Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Psychon Bull Rev. Author manuscript; available in PMC 2010 October 1.
Published in final edited form as:
Psychon Bull Rev. 2009 October; 16(5): 920–924.
doi:  10.3758/PBR.16.5.920
PMCID: PMC2829667

Five-year-olds’ beliefs about the discreteness of category boundaries for animals and artifacts


Research on adult concepts indicates that category structure varies by domain; adults view membership in animal categories as absolute, but membership in artifact categories as graded. This study examined domain differences in beliefs about category boundaries among young children (5-year-olds). Results indicated that young children, like adults, were less likely to endorse graded category membership for animal than for artifact categories. These domain differences could not be attributed to domain differences in typicality. Implications for conceptual development, and for models of domain specificity in adult cognition, are discussed.

Research on adult concepts indicates that category structure varies by domain of knowledge. These domain differences are thought to reflect systematic ontological commitments (Gelman & Coley, 1991). Particularly, adults’ belief that many animal categories are natural kinds (i.e., real categories that are discovered in nature) indicates that even atypical members (e.g., birds that cannot fly) are full category members by virtue of being included in the kind defined by nature. Thus, adults view animal categories as having discrete, objectively-defined boundaries, and view membership in animal categories as absolute (Diesendruck & Gelman, 1999; Estes, 2003, 2004; see also Kalish, 1995, and discussion in Gelman & Hirschfeld, 1999). In contrast, adults believe that artifact categories are created by people, and therefore are subjective and flexible (Kalish, 1998; Malt, 1990). These beliefs allow for partial membership; thus, membership in artifact categories is viewed as graded, and typicality becomes an important predictor of category membership. For artifacts, atypical items (e.g., a beanbag chair for the category furniture) are considered to be partial members (e.g., “sort of” furniture), and to be less full members than more typical items (e.g., couches). The goal of the present study was to examine whether young children share adults’ intuitions that the discreteness of category boundaries varies by domain.

There is considerable debate as to whether young children incorporate abstract beliefs, including ontological commitments, into their categories. Traditional perspectives on cognitive development suggested that younger children’s categories lack abstract beliefs, are grounded in superficial perceptual features (e.g., color, shape), and are unstable across contexts (Piaget, 1929). More recently, Sloutsky and colleagues have also argued that young children do not have conceptually rich categories. They suggest that conceptual knowledge, which includes domain-specific beliefs about the nature of categories (e.g., that natural kind categories are homogeneous and have sharp boundaries, Sloutsky, Kloos, &Fisher, 2007), is the product of an extended period of associative learning, and does not emerge until middle childhood (ages 8–9; see also Smith, 1989). Instead, they propose that young children (ages 4–5) rely on perceptual properties, instead of categories, for a variety of cognitive processes (e.g., for encoding and retrieval, Fisher & Sloutsky, 2005; Sloutsky & Fisher, 2004; for induction, Sloutsky et al., 2007). On this account, younger children’s categories should not incorporate abstract ontological commitments; such beliefs are expected to be the product of extended developmental processes. Thus, in early childhood (e.g., before age 8), beliefs about the discreteness of category boundaries should not vary systematically by domain.

In contrast, others have argued that children incorporate abstract beliefs into their categories from a very young age, and that these abstract beliefs guide and propel conceptual development (Gelman, 2003). This view accounts for and predicts a range of findings indicating the early emergence of ontological commitments. For example, children discriminate between animals and artifacts in early infancy (Mandler & McDonough, 1993). By preschool, children know that animals, but not artifacts, move on their own (R. Gelman, 1990), that artifacts are made by people whereas animals originate in nature (Gelman & Kremer, 1991), and that animals, not artifacts, are subject to biological causal mechanisms (e.g., fatigue, hunger; Hatano & Inagaki, 1994). Also, preschoolers believe that category identity can change over time for artifacts (Keil, 1989; Siegel & Callanan, 2007) but is stable for animals (Gelman & Wellman, 1991). Preschoolers are also more likely to express information about novel animals than artifacts in generic form, indicating perhaps that they have a general expectation that animal categories are more coherent and richly-structured than artifact categories (Brandon & Gelman, 2009).

