Regarding the complexity of additive clustering models: Comment on Lee (2001)

The additive clustering approach to modeling pairwise similarity of entities is a powerful tool for deriving featural stimulus representations. In a recent paper, Lee (2001) proposes a statistically principled measure for choosing between clustering models that accounts for model complexity as well as data fit. Importantly, complexity is understood to be a property, not merely of the number of clusters, but also their size and pattern of overlap. However, some caution is required when interpreting the measure, with regard to the applicability of the Hadamard inequality to the complexity matrix.     

Processes of similarity judgment

Similarity underlies fundamental cognitive capabilities such as memory, categorization, decision making, problem solving, and reasoning. Although recent approaches to similarity appreciate the structure of mental representations, they differ in the processes posited to operate over these representations. We present an experiment that differentiates among extant structural accounts of similarity in their ability to account for patterns of similarity ratings. These data pose a challenge for transformation-based models and all but one mapping-based model, the Similarity as Interactive Activation and Mapping (SIAM) model of similarity.     

What makes words sound similar?

Although similarity plays an important role in accounts of language processing, there are surprisingly few direct empirical studies of the phonological similarity between words, and it is therefore not clear whether similarity comparisons between words involve processes similar to those involved in other cognitive domains. In five experiments, participants chose which of two monosyllabic pseudo-words sounded more similar to a target pseudo-word. Our results are generally consistent with the structural alignment theory of comparisons between complex mental representations, suggesting that phonological word similarity parallels similarity involving other kinds of information including visual objects and scenes, events, and word meanings. We use our results to test new metrics of word similarity, and identify predictions for future similarity research both in the domain of word sounds and in other cognitive domains.     

Physical similarity (and not quantity representation) drives perceptual comparison of numbers: evidence from two Indian notations

Numerical quantity seems to affect the response in any task that involves numbers, even in tasks that do not demand access to quantity (e.g., perceptual tasks). That is, readers seem to activate quantity representations upon the mere presentation of integers. One important piece of evidence in favor of this view comes from the finding of a distance effect in perceptual tasks: When one compares two numbers, response times (RTs) are a function of the numerical distance between them. However, recent studies have suggested that the physical similarity between Arabic numbers is strongly correlated with their numerical distance, and that the former could be a better predictor of RT data in perceptual tasks in which magnitude processing is not required (Cohen, 2009a). The present study explored the Persian and Arabic versions of Indian numbers (Exps. 1 and 2, respectively). Na?ve participants (speakers of Spanish) and users of these notations (Pakistanis and Jordanians) participated in a physical same–different matching task. The RTs of users of the Indian notations were regressed on perceptual similarity (estimated from the Spanish participants’ RTs) and numerical distance. The results showed that, regardless of the degree of correlation between the perceptual similarity function and the numerical distance function, the critical predictor for RTs was perceptual similarity. Thus, participants do not automatically activate Indian integers’ quantity representations, at least not when these numbers are presented in simple perceptual tasks.     

Exploring the breadth of the top-down representations that control attentional disengagement

Recent studies attribute top-down control as the primary determinant of the speed with which attention can be disengaged from an object, and disengagement has received increased interest as a means to distinguish between top-down and bottom-up accounts of attention capture. We present the results of three experiments exploring the breadth of the representations that delay attentional disengagement based on top-down goals. Experiments 1 and 2 examined whether objects similar to an observers' search target, but differing in luminance and hue, delayed the reallocation of attention in a search paradigm designed to isolate disengagement time. Experiment 3 explored whether the representations that delay disengagement are based on absolute similarity with the search target, or are tuned based on target/nontarget relationships. These three studies confirmed the role of top-down goals in automatically contributing to dwell times and revealed that the representations that underlie disengagement effects are broad (automatically delaying disengagement for items similar, but not identical to, the search target). In some cases, attention sets appeared to be graded in nature, but in others target–distractor relationships influenced the degree to which an irrelevant item held attention. Implications for theories of attention capture and potential functional significance of these automatic effects are discussed.     

