
5-year-olds 0.06
b
—0.24
b
7-year-olds — — 0.52
9-year-olds — — 0.72
b
a
Significantly above chance, p<0.05.
b
Significantly below chance, p<0.05.The children selected the
object-same choice significant above chance.
Q5
Q6
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 460
Given that all three stimuli were ongoing video events, this may have overloaded
the 3-year-olds’ information processing capacity.
To circumvent the processing load problem, we replicated the original study
using a yes-no paradigm. That is, instead of selecting one of the two scenes, chil-
dren saw one test event at a time while watching the standard next to it.Thus, the
demand of information processing should have been greatly decreased in this task.
To examine the three possibilities introduced above, we included a scene of the
object lying still on a table (still object, henceforth SO) in addition to the AS and
OS events in the test. If the 3-year-olds thought that the new verb would refer to
the object, they would extend the word to the SO scene but not to the action
with a different object (AS). If they thought that a novel verb could refer to either
an action or an object involved in the action, they should accept all of the test
items, including the SO test. On the other hand, if they thought that a novel verb
refers to an action with a particular object, they should reject any of the test stim-
uli as a referent of the verb in this study.
It turned out that the Japanese 3-year-olds clearly rejected the still object,
which means that they did not think that a verb could refer to an object. The rate
of “yes” responses was not different across the AS test and OS test, both of which
fell at the chance level. These results suggest that Japanese 3-year-olds assume that
verbs refer to the action-object interaction. In their verb meaning representation,
the core meaning (i.e., the action) is not separated from the theme object, and as a
result, their generalization of novel verbs is overly conservative: they do not ex-
tend a novel verb to the same action if the object involved in the action is replaced
with a different object.
Given these results, in another follow-up study we examined whether Japan-
ese 3-year-olds would extend a verb to the same action when the agent was
changed but the same object was retained. This question is worth examining, since
Maguire et al. (2002) recently reported that the 18-month-olds did not extend
the verb to the exact same action done by a new person even after hearing the
verb with the identical action acted by four different people. We thus tested
whether 3-year-olds would extend a verb to a scene in which a different actor was
doing the same action as the original event with the same object, again using the
yes-no procedure (Imai et al., in press, Study 3). In this case, the children had no
problem extending the verbs to the same action.
In summary, the pattern of the results from Japanese children suggest that 3-
year-olds do tolerate a change in the actor but are unwilling to extend a newly
learned verb to a new instance when the theme object is changed. This indicates
that they do not fully understand the basic principle for verb extension, that
verbs are extended on the basis of the action independent of the object. Five-year-
olds did seem to understand this principle well and were able to apply it immedi-
ately in a novel verb learning situation. Interestingly, however, they were able
to do so when the arguments of the verb were omitted but not when they were
explicitly mentioned. (We will discuss the possible reason later in the chapter.)
REVISITING THE NOUN-VERB DEBATE 461
Q7
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 461
English-Speaking Children
In spite of the linguistic differences between English and Japanese, English-speaking
children’s performance in the novel verb extension task was overall very similar
to that of Japanese children: 3-year-olds showed chance-level performance, while
5-year-olds were able to extend a novel verb to the AS test (Mayer et al., 2003).
There was one important difference between Japanese-speaking and English-
speaking groups, however. Unlike Japanese children, who performed above chance
in the bare verb condition but not in the full argument verb condition, English-
speaking 5-year-olds were able to extend the verb to the AS test reliably above
chance only when the verb arguments were specified (“Look, she is X-ing it”). They
selected the AS tests only 55.6% of the time when the verb arguments were omitted.
This difference suggests that the structural characteristics of children’s native lan-
guage might influence the structural form in which children expect to hear a verb.
Chinese-Speaking Children and Adults
The results from Chinese-speaking children were utterly surprising. Unlike
Japanese- and English-speaking children, both 3- and 5-year-olds selected the OS
test at highly above chance level in the full argument verb condition. This means
that they mapped the novel verb to the object instead of the action. As shown
in table 17.2, the Chinese-speaking 3- and 5-year-olds consistently selected the
OS test regardless of whether the word was presented as a noun or a verb.
Given these surprising results from Chinese-speaking children, we tested
monolingual Mandarin-speaking adults living in Beijing, China, to see how they
performed in the task. The Chinese-speaking adults who were assigned to the verb
(with full arguments) condition selected the AS test 100% of the time. These re-
sults suggest (1) that it was perfectly clear to Chinese-speaking adults that the tar-
get novel word presented in the full argument verb condition was indeed a verb
and (2) that there was a large developmental shift from an object-naming bias to
an action-naming bias in Chinese speakers.
To identify the age at which this shift takes place, we further tested 7- and 9-
year-old Mandarin Chinese–speaking children in the full argument verb condition
and bare verb condition. In the full argument verb case, the 7-year-olds selected
the AS test at chance (52.2%).At 9 years of age, Chinese-speaking children finally
extended a novel verb to the AS test significantly above chance level (72%).
Did Extralinguistic Cues Help?
Given the surprising results from the Chinese speakers, we conducted a few dif-
ferent versions of the full argument verb condition, trying to find a condition
under which Chinese children (at least 5-year-olds) could reliably extend the verb
to the action even when the object is changed.
