United States

A key aspect of understanding how children learn the meanings of words involves understanding how they mine different sources of information (e.g., observational, linguistic) in the service of learning. According to one dominant view, there exists a class of words (i.e., “hard words”; Gleitman et al., 2005) for which learning their meaning requires access to information beyond the observational contexts in which those words occur. By making alterations to the Human Simulation Paradigm, a paradigm commonly used for investigating vocabulary learning, this study revisits the role of observational contexts for the acquisition of one class of hard words: nouns that denote non-basic level object categories (or “hard nouns”; see Kako, 2005). These data reveal that although observational contexts may not be sufficient to yield learning of precise hard noun meanings, they allow learners to extract systematic partial knowledge, knowledge that may lay a critical foundation for meaning acquisition.

CogSci 2023

Laying the Foundation: Extracting Partial Meanings of Hard Nouns via Observational Contexts

poster

The 45th Annual Meeting of the Cognitive Science Society will take place in Sydney, Australia as a modified hybrid conference. Presentations from the main-session room will be live-streamed every day for all events. This will include all invited talks and award sessions. Recordings of all CogSci 2023 sessions will be made available for on-demand viewing after the conference.<br>
The in-person component will be held at the International Convention Centre Sydney at Darling Harbour. The Conference will take place Wednesday July 26th – Saturday July 29th, 2023 (Wednesday, July 26th is the Workshops Day).

Join us!

<br>[**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2023/07/COgSci-PAAG-V2.pdf)
<br>[**The Detailed Program can be found here**](https://online.fliphtml5.com/ebtyf/mswz/)
<br><br>

#### **CogSci 2023 Program Book**
<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/tzgk/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>
<br>

The 45th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

Our knowledge about the world is populated with categories like dog, chair, and cup. Yet much of what we understand about how we acquire this knowledge comes from studies of learning in circumstances that little resemble real-world experience. In the lab, category learning typically involves pursuing an explicit goal to learn categories that prompts a search for just one or a few features diagnostic of category membership. In contrast, everyday experience is full of incidental encounters that allow us to observe how features cluster together in categories, such observing the co-occurrence of four legs, tail, and snout in dogs we happen to pass on the street. Here, we investigated how incidental exposure shapes category learning using a combined behavioral, eye tracking, and computational modeling approach. We found that learners picked up on the way features clustered together in categories just from incidental exposure, with pronounced downstream consequences for category learning.

Without even Trying: How Incidental Exposure Shapes Category Learning

The category variability effect describes assigning objects to a high-variability category rather than to a low-variability category. We show that similarity-based categorization theories can predict the category variability effect by adapting the similarity computation and conduct a rigorous empirical test. In an optimized categorization experiment, participants learned to assign geometrical figures to a high-variability and a low-variability category and then categorized previously unseen transfer stimuli located between the categories. We compared a formal model that ignores the category variability (Euclidean model) to one that considers the category variability (Mahalanobis model) during similarity computation. The data (N = 43) revealed that most participants did not show the category variability effect, in line with the Euclidean model. Nevertheless, the Mahalanobis model, which considers category variability, consistently described some participants. This demonstrates that—contrary to previous claims—similarity can account for the category variability effect; however, most people do not show the effect.

Modeling the Category Variability Effect in an Exemplar-Similarity Framework

Prior knowledge has long been known to influence retention of newly learned information. In particular, known semantic associations across items facilitates subsequent retrieval for these items, and this effect has been shown to increase with measures of semantic relatedness. In the field of concepts and categories, the processing of taxonomic (e.g., cup-fork, dog-bird) versus thematic (e.g., cup-drink, dog-leash) conceptual relations has been a topic of extensive investigation. Recent work indicates that these two types of conceptual relations can be dissociated at the behavioral and neural levels. However, while semantic relatedness has been explored more broadly, the precise effects of these distinct conceptual relations on associative memory remain unresolved. In this study, we demonstrate that these relations might provide a principled approach to understanding the influence of prior semantic knowledge on memory. We used a stimulus set which consists of word pairs that are either taxonomically or thematically related, while controlling a variety of lexical features relevant to memory performance (e.g., written frequency, number of syllables, concreteness, etc.). We additionally generated semantically unrelated word pairs (e.g., cookie-slate, wool-basement) using word-embeddings from a pre-trained autoencoder model. Participants (n=79) were sequentially shown word pairs (taxonomic, thematic, and unrelated) and subsequently given a cued memory test, in which they were prompted to type the word previously paired with a probe word. We found that word pairs with thematic relations led to better memory performance and shorter reaction times, followed by taxonomic relations, and finally unrelated pairs. These findings may indicate that thematic relations are better remembered due to their known functional or causal properties. This study showed that different conceptual relations differ in the extent to which they facilitate memory performance, providing new insights regarding the influence of prior semantic knowledge on memory.


Thematic relations outperform taxonomic relations in a cued recall task

Key questions in the study of categorization are how individuals form categories from experience and extend that knowledge to assess membership of novel examples. Popular accounts predict generalization to be based on either similarity to reference points or the application of rules or bounds. However, recent data show that some categorization behavior defies the predictions of leading accounts. Expanding on these findings, in the present experiment participants learned a one or two dimensional alternating category structure and were then tested on near and far transfer tasks. Findings reveal that individuals can extend a learned alternating category structure across multidimensional spaces and increasingly distant generalization regions. A new far transfer task suggests that alternation is learned in the form of a relational category at the level of the domain space within which subjects have learned an attribute-based categorization scheme operating at the item level.

