Content uploaded by Anne Christophe
Author content
All content in this area was uploaded by Anne Christophe
Content may be subject to copyright.
LANGUAGE LEARNING AND DEVELOPMENT, 3(4), 325–341
Copyright © 2007, Lawrence Erlbaum Associates, Inc.
frHLLD1547-54411547-3341Language Learning and Development, Vol. 3, No. 4, August 2007: pp. 1–33Language Learning and Development
Syntax Constrains the Acquisition
of Verb Meaning
Early verb acquisitionBERNAL ET AL.
Savita Bernal
Laboratoire de Sciences Cognitives et Psycholinguistique
EHESS / CNRS / DEC-ENS, Paris, France
Jeffrey Lidz
Cognitive Neuroscience of Language Laboratory, Department of Linguistics
University of Maryland
Séverine Millotte
Laboratoire de Psycholinguistique Expérimentale, Geneva, Switzerland
and Laboratoire de Sciences Cognitives et Psycholinguistique
EHESS / CNRS / DEC-ENS, Paris, France
Anne Christophe
Laboratoire de Sciences Cognitives et Psycholinguistique
EHESS / CNRS / DEC-ENS, Paris, France
and Maternité Port-Royal, AP-HP, Faculté de Médecine
Université Paris-Descartes
Can infants use the syntactic context of an unknown word to infer that it is a verb, and
thus refers to an action? Twenty-three-month-old French infants watching a moving
object were taught novel verbs, within sentences that contained only function words
(e.g. “il poune par là” / “it’s pooning there”). Infants then watched two instances of the
object undergoing either the familiar or a novel action and were asked to point towards
the screen matching the novel verb. Infants correctly pointed more often towards the
familiar action. To check that they did not simply perseverate in pointing at the familiar
scene, control infants were taught novel nouns on the same visual stimuli (e.g. “un
The authors wish to thank three anonymous reviewers for their helpful comments.
Correspondence should be addressed to Savita Bernal, LSCP, ENS, 29, rue d’Ulm, 75005 Paris,
France. E-mail: savita.bernal@ens.fr
326 BERNAL ET AL.
poune est là”/ “a poon is here”). Contrary to verb-learning infants, noun-learning infants
pointed more often to the novel action. These results confirm the hypothesis that func-
tion words, and more generally syntactic structure, support early lexical acquisition.
Learning novel words is not an easy task. Assigning a meaning to a word, even in
the simplest situation (someone saying “look, that’s a cat!” while pointing at a
furry four-legged animal) is not straightforward. How is a child to know that this
sequence of sounds refers to an animal and not, for instance, to its color or its
movement (Quine, 1960)? Different hypotheses have been formulated as to how
a child may restrict the set of hypotheses she entertains about the meaning of a
novel word (for reviews, see Waxman & Lidz, 2006; Woodward & Markman,
1998). Among these, the Syntactic Bootstrapping hypothesis postulates that there
exists a principled relationship between the meaning of a word and the syntactic
structure in which it occurs and that these principles guide children’s hypotheses
about novel word meaning (Gleitman, 1990; Landau & Gleitman, 1985). On this
view, some basic knowledge about syntax may help children identify the mean-
ings of novel words. This hypothesis has received support from adult experi-
ments (Gillette, Gleitman, Gleitman, & Lederer, 1999), as well as from toddlers
(Fisher, Hall, Rakowitz, & Gleitman, 1994; Lidz, Gleitman, & Gleitman, 2003)
and infants (Fisher, 2002; Naigles & Kako, 1993; Waxman & Booth, 2001). In
all of these studies, the number and type of syntactic arguments provide informa-
tion that is used by children to identify the meaning of a novel verb (except for
Waxman & Booth, 2001, which deals with the nouns/adjective distinction).
While the existing literature indicates that children can use syntactic informa-
tion to choose one event from several as the referent of a novel verb, the ability to
do so depends on the child’s having identified the verb category as one that poten-
tially refers to events. A precursor to being able to use the number and type of
arguments to determine which event a novel verb labels is to know that verbs pro-
totypically refer to events. In the present study, we ask whether infants know that
the syntactic category of a word is a cue to its meaning. Nouns generally refer to
object categories, while verbs typically refer to event categories and adjectives to
properties of objects (Brown, 1957; Grimshaw, 1981; Macnamara, 1982; Maratsos &
Chalkley, 1980; Pinker, 1987). When do children know these patterns and what
information do they use to identify the syntactic category of a word?
Syntactic categories are, by definition, distributional categories. What makes a
set of words a category is that they occur in the same linguistic environments.
One of the most reliable distributional properties of a syntactic category is the set
of function words co-occurring with it. In French, for example, nouns generally
co-occur with articles while verbs co-occur with pronouns and auxiliaries.
Infants have been shown to be sensitive to function words very early on (Shi,
Werker, & Morgan, 1999), to recognize the function words of their language
EARLY VERB ACQUISITION 327
towards the end of their first year (Shafer, Shucard, Shucard, & Gerken, 1998;
Shi, Werker, & Cutler, 2006) and to use them to facilitate word segmentation
(Shi & Gauthier, in press), sentence processing (Gerken, 1994; Gerken, Landau,
& Remez, 1990; Kedar, Casasola, & Lust, 2006) and noun or verb categorization
(Höhle, Weissenborn, Kiefer, Schulz, & Schmitz, 2004; Mintz, 2006). Recent
corpus analyses have also supported the feasibility of categorizing content words
using distributional co-occurrences between function and content words
(Cartwright & Brent, 1997; Chemla, Mintz, Bernal, & Christophe, in press;
Mintz, 2003; Mintz, Newport, & Bever, 2002).
