Figure 1. Two Views of Attention.

Note. (A) Contemporary categorization models tend to sidestep questions related to how decision makers sample information from the world. Instead, their emphasis is on how multidimensional stimulus representations are are “contorted” by selective attention (e.g., Kruschke, 1992; Love et al., 2004; Nosofsky, 1986). In the example on the left, three stimulus dimensions (Size, Color, and Shape) are equally attended. On the right, “Size” is given greater attentional weight than “Shape” or “Color”. (B) Active sampling requires decisions, not only about the appropriate final choice but also about what samples should be selected. In the category structure depicted at left (Blair et al., 2009), the optimal sampling strategy is to first sample Dimension 1, and then, depending on its value, sample either D2 or D3 (gray rectangles denote informative samples). This temporally ordered sequence is illustrated at right. It is never necessary to sample all three dimension if D1 is sampled first.