Evidence from machines that learn and think like people.

We agree with Lake et al.'s trenchant analysis of deep learning systems, including that they are highly brittle and that they need vastly more examples than do people. We also agree that human cognition relies heavily on structured relational representations. However, we differ in our analysis of human cognitive processing. We argue that (1) analogical comparison processes are central to human cognition; and (2) intuitive physical knowledge is captured by qualitative representations, rather than quantitative simulations.

Prelinguistic Relational Concepts: Investigating Analogical Processing in Infants.

This research asks whether analogical processing ability is present in human infants, using the simplest and most basic relation-the same-different relation. Experiment 1 (N = 26) tested whether 7- and 9-month-olds spontaneously detect and generalize these relations from a single example, as previous research has suggested. The attempted replication failed. Experiment 2 asked whether infants could abstract the relation via analogical processing (Experiment 2, N = 64). Indeed, with four exemplars, 7- and 9-month-olds could abstract the same-different relation and generalize it to novel pairs. Furthermore, prior experience with the objects disrupted learning. Facilitation from multiple exemplars and disruption by individual object salience are signatures of analogical learning. These results indicate that analogical ability is present by 7 months.

Finding faults: analogical comparison supports spatial concept learning in geoscience.

A central issue in education is how to support the spatial thinking involved in learning science, technology, engineering, and mathematics (STEM). We investigated whether and how the cognitive process of analogical comparison supports learning of a basic spatial concept in geoscience, fault. Because of the high variability in the appearance of faults, it may be difficult for students to learn the category-relevant spatial structure. There is abundant evidence that comparing analogous examples can help students gain insight into important category-defining features (Gentner in Cogn Sci 34(5):752-775, 2010). Further, comparing high-similarity pairs can be especially effective at revealing key differences (Sagi et al. 2012). Across three experiments, we tested whether comparison of visually similar contrasting examples would help students learn the fault concept. Our main findings were that participants performed better at identifying faults when they (1) compared contrasting (fault/no fault) cases versus viewing each case separately (Experiment 1), (2) compared similar as opposed to dissimilar contrasting cases early in learning (Experiment 2), and (3) viewed a contrasting pair of schematic block diagrams as opposed to a single block diagram of a fault as part of an instructional text (Experiment 3). These results suggest that comparison of visually similar contrasting cases helped distinguish category-relevant from category-irrelevant features for participants. When such comparisons occurred early in learning, participants were more likely to form an accurate conceptual representation. Thus, analogical comparison of images may provide one powerful way to enhance spatial learning in geoscience and other STEM disciplines.

Children's Early Spontaneous Comparisons Predict Later Analogical Reasoning Skills: An Investigation of Parental Influence.

Laboratory studies have demonstrated beneficial effects of making comparisons on children's analogical reasoning skills. We extend this finding to an observational dataset comprising 42 children. The prevalence of specific comparisons, which identify a feature of similarity or difference, in children's spontaneous speech from 14-58 months is associated with higher scores in tests of verbal and non-verbal analogy in 6th grade. We test two pre-registered hypotheses about how parents influence children's production of specific comparisons: 1) via modelling, where parents produce specific comparisons during the sessions prior to child onset of this behaviour; 2) via responsiveness, where parents respond to their children's earliest specific comparisons in variably engaged ways. We do not find that parent modelling or responsiveness predicts children's production of specific comparisons. However, one of our pre-registered control analyses suggests that parents' global comparisons-comparisons that do not identify a specific feature of similarity or difference-may bootstrap children's later production of specific comparisons, controlling for parent IQ. We present exploratory analyses following up on this finding and suggest avenues for future confirmatory research. The results illuminate a potential route by which parents' behaviour may influence children's early spontaneous comparisons and potentially their later analogical reasoning skills.

Spatial alignment supports comparison of life science images.

Visual comparisons are pervasive in science, technology, engineering, and mathematics (STEM) instruction and practice. In previous work, adults' visual comparisons of simple stimuli were faster and more accurate when the layout of a display facilitated alignment of corresponding elements-the spatial alignment principle (Matlen et al., 2020). Here, we asked whether the spatial alignment principle extends to rich, educationally relevant stimuli, and how prior experience and spatial skill relate to spatial alignment effects. Participants were asked to find an incorrect bone within a skeleton, presented individually or paired with a correct skeleton in a layout that did (direct placement) or did not (impeded placement) support alignment (Kurtz & Gentner, 2013). Consistent with the spatial alignment principle, undergraduates (Study 1) showed an advantage of direct over impeded placement. Middle schoolers (Study 2) showed a direct advantage on items presented in atypical orientations. That atypical items showed the strongest effects suggests that direct placement may help most when materials are less familiar. However, neither individual differences in undergraduates' STEM course history, nor undergraduates' or middle schoolers' spatial skills moderated spatial alignment effects. Thus, applying the spatial alignment principle in science, technology, engineering, and mathematics has potential to improve visual comparisons, especially those that are challenging, for students of all spatial skill levels. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Verb Metaphoric Extension Under Semantic Strain.