Building on early work by Roger Brown (1957), showing that preschoolers
can use the morphological markers surrounding a novel word as a cue to its
meaning, we asked whether 23-month-old infants would attribute a different
meaning to a novel word depending on the syntactic context in which it occurred.
We chose 23-month-olds for two reasons. First, while children start to produce
words in the vicinity of their first birthday, it is not until 24 to 30 months that
infants begin to produce a sizeable number of verbs, and to use them systemati-
cally to refer to actions (e.g., eat, run), mental states (e.g., want, see) and relations
(e.g., touch, move) (Bates et al., 1994; Caselli et al., 1995; Fenson et al., 1994;
Gentner, 1981; Nelson, 1973).
Second, in many experimental verb-learning tasks, learners show surprising
difficulties, many of which persist as late as three to five years of age, despite the
fact that by 24 months they have begun to use verbs spontaneously and appropri-
ately (Abbot-Smith, Lieven, & Tomasello, 2004; Behrend, 1995; Childers &
Tomasello, 1996; Forbes & Poulin-Dubois, 1997; Hirsch-Pasek, Golinkoff,
Maguire, & Imai, 2005; Imai, Haryu, & Okada, 2005; Kersten & Smith, 2002;
Rice, 1980; Theakston, Lieven, Pine, & Rowland, 2001). This difficulty suggests
that targeting children at the earliest stages of verb production could be highly
informative about later difficulties in experimental tasks.
We defined the syntactic context of our novel words by varying the functional
elements co-occurring with the word. In particular, we asked whether a novel
word presented in a verb context would be taken to refer to an action. Twenty-
three-month-old French infants were presented with video sequences featuring an
object undergoing a simple action. Simultaneously, they heard sentences contain-
ing a novel word together with function words (as well as some attention-getters
such as ‘regarde!’/ ‘look!’ that did not refer to any particular object or action in
the visual display). There were only two potential referents for this novel word in
the visual scene: the object and its action. If infants can use the syntactic context
to identify the grammatical category of a novel word and thus infer its meaning,
then they should think that it refers to the action when presented as a verb (ex:
“Regarde, il poune” i.e. “Look, it(‘s) blick(ing)”) but to the object when pre-
sented as a noun (ex: “Regarde le poune” i.e. “Look at the blick”). To test
this, we used a paradigm derived from the well-known inter-modal preferential
328 BERNAL ET AL.
looking paradigm (IPL); in this paradigm, we replaced the looking response with
a pointing one.
MATERIAL AND METHOD
Subjects
32 infants from monolingual French-speaking homes participated in this experi-
ment (18 girls, mean age 23;02 months, range 22;18 to 23;15), 16 in each of two
groups (Experimental and Control). They had no known neurological or hearing
deficits. The project was accepted by the Cochin Hospital Commission for
Human Participants. Parents gave their informed consent before the study and
completed a French version of the MacArthur CDI test (Kern, 2007). There was
no difference between the experimental and control groups in vocabulary scores
(Experimental group, 223.6, Control Group, 225, t(31) = 1.13, p = 0.27). The
results from 31 additional subjects were not analyzed for the following reasons:
not pointing during warm-up (9, generally unwilling to interact with the experi-
menter), not pointing during training (12, generally losing interest in the setup),
crying/fussing/not completing the experiment (6), mother speaking (1), technical
problem (2), experimenter error (1).
Set up
Infants were seated in a high-chair, facing a large screen (1.5m by 2m), situated
about 2m away, on which the stimuli were projected at the infant’s eye-level.
When two scenes were projected simultaneously on the screen (one on the left
and one on the right), the angle between the center of both stimuli, from the
infant’s viewpoint, was about 30°. Infants were videotaped from a camcorder sit-
uated right in front of them, just above the top of the screen. Parents were seated
behind the infants and were instructed not to talk to or distract infants. They wore
headphones playing masking noise. A table (1m by 2m) was set between the
infant and the screen. On the right of that table (from the infant’s viewpoint), a
chair was placed, that allowed an experimenter to sit with her back to the screen,
facing the infant if she turned her head slightly to her right. This experimenter
manipulated the puppet that asked the test questions. She did not wear masking
headphones (since she had to speak in the experiment, she had to know when to
take her turn), however, she was unable to see the screen and could therefore not
influence the infant’s reponses. A second experimenter, in another part of the
room delimited by a curtain, out of sight of the infant, launched the various mov-
ies at the appropriate times. She could watch the infant on a monitor and check
that the infant’s pointing responses were visible on the video recording.
EARLY VERB ACQUISITION 329
Procedure
Before the experimental session itself, during which infants were actually taught
new words and tested on their comprehension, they went through two prelimi-
nary sessions to familiarize them with the set-up and the pointing task: warm-up
and training. The warm-up session was aimed at eliciting pointing towards plas-
tic animal toys (a dog, a cat, a hen and a giraffe, words known to infants of this
age). During that phase only, the experimenter manipulating the puppet knelt
behind the table, facing the infant (with the screen at her back), so that her face
and shoulders were visible. She named all 4 animals, then placed 2 cardboard
boxes on the table, their opening facing the infant, so that he could look inside
but the experimenter (and the puppet) could not. Speaking in puppet-voice, the
experimenter announced that the puppet loved to play hide-and-seek, and was
going to play with its friends the animals. The puppet then made a show of leav-
ing the scene (by going under the table), so that it would not see where the ani-
mals were hiding. Meanwhile, the experimenter, speaking in her normal voice (in
a conspiratory tone), chose two of the animal toys, named them, and put them in
each of the two boxes, speaking all the time (assuming she spoke to an infant
named Tom: “Look, Tom, this is the dog! The dog is going in that box! Do you
see the dog here, Tom?” “Oh, and this is the cat!” etc.; she used the infant’s name
a lot, to keep him engaged in the interaction). The experimenter then made the
puppet come back and ask for one of the animals, speaking in puppet-voice again
(“Hey, Tom, here I am again! I am looking for my friend the dog. Can you help
me find the dog, Tom? I can’t see him! Where is he? Can you show me where the
dog is hiding, please, Tom?” etc.). The experimenter controlling the puppet went
on asking questions until she managed to elicit a pointing response from the
infant. If the infant did not point, the experimenter enrolled the parent to help,
saying “Maybe Mummy knows where the dog is? Could you show me the dog,
please, Mummy?”, so that the parent would demonstrate what was expected from
the infant. If the infant pointed towards the right animal, the puppet would thank
him profusely, give him the toy as a reward and then go on to ask for the other
animal. If the infant could point correctly to both objects in three instances of the
game (not necessarily consecutive), he was allowed to continue with the training
session. If not, he was excluded from this study. Throughout the warm-up phase,
no effort was spared to make the infant point, and that phase lasted as long as the
infant was not bored or restless.
