F ~ I - Language,
s/
17 (19971, 0 5 7 4 7 4 . Printed in England
Speaker perspective and reference in young children*
EVE V. CLARK
TRISHA A. SVAIB
) Stanford University
ABSTRACT
When speakers choose a word, they choose the perspective froin
which they wish to present an entity or an event. In the present
study we tested the hypothesis that young children accept multiple
perspectives from an early age. That is, they know that two terms
can refer to the same entity, as shown by their comprehension and
production of multiple terms for the same referent. A total of 36
children aged 2;2 to 4;8 answered questions about pictures that
required shifts in level (e.g., from cat to animal or the reverse) or
domain (e.g., from dog to sailor or the reverse). Even the
youngest two-year-olds managed both tasks: they understood
multiple terms for the same referents and also produced second
terms for referents that had already been labelled. These data
support a many-perspectives view of lexical acquisition over a
single-perspective view.
Recent research on word acquisition in young children has focused on
possible semantic constraints that, by limiting the options, make the
initial acquisition of new word meanings simpler (Markman 1989).
One assumption has been that children make things easier by limiting
their lexicon to one term per referent. The evidence for this comes from
two sources: first, from anecdotal reports of spontaneous rejections of
This research was supported in part by a grant to the Cognitive Science Group at
Stanford University from the Sloan Foundation, and in part by a National Science
Foundation Graduate Fellowship to the second author. A preliminary version was
presented at the 1991 Seminar on Child Language, University of Manchester, and
we thank participants at the meeting for their helpful questions. We would like to
thank the staff of Bing Nursery School, Stanford, for their co-operation, and all
the children who participated so willingly in our study; we also thank Caroline
Cuevas and Sandy H.-S. Lai for help in collecting the data, and Judith Wasow,
Infant Studies Laboratory, for aid in recruiting two-year-olds. Address for
colrespondence: Eve V. Clark, Department of Linguistics, Stanford University,
Stanford, CA 94305-2150, USA. E-mail: [email protected]
0142-7237/97/170 1-0057 $5.00 O Alpha Academic
58
FIRST LANGUAGE
second terms for referents for which children already have words (e.g.,
Macnamara 1982, Valentine 1942); and second, from experimental
studies of three- and four-year-olds where the children often reject a
second term that has been taught for an already-labelled referent (e.g.,
Markman 1987, Markrnan & Wachtel 1988, Merriman 1986, Merriman
& Bowman 1989). A child who uses the term dog in referring to dogs
should reject uses of animal or spaniel on the one hand, or of pet or
hunter on the other, in reference to dogs. In short, this research
suggests that if children can at first manage just one term to refer to
instances of a particular kind, they must be taking just one perspective
on that type. Later, of course, children must over-ride such constraints
and learn that they can apply multiple terms to a singIe referent - a dog
can be the dog, that animal,Tim'spet, or the spaniel, according to the
perspective chosen by the speaker on each occasion. En lhis account,
then, children start out with a one-perspective view, and only at around
three or four begin to master additional perspectives (with the
additional terms) on the things they talk about.
For adult speakers, though, we take for granted a many-perspectives
view: words offer speakers flexible ways of categorizing and presenting
the world to others. They can talk about a tree stump by referring to the
stump on one occasion, to my chair on another, or to that table on a third
(Miller 1978). On each occasion, the speaker takes a different conceptual
perspective: in the first reference the tree smmp is presented as a stump,
in the second as a seat, and in the third as a supporting surface. With
each categorization, the speaker's perspective i s marked by choice of
the relevant term (CIark 1996). Under this view of lexical choices for
reference, speakers can take many different perspectives on the same
entity, with each perspective reflecting a different categorization.
Even when adults talk to one- and two-year-olds, they offer multiple
perspectives from the start. As Brown (1958) pointed out, adults use a
variety of terms for the same entity, with their choices typically
determined by what appears most useful in context: a n apple vs. some
fiuit vs. something fo cat vs. a slice. Adults reword utterances to oneand two-year-olds and, in doing so, often shift perspective in their
efforts to get the children to attend to directives, as in 'Pick up the
blocks. Put your blocks in the box. Put them in here. The blocks go
back in the box. Can you put them in?' (Shatz 1978, Snow 1972). And
they offer explicit directions to young children when they introduce
new words for familiar, already-labelled, referents, with kind of, for
example, as in 'A ladle is a kind of spoon'; or when they introduce
superordinate terms as in 'Cats and dogs and bears are all animals'
(e.g., Blewitt 1983, Callanan 1985, Shipley, Kuhn & Madden 1983).