462 HOW LANGUAGE INFLUENCES VERB LEARNING
Q8
Q9
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 462
Table 17.3 Proportion of the AS response in Chinese-speaking children in the noun, bare verb, and full argument verb conditions tested with
the original stimulus set and the revised stimulus set in which the object-holding segment was removed.
Full Argument Verb Full Argument Verb Full Argument Verb
2-syllable word, 1 syllable word, 1-syllable word,
Stimuli Age Noun zai only zai only 3-sentence frames
Original 3-year-olds 0.25
b
0.06
b
0.25
b
0.34
5-year-olds 0.06
b
0.24
b
0.34 0.39
7-year-olds — 0.52 — —
9-year-olds — 0.72
a
——
Object-holding 3-year-olds 0.32
b
—— 0.41
part removed 5-year-olds 0.27
b
—— 0.88
a
a
Significantly above chance, p<0.05.
b
Significantly below chance p<0.05. The children selected the object-same choice significant above chance.
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 463
First, the number of syllables in the word was changed. In the original study,
we prepared novel words (both nouns and verbs) with two syllables. This was be-
cause two-syllable words were most common for both nouns and verbs. However,
verbs referring to simple actions, such as jump,kick,and run,tend to be monosyl-
labic words. Thus, we constructed monosyllabic nonsense words and replicated
the full argument verb condition with them. Although this manipulation lifted
the AS response a little, no statistically reliable difference was obtained.
We then provided additional linguistic cues to indicate that the novel word was
a verb. In the original instruction in the full argument verb condition, the experi-
menter said “Ayi (girl) zai (progressive) X(novel word) yi (one) ge (classifier) dongxi
(thing) ne (mode marking particle)” (She is X-ing something). In this instruction, the
novel word Xcould be unambiguously identified as a verb by the structure of the
sentence, in particular, by the word order and the presence of the aspect marker zai.
However, zai is also used as a verb, meaning roughly “to exist” or “to be present (at a
place).” In this case, the word that comes after zai is usually a noun. Young children
thus could have been confused because of this homonymous use of zai and mistak-
enly assumed that the word was a noun. We thus presented the verb in three differ-
ent sentences using three different auxiliaries, namely, zai,zhengzai,and yizhizai,all
of which mark the progressive aspect, to provide even clearer and stronger clues that
the novel word was a verb. However, again, this manipulation did not bring a statisti-
cally reliable increase in the Chinese-speaking children’s performance.
Chinese-Speaking Children Are Sensitive to Subtle
Extralinguistic Cues
So far, the results suggested that Chinese children as old as 5 years of age could
not extend newly learned verbs to the same action in the face of a change in the
object even when a novel word was presented in such a way as to make it clear
that it was a verb. It is possible that the lack of morphological distinction between
nouns and verbs makes it difficult for Chinese children to extract the extension
principle for verbs, in contrast to the general assumption in the literature that Chi-
nese is a verb-friendly language.At the same time, there must be conditions under
which Chinese preschoolers, especially 5-year-olds, can extend to novel verbs to
the action in the face of a change in the object. What cue do they need in addition
to linguistic cues? We suspected that that the difficulty in identifying a word’s
grammatical form class solely from structural cues such as morphological marking
or word order leads Chinese children to rely heavily on extralinguistic cues.
Upon reflection, in this light, there is one property of our stimuli that may
have given Chinese children a subtle cue that the object is the one that should be
attended to in the event. We created the standard video clips in such a way that
the actor holds the object for a moment (for about half a second) before starting
the action. We did so to make sure that children see the object clearly; the details
of the object may not be clearly observable when it is in motion. Of course, the
464 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 464
novel word was presented after the action started whether it was presented as a
noun or a verb. It should be stressed that the object was not unnaturally high-
lighted in the original stimuli, and it did not affect Japanese- or English-speaking
children. However, if Chinese children were very sensitive to extralinguistic, situa-
tional cues, this first segment of the video might have lead Chinese children to
think that the object was in a way “topicalized.”
To test this possibility, we removed the segment of the video clip in which the
actor was holding the object. In the new video, thus, the object is already in mo-
tion at the very start of the event presentation. We replicated the full argument
verb condition with Chinese-speaking 3- and 5-year-olds with this version of the
stimuli. We again presented the monosyllabic nonsense words in three sentences
with three different aspect marking auxiliaries to highlight that the word was a
verb to give the children as much linguistic support as possible.
Consistent with our expectation, this manipulation—removing the half-second
segment of the video clip in which the object was held still—indeed brought a dras-
tic change in Chinese-speaking children’s performance in the verb learning task and
their performance was now equivalent to the level of performance by Japanese-
or English-speaking children. The Chinese-speaking 3-year-olds were now at the
chance level, just like Japanese- and English-speaking 3-year-olds, and the Chinese-
speaking 5-year-olds now selected the AS test above chance level, just like their
Japanese- and English-speaking counterparts. We then conducted the noun condi-
tion with Chinese-speaking 3- and 5-year-olds using this revised stimuli to see
whether they could still select the OS test and confirmed that they had no problem
in doing so. Thus, it was not the case that Chinese-speaking children mapped the
novel word simply to the most salient component of the event, whether it was a
noun or a verb.They were able to extend a novel verb to the same action only when
the action was maximally salient, but even under this condition, they had no prob-
lem in mapping a novel noun to the object. Taken together, this shows that Chinese-
speaking 5-year-olds can extend novel verbs to the same action with a different
object, but they need support from contextual or perceptual cues in order to do so.