This paper explores the processes underlying verb metaphoric extension. Work on metaphor processing has largely focused on noun metaphor, despite evidence that verb metaphor is more common. Across three experiments, we collected paraphrases of simple intransitive sentences varying in semantic strain-for example, The motor complained → The engine made strange noises-and assessed the degree of meaning change for the noun and the verb. We developed a novel methodology for this assessment using word2vec. In Experiments 1 and 2, we found that (a) under semantic strain, verb meanings were more likely to be adjusted than noun meanings; (b) the degree of verb meaning adjustment-but not noun meaning adjustment-increased with semantic strain; and (c) verb meaning extension is primarily driven by online adjustment, although sense selection also plays a role. In Experiment 3, we replicated the word2vec results with an assessment using human subjects. The results further showed that nouns and verbs change meaning in qualitatively different ways, with verbs more likely to change meaning metaphorically and nouns more likely to change meaning taxonomically or metonymically. These findings bear on the origin and processing of verb metaphors and provide a link between online sentence processing and diachronic change over language evolution.

Perceptual alignment contributes to referential transparency in indirect learning.

Young children can sometimes acquire new word meanings-even for property terms-through incidental learning (e.g. Carey & Bartlett, 1978). We propose that an important support for this process is spontaneous perceptual comparison processes that lead children to notice key commonalities and differences. Specifically, we hypothesize that when the target property appears as a difference between two highly similar and alignable objects, spontaneous comparison processes operate to highlight the property. The property may then be linked to an accompanying word, even if the child has no prior intention to learn the word. To test this, we revisited the Carey and Bartlett paradigm, varying the perceptual alignability of the objects that 3- and 4-year-olds saw while hearing a novel color word, chromium. In Experiments 1 and 2, children in the High Alignment condition were able to identify chromium objects in a subsequent task, whereas those in the Low Alignment condition were not. Experiment 3 showed that direct instructions to learn the word led to a different pattern of results. Experiment 4 showed that the incidental learning persisted over delay and transferred to new objects. We conclude that perceptual alignment contributes to referential transparency and to incidental learning of word meanings. Implications for hypothesis-testing theories of word learning are discussed.

Spatial Alignment Facilitates Visual Comparison in Children.

Visual comparison is a key process in everyday learning and reasoning. Recent research has discovered the spatial alignment principle, based on the broader framework of structure-mapping theory in comparison. According to the spatial alignment principle, visual comparison is more efficient when the figures being compared are arranged in direct placement-that is, juxtaposed with parallel structural axes. In this placement, (1) the intended relational correspondences are readily apparent, and (2) the influence of potential competing correspondences is minimized. There is evidence for the spatial alignment principle in adults' visual comparison (Matlen et al., 2020). Here, we test whether it holds for children. Six- and eight-year-old children performed a same-different task over visual pairs. The results indicated that direct placement led to faster and more accurate comparison, both for concrete same-different matches (matches of both objects and relations) and for purely relational matches-evidence that the same structural alignment process holds for visual comparison in 6- and 8-year-olds as in adults. These findings have implications for learning and education.

No evidence for language benefits in infant relational learning.

Recent studies have found that infants show relational learning in the first year. Like older children, they can abstract relations such as same or different across a series of exemplars. For older children, language has a major impact on relational learning: labeling a shared relation facilitates learning, while labeling component objects can disrupt learning. Here we ask: Does language influence relational learning at 12 months? Experiment 1 (n = 64) examined the influence of a relational label on learning. Prior to the study, the infants saw three pairs of objects, all labeled "These are same" or "These are different". Experiment 2 (n = 48) examined the influence of object labels prior to the study, with three objects labeled (e.g., "This is a cup, this is a tower."). We compared the present results with those of Ferry et al. (2015), where infants abstracted same and different relations after undergoing a similar paradigm without prior labels. If the effects of language mirror those in older children, we would expect that infants given relational labels (Experiment 1) will be helped in abstracting same and different compared to infants not given labels and that infants given object labels (Experiment 2) will be hindered relative to those not given labels. We found no evidence for either prediction. In Experiment 1, infants who had heard relational labels did not benefit compared to infants who had received no labels (Ferry et al., 2015). In Experiment 2, infants who had heard object labels showed the same patterns as those in Ferry et al. (2015), suggesting that object labels had no effect. This finding is important because it highlights a key difference between the relational learning abilities of infants and those seen in older children, pointing to a protracted process by which language and relational learning become entwined.

Structure mapping and relational language support children's learning of relational categories.

