Gaze and informativeness during category learning: Evidence for an inverse relation

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

“…Work in this area has explored whether different concepts can be ordered by their relative learning difficulty (e.g., Feldman, 2000;Kruschke, 1992;Nosofsky, Gluck, Palmeri, McKinley, & Glauthier, 1994;Vigo, 2013). Indeed, for certain categories of objects (i.e., categorical stimuli), research has shown that robust learning difficulty orderings can be established in terms of how accurately subjects classify the objects as members of the category (Kruschke, 1992;Nosofsky, Gluck, et al, 1994;Vigo, 2013Vigo, , 2014Vigo, Evans, & Owens, 2015;Vigo, Zeigler, & Halsey, 2013). Such relative difficulty between categories tells us something fundamental about our conceptual system.…”

“…In these tasks, the number of trials/blocks to criterion is the primary dependent variable (e.g., Nosofsky, Gluck, et al, 1994). Alternatively, research using learning accuracies typically takes mean proportion of errors as the primary dependent variable (e.g., Feldman, 2000;Vigo, 2013;Vigo et al, 2013). Additionally, the process by which concepts are learned can be studied by obtaining the rate of learning throughout an experimental session, and models can be subsequently compared by quantitatively predicting block-by-block changes in performance, providing a more comprehensive comparison than measures of aggregate performance.…”

“…Classification performance for sets of eight object stimuli that vary along three binary-valued separable dimensions has been the focus of much empirical work (Kruschke, 1992;Kurtz, Levering, Stanton, Romero, & Morris, 2013;Nosofsky, Gluck, et al, 1994;Rehder & Hoffman, 2005;Shepard et al, 1961;Vigo, 2013;Vigo et al, 2013). In classification tasks, this set is divided into two categories of four objects.…”

“…Figure 2 also shows example instances of each category type using the stimuli depicted in the cube. These flask-like stimuli are used in the current experiments and have been used in several previous studies (Vigo, 2013(Vigo, , 2014Vigo, Doan, Doan, & Pinegar, 2018;Vigo et al, 2013). Using stimuli that have been used in previous work provides the benefit of a common frame of reference in which meaningful comparisons can be made between studies.…”

“…In addition to these replications of serioinformative tasks, the ordering has also been replicated using parainformative tasks (Vigo, 2013;Vigo et al, 2013), where participants are able to study all category members (and nonmembers) in a learning phase before classifying objects one at a time. 1 Table 1 shows the reported mean classification errors per type in studies that used a parainformative task without feedback.…”