The training session was aimed both at familiarizing the infant with the game
on the screen and at teaching him to point towards an action. From that point on,
the experimenter manipulating the puppet sat in her chair on the right of the
infant, with her back to the screen. She had to rotate her torso and twist her neck
to be able to glimpse at the screen and ask the right questions to the infant. Three
short sequences displaying a woman performing familiar actions (eating, drinking
330 BERNAL ET AL.
and giving) were shown to the infant while the puppet labelled the actions in a
lively voice (cf. Table 1, “Look, that lady is drinking! Can you see? She’s drink-
ing! Oh! All gone!). Then the pointing game resumed. Two pictures of the same
toys used during warm-up were shown simultaneously on either side of the
screen. The puppet asked the infant to show it one of the animals (“Can you show
me the dog, please, Tom? Where is the dog?”). If the infant pointed correctly on
three instances, the pictures were replaced by two sequences of the familiar
actions (e.g. a woman drinking on the left side and eating on the right side). The
puppet asked the infant to point towards one action (“Can you show me the lady
who is drinking? Show me the one who is drinking, please, Tom!”). When the
infant had pointed correctly towards three actions, the experiment per se began.
During the experimental session, infants were taught novel words while look-
ing at a video clip showing an object performing an intransitive action (e.g., a
flower rotating). This session consisted of 4 trials, during which 4 novel words
were presented together with 4 different object-action pairs (see ‘materials’ for
details). Each trial included three phases: familiarization, contrast and test (see
Table 2, design similar to Booth & Waxman, 2003). During the familiarization
phase, infants were presented with three instances of the same object (e.g. a
flower) undergoing the same motion (e.g. rotating). Meanwhile, they were taught
a new verb, using short sentences that contained only that verb, some function
words, and attention-getters (e.g. ‘Look! it’s pooning there’; see Table 2, left-
hand column, for the full set of sentences in French). These sentences were
uttered by a pre-recorded voice coming from loudspeakers. During the contrast
phase, a different object was presented doing a different action (e.g. a frog slid-
ing), produced with a negative sentence (‘uh-oh, this one is not pooning!’) and
followed by the first object doing the familiar action, accompanied by an affirma-
tive sentence (‘Yeah! This one is pooning!’). In the test phase, infants were
TABLE 1
Script for the Training Session
For the verb warm-up Regarde, la dame boit! Tu as vu? Elle boit! Hmm c’est bon…Oh! Elle
est partie!
Look, the lady is drinking! Can you see? She’s drinking. It looks
good…Oh! all gone!”
For the noun training Oh! Tu peux me montrer le chien, s’il te plaît, Tom? Lequel est un
chien? Il est où, le chien? Tu vois le chien? Montre-le moi! etc.
Oh, can you show me the dog, please, Tom? Which one is a dog? Where
is the dog? Can you see the dog ? Show it to me!
For the verb training Oh! Tu peux me montrer la dame qui boit, s’il te plaît Tom? C’est
laquelle, qui boit? Elle est où, celle qui boit? Montre-la moi! etc.
Oh! Can you show me the lady who’s drinking, please, Tom? Which one
is drinking? Where is the one who’s drinking? Show her to me!
EARLY VERB ACQUISITION 331
presented with two simultaneous pictures of the now-familiar object (the flower).
In one, the motion matched that of the familiarization phase (rotating); in the
other, the motion was novel (e.g. moving up and down). At that point, the exper-
imenter holding the puppet started to ask questions on-line; she asked the infant
to point towards the object/action pair matching the newly-learned verb.
The test phase was 20s long and the experimenter tried to elicit pointing by
asking questions throughout this period of time. She always started with the same
TABLE 2
Example of Trials for Each Group of Subjects
Session Video Experimental Group Control Group
FAMILIARIZATION
(off voice, 20s)
Flower
rotating (left)
Regarde, elle poune!
Tu as vu ? Elle poune!
Look! It’s pooning!
Do you see? It’s pooning!
Regarde, c’est une poune
Tu as vu ? Une poune!
Look! This is a poon!
Do you see? A poon!
Flower
rotating (right)
Regarde maintenant!
Elle poune par là!
Tu vois comment elle poune?
Now look! It’s pooning there!
Do you see how it’s pooning?
Regarde maintenant!
La poune est là !
Tu vois ce que fait la poune?
Now look! The poon is there!
Do you see what the poon is
doing?
Flower
rotating (left)
Oh, elle poune encore là!
Tu la vois qui poune?
Oh! It’s pooning there again!
Can you see it pooning?
Oh, la poune est encore là!
Tu la vois cette poune?
Oh! The poon is there again!
Can you see this poon?
CO
NTR
AS
T
(off voice, 10s)
Frog
sliding (center)
Oh oh, celle-là ne poune pas!