Although prior developmental work has documented numerous distinctions in children’s beliefs about animals and artifacts, all prior work directly examining beliefs about the discreteness of category boundaries has focused on adult populations. One prior developmental study, however, is directly relevant to this issue. Keil (1989) presented 5-year-olds with a series of 5 pictures that included an animal of one category (e.g., a tiger), an animal of another category (e.g., a lion), and three intermediary pictures that were described as showing the animal going through a transformation from being a tiger to becoming a lion (e.g., “some scientists took some dye that never comes off and dyed his black stripes orange,” p. 240). For each picture, children were asked to judge the category identity of the animal. Keil found that children almost always responded that the animal was one kind or another. Only one five-year-old raised the possibility that an intermediary example had partial category membership (e.g., part lion, part tiger).

Keil’s study suggests that young children, like adults, may view animal category boundaries as absolute. However, there are several reasons why this work does not provide a definitive answer to this question (nor was it designed to do so). First, this study included only animal categories, so it is not possible to determine whether children’s beliefs about category boundaries vary by domain. Second, partial category membership was not explicitly suggested to children. Thus, one possibility is that children had difficulty articulating beliefs about partial membership, but would endorse such an option if offered. Third, this work presented highly unusual transformations; it is not clear how these findings will relate to judgments about real, but unusual category members (e.g., an ostrich for the category bird).

In the present study, 5-year-olds were assigned to either a category membership or a typicality condition. In the membership condition, children judged whether an item was a member of a category on a three-point scale (yes, sort of, no). In the typicality condition, children were asked to judge whether an item was a typical member of a category, using the same scale. Our primary hypothesis concerned the membership condition; here, if children have domain-specific ontological commitments, 5-year-olds should endorse the graded response (“sort of”) less often for animals than for artifacts. The typicality condition had two purposes. First, we should find no differences in typicality by domain, which will confirm that domain effects in the membership condition cannot be attributed to domain differences in typicality. Second, the typicality condition will allow us to document that children do identify some animals as atypical, thus supporting the interpretation that even atypical animals are treated as full category members (or not category members) by children.



Participants included 32 children (18 male, 14 female; M age = 5.5 years, range = 4.95 – 6.0 years) recruited via parent consent letters from kindergartens in a midsize city in the midwestern United States. The sample was approximately 80% white, with the remainder of participants from diverse racial-ethnic backgrounds.


Children were randomly assigned to the membership (n = 16) or typicality (n = 16) condition. Within each condition, children completed items in two domains: animals and artifacts. The domains were presented in a counterbalanced order across participants. There were two sets of stimuli; half of children received items about birds in the animal domain and clothing in the artifact domain (stimulus set A), the other half received items about fish in the animal domain and tools in the artifact domain (stimulus set B; see Table 1). These categories were selected because they are familiar to young children, and have been the subject of prior work on category structure with adults. Each domain included 12 items. Based on typicality ratings obtained from adults in prior work (Diesendruck & Gelman, 1999; Estes, 2003; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976), for each category, 4 items were typical exemplars, 4 items were atypical exemplars, and 4 items were thematically related to the category but not category members. The items within domains were given in 1 of 2 pre-set random orders. To summarize the design, the following factors were crossed across participants: assignment to the typicality or membership task, order of presentation of the domains, stimulus set, and item order.

Table 1
Stimuli by category


Children completed the task in individual sessions with trained undergraduate research assistants in a quiet area of their elementary school.