What can the same–different task tell us about the development of magnitude representations?

We examined the development of magnitude representations in children (Exp 1: kindergartners, first-, second- and sixth graders, Exp 2: kindergartners, first-, second- and third graders) using a numerical same-different task with symbolic (i.e. digits) and non-symbolic (i.e. arrays of dots) stimuli. We investigated whether judgments in a same-different task with digits are based upon the numerical value or upon the physical similarity of the digits. In addition, we investigated whether the numerical distance effect decreases with increasing age. Finally, we examined whether the performance in this task is related to general mathematics achievement. Our results reveal that a same-different task with digits is not an appropriate task to study magnitude representations, because already late kindergarteners base their responses on the physical similarity instead of the numerical value of the digits. When decisions cannot be made on the basis of physical similarity, a similar numerical distance effect is present over all age groups. This suggests that the magnitude representation is stable from late kindergarten onwards. The size of the numerical distance effect was not related to mathematical achievement. However, children with a poorer mathematics achievement score seemed to have more difficulties to link a symbol with its corresponding magnitude.     

The Role of Co-Occurrence Statistics in Developing Semantic Knowledge

The organization of our knowledge about the world into an interconnected network of concepts linked by relations profoundly impacts many facets of cognition, including attention, memory retrieval, reasoning, and learning. It is therefore crucial to understand how organized semantic representations are acquired. The present experiment investigated the contributions of readily observable environmental statistical regularities to semantic organization in childhood. Specifically, we investigated whether co-occurrence regularities with which entities or their labels more reliably occur together than with others (a) contribute to relations between concepts independently and (b) contribute to relations between concepts belonging to the same taxonomic category. Using child-directed speech corpora to estimate reliable co-occurrences between labels for familiar items, we constructed triads consisting of a target, a related distractor, and an unrelated distractor in which targets and related distractors consistently co-occurred (e.g., sock-foot), belonged to the same taxonomic category (e.g., sock-coat), or both (e.g., sock-shoe). We used an implicit, eye-gaze measure of relations between concepts based on the degree to which children (N = 72, age 4–7 years) looked at related versus unrelated distractors when asked to look for a target. The results indicated that co-occurrence both independently contributes to relations between concepts and contributes to relations between concepts belonging to the same taxonomic category. These findings suggest that sensitivity to the regularity with which different entities co-occur in children's environments shapes the organization of semantic knowledge during development. Implications for theoretical accounts and empirical investigations of semantic organization are discussed.     

Similarity, kernels, and the triangle inequality

Similarity is used as an explanatory construct throughout psychology and multidimensional scaling (MDS) is the most popular way to assess similarity. In MDS, similarity is intimately connected to the idea of a geometric representation of stimuli in a perceptual space. Whilst connecting similarity and closeness of stimuli in a geometric representation may be intuitively plausible, Tversky and Gati [Tversky, A., & Gati, I. (1982). Similarity, separability, and the triangle inequality. Psychological Review, 89(2), 123-154] have reported data which are inconsistent with the usual geometric representations that are based on segmental additivity. We show that similarity measures based on Shepard’s universal law of generalization [Shepard, R. N. (1987). Toward a universal law of generalization for psychologica science. Science, 237(4820), 1317-1323] lead to an inner product representation in a reproducing kernel Hilbert space. In such a space stimuli are represented by their similarity to all other stimuli. This representation, based on Shepard’s law, has a natural metric that does not have additive segments whilst still retaining the intuitive notion of connecting similarity and distance between stimuli. Furthermore, this representation has the psychologically appealing property that the distance between stimuli is bounded.     

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.     

