Perceptual to Category Transition 1 Running Head: a Transition from Perceptual to Category Induction
چکیده
We examined whether inductive reasoning development is better characterized by accounts assuming an early category bias versus an early perceptual bias. We trained 264 children aged 3 to 9 years to categorize novel insects using a rule that directly pitted category membership against appearance, followed by an induction task with perceptual distractors at different levels of featural similarity; a further 52 children aged 4-7 were tested on a maturity-grouping induction task. Categorization performance was consistently high, however we found a gradual transition from a perceptual bias in our youngest children to a category bias around age 6-7. In addition, children of all ages were equally distracted by higher levels of featural similarity. The transition of strategy is unlikely to be due to an increased ability to inhibit perceptual distractors. Instead, we argue that the transition is driven by a fundamental change in children’s understanding of category membership. Perceptual to category transition 3 Human reasoning depends crucially on our ability to attend to relevant information when making generalizations. This becomes most challenging when the relevant information is not immediately obvious. For example, within the natural world, subtle cues are sometimes better predictors of behavior than overall appearance (e.g., Canadian goslings can only be distinguished from mallard ducklings by subtle differences in their bills and feet, even though the two species have different diets and migration habits, and grow up to look highly dissimilar). Two broad accounts have been proposed to explain the development of category induction and analogical reasoning in young children. The first set of accounts hold that children have an early bias to attend to hidden, non obvious properties and therefore, make generalizations based on kind-information over appearance (category-bias accounts; Gelman, 2003). Similarly, these accounts suggest that young children are capable of reasoning analogically, although are constrained by their knowledge of the relevant relations (Goswami, 1992; 2001). The second set of accounts state that young children intuitively focus on perceptual information when making induction decisions (perceptual-bias accounts; Sloutsky, Kloos & Fisher, 2007), and surface characteristics in analogical reasoning tasks (Gentner, 1988). Children maintain this perceptual / surface bias until they have developed the knowledge base necessary to support a shift towards a focus on category and relational information (a relational shift; Gentner, 1988; Gentner & Rattermann, 1991; Rattermann & Gentner, 1998). Although these two accounts make clear predictions for young children’s reasoning strategies, evidence exists to support both category (e.g., Gelman & Markman, 1986) and perceptual (e.g., Sloutsky et al., 2007) biases in young children, which currently makes it impossible to resolve this conflict. Studies investigating early reasoning biases Triad paradigms are commonly used to investigate induction strategies since they can directly pit perceptual against category choices (Bulloch & Opfer, 2009; Gelman & Perceptual to category transition 4 Markman, 1986; Hayes & Thompson, 2007; Opfer & Bulloch, 2007; Sloutsky et al., 2007). A target item is presented alongside two test items, one of which matches the target in category membership but not overall appearance, whereas the other matches the target in overall appearance but not category membership. The child is told a hidden property of the two test items “this puppy hides bones in the ground and this fox hides food in the ground” and asked to generalize this property onto the target item (Gelman & Markman, 1986). Some studies using triad paradigms have indicated that young children make induction decisions consistent with category context, even when the relationship is not perceptually obvious (e.g. generalizing “bird properties” to a dissimilar looking bird; Gelman & Markman, 1986; see also Gopnik & Sobel, 2000). Category preference has been shown in children as young as 2 years old when items have shared category labels (Gelman & Coley, 1990; also Deák & Bauer, 1996), and children aged 4-5 years have shown a clear preference for category information, even when items were presented with dissimilar labels (for example ‘rabbit’ and ‘bunny’; Gelman & Markman, 1986). Other research has indicated that young children are capable of varying their strategy, depending on context. Opfer and Bulloch (2007) found that when category information was available, children were capable of using this information to make a category-based induction decision. However if no category information was available, then children relied on perceptual similarity (see also Hayes & Thompson, 2007). Together, these studies suggest that children may have an early tendency to focus on category cues beyond overall appearance. Nevertheless, even atypical category members share some perceptual characteristics (Heit & Hayes, 2005). Thus, the apparent category-based decisions observed in the studies described above may have been influenced by perceptual information (Jones & Smith, 1993; McClelland & Rogers, 2003; Rakison, 2000; Rakison & Hahn, 2004; Sloutsky & Fisher, 2004). In order to address this issue, Sloutsky et al (2007) investigated children’s default strategies when category and perceptual Perceptual to category transition 5 cues conflict, allowing perceptual-based versus category-based decisions to be disambiguated. Children aged 4 to 5 were trained to categorize novel biological stimuli based on a non-obvious rule (relative number of “fingers” to “buttons”). All other perceptual features were non-predictive of category membership. Although children were highly accurate in categorization tasks, they made significantly more perceptual than category induction choices, thus demonstrating a clear perceptual induction bias. However, the validity of Sloutsky et al’s design has been questioned. Although Sloutsky et al demonstrated that the children could use the category rule very accurately in the initial and final categorization tasks, it was unclear whether the children recognized the biological relevance of the category rule, and therefore may not have been sufficiently motivated to apply this rule in the induction task. Thus, the apparent perceptual bias Sloutsky et al observed may have been an artifact of unrealistic stimuli or arbitrary category rules (Gelman & Waxman, 2007). Evidence for a developmental transition in reasoning strategy According to perceptual-bias accounts, young children’s default strategy is for perceptual induction, and only once they develop the sufficient knowledge of causal relations within a domain, can they shift their focus to category and relational cues (Gentner, 1988; Gentner & Toupin, 1986; Sloutsky et al., 2007). For example, when presented with a straw and a plant stem and asked how the two items were similar, children aged 5 focused on the perceptual similarity and said that both items were long and thin, but by age 9 children could shift their focus to relational similarity and said that both could also carry water: a relational shift (Gentner, 1988; also see Gentner & Rattermann, 1991; Halford, 1993). As discussed in Ratterman and Gentner (1998), this shift reflects changes in knowledge representation. Specifically, young children have better representations of object properties than of the relations between objects. Only once children have developed sufficient knowledge of the Perceptual to category transition 6 causal relations within a domain, can they shift their attention away from common object properties towards common relational structure. However, there are several alternative explanations for apparent inductive transitions in reasoning. Firstly, this may reflect a qualitative change in the nature of children’s thinking, rather than a gradual transition with increasing knowledge, perhaps as a result of a stage-like shift in cognitive maturity, akin to a Piagetian shift from preoperational to concrete operational thinking (Inhelder & Piaget, 1958; Piaget, 1964). Secondly, according to Goswami (1992, 2001), young children’s natural default is to actively seek relational interpretations, although they may be held back by performance factors, such as knowledge about the relevant relational comparisons. Thus, once children understand the specific relationships that are relevant in a particular context, they will naturally use these relationships over surface characteristics to inform their decisions. Thirdly, a transition could be due to a boost in children’s understanding of cue validity: if a child does not recognize the relational connection as a reliable predictor, then they will refer back to perceptual cues (Bulloch & Opfer, 2009). Thus, again, young children may be capable of using relational knowledge, but only use this information when they recognize the relevance of this relational connection to the task. Finally, a child may be capable of using relational knowledge and making induction decisions based on category information, yet does not have the cognitive control to successfully initiate the appropriate response. Even for adults, the likelihood of generalizing properties from one item to another depends on the strength of within-category similarity. For example, Osherson, Smith, Wilkie, López and Shafir (1990) found that the more perceptually similar the items (e.g., robin, bluejay and sparrow versus robin, bluejay and goose), and the more specific the category (e.g., falcon and falcon – shared level: falcon, versus falcon