Membership Condition

First, children completed a warm-up activity. They were introduced to a scale that presented three cartoon thumbs in various positions (one pointed up, one pointed to the side, and one pointed down). For warm-up, children were told, “I’m going to show you some pictures and ask you some questions about them. We’ll practice with some questions about sports. I’ll show you some pictures, and you tell me if the thing in the picture is a sport or not. You can tell me ‘thumbs up!’ it is definitely a sport; ‘thumbs to the side!’ it is sort of a sport; or ‘thumbs down!’ it is not a sport at all.” Children were then given three example questions, each accompanied by a colorful picture of the activity: “This is baseball. Is baseball a sport? This is reading. Is reading a sport? This is hopscotch. Is hopscotch a sport?” Children were encouraged to respond verbally and by pointing to the correct picture on the scale. Children were expected to say “yes” to the first question, “no” to the second question, and “sort of” to the third question. If children did not give the expected response, or failed to respond, the experimenter supplied the correct answer. For example, the experimenter would say: “I think hopscotch is sort of a sport. Can you point to the thumb that shows ‘sort of’?” No corrective feedback was supplied after the warm-up period.

To begin the task, for a bird set (for example), the experimenter said, “Now we’ll do some questions about animals. I’ll show you some pictures, and for each one, you tell me if the thing in the picture is a bird. Remember, use the thumbs to tell me: yes- it is a bird; sort of- it is sort of a bird; no- it is not a bird at all.” For each item, the experimenter presented a colorful photograph of the test object (e.g., of a robin), and said, for example, “Is a robin a bird?” Children could respond verbally, or by pointing to a picture on the scale.

For stimulus set B, the word “fish” was substituted for “bird” in the introduction. An example item is, “Is a trout a fish?” On artifact sets, the procedure was identical, except the word “thing” was substituted for “animal”, and the words “clothing” and “tool” were used as appropriate. Example items are: “Is a dress clothing”, “Is a hammer a tool?”

Typicality Condition

The warm-up for the typicality condition was similar. Children were shown the same scale and told “…I’ll show you some pictures, and you tell me if the thing in the picture is a good activity to show sports. You can tell me ‘Thumbs up- Yes! That is a good one to show what a sport is,’ ‘Thumbs to the side- Sort of, that is a sort of good one to show what a sport is,’ or ‘Thumbs down- No, that is not a good one to show what a sport is at all.”’ They were also given three warm-up trials: “This is baseball. Is baseball a good one to show what a sport is? This is reading. Is reading a good one to show what a sport is? This is hopscotch. Is hopscotch a good one to show what a sport is?” As in the membership condition, children were expected to say “yes” to the first question, “no” to the second question, and “sort of” to the third question. Corrective feedback was provided as needed.

To begin a set, for a bird trial, the experimenter said “…for each one, you tell me if the thing in the picture is a good one to show what a bird is. Remember, use the thumbs to tell me: yes- it is a good one to show what a bird is; sort of- it is sort of a good one to show what a bird is; no- it is not a good one to show what a bird is at all.” For each item, the experimenter said, for example “Is a [robin] a good one to show what a [bird] is?” The fish, clothing, and tool sets followed the same structure. All other procedures were the same as in the membership condition, and the same visual stimuli were used in both conditions.


Preliminary Analyses

Preliminary analyses revealed no main or interactive effects of participant gender, domain order, item order, or stimulus set. Therefore, these factors were not considered further. For stimulus set, this means that ratings for birds and fish were collapsed, as were items for tools and clothing.

To confirm that the stimuli varied as intended in terms of typicality, we first examined the data from the typicality condition by analyzing responses as mean scale scores (items were coded from 0–2, such that higher numbers indicate higher ratings of typicality). As intended, typicality did not vary by domain, for either the typical items (Animals, M = 1.88, SD = .29; Artifacts, M = 1.88, SD =.27; t(15) < .01, p > .99), the atypical items (Animals, M = 1.03, SD = .47; Artifacts, M = 1.03, SD =.54; t(15) <.01, p > .99), or the not-member items (Animals, M = .41, SD = .44; Artifacts, M = .30, SD = .39; t(15) = .88, p > .35). Also as intended, the typical items were rated as more typical than the atypical items, for both animals, t(15) = 8.29, p < .001, and artifacts, t(15) = 6.09, p < .001, and the atypical items were rated as more typical than the not-members for both animals, t(15) = 4.84, p < .001, and artifacts, t(15) = 4.40, p =.001. Thus, these analyses confirm that typicality varied as intended across items, and did not vary by domain.

