The benefits and costs of comparisons in a novel object categorization task: Interactions with development

How children generalize novel nouns: An eye-tracking analysis of their generalization strategies

Recent research has shown that comparisons of multiple learning stimuli which are associated with the same novel noun favor taxonomic generalization of this noun. These findings contrast with single-stimulus learning in which children follow so-called lexical biases. However, little is known about the underlying search strategies. The present experiment provides an eye-tracking analysis of search strategies during novel word learning in a comparison design. We manipulated both the conceptual distance between the two learning items, i.e., children saw examples which were associated with a noun (e.g., the two learning items were either two bracelets in a "close" comparison condition or a bracelet and a watch in a "far" comparison condition), and the conceptual distance between the learning items and the taxonomically related items in the generalization options (e.g., the taxonomic generalization answer; a pendant, a near generalization item; versus a bow tie, a distant generalization item). We tested 5-, 6- and 8-year-old children's taxonomic (versus perceptual and thematic) generalization of novel names for objects. The search patterns showed that participants first focused on the learning items and then compared them with each of the possible choices. They also spent less time comparing the various options with one another; this search profile remained stable across age groups. Data also revealed that early comparisons, (i.e., reflecting alignment strategies) predicted generalization performance. We discuss four search strategies as well as the effect of age and conceptual distance on these strategies.

Children's generalization of novel names in comparison settings: The role of semantic distance during learning and at test

There is considerable evidence showing that, in novel noun learning and generalization tasks, comparisons of several learning stimuli lead to more taxonomically based generalizations of novel nouns than single stimulus presentations. This study investigated the role of semantic distance (close vs. far) between learning examples and between learning examples and transfer items (near vs. distant) in comparison designs. In two experiments, we investigated the case of object nouns (e.g., foods, Experiment 1) and relational nouns (e.g., is the cutter for, Experiment 2) in 4- to 6-year-old and 3- to 4-year-old children, respectively. As predicted, the comparison conditions led to better results than the no-comparison conditions. In comparison conditions, far training items and near generalization items gave the best performance. Semantic distance effects are discussed in terms of abstracted representations during learning as well as in terms of cognitive constraints on generalization. It is argued that both object nouns and relational nouns are construed in the light of the type of example used during learning (i.e., single or multiple). Depending on the distance between learning and generalization items, children build different categories and are more or less likely to accept distant referents.

Cleaner wrasse Labroides dimidiatus perform above chance in a "matching-to-sample" experiment

Concept learning have been studied widely in non-human animal species within or not an ecological context. Here we tested whether cleaner fish Labroides dimidiatus, which show generalised rule learning in an ecologically relevant context; they generalise that any predator may provide protection from being chased by other fish; can also learn a general concept when presented with abstract cues. We tested for this ability in the matching-to-sample task. In this task, a sample is shown first, and then the subject needs to choose the matching sample over a simultaneously presented different one in order to obtain a food reward. We used the most general form of the task, using each stimulus only once in a total of 200 trials. As a group, the six subjects performed above chance, and four individuals eventually reached learning criteria. However, individual performance was rather unstable, yielding overall only 57% correct choices. These results add to the growing literature that ectotherms show the ability of abstract concept learning, though the lack of stable high performance may indicate quantitative performance differences to endotherms.

A Developmental Perspective on Young Children's Understandings of Paired Graphics Conventions From an Analogy Task

The present study investigated children’s understanding development of multiple graphics, here paired conventions commonly used in primary school textbooks. Paired graphics depicting everyday objects familiar to the children were used as the basis for an analogy task that tested their comprehension of five graphics conventions. This task required participants to compare pictures in a base pair in order to complete a target pair by choosing the correct picture from five alternative possibilities. Four groups of children aged 5, 6, 8, and 10 years old respectively (total N = 105), completed 45 analogy task items built around nine conceptual domains. Results showed mainly an overall increase of comprehension performance with age for all the tested conventions. There were also differences between the five conventions and an interaction between age and convention type. Further, children’s explanation of the conventions (justification of the choices in the analogy task) were also analyzed. This investigation showed the analogy task answers were a more reliable measure of the “actual” level of understanding of the conventions than the justification themselves. The findings show that younger students tried to actively compare the pictures of the pairs and to search for a relevant meaning of the pairs, however, the youngest children have a limited capacity to interpret paired graphic conventions and our results suggests that this aspect of graphic conventions develops slowly but effectively over the course of children’s schooling. Because “graphicacy” knowledge and skills are not typically taught in primary school classrooms (in contrast with literacy and numeracy), its development is likely acquired incidentally with increasing exposure to varied paired graphics during primary school education. Given the high reliance of today’s educational resources on graphics-based explanations, the results from this study may signal a need for (i) for more attention to learning graphics conventions (and more generally to graphics explanations) from teachers in primary school and (ii) for a better design of the graphics with their contextual accompanying texts and captions, from designers.

Do typically and atypically developing children learn and generalize novel names similarly: The role of conceptual distance during learning and at test

There is a large body of evidence showing that comparison of multiple stimuli leads to better conceptualization and generalization of novel names than no-comparison settings in typically developing (TD) children. By contrast, the evidence regarding this issue remains scarce in children with intellectual disabilities (ID). Children with intellectual disabilities (ID) and TD children matched on mental age with the Raven's coloured progressive matrices were tested in several novel name learning comparison conditions, with familiar objects. We manipulated the conceptual distance between the learning stimuli in the learning phase and between the learning and generalization phase stimuli for object and relational nouns. Results showed that both populations had rather similar performance profile when matched on their cognitive skills (low- vs. high-functioning). Unexpectedly, ID children's performance was equivalent for relations and better for objects compared to their TD peers' performance. However, when controlling for chronological age, the difference between ID and TD children disappeared in the case of object categories and was better understood by TD children in the case of relations. We discuss the role of conceptual distance on participants' conceptual generalization as a function of their intellectual abilities and cognitive functioning.

Children prefer pattern over shape during complex categorization

Numerous studies have shown that children tend to view objects with similar shapes as having the same category. However, these studies often adopt simple categorization tasks and ignore the perceptual dimension (e.g., surface pattern of objects) that likely attract children's attention. The purpose of this study was to test how children categorize when pattern competes against shape. In Experiment 1a children were presented with a target and several testing objects that shared the same shape, color, or texture as the target. The results indicated that children preferentially selected the shape-sharing objects. However, when the texture was replaced by pattern (Experiment 1b), there was no significant difference between shape and pattern choices. When shared features were intricately overlapped between different pairs of stimuli (Experiment 2), children preferentially chose objects that shared patterns over those that shared shapes. These findings are the first to reveal children's pattern preference in categorization, supporting the view that children's categorization is flexible.

Generating explanations via analogical comparison

Generating explanations can be highly effective in promoting learning in both adults and children. Our interest is in the mechanisms that underlie this effect and in whether and how they operate in early learning. In adult reasoning, explanation may call on many subprocesses-including comparison, counterfactual reasoning, and reasoning by exclusion; but it is unlikely that all these processes are available to young children. We propose that one process that may serve both children and adults is comparison. In this study, we asked whether children would use the results of a comparison experience when asked to explain why a model skyscraper was stable. We focused on a challenging principle-that diagonal cross-bracing lends stability to physical structures (Gentner et al., Cognitive Science, 40, 224-240, 2016). Six-year-olds either received no training or interacted with model skyscrapers in one of three different conditions, designed to vary in their potential to invite and support comparison. In the Single Model condition, children interacted with a single braced model. In the comparison conditions (Low Alignability and High Alignability), children compared braced and unbraced models. Following experience with the models, children were asked to explain why the braced model was stable. They then received two transfer tasks. We found that children who received highly alignable pairs were most likely to (a) produce brace-based explanations and (b) transfer the brace principle to a dissimilar context. This provides evidence that children can benefit from analogical comparison in generating explanations and also suggests limitations on this ability.

Analogical reasoning in children with specific language impairment: Evidence from a scene analogy task

Analogical reasoning is a human ability that maps systems of relations. It develops along with relational knowledge, working memory and executive functions such as inhibition. It also maintains a mutual influence on language development. Some authors have taken a greater interest in the analogical reasoning ability of children with language disorders, specifically those with specific language impairment (SLI). These children apparently have weaker analogical reasoning abilities than their aged-matched peers without language disorders. Following cognitive theories of language acquisition, this deficit could be one of the causes of language disorders in SLI, especially those concerning productivity. To confirm this deficit and its link to language disorders, we use a scene analogy task to evaluate the analogical performance of SLI children and compare them to controls of the same age and linguistic abilities. Results show that children with SLI perform worse than age-matched peers, but similar to language-matched peers. They are more influenced by increased task difficulty. The association between language disorders and analogical reasoning in SLI can be confirmed. The hypothesis of limited processing capacity in SLI is also being considered.

Young children's learning of relational categories: multiple comparisons and their cognitive constraints

Relational categories are notoriously difficult to learn because they are not defined by intrinsic stable properties. We studied the impact of comparisons on relational concept learning with a novel word learning task in 42-month-old children. Capitalizing on Gentner et al. (2011), two, three or four pairs of stimuli were introduced with a novel relational word. In a given trial, the set of pairs was composed of either close or far pairs (e.g., close pair: knife1-watermelon, knife2-orange, knife3-slice of bread and knife4-meat; far pair: ax-evergreen tree, saw-log, cutter-cardboard, and knife-slice of bread, for the “cutter for” relation). Close pairs (2 vs. 3 vs. 4 pairs) led to random generalizations whereas comparisons with far pairs gave the expected relational generalization. The 3 pair case gave the best results. It is argued that far pairs promote deeper comparisons than close pairs. As shown by a control experiment, this was the case only when far pairs display well known associations.

Revisiting Wittgenstein's puzzle: hierarchical encoding and comparison facilitate learning of probabilistic relational categories

Kittur et al. (2004, 2006) and Jung and Hummel (2011, 2014) showed that people have great difficulty learning relation-based categories with a probabilistic (i.e., family resemblance) structure, in which no single relation is shared by all members of a category. Yet acquisition of such categories is not strictly impossible: in all these studies, roughly half the participants eventually learned to criterion. What are these participants doing that the other half are not? We hypothesized that successful participants were those who divided the nominal categories into two or more sub-categories, each of which individually had a deterministic structure. We report three experiments testing this hypothesis: explicitly presenting participants with hierarchical (category and sub-category) structures facilitated the acquisition of otherwise probabilistic relational categories, but only when participants learned the subordinate-level (i.e., deterministic) categories prior to learning the nominal (i.e., probabilistic) categories and only when they were permitted to view multiple exemplars of the same category simultaneously. These findings suggest that one way to learn natural relational categories with a probabilistic structure [e.g., Wittgenstein’s (1953), category game, or even mother] is by learning deterministic subordinate-level concepts first and connecting them together under a common concept or label. They also add to the literature suggesting that comparison of multiple exemplars plays an instrumental role in relational learning.