Similarity Judgment Within and Across Categories: A Comprehensive Model Comparison

Similarity is one of the most important relations humans perceive, arguably subserving category learning and categorization, generalization and discrimination, judgment and decision making, and other cognitive functions. Researchers have proposed a wide range of representations and metrics that could be at play in similarity judgment, yet have not comprehensively compared the power of these representations and metrics for predicting similarity within and across different semantic categories. We performed such a comparison by pairing nine prominent vector semantic representations with seven established similarity metrics that could operate on these representations, as well as supervised methods for dimensional weighting in the similarity function. This approach yields a factorial model structure with 126 distinct representation-metric pairs, which we tested on a novel dataset of similarity judgments between pairs of cohyponymic words in eight categories. We found that cosine similarity and Pearson correlation were the overall best performing unweighted similarity functions, and that word vectors derived from free association norms often outperformed word vectors derived from text (including those specialized for similarity). Importantly, models that used human similarity judgments to learn category-specific weights on dimensions yielded substantially better predictions than all unweighted approaches across all types of similarity functions and representations, although dimension weights did not generalize well across semantic categories, suggesting strong category context effects in similarity judgment. We discuss implications of these results for cognitive modeling and natural language processing, as well as for theories of the representations and metrics involved in similarity.     

Investigating linguistic relativity through bilingualism: the case of grammatical gender

The authors investigated linguistic relativity effects by examining the semantic effects of grammatical gender (present in Italian but absent in English) in fluent bilingual speakers as compared with monolingual speakers. In an error-induction experiment, they used responses by monolingual speakers to establish a baseline for bilingual speakers and show that gender affects the semantic substitution errors made by monolingual Italian speakers compared with monolingual English speakers. They then showed that Italian-English bilingual speakers behave like monolingual English speakers when the task is in English and like monolingual Italian speakers when the task is in Italian, hence exhibiting appropriate semantic representations for each language. These results show that for bilingual speakers there is intraspeaker relativity in semantic representations and, therefore, that gender does not have a conceptual, nonlinguistic effect. The results also have implications for models of bilingual semantic memory and processing.     

Rhesus monkeys (Macaca mulatta) immediately generalize the uncertain response

Rhesus monkeys (Macaca mulatta) have learned, like humans, to use an uncertain response adaptively under test conditions that create uncertainty, suggesting a metacognitive process by which human and nonhuman primates may monitor their confidence and alter their behavior accordingly. In this study, 4 rhesus monkeys generalized their use of the uncertain response, without additional training, to 2 familiar tasks (2-choice discrimination learning and mirror-image matching to sample) that predictably and demonstrably produce uncertainty. The monkeys were significantly less likely to use the uncertain response on trials in which the answer might be known. These results indicate that monkeys, like humans, know when they do not know and that they can learn to use a symbol as a generalized means for indicating their uncertainty.     

Conceptual and linguistic representations of kinds and classes

We investigate the hypothesis that our conceptual systems provide two formally distinct ways of representing categories by investigating the manner in which lexical nominals (e.g., tree, picnic table) and phrasal nominals (e.g., black bird, birds that like rice) are interpreted. Four experiments found that lexical nominals may be mapped onto kind representations, whereas phrasal nominals map onto class representations but not kind representations. Experiment 1 found that phrasal nominals, unlike lexical nominals, are mapped onto categories whose members need not be of a single kind. Experiments 2 and 3 found that categories named by lexical nominals enter into both class inclusion and kind hierarchies and thus support both class inclusion (is a) and kind specification (kind of) relations, whereas phrasal nominals map onto class representations which support only class inclusion relations. Experiment 4 showed that the two types of nominals represent hierarchical relations in different ways. Phrasal nominals (e.g., white bear) are mapped onto classes that have criteria of membership in addition to those specified by the class picked out by the head noun of the phrase (e.g., bear). In contrast, lexical nominals (e.g., polar bear) specify one way to meet the criteria specified by the more general kind concept (e.g., bear). Implications for the language-conceptual system interface, representation of hierarchical relations, lexicalization, and theories of conceptual combination are discussed.     

