Fig 3 -

Logistic analysis of the bias (A) and adaptation-like (B) effects in the stimulus-based condition, as a function of pre-test tone position (−1 = the tone just prior to the test tone, etc.). Values along the ordinate are the medians of per-subject beta coefficients from the logistic model (points), and the error bars are bootstrapped 95% confidence intervals (computed using 10,000 bootstraps). These beta coefficients represent the effects of a pre-test tone presented at that position on choice, measured as log-odds ratios biasing choices toward (positive values) or away from (negative values) the frequency of the pre-test tone at the given position. In B, the two symbols (see legend) correspond to effects measured for different SNR values of the test tone. Note that the total effect of a given pre-test tone is found by summing the bias and adaptation-like effects, for a given test-tone SNR.