A dimensional summation account of polymorphous category learning

Abstract

Polymorphous concepts are hard to learn, and this is perhaps surprising because they, like many natural concepts, have an overall similarity structure. However, the dimensional summation hypothesis (Milton and Wills Journal of Experimental Psychology: Learning, Memory and Cognition, 30, 407-415 2004) predicts this difficulty. It also makes a number of other predictions about polymorphous concept formation, which are tested here. In Experiment 1, we confirm the theory's prediction that polymorphous concept formation should be facilitated by deterministic pretraining on the constituent features of the stimulus. This facilitation is relative to an equivalent amount of training on the polymorphous concept itself. In further experiments, we compare the predictions of the dimensional summation hypothesis with a more general strategic account (Experiment 2), a seriality of training account (Experiment 3), a stimulus decomposition account (also Experiment 3), and an error-based account (Experiment 4). The dimensional summation hypothesis provides the best account of these data. In Experiment 5, a further prediction is confirmed-the single feature pretraining effect is eliminated by a concurrent counting task. The current experiments suggest the hypothesis that natural concepts might be acquired by the deliberate serial summation of evidence. This idea has testable implications for classroom learning.

On the importance of feedback for categorization: Revisiting category learning experiments using an adaptive filter model.

“…To that end, we implemented four conditions used in Lea et al (2006), where nonhuman animal subjects were used, and replicated their qualitative pattern of results in three out of four conditions. Results from Simulation II suggest that the relative difficulty of polymorphous structures can indeed be explained by the use of an associative process, rather than by an explicit logical rule (as hypothesized in Wills et al, 2020). In our last Simulation III, we confirmed that the ALF model can predict that NLS categories may be easier to learn than LS categories, replicating results from a previous configural-cue model (Gluck, 1991).…”

Applying Insights on Categorisation, Communication, and Dynamic Decision-Making: A Case Study of a ‘Simple’ Maritime Military Decision

“…For instance in categorisation the evidence suggests participants are most likely to use simple rules based on a subset of the available information (e.g. Edmunds et al, 2015Edmunds et al, , 2018Edmunds et al, , 2019Rehder & Hoffman, 2005a;Wills et al, 2015Wills et al, , 2020. Similarly, in dynamic decision-making participants will systematically vary one variable at a time to work out the underlying causal structure of a dynamic task (e.g.…”

“…The dominant explanation, seen in many different theoretical accounts of categorisation (Pothos & Wills, 2011), is that operators compare information from every stimulus dimension to representations of every category (Nosofsky, 1986;Wills et al, 2020). The entity would then be assigned to the category to which it is most similar, where 'most similar' is typically determined geometrically, in terms of minimising the psychological distance between the stimuli and the representation of the category (e.g.…”

A cognitive category-learning model of rule abstraction, attention learning, and contextual modulation.

“…Conceptually, we view stimulus memorization as demanding cognitive resources to bind (all) separately perceived stimulus features into one representation (Treisman, 1998;Unsworth, 2019). In category learning, similar views are referred to as 'Combination Theory', which conceives configural memories as a result of more complex brain processes, relative to learning simple rules which we view as dominant route (see Lamberts, 1995;Wills, Ellett, Milton, Croft, & Beesley, 2020;Wills et al, 2015).…”

“…Conceptually, we view stimulus memorization as demanding cognitive resources to bind (all) separately perceived stimulus features into one representation (Treisman, 1998; Unsworth, 2019). In category learning, similar views are referred to as “Combination Theory,” which conceives configural memories as a result of more complex brain processes, relative to learning simple rules which we view as dominant route (see Lamberts, 1995; Wills et al, 2015, 2020). Consequently, similar but not identical to the error-based formation of clusters in SUSTAIN (Love et al, 2004) or exception learning in RULEX (Nosofsky, Palmeri, et al, 1994), we assume that prediction errors of CAL’s rules (i.e., if modulation fails) lead to feature combination beyond contextual modulation.…”