and pelican – shared level: bird), then the greater the strength of the inductive inference (see also Sloman, 1993). In a triad paradigm in which perceptual similarity Perceptual to category transition 7 conflicts with category membership, the most salient, perceptually similar choice must be inhibited before a category induction decision can be made. In this situation, overcoming highly similar perceptual distractors is likely to be challenging for young children, even if they understand the importance of category membership information (Richland, Morrison & Holyoak, 2006). Thus, there may be no genuine “relational shift”. Instead, even young children may be seeking relational information, and become increasingly successful in initiating the appropriate response as they develop in their executive control. Current study Each of the broad accounts we have outlined can accommodate evidence that children’s responses in reasoning tasks change between ages 2 and 9. In particular, over time, children focus more on category and relational information, and less on salient perceptual characteristics. However, as outlined above, the perceptual-bias and category-bias accounts make different predictions about the mechanisms that drive these developmental changes. We aim to examine developmental changes in children’s category induction, using a task which is designed to elicit different predictions from these two broad accounts. Importantly, unlike previous research, this study tested children from multiple year groups (nursery: 3-4 years, through to year four: 8-9 years). This will enable us to examine the developmental trajectory of children’s induction strategies, providing a stronger test of category-bias and perceptualbias accounts. As in Sloutsky et al (2007), we created novel examples of biological stimuli in order to examine children’s early induction biases when the salient perceptual cues are not informative of category membership. Children’s understanding of the category rule was confirmed in a stringent test of their categorization performance before and after the induction task. Perceptual and category induction choices were disambiguated by using a triad paradigm in which children were asked to generalize properties from a target item to Perceptual to category transition 8 either a perceptual or a category choice (as in Bulloch & Opfer, 2009; Gelman & Markman, 1986; Opfer & Bulloch, 2007; Sloutsky et al., 2007). In order to address concerns about the validity of Sloutsky et al’s design (Gelman & Waxman, 2007), we used biologically plausible stimuli (novel insects) and demonstrated a familiar relational connection through an animated transition from juvenile to adult. Previous research demonstrates that young children understand the aging process (see Inagaki & Hatano, 1996 for a review). Therefore, growth from infancy to adulthood provides a simple way to demonstrate a relational match, whilst providing a biologically plausible context for changes in perceptual features of the items. Thus, the category-bias account would predict that as long as children clearly understand this relationship, then even the youngest children should generalize properties from an adult insect to the same item depicted as a juvenile, rather than to a different-category adult. In order to demonstrate that the relational connection is a reliable predictor, the transition from juvenile to adult will be shown prior to every induction trial. In Experiment 1a, we examine whether improvements in cognitive control drive a potential perceptual-category transition, by varying the level of featural similarity between the target and perceptual distractors. If a pattern of decreasing “distractibility” is observed along with a parallel transition from perceptual to category induction, then this would provide an explanation of the transition consistent with category-bias accounts. Specifically, young children understand the importance of category membership, yet fail to demonstrate their knowledge due to underdeveloped cognitive control. This outcome would be manifested in an interaction between distractor-similarity and age: the level of distractor similarity should affect induction decisions for younger children more than older children. In contrast, if the transition is independent of featural distraction, then this would support perceptual-bias accounts, suggesting that children’s initial failure to make category-based decisions reflects a lack of understanding about the importance of category membership rather than an inability Perceptual to category transition 9 to ignore perceptual distractors. Thus, children only start to use category membership to inform their induction decisions once they’ve learned to shift their focus away from perceptual features, and towards relational structure (Gentner, 1988; Gentner & Rattermann, 1991; Sloutsky