Ceci ne poune pas.
Uh-Oh! That one is not
pooning.
This is not pooning.
Oh oh, celle-là n’est pas une
poune!
Ceci n’est pas une poune.
Uh-Oh! That one is not a poon.
This is not a poon.
Flower
rotating (center)
Oui! Celle-là poune!
Tu as vu ça! Elle poune!
Yeah! This one is pooning!
Did you see that? It is pooning!
Oui! Celle-là est une poune!
Tu as vu ça! Une poune!
Yeah! This one is a poon!
Did you see that? A poon!
EXPERIMENT
A
L
(puppet voice, 20s)
Flower
rotating (right)
Vs.
Flower jumping
(left)
Laquelle poune ?
Montre-moi celle qui poune!
Elle est où, celle qui poune?
(etc..)
Which one is pooning?
Show me the one that’s pooning!
It is where, the one that’s pooning?
Laquelle est une poune?
Montre-moi la poune!
Elle est où, la poune?
(etc…)
Which one is a poon?
Show me the poon!
It is where, the poon?
Note. Sentences Used for Both Groups were Closely Matched for Length and Position of the Tar-
get Word. In French, the Word is Pronounced Exactly the Same Way as When it is a Noun and When
it is a Verb (3rd Person Singular Present Tense). A Second Contrast Phase and a Second Test Phase
Followed (Identical Except for the Left/Right Ordering of the Test Phase, that Varied Quasi-
Randomly).
332 BERNAL ET AL.
two questions (‘Which one is pooning? Show me the one that’s pooning!), then
went on asking variants of these questions, interspersed with ‘Please, Tom’, and
occasionally ‘My little eyes cannot see very well!’. She stopped asking as soon as
the infant pointed, or at the end of the trial (she knew the trial ended when she
heard a baby laugh that separated one trial from the next). All pointing responses
were reinforced by thanking the infant, whatever the direction of pointing was
(‘Thank you, Tom. Thank you very much!’). The rest of the trial was fast-
forwarded to avoid a long ‘blank’ interval for infants who pointed fast. Mean-
while, the puppet filled the silence with praise (‘You’re such a great help to me,
Tom! You show very well! Thank you so much!’). Table 2 shows the complete
script for one trial, in French, together with the English glosses (left-hand col-
umn). The same sentences were used in all trials (changing only the relevant
verb). The contrast and test phases were presented twice (overall: familiarization,
contrast 1, test 1, contrast 2, test 2), so that the infants had 8 opportunities to
respond to questions by pointing.
If infants correctly map the novel verb to the action, they should point more
towards the familiar action (flower rotating), than towards the novel one (flower
moving up and down). However, infants might also point towards the familiar
action for either one of the following wrong reasons: (1) because they mapped
the new word to the whole object/action pair; (2) because they have a tendency to
perseverate and point towards the most familiar thing in their environment; or (3)
because they mapped the new word to the precise object seen during familiariza-
tion (like a proper name), and, being faced with 2 identical objects during the test
phase, they assumed that the object continued to perform the same action. To test
for these alternative interpretations, a control group of infants was run: these
infants watched the exact same sequence of events, but were taught a novel noun
instead of a novel verb. During familiarization, they heard ‘the poon is there!’;
during contrast, ‘uh-oh, this one is not a poon! Yeah, this one is a poon’, and dur-
ing test, they were asked on-line ‘Which one is a poon? Show me the poon!’ (see
Table 2, right-hand column, for the complete set of sentences used in the control
condition). Since the screen showed two instances of the same object performing
two different actions, this question is pragmatically inappropriate for adults; the
new noun, ‘poon’, matches both objects. If infants, like adults, interpret the new
word as a label for the category of objects, they may either point to both objects,
point at chance, or point more often to the novel action (just because it is more
interesting)
1
. What they clearly should not do, is point more towards the familiar
action. In contrast, if they use any of the strategies outlined above, they should
1
The latter option was more likely according to data collected using the IPL paradigm with the
same stimuli, see Lidz, Bunger, Leddon, & Waxman (2006).
EARLY VERB ACQUISITION 333
show the exact same behavior as the experimental group, and point more often
towards the familiar action.
The order of presentation of items was counterbalanced between participants
(four different orders were used in four sub-groups of infants). For each infant,
the familiar action appeared on the right side of the screen half the time, and on
the left side half the time. The left-right order was quasi-random (with no more
than 3 trials in a row with the familiar action on the same side).
Experimental Stimuli
The visual stimuli consisted of 3D objects made using the Amorphium software
database. Four different objects and four different actions were used. All actions
were rigid movements (that is, the object moved without undergoing any defor-
mations). The following 4 object/action pairs were used: a flower rotating on a
horizontal axis, a penguin jumping from left to right while rotating on a horizon-
tal axis, an apple rotating on a vertical axis, and a pot moving up and down.
During test, the same object appeared with two actions: its own action (familiar)
and a novel one that was one of the other three actions. As a result, each action
served once as familiar action and once as novel action (this ensures no bias in
the stimuli, if one action for instance was particularly interesting and attracted
more points than the others). Two other objects and actions were used for the
contrast phases, yielding 4 new object-action combinations, one for each trial: a
frog sliding, a frog jumping from left to right, a hippopotamus sliding, and a
hippopotamus jumping from left to right.
