The Goodness of Garner dot patterns has been shown to influence same-different response times in a specific way, which has led to the formulation of a memory search model of pattern comparison. In this model, the space of possible variations of each pattern is searched separately for each pattern in the comparison, resulting in faster response times for patterns that have fewer alternatives. Compared to an alternative explanation based on stimulus encoding plus mental rotation, however, the existing data strongly favor this explanation. To obtain a more constraining set of data to distinguish between the two possible accounts, we extended the original paradigm to a situation in which participants needed to compare three, rather than two patterns and varied the way the stimuli were presented (simultaneously or sequentially). Our findings suggest that neither the memory search nor the encoding plus mental rotation model provides a complete description of the data, and that the effects of Goodness must be understood in a combination of both mechanisms, or in terms of cascades processing.