Categorical <Emphasis Type="Italic">and</Emphasis> associative relations increase false memory relative to purely associative relations

The goal of the present study was to examine the contributions of associative strength and similarity in terms of shared features to the production of false memories in the Deese/Roediger–McDermott list-learning paradigm. Whereas the activation/monitoring account suggests that false memories are driven by automatic associative activation from list items to nonpresented lures, combined with errors in source monitoring, other accounts (e.g., fuzzy trace theory, global-matching models) emphasize the importance of semantic-level similarity, and thus predict that shared features between list and lure items will increase false memory. Participants studied lists of nine items related to a nonpresented lure. Half of the lists consisted of items that were associated but did not share features with the lure, and the other half included items that were equally associated but also shared features with the lure (in many cases, these were taxonomically related items). The two types of lists were carefully matched in terms of a variety of lexical and semantic factors, and the same lures were used across list types. In two experiments, false recognition of the critical lures was greater following the study of lists that shared features with the critical lure, suggesting that similarity at a categorical or taxonomic level contributes to false memory above and beyond associative strength. We refer to this phenomenon as a “feature boost” that reflects additive effects of shared meaning and association strength and is generally consistent with accounts of false memory that have emphasized thematic or feature-level similarity among studied and nonstudied representations.     

The comprehension of anomalous sentences: evidence from structural priming

We report three experiments investigating how people process anomalous sentences, in particular those in which the anomaly is associated with the verb. We contrast two accounts for the processing of such anomalous sentences: a syntactic account, in which the representations constructed for anomalous sentences are similar in nature to the ones constructed for well-formed sentences; and a semantic account, in which the representations constructed for anomalous sentences are erroneous, or altogether missing, and interpretation is achieved on the basis of semantic representations instead. To distinguish between these accounts, we used structural priming. First, we ruled out the possibility that anomaly per se influences the magnitude of the priming effect: Prime sentences with morphologically incorrect verbs produced similarly enhanced priming (lexical boost) to sentences with the same correct verbs (Exp. 1). Second, we found that prime sentences with a novel verb (Exp. 2) or a semantically and syntactically incongruent verb (Exp. 3) produced a priming effect, which was the same as that produced by well-formed sentences. In accord with the syntactic account, we conclude that the syntactic representations of anomalous sentences are similar to those constructed for well-formed sentences. Our results furthermore suggest that lexically-independent syntactic information is robust enough to produce well-formed syntactic representations during processing without requiring aid from lexically-based syntactic information.     

A study of metrics of distance and correlation between ranked lists for compositionality detection

Compositionality in language refers to how much the meaning of some phrase can be decomposed into the meaning of its constituents and the way these constituents are combined. Based on the premise that substitution by synonyms is meaning-preserving, compositionality can be approximated as the semantic similarity between a phrase and a version of that phrase where words have been replaced by their synonyms. Different ways of representing such phrases exist (e.g., vectors (Kiela and Clark, 2013) or language models (Lioma, Simonsen, Larsen, and Hansen, 2015)), and the choice of representation affects the measurement of semantic similarity.We propose a new compositionality detection method that represents phrases as ranked lists of term weights. Our method approximates the semantic similarity between two ranked list representations using a range of well-known distance and correlation metrics. In contrast to most state-of-the-art approaches in compositionality detection, our method is completely unsupervised. Experiments with a publicly available dataset of 1048 human-annotated phrases shows that, compared to strong supervised baselines, our approach provides superior measurement of compositionality using any of the distance and correlation metrics considered.     

Theories of artificial grammar learning

Artificial grammar learning (AGL) is one of the most commonly used paradigms for the study of implicit learning and the contrast between rules, similarity, and associative learning. Despite five decades of extensive research, however, a satisfactory theoretical consensus has not been forthcoming. Theoretical accounts of AGL are reviewed, together with relevant human experimental and neuroscience data. The author concludes that satisfactory understanding of AGL requires (a) an understanding of implicit knowledge as knowledge that is not consciously activated at the time of a cognitive operation; this could be because the corresponding representations are impoverished or they cannot be concurrently supported in working memory with other representations or operations, and (b) adopting a frequency-independent view of rule knowledge and contrasting rule knowledge with specific similarity and associative learning (co-occurrence) knowledge.     