Are phonological representations of printed and spoken language isomorphic? Evidence from the restrictions on unattested onsets

Are the phonological representations of printed and spoken words isomorphic? This question is addressed by investigating the restrictions on onsets. Cross-linguistic research suggests that onsets of rising sonority are preferred to sonority plateaus, which, in turn, are preferred to sonority falls (e.g., bnif [symbol: see text] bdif [symbol: see text] lbif). Of interest is whether these grammatical preferences constrain the recognition of auditory and printed words by speakers of English--a language in which such onsets are unattested. Five experiments compare phonological lexical decision responses to nonwords, including unattested onsets, through either aural or visual presentation. Results suggest that both hearers and readers are sensitive to the phonotactics of unattested onsets. However, the phonotactic generalizations of hearers and readers differ on their scope and source. Hearers differentiated all three types of onsets (e.g., bnif, bdif, and lbif), and their behavior implicated both grammatical and statistical constraints. In contrast, readers were able to differentiate only those structures similar to attested English onsets from dissimilar structures (i.e., bnif vs. bdif or lbif), and their preferences reflected statistical knowledge alone. These findings suggest that the phonological representations informing lexical decision to spoken and printed words are not isomorphic.     

Syntactical knowledge in a case of agrammatism: Evidence from transcoding Roman and Arabic numerals

The ability of an aphasic subject with agrammatism in both comprehension and production to transcribe quantities from Roman numerals to Arabic and the reverse was investigated. Systematic errors in the transcoding processes were observed that could not be accounted for by the peculiarities of the two ideographic coding systems or by difficulties with direct transcoding rules. The results are discussed in the framework of the current debate on preserved/impaired hierarchical syntactical knowledge of agrammatic subjects. The findings paralleled the results of previous studies on the transcoding skills of agrammatics from/to alphabetic numerals to/from digital forms. In the case of this particular patient, it is therefore tentatively concluded in favor of preserved syntactical knowledge.     

The benefit of symbols: monkeys show linear, human-like, accuracy when using symbols to represent scalar value

When humans and animals estimate numbers of items, their error rate is proportional to the number. To date, however, only humans show the capacity to represent large numbers symbolically, which endows them with increased precision, especially for large numbers, and with tools for manipulating numbers. This ability depends critically on our capacity to acquire and represent explicit symbols. Here we show that when rhesus monkeys are trained to use an explicit symbol system, they too show more precise, and linear, scaling than they do using a one-to-one corresponding numerosity representation. We also found that when taught two different types of representations for reward amount, the monkeys systematically undervalued the less precise representation. The results indicate that monkeys, like humans, can learn alternative mechanisms for representing a single value scale and that performance variability and relative value depend on the distinguishability of each representation.     

The KEY to the ROCK: Near-homophony in nonnative visual word recognition

To test the hypothesis that native language (L1) phonology can affect the lexical representations of nonnative words, a visual semantic-relatedness decision task in English was given to native speakers and nonnative speakers whose L1 was Japanese or Arabic. In the critical conditions, the word pair contained a homophone or near-homophone of a semantically associated word, where a near-homophone was defined as a phonological neighbor involving a contrast absent in the speaker’s L1 (e.g., ROCK-LOCK for native speakers of Japanese). In all participant groups, homophones elicited more false positive errors and slower processing than spelling controls. In the Japanese and Arabic groups, near-homophones also induced relatively more false positives and slower processing. The results show that, even when auditory perception is not involved, recognition of nonnative words and, by implication, their lexical representations are affected by the L1 phonology.     

Seeing Like a Geologist: Bayesian Use of Expert Categories in Location Memory

Memory for spatial location is typically biased, with errors trending toward the center of a surrounding region. According to the category adjustment model (CAM), this bias reflects the optimal, Bayesian combination of fine‐grained and categorical representations of a location. However, there is disagreement about whether categories are malleable. For instance, can categories be redefined based on expert‐level conceptual knowledge? Furthermore, if expert knowledge is used, does it dominate other information sources, or is it used adaptively so as to minimize overall error, as predicted by a Bayesian framework? We address these questions using images of geological interest. The participants were experts in structural geology, organic chemistry, or English literature. Our data indicate that expertise‐based categories influence estimates of location memory—particularly when these categories better constrain errors than alternative (“novice”) categories. Results are discussed with respect to the CAM.     

Responses toward a trapped animal by wild bonobos at Wamba

Chimpanzees and bonobos are the closest living relatives of humans and diverged relatively recently in their phylogenetic history. However, a number of reports have suggested behavioral discrepancies between the two Pan species, such as more cooperative and tolerant social interaction and poorer tool-using repertoires in bonobos. Concerning hunting behavior and meat consumption, recent studies from the field have confirmed both behaviors not only in chimpanzees but also in bonobos. The present study reports an encounter by wild bonobos at Wamba with a duiker trapped in a snare. Bonobos interacted with the live duiker for about 10 min but did not eventually kill the animal. They showed fear responses when the duiker moved and exhibited behaviors related to anxiety and stress such as branch-drag displays and self-scratching. Although bonobos manipulated nearby saplings and parts of the snare, they did not use detached objects to make indirect contact with the duiker. Juveniles and adults of both sexes engaged in active interactions with the trapped duiker. Overall, bonobos’ behavioral responses indicated species-specific cognitive characteristics largely different from those of chimpanzees.     