PERSPECTIVE AND REFERENCE
59
Such introductions make clear what the relation is between two (or
more) terms produced, and hence what kind of shift in perspective is
required in the move from one term to the other. In short, adult
speakers tacitly assume that even very young children are capable of
realizing that an entity can be considered from more than one
perspective. The question is, do young children start with a oneperspective view and then move on gradually to a many-perspective
view, as suggested by many of the experimental studies of early wordlearning? Or do children start out by assuming many perspectives - and
hence multiple terms for the same referent?
There is some evidence that the latter might be the case. First, in
their spontaneous speech one-year-olds may refer to a toy horse on
wheels with the term horse on one occasion when pulling it along, but
with chair on another when sitting down on it. Or they may refer to a
waste-basket with the term basket when looking at it, but shift to hider
when putting it over their heads (Clark 1995a, 1996). Second, twoyear-olds readily construct second terms - subordinate terms - for
subtypes in familiar categories: they coin terms like Dalmatian-dog and
poodle-dog for kinds of dogs, otherwise referred to with dog, or Volvocar and Ford-car for kinds of cars, otherwise referred to with the term
car (e.g., Clark, Gelman & Lane 1985; see also Waxman & Senghas
1992). Third, when they reject second terms, they are probably
rejecting the perspectives thus presented and not the words themselves.
For instance, a child who knows the two terms bird and pal-rot, may on
occasion reject bird in favour of parrot - presumably because he is
concerned with the particular kind of bird rather than just with whether
something is a bird or not. Detailed analysis of all the rejections
recorded in a diary study (Clark 1993) showed that the majority were
rejections of familiar terms already known to and produced by the
child. The child's rejections, then, could not have been rejections of the
words as second terms for the referents in question; instead, they
appeared to be rejections of those perspectives.'
The rejections in the anecdotal reports, however, seem to have been rejections
based on the meanings of the words rejected. Grieve (1975), Macnamara (1982)
and Valentine (1942) reported similar episodes: in two, children rejected use of
thc superortlinate term rrnimal when it referred to a single instance, but accepted
i t whcn it referred to a group of animals. Thcse rcjeclions. then, appear to have
been based on their meaning for atrimctl, which ror them denoted a collection and
not a kind. As a rcstllr. it could not apply to single instances. The third case was
similar, with a child treating col- as a collection term ancl therefore as inapplicable
lo single cars. In short, the children's meanings for nnimc~land car. were not yet
equivalent to the conventional superordinate meanings assumed by adults. and
their collection meanings precluded their being used to refer to single instances.
60
FIRST LANGUAGE
The present study was designed to test the hypothesis that young
children start out with a many-perspectives view. The manyperspectives view hypothesizes that, from the start, children take for
granted that speakers use multiple terms for things, and realize that
different words serve to present the same object or event from different
perspectives. In the same way that in early play a pebble can be
presented as a pebble or as a car, so a block can be referred to with the
term block when building or with apple when handed to an adult in
play. The two terms, block and apple, mark two different perspectives
(two different categorizations) on the part of the speaker. This view
therefore predicts that even very young children should accept multiple
terms for the same referent and should be willing to produce a second
term for a referent that has already been identified with a first one. In
contrast, the one-perspective view hypothesizes that two- to four-yearolds will reject multiple terms applied to the same referent under the
assumption that they implicitly apply an economy metric in the early
stages of lexical acquisition. The metric says: 'You only need one term
for an object; don't bother with any more than one.' Children should
also be unwilling, according to this view, to produce a second term for
anything that has already been labelled.
To test these predictions, we chose a simple book-reading task where
children had to identify specific characters when asked to point to
instances of various categories, and to produce a second appropriate
term for each referent. Two types of perspective potentially accessible
even to young children are the perspectives involved in shifts of Level
(dog vs. animal, or oak vs. tree) and shifts in Domain (boy vs. friend,
or goldfish vs. pet). But two-year-olds generally have a rather limited
vocabulary and the ability to recognize and make use of different
perspectives depends on vocabulary, so it was unclear whether children
would find it easier to deal with terms from two distinct levels or terms
from two distinct domains. The present study examined both
possibilities in a pragmatically natural task already familiar to the
youngest children. The age range was chosen to include the full range
for which researchers have assumed that constraints apply on word
acquisition, namely two- to four-year-olds.
METHOD
Procedure
Children were interviewed individually in a small room adjacent to
their classroom in the nursery school. The experimenter asked each
child to take part in two tasks by looking at a picture-book with her.
PERSPECTIVE AND REFERENCE
61
The book contained eight large pages, each covered with Richard
Scarry-like drawings of animals, similar to those in Fig. 1. For the
Level task, children saw four pages. each depicting a general activity
scene where three different kinds of animals were all engaged in the
same event, for example, all helping build a house or all playing with
toys. A11 the animals were clothed and carried or were accompanied by
accessories such as tools, hats or balloons. For the Domain task,
children also saw four pages with four vignettes on each. In each
vignette two different animals (e.g., a pig and a cat) were shown
engaged in the same activity (e.g., sailing, flying, baking). In the Level
task the experimenter asked questions designed to test comprehension
and to elicit terms from two levels, superordinate and basic, for the
animal-types pictured. For example, she might ask whether a character
referred to as the bear was also an animal, or what that animal was (a
cat). In the Domain task, she asked questions about two different
domains, professions and animal-types. For example, she might refer to
one character as both the cat and the pilot, or ask of thcfrshertnan what
he was (a rabbit). The first page in each set of four was used to
familiarize the children with that task.
The experimenter introduced the Level task by showing each child
the first activity page with 12 animals on it and asking some questions
about it. (These questions were not scored or analysed.) She first asked
the child to show her 'all the animals'. If the child did not point to all
the animals on the page, she asked 'Are there any more? Can you point
to them?' She then asked the child to tell her what each animal was
('What's this?' 'What's this one?'), to make sure that all the animals
pictured were familiar. If a child mislabelled an animal, the child's term
was used in all subsequent questions about that animal-type. For
instance, a few children called the dogs in some pictures hears. When
this occurred, the term hear was used from then on for those children.
Lastly, the child was asked to show the experimenter the animals
wearing hats. This was done lo familiarize the children with pointing
out a specific subset of the animals.
After turning over the first page, the experimenter then asked a
systematic series of questions about the animals on the three remaining
activity p a p s . After an introductory request (e.g.. 'Show me all cats').
she asked For each animal type 'What kind of animal is this?' to elicit
specific animal terms. Also. to check on comprehension and acceptance
of two terms for the same referent, she asked questions like 'Is this dog
an animal?' In all, each child answered six comprehension and six
production questions. Finally. as a check on how the child was applying
the word animal (versus car, say), she asked if non-animate objects
62
FIRST LANGUAGE
PERSPECTIVE AND REFERENCE
1
Fig. 1. Sample pictures for (A) the Level task, and (B) the Domain task
63
were animals (e.g., 'Is this hammer an animal?'). In this way, all the
animals in each picture were either identified as referents of two
distinct terms or labelled with a second term, by each child, in response
to questions checking on multiple terms for the same referents.
In the Domain task, the experimenter again used four vignette pages,
each with four small scenes showing two animals performing some
activity. Here too she asked children a preliminary set of questions (not
analysed) about the first page. She asked the child to find a certain type
of animal, 'Can you find a cat?', then she asked, 'What is that cat?' If
the child responded by naming the activity (for example, 'He's
sailing'), rather than by giving a noun for the occupation, the
experimenter prompted by asking, 'What do you call someone who
does that?' If the child still did not give an occupation term, the
experimenter followed up a second time with, for example, 'Is that cat
a sailor?' This procedure was then repeated a second time to make sure
the child was attending both to the occupations and the kinds of
animals pictured. (The occupations pictured and the terms for each
were selected after pilot-testing with terms listed as common in threeyear-old vocabularies; Edwards & Gibbon 1964, Rinsland 1945.) For
the three remaining pages, the experimenter asked a systematic series
of questions: first, to find a specific kind of animal, 'Can you find a
dog?' and then to identify that animal's occupation, 'What is this dog?'
(or the reverse). Then she asked the child to find all the animals with a
certain occupation, 'Can you find all the nurses?' and then (of the two
animals identified) asked 'What is this nurse?' where the response
expected was the term for the kind of animal (a rabbit, say). Then she
asked further questions with the form: 'Is this X a Y?' in which the
terms for occupation and animal-type alternated as the first term, e.g.,
'Is this cat a fireman?' and 'Is this nurse a dog?' Two of the questions
were designed to be answered in the negative to ensure that children
were not simply agreeing with whatever was asked. When the child
answered 'No', the experimenter followed up with a further question,
for example, for a cat that was not a fireman but a painter, 'What is this
cat then?' Each child answered 12 comprehension and 12 production
questions. Both tasks were recorded for later transcription.
Materials
The materials consisted of a book with eight pages of pictures. All
pictures were drawings of animals modelled on those in the Richard
Scarry series of children's books (e.g., Scarry 1968). This type of
drawing was used because most children are familiar with these
books and are used to looking at such pictures and talking about the
64
FIRST LANGUAGE
activities depicted. The animals in the pictures were dogs, cats, rabbits
and pigs.
For the Level task, there were four pages, each depicting some
general activity designed to elicit children's comprehension and
production of terms for the participants from different levels in the
same domain. The first page contained 12 animals (three of each of the
four kinds), several wearing hats, and surrounded by different types of
fruit. This page was used as a pretest for the remaining three pages
which each contained six animals (two each of three kinds), with each
picture depicting a joint activity - all six animals helping to build a
house, all six playing with toys and all six driving vehicles. Each child
answered six questions about each page (two checking on
comprehension, two eliciting production, and one plus a follow-up as a
check on children's general propensity to answer 'yes').
The other four pages for the Domain task were vignette pages
designed to find out about children's knowledge of terms from two
domains. Each page contained four vignettes, with each vignette
depicting an activity with two different animals dressed appropriately
for that activity. Each page therefore showed eight animals in all, two
of each of four kinds. The first page was used to introduce the task and
the remaining three for elicitation. Each child answered 12 questions
per page (four to identify and Iabel: subsets, as in 'Can you find the
rabbits?' followed by 'What are they?', another four on comprehension
as in 'Is this cat a captain?', with two questions requiring positive 'yes'
answers m d two requiring 'no' answers. and four on production as in
'What is rhe other cook?').
The order of presentation of the Level and the Domain tasks was
counterbalanced across subjects within each age group.
Subjects
Thirty-six children participated in the study. They were divided into three
age groups of 12 each (with six boys and six girls at each age level):
two-year-olds (2;O to 2;11, mean age 2;5), three-year-olds (3;O to 3;10,
mean age 3 3 ) and four-year-olds (4;O to 4;8, mean age 4;3). All the
children were from middle-class backgrounds and were learning English
as their first language. Half the two-year-olds and all the three- and fouryear-olds were attending the same nursery school; the other two-yearolds were contacted through the Infant Studies Laboratory at Stanford.
RESULTS
All the children, regardless of age, succeeded in the two tasks and
PERSPECTIVE A N D REFERENCE
i
)
1
65
successfully answered nearly all the questions put to them. In so doing,
they accepted multiple terms and produced second terms in addition to
an initial term already presented and accepted. The children's responses
in the two tasks were scored first for the appropriateness of the
response to each question asked. Some questions required children's
acceptance of two terms for the same referent; others elicited a second
term for a referent already identified by means of a first term. All the
children at each age level succeeded in answering some, and often all,
of these questions appropriately in both tasks. None of the children
rejected two terms for the same object, and nearly all of them were able
to produce a second term for an already-labelled object when asked.
All the children answered all the set and subset questions correctly,
so these will not be analysed further. In response to the comprehension
and production questions, children offered appropriate answers 96% of
the time overall, with little difference between the two tasks: the Level
task received 97% appropriate responses and the Domain task 96%.
Even the two-year-olds were almost at ceiling, with 93% of their
responses overall being appropriate. When they did not give
appropriate responses, they typically said 'Don't know', gave no
response at all, or, on rare occasions, gave an irrelevant response.
Three- and four-year-olds gave appropriate responses 98% and 100%
of the time, respectively.
The Level task
In the Level task, the experimenter asked children six questions (two
per page) that tapped their acceptance (comprehension) of two terms in
the same utterance for the target referent, and six questions that elicited
their own production of a second term. Overall all 12 children at each
of the three ages answered nearly all of these questions appropriately.
The percentages of these responses by age are shown in Table 1. There
were no significant differences with age in either comprehension or
production. That is, even the two-year-olds supplied appropriate
responses to these questions over 90% of the time.
In the Level task, children were also asked two further questions per
page, designed to check on whether they understood that the term
animal applied only to animals and not to inanimate objects. For
example, the experimenter might point at a hammer in the picture and
ask 'Is this an animal?' and follow up with 'Why not?' (or 'Why?').
The first of these two questions therefore required the answer 'no'. This
also served as a check on children's tendency to give compliant 'yes'
answers to comprehension questions. Many of the two-year-olds found
this question hard. But six of the 12 answered 'no' on at least one
66
TABLE 1.
67
PERSPECTIVE AND REFERENCE
FIRST LANGUAGE
Understanding and producing multiple terms in the Level task
TABLE 2. Understanding and producing multiple terms in the
Domain task
Age
Understood (96)
Produced (%)
3;5
92
97
99
99
4;3
100
100
96
99
2;5
Mean
Age
No. children
2;5
3;5
4:3
12
12
12
95
99
100
No. children
11
12
12
98
Mean
Note. Each percentage is based on 72 data points.
Understood (%)
Produced (%)
86
99
100
95
Note. Each percentage is based on 144 data points.
occasion (out of three), another five relied on 'Don't know' or refused
to answer, Overall, two-year-olds offered appropriate responses here
only 28% of the time, compared with 92% for the three-year-olds and
100%for the four-year-oIds. At the same time, all the two-year-olds (as
we11 as the older children) responded appropriately on the introductory
page when asked to pick out, say, 'all the animals' vs. 'all the cats'.
They never included non-animate objects in their choices and they did
include all the possible target referents. These data suggest that even
the two-year-olds had a good understanding of the nature of the
perspective shifts from cat to animal, say, and from animal to dog. But,
unlike the three- and four-year-olds, they did have difficulty with the
question requiring a 'no' answer and with providing reasons for such
an answer. At the same time, their general acceptance of multiple labels
in the other comprehension questions (Table 1) is further backed up by
their ability to produce second terms.
The Domain task
In the Domain task, the experimenter asked 12 questions (four per
page) designed to check on children's understanding and acceptance of
two terms, both referring to the same character, and another 12
questions (again, four per page) to elicit production of a second term
for a previously labelled character. The percentages of appropriate
responses are shown in Table 2. Again, as in the Level task, even the
two-year-olds did very well on both types of questions and gave
appropriate responses 95% of the time for comprehension questions
and 86% for production ones. While there were no differences with age
for comprehension, there was a significant difference for production
(F(2,33) = 27.05, p < 0.001). with the two-year-olds providing
significantly fewer twms (86%) than the three- or four-year-olds (99%
and 10096, respectively). This suggests that the larger vocabulary
required by the Domain task was somewhat more demanding than the
Level task, in production, for the youngest children: they Ilad to be able
TABLE 3. Percentage of appropriate responses to comprehension
questions requiring 'Yes' vs. 'No' answers in the Domain task
I
Age
'Yes'
'No'
Note. Each percentage is based on 72 data points.
to come up with terms for occupations as well as for animal subkinds,
and this was sometimes difficult.
In this task, half the comprehension questions were designed to
require a 'yes' answer in response to questions such as 'Is that cat a
painter?' (where this was true), and half were designed to require a 'no'
answer, again as a check on any tendency to give only 'yes' answers.
Ten of the 12 two-year-olds answered all six 'yes' questions
appropriately; one answered five of the six questions appropriately, and
one answered three. With the 'no' questions, all 12 two-year-olds
answered at least one appropriately, four answered all six appropriately,
three answered five, two answered four, two answered three, and one
answered only one appropriately. In short, even the two-year-olds
showed here that they could answer 'no' questions appropriately.
Among the three- and four-year-olds, all 12 children at each age
answered both the 'yes' and the 'no' comprehension questions
correctly. Table 3 gives the percentage of correct responses for 'yes'
and 'no' comprehension questions separately, by age.
In summary, this pattern of responses, with even the youngest
children (aged two) accepting and producing multiple terms of the
same referents nearly all the time, is predicted by the many-perspectives
68
FIRST LANGUAGE
account, but not by the one-perspective one. Although two-year-olds
had difficulty with comprehension questions requiring the answer 'no'
in the Level task, data from the same children in the Domain task
showed that they could answer questions requiring 'no' appropriately
76% of the time overall, with 11 of the 12 two-year-olds giving
appropriate 'no' answers at least half the time.
Finding the right word
To shift from a term at one level to a term at another, children need to
know the conventional words for the categories in question and to
understand the relation between the words at different levels. In the
Level task, they needed to know the terms animal, cat, rabbit, bear and
dog. In the Domain task, they needed to know rather more - not only
terms for the animal-types, as in the Level task, but also terms for
occupations such as cowboy, cook, sailor and pilot. In order to respond
appropriately to the questions, either by pointing out the relevant
characters or by answering 'yes' or 'no' or by producing an appropriate
second term for some character, children had to be able to shift
perspective from basic to superordinate level (or the reverse) in the
Level task and from animal-type to occupation (or the reverse) in the
Domain task. To do this, children have to know the appropriate words.
The high number of responses at all three ages (Tables 1 and 2) showed
that children as young as two years can both understand and produce
multiple terms for the same referent; and on most occasions, they
appeared to have no difficulty in doing this.
But they did sometimes fail to find the right words. On such
occasions, children sometimes coined a new word for the category in
question or, having proposed one term, repaired it with another term.
They were more likely to coin words and to make repairs when shifting
perspective in the Domain task than in the Level task, presumably
because they needed to draw on a greater number of terms for the
Domain task.
Children coined new words on only four occasions (four different
children) in the Level task, but on 45 occasions (19 different children)
in the Domain task, for an 8% to 92% split (significant at p < 0.001 by
Mann-Whitney). In these coinages, children nearly always constructed
compound nouns (90%) rather than derived forms (10%) (see Clark
1993). Typical innovations from the Domain task include: jump-roper
(2;5, for target 'cowboy' [a cat holding a lasso]), rabbit-drummer (2;8,
for target 'soldier'); boat-rider (3; 1, for 'sailor'), fish-guy (3; 10, for
'fisherman'), parader (3;8, for 'soldier'); airplane-driver (4;0, for
'pilot'), fisher (4;0, for 'fisherman'), and drummist (4;6, for 'soldier').
I
PERSPECTIVE AND REFERENCE
I
I1
I
I
I
69
The rare coinages in the Level task are exemplified by pig-animal (2;4,
for 'pig' as a subtype). This compound is similar to forms like Volvocar or poodle-dog coined by two-year-olds for subkinds (Clark et a/.
1985). Most of their novel compounds identified an agent as the head
noun (-man, -guy, -driver, -rider, etc.) and an object or location as the
modifier (e.g., drum-,fish-, boat-).
Children at every age also distinguished level shifts from domain
shifts in the number of repairs they made to their responses. Although
repairs to the term being elicited were not frequent, seven children
made at least one repair, for a total of eight in the Level task, while 18
children made repairs, for a total of 29 in the Domain task, for a 22% to
78% split (significant at p < 0.01, by Mann-Whitney). Four-year-olds
made more repairs than the two- or three-year-olds in the Domain task,
apparently in an effort to be more precise. One four-year-old, for
instance, in response to the experimenter's asking 'What is this
rabbit?', replied 'Drummer. Oh no no no. A soldier.' In their repairs to
domain-shift responses, four-year-olds offered a more specific label
65% of the time, compared with simply changing the label offered the
rest of the time. Overall, then, children made more repairs and also
produced more lexical innovations in the Domain task than in the Level
task.
These differences between the two tasks, though small, are
consistent with the greater demands placed on the children in the
Domain task when they were asked to produce terms for both
occupations and animal subkinds. They are also consistent with the one
age difference observed in production: in the Domain task, two-yearolds produced significantly fewer appropriate terms than the three- and
four-year-olds.
In summary, even the two-year-olds were very close to ceiling in
their responses to both tasks. They could all deal appropriately with
two distinct terms for the same referent nearly all the time, and there
were no significant differences with age for comprehension in either
task. All the children, regardless of age, gave over 90% appropriate
responses. In production, two-year-olds did slightly less well in the
Domain task, but were no different from three- and four-year-olds in
the Level task. But the Domain task was slightly more demanding
overall in that it also elicited more lexical innovations and more repairs
than the Level task. Overall, children at all three ages showed
considerable skill in shifting perspective, whether in responding to
comprehension questions that contained multiple terms or in producing
further terms of their own (Tables 1 and 2). These results are consistent
with the predictions of the many-perspectives hypothesis: even the
70
FIRST LANGUAGE
PERSPECTIVE AND REFERENCE
youngest children tested were virtually at ceiling in their responses,
whether for comprehension of multiple terms for the same referent or
for production.
DISCUSSION
The present findings offer strong support for the view that children are
able to use two distinct terms for the same entity: from age two, they
accept two distinct terms for the same referent, with no rejections of the
second term. And, also from age two, they are able to produce a second
term for referents that have already been labelled with a first one.
Children succeeded in this in the Level task, where they used the
superordinate term animal and various animal-type terms (e.g., bear,
rabbit, pig), and in the Domain task, where they used terms for
occupations (e.g., pilot, sailor) and for animal-types (e.g., cat, dog). In
short, these data support a many-perspective view of children's ability
to use words, even in the earliest stages of language acquisition, over a
one-perspective view which assumes that children can apply just one
term per referent.
The present data, then, are consistent with and add to earlier
findings, both from children's spontaneous speech and from elicitation
tasks in which children showed they could use multiple terms for the
same object. In a new-word learning task, for example, Waxman &
Senghas (1992) showed that two-year-olds (mean age 2;l) readily
learned two terms, differing in level, for the same referent. Two-yearolds also readily construct new compound nouns for subcategories of
familiar categories, and produce nouns like snow-car for a subkind of
car (otherwise referred to with car), or poodle-dog for a subkind of dog
(Clark et al. 1985; also Berman & Clark 1989, Gelman, Wilcox &
Clark 1989, Waxman, Shipley & Shepperson 1991).
These findings are also consistent with findings that three- and fouryear-olds can use multiple terms for the same referent. Waxman &
Hatch (1992), for example, elicited terms for pictures from three- and
four-year-olds by asking appropriately worded questions (e.g., of a
rose, asking 'Is this a dandelionla treeIan animal?') to see whether
children could produce more than one term for the same referent. They
found that most three-year olds (75%) and all but one four-year-old
produced more than one term for the same object on at least 50% of
trials (see also Au & Glusman 1990).
The findings of these studies, together with the present results, offer
strong evidence that children can understand and produce more than
one label for an object, and they can do so as young as age two. But for
!
I
I
I
I
I
71
this to occur, use of a second (or subsequent) term needs to be
motivated, to make sense to the child either within the context of the
task itself, as in the book-reading tasks, or through use of added
pragmatic directions about the relation between an earlier, perhaps
familiar, term and the later one. Such directions may be non-linguistic,
in the form of heightened attention from the speaker and other signals
of interest marked by eye-gaze, physical stance, gestures, pointing,
demonstrations of action or function, and so on. Or the speaker may
draw the child's attention to how certain objects or actions differ in the
same setting (e.g., Akhtar & Tomasello 1996, Tomasello & Akhtar
1995, Tomasello & Barton 1994, Waxman & Senghas 1992). Other
pragmatic directions, of course, are linguistic, as when the adult
interlocutor specifies the relation between the term already known and
a new term being offered, e.g., 'An X is a kind of Y' or 'Xs are baby
Ys' (Blewitt 1983, Callanan 1985, 1989, Shipley et al. 1983). In
conversation, children are usually offered both non-linguistic and
linguistic clues to the relations among the different terms being used by
the speaker for the same referent (Clark 1995b).
Yet experimental studies of early word learning have suggested that
three- and four-year-olds often fail to accept second terms for referents
already labelled. Why should this happen? We suggest that one reason
for this is that children lack information about how the second term is
connected to the first. Use of a second term without any pragmatic
directions, non-linguistic or linguistic, present children with a problem:
Is the second term being introduced a repair made to replace the first
term? Is the second term more specific in meaning than the first? (Or
more general?) Is the second term drawn from a different domain? In
everyday conversation, adult speakers are likely to offer explicit
information about such options, but just this information is typically
missing from the word-learning contexts. So children have to decide,
on their own, what the status of a second term is. If it is a mistake, for
example, then adherence to a prior term (with implicit rejection of any
later one) makes a reasonable strategy for responding. When the
referent objects are unfamiliar as well, children may take a different
route instead and assume that the first term was a mistake and so stick
only with the second one proffered by the speaker. Both interpretations
would be reasonable where the speaker has given no indication of how
the two terms are related. And in either case children will appear to
accept only one term (e.g., Merriman 1986), and hence only one
perspective, on the referent in question. In short, children try to make
sense of experimental tasks in the light of what they know about
everyday conversation (e.g., Siegal, Waters & Dinwiddy 1988). Where
72
FIRST LANGUAGE
these tasks do not conform to everyday pragmatic expectations,
children may have a difficult time working out what the intended
relations are among the new words they are presented with, and
between familiar terms and new terms proposed by the experimenter.
Finally, one might suppose that adults would refrain from using two
labels for the same referent in talking to very young children, but this is
not the case. In one study where parents were asked to teach their
infants words for items in three domains, they offered second terms
17% of the time to children with fewer than 50 words of vocabulary in
production (mean words produced = 21, mean age = 16 months), and
did so over 30% of the time to slightly older children with vocabularies
of at least 100 words (mean words produced = 299, mean age 23
months) (Callanan, Sabbagh, Perez & Cervantes 1995). That is, adults
rely on more than one term for the same referent even in speaking to
one-year-olds, but they link the meanings of such terms to words
already known to their young interlocutors.
The speech adults address to young children is consistent with the
general nature of the lexicon and the uses to which it can be put by
speakers. The same referent can be viewed from a variety of
perspectives, and speakers signal their choice in this regard in the
words they use (Clark 1996). Entities in some domains are organized
hierarchically, so terms lower in the hierarchy are included in those
higher up: a speaker can refer to her spaniel on one occasion as the
spaniel, on another as the dog, and on yet others as the animal. The
same entity may be also be referred to variously as the guard-dog, the
barker, the garbage-disposal or the dribbler, depending on which
perspective is relevant to the speaker's conversational goals (Ravn
1988). In short, the lexicon offers great flexibility to speakers by
allowing them to use words with very different senses to refer to the
same entity. Adult speakers expose children to this property of words
from the start.
SUMMARY
The present study looked at young children's ability to use more than
one term for the same referent in two tasks - one where the second
term belonged to the same semantic domain and was hierarchically
related to the first, and one where the second term belonged to a
different domain. Even the youngest children, aged two, had no
difficulty and showed that they could readily deploy two distinct terms
for the same referent. That is, they were able to take more than one
perspective on the same entity and switch from one to another. These
PERSPECTIVE AND REFERENCE
73
data support the many-perspective view of lexical acquisition and offer
evidence against a one-perspective account.
REFERENCES
Akhtar, N. & Tomasello, M. (1996). Twenty-four-month-old children learn words for
absent referents. British Journal of Developmental Psychology, 14, 79-93.
Au, T. K.-F. & Glusman, M. (1990). The principle of mutual exclusivity in word
learning: to honor or not to honor? Child Development, 61, 1474-90.
Berman, R. A. & Clark, E. V. (1989). Learning to use compounds for contrast. First
Language, 9,247-70.
Blewitt, P. (1983). Dog vs collie: vocabulary and speech to young children.
Developmental Psychology, 19, 602-9.
Brown, R. (1958). How shall a thing be called? Psychological Reviewl. 65, 14-21.
Callanan, M. A. (1985). How parents label objects for young children: the role of
input in the acquisition of category hierarchies. Child Development, 56,508-23.
- (1989). Maternal speech strategies and children's acquisition of hierarchical
category labels. Genetic Epistemologist, XVII (2), 3-12.
