Response Bias and the Measurement of Choice Alternation

Abstract

The problem of devising a measure of alternation of choices that is independent of response bias is considered. Previous solutions are equivalent to the correction for chance guessing in the Yes/No experiment of psychophysics, and require an estimate of response bias, usually from choices on the first of two daily trials. This involves the questionable assumption that response bias is homogeneous over time or across subjects. It is suggested that spontaneous alternation is analogous to the two alternative forced choice situation of psychophysics, and the fourfold point correlation is proposed as an index of alternation that is statistically independent of response bias. Implications for interpreting the effect upon spontaneous alternation of hippocampal and other lesions are considered.