The acoustic stimuli for the familiarization and contrast phases consisted of
child-directed sentences uttered by a female native speaker of French (different
from the puppet’s voice). Four CVC pseudo-words were used as new verbs and
nouns: ‘poune’ /pun/, ‘nuve’ /nyv/, ‘dase’ /daz/ and ‘fome’ /fɔm/. All of these
respect the phonotactics of French and are perfectly plausible both as nouns and
as verbs (3rd person singular present tense; the infinitive forms would be respec-
tively: ‘pouner’ /pune/, ‘nuver’ /nyve/, ‘daser’ /daze/ and ‘fomer’ /fome/). The
full set of the sentences that were used is presented in Table 2 (only the pseudo-
words were changed from one trial to the next). As can be seen from the table,
the noun and verb sentence frames were as well matched as possible as regards
the position of the target word
2
. In addition, they were pronounced with similar
intonations, so that the target word received as much stress and emphasis when it
occurred as a noun or as a verb. Sentences were recorded using a SONY microphone
2
The pseudo-noun often came one syllable later than its verb counterpart, due to the insertion of
an article, and the negative sentences used during the contrast phase do not allow a similar position
for nouns and verbs in French.
334 BERNAL ET AL.
connected to a PC and were sampled at 16 KHz. Visual and auditory stimuli were
combined using Final Cut Pro software into mpeg video clips of 6min 8s,
containing 4 complete trials. The clips were run using Quick Time Player, on a
Macintosh Power G5 computer.
Data analysis
Infants’ behavior was recorded using a Canon MVX10i video camcorder on a
digital cassette. A pencil line drawn in the middle of the highchair’s tablet facili-
tated the coding. Coders identified the direction of pointing for each trial, using
the following 6 categories: ‘left’ (clear pointing response towards the left side),
‘right’ (same for the right side), ‘both’ (simultaneous pointing response to both
sides, using both arms), ‘out’ (pointing outside the screen, at oneself, mum, the
puppet), ‘unclear’ (a pointing response whose direction was ambiguous), and
none (no pointing response during the full 20 seconds of the test trial). Whenever
an infant gave more than one pointing response (e.g. pointing left, then right,
then left again), then the first response was coded. After coding was finished,
all clear pointing responses (‘left’ and ‘right’ categories) were then categorized
into ‘pointing towards the familiar action’ and ‘pointing towards the novel
action’. Twenty-five percent of the infants were double-coded by an indepen-
dent coder (4 infants in each group), with a good inter-coder reliability
(Cohen’s Kappa = 0.81).
RESULTS
The mean numbers of clear pointing responses per infant are shown on Figure 1.
They were analysed in an ANOVA with the between-subject factor Group
(Experimental vs. Control) and the within-subject factor Familiar/Novel (familiar
action vs. novel action). The mean number of clear pointing responses was 5.35
(out of 8 possible responses)
3
, with no main effect of Group
4
or Familiar/Novel
(both F(1,30) < 1). Crucially, the interaction between Group and Familiar/Novel
was significant (F(1,30) = 11.2; p < 0.01). Infants in the Experimental group
pointed significantly more often towards the Familiar action than to the Novel
3
In 2.65 cases out of 8, children did not give a clear pointing response: 0.05 responses were ‘out’,
0.25 responses were ‘both’ (simultaneous pointing to both sides, using both arms), 0.65 were
‘unclear’, and for the rest of the trials, 1.7, the infants did not point at all (there were no significant
differences between the Experimental and the Control groups on any of these variables).
4
The absence of a main effect of Group indicates that children pointed equally often whether they
were taught a new verb or a new noun. It might have been the case that infants in the Control group
pointed less often, since they were asked a puzzling question. But they did not.
EARLY VERB ACQUISITION 335
one (3.3 vs. 2.1, F(1,15) = 6.7; p < 0.02). In contrast, infants in the Control group
pointed significantly more towards the Novel action than to the Familiar one (3.3
vs. 2.0, F(1,15) = 4.8; p < 0.05)
5
.
Thus, infants in the Experimental group, who were taught a new verb, pointed
more often towards the correct answer, the familiar action, in contrast to infants
in the Control Group, who were taught a new noun and showed a novelty prefer-
ence. The significant difference in behavior between the two groups of infants
can only be due to the difference in the sentence frames in which they heard the
novel word (novel verb vs. novel noun).
5
Another way to look at these data is to compute the percentage of points towards the familiar
action, for each infant (number of points towards familiar action, divided by total number of clear
pointing responses for that infant). This percentage analysis gives the same results, with a significant
preference for the familiar action in the Experimental Group (65.9%, significantly greater than
chance, 50%, t(15) = 3.2, p < 0.01); and a significant preference for the novel action in the Control
Group (35.5% points towards the familiar action, significantly smaller than 50%, t(15) = −2.1, p <
0.05), and the behaviour of the two groups was significantly different: t(31) = 3.6, p < 0.001.
FIGURE 1 Number of pointing responses towards the familiar and unfamiliar actions for
infants in the Experimental and Control groups. Error bars represent standard errors of the means.
0
1
2
3
4
5
Experimental Group Control Group
Mean number of pointing responses
Familiar Action
Novel Action
336 BERNAL ET AL.
DISCUSSION
Our aim was to evaluate whether children could use syntactic context to infer that a
word is a verb, and thus refers to an action. In this study, infants were taught novel
words while looking at an object performing an action. Results showed that when
infants heard novel verbs presented in very short sentences, they selectively associ-
ated them to actions. Indeed, when given a choice between the same object per-
forming either the familiar or a novel action, and being asked to point towards “the
one that’s pooning” (if ‘poon’ was the novel verb), they reliably pointed towards
the object performing the familiar action. To ascertain that infants were not simply
perseverating at pointing at the picture they had already seen, infants in the control
group were taught novel nouns on the same visual stimuli. In stark contrast to verb-
learning infants, noun-learning infants pointed more often to the object performing
the new action when asked to point to “the poon” (if ‘poon’ was the novel noun).
