Perceptual variability: Implications for learning and generalization

Abstract

The generalization of learned behavior has been extensively investigated, but accounting for variance in generalized responding remains a challenge. Based on recent advances, we demonstrate that the inclusion of perceptual measures in generalization research may lead to a better understanding of both intra-and interindividual differences in generalization. We explore various ways through which perceptual variability can influence generalized responding. We investigate its impact on the ability to discriminate between stimuli and how similarity between stimuli may be variable, rather than fixed, because of it. Subsequently, we argue that perceptual variations can yield different learning experiences and that interindividual differences in generalized responding may be understood from this perspective. Finally, we point to the role of memory and decision-making within this context. Throughout this paper, we argue that accounting for perception in current generalization protocols will improve the precision of obtained generalization gradients and the ability to infer latent mechanisms. This can inspire future attempts to use generalization gradients as a (clinical) predictor or to relate them to individual traits and neural correlates and, ultimately, may lead to new theoretical and clinical insights.

To perceive or not to perceive? Individual differences in perception predict rule induction and response generalization

“…Therefore, poor performance on the current task may actually reflect poor CS memory. In a recent study, we demonstrated that increased perceptual memory uncertainty about the CS was associated with broader generalization gradients, much like those we found for the bad perceivers (Zenses et al, 2020). Other studies have demonstrated that associative learning is capable of altering the ability to discriminate between stimuli (Laufer & Paz, 2012;Resnik, Sobel, & Paz, 2011;Schechtman, Laufer, & Paz, 2010) with inter-individual differences in the extent of learning-induced changes (Laufer, Israeli, & Paz, 2016).…”

“…How we respond to novel stimuli is the outcome of a process that is luckily not based solely on randomness but relies on different sources of information to generate an adaptive response (Zaman, Chalkia, et al, 2020). This ability to generalize from previous learning is highly functional as it makes optimal use of acquired knowledge in order to avoid unnecessary (re)learning and costly mistakes.…”

“…Despite abundant examples of response generalization within the laboratory and real-life (Enquist & Ghirlanda, 2005;Ghirlanda & Enquist, 2003;Mednick & Freedman, 1960), insight into the underlying mechanisms remains unclear for several reasons (Struyf, Zaman, Vervliet, & Van Diest, 2015;Zaman, Chalkia, et al, 2020). The first issue is that current evidence suggests that there are multiple mechanisms through which generalized response can occur, including a failure to discriminate or inductive extrapolation (Lovibond, Lee, & Hayes, 2020;Zaman, Chalkia, et al, 2020). In other words, response generalization may be an interaction between perceptual and cognitive processes.…”

“…The identification and differentiation regarding these and other mechanisms based solely on patterns of generalized responding, however, remains difficult as different mechanisms can yield the same behavioral phenotype (Struyf, Zaman, Vervliet, & Van Diest, 2015). The second complexity arises from the fact that response generalization in humans displays a high amount of variability (Lee, Mills, Hayes, & Livesey, 2020;Zaman, Chalkia, et al, 2020), with large individual differences in inductive rules (Lee et al, 2018;Lovibond et al, 2020;Wong & Lovibond, 2017) and perceptual ability (Struyf, Zaman, Hermans, & Vervliet, 2017;Struyf et al, 2015;Zaman, Ceulemans, Hermans, & Beckers, 2019;Zaman, Struyf, Ceulemans, Beckers, & Vervliet, 2019;Zaman, Chalkia, et al, 2020;Zaman, stimulus features. This leads to the prediction that more accurate stimulus perception should lead to a higher likelihood of deriving a similarity over a linear generalization rule.…”

Representing absence of evidence: why algorithms and representations matter in models of language and cognition

“…In this regard, it is notable that the problem of specifying input representations for models of neural/cognitive function appears to apply more generally to many different fields of research. For instance, recent work in category learning (Roark et al, 2020), perceptual learning (Zaman et al, 2020) and visual memory (Schurgin et al, 2020) supports a more general case that many well-established theories about underlying processes have made assumptions about the representation of physical stimulus properties as inputs that on examination are problematic. Firstly, it has been noted that assumptions made in one stimulus domain (e.g.…”