Object Shape, Object Function, and Object Name

Page created by Rebecca Lee
 
CONTINUE READING
JOURNAL OF MEMORY AND LANGUAGE       38, 1–27 (1998)
ARTICLE NO.   ML972533

                  Object Shape, Object Function, and Object Name

                                                Barbara Landau
                                              University of Delaware

                                                        and
                                       Linda Smith and Susan Jones
                                                 Indiana University

            We investigated the roles of shape and function in object naming. Two-, three-, and five-year-
          olds and adults heard novel or familiar objects named; some participants also were instructed
          about the objects’ functions. Then they were asked to generalize the names to new objects that
          preserved shape or functional capability; some participants also judged the objects’ potential for
          carrying out the designated function. Children generalized the name by object shape regardless
          of instruction, but adults did so only in the absence of instruction or for familiar objects.
          Knowledge of function independent of naming became increasingly stronger and diverse over
          age. The strong developmental changes in the role of function bear on mechanisms of object
          naming. q 1998 Academic Press

   Common sense tells us that if we know                        function. As Miller and Johnson-Laird (1976)
what an object is we often know what it does.                   argued, function is a critical aspect of catego-
If asked ‘‘What is a chair?’’, one might re-                    rizing most human artifacts, so much so that
spond ‘‘Something you sit on’’; if asked                        ‘‘function—at least for artifacts—is more ba-
‘‘What is a knife?’’, a plausible answer might                  sic to the definition than form’’ (p. 229).
begin ‘‘Something you cut with.’’ Indeed, the                      In this paper, we consider the role of object
importance of function in our knowledge of                      form and function in the early development
named artifact categories is so deeply in-                      of object naming—specifically, the naming of
grained that it often serves as a starting as-                  artifacts. Interest in the relative roles of object
sumption in discussions of the ontogeny and                     appearance and function has a long history in
nature of naming: Assuming that artifact                        both the developmental and adult literatures
names are cover terms for artifact categories,                  on object naming and concept representation
and that function is a critical component of                    (Bloom, 1996; Gentner, 1978; Malt & John-
many such categories, it follows that artifact                  son, 1992), partly because these properties
names should be generalized on the basis of                     tend to be viewed as major criteria for mem-
                                                                bership in artifact categories. In our view,
   This research was supported by NICHD Grant PHS               however, the contrast is an especially interest-
RO1 HD-28675, NIMH Grant PHS RO1 MH-55240, and                  ing one because sensitivity to the properties
Social and Behavioral Sciences Research Grant 12-FY93-          themselves stems from different sources, their
0723 from the March of Dimes. We thank C. E. Wright
for advice on statistical matters, the University Montes-
                                                                relationship is complex, and there may be de-
sori, Irvine Child Development Center, and the University       velopmental changes between them that yield
of Delaware Preschool for cooperating with us, and Diane        insight into the development and mechanisms
Beck, Jamie Chosak, Jennifer Nolan, and Melissa                 of object naming. Although function might be
Schweisguth for help in conducting the studies.                 more important than form in adults’ lexical
   Address reprint requests and correspondence to Barbara
Landau, Department of Psychology, Wolf Hall, Univer-
                                                                categorization of artifacts, it may not be in
sity of Delaware, Newark, Delaware 19716. E-mail:               children’s. Further, tracking possible develop-
blandau@udel.edu.                                               mental changes in form and function in nam-
                                                            1                     0749-596X/98 $25.00
                                                                                  Copyright q 1998 by Academic Press
                                                                                  All rights of reproduction in any form reserved.

AID     JML 2533          /    a010h$$$81         12-17-97 08:40:26              jmla           AP: JML
2                                LANDAU, SMITH, AND JONES

ing may yield new insights into ontogenetic       Landau, Smith, & Jones, 1992; Smith, Jones, &
process. Information about form originates        Landau, 1992). In these studies, adults show
principally through the visual system, and        an extreme bias: They reject objects that vary
function through an understanding of the          even a small amount in shape while accepting
physical principles (such as gravity or force)    objects that are radically different from the
involved in the interactions of people with ob-   standard in either texture or size.
jects and objects with each other.                   The existence of a shape bias in no way
   One hypothesis about the relationship be-      implies that it is the only mechanism at work
tween these sources of information is that the    during object naming. On the contrary, the
early development of object naming princi-        bias is highly context dependent, occuring in
pally engages the perceptual systems, and only    the specific context of object naming, but not
secondarily the systems of general world-         in nonnaming contexts that also require judg-
knowledge about objects. Because of this,         ments of similarity (see Jones & Smith, 1993;
early object naming may be relatively ‘‘im-       Smith, 1994, for reviews). Furthermore, in
mune’’ to the influences of young children’s      contexts where similarity of shape is pitted
knowledge about objects, including for arti-      directly against taxonomic category member-
facts an object’s functions (Smith, Jones, &      ship (an unusual situation in the real world),
Landau, 1996). We ask whether this early im-      the importance of shape declines with age (be-
munity occurs only in naming tasks, whether       tween three and five) and with familiarity of
it occurs over a variety of object types, and     category (Baldwin, 1992; Imai, Gentner, &
whether and how lexical categorization comes      Uchida, 1994). Thus, while shape is not the
to incorporate information about function.        only information that children use to establish
   In the following sections, we review evi-      lexical membership, it does seem to enjoy a
dence supporting the importance of shape in       special status during children’s first encoun-
early object naming, the importance of func-      ters with newly named objects.
tion in mature object naming, and ways of            The contextual dependence of the shape
evaluating the relative importance of these       bias and its decline in the context of familiar
over development.                                 objects suggests that an emphasis on shape is
                                                  modifiable when alternative bases for general-
The Importance of Shape in Early Object           ization become known. Object function pres-
  Naming                                          ents itself as an obvious source of information
   In a variety of studies, we have shown that    which might critically influence lexical cate-
young children and adults tend to generalize      gory formation.
a novel artifact’s name to other objects that
are similar to the exemplar object in shape       The Importance of Function in Mature
rather than texture, substance, or size. The        Object Naming
findings of preference for same-shape objects        There is considerable evidence consistent
are consistent with the findings on the taxo-     with the idea that functional information is a
nomic basis for categorization in young chil-     critical, even core, aspect of our mature named
dren and infants (Markman & Hutchinson,           object categories. Rosch, Mervis, Gray, and
1984; Waxman, 1995), as many objects in the       Boyes-Braem (1976) found that basic level
same taxonomic category do share overall          objects share a high degree of overlap in func-
shape (Imai, Gentner, & Uchida, 1994; Rosch       tion. Malt and Johnson (1992) showed that
et al., 1976) and part structure (Tversky,        object function—while not sufficient—can
1989). The preference for same shape in ob-       play an important role in adults’ decisions
ject naming appears in fragile form by the age    about category membership for novel arti-
of 2, becomes stronger by the age of 3, and       facts, for example, their judgments of when
is very strong in 5-year-olds (Jones, Smith, &    an object can legitimately be called a ‘‘boat.’’
Landau, 1991; Landau, Smith, & Jones, 1988;       Finally, Barsalou (1989) argued that adults