When contextual cues are in conflict with linguistic cues, it appears that Chinese-
speaking preschoolers rely more heavily on extralinguistic cues than linguistic cues,
unlike Japanese- or English-speaking children. It may be that the lack of obvious
morphological distinction between nouns and verbs leads Chinese-speaking chil-
dren to be more attentive to extralinguistic cues than Japanese or English-speaking
children are.
Implications for Theories of Lexical Development
and Verb Learning
In this chapter, we have approached the question of whether learning of nouns
(object names) is universally privileged over learning of verbs by asking how well
REVISITING THE NOUN-VERB DEBATE 465
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 465
children from three different language groups—Japanese, Chinese, and English—
learn novel nouns and verbs introduced during ongoing action events. Two find-
ings from the cross-linguistic studies were particularly important for the question:
(1) children in all three language groups succeeded in extending a novel noun to
the same object appearing in a different action at 3 years of age but did not suc-
ceed in generalizing a novel verb to the same action involving a different object
until 5 years of age or later; (2) 5-year-olds succeeded in the verb learning task,
but the condition under which they showed the best performance differed across
languages. We now discuss the implications we might draw from our results for
theories of verb learning as well as theories of lexical development in general.
Comparison of Novel Noun Learning and Novel Verb Learning
in Experimental Settings
In the studies reported in this chapter, 3-year-olds learning three different languages
could extend a newly introduced noun to the same object used in a different action,
while in no language group could 3-year-olds extend a newly introduced verb to the
same action carried out with a different object. A very similar pattern of results was
reported by Kersten and Smith (2002) with English-speaking children.As reviewed
earlier, in their study, English-speaking 3-year-olds were unwilling to extend a novel
verb to the same motion when the agent object was changed. Yet, parallel to the re-
sults from our own studies, the 3-year-olds in their study were willing to apply the
same noun to the same object even though it appeared in a different motion.
Our cross-linguistic novel noun/verb learning study and Kersten and Smith’s
(2002) study with English-speaking children both showed that young children
learn novel nouns more easily than novel verbs when the ease or difficulty was
measured by how well and how willingly young children extend newly learned
words to new instances (see also Golinkoff, Jacquet, Hirsh-Pasek, & Nandakumar,
1996). The advantage of novel noun learning over novel verb learning is also ob-
served when we compare the ease with which young children form object-label
associations and action-label associations. Werker, Cohen, Lloyd, Casasola, and
Stager (1998) demonstrated that 14-month-old infants were able to form asso-
ciations between novel labels and novel objects. Using the same experimental
paradigm as in Werker et al. (1998), Casasola and Cohen (2000) tested whether
14- and 18-month-old infants were able to form associations with novel labels and
novel actions and found that it was not until 18 months that infants could associ-
ate a novel action with a novel label.
With slightly older children, Childers and Tomasello (2002) also showed that
children learn the noun-object link easier than the verb-action link. In one condi-
tion, they showed 2-1/2-year-old English-speaking children three different novel
objects and taught them their noun labels. In another condition, they showed the
children three novel actions performed with three novel objects and taught them
three novel verbs corresponding to the three actions. In the third condition, they
466 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 466
simply taught three novel actions with three novel objects. The 2-year-olds re-
membered both objects and actions very well, yet their memory of the object-
noun link was twice as good as their memory of the action-verb link.
The contrast between novel noun learning and novel verb learning in young
children becomes even more prominent when we compare the conservatism chil-
dren repeatedly show in extending novel verbs to the liberal, yet principled, fash-
ion in which same-age or even younger children generalize newly learned nouns
(e.g., Golinkoff, Hirsh-Pasek, Bailey, & Wenger, 1992; Haryu & Imai, 2002; Imai &
Gentner, 1997; Imai, Gentner, & Uchida, 1994; Imai & Haryu, 2001; Landau,
Smith, & Jones, 1988; Markman, 1989). In particular, in different studies we have
demonstrated that Japanese-speaking 2-year-old children are able to flexibly map
a novel noun not only to a basic-level object category but also to a subordinate
category, a substance, or to a unique individual, depending on the perceptual or
conceptual nature of the named entity and its familiarity. This flexible pattern of
noun extension should be noted all the more because ontologically distinct sub-
classes of nouns—nouns denoting object kinds, nouns denoting substance kinds,
and nouns denoting unique individuals—are not grammatically distinguished in
Japanese (Haryu & Imai, 2002; Imai & Gentner, 1997; Imai & Haryu, 2001; see
also Imai & Haryu, 2004).3
In summary, previous research has shown that children map a novel noun to
its referent more easily than they map a novel verb to its referent (Cassasola &
Cohen, 2000; Childers & Tomasello, 2002; Werker et al., 1998). Furthermore, it
appears that young children find it easier to extend novel nouns than to extend
novel verbs, as they extend a newly learned noun to instances other than the orig-
inally named object in principled ways, while they are reluctant to extend a newly
learned verb to other instances with only a change in the object involved in the
action (Imai et al., in press; Kersten & Smith, 2002; Maguire et al., 2002). Thus,
the pattern of results from novel noun and verb learning seems to converge on the
conclusion that novel noun learning is easier than novel verb learning.