Alternation as a Relational Category

Some people say they routinely think in an “inner voice,” while others assume this phrase is just a figure of speech. Across two preregistered experiments (N = 190), we investigated whether such subjective differences in internal verbalization predict attested effects of language on perceptual discrimination. Participants completed a canonical discrimination task assessing language-driven categorical perception of facial expressions, plus a self-report measure of internal verbalization and other modes of mental representation. The results replicated the behavioral signatures of categorical perception—faster and more accurate responses for cross-category than within-category discriminations. However, the strength of these effects was unrelated to participants’ propensity for internal verbalization. Our findings suggest that subjective experiences of inner voice use are not a reliable indicator of linguistic involvement in perceptual discrimination. These findings add to an emerging body of research showing that even large differences in mental experiences often lack clear behavioral consequences.

Does Experiencing an “Inner Voice” Predict Language-Driven Categorical Perception?

When people make decisions, these are often sequential, and later decisions frequently depend on the outcome of the previous decision. For instance, a doctor will first decide on a patient's treatment and then its duration. Grammar learning shows that humans can learn temporal regularities between such sequential information. However, little is known about sequential category learning tasks. Here, we investigate learning of regularities between two categorization tasks and generalization to novel objects. In an experiment, we varied whether a contingency between the outcome of two categorization tasks existed and whether they were adjacent (the tasks followed each other) or non-adjacent (an intervening task in between). Participants learned to categorize the objects of the second task in adjacent condition better than other conditions. However, both regularity conditions were beneficial to categorize novel objects (generalization). The results show the importance of considering temporal regularities between decision tasks in theories of category learning.

Learning Regularity in a Sequence of Decision-Making Tasks

In this study we conducted further tests of the ability of a computational model to fit human category learning performance. The Divergent Autoencoder (DIVA) is a generative-connectionist model of human categorization that learns through error-driven adjustments to predict class membership on the basis of the fidelity of auto-associative item reconstructions with respect to each category (Kurtz 2007). DIVA was tested on the classic SHJ category structures resulting in new qualitative and quantitative fits to both the original and revised SHJ orderings that show the impact of proposed design modifications and parameters. Theoretical implications for psychological accounts of category learning are discussed.

Modeling Human Performance on SHJ Category Structures with a Divergent Autoencoder

It is well known that continuous encoding and recall of similar semantic content leads to memory deficits caused by the build-up of proactive interference. This build-up can be released if the content is switched to a different semantic category. While release from proactive interference has been studied using non-semantic content, it has yet to be studied with relational categories (i.e., categories that cohere based on fulfilling a particular semantic relationship such as obstacles). Recent findings suggest that relational information may be stored as individual entities similar to traditional entity categories. To assess the extent to which entity and relational content are similarly stored and retrieved, participants repeatedly encode and recall lists of word singletons in either an entity or relational category condition. Recall performance is evaluated for each list to compare the build-up and release of proactive interference for participants exposed to a semantic content switch or corresponding no-switch controls.

Release from Proactive Interference with Relational Categories Versus Traditional Entity Categories

Humans can readily generalize their learning to new visual concepts, and infer their associated meanings. How do people align the different conceptual systems learned from different modalities? In the present paper, we examine emojis— pictographs uniquely situated between visual and linguistic modalities—to explore the role of alignment and multimodality in visual and linguistic semantics. Simulation experiments show that relational structures of emojis captured in visual and linguistic conceptual systems can be aligned, and that the ease of alignment increases as the number of emojis increases. We also found that emojis with subjective impressions of high popularity are easier to align between their visual and linguistic representations, suggesting the expressive power of these emojis is likely associated with their popularity. A behavioral experiment was conducted to measure similarity patterns between 48 emojis, and to compare human similarity judgments with three models. We found that the model trained with multimodal data by aligning visual and semantic spaces best accounts for human judgments.

Human similarity judgments of emojis support alignment of conceptual systems across modalities

This work tests three coping mechanisms for time pressure in similarity-based inferences from multiple features. Specifically, we test whether time pressure makes people attend to fewer object features, respond less precisely, or simplify the psychological similarity computation. We ran two preregistered time pressure experiments (Ns = 61 and 175), in which people classified objects or rated the similarity of object pairs. Inferential statistics and computational modeling in an exemplar-similarity framework indicate that time pressure mostly reduced participants’ response precision, especially concerning their similarity judgments. Specifically, participants’ categorizations were less deterministic and their similarity judgments became more variable as time pressure increased. In contrast, we do not find evidence that participants attend to fewer features or simplify the similarity computation. These results add to the evidence that time pressure and other cognitive load do not necessarily affect cognitive processes themselves but rather lower the precision of response selection.

Premium content

Laying the Foundation: Extracting Partial Meanings of Hard Nouns via Observational Contexts

Next from CogSci 2023

Without even Trying: How Incidental Exposure Shapes Category Learning

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES