The Effects of Categorization on Perceptual Judgment are Robust across Different Assessment Tasks

Abstract

Learned visual categorical perception (CP) effects were assessed using three different measures (similarity rating, same-different judgment, and an XAB task) and two sets of stimuli differing in discriminability and varying on one category-relevant and one category-irrelevant dimension. Participant scores were converted to a common scale to allow assessment method to serve as an independent variable. Two different analyses using the Bayes Factor approach produced patterns of results consistent with learned CP effects: compared to a control group, participants trained on the category distinction could better discriminate between-category pairs of stimuli and were more sensitive to the category-relevant dimension. In addition, performance was better in general for the more highly discriminable stimuli, but stimulus discriminability did not influence the pattern of observed CP effects. Furthermore, these results were consistent regardless of how performance was assessed. This suggests that, for these methods at least, learned CP effects are robust across substantially different performance measures. Four different kinds of learned CP effects are reported in the literature singly or in combination: greater sensitivity between categories, reduced sensitivity within categories, increased sensitivity to category-relevant dimensions, and decreased sensitivity to category-irrelevant dimensions. The results of the current study suggest that the cause of these different patterns of CP effects is not due to either stimulus discriminability or assessment task. Other possible causes of the differences in reported CP findings are discussed.

Disentangling the effects of attentional weighting and associative mediation in perceptual learning reveals no evidence for associative mediation.

“…Pothos and Reppa (2014) observed increases in within-category similarity following category learning on a difficult, but not on an easy, problem where the stimuli were arrows which differed in height and width and task difficulty was manipulated by varying the separation of the stimuli along these dimensions. de Leeuw et al (2016) reported a similar effect of task difficultly on acquired equivalence and distinctiveness effects using more complex stimuli. Conversely, Pérez-Gay et al (2017) did not find an effect of stimulus discriminability on categorical perception effects, but their stimuli were textures generated by tessellation of simple checkerboard patterns, and it is arguable that even their most discriminable stimuli were substan-tially more similar than any of the stimuli used by Hall et al (2003) or by Meeter et al (2009).…”

“…These experiments, however, employed simple stimuli and transfer tests where the effects of initial training were observed on the acquisition of a subsequent discrimination. The majority of recent studies of acquired equivalence and distinctiveness in humans have used more complex stimuli and assessed the effects using measures such as similarity ratings, same-different judgements, or an X-AB (or match-to-sample) task (e.g., de Leeuw et al, 2016;Goldstone, 1994;Goldstone & Steyvers, 2001;Livingstone et al, 1998;Notman et al, 2005). In these experiments, participants are typically trained to categorize stimuli that differ along two dimensions where the category boundary lies at the middle of one of the dimensions while the other dimension is irrelevant.…”

“…Commonly cited examples include wine (Bende & Nordin, 1997; Walk, 1966) and beer tasting (Peron & Allen, 1988), the detection of features in X-ray images (Sowden, Davies, & Roling, 2000), sexing of day-old chicks (Biederman & Shiffrar, 1987), the identification of faces (Shapiro & Penrod, 1986), and differentiation between speech sounds (Liberman, Harris, Hoffman, & Griffith, 1957). Typical findings are that category learning can increase the similarity of stimuli belonging to the same category (acquired equivalence; e.g., Livingston, Andrews, & Harnad, 1998), and/or decrease the similarity of stimuli belonging to different categories (acquired distinctiveness; e.g., de Leeuw, Andrews, Livingstone, & Chin, 2016; Notman, Sowden, & Özgen, 2005). For example, Goldstone (1994) trained participants to sort squares differing in size and brightness into two categories based on one or other of those perceptual dimensions.…”

Perceptual Kinds as Supervening Sortals

“…The effects of perceptual learning are restricted to comparing the members of a newly learned category with objects that do not belong to it, and so they do not concern a change in the general ability to compare certain properties (Furmanski and Engel, ). Furthermore, the aforementioned results have been replicated using various experimental designs (see de Leeuw et al ., ; Goldstone and Hendrickson, , for reviews). In consequence, such behavioural data provide a strong reason to believe that even the simplest forms of perceptual learning, which do not involve an obvious phenomenal change, involve a modification in represented lower‐level properties.…”

The Influences of Category Learning on Perceptual Reconstructions

“…In particular, dimensions or feature ranges relevant for the categorical discrimination are encoded more accurately than the ones that are irrelevant to the categorization task (Bates, Lerch, Sims, & Jacobs, 2019; de Beeck, Wagemans, & Vogels, 2003; Folstein, Palmeri, & Gauthier, 2013; Goldstone & Steyvers, 2001; Gureckis & Goldstone, 2008; Hockema, Blair, & Goldstone, 2005; Livingston et al., 1998; Notman, Sowden, & Özgen, 2005). The effects of learned categorical perception have been shown to be robust across different measurement methods, such as similarity ratings, same‐different judgments, and XAB tasks (de Leeuw, Andrews, Livingston, & Chin, 2016).…”