Influence of Language-Specific Properties on Verb Learning
So far, we have argued for the universal advantage of noun learning over verb
learning. Furthermore, there was a striking cross-linguistic similarity in the devel-
opmental pattern in novel verb learning. The 3-year-olds in any of the three lan-
guage groups were not successfully able to extend a novel verb to the same action
when the patient object was changed, but in all three languages in the optimal
condition, 5-year-olds succeeded in extending novel verbs. At the same time, how-
ever, the condition in which 5-year-olds succeeded as well as the ease with which
children learn novel verbs appears to be different across the three languages.
Following the common assumption in the literature that learning an argument-
dropping language gives an advantage to verb learning (Choi & Gopnik, 1995;Tardif,
1996), we had expected that Chinese- and Japanese-speaking children might perform
REVISITING THE NOUN-VERB DEBATE 467
Q10
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 467
better than English-speaking children in the novel verb learning task. Furthermore,
we had suspected that Chinese-speaking children might show even higher perfor-
mance than Japanese-speaking children because of the morphological simplicity of
Chinese verbs (Tardiff, 1996). Contrary to these predictions, Chinese-speaking chil-
dren did not perform any better than Japanese- or English-speaking children. In fact,
in conditions in which the action was not made particularly salient over the object,
Chinese-speaking children showed greater difficulty in learning novel verbs than
English- or Japanese-speaking children. Chinese-speaking children were extremely
sensitive to contextual cues when learning novel verbs for action events, and unless
the action was made very salient, Chinese-speaking 5-year-olds were not able to
map a novel verb to the action. It should be noted that Chinese-speaking children
did not determine the novel word form class solely based on contextual (or per-
ceptual) saliency of the event, as they were able to map novel nouns to the objects
under the action-salient situation. Given this, it seems reasonable to conclude
that nouns are easier to learn than verbs for Chinese-speaking children, just as for
Japanese- or English-speaking children. Furthermore, verb learning may be even
more difficult for Chinese-speaking children than Japanese- or English-speaking
children in conditions in which children must infer the meaning of a verb under cir-
cumstances in which strong contextual cues are not provided.
At present, we can only speculate on the reason Chinese-speaking children
were so sensitive to contextual cues, even to the extent that linguistic cues that are
apparent to Chinese-speaking adults were bluntly overridden. As discussed earlier,
one important linguistic property that sets Chinese against Japanese and English
is the lack of morphological distinction between nouns and verbs. Thus, unlike the
case with Japanese or English, Chinese speakers cannot determine the grammatical
form class of a word by morphological markings. Furthermore, even though word
order provides a cue for determining the form class of each word in the sentence, it
is only probabilistic: Although the basic word order is SVO, there other word or-
ders: OSV, SOV, and VOS are also found in the spoken language (Li et al., 1993).
Thus, to identify the grammatical class of each word in the sentence and assign its
thematic roles to it, Chinese speakers have to coordinate semantic, syntactic, semi-
morphological grammatical cues such as aspect markers, object markers, and passive
markers in “a complex system of mutual constraints” (Li et al., 1993, p. 193). This
linguistic property may lead Chinese-speaking children to rely more on extralinguis-
tic, contextual cues than on linguistic cues in novel word learning.
It is also noteworthy that the condition in which children performed best in
our novel verb extension task was different for English- and Japanese-speaking
children. The action events used in our research involve only three elements, an
actor, an action, and an object. Thus, even when children heard a verb without the
explicit mention of the subject and the object of the sentence, it should have been
easy to infer what the dropped arguments would have been. In Japanese, it is nat-
ural to drop the arguments when the speaker thinks that the hearer can infer
them from the observational or pragmatic cues. From the Japanese point of view,
468 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 468
it was obvious that the subject was the actor and the theme object was the novel
object, and hence it was more natural that the arguments be dropped in this case.
Japanese children in fact could have been distracted by hearing this unnecessary
information. In sharp contrast, English-speaking 5-year-olds extended the verb to
the AS test only when the verb was accompanied by the pronouns she and it.It
appears that the English-speaking children would not extend a novel verb when
the verb was presented in an unusual structural form, even though the arguments
of the verb could have been easily inferred from observation of the event.
Taken together, universally shared cognitive factors and language-specific lin-
guistic factors both matter for early word learning, but it appears that the former
is more prominent than the latter, given the striking similarity in the developmen-
tal pattern in the novel noun/verb learning task across the three languages of very
different linguistic properties.
Progressive Development of Verb Meanings
The results of our cross-linguistic novel noun/verb learning study support the
view that object naming is advantaged over verb learning, as discussed above. At the
same time, it needs to be explained why children as old as 3 years had so much dif-
ficulty in our verb extension task even though they comprehend and produce many
verbs (e.g., Choi & Bowerman, 1991; Golinkoff, Hirsh-Pasek, & Schweisguth, 2001;
Gopnik & Choi, 1995;Tardif, 1996). We have no intension of claiming that children
cannot learn verbs until 5 years of age. The point we would like to make instead is
that young children’s verb meanings develop progressively and that it takes a sub-
stantial amount of time before they finally obtain fully adult-like representations of
verb meanings. Remember the example of the Japanese verb ageru we described
earlier. The child started using the verb when she was 26 months old, but at 4-1/2,
she still had not acquired the fully adult-like meaning of it.