Analyses of Graded Responses

Our central research question involved the proportion of times that children selected the graded response, “sort of”. To examine these data, we conducted a 2 (domain: animal, artifact) × 2 (task: membership, typicality) factorial analysis using the Generalized Estimating Equations (GEE) procedure in SPSS 16 (Hardin & Hilbe, 2003), with domain as a within-subjects variable, and task as a between-subjects variable. The dependent variable was the number of times children selected the graded response, out of the total possible, in each domain. The proportions of graded responses by task, typicality level, and domain, are presented in Table 2.

Table 2
Mean proportions of graded responses by task, domain, and typicality level, Study 1.

This analysis revealed a main effect of domain, χ2(1) = 6.99, p < .01, and an interaction between domain and task, χ2(1) = 15.44, p < .001. As shown in Table 2, for judgments about category membership, children gave more graded responses for artifacts than for animals, whereas there were no domain differences in graded response for typicality. Also, for animals, children gave more graded responses when judging typicality than membership, but for artifacts, they gave equivalent proportions of graded responses for both.

We also reran these analyses separately for items of each typicality level. Of most interest were the atypical items, as these were expected to elicit the most graded responses. For the atypical items, graded responses were fairly common (33% of total responses). As in the analyses of the combined data, there was a significant interaction between domain and task, χ2(1) = 6.61, p = .01. As shown in Table 2, for judgments about category membership, children gave more graded response for artifacts than for animals, whereas the proportion of graded responses for typicality did not vary by domain. Also, for animals, children gave more graded responses when judging typicality than when judging membership, but for artifacts, they gave equivalent proportions for both.

For the typical items and the non-member items, graded responses were quite rare (6% for typical items, 13% for non-member items). Yet, the predicted pattern was still found for both item sets. There were significant interactions between domain and task, for typical items, χ2(1) = 3.93, p < .05, and for non-member items: χ2(1) = 4.26, p < .05 (see Table 2).

Study 1b

The findings from Study 1 are consistent with prior work in adult populations (Diesendruck & Gelman, 1999; Estes, 2003). To confirm that children’s performance was adult-like using our particular items and procedure, we conducted Study 1b, with a sample of older participants, using the same stimuli and methods. Data were collected from 39 high school students (18 male, 21 female; M age = 17.7 years, range = 15.72 – 19.07), who attended a high school in the same school district as the children who participated in Study 1. Students completed a paper-and-pencil version of the task independently during a regularly scheduled class. Participants were randomly assigned to the typicality (n = 20) or membership (n = 19) task.

As in Study 1, we found a domain × task interaction, χ2(1) = 13.46, p < .001, for the total number of graded responses. When judging category membership, participants gave more graded responses for artifacts (M = .19, SD = .13) than for animals (M = .07, SD = .10), whereas the pattern for typicality was reversed (Animals, M = .21, SD = .14; Artifacts, M = .13, SD = .10). Also, for animals, participants gave more graded responses for typicality than for membership, whereas for artifacts, there was no effect of task. We next combined data from Studies 1 and lb, and ran a 2 (age) × 2 (task) × 2 (domain) GEE analysis on the proportion of graded responses. This analysis revealed no main or interactive effects of age, p > .15. The three-way interaction involving age did not approach significance, p > .90.

Study 1c

The goal of Study 1 c was to test more fully whether children believe that animal categories have absolute boundaries, by asking them to reason about less familiar categories and exemplars. For the items in Study 1, it is possible that children had previously heard some of the atypical animals labeled as category members (e.g., they may have heard penguins labeled as birds). If so, such experiences could have increased their certainty that atypical category members may be considered full members. Thus, an important question is whether children will infer that atypical animals are still either full members (or not members) of specific categories when they are uncertain about an individual animal’s identity and have never experienced the animal’s label.