AID    JML 2533      /   a010h$$$82      12-17-97 08:40:26        jmla      AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                     3

readily create novel ‘‘ad hoc’’ categories          ‘‘hat’’ to other objects—such as keys—as he
based on function, e.g., ‘‘things to take on a      places them on his head. Nelson suggested
picnic.’’                                           that once such categories are formed on the
   Evidence also shows that school-aged chil-       basis of action, the corresponding category la-
dren tend to rely on an object’s function in        bels may be generalized on the basis of per-
making lexical category assignments (Keil,          ceptual properties such as form. Thus function
1989; Merriman, Scott, & Marazita, 1993;            would serve as the core of the infant’s con-
Richards, Goldfarb, Richards, & Hassen,             cept, while perceptual similarity would follow
1989). Keil found that such older children          as the basis for generalizing the object’s name.
would select as a ‘‘hammer’’ an object that            Knowledge of object function is indeed ac-
was heavy enough to pound objects rather than       quired before or during the earliest object
one that had the overall appearance of a ham-       naming, making plausible the idea that func-
mer but could not carry out the standard flat-      tion might be the core for object naming. In
tening function due to a hole placed strategi-      habituation tasks, Kolstad and Baillargeon
cally in the hammer’s head. Richards et al.         (1991) found that 10- and 12-month-old in-
(1989) found that fourth to sixth graders and       fants who were given experience with a func-
adults judged object category membership on         tional property (bottoms that serve to contain
the basis of function rather than form. For         a substance) then generalized on the basis of
example, they were unlikely to accept an ob-        that property rather than overall similarity of
ject that was ‘‘just like a fork but had no         appearance. Madole and Cohen (1992) found
tines’’ as ‘‘still a fork.’’                        that 14-month-olds were sensitive to form–
   Might object function also serve as a critical   function correlations: After habituating to ob-
basis for naming during the early years of          jects exhibiting a given form together with a
word learning? One framework, which we will         given function (carried out by particular object
call ‘‘action-based,’’ sees the foundation for      parts), they dishabituated to objects pos-
categories as the product of sensory and motor      sessing novel form–function combinations.
schemas that develop as the infant grasps, ma-         In nonhabituation tasks, there are similar
nipulates, and otherwise discovers an object’s      findings. Brown (1990) reported that as early
properties through action. According to Piaget      as 18 months, children will select an object to
(1954), the infant’s sensori-motor activities       perform a function on the basis of its function-
become organized over the first year of life        ally relevant properties rather than other sa-
as the infant discovers enduring properties of      lient features. In one experiment, children
objects such as their shape, size, and mallea-      were enticed to retrieve a toy from some dis-
bility. Discovering the physical effects of ob-     tance away using a long rigid cane. Then they
jects on each other constitutes some of the         were given a choice of several objects to use
infant’s earliest knowledge, and could thus         as a tool. The children chose objects that pos-
serve as a first step in the acquisition of names   sessed both sufficient length and rigidity for
for objects.                                        the task in preference to objects that looked
   In this vein, Nelson (1974) argued that ob-      more like the standard that had previously
ject categories underlying early words are ini-     been shown to work. Brown argued that chil-
tially formed on the basis of action—what the       dren bring to this problem-solving task an un-
infant does with an object (see also Werner &       derstanding of what is relevantly ‘‘similar’’
Kaplan, 1963). These functional categories are      to the standard: here, length and rigidity, not
the prelinguistic categories that serve as the      surface texture or overall shape.
basis for first words, which are often words           These ideas about the early importance of
for moveable and manipulable objects. Nelson        function are also consistent with current ‘‘the-
gave the example of the child who first utters      ory-driven’’ views of word learning, which
the word ‘‘hat’’ in the context of placing it on    show that even young children appreciate the
his head. Subsequently, he might generalize         deeper properties underlying an object’s

AID    JML 2533       /   a010h$$$82       12-17-97 08:40:26        jmla      AP: JML
4                                   LANDAU, SMITH, AND JONES

named category and that this understanding            shape-based generalization among 2- to 5-
allows them to bypass superficial (often ap-          year-olds and adults (see Jones & Smith,
pearance-based) similarities (e.g., Gelman,           1993; Landau, 1994, for review). In Gentner’s
1988). Of particular relevance is a study by          study, the only indication of function-based
McCarrell and Callanan (1995) which showed            responses in adults was their spontaneous ten-
that 2- and 4-year-olds were sensitive to corre-      dency to call the hybrid a ‘‘jiggy-zimbo’’
spondences between an animal’s form and               upon first seeing it. Their use of the function-
its ‘‘function’’ (here, a behavior such as            based ‘‘zimbo’’ as the head noun suggests that
‘‘stretches tall to eat cherries’’). This sensitiv-   they thought the object was a member of
ity was sufficient to promote inductive infer-        the ‘‘zimbo’’ category, but one with ‘‘jiggy’’
ences about the behavior of new category              properties (Lyons, 1977).
members, suggesting that knowledge of func-              Other studies also have found a dominance
tion may play a critical role in tasks of catego-     for an object’s overall appearance compared
rization other than naming. For artifacts, func-      to its function among young children and have
tion may be the key conceptual property (see          suggested a developmental increase in the im-
Keil, 1989) and thus within this general view         portance of function. Tomikawa and Dodd
of word learning, an emphasis on function             (1982) found that 2- and 3-year-olds more
over shape might be expected in naming as             readily learned names for perceptually similar
well as other kinds of categorization.                objects (that had different functions) than for
                                                      objects sharing a common function (but hav-
                                                      ing different shapes). Merriman (1993) found
Evaluating the Importance of Shape and                an increase between ages 3 and 6 in the ten-
  Function in Early Object Naming                     dency to generalize a novel name on the basis
   Previous studies on naming suggest that the        of function (as opposed to appearance), but
two different kinds of information may inter-         even in 6-year-olds function was weighted
act differently at different points in develop-       only slightly more strongly than appearance.
ment. In one of the earliest studies pitting ap-      Although there are no comparable studies of
pearance against function, Gentner (1978) in-         novel object naming among older children and
troduced children and adults to two different         adults, related findings suggest that at these
complex novel objects each possessing a very          ages function may often prevail in the naming
salient function (a lever dispensed jellybeans        of familiar objects (Richards et al., 1989;
for the ‘‘jiggy’’; an identical lever moved           Rosch et al., 1976).
parts of a face for the ‘‘zimbo’’). Participants         These findings suggest that overall appear-
were encouraged to manipulate each object             ance, primarily similarity of shape, dominates
repeatedly, viewing the function; during this,        function in early object naming. However,
each object was labeled. Later, participants          there is an alternative interpretation. It is pos-
were shown a hybrid object which possessed            sible that, given the particular stimuli used,
the form of the zimbo but produced jellybeans         the appearances (shapes) of the objects were
like the jiggy. They were asked whether it            much more salient overall than the properties
should be called a ‘‘jiggy’’ (following the ob-       supporting the functions. If so, children might
ject’s overall appearance) or a ‘‘zimbo’’ (fol-       have generalized on the basis of object shape
lowing its function).                                 no matter what kind of task they were given.
   Gentner found a U-shaped function: 2- to           To argue that similarity of shape not only
5-year-olds and adults chose strongly on the          dominates object function, but specifically
basis of appearance, but 5- to 15-year-olds           does so in the context of object naming, one
chose more than half the time on the basis of         would need a test that discovers whether chil-
function. The strong reliance on appearance           dren are sensitive to the objects’ functional
among 2- to 5-year-olds and adults is consis-         properties and then a measure of whether this
tent with our previous findings showing robust        sensitivity enters into the object naming task.

AID    JML 2533        /   a010h$$$82        12-17-97 08:40:26         jmla      AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                       5

   A recent set of experiments assessed this         knowledge of the objects’ functions. We did
possibility. Smith, Jones, and Landau (1996)         this using simple novel artifacts with simple
showed 3-year-olds and adults highly complex         novel functions (Experiment 1). Keeping in
novel artifacts built of a distinctive base object   mind that the findings may vary widely with
with salient appended parts. In a Name task,         object and function type, we then proceeded
participants were asked whether test objects         to ask the same questions using functions
had the same name as the exemplar. In a Simi-        which are known to be accessible to very
larity task, participants were asked whether         young children (Experiment 2). Finally, be-
test objects were ‘‘the same as’’ the exemplar.      cause the form and function relationship in
Half of the participants were also shown a           naming may be different for familiar than for
function for either the base object or the sa-       novel objects (Merriman, 1993), we investi-
lient parts. Note that both tasks—Name and           gated these roles in children’s and adults’
Similarity—required participants to make             knowledge of familiar everyday objects (Ex-
judgments on the basis of some kind of simi-         periment 3).
larity. The question was whether the basis
would be the same for both tasks.                                  EXPERIMENT 1
   Results showed that adults’ judgments in             We tested children’s and adults’ attention to
the Name and Similarity tasks were influenced        shape over material in object naming (Name
by the functional information. Three-year-           task), where the object’s function could be
olds’ judgments in the Similarity task were          carried out by virtue of its material. All parti-
also influenced by the functions. However,           cipants heard a novel object named. Half of
children’s judgments in the Name task were           them were explicitly shown a material-based
immune to influence from information about           function and half were not, and then all were
function. Instead, their judgments were influ-       asked to extend the standard object’s name
enced by changes in the relative perceptual          to new objects. Those who were shown the
salience of the base vs the parts. These results     function were also directly asked whether
led Smith et al. to suggest that early object        each object could be used to carry out the
naming may in some way be so tightly tied            designated function (Function task). Our ques-
to certain aspects of perception that it is cut      tion was whether the instructions about object
off from other influences of world knowledge         function could influence participants’ atten-
that the child has.                                  tion to shape vs material, and whether the pat-
   These results suggest a strong hypothesis         tern of responses in the Name task would dif-
for the development of object naming. Spe-           fer from those of the Function task.
cifically, knowledge of function—while criti-
cal for the formation and use of mature object       Method
representations that underlie naming—may                Participants. Seventy-two participants were
play a quite minor role in early object naming.      tested, including twenty-four 3- and 5-year-
Further, the results suggest that immunity to        olds and adults, with each age group balanced
functional knowledge should be seen earlier          for gender. Mean ages of the children were 3;
than age 3, but may begin to change with             5 and 5; 4 (Ranges Å 3; 0–3; 10, 4; 11–5;
development after this point.                        11). Children were drawn from preschools in
   The purpose of the present experiments was        the area around Irvine, California, and adults
to begin to test this account, using a wider         were undergraduates at the University of Cali-
developmental window and a broader range             fornia, Irvine, who participated for class
of objects and functions than those used by          credit. Half of each age group was randomly
Smith et al. (1996). We began by challenging         assigned to each of the two experimental con-
children’s and adults’ attention to object shape     ditions described below.
in a naming task and comparing it to their              Design and materials. In the No-Function
responses in a task that directly assessed their     condition, participants were shown the stan-

AID    JML 2533        /   a010h$$$83       12-17-97 08:40:26         jmla      AP: JML
6                                            LANDAU, SMITH, AND JONES