The pattern of success and failure in young children’s verb learning in experi-
mental settings also supports this idea. First, as demonstrated by Casasola and Co-
hen (2000), children seem to become able to associate a novel causal action to a
novel label at around 18 months of age. However, it is highly unlikely that 18-
month-olds would be willing to extend a newly learned verb to the same action
when one of the two objects (either the agent or the patient) or both objects are
changed, given the results by Maguire et al. (2002). As reviewed earlier, these re-
searchers demonstrated that English-speaking 18-month-olds were not willing to
extend a newly learned verb to the same intransitive action when the actor was
changed from the original scene even after they had been trained on the verb-
action association with four different agents. Golinkoff et al. (1996) demonstrated
that 3-year-old English-speaking children could extend a novel verb to the same
intransitive action by a different actor, but the 3-year-olds in Kersten and Smith’s
(2002) study failed to do so.4In our studies (Haryu et al., 2004; Imai et al., in
press; Mayer et al., 2003), in none of the three languages were 3- year-olds willing
REVISITING THE NOUN-VERB DEBATE 469
Q11
Q12-13
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 469
to extend a verb to the same action when an object (the agent in their case and
the theme object in our case) was changed.
In summary, previous results from novel verb learning studies suggest that
children do learn verbs as early as 18 months, but their representation of verb
meanings at initial stages is incomplete and fragile. It seems to take some time for
children to acquire the full, adult-like meaning for many verbs. It also seems to
take quite some time for the basic principle for verb meaning extension—that
verbs get extended on the basis of the same action or relation while the objects are
variables that can be changed from situation to situation—to become solidified
enough to be applicable under very stringent circumstances when external aids
such as social and perceptual support are not provided, and a new verb is associ-
ated only with a single instance.
Fast-Mapping May Not Be Beneficial for Verb Learning
What does it take for a word learner to infer the meaning of a verb? What do they
need to know to extend a verb to other instances correctly? The principle of verb
extension we dealt with in this chapter—that verbs get extended to a new situation
with a different agent or a different object, as long as the action is the “same”—is a
very basic extension principle that holds for almost all verbs, and it is an important
first step toward adult-like representation of verb meanings. However, word learners
also need to know that different types of verbs employ different criteria for exten-
sion. Some verbs should be extended on the basis of sameness of manner, while
others should be extended solely on the sameness of results, yet others should be
extended on the basis of the sameness of direction or trajectory of motion. Further-
more, here, the notion of “being the same” does not mean “being identical,” as, for
example, there are many ways of walking, turning, throwing, hopping, climbing, as-
cending, and so on. Children thus first need to know which semantic aspect (such as
manner, path/direction/trajectory, and result) they should attend to in extending the
particular verb they are learning, and they then need to extract what constitutes the
“sameness” for the given semantic aspect in the verb in question (see Maguire et al.,
2002, for a similar point). This process of extracting sameness—or discovering the
“invariant component” in other words—and using it as the basis for extension is ex-
actly like analogical mapping based on abstract higher-order relations.
Seen this way, the way children learn nouns and verbs may be very different,
and for a good reason. Children as young as 2 years of age are willing to fast-map
the meaning of a noun after only observing it to get associated with a single refer-
ent (e.g., Golinkoff et al., 1992; Haryu & Imai, 2002; Imai & Gentner, 1997; Imai
et al., 1994; Imai & Haryu, 2001; Landau et al., 1988; Markman, 1989). This may
be in large part because the structure of the noun lexicon is hierarchical and coher-
ent, and also because the meaning of a noun is largely predictable from the percep-
tual nature of the named entity due to a high correlation between conceptual and
perceptual properties in concrete entities. For example, the (concrete) noun lexicon
470 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 470
is divided into two distinct ontological classes, the class of objects and the class
of substances, and determination of the class membership of a given entity (i.e.,
whether this thing is an object or a substance) is largely, if not completely, sup-
ported by the entity’s perceptual nature, such as its solidity, boundedness, and
shape complexity (Imai & Gentner, 1997; Imai & Mazuka, 2003; Smith, Columga,
& Yoshida, 2003). Once the named entity’s ontological class membership is deter-
mined, the extension principles are rather simple (or at least much simpler than
those for verbs). When the word is determined to be an object name, it should be
extended by shape; when it is determined to be a substance name, it should be ex-
tended by material (Soja, Carey, & Spelke, 1991). Given this coherent and cleanly
structured noun lexicon, it is relatively easy for children to come up with assump-
tions about the noun lexicon (e.g., what kinds of nouns are there in the noun lexi-
con, what kinds of noun correspond to what kind of conceptual classes, how
different kinds of nouns are syntactically marked, what perceptual and conceptual
factors are most useful in inferring word meanings, etc.), and with these assump-
tions, it is likely that children can successfully fast-map a new noun to its meaning
with only a single instance (Imai & Haryu, 2004).
The verb lexicon is not as cleanly structured as the noun lexicon, and the
cross-linguistic variability is much greater than the noun lexicon (Gentner, 1982).
Even though there seems to be only a small set of semantic components that are
universally lexicalized in motion/action events (e.g.,such as figure, ground,manner,
and path), the component that is most likely to be packaged into verb meanings
varies across different languages (Talmy, 1985). Furthermore, within each language,
the lexicalization pattern is only probabilistic. For example, even though English is
a manner-dominant language, there are quite a few verbs that lexicalize path of the
motion (e.g., enter,ascend,descend). Syntactic argument structures provide clues
for the inference of verb meanings at a global level (e.g., whether the action in-
volves change of state of the patient object, whether it is spontaneous and self-
moved, or whether it only happens by some external force, etc.). However, as there
are so many different classes of verbs that take different argument alternation pat-
terns in different semantic domains (Levin, 1993), inferring the specific meaning of
a verb from a single instance with one syntactic argument structure would be diffi-
cult and could even be harmful (Naigles, 1996).
Seen in this light, the different strategies children take in learning nouns and
verbs may be quite reasonable and optimal. In learning nouns, they generate as-
sumptions about noun extension at very early stages of lexical development, and
by applying these assumptions, they fast-map a newly encountered noun to its
meaning with a single instance. In learning verbs, in contrast, they do not seem to
easily generate such assumptions about verb extension. Children extract abstract,
widely applicable rules for verb meaning extension quite slowly, only after learn-
ing different verbs in an “island-like” fashion for a substantial period of time (com-
pare Tomasello, 2000). Considering the structure of the verb lexicon, this is
probably a more beneficial strategy than fast-mapping.
REVISITING THE NOUN-VERB DEBATE 471
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 471
472 HOW LANGUAGE INFLUENCES VERB LEARNING
Concluding Remarks: Interaction Between Universal and
Language-Specific Factors in Early Word Learning
One of the important questions in the literature of lexical development is
whether a particular word class is easier for children to learn over others, and what
factors determine the relative ease or difficulty of word learning. The results from
the cross-linguistic novel noun/verb learning study reported in this chapter sup-
port the view that noun learning is universally advantaged over than verb learning,
as children of three very different languages performed better in learning novel
nouns than in learning novel verbs. At the same time, the cross-linguistic pattern
of novel verb learning points to the influence of linguistic properties on the ease of
novel verb learning as well as on cues children utilize in inferring verb meanings.
Interestingly, however, it was not the distributional property of nouns vs. verbs,
but the structural properties of the language that had a greater influence here.
Contrary to the general assumption in the literature (Gentner, 1982; Tardif, 1996),
morphological simplicity in the Chinese language may magnify the difficulty chil-
dren experience in learning a new verb (see Erbaugh, 1983). Note, however, that
the lack of the form class information within the subclasses of nouns does not
seem to make novel noun learning difficult for Japanese-speaking children (see
note 3; see also Imai & Gentner, 1997; Imai & Haryu, 2001, 2004). This suggest
that availability of cues from syntax for the inference of word meanings interacts
with the nature of the concepts for a given word class.
Early word learning takes place within a dynamic interaction among chil-
dren’s universal cognitive disposition, distributional and structural properties of
the language they are learning, and nature of concepts (e.g., the degree of abstract-
ness, complexity of meaning, perceptual accessibility, etc.) denoted by words. In
this interaction, the relative dominance among these factors seems to be hierarchi-
cally ordered. Based on the pattern of results in the word learning literature, it is
probable that conceptual factors take precedence over linguistic factors in the hi-
erarchy. As we reviewed throughout the chapter, it has been repeatedly observed
that, across different languages, children learn labels of objects more readily and
easily than labels of actions, and that they generalize nouns more willingly than
verbs. Linguistic factors, either structural or distributional, do affect word learning
but not to the degree that they can override conceptual constraints.5
Taken together, what is important for future research is not so much the ques-
tion of which of the two factors—universal conceptual constraints and language
specific properties—is more important, but the question of how the two factors
interact with one another.
Acknowledgments This research is supported by the Ministry of Education Grant #
and the Keio University Mori Grant to Mutsumi Imai. We thank Kathy Hirsh-
Pasek, Roberta Golinkoff, Hua Shu, and Victoria Muehleisen for insightful com-
ments on earlier versions of the chapter.
Q14
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 472
Notes
1. Of course, verbs put some constraints on the types of arguments they can be
used with. A piece of fabric cannot be smashed, so the verb smash cannot take fabric as
an argument. However, within the range of the semantic constraints, different objects
can occur as the argument of the verb.
2. However, we did not set up the “zai +verb” alone (without the arguments) pat-
tern, since native speakers of Chinese judged this pattern unnatural.
3. For example, when a novel individuated object (either animate or inanimate)
was labeled with a novel noun, Japanese-speaking 2-year-old children spontaneously
generalized the noun to other objects that were similar to the original in shape (but
not in other perceptual dimensions) assuming it to be a basic-level object category
name. At the same time, they could relax this default assumption quite readily. When
anovel noun was associated with a substance, they generalized it on the basis of mate-
rial identity, ignoring the sameness in shape (Imai & Gentner, 1997). When a familiar
animal was named, they interpreted it to be a proper name of the named animal (Imai
& Haryu, 2001).When a named object was inanimate and was a typical member of the
familiar category, they mapped the new noun to a category subordinate to a old, famil-
iar one, but when the inanimate object was an atypical member of the familiar cate-
gory, they mapped the new label to a new basic-level category, restructuring the
boundary of the old, familiar category by excluding the named object from it (Haryu &
Imai, 2002).
4. This difference in the results could be attributed to the difference in the stim-
uli. While Golinkoff et al. (1996) used still pictutes, Kersten and Smith (2002) used
animated events.