This behavior can be considered as a classical novelty preference as observed in
looking experiments (see Bernal, Lidz, Waxman, Dutat, & Christophe, 2005). A
bias to point to the novel action may also arise because infants typically point to
attract their parents’ attention to novel things in their environment (deictic point-
ing). Importantly, infants in the Experimental Group managed to overcome their
novelty preference in order to provide the correct answer to the question they were
asked. The difference in behavior between infants from the experimental and the
control group shows that 23-month-old infants can exploit the syntactic context of
a word online and use it to infer some of its semantic properties.
In this study, infants could only have inferred the word category by using its
co-occurring function words. Mintz (2003) has recently suggested that syntactic
categorization might arise from the use of “frequent frames” (i.e. pairs of words
that co-occur frequently with one word between them). For instance, in English,
the frame [the _ is] allows the algorithm to spot nouns (as in ‘the dog is’, ‘the car
is’, etc); while [I _ it] frames exclusively verbs (as in ‘I see it’, ‘I put it’). This
work has recently been replicated in French (Chemla, Mintz, Bernal, & Chris-
tophe, in press), in which frames such as [ne _ pas] pick out verbs, while [le _ de]
picks out nouns. Some of the sentences we used during familiarization featured
some of the most frequent frames of French (such as [ne _ pas] for verbs, [le _
est] for nouns). The simple co-occurrence of the preceding function word with
the new content word can also give some information about its category. German
infants have recently been shown to use a preceding article to categorize nouns
(Höhle, Weissenborn, Kiefer, Schulz, & Schmitz, 2004).
The ability to use function words as categorizers may arise quite early in
development. Infants distinguish function and content words since the first
months of life (Shi & Werker, 2001; Shi, Werker, & Morgan, 1999) and use them
as a cue to segment continuous speech (Shi & Lepage, in press). Toddlers also
understand sentences better when they contain real function words than when
EARLY VERB ACQUISITION 337
these are replaced by nonsense or misplaced functors (Gerken, 1994; Gerken &
McIntosh, 1993; Kedar, Casasola, & Lust, 2006; Shipley, Smith, & Gleitman,
1969; Zangl & Fernald, 2007). Our work shows that 2-year-olds can use function
words to categorize novel content words and infer their probable meaning. These
results differ from Brown (1957) in two ways. First, in our materials the only
information about the syntactic category of the novel word was carried by the co-
occurring function words. In Brown’s study, both function words and inflectional
morphology were available to distinguish words from different grammatical cate-
gories. Second, the participants in our study were considerably younger than the
preschoolers tested in Brown’s seminal work.
The fact that our participants were able to infer the meaning of a novel verb on the
basis of co-occurring function words raises the question of how infants come to deter-
mine which function words are attached to which category of content words. One
possibility is that infants first group words together by using some kind of distribu-
tional strategy (such as the ‘frequent frames’ mentioned above). This done, they
could ‘label’
6
whole categories as ‘nouns’ or ‘verbs’ by identifying the category of
just a few nouns and verbs amongst them. One way to do this would be to guess the
meaning of a few very frequent words referring to concrete objects and events (see
also Chemla, Mintz, Bernal, & Christophe, in press). In real life, infants probably
combine different strategies to acquire novel words depending on the situation. For
example, Tomasello and Akhtar (1995) presented novel words to 27-month-olds
without any linguistic context but within a meaningful socio-pragmatic context. In
that case, infants were able to link the novel word either to the action or to the object
by relying on the pragmatics of the situation (novel object or novel action). Further
work is needed to understand better how the child combines different strategies in
word learning and what their developmental trajectories are.
Moreover, our experiment widens the scope of the syntactic bootstrapping
hypothesis as a linguistic strategy for lexical acquisition. The present research shows
that infants can use syntax to infer the meaning of different categories of words
(nouns vs. verbs) and not only of transitive vs. intransitive verbs (Fisher, 2002;
Naigles, 1990; Naigles & Kako, 1993). Recently, some studies have also shown that
different sentential contexts (noun vs. adjective) change 14 month-old expectations
about the meaning of a word (Booth & Waxman, 2003). Fisher, Klingler, & Song
(2006) have also suggested that children can use argument structure, in that case the
presence of a noun phrase, to decide whether a word should be considered as a prep-
osition or a noun (“this is acorp my box”, acorp = preposition vs. “this is a corp”, a
corp = NP), and hence whether its meaning is relational. In our experiment, there
6
By labelling, we mean the ability to generalize the properties of one of the members of the cate-
gory to the others. We do not assume that children have conscious access to abstract categories such
as nouns and verbs.
338 BERNAL ET AL.
were no content words, so that only function words could be relied on to compute
the argument structure. All these experiments point to the conclusion that syntactic
structure is crucial for word categorization and lexical acquisition.
Finally, we would like to say a few words about the use of a pointing response
(see also Casenhiser & Goldberg, 2005; Fernandes, Marcus, DiNubila, &
Vouloumanos, 2006; Fisher, 2002). Before running this experiment, we tried to use
the IPL paradigm (using essentially the design of Waxman & Booth, 2003) to test
our hypothesis with similar materials at the same age: the looking data we obtained
was extremely noisy (Bernal, Lidz, Waxman, Dutat, & Christophe, 2005). In these
studies, a few infants spontaneously pointed towards the correct answer. However,
as soon as they had responded, their gaze switched back and forth between the left
and right image, giving rise to very noisy looking data. In the present experiment,
pointing proved to be a reliable methodology, in which infants were keen to engage.
Deictic pointing responses had already been observed in experimental conditions in
12 month-olds (Liszkowski, Carpenter, Henning, Striano, & Tomasello, 2004) and
we built on this ability to train the children to answer questions by pointing.