Evaluating the random representation assumption of lexical semantics in cognitive models

A common assumption implicit in cognitive models is that lexical semantics can be approximated by using randomly generated representations to stand in for word meaning. However, the use of random representations contains the hidden assumption that semantic similarity is symmetrically distributed across randomly selected words or between instances within a semantic category. We evaluated this assumption by computing similarity distributions for randomly selected words from a number of well-known semantic measures and comparing them with the distributions from random representations commonly used in cognitive models. The similarity distributions from all semantic measures were positively skewed compared with the symmetric normal distributions assumed by random representations. We discuss potential consequences that this false assumption may have for conclusions drawn from process models that use random representations.     

Similarity chains in the transformational paradigm

Rips’ (1989) results with the transformational paradigm have often been cited as supporting accounts of categorisation not based on similarity, such as involving necessary or sufficient features (or a belief in such features), which guarantee a categorisation outcome once their presence has been established. We discuss a similarity account of the transformational paradigm based on similarity chains, which predicts that when the transformation is more gradual the identity of the transformed object is less likely to change. Conversely, we suggest that an essentialist approach to categorisation predicts that essences are more likely to change in gradual transformations, across generations, as is the case with evolutionary change of species. In two experiments we examined the scope of the similarity versus the essentialist account in the transformational paradigm. With space aliens, the similarity account was superior to the essentialist one, but the converse was true with earth creatures. We suggest that an essentialist mode of categorisation is more likely than a similarity one for stimuli that are in better correspondence with our naïve understanding of the world.     

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.     

Can We Make Sense of the Notion of Trustworthy Technology?

In this paper we raise the question whether technological artifacts can properly speaking be trusted or said to be trustworthy. First, we set out some prevalent accounts of trust and trustworthiness and explain how they compare with the engineer’s notion of reliability. We distinguish between pure rational-choice accounts of trust, which do not differ in principle from mere judgments of reliability, and what we call “motivation-attributing” accounts of trust, which attribute specific motivations to trustworthy entities. Then we consider some examples of technological entities that are, at first glance, best suited to serve as the objects of trust: intelligent systems that interact with users, and complex socio-technical systems. We conclude that the motivation-attributing concept of trustworthiness cannot be straightforwardly applied to these entities. Any applicable notion of trustworthy technology would have to depart significantly from the full-blown notion of trustworthiness associated with interpersonal trust.     

Dynamical Models of Sentence Processing

We suggest that the theory of dynamical systems provides a revealing general framework for modeling the representations and mechanism underlying syntactic processing. We show how a particular dynamical model, the Visitation Set Gravitation model of Tabor, Juliano, and Tanenhaus (1997), develops syntactic representations and models a set of contingent frequency effects in parsing that are problematic for other models. We also present new simulations showing how the model accounts for semantic effects in parsing, and propose a new account of the distinction between syntactic and semantic incongruity. The results show how symbolic structures useful in parsing arise as emergent properties of connectionist dynamical systems.     

Evidence for working memory storage operations in perceptual cortex

Isolating the short-term storage component of working memory (WM) from the myriad of associated executive processes has been an enduring challenge. Recent efforts have identified patterns of activity in visual regions that contain information about items being held in WM. However, it remains unclear (1) whether these representations withstand intervening sensory input and (2) how communication between multimodal association cortex and the unimodal perceptual regions supporting WM representations is involved in WM storage. We present evidence that the features of a face held in WM are stored within face-processing regions, that these representations persist across subsequent sensory input, and that information about the match between sensory input and a memory representation is relayed forward from perceptual to prefrontal regions. Participants were presented with a series of probe faces and indicated whether each probe matched a target face held in WM. We parametrically varied the feature similarity between the probe and target faces. Activity within face-processing regions scaled linearly with the degree of feature similarity between the probe face and the features of the target face, suggesting that the features of the target face were stored in these regions. Furthermore, directed connectivity measures revealed that the direction of information flow that was optimal for performance was from sensory regions that stored the features of the target face to dorsal prefrontal regions, supporting the notion that sensory input is compared to representations stored within perceptual regions and is subsequently relayed forward. Together, these findings indicate that WM storage operations are carried out within perceptual cortex.