5. Interestingly, an analogous pattern has been witnessed in the learning of object
names and substance names. Imai and Gentner (1997; see also Imai & Mazuka, 2003)
examined how Japanese- and English-speaking children project word meanings when a
novel noun was presented in association with a range of entities with different levels
of individuation (i.e., highly individuated complex-shaped objects, simple-shaped solid
substances that can be construed either as individuated objects or unindividuated solid
substance, and clearly unindividuated substances). Imai and Gentner showed that the
ontological distinction between objects and substances is universally available at the be-
ginning stage of word learning, and this conceptual understanding constrains noun
learning whether or not this ontological distinction is marked by the grammar of the
language children are learning. At the same time, the way in which language divides
the individuation continuum affected the pattern of word meaning projection for sim-
ple shaped solid substances, whose status of individuation was perceptually ambiguous.
References
Au,T.K.-F., Dapretto, M., & Song, Y. K. (1994). Input vs. constraints: Early word ac-
quisition in Korean and English. Journal of Memory and Language,33,567–582.
Bowerman, M. (1980). The structure and origin of semantic categories in the language
learning child. In M. L. Foster & S. H. Brandes (Eds.), Symbol as a sense: New ap-
proaches to the analysis of meaning.New York: Academic Press.
REVISITING THE NOUN-VERB DEBATE 473
Q15
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 473
Casasola, M., & Cohen, L. B. (2000). Infants’ association of linguistic labels with causal
actions. Developmental Psychology,36,155–168.
Childers, J., & Tomasello, M. (2002). Two-year-olds learn novel nouns, verbs, and con-
ventional actions from massed or distributed exposures. Developmental Psychology,
37,698–705.
Choi, S. (2000). Caregiver input in English and Korean: Use of nouns and verbs in
book-reading and toy-play contexts. Journal of Child Language,27,69–96.
Choi, S., & Bowerman, M. (1991). Learning to express motion events in English and
Korean: The influence of language-specific lexicalization patterns. Cognition,42,
83–121.
Choi, S., & Gopnik, A. (1995). Early acquisition of verbs in Korean: A cross-linguistic
study. Journal of Child Language,22,497–529.
Dromi, E. (1987). Early lexical development. Cambridge: Cambridge University Press.
Erbaugh, M. (1983). Acquisition of Mandarin syntax: “Less” grammar isn’t easier. Chi-
nese Language Teatchers Association Journal,22,51–63.
Erbaugh, M. (1992). The acquisition of Mandarin. In D. S. Slobin (Ed.), The crosslinguis-
tic study of language acquisition,Vol. 3.Hillsdale, NJ: Lawrence Erlbaum.
Fisher, C. (1996). Structural limits on verb meaning: The role of analogy in children’s
interpretations of sentences. Cognitive Psychology,31,41–81.
Forbes, J., & Poulin-Dubois, D. (1997). Representational change in young children’s
understanding of familiar verb meaning. Journal of Child Language,24,389–406.
Gentner, D. (1982). Why nouns are learned before verbs: Linguistic relativity versus
natural partitioning. In S. A. Kuczaj (Ed.), Language development: Vol. 2. Language,
thought, and culture (pp. 301–334). Hillsdale, NJ: Erlbaum.
Gentner, D., & Boroditsky, L. (2001). In individuation, relativity, and early word learn-
ing. In M. Bowerman & S. Levinson (Eds.), Language acuiqisition and conceptual
development (pp. 257–283). Cambridge: Cambridge University Press.
Gillette, J., Gleitman, H., Gleitman, L., & Lederer, A. (1999). Human simulations of
vocabulary learning. Cognition,73,135–176.
Golinkoff,R., Hirsh-Pasek, K., Bailey, L. M., & Wenger, N.R. (1992).Young children and
adults use lexical principles to learn new nouns. Child Development,28,99–108.
Golinkoff, R., Hirsh-Pasek, K., & Schweisguth, M. (2001). A reappraisal of young chil-
dren’s knowledge of grammatical morphemes. In J. Weissenborn & B. Höhle
(Eds.), Approaches to bootstrapping: Phonological, lexical, syntactic and neurophysio-
logical aspects of early language acquisition.Amsterdam: John Benjamins.
Golinkoff, R., Jacquet, R. C., Hirsh-Pasek, K., & Nandakumar, R. (1996). Lexical prin-
ciples may underlie the learning of verbs. Child Development,67,3101–3119.
Haryu, E., & Imai, M. (2002). Reorganizing the lexicon by learning a new word: Japan-
ese children’s inference of the meaning of a new word for a familiar artifact. Child
Development, 73,1378–1391.
Haryu, E., Imai, M., Okada, H., Li, L., Meyer, M., Hirsh-Pasek, K., & Golinkoff, R.
(2004, November). Noun bias in Chinese children: Novel noun and verb learning in
Chinese, Japanese and English preschoolers.Paper presented at the 29th Annual
Boston University Conference on Language Development, Boston, MA.
Huttenlocher, J., Smiley, P., & Charney, R. (1983). Emergence of action categories in
the child: Evidence from verb meanings. Psychological Review,90,72–93.
Imai, M., & Gentner, D. (1997). A crosslinguistic study on constraints on early word
meaning: Linguistic influence vs. universal ontology. Cognition,62,169–200.
Imai, M., Gentner, D., & Uchida, N. (1994). Children’s theories of word meanings: The
role of shape similarity in early acquisition. Cognitive Development,9,45–75.
474 HOW LANGUAGE INFLUENCES VERB LEARNING
Q16
Q17
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 474
Imai, M., & Haryu, E. (2001). Learning proper nouns and common nouns without
clues from syntax. Child Development,72,787–802.