                                                                  and therefore provide a comparison point to
                                                                  previous findings of a shape bias in the context
                                                                  of naming. However, the materials were novel
                                                                  and designed specifically to support clear
                                                                  functions.
                                                                     The test set for each standard included nine
                                                                  objects: two subsets of four plus a replica of
                                                                  the standard. Each subset contained two
                                                                  Same-shape and two Same-material objects.
                                                                  The Same-material objects were the same ma-
                                                                  terial as the standard but different shapes; and
                                                                  the Same-shape objects were the same shape
                                                                  as the standard but were made of different
                                                                  materials—ones that could not support the
                                                                  standard’s function. In the Rif set, the novel
                                                                  materials were nonporous, either styrofoam or
                                                                  a tightly woven, glossy fabric covering a thick
                                                                  soft pad. In the Dax set, these materials were
                                                                  hard and impenetrable, either brick-shaped
                                                                  Lego parts assembled into the U shape or plas-
                                                                  ter painted to a high gloss. In both the Rif
                                                                  and Dax sets, each Same-shape object was
                                                                  matched in color to one Same-material object.
                                                                     Procedure. Participants in the No-Function
   FIG. 1. Objects used in Experiment 1. The ‘‘standard’’         condition were shown one of the standards
Rif and Dax are shown at the left. Test objects included
two Same-shape (different material) and two Same-mate-
                                                                  and told, ‘‘See this? This is a dax.’’ Then the
rial (different shape) objects (see text for full description).   standard was placed to the side, still in full
                                                                  view, and each of the nine test objects was
                                                                  removed from a box one at a time and shown
dard object and heard it labeled, but no in-                      to the participant, who was asked ‘‘Is this a
formation about function was provided. In                         rif/dax?’’ (Name question). After he or she
the Function condition, participants saw the                      answered yes or no to each object, it was put
standard and heard it labeled, but they were                      back in the box and the next object was
also told and shown what the standard was                         brought out. Each object was queried twice,
used for.                                                         followed by the second standard and test set,
   Two sets of objects were constructed (see                      for a total of 18 test trials per set and 36 in
Fig. 1). The standard for the Rif set was a                       total.
roughly 2’’ 1 2’’ roundish object made of                            Participants in the Function condition un-
pale yellow sponge used in standard sink                          derwent the same procedure, with the follow-
sponges. The Rif’s function was to wipe up                        ing exception. They were shown the standard
water. The standard for the Dax set was a 2’’                     and told, ‘‘See this? This is a rif. Rifs are
1 2’’ squarish U-shaped object made of dark                       made by a special company so they can do
brown cork, similar to that used for bulletin                     this (demonstrate function). See? So rifs are
boards. The Dax’s function was to support                         made so you can mop up water with them.’’
stick pins of the type used to hold messages                      Exactly parallel instructions were given for
on cork boards. The shapes of these objects                       the Dax standard. To demonstrate the function
were the same as those used in previous stud-                     for the Rif set, the experimenter took a small
ies of the shape bias (e.g., Jones, Smith, &                      bottle of water, spilled some on the table, and
Landau, 1991; Landau, Smith, & Jones, 1988),                      mopped up the puddle with the object, saying

AID      JML 2533            /    a010h$$$83            12-17-97 08:40:26         jmla      AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                       7

‘‘I can mop up water with it.’’ The Rif was           The reliability of these findings was con-
then set next to the bottle of water, where it     firmed by a mixed analysis of variance con-
remained during the test trials. To demonstrate    ducted on the numbers of ‘‘yes’’ responses,
the function for the Dax set, the experimenter     with age and condition as between-subjects
took a push pin and stuck it into the surface      factors and property (Same shape/Same mate-
of the object, saying ‘‘I can stick pins into it   rial) the within-subjects factor. This analysis
and they’ll hold.’’ The object with the embed-     revealed a main effect of property, F(1,66) Å
ded pin was left visible during the test trials.   35.04, p õ .01, and interactions between age
   After the entire naming procedure was com-      and property, F(2,66) Å 18.52, p õ .01, and
plete, participants in the Function condition      condition and property, F(1,66) Å 58.15, p
were asked directly whether they thought each      õ .01, which were subsumed by a three-way
of the test objects could support the designated   interaction among age, condition, and prop-
function of the relevant standard (Function        erty, F(2,66) Å 11.86, p õ .01. Planned com-
question). Participants were directed to look      parisons within age showed reliable prefer-
at the standard and were asked ‘‘Do you re-        ences for Same-shape over Same-material ob-
member what I did with this one?’’ If they         jects for the 3’s in both the No-Function and
did not respond promptly, the experimenter         Function conditions, t’s (66) Å 5.91, 4.04, re-
reminded them, saying ‘‘I mopped up water          spectively, p’s õ .05, and for the 5’s and
with it (stuck a pin in it so it would hold).’’    adults in the No Function condition, t’s (66)
Then they were asked either ‘‘Could you mop        Å 5.78, 4.79, respectively, p’s õ .05. The
up water with this one?’’ (for the Rif set) or     adults in the Function condition showed a reli-
‘‘Could you stick a pin in this one so it would    able preference for material over shape, t(66)
hold?’’ (for the Dax set). In this part of the     Å 06.66, p õ .05. Planned comparisons were
procedure, each test object was queried just       also carried out within age to compare shape
once. The test items were presented in one of      and material responses across No-Function
two randomized orders, and the presentation        and Function conditions. Only the adults
order of object sets was counterbalanced over      showed a reliable difference between the two
participants.                                      conditions, although the fives approached sig-
                                                   nificance, t’s (66) Å .66, 1.86, 4.06 for the
Results and Discussion                             3’s, 5’s, and adults, critical t Å 1.99, p õ .05.
   Table 1 shows the mean proportions of              An examination of individual performances
‘‘yes’’ responses to the Name question for         provides greater insight into the develop-
Same-shape and Same-material objects over          mental trend. Figure 2 shows the scatterplots
age and condition (No Function and Func-           of individual performances—the number of
tion). The table also shows the results for the    ‘‘yes’’ responses to Same-shape vs Same-ma-
Function question (Function condition only).       terial responses for each individual over age
Preliminary analyses revealed no differences       and condition. The left panel shows the data
between object sets, so the results for both       for the Name question. Each dot represents
questions were collapsed across object set         one child’s performance. Individuals who said
(Dax vs Rif).                                      ‘‘yes’’ to more Same-shape than Same-mate-
   Name question. As is apparent, 3-year-olds      rial objects fall in the lower half of the figure,
generalized the name strongly on the basis         whereas those who did the reverse fall in the
of shape in both No-Function and Function          upper half. Individuals who accepted an equal
conditions. Adults, in contrast, generalized on    number of the two object types fall along the
the basis of shape in the No-Function condi-       diagonal.
tion but material in the Function condition.          Three-year-olds in both conditions and 5-
Five-year-olds generalized strongly on the ba-     year-olds and adults in the No Function condi-
sis of shape in No-Function condition and          tion accepted more Same-shape than Same-
weakly in the Function condition.                  material objects. The dominance of shape over