Exemplars, prototypes, similarities, and rules in category representation: an example of hierarchical bayesian analysis

This article demonstrates the potential of using hierarchical Bayesian methods to relate models and data in the cognitive sciences. This is done using a worked example that considers an existing model of category representation, the Varying Abstraction Model (VAM), which attempts to infer the representations people use from their behavior in category learning tasks. The VAM allows for a wide variety of category representations to be inferred, but this article shows how a hierarchical Bayesian analysis can provide a unifying explanation of the representational possibilities using 2 parameters. One parameter controls the emphasis on abstraction in category representations, and the other controls the emphasis on similarity. Using 30 previously published data sets, this work shows how inferences about these parameters, and about the category representations they generate, can be used to evaluate data in terms of the ongoing exemplar versus prototype and similarity versus rules debates in the literature. Using this concrete example, this article emphasizes the advantages of hierarchical Bayesian models in converting model selection problems to parameter estimation problems, and providing one way of specifying theoretically based priors for competing models.     

Structure of the Indonesian Emotion Lexicon

Based on a prototype approach to emotion concepts, two studies were conducted: (1) to identify the mental state words that Indonesian speakers are most certain name emotions ( perasaan hati ) and (2) to map the hierarchical and family-resemblance structure of the top 124 emotion concepts. As in an earlier study of emotion terms in American English (Shaver, Schwartz, Kirson, & O'Connor, 1987), cluster analysis of sorting data collected in Indonesia revealed five basic-level emotion categories: cinta (love), senang (happiness), marah (anger), kawatir/takut (anxiety/fear), and sedih (sadness). Also in line with the American results, the five basic-level categories formed two large categories at the superordinate level: positive emotions and negative emotions. Each of the five basic-level categories contained several subordinate-level categories, totaling 31 in all. The results suggest that the emotion lexicons, and corresponding conceptualizations of the emotion domain, in Indonesia and the U.S.A. are similar at the superordinate and basic levels but somewhat variable at the subordinate level. This outcome – like other kinds of psychological research on emotions and emotion concepts – suggests that the gross structure of representations of the emotion domain are similar worldwide, perhaps for biological reasons, but that different cultures make different fine-grained distinctions and emphasize different subordinate-level emotion concepts.     

Event structure in perception and conception

Events can be understood in terms of their temporal structure. The authors first draw on several bodies of research to construct an analysis of how people use event structure in perception, understanding, planning, and action. Philosophy provides a grounding for the basic units of events and actions. Perceptual psychology provides an analogy to object perception: Like objects, events belong to categories, and, like objects, events have parts. These relationships generate 2 hierarchical organizations for events: taxonomies and partonomies. Event partonomies have been studied by looking at how people segment activity as it happens. Structured representations of events can relate partonomy to goal relationships and causal structure; such representations have been shown to drive narrative comprehension, memory, and planning. Computational models provide insight into how mental representations might be organized and transformed. These different approaches to event structure converge on an explanation of how multiple sources of information interact in event perception and conception.     

Conversation level syntax similarity metric

The syntax and semantics of human language can illuminate many individual psychological differences and important dimensions of social interaction. Accordingly, psychological and psycholinguistic research has begun incorporating sophisticated representations of semantic content to better understand the connection between word choice and psychological processes. In this work we introduce ConversAtion level Syntax SImilarity Metric (CASSIM), a novel method for calculating conversation-level syntax similarity. CASSIM estimates the syntax similarity between conversations by automatically generating syntactical representations of the sentences in conversation, estimating the structural differences between them, and calculating an optimized estimate of the conversation-level syntax similarity. After introducing and explaining this method, we report results from two method validation experiments (Study 1) and conduct a series of analyses with CASSIM to investigate syntax accommodation in social media discourse (Study 2). We run the same experiments using two well-known existing syntactic metrics, LSM and Coh-Metrix, and compare their results to CASSIM. Overall, our results indicate that CASSIM is able to reliably measure syntax similarity and to provide robust evidence of syntax accommodation within social media discourse.     