To conclude, this work shows the ability of French 23-month-olds to use the
syntactic context in which a novel word occurs to infer its meaning. It empha-
sises the crucial role of function words in the lexical categorization of nouns and
verbs. This makes sense since infants start having some knowledge of function
words early in development, possibly towards the end of the first year of life.
AUTHORS’ NOTE
The work presented in this paper was supported by a PhD grant from the French
Ministry of Research to the first author, by a CNRS Poste Rouge to the second
author (in 2004), by two grants from the Direction Générale de l’Armement to
the third author (PhD and post-doctoral fellowships), by a grant from the French
ministry of Research to the fourth author (ACI n° JC6059), by a grant from the
National Institute of Health (R03-DC006829), by a grant from the Agence Natio-
nale de la Recherche (‘Acquisition précoce du lexique et de la syntaxe’, n° ANR-
05-BLAN-0065-01), and by the European Commission FP6 Neurocom project.
REFERENCES
Abbot-Smith, K., Lieven, E., & Tomasello, M. (2004). Training 2;6-year-olds to produce the transi-
tive construction: The role of frequency, semantic similarity and shared syntactic distribution.
Developmental Science, 7(1), 48–55.
Bates, E., Marchman, V. A., Thal, D. J., Fenson, J., Dale, P. S., Reznick, J. S., et al. (1994). Develop-
mental and stylistic variation in the composition of early vocabulary. Journal of Child Language,
21(1), 85–123.
EARLY VERB ACQUISITION 339
Behrend, D. A. (1995). Processes involved in the initial mapping of verb meanings. In M. Tomasello
& W. E. Merriman (Eds.), Beyond names for things: Young children’s acquisition of verbs (pp.
251–273). Hillsdale, NJ: Lawrence Erlbaum Associates.
Bernal, S., Lidz, J., Waxman, S., Dutat, M., & Christophe, A. (2005, July). Role of syntactic cues in
verb acquisition: Results from 23-month old infants. Paper presented at the 10
th
International Con-
ference for the Study of Child Language, Berlin, Germany.
Booth, A. E., & Waxman, S. (2003). Mapping words to the world in infancy: Infants’ expectations for
count nouns and adjectives. Journal of Cognition and Development, 4(3), 357–381.
Brown, R. (1957). Linguistic determinism and the part of speech. Journal of Abnormal and Social
Psychology, 55, 1–5.
Cartwright, T. A., & Brent, M. R. (1997). Syntactic categorization in early language acquisition: For-
malizing the role of distributional analysis. Cognition, 63(2), 121–170.
Caselli, M., Bates, E., Casadio, P., Fenson, J., Fenson, L., Sanderl, L., et al. (1995). A cross-linguistic
study of early lexical development. Cognitive Development, 10, 159–199.
Casenhiser, D. M., & Goldberg, A. E. (2005). Fast mapping between a phrasal form and meaning.
Developmental Science, 8, 500–508.
Chemla, E., Mintz, T. H., Bernal, S., & Christophe, A. (in press). Categorizing words using “Fre-
quent Frames”: What cross-linguitsic analyses reveal about core principles. Developmental
Science.
Childers, J. B., & Tomasello, M. (2001). The role of pronouns in young children’s acquisition of the
English transitive construction. Developmental Science, 37, 739–748.
Fenson, L., Dale, P. S., Reznick, J. S., Bates, E., Thal, D. J., & Pethick, S. J. (1994). Variability in early
communicative development. Monographs of the Society for Research in Child Development, 59.
Fernandes, K., Marcus, G. F., DiNubila, J. A., & Vouloumanos, A. (2006). From semantics to syntax
and back again: Argument structure in the third year of life. Cognition, 100, B10–B20.
Fisher, C. (2002). Structural limits on verb mapping: The role of abstract structure in 2.5-year-olds’
interpretations of novel verbs. Developmental Science, 5(1), 56–65.
Fisher, C., Hall, D. G., Rakowitz, S., & Gleitman, L. (1994). When it is better to receive than to give:
Syntactic and conceptual constraints on vocabulary growth. Lingua, 92, 333–375.
Fisher, C., Klingler, S. L., & Song, H. (2006). What does syntax say about space? 2-year-olds use
sentence structure to learn new prepositions Cognition, 101, B19–B29.
Forbes, J. N., & Poulin-Dubois, D. (1997). Representational change in young children’s understand-
ing of familiar verb meaning. Journal of Child Language, 24(2), 389–406.
Gentner, D. (1981). Some interesting differences between verbs and nouns. Cognition and Brain
Theory, 4(161–178).
Gerken, L. (1994). Young children’s representation of prosodic phonology: evidence from English-
speakers’ weak syllable production. Journal of Memory and Language, 33, 19–38.
Gerken, L., Landau, B., & Remez, R. E. (1990). Function morphemes in young children’s speech per-
ception and production. Developmental Psychology, 27, 204–216.
Gerken, L., & McIntosh, B. (1993). Interplay of function morphemes and prosody in early language.
Developmental Psychology, 29(3), 448–457.
Gillette, J., Gleitman, H., Gleitman, L., & Lederer, A. (1999). Human simulations of vocabulary
learning. Cognition, 73(2), 135–176.
Gleitman, L. (1990). The structural sources of verb meaning. Language Acquisition, 1(1), 3–55.
Grimshaw, J. (1981). Form, function and the language acquisition device. In C. L. Baker & J. J.
McCarthy (Eds.), The logical problem of language acquisition. Cambridge, MA: The MIT
Press.
Hirsch-Pasek, K. A., Golinkoff, R. M., Maguire, M. J., & Imai, M. (2005, April). Facilitating verb
learning: What makes verb learning so difficult? Paper presented at the Biennial Meeting of the
Society for Research on Child Development, Atlanta, GA.