AID    JML 2533       /   a010h$$$83      12-17-97 08:40:26         jmla      AP: JML
8                                            LANDAU, SMITH, AND JONES

                                                     TABLE 1
        Mean Proportions (SE) of ‘‘Yes’’ Responses to Same-Shape and Same Material Objects (Experiment 1)

                               No function                                        Function

                          Name question                     Name question                        Function question
Condition,a
 question,            Same               Same           Same            Same                  Same             Same
 and age              shape             material        shape          material               shape           material

     3 yr            .92 (.04)          .18 (.06)      .76 (.07)      .25 (.08)              .38 (.11)       .58 (.12)
     5 yr            .82 (.08)          .09 (.05)      .41 (.10)      .34 (.08)              .04 (.04)       .78 (.10)
    Adults           .69 (.09)          .09 (.05)      .06 (.03)      .89 (.06)              .06 (.03)       .93 (.05)

    a
   n Å 12 in each age group and condition. Participants in the No-Function condition answered only the Name
question; participants in the Function condition answered both the Name and Function questions.

material among 3’s was reliable in both condi-              whether the test objects could carry out the
tions, n Å 12 vs 0 and 10 vs 0, as it was among             designated functions. Results from this ques-
5’s and adults in the No-Function condition, n              tion are also shown in Table 1. The first issue
Å 10 vs 0 and 9 vs 1, respectively, p’s õ .05,              was whether the Function question elicited the
Binomial tests (assuming a random probabil-                 same generalization pattern as the Name ques-
ity of .50 in this and all further Binomials).              tion. We answered this by directly comparing
Adults also showed a reliable dominance of                  responses to the Name and Function questions
material over shape in the Function condition,              among participants who had been instructed
n Å 11 vs 0, p õ .05, Binomial test, and the                about function (the Function condition). Pre-
5’s showed no preference in that condition,                 liminary analyses revealed no effects of object
n Å 4 shape, 4 material, 4 equivocal. The                   set, so the data were collapsed over this factor.
preference for Same-shape objects differed re-              The mean proportions of ‘‘yes’’ responses to
liably across the two conditions for the 5-year-            Same-shape and Same-material objects were
olds and adults, p’s õ .05, Fisher Exact tests.             submitted to a mixed analysis of variance with
Thus, the analysis of individuals was consis-               age the between-subjects factor and property
tent with the group analysis.                               and question type (Name vs Function) as
   To summarize the results of the Name ques-               within-subjects factors. The results showed a
tion, when a novel object was introduced and                main effect of property, F(1,33) Å 49.10, p
named with no further information, young                    õ .01, and interactions between age and prop-
children and adults relied on shape as the basis            erty, F(2,33) Å 34.93, p õ .01, and question
for generalizing the name. When competing                   type and property, F(1,33) Å 38.57, p õ .01,
information about the functional importance                 both subsumed by a three-way interaction
of material was provided, adults switched                   among age, question type, and property,
from naming by shape to naming by material.                 F(2,33) Å 8.46, p õ .01. Planned comparisons
The 5-year-olds were affected by instructional              revealed several reliable differences across
condition, with a different pattern of prefer-              question type: The 3- and 5-year-olds pre-
ences when given functional instruction. Un-                ferred shape when asked the Name question
like adults, however, the 5-year-olds did not               but not when asked the Function question; the
reliably shift to material when told the objects’           adults showed no such difference, as they pre-
functions.                                                  ferred material when asked either question, t’s
   Name vs function question. Participants                  (33) Å 2.5, 2.86, .14, p’s õ .05. Thus the two
who were given instructions about the objects’              questions elicited different responses in 3- and
functions were also queried directly about                  5-year-olds, but not adults.

AID       JML 2533         /      a010h$$$83        12-17-97 08:40:26        jmla             AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                                   9

      FIG. 2. Individual participants’ choices of Same-shape vs Same-material objects for the Name Question
    and the Function question in Experiment 1. Closed circles represent individuals who were not instructed
    about function (No-Function condition) and were asked the Name question only. Open circles represent
    individuals who were instructed about function and were asked both the Name and Function questions
    (Function condition). Individuals who chose on the basis of shape fall into the lower half of the plot whereas
    those who chose on the basis of material fall into the upper half.

   The second issue was whether children un-                and adults reliably preferred material, t’s (33)
derstood the standard objects’ functions,                   Å 05.3, 06.2, respectively, p’s õ .05. Thus,
whether they generalized correctly in the                   although both the 3’s and 5’s showed different
Function question (i.e., on the basis of mate-              response patterns in the Name vs Function
rial), and most important, whether this under-              questions, only the 5’s and adults reliably re-
standing affected their generalizations in the              sponded reliably ‘‘correctly’’ to the Function
Name question. Planned comparisons for the                  question—that is, on the basis of material.
Function question showed that 5-year-olds                   Comparisons for the Name question showed

AID    JML 2533          /    a010h$$$84          12-17-97 08:40:26              jmla        AP: JML
10                                 LANDAU, SMITH, AND JONES

