Skip to main content
. Author manuscript; available in PMC: 2015 Jul 1.
Published in final edited form as: Psychol Rev. 2014 Jul;121(3):526–558. doi: 10.1037/a0037018

Figure 2.

Figure 2

A schematic representation of the generative process assumed by the model. In trial t of a compound generalization experiment, each active (shaded) latent cause zk for k=1, 2,...K (Step 1) generates a number nkt of observed stimuli (Step 2). The values of each of these stimuli on several dimensions are sampled from a consequential region ck associated with the latent cause (Step 3). The active latent causes also produce an outcome with magnitude rt, which is sampled from a normal distribution around the sum of weights associated with all active latent causes (Step 4).