et al., 2007). Experiment 1a Method Participants Two hundred and sixty four primary school children participated: 57 Nursery (3-4 years; range 3.01-4.04 years; mean = 3.10), 65 Reception (4-5 years; range 4.09-5.09; mean = 5.04), 43 Year 1 (5-6 years; range 5.09-6.09; mean = 6.03), 38 Year 2 (6-7 years; range 6.097.09; mean = 7.03), 31 Year 3 (7-8 years; range 7.05-8.03; mean = 7.10), and 30 Year 4 (8-9 years; range 8.05-9.03; mean = 8.11); 133 males and 131 females. An extended age range of six year groups was used in order to better observe a developmental trajectory of the induction shift with age, and whether perceptual distraction decreases. Forty-nine primary school children participated in a similarity pre-test: 10 Reception (4-5 years), 27 Year 1 (5-6 years), and 12 Year 2 (6-7 years); range 4.11-7.10 years; 26 males and 23 females. Twenty three adults (mean age = 26.0 years, range 18-58; 7 males and 16 females) participated in a domain categorization pre-test. Stimuli One hundred and ninety two novel insect stimuli were created (42 juveniles, 42 adult targets, 12 High Similarity Distractors (HSDs; these distractors only differed from the relevant adult target by head shape, see Fig. 1a), 12 Low Similarity Distractors (LSDs; these distractors differed from the relevant target on all dimensions apart from the overall size and shade, see Fig. 1b) and 84 transitional images used in the juvenile to adult animations, see Fig. 1c). Most (83%) of the stimuli were only shown once during the experiment, but some Perceptual to category transition 10 overlap was allowed between initial categorization and induction phases. In order to allow every stimulus to be unique, they all differed on at least one dimension (head shape: round or angled; eyes: orange or white; body: round or triangular; color: purple or green; color of markings: black or white; number of markings: two or four; overall size: small juvenile or large adult; overall shade: pale juvenile or dark adult). In the transition from juvenile to adult, stimuli changed in size, overall shading and grew legs (see Figure 1a). [Insert Figure 1a here] During the induction task in the main experiment, children saw 24 triads consisting of an adult target, a same-category but perceptually dissimilar juvenile, and a different-category but perceptually similar adult (see Figures 1b and 1c). To make certain that the target item was perceptually more similar to the perceptual distractor test item than the same-category test item, the similarity between test and target items was validated using child ratings. [Insert Figures 1b and 1c here] On 24 triads, children had to choose which test item looked most like the target item. One-sample t-tests confirmed that the perceptual choice was chosen significantly more often than chance. In triads with HSDs, t (48) = -205.04; p 25 References Bulloch, M. J., & Opfer, J. E. (2009). What makes relational reasoning smart? Revisiting the perceptual-to-relational shift in the development of generalization. Developmental Science, 12 (1), 114-122. Deák, G., & Bauer, P. J. (1996). The dynamics of preschoolers’ categorization choices. Child Development, 67, 740-767. Gelman, S. A. (2003). The essential child: origins of essentialism in everyday thought. Oxford: Oxford University Press. Gelman, S. A., & Coley, J. D. (1990). The importance of knowing a dodo is a bird: categories and inferences in 2-year old children. Developmental Psychology, 26, 796- 804. Gelman, S. A., & Markman, E. M. (1986). Categories and induction in young children. Cognition, 23, 183-209. Gelman, S. A., & Waxman, S. R. (2007). Looking beyond looks: Comments on Sloutsky, Kloos and Fisher. Psychological Science, 18 (6), 554-555. Gentner, D. (1988). Metaphor as structure mapping: the relational shift. Child Development, 59, 47-59. Gentner, D., & Rattermann, M. J. (1991). Language and the career of similarity. In S. A. Gelman & J. P. Byrnes (Eds.), Perspectives on thought and language: Interrelations in development (pp. 225-277). London: Cambridge University Press. Gentner, D., & Toupin, C. (1986). Systematicity and surface similarity in the development of analogy. Cognitive Science, 10, 277-300. Gopnik, A., & Sobel, D. M. (2000). Detecting blickets: How young children use information about novel causal powers in categorization and induction. Child Development, 26, 295- 320. Perceptual to category transition 26 Goswami, U. (1992). Analogical reasoning in children. Hillsdale, NJ: Lawrence Erlbaum. Goswami, U. (2001). Analogical reasoning in children. In D. Gentner, K. J. Holyoak & B. N. Kokinov (Eds.), The analogical mind: Perspectives from cognitive science (pp. 437-470). Cambridge, MA: MIT Press. Halford, G. S. (1993). Children’s understanding: the development of mental models. Hillsdale, NJ: Erlbaum. Hayes, B. K., & Thompson, S. P. (2007). Causal relations