Classification errors and response times over multiple distributed sessions as a function of category structure

Abstract

Learning difficulty orderings for categorical stimuli have long provided an empirical foundation for concept learning and categorization research. The conventional approach seeks to determine learning difficulty orderings in terms of mean classification accuracy. However, it is relatively rare that the stability of such orderings is tested over a period of extended learning. Further, research rarely explores dependent variables beyond classification accuracy that may also indicate relative learning difficulty, such as classification response times (RTs). Using a family of category structures defined over three binary dimensions and four positive examples that is well-known for its robust learning difficulty ordering, we report the results of two experiments that test the stability of the ordering (in terms of both errors and RTs) over multiple category learning sessions. The experimental stimuli consisted of instantiations of each of the six category structures in the family. These take the form of categories consisting of four "flasks" that vary along the binary features of size (large or small), shape (circular or triangular), and color (black or white). Experiment 1 shows that when participants are randomly presented instances of all six types, the difficulty ordering remains stable across all three sessions. This stability is present in terms of mean accuracy (errors) as well as mean RTs. In Experiment 2, participants were repeatedly exposed to category instances of a single type. In terms of errors, the ordering is revealed in the first session and disappears in later sessions. The opposite trend is observed for classification RTs: The ordering is not present in the first session but is revealed in later sessions. This suggests that even when individuals reach a relative degree of expertise in terms of reduced errors, the original degree of difficulty continues to influence processing. We interpret these results in the context of the concept learning and perceptual expertise literatures.

Context, structure, and informativeness judgments: An extensive empirical investigation

“…Typically, degrees of learning difficulty for each of these six structure types is operationalized as the average proportion of errors in sequentially classifying the members belonging to the category. With regard to the six distinct structure types of the 3 2 [4] family, the highly replicated learning difficulty ordering in terms of average proportion of errors is I < II < [III, IV, V] < VI (Nosofsky et al, 1994;Rehder & Hoffman, 2005;Shepard et al, 1961;Vigo, 2013Vigo, , 2016Vigo, Evans, & Owens, 2014;Zeigler & Vigo, 2018; but see Kurtz et al, 2012). The question the current research addresses is how individuals perceive the informativeness of each of the objects (e.g., concept cues) comprising each of these distinct structure types.…”

Classification of three-dimensional integral stimuli: Accounting for a replication and extension of Nosofsky and Palmeri (1996) with a dual discrimination invariance model.

“…There is ample reason for implementing a serioinformative classification task as this approach is one of the most often utilized classification approaches in the literature (Ashby et al, 1999; Kurtz, Boukrina, & Gentner, 2013; Kurtz, Levering, et al, 2013; Little et al, 2013; Nosofsky et al, 1994; Shepard et al, 1961). Researchers, however, have replicated the robust 3 2 [4] LDO for separable dimension stimuli with a different empirical paradigm, referred to as a “parainformative” task (Vigo, 2013, 2015; Vigo et al, 2015; Zeigler & Vigo, 2018). In theory, this procedure lessens the impact short- and long-term memory have on subsequent categorization behavior and instead promotes conceptual and pattern recognition cognitive processes.…”

“…This is in contrast to the sequential presentation and classification of individual object stimuli with the more popular serioinformative approach. Notwithstanding, classification of separable dimension stimuli is robust to these differences in empirical protocol (Feldman, 2000; Vigo, 2013, 2015; Vigo et al, 2015; Zeigler & Vigo, 2018). Following, the current experiments were designed to assess this robustness for the integral results obtained with the serioinformative paradigm employed in the first two experiments.…”

“…2We note here that every time we have replicated the classic ordering (I < II < [III, IV, V] < VI), we have not stressed the need to find a “rule” in the instructions (Vigo, 2013; Vigo, 2015, 2013; Zeigler & Vigo, 2018). Rather, we often simply instruct each participant to learn the preferences of a particular individual, such as an “art collector.” Thus, we suggest that the generality of Kurtz, Levering, et al’s (2013) finding regarding the impact of “rule” instructions on Type II learning should be tempered until additional empirical support is gathered.…”

Premise typicality as feature inference decision-making in perceptual categories

“…And, as discussed above, learning such sophisticated categories involves a complex interplay with prior knowledge making it difficult to assess the representational basis of these categories, how people learn and use them, because so much is unknown or hard to characterize. To control for these complexities, the perceptual categorization paradigm uses novel, carefully constrained stimuli and newly constructed categories as a way to assess the basic mechanisms of category learning and decision-making(e.g., Griffiths et al, 2012;Honke et al, 2016;Johansen & Kruschke, 2005;Love, 2002;Medin & Schaffer, 1978;Medin & Schwanenflugel, 1981;Nosofsky & Zaki, 2002;Shepard et al, 1961;Yamauchi & Markman, 1998;Zeigler & Vigo, 2018). So, the paradigm facilitates evaluating how people represent new concepts and make inference decisions using those concepts by simplifying and controlling the categories and feature instances.…”