that 3-year-olds reliably preferred shape and       quadrant of the right panels in Fig. 2). This
that adults reliably preferred material, t’s (33)   group included four 3-year-olds and nine 5-
Å 3.6, 05.9, respectively, p’s õ .05, just as       year-olds. The second group included the re-
in the previous analysis of the Name question.      maining 11 children. Given this division, we
   Individual participants’ pattern of shape vs     analyzed the number of shape vs material-
material responses for the two questions con-       based responses in the Name question, using
firmed these patterns (see Fig. 2, left and right   an analysis of variance with age and group
panels). Three-year-olds reliably preferred         as the between-subjects factors and property
shape when asked the Name question, n Å 10          (shape, material) as the within-subjects factor.
shape, 0 material preference, p õ .01, Bino-        There was a main effect of property only, with
mial test, but showed no reliable preference        responses by shape reliably greater than by
when asked the Function question, n Å 7 ma-         material, M’s Å .58, .29, respectively, F(1,20)
terial, 3 shape preference, p Å .17, Binomial       Å 5.5, p õ .05. There were no interactions.
test. Seven of the 3-year-olds switched from           These results suggest that object function
a shape preference when asked the Name              may not enter into object naming early in de-
question to a material preference when asked        velopment, even when children understand the
the Function question; one did the reverse,         object properties related to function. First,
confirming a reliable difference in their re-       there were no reliable effects of instructional
sponses across the two questions, p Å .03,          condition on the 3-year-olds’ responses to the
Binomial test. Five-year-olds showed no pref-       Name question and weak effects on the 5-
erence when asked the Name question, n Å 4          year-olds’ responses; only adults were reliably
shape, 4 material, but reliably preferred mate-
                                                    and strongly affected, generalizing names dif-
rial when asked the Function question, n Å
                                                    ferently when given instruction about func-
11 material, 0 shape, p õ .01, Binomial test.
                                                    tion. Second, even children who generalized
Three of the 5-year-olds switched from a
                                                    on the basis of material in the Function ques-
shape preference in the Name question to a
                                                    tion still generalized by shape in the Name
material preference in the Function question;
                                                    question. Thus, although some children under-
none did the reverse. Adults reliably preferred
                                                    stood that material was critical to carrying out
material for both questions, n Å 11 material,
                                                    the function, this knowledge did not enter into
0 shape for the Name question, n Å 12 mate-
                                                    their naming responses.
rial, 0 shape for the Function question, p’s õ
.01, Binomial test. Thus, although there were          It is possible, however, that functional
reliable differences across question type for       knowledge did not affect naming because the
both the 3- and 5-year-olds, the nature of the      objects and functions used in Experiment 1
change was different. Three-year-olds showed        were quite novel and perhaps not part of natu-
shape-based responses when asked the Name           ral, early developing systems of knowledge.
question, and diffuse responses when asked          Only the adults may have truly understood the
the Function question. Five-year-olds showed        functions, which were surely familiar to them.
divided responses when asked the Name ques-         Accordingly, in Experiment 2 we asked the
tion and material-based responses when asked        Name and Function questions using objects
the Function question.                              whose functions should be understood by age
   Did the children’s varying understanding of      2: retrieval of one object by another (Brown,
the objects’ functions affect their responses to    1990) and containment (Kolstad & Baillar-
the Name question? To address this, we di-          geon, 1991). If the results of Experiment 1
vided the children into two groups. One group       were due to the unusual or complex nature of
responded correctly to the Function question,       the functions we devised, then more familiar
specifically, accepting greater than 50% of the     functions might be more likely to enter into
time on material and less than 50% of the time      naming for 3- and 5-year-olds and perhaps
on shape (locating them in the upper left-hand      even for younger children. To test this possi-

AID    JML 2533       /   a010h$$$84       12-17-97 08:40:26        jmla      AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                      11

bility, we extended our study downward             function: One had a large hole in the bottom,
in age.                                            a second was perforated with holes throughout
                                                   the surface, and the third was made of flimsy
              EXPERIMENT 2                         tissue paper. The object with the hole in the
                                                   bottom violated the containment function in a
Method
                                                   way similar to the containers used by Kolstad
   Participants. Ninety-six participants were      and Baillargeon (1991). Each of the Same-
tested, 24 each of 2-year-olds (M age 2;6,         function objects was a different shape from
range 24–35 months), 3-year-olds (M age 3;7,       the standard. In each set, each Same-shape
range 3; –3;11), 5-year-olds (M age 5;5, range     object was matched in color to one Same-
5;1–5;11), and adults. Participants were           function object.
drawn from the same populations as in Experi-         Procedures were identical to those of Ex-
ment 1 and age groups were balanced for            periment 1, except for appropriate modifica-
gender.                                            tion of the Function condition instructions for
   Design, materials, and procedure. The           the new object sets. In addition, the instruc-
overall design was identical to Experiment 1,      tions were simplified a bit to ensure that even
with half of the participants in each age group    the youngest children would understand them.
randomly assigned to the No-Function condi-        For the Cane set, participants were told ‘‘See
tion and half to the Function condition. How-      this? This is a dax (rif). And this is what I
ever, all participants were asked both the         can do with it. I can pull toys with it.’’ The
Name question and the Function question.           experimenter took a small toy and placed it at
   There were two sets of objects (see Fig. 3).
                                                   the far end of a table, several feet away, then
The standard for the Cane set was a 6’’ long
                                                   hooked it with the head of the cane and pulled
blue cane made of hardened clay. The function
                                                   it toward herself. For the Container set, the
of this standard was to retrieve small toys from
                                                   experimenter introduced the object in the
several feet away by hooking and pulling
                                                   same way, then poured a small amount of wa-
them. The test items included a replica of the
                                                   ter into the standard and lifted the container,
standard plus six additional items. Three of
                                                   moving it along a path and saying ‘‘I can carry
these were the Same-shape as the standard,
                                                   water with it.’’
but could not support the designated function:
One was too short (2’’) and the remaining             Following this introduction, participants
two were made of either lightweight fabric or      were tested as in Experiment 1 on the Name
beads strung on a cotton thread, each too          question (‘‘Is this a dax?’’) with two trials for
flimsy to secure and pull small toys. These        each of six objects per set, a total of 24 trials.
objects violated the relevant function in ways     After all items in both sets were queried, parti-
similar to those in Brown’s (1990) study. The      cipants were asked the Function question.
remaining three test items could support the       They were shown the standard once more.
Same-function as the standard, being the same      Those hearing Function instructions were
length and material. However, their handle         asked ‘‘Do you remember what I did with this
pieces and heads were shaped differently.          one?’’ In most cases, participants answered
   The standard for the Container set was a        correctly. If they did not respond spontane-
2’’ 1 2’’ 1 2’’ trapezoidal-type container         ously, the experimenter prompted by saying
made of hardened clay. It stood on short legs      ‘‘I pulled the toy (carried water) with it.’’ All
and was completely open at the top. Its desig-     participants then were asked for each test item,
nated function was to carry water. The test        ‘‘Could you pull a toy from over there with
items included a replica of the standard plus      this?’’ or ‘‘Could you carry water with this?’’
three Same-shape and three Same-function ob-       Note that although the No-Function partici-
jects. The Same-shape objects were identical       pants were not given information about the
in shape but could not support the designated      standard’s function, they could plausibly

AID    JML 2533       /   a010h$$$84      12-17-97 08:40:26         jmla      AP: JML
12                                       LANDAU, SMITH, AND JONES

        FIG. 3. Objects used in Experiment 2. The ‘‘standard’’ for each set is shown at the left. Test objects
     included three Same-shape (different function) and three Same-function (different shape) objects. The
     specified function for the Cane set was to pull toys from across a table. The function for the Container set
     was to carry water. The Same-shape objects for the Cane set violated the specified function by being too
     short or too flimsy; those for the Container set violated function by possessing holes in the surface or being
     composed of thin paper.