Learning relational categories--whose membership is defined not by intrinsic properties but by extrinsic relations with other entities--poses a challenge to young children. The current work showed 3-, 4- to 5-, and 6-year-olds pairs of cards exemplifying familiar relations (e.g., a nest and a bird exemplifying home for) and then tested whether they could extend the relational concept to another category (e.g., choose the barn as a home for a horse). It found that children benefited from (a) hearing a (novel) category name in a relational construction and (b) comparing category members. The youngest group--3-year-olds--learned the category only when given a combination of relational language and a series of comparisons in a progressive alignment sequence.

Analogy Generation in Science Experts and Novices.

There is a critical inconsistency in the literature on analogical retrieval. On the one hand, a vast set of laboratory studies has found that people often fail to retrieve past experiences that share deep relational commonalities, even when they would be useful for reasoning about a current problem. On the other hand, historical studies and naturalistic research show clear evidence of remindings based on deep relational commonalities. Here, we examine a possible explanation for this inconsistency-namely, that remindings based on relational principles increase as a function of expertise. To test this claim, we devised a simple analogy-generation task that can be administered across a wide range of expertise. We presented common events as the bases from which to generate analogies. Although the events themselves were unrelated to geoscience, we found that when the event was explainable in terms of a causal principle that is prominent in geoscience, expert geoscientists were likely to spontaneously produce analogies from geoscience that relied on the same principle. Further, for these examples, prompts to produce causal analogies increased their frequency among nonscientists and scientists from another domain, but not among expert geoscientists (whose spontaneous causal retrieval levels were already high). In contrast, when the example was best explained by a principle outside of geoscience, all groups required prompting to produce substantial numbers of analogies based on causal principles. Overall, this pattern suggests that the spontaneous use of causal principles is characteristic of experts. We suggest that expert scientists adopt habitual patterns of encoding according to the key relational principles in their domain, and that this contributes to their propensity to spontaneously retrieve relational matches. We discuss implications for the nature of expertise and for science instruction and assessment.

The role of comparison in preschoolers' novel object categorization.

We examined the role of the comparison process and shared names on preschoolers' categorization of novel objects. In our studies, 4-year-olds were presented with novel object sets consisting of either one or two standards and two test objects: a shape match and a texture match. When children were presented with one standard, they extended the category based on shape regardless of whether the objects were named. When children were presented with two standards that shared the same texture and the objects were named with the same noun, they extended the category based on texture. The opportunity to compare two standards, in the absence of shared names, led to an attenuation of the effect of shape. These findings demonstrate that comparison plays a critical role in the categorization of novel objects and that shared names enhance this process.

Relational labeling unlocks inert knowledge.

Insightful solutions often come about by recalling a relevant prior situation-one that shares the same essential relational pattern as the current problem. Unfortunately, our memory retrievals often depend primarily on surface matches, rather than relational matches. For example, a person who is familiar with the idea of positive feedback in sound systems may fail to think of it in the context of global warming. We suggest that one reason for the failure of cross-domain relational retrieval is that relational information is typically encoded variably, in a context-dependent way. In contrast, the surface features of that context-such as objects, animals and characters-are encoded in a relatively stable way, and are therefore easier to retrieve across contexts. We propose that the use of relational language can serve to make situations' relational representations more uniform, thereby facilitating relational retrieval. In two studies, we find that providing relational labels for situations at encoding or at retrieval increased the likelihood of relational retrieval. In contrast, domain labels-labels that highlight situations' contextual features-did not reliably improve domain retrieval. We suggest that relational language allows people to retrieve knowledge that would otherwise remain inert and contributes to domain experts' insight.

Spatial alignment facilitates visual comparison.

Humans have a uniquely sophisticated ability to see past superficial features and to understand the relational structure of the world around us. This ability often requires that we compare structures, finding commonalities and differences across visual depictions that are arranged in space, such as maps, graphs, or diagrams. Although such visual comparison of relational structures is ubiquitous in classrooms, textbooks, and news media, surprisingly little is known about how to facilitate this process. Here we suggest a new principle of spatial alignment, whereby visual comparison is substantially more efficient when visuals are placed perpendicular to their structural axes, such that the matching components of the visuals are in direct alignment. In four experiments, this direct alignment led to faster and more accurate comparison than other placements of the same patterns. We discuss the spatial alignment principle in connection to broader work on relational comparison and describe its implications for design and instruction. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Analogical Comparison Promotes Theory-of-Mind Development.