Study 1c followed the procedures of the bird and fish trials from the category membership task of Study 1, except that we examined children’s (n = 12, 6 male, 6 female, age, M = 5.9 years) responses to the category “mammal”, which we suspected that children know less about and hear labeled less often. Also, the atypical items were unfamiliar mammals, and we did not label any of the items. Individual items were presented visually, and children were asked “Is this a mammal?” As in Study 1, the answer choices were “yes”, “sort of”, and “no.” Items included four typical mammals (pig, monkey, cat, dog), four atypical and unfamiliar mammals (armadillo, tarsier, matschie, pangolin), and four animals which were not mammals (penguin, frog, shark, turtle). After children provided ratings for category membership for all items, we returned to the items, and asked them to label each one. For example, they were shown the picture of the dog a second time and simply asked, “What is this?”

First, we examined children’s overall performance for evidence that children had high levels of uncertainty about the category mammals. Indeed, children’s performance indicated that they were not very familiar with this category. For the typical mammals, children correctly responded that the animals were mammals on only 52% of trials (in comparison, children in Study 1 correctly identified the typical birds and fish as category members on 70% of trials). Children also identified the atypical mammals as mammals on 42% of trials, and incorrectly identified the not category members as mammals on 40% of trials. Thus, these children appeared relatively poor at identifying mammals. Also, the data from the labeling phase at the end of the trial (when children were shown each item, and asked to identify it) confirmed that children were not familiar with the atypical mammals. They labeled the typical mammals correctly on 100% of trials, the non-mammals correctly on 96% of trials, but the atypical mammals correctly on only 8% of trials. Thus, this preliminary look at the data indicated that children had a high level of uncertainty, both about how to identify mammals, and about the identity of the individual atypical exemplars.

Despite this uncertainty, children rarely endorsed graded category membership for mammals (Total, M = .03, SD = .07). They failed to endorse the graded response across each level of typicality (Typical mammals, M = .04, SD = .14; Atypical mammals, M = .04, SD =.10; Not mammals, M = 00, SD = 00). We compared these proportions to the equivalent proportions from Study 1, and found that despite the increased uncertainty present in Study 1c, children did not endorse the graded response more often, Z = 1.66, p > .47. Thus, this follow-up study indicates that children believe that animal categories have strict boundaries even when they do not know much about the relevant categories or exemplars (see Estes 2004, for a similar finding with adults).

General Discussion

By age five, children have domain-specific intuitions about the discreteness of category boundaries. In particular, children viewed animal categories as having more absolute boundaries than artifact categories. Children recognized that certain animals were atypical; the proportion of times that they endorsed the graded response (“sort of”) did not vary by domain in the typicality condition. Thus, although children recognized that an ostrich was an atypical bird, for example (much as they recognized that a headband was an atypical piece of clothing), they were unlikely to view an ostrich as “sort of” a bird. They endorsed partial category membership significantly more often for artifacts. Overall, these data are consistent with the proposal that young children incorporate ontological commitments into their categories (Gelman, 2003).

One limitation of the present work is the small number of categories examined. These categories were selected because they are familiar to young children (with the exception of mammals, in Study 1 c) and have been the subject of prior work on category structure with adults (e.g., Rosch et al., 1976). In future work, however, it will be necessary to include a broader range of animal and artifact categories in order to determine whether children’s approach to categorization is consistent across more diverse categories within a domain.

These findings are also relevant to questions regarding domain-specificity in adult cognition. The present data are consistent with the proposal that domain specificity is cognitively primitive (e.g., the result of innate modules or domain-specific learning mechanisms; Atran, 1998). Alternate theories propose that domain-specificity in cognition is the product of domain-general learning mechanisms and histories of causal learning (Sloman, Lombrozo, & Malt, 2007). The present findings indicate that a challenge for such models is to explain how domain-general learning mechanisms lead to the development of abstract, domain-specific ontological beliefs prior to age five. In future work, data on the ontological beliefs of even younger children may help to distinguish between various theoretical accounts of domain specificity in adult cognition.


Funding was provided by NICHD grant HD-36043 to the second author. We are grateful to the children, teachers, and parents who participated in this research, as well as to Amanda Karp and Christine Shenouda for their assistance with data collection.