Callanan, M. A., Sabbagh, M., Perez, D. & Cervantes, C. (1995). Names for objects in
conversations between young children and their parents. Symposium on
'Constructing meanings in context: constraints and pragmatics in semantic
development', biennial meeting of the Society for Research in Child
Development, Indianapolis, IN, March 1995.
Clark, E. V. (1993). The Le-vico~zin Acquisition (Cambridge: Cambridge University
Press).
- (19950). Speaker perspective in lexical acquisition. In S. Gahl, A. Dolbey & C.
Johnson (eds), Proceedings of the 20th Annual Meeting of the Berkeley
Linguistics Society (1994) (Berkeley, CA: Department of Linguistics, University
of California at Berkeley), 125-33.
- (19956). Pragmatics and meaning acquisition. Symposium on 'Constructing
meanings in context: constraints and pragmatics in semantic development,'
biennial meeting of the Society for Research in Child Development, Indianapolis,
IN, March 1995.
- (1996). Conceptual perspective and word choice in acquisition. Unpuiblished
MS, Stanford University.
Clark, E. V., Gelman, S. A. & Lane, N. M. (1985). Noun compounds and category
structure in young children. Child Development, 56, 84-94.
Edwards, R. P. A. & Gibbon, V. (1964). Words your- Children use: A Survey of the
Wol-ds used by Children in Infants' Schools with the Resultant GI-aded
Vocabulary (London: Burke Books).
Gelman, S. A,, Wilcox, S. A. & Clark, E. V. (1989). Conceptual and lexical
hierarchies in young children. Cognitive Development, 4, 309-26.
Grieve, R. (1975). Problems in the study of early semantic development. In G.
Drachman (ed.), Salzburger Beitrage zur Linguistik 11 (Tiibingen: Gunter Narr).
Macnamara, J. (1982). Names for Things: A Study of Human Learning (Cambridge,
MA: Harvard University Press).
Markman, E. M. (1987). How children constrain the possible meanings of words. In
Ulric Neisser (ed.), Concepts and Conceptual Development: Ecological and
Intellectual Factors in Categorization (Cambridge: Cambridge University Press),
255-87.
- (1989). Categor-ization and Naming in Children: Problems of Inductio~z
(Cambridge, MA: MIT PressBradford Books).
Markman, E. M. & Wachtel, G. F. (1988). Children's use of mutual exclusivity to
constrain the meanings of words. Cognitive Psychology, 20, 121-57.
74
FIRST LANGUAGE
Merriman, W. E. (1986). Some reasons for the occurrence and eventual correction of
children's naming errors. Child Development, 57, 942-52.
Merriman, W. E. & Bowman, L. L. (1989). The mutual exclusivity bias in children's
word learning. Monographs of the Society for Research in Child Development, 54
(3-4) (Serial No. 220).
Miller, G. A. (1978). Practical and lexical knowledge. In E. Rosch & B. B. Lloyd
(eds), Cognition and Categorization (Hillsdale, NJ: Lawrence Erlbaum), 305-14.
Ravn, K. (1988). On calling things names. Unpublished PhD dissertation, Stanford
University.
Rinsland, H. D. (1945). A Basic Vocabulary
. o"f Elementary School Children (New
York: Macmillan).
Scarry, R. (1968). What do People do All Day? (New York: Random House).
Shatz, M. (1978). On the develo~mentof communicative understanding: an early
strategy for interpreting and responding to messages. Cognitive ~ s ~ c h o l 10,
o~~,
271-301.
Shipley, E. F., Kuhn, I. F. & Madden, E. C. (19833. Mothers' use of superordinate
category terms. Journal of Child Language, 10, 571-88.
Siegal, M., Waters, L. J. & Dinwiddy, L. S. (1988). Misleading children: causal
attributions for inconsistency under repeated questioning. Journal of
Experimental Child Psychology, 45,438-56.
Snow, C. E. (1972). Mothers' speech to children learning language. Child
Development, 43, 549-65.
Tomasello, M. & Akhtar, N. (1995). Two-year-olds use pragmatic cues to differentiate
reference to objects and actions. Cognitive Development, 10, 201-24.
Tomasello, M. & Barton, M. (1994). Learning words in nonostensive contexts.
Developmental Psychology, 30, 639-50.
Valentine, C. W. (1942). The Psychology of Early Childhood (London: Methuen).
Waxman, S. R. & Hatch, T. (1992). Beyond the basics: preschool children label
objects flexibly at multiple hierarchical levels. Journal of Child Language, 19,
153-66.
Waxman, S. R. & Senghas, A. (1992). Relations among word meanings in early
lexical development. Developmental Psychology, 28, 862-73.
Waxman, S. R., Shipley, E. F. & Shepperson, B. (1991). Establishing new
subcategories: the role of category labels and existing knowledge. Child
Development, 62, 127-38.
1
7
1