Theory-of-mind (ToM) is an integral part of social cognition, but how it develops remains a critical question. There is evidence that children can gain insight into ToM through experience, including language training and explanatory interactions. But this still leaves open the question of how children gain these insights-what processes drive this learning? We propose that analogical comparison is a key mechanism in the development of ToM. In Experiment 1, children were shown true- and false-belief scenarios and prompted to engage in multiple comparisons (e.g., belief vs. world). In Experiments 2a, 2b, and 3, children saw a series of true- and false-belief events, varying in order and in their alignability. Across these experiments, we found that providing support for comparing true- and false-belief scenarios led to increased performance on false-belief tests. These findings show that analogical comparison can support ToM learning.

Cognitive Science Is and Should Be Pluralistic.

Núñez et al (2019) argue (1) that the field of Cognitive Science has failed, in that it has not arrived at a cohesive theory, and (2) that this is contrary to the intentions of the founders. Their survey of publication and citation patterns bears out the lack of a cohesive theory and also provides corroboration for (3) the concern that the field is becoming unbalanced, with psychology overweighted (Gentner, 2010). I will argue against points (1) and (2), but agree with point (3). My central claim is that cognitive science was never meant to have one unified theoretical framework, nor should it have.

The career of measurement.

Units as they exist today are highly abstract. Meters, miles, and other modern measures have no obvious basis in tangible phenomena and can be applied broadly across domains. Historical examples suggest, however, that units have not always been so abstract. Here, we examine this issue systematically. We begin by analyzing linear measures in the Oxford English Dictionary (OED) and in an ethnographic database that spans 114 cultures (HRAF). Our survey of both datasets shows, first, that early length units have mostly come from concrete sources-body parts, artifacts, events, and other tangible phenomena-and, second, that they have often been tied to particular contexts. Measurement units have thus undergone a shift from highly concrete to highly abstract. How did this shift happen? Drawing on historical surveys and case studies-as well as data from the OED and HRAF-we next propose a reconstruction of how abstract units might have evolved gradually through a series of overlapping stages. We also consider the cognitive processes that underpin this evolution-in particular, comparison. Finally, we discuss the cognitive origins of units. Units are not only slow to emerge historically, they are also slow to be acquired developmentally, and mastering them appears to have cognitive consequences. Taken together, these observations suggest that units are not inevitable intuitions, but are best thought of as culturally evolved cognitive tools. By analyzing the career of measurement in detail, we illustrate how such tools-abstract as they are today-can arise from concrete, often bodily origins.

Explanation recruits comparison in a category-learning task.

Generating explanations can be highly effective in promoting category learning; however, the underlying mechanisms are not fully understood. We propose that engaging in explanation can recruit comparison processes, and that this in turn contributes to the effectiveness of explanation in supporting category learning. Three experiments evaluated the interplay between explanation and various comparison strategies in learning artificial categories. In Experiment 1, as expected, prompting participants to explain items' category membership led to (a) higher ratings of self-reported comparison processing and (b) increased likelihood of discovering a rule underlying category membership. Indeed, prompts to explain led to more self-reported comparison than did direct prompts to compare pairs of items. Experiment 2 showed that prompts to compare all members of a particular category ("group comparison") were more effective in supporting rule learning than were pairwise comparison prompts. Experiment 3 found that group comparison (as assessed by self-report) partially mediated the relationship between explanation and category learning. These results suggest that one way in which explanation benefits category learning is by inviting comparisons in the service of identifying broad patterns.

Spatial cognition in apes and humans.

The debate on whether language influences cognition is sometimes seen as a simple dichotomy: cognitive development is governed either by innate predispositions or by influences of language and culture. In two recent papers on spatial cognition, Haun and colleagues break new ground in bringing together a comparative cognition approach with a cross-linguistic framework to arrive at a third position: that humans begin with the same spatial reference frames as our near relatives, the great apes, and diverge later owing to the influence of language and culture.

Developmental changes in children's understanding of the similarity between photographs and their referents.

In a series of three experiments, we investigated the development of children's understanding of the similarities between photographs and their referents. Based on prior work on the development of analogical understanding (e.g. Gentner & Rattermann, 1991), we suggest that the appreciation of this relation involves multiple levels. Photographs are similar to their referents both in terms of the constituent objects and in terms of the relations among these objects. We predicted that children would appreciate object similarity (whether photographs depict the same objects as in the referent scene) before they would appreciate relational similarity (whether photographs depict the objects in the same spatial positions as in the referent scene). To test this hypothesis, we presented 3-, 4-, 5-, 6-, and 7-year-old children and adults with several candidate photographs of an arrangement of objects. Participants were asked to choose which of the photographs was 'the same' as the arrangement. We manipulated the types of information the photographs preserved about the referent objects. One set of photographs did not preserve the object properties of the scene. Another set of photographs preserved the object properties of the scene, but not the relational similarity, such that the original objects were depicted but occupied different spatial positions in the arrangement. As predicted, younger children based their choices of the photographs largely on object similarity, whereas older children and adults also took relational similarity into account. Results are discussed in terms of the development of children's appreciation of different levels of similarity.