FIG. 3.

Analysis of the response errors as a function of ${\mathrm{TMR}}_{\mathrm{be}}$ for each kind of error, averaged across subjects. The dependence of each type of response error on ${\mathrm{TMR}}_{\mathrm{be}}$ differs, indicating differences in the kind of masking responsible for the errors. Error bars show the across-subject standard error of the mean. Filled symbols show results for the target at 0°, open symbols for the target at 90°. Results for spatially separated target and masker are shown with dashed lines. Results for co-located sources are shown with solid lines. Labels above the panels indicate the posited kind of perceptual interference thought to contribute to the different types of response errors. (a) Drop-1 errors, where listeners report one target word and one word not present in either the target or masker. (b) Drop-2 errors, where listeners report two words not present in either the target or masker. (c) Mix errors, where listeners report one word from the target and one from the masker. (d) Masker errors, where listeners report both words from the masker. (e) Combination errors, which encompass all incorrect responses that are not in the above categories (e.g., reporting one masker word and guessing one word).