Chapter 5 Highlighting

Highlighting is a perplexing effect in learning, in which shared features are more strongly associated with early learned outcomes but distinctive features are more strongly associated with later learned outcomes. The effect has been widely observed with different stimuli, procedures, and application domains. It continues to discomfit many theories of learning. This chapter provides results from a "canonical" design in which the base rates of early and late outcomes are equalized. This balanced design yields data that pose a challenge to models that have relied on differential base rates of past designs to mimic highlighting. The data are available at the author's Web site as a test bed for models. A Bayesian data analysis is also reported that provides explicit posterior distributions over choice probabilities. The posterior distribution is also available online.