Imai, M., & Haryu, E. (2004). The nature of word learning biases and their roles for
lexical development: From a cross-linguistic perspective. In D. G. Hall & S. R.
Waxman (Eds.), Weaving a lexicon (pp. 411–444). Cambridge, MA: MIT Press.
Imai, M., Haryu, E., & Okada, H. (in press). Mapping novel nouns and verbs onto dy-
namic action events: Are verb meanings easier to learn than noun meanings for
Japanese children? Child Development.
Imai, M., & Mazuka, R. (2003). Re-evaluation of linguistic relativity: Language-specific
categories and the role of universal ontological knowledge in the construal of indi-
viduation. In D. Gentner & S. Goldin-Meadow (Eds.), Language in mind:Advances
in the issues of language and thought (pp. 430–464). Cambridge, MA: MIT Press.
Kersten, A. W., & Smith, L. (2002). Attention to novel objects during verb learning.
Child Development,73,93–109.
Kim, M., McGregor, K. K., & Thompson, C. K. (2000). Early lexical development in
English- and Korean-speaking children: Language-general and language-specific
patterns. Journal of Child Language,27,225–254.
Landau, B., Smith, L. B., & Jones, S. S. (1988).The importance of shape in early lexical
learning. Cognitive Development,3,299–321.
Levin, B. (1993). English verb classes and alternations.Chicago: University of Chicago
Press.
Li, P., Bates, E., & MacWinney, B. (1993). Processing a language without inflections: A
reaction time study of sentence interpretation in Chinese. Journal of Memory and
Language,32,169–192.
Maguire, M., Hennon, E., Hirsh-Pasek, K., Golinkoff, R., Slutzky, C., & Sootsman, J.
(2002). Mapping words to actions and events: How do 18-month-olds learn a
verb? Proceedings of the 26th Annual Boston University Conference on Language De-
velopment.
Markman, E. M. (1989). Categorization in children: Problems of induction.Cambridge,
MA: MIT Press.
Meyer, M., Leonard, S., Hirsh-Pasek, K., Imai, M., Haryu, E., Pulverman, R., & Addy,
D. (2003, November). Making a convincing argument: A crosslinguistic comparison
of noun and verb learning in Japanese and English.Paper presented at the 28th
Annual Boston Conference on Language Development, Boston University,
Boston, MA.
Naigles, L. (1996). The use of multiple frames in verb learning via syntactic bootstrap-
ping. Cognition,58,221–251.
Ogura, T. (2001). Meishi yuui, doushi yuui ni oyobosu hahaoya no gengo nyuryoku no
kentou.(Influence of maternal input in the distribution of nouns and verbs in early
vocabulary.) Ministery of Education and Science grant report. Kobe, Japan: Kobe
University.
Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.), Cogni-
tion and categorization.Hillsdale, NJ: Erlbaum.
Smith, L., Columga, E., & Yoshida, H.(2003). Making an ontology: Crosslinguistic evi-
dence. In D. H. Rakison & L. M. Oakes (Eds.), Early category and concept develop-
ment: Making sense of the blooming, buzzing conflusion (pp. 275–302). New York:
Oxford University Press.
Soja, N., Carey, S., & Spelke, E. (1991). Ontological categories guide young children’s
inductions of word meaning: Object terms and substance terms. Cognition,38,
179–211.
REVISITING THE NOUN-VERB DEBATE 475
Q18
Q19
Q20
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 475
Talmy, L. (1985). Lexicalization patterns: Semantic structure in lexical forms. In T.
Shopen (Ed.), Language typology and syntactic description (Vol. 3, pp. 57–149).
New York: Cambridge University Press.
Tardif, T. (1996) Nouns are not always learned before verbs: Evidence from Mandarin
speakers’ early vocabulary. Developmental Psychology,32,492–504.
Tardif, T., Gelman, S., & Xu, F. (1999). Putting the “noun bias” in context: A compari-
son of English and Mandarin. Child Development,70, 620–635.
Tomasello, M. (2000). The item based nature of children’s early syntactic develop-
ment. Trends in Cognitive Sciences,4,156–163.
Werker, J. F., Cohen, L. B., Lloyd, V. L., Casasola, M., & Stager, C. L. (1998). Acquisi-
tion of word-object associations by 14-month-old infants. Developmental Psychol-
ogy,34,1289–1309.
Yamashita, Y. (1997). The effect of saliency in input on the acquisition of nouns and
verbs: Innate constraints or language-specific input? Naruto Eigo Kyoiku (Naruto
College of Education Research Papers),11.
476 HOW LANGUAGE INFLUENCES VERB LEARNING
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 476
Chapter 17: Author Queries
[Q1] Add page no. of quote.
[Q2] au: update?
[Q3] au: please check spelling; reference has Meyer.
[Q4] au: update?
[Q5] au: update?
[Q6] au: update?
[Q7] au: update?
[Q8] au: check spelling.
[Q9] Should this be table 17.3? There is no mention of 17.3 in the text.
[Q10] au: update.
[Q11] au: please add reference or delete citation.
[Q12] au: update?
[Q13] au: check spelling.
[Q14] au: please give grant number.
[Q15] Pages?
[Q16] Pages?
[Q17] Pages?
[Q18] au: update?
[Q19] Add editors, pages, publisher & location.
[Q20] Pages?
17Hirsh-Pasek_ch17 450-476.qxd 31:10:2005 21:41 PM Page 477
