Using statistical reasoning performance to reveal information parsing preferences in the mind

Many cognitive tasks require the parsing of information into smaller, discrete units in order to enable effective information processing. This parsing can, broadly speaking, be done along either situationally ad hoc dimensions or done preferentially along ecologically and evolutionarily relevant dimensions. The present research systematically evaluates these two possibilities within a statistical reasoning context. While replicating results that appear to support the partition-edit-count hypothesis (that item parsing is equipotential, based on subtle linguistic cues), this result was found to be in large part due to confounds in the nature of the tasks rather than the partitioning manipulations (Experiment 1). Additionally, a frequency presentation of the same task not only eliminated the earlier confounds but also improved performance directly and as predicted by the alternative hypothesis (Experiment 2). Attempts to reintroduce a biasing partition frame (Experiment 3) and a process study of participants' task representation (Experiment 4) also both failed to support the partition-edit-count hypothesis. These results favour an ecological rationality perspective and the associated frequency and individuation hypotheses regarding statistical reasoning (i.e., a privileged status for frequency representations to guide parsing of objects, events, and locations into easily countable units).