AID     JML 2533          /    a010h$2533          12-17-97 08:40:26              jmla        AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                     13

judge whether or not each of the test items        tions preferred Same-function over Same-
could carry out the specified function.            shape objects, Bonferroni critical d Å .24,
  All test items were presented in one of two      p Å .05. The adult pattern in the Name ques-
randomized orders, and the presentation order      tion was responsible for the only reliable pair-
of the Cane or Container set was counterbal-       wise effects of condition: They preferred
anced.                                             Same-shape objects in the No-Function condi-
                                                   tion, and Same-function objects in the Func-
Results and Discussion                             tion condition, Bonferroni critical d Å .24,
   In Table 2 are shown the mean proportions       p Å .05.
of ‘‘yes’’ responses to Same-shape and Same-          To assess the differential effects of question
function objects over age and condition for        type among children, we conducted a separate
the Name question (‘‘Is this a dax?’’) and the     four-way analysis of variance (age, condition,
Function question (‘‘Could you pull a toy/         question type, and property) on the children’s
carry water with this?’’). When asked the          data only. The relevant means, shown in Table
Name question, participants of all ages pre-       2, indicate differences between the two ques-
ferred the Same-shape objects, except for the      tions among 3- and 5-year-olds. This was re-
adults in the Function condition, who general-     flected in the principal result from the analy-
ized on the basis of function. When asked          sis, a three-way interaction among age, ques-
the Function question, 3- and 5-year-olds and      tion type, and property, F(2,66) Å 9.56, p õ
adults preferred the Same-function objects         .05. This interaction subsumed all other ef-
whereas 2-year-olds showed a slight prefer-        fects, which included main effects of age,
ence for Same-shape objects. Thus there were       F(2,66) Å 7.3, p õ .01, question type, F(1,66)
qualitatively different patterns of response for   Å 12.39, p õ .01, and property, F(1,66) Å
the two questions, and within each question,       10.35, p õ .01, two-way interactions between
the developmental pattern was different.           age and question type, F(2,66) Å 4.28, p õ
   Preliminary analyses suggested that there       .05, age and property, F(2,66) Å 3.74, p õ
were different interactions involving object       .05, and question type and property, F(1,66)
set for the Name and the Function questions.       Å 119.83, p õ .01. There was no effect of or
We examined each of these interactions and         interactions with condition.
found that none of them was relevant to or            Planned comparisons evaluated the three-
mitigated the results relevant to the principal    way interaction by assessing the children’s
hypotheses. Therefore, we collapsed over ob-       preferences across question type. The 3- and
ject set for further analyses. Preliminary anal-   5-year-olds both accepted more Same-shape
yses also indicated a four-way interaction         objects when asked the Name question than
among age, question type, condition, and           the Function question, t’s (66) Å 3.62, 3.0,
property, F(3,88) Å 4.03, p õ .01, reflecting      p’s õ .05, and more Same-function objects
the qualitatively different patterns of re-        when asked the Function question than the
sponse for the two questions.                      Name question, t’s (66) Å 03.25, 06.75, p’s
   Planned comparisons were conducted to as-       õ.05. Two-year-olds did not accept same-
sess these patterns, specifically examining the    shape objects differentially over the two ques-
different effects of condition and property        tions, but did accept more Same-function ob-
over age, separately for the two questions.        jects when asked the Function question than
When asked the Name question, participants         the Name question, t’s (66) Å 0.75, 03.87,
in all age groups and in both conditions pre-      p õ .05.
ferred Same-shape over Same-function ob-              Individual participant plots for the two
jects, except for the adults in the Function       questions are shown in Fig. 4. The overall
condition, who reliably preferred Same-func-       pattern supports the group analysis: When
tion to Same-shape. When asked the Function        asked the Name question, 2-, 3-, and 5-year-
question, 5-year-olds and adults in both condi-    olds in both conditions showed a reliable pref-

AID    JML 2533       /   a010h$$$84      12-17-97 08:40:26         jmla      AP: JML
14                                           LANDAU, SMITH, AND JONES

                                                         TABLE 2
      Mean Proportions (SE) of ‘‘Yes’’ Responses to Same-Shape and Same-Function Objects (Experiment 2)

                               No function                                Function                           M

Conditiona          Same                  Same                Same                    Same          Same          Same
 and age            shape                function             shape                  function       shape        function

                                                       Name question
   2 yr            .73   (.08)          .41   (.12)         .76   (.08)              .41   (.12)    .74          .41
   3 yr            .72   (.11)          .25   (.13)         .73   (.07)              .32   (.13)    .72          .28
   5 yr            .49   (.11)          .08   (.08)         .52   (.11)              .19   (.08)    .51          .14
  Adults           .69   (.09)          .33   (.10)         .26   (.06)              .78   (.06)

                                                      Function question
   2 yr            .73   (.08)          .62   (.10)         .86   (.05)              .82   (.10)    .80          .72
   3 yr            .40   (.10)          .59   (.11)         .46   (.09)              .50   (.10)    .43          .55
   5 yr            .22   (.05)          .68   (.07)         .33   (.08)              .69   (.08)    .27          .68
  Adults           .39   (.06)          .94   (.04)         .19   (.03)              .92   (.04)    .29          .93

  a
    n Å 12 in each age group and condition (No Function, Function). All participants answered both the Name and
Function questions.