  • Atran S. Folk biology and the anthropology of science: Cognitive universals and cultural particulars. Behavioral and Brain Sciences. 1998;31:547–609. [PubMed]
  • Brandone AC, Gelman SA. Differences in preschoolers’ and adults’ use of generics about novel animals and artifacts: A window onto a conceptual divide. Cognition. 2009;110:1–22. [PMC free article] [PubMed]
  • Diesendruck G, Gelman SA. Domain differences in absolute judgments of category membership: Evidence for an essentialist account of categorization. Psychonomic Bulletin and Review. 1999;6:338–346. [PubMed]
  • Estes Z. Domain differences in the structure of artifactual and natural categories. Memory and Cognition. 2003;31:199–214. [PubMed]
  • Estes Z. Confidence and gradedness in semantic categorization: Definitely somewhat artifactual, maybe absolutely natural. Psychonomic Bulletin and Review. 2004;11:1041–1047. [PubMed]
  • Fisher AV, Sloutsky VM. When induction meets memory: Evidence for gradual transition from similarity-based to category-based induction. Child Development. 2005;76:583–597. [PubMed]
  • Gelman R. First principles organize attention to and learning about relevant data: Number and the animate-inanimate distinction as examples. Cognitive Science. 1990;14:79–106.
  • Gelman SA. The essential child: Origins of essentialism in everyday thought. New York: Oxford University Press; 2003.
  • Gelman SA, Coley JD. Language and categorization: The acquisition of natural kind terms. In: Gelman SA, Byrnes JP, editors. Perspectives on language and thought: Interrelations in development. Cambridge: Cambridge University Press; 1991. pp. 146–196.
  • Gelman SA, Hirschfeld LA. How biological is essentialism? In: Atran S, Medin D, editors. Folkbiology. Cambridge, MA: MIT Press; 1999.
  • Gelman SA, Kremer KE. Understanding natural cause: Children’s explanations of how objects and their properties originate. Child Development. 1991;62:396–414. [PubMed]
  • Gelman SA, Wellman HM. Insides and essences: Early understandings of the nonobvious. Cognition. 1991;38:213–244. [PubMed]
  • Hardin JW, Hilbe JM. Generalized estimating equations. Boca Raton, FL: CRC Press; 2003.
  • Hatano G, Inagaki K. Young children’s naïve theory of biology. Cognition. 1994;50:171–188. [PubMed]
  • Kalish CW. Essentialism and graded membership in animal and artifact categories. Memory and Cognition. 1995;23:335–353. [PubMed]
  • Kalish CW. Natural and artificial kinds: Are children realists or relativists about categories? Developmental Psychology. 1998;34:376–391. [PubMed]
  • Keil FC. Concepts, kinds, and cognitive development. Cambridge, MA: Bradford Book/MIT Press; 1989.
  • Malt BC. Features and beliefs in the mental representation of categories. Journal of Memory and Language. 1990;29:289–315.
  • Mandler JM, McDonough L. Concept formation in infancy. Cognitive Development. 1993;8:291–318.
  • Piaget J. The child’s conception of the world. London: Routledge and Kegan Paul; 1929.
  • Rosch E, Mervis CB, Gray WD, Johnson DM, Boyes-Braem P. Basic objects in natural categories. Cognitive Psychology. 1976;8:382–439.
  • Siegel DR, Callanan MA. Artifacts as conventional objects. Journal of Cognition and Development. 2007;8:183–203.
  • Sloman S, Lombrozo T, Malt B. Ontological commitments and domain specific categorization. In: Roberts MJ, editor. Integrating the mind: Domain general versus domain specific processes in higher cognition. Hove, UK: Psychology Press; 2007. pp. 105–129.
  • Sloutsky VM, Fisher AV. Induction and categorization in young children: A similarity-based model. Journal of Experimental Psychology: General. 2004;133:166–188. [PubMed]
  • Sloutsky VM, Kloos H, Fisher AV. When looks are everything: Similarity versus kind information in early induction. Psychological Science. 2007;18:179–185. [PubMed]
  • Smith LB. A model of perceptual classification in children and adults. Psychological Review. 1989;96:125–144. [PubMed]