Typological Analysis of MMPI Personality Patterns of Drug Dependent Men

This study replicated Types I and II of Berzins, Ross, English and Haley's (1974) initial typological research with a drug dependent sample. In this research, Minnesota Multiphasic Personality Inventory (MMPI) profiles of 447 drug dependent men were submitted to Johnson's (1967) hierarchical clustering procedure. The hierarchical clustering procedure yielded nine small but clinically meaningful personality subtypes. These types were compared to Berzins's (Berzins et al.) profiles and categorized into three diagnostic classes: (a) character disorder, 42% of the sample; (b) neurotic characters, 35% of the sample; and (c) borderline disorder, 23% of the sample. The profiles were discussed with respect to the scale similarity and differences within the diagnostic class. The results were discussed, with implications for treatment and further research efforts with the drug dependent population.     

Expectancy and stimulus frequency: a comparative analysis in rats and humans.

We examined whether expectancy, one of several factors influencing attention, is similarly affected in rats and humans by manipulation of relative stimulus frequency. A two-choice reaction time (RT) task was developed for rats, and an analogous task was used for humans. Errors, RTs, discriminability, and response bias were measured. Both rats and humans shifted their response bias to the more frequent stimulus, with no change in overall discriminability. As stimulus probability or stimulus repetition increased, RTs and errors decreased. These results illustrate the similarity of expectancy in rats and humans. This two-choice RT task for rats can be used in future studies to examine the neuronal mechanisms of expectancy and attention.     

Representation at different levels in a conceptual hierarchy

The present study examines the influence of hierarchical level on category representation. Three computational models of representation – an exemplar model, a prototype model and an ideal representation model – were evaluated in their ability to account for the typicality gradient of categories at two hierarchical levels in the conceptual domain of clothes. The domain contains 20 subordinate categories (e.g., trousers, stockings and underwear) and an encompassing superordinate category (CLOTHES). The models were evaluated both in terms of their ability to fit the empirical data and their generalizability through marginal likelihood. The hierarchical level was found to clearly influence the type of representation: For concepts at the subordinate level, exemplar representations were supported. At the superordinate level, however, an ideal representation was overwhelmingly preferred over exemplar and prototype representations. This finding contributes to the increasingly dominant view that the human conceptual apparatus adopts both exemplar representations and more abstract representations, contradicting unitary approaches to categorization.     

Structural Alignment during Similarity Comparisons

Similarity comparisons are a basic component of cognition, and there are many elegant models of this process. None of these models describe comparisons of structured representations, although mounting evidence suggests that mental representations are well characterized by structured hierarchical systems of relations. We propose that structured representations can be compared via structural alignment, a process derived from models of analogical reasoning. The general prediction of structural alignment is that similarity comparisons lead subjects to attend to the matching relational structure in a pair of items. This prediction is illustrated with a computational simulation that also suggests that the strength of the relational focus is diminished when the relational match is impoverished, or when competing interpretations lead to rich object matches. These claims are tested in four experiments using the one-shot mapping paradigm, which places object similarity and relational similarity in opposition. The results support the hypothesis that similarity involves the alignment of structured representations.     

What Are You Waiting For? Real‐Time Integration of Cues for Fricatives Suggests Encapsulated Auditory Memory

Speech unfolds over time, and the cues for even a single phoneme are rarely available simultaneously. Consequently, to recognize a single phoneme, listeners must integrate material over several hundred milliseconds. Prior work contrasts two accounts: (a) a memory buffer account in which listeners accumulate auditory information in memory and only access higher level representations (i.e., lexical representations) when sufficient information has arrived; and (b) an immediate integration scheme in which lexical representations can be partially activated on the basis of early cues and then updated when more information arises. These studies have uniformly shown evidence for immediate integration for a variety of phonetic distinctions. We attempted to extend this to fricatives, a class of speech sounds which requires not only temporal integration of asynchronous cues (the frication, followed by the formant transitions 150–350 ms later), but also integration across different frequency bands and compensation for contextual factors like coarticulation. Eye movements in the visual world paradigm showed clear evidence for a memory buffer. Results were replicated in five experiments, ruling out methodological factors and tying the release of the buffer to the onset of the vowel. These findings support a general auditory account for speech by suggesting that the acoustic nature of particular speech sounds may have large effects on how they are processed. It also has major implications for theories of auditory and speech perception by raising the possibility of an encapsulated memory buffer in early auditory processing.     

Tool use induces complex and flexible plasticity of human body representations

Plasticity of body representation fundamentally underpins human tool use. Recent studies have demonstrated remarkably complex plasticity of body representation in humans, showing that such plasticity (1) occurs flexibly across multiple time scales and (2) involves multiple body representations responding differently to tool use. Such findings reveal remarkable sophistication of body plasticity in humans, suggesting that Vaesen may overestimate the similarity of such mechanisms in humans and non-human primates.