Table 1.

Summary of the results of our simulations. Each cell provides information about whether a model can reproduce a specific empirical result and, if it can, the assumptions that are required for its success.

Empirical Result	Latent Causes Model	REM	ECM
Effect of modality on simple summation (Kehoe et al., 1994).	YES Noise and tone varied in timbre; no other assumptions are necessary.	YES Assuming higher replacement for unimodal than multimodal stimuli.	YES Assuming more configurality for unimodal than multimodal stimuli.
Effect of temporal separation on simple summation (Rescorla & Coldwell, 1995).	YES Assuming that temporal separation is separable from visual dimensions.	NO Unless the model is developed further.	NO Unless the model is developed further.
Effect of modality on differential summation (Myers et al., 2001; Ploog, 2008).	YES Assuming that color and size are integral for pigeons.	YES Assuming that more similar stimuli produce higher replacement.	YES Assuming that more similar stimuli produce more configural processing.
Summation with a recovered inhibitor not affected by modality (Pearce & Wilson, 1991).	YES No other assumptions are necessary	NO Unless it is assumed that modality does not affect replacement, which contradicts first entry above.	NO Unless it is assumed that modality does not affect configurality, which contradicts first entry above.
Effect of the spatial position of stimuli on the blocking effect (Glautier, 2002; Martin & Levey, 1991).	YES Assuming that spatial position and color are separable dimensions. Glautier's results also require assuming that color and shape are integral dimensions.	NO	NO
Prevention of a blocking effect by random variation in position of cues (Dibbets et al., 2000).	YES Assuming that stock identity can be encoded as a single dimension, separable from spatial position.	NO	NO
Asymmetrical generalization decrement after addition and subtraction of cues to/from compound (e.g., Brandon et al., 2000).	YES Assuming sufficiently dissimilar stimuli.	YES Assuming replacement value larger than zero.	NO
External inhibition effect (e.g., Brandon et al., 2000)	YES Assuming a prior on the outcome with mean lower than zero.	YES Assuming replacement value larger than zero.	YES Assuming suffficient configurality.