erence for Same-shape objects, n Å 19 vs 0,                       the analysis. In the group that accepted Same-
18 vs 1, 19 vs 1 over the three age groups,                       function objects greater than 50% of the time
p’s õ .05, Binomial test. Adults showed a                         and Same-shape objects less than 50% of the
reliable preference for Same-shape objects in                     time, there were eight 3-year-olds and four-
the No-Function condition, n Å 10 of 12, but                      teen 5-year-olds; these were divided evenly
for Same-function objects in the Function con-                    between the No-Function and Function condi-
dition, n Å 11 of 12, x2 (1) Å 13.59, p õ .05.                    tions. In the second group (who did not meet
When asked the Function question, 2-year-                         the criterion) were the remainder of the chil-
olds did not show any distinct preference for                     dren, including sixteen 3-year-olds and ten 5-
Same-shape vs Same-function objects, n Å 8                        year-olds. Again, these were divided evenly
vs 7, respectively, whereas 3’s, 5’s, and adults                  between the conditions. Given this division,
preferred Same-function objects. n Å 12 vs 6,                     we analyzed the responses to the Name ques-
19 vs 2, 24 vs 0 for each age group, p’s õ                        tion, using an analysis of variance with age
.05, for the 5’s and adults, p Å .07 for the 3’s,                 and group as between-subjects factors and
Binomial tests. The only difference from the                      property (Same-shape, Same-function) as the
overall analysis is the addition that 3-year-                     within-subjects factor. The results showed a
olds reliably preferred Same-function objects                     main effect of property only, with children
when asked the Function question.                                 accepting Same-shape objects more often than
   As in Experiment 1, we asked whether chil-                     Same-function objects, M proportions Å .62,
dren’s varying understanding of the functions                     .19, respectively, F(1,44) Å 49.45, p õ .01.
affected their responses to the Name question.                       To summarize, the results of this experi-
We again divided children into two groups in                      ment were consistent with those of Experi-
accord with their performance on the Function                     ment 1. As in that experiment, the Name ques-
question (using the same criterion as in Exper-                   tion (‘‘Is this an X?’’)elicited a preference for
iment 1). There were no 2-year-olds who an-                       Same-shape objects among both children and
swered the Function question correctly by this                    adults when no functional instruction about
criterion, so these children were dropped from                    the objects was provided. When functional in-

AID     JML 2533           /     a010h$$$84           12-17-97 08:40:26                     jmla   AP: JML
OBJECT SHAPE, OBJECT FUNCTION                                                 15

      FIG. 4. Individual participants’ choices of Same-shape vs Same-function objects for the Name question
    and the Function question in Experiment 2. Closed circles represent individuals who had not been instructed
    about function (No-Function condition); open circles those who were instructed about function (Function
    condition). All participants were asked the Name question and the Function question.

formation was provided, adults showed a                    their relationship to the Name question also
strong bias for objects that could support the             confirmed and extended the results of Experi-
same function. These functions depended on                 ment 1. As in Experiment 1, adults in the
material in Experiment 1 and material, length,             Name question were strongly affected by in-
and global object structure in the present ex-             struction about function, whereas they an-
periment.                                                  swered the Function question correctly
   The results of the Function question and                whether instructed or not. Clearly, the adults

AID    JML 2533          /   a010h$$$85          12-17-97 08:40:26             jmla       AP: JML
16                                 LANDAU, SMITH, AND JONES

could determine by inspection whether the ob-        of a novel object name. This is consistent with
jects could carry out a given function. Also as      two possibilities. One is that functional
in Experiment 1, 3- and 5-year-olds responded        knowledge must be extremely robust to be
differently to the Function question than they       incorporated into object naming generaliza-
had to the Name question, and there was an           tions. The second is that the initial links be-
increase between age 3 and 5 in the strength         tween perception and naming are so strong
and accuracy of responses to the Function            that even extremely robust functional knowl-
question. Two-year-olds showed no evidence           edge is excluded from naming judgments.
of being able to correctly judge whether the         These two possibilities can be tested using
objects could carry out their functions. Cru-        common objects whose functions are well un-
cially, the developmental changes in the 3-          derstood.
and 5-year-olds’ responses to the Function
question did not appear to affect their re-                        EXPERIMENT 3
sponses to the Name question: There was an
                                                     Method
overall preference for Same-shape objects
among children who did respond correctly to             Participants. Ninety-two participants were
the Function question as well as those who           tested, twenty-four 2-year-olds (M age 2;6,
did not.                                             range 29–31 months), twenty-two 3-year-olds
   Although there was no effect of being in-         (M age 3;4, range 36–46 months), twenty-two
structed about function among 5-year-olds in         5-year-olds (M age 5;3, range 60–70 months),
the present experiment, these children were          and twenty-four adults. They were drawn from
rather conservative in their generalization of       the same populations as in previous experi-
the name (compared to levels of generaliza-          ments, as well as from preschools in the area
tion in Experiment 1). Why didn’t they gener-        around Bloomington, Indiana and Newark,
alize strongly to all Same-shape objects?            Delaware. All age groups were balanced for
Spontaneous comments suggest that their              gender.
knowledge about named objects might have                Design. The overall design and procedure
played a role. For example, in the No-Function       were similar to the previous experiments, ex-
condition, the cloth and bead canes were             cept that question type was a between-subjects
sometimes rejected with comments such as             variable, with roughly half of the participants
‘‘it’s a piece of towel’’ or ‘‘it’s like a neck-     in each age group asked the Name question
lace’’ and the pyramid container with ‘‘looks        and half the Function question. Participants
like a volcano.’’ In the Function condition,         who were asked the Name question were
the perforated and paper containers were             shown a standard, heard it labeled with its real
sometimes rejected with comments such as             name, and heard its function described. Then
‘‘has holes in it,’’ ‘‘it’s too soft to pull,’’ or   they were shown a series of test objects and
‘‘it’s not hard.’’ Although these comments           were asked whether each had the same name
were not frequent enough to analyze, they sug-       as the standard. Participants who were asked
gest that the 5-year-olds’ decisions to reject       the Function question observed as the experi-
some of the Same-shape objects were based            menter showed them how the standard carried
either on their similarity to other known ob-        out its function and then were asked whether
jects or, in the Function condition, the inabil-     the test objects could be used for that function.
ity of the object to support the designated          However, they never heard any of the objects
function.                                            named.
   The findings thus support the notion that            Materials. There were two sets of objects,
there are strong developmental changes in            one with a well-known standard and the other
children’s knowledge of objects’ functions,          with a less well-known standard. The standard
and that this knowledge is only slowly incor-        for the Comb set was a 4’’ 1 3’’ blue plastic
porated into children’s initial generalizations      10-prong comb with no handle. The test set

AID    JML 2533        /   a010h$$$85       12-17-97 08:40:26         jmla      AP: JML
You can also read