340 BERNAL ET AL.
Höhle, B., Weissenborn, J., Kiefer, D., Schulz, A., & Schmitz, M. (2004). Functional elements in
infants’ speech processing: The role of determiners in the syntactic categorization of lexical
elements. Infancy, 5(3), 341–353.
Imai, M., Haryu, E., & Okada, H. (2005). Mapping novel nouns and verbs onto dynamic action
events: Are verb meanings easier to learn than noun meanings for Japanese children? Child Devel-
opment, 76, 340–355.
Kedar, Y., Casasola, M., & Lust, B. (2006). Getting there faster: 18- and 24-month-old infants’ use of
function words to determine reference. Child Development, 77(2), 325–338.
Kern, S. (2007). Lexicon development in French-speaking infants. First Language, 8, 227–250.
Kersten, A. W., & Smith, L. B. (2002). Attention to novel objects during verb learning. Child Devel-
opment, 73, 93–109.
Landau, B., & Gleitman, L. (1985). Language and experience: Evidence from the blind child.
Cambridge, MA: Harvard University Press.
Lidz, J., Bunger, A., Leddon, E., & Waxman, S. (2006, March). When one cue is better than two:
Syntactic vs. lexical information in infant verb learning. Paper presented at the CUNY Workshop
on on-line methods in children’s sentence processing.
Lidz, J., Gleitman, H., & Gleitman, L. (2003). Understanding how input matters: Verb learning and
the footprint of universal grammar. Cognition, 87(3), 151–178.
Liszkowski, U., Carpenter, M., Henning, A., Striano, T., & Tomasello, M. (2004). Twelve-month-olds
point to share attention and interest. Developmental Science, 7(3), 297–307.
Macnamara, J. (1982). Names for things: A study of human learning. Cambridge, MA: The MIT
Press.
Maratsos, M., & Chalkley, M. A. (1980). The internal language of children’s syntax: The ontogenesis
and representation of syntactic categories. In K. Nelson (Ed.), Children’s language (Vol. 2). New
York: Gardner Press.
Mintz, T. H. (2003). Frequent frames as a cue for grammatical categories in child directed speech.
Cognition, 90(1), 91–117.
Mintz, T. H. (2006). Finding the verbs: distributional cues to categories available to young learners.
In K. A. Hirsch-Pasek & R. M. Golinkoff (Eds.), Action meets words: How children learn verbs:
Oxford University Press.
Mintz, T. H., Newport, E. L., & Bever, T. G. (2002). The distributional structure of grammatical cat-
egories in speech to young children. Cognitive Science, 26, 393–424.
Naigles, L. (1990). Children use syntax to learn verb meanings. Journal of Child Language, 17(2),
357–374.
Naigles, L., & Kako, E. T. (1993). First contact in verb acquisition: defining a role for syntax. Child
Development, 64(6), 1665–1687.
Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society for Research
in Child Development, 38(1–2), 136.
Pinker, S. (1987). The bootstrapping problem in language acquisition. In B. Mc Whinney (Ed.),
Mechanisms of language acquisition. Hillsdale, NJ: Erlbaum.
Quine, W. V. (1960). Word and object. Cambridge, MA: The MIT Press.
Rice, M. (1980). Cognition to language: Categories, word meanings, and training. Baltimore, MD:
University Park Press.
Shafer, V. L., Shucard, D. W., Shucard, J. L., & Gerken, L. (1998). An electrophysiological study of
infants’ sensitivity to the sound patterns of English speech. Journal of Speech, Language and
Hearing Research, 41(4), 874–886.
Shi, R., & Lepage, M. (in press), The effect of functional morphemes on word segmentation in pre-
verbal infants. Developmental Science.
Shi, R., Werker, J., & Cutler, A. (2006). Recognition and representation of function words in English-
learning infants. Infancy, 10, 187–198.
EARLY VERB ACQUISITION 341
Shi, R., & Werker, J. F. (2001). Six-months old infants’ preference for lexical words. Psychological
Science, 12(1), 71–76.
Shi, R., Werker, J. F., & Morgan, J. L. (1999). Newborn infants’ sensitivity to perceptual cues to lex-
ical and grammatical words. Cognition, 72(2), B11–21.
Shipley, E. F., Smith, C. S., & Gleitman, L. (1969). A study in the acquisition of language: Free
responses to comments. Language, 45(2), 322–342.
Theakston, A. L., Lieven, E., Pine, J. M., & Rowland, C. F. (2001). The role of performance limita-
tions in the acquisition of verb-argument structure: An alternative account. Journal of Child
Language, 28, 127–152.
Tomasello, M., & Akhtar, N. (1995). Two-year-olds use pragmatic cues to differentiate reference to
objects and actions. Cognitive Development, 10, 201–224.
Waxman, S., & Booth, A. E. (2001). Seeing pink elephants: Fourteen-month-olds’ interpretations of
novel nouns and adjectives. Cognitive Psychology, 43(3), 217–242.
Waxman, S., & Booth, A. E. (2003). The origins and evolution of links between word learning and
conceptual organization: New evidence from 11-month-olds. Developmental Science, 6, 128–135.
Waxman, S., & Lidz, J. (2006). Early word learning. In D. Kuhn & R. Siegler (Eds.), Handbook of
child psychology: Cognition, perception and language (6th ed., pp. 299–325). New York: John
Wiley & Sons.
Woodward, A. L., & Markman, E. M. (1998). Early word learning. In W. Damon, D. Kuhn & R. Siegler
(Eds.), Handbook of child psychology, Volume 2: Cognition, perception and language (pp. 371–420).
New York: John Wiley and Sons.
Zangl, R., & Fernald, A. (2007). Increasing flexibility in chidlren’s online processing of grammatical
and nonce determiners in fluent speech. Language Learning and Development, 3, 199–231.