and feature similarity in children’s inductive reasoning. Journal of Experimental Psychology, 136 (3), 470-484. Heit, E., & Hayes, B. K. (2005). Relations between categorization, induction, recognition, and similarity: Comment on Sloutsky and Fisher (2004). Journal of Experimental Psychology: General, 134, 596-605. Inagaki, K., & Hatano, G. (1996). Young children’s recognition of commonalities between animals and plants. Child Development, 67(6), 2823-2840. Inhelder, B., & Piaget, J. (1958). The growth of logical thinking from childhood to adolescence. New York: Basic Books. Jones, S., & Smith, L. B. (1993). The place of perception in children’s concepts. Cognitive Development, 8, 113-139. Medin, D., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.), Similarity and analogical learning (pp.179-195). New York: Cambridge University Press. Opfer, J. E., & Bulloch, M. J. (2007). Causal relations drive young children’s induction, naming, and categorization. Cognition, 105, 206-217. Osherson, D. N., Smith, E. E., Wilkie, O., López, A., & Shafir, E. (1990). Category-based induction. Psychological Review, 97, 185-200. Perceptual to category transition 27 Piaget, J. (1964). Part I: Cognitive development in children: Piaget development and learning. Journal of Research in Science Teaching, 2(3), 176-186. Rakinson, D. H. (2000). When a rose is just a rose: The illusion of taxonomies in infants’ categorization. Infancy, 1, 77-90. Rakinson, D. H., & Hahn, E. (2004). The mechanisms of early categorization and induction: smart or dumb Infants? In R. Kail (Ed.), Advances in child development and behavior (pp. 281-322). New York: Academic Press. Rattermann, M. J., & Gentner, D. (1998). More evidence for a relational shift in the development of analogy: children’s performance on a causal-mapping task. Cognitive Development, 13, 453-478. Richland, L. E., Morrison, R. G., & Holyoak, K. J. (2006). Children’s development of analogical reasoning: Insights from scene analogy problems. Journal of Experimental Child Psychology, 94, 249-273. Sloman, S. A. (1993). Feature-based induction. Cognitive Psychology, 25, 231-280. Sloutsky, V. M., & Fisher, A. V. (2004). Induction and categorization in young children: a similarity-based model. Journal of Experimental Psychology: General, 133, 166-188. Sloutsky, V. M., Kloos, H., & Fisher, A. V. (2007). When looks are everything: Appearance similarity versus kind information in early induction. Psychological Science, 18 (2), 179- 185. Waxman, S. R., Lynch, E. B., Casey, K. L., & Baer, L. (1997). Setters and Samoyeds: The emergence of subordinate level categories as a basis for inductive inference. Developmental Psychology, 33, 1074-1090. Perceptual to category transition 28 Tables Table 1 Mean scores for the percentage of perceptual choices during LSD and HSD trials, for each year group Year Group Mean % LSD choices (SD) Mean % HSD choices (SD) Reception 69% (4.79) 91% (2.81) Year 1 74% (3.81) 99% (.27) Year 2 67% (3.49) 99% (.29) Table 2 Mean scores for the initial and final categorization tasks for each year group Year Group Mean initial score (SD) Mean final score (SD) Nursery 96% (4.00) 97% (3.10) Reception 98% (4.00) 97% (4.53) Year 1 99% (2.65) 98% (4.55) Year 2 99% (3.02) 98% (4.02) Year 3 99% (1.44) 99% (1.00) Year 4 100% (0.00) 99% (0.73) Perceptual to category transition 29 Table 3 The percentage of category induction choices for each year group. Year Group Mean initial score (SD) Nursery 26% (28.10) Reception 43% (43.60) Year 1 61% (38.14) Year 2 72% (37.57) Year 3 71% (38.07) Year 4 75% (37.64) Perceptual to category transition 30 Figure Captions Figure 1a. The juvenile to adult transformation: in the category learning and the initial and final categorization tasks, participants completed randomized trials showing juvenile to adult transformations and were asked whether each juvenile and adult was a Sandbug or a Rockbug. Figure 1b. An example of an induction trial with a High Similarity Distractor: the juvenile is transformed into an adult target (transformation), then the adult target is shown with a category choice (bottom figure) and a perceptual choice (top figure; induction triad). Figure 1c. An example of a category induction trial with a Low Similarity Distractor: the juvenile is transformed into an adult target (transformation), then the adult target is shown with a category choice (bottom figure) and a perceptual choice (top figure; induction triad). Figure 2. The percentage of children showing each type of bias across the six year groups. Figure 3. The mean percentage of category choices made when distractor was either a HSD or a LSD for each age group. Figure 4. The percentage of children showing each type of bias across the three year groups. Figure 5. The mean percentage of category choices made when the category choice was either a juvenile or an adult.