Figure 7 shows the mean values of response bias (C) for performance on the peripheral color judgment task. C was estimated from the normalized hit rates and the normalized false alarm rates of performance in each condition. Values of C that fall below 0 indicate that observers tended to answer “full-color” to the chimera image; values of C greater than 0 indicate a tendency to answer “chimera” to the full-color image.
We conducted a 3 × 5 within-subjects ANOVA on C for presentation time (83 ms, 117 ms, and 150 ms) × size of colored area (9, 13, 17, 21, and 25 degrees). This ANOVA revealed a significant main effect of presentation time (F (2, 86) = 9.3, p < 0.001, ηp2 = 0.03). Multiple comparison tests (with Bonferroni correction) revealed that response bias was lower for 117 ms than for 83 ms and 150 ms presentation times (p < 0.001 for 83 ms and p = 0.014 for 150 ms), and 83 ms and 150 ms did not differ significantly (p = 0.08). The main effect of size of the colored area was also significant (F (4, 172) = 135.3, p < 0.001, ηp2 = 0.49), with response bias reliably lower for larger colored areas. The interaction between presentation time and size of the colored area was also significant (F (8, 344) = 3.6, p < 0.001, ηp2 = 0.02). The analysis of simple main effects yielded significant effects of presentation time for the 17 degrees condition (F (2, 430) = 16.2, p < 0.001), 21 degrees condition (F (2, 430) = 7.7, p < 0.01), and 25 degrees condition (F (2, 430) = 3.1, p = 0.05). On the other hand, the simple main effects of presentation time were not significant for the 9 degrees and 13 degrees conditions (F (2, 430) = 0.7, p = 0.49, for the 9 degrees condition; F (2, 430) = 2.1, p = 0.13, for the 13 degrees condition). The analysis of simple main effects also yielded significant effects of the size of colored area for all presentation time conditions (F (4, 516) = 49.1, p < 0.001, for the 83 ms condition; F (4, 516) = 66.6, p < 0.001, for the 117 ms condition; and F (4, 516) = 74.2, p < 0.001, for the 150 ms conditions).
These results suggest that C decreased as the size of the colored area consistently increased; that is, response bias shifted to a more liberal criterion as the size of the colored area grew larger. This result is in line with previous findings of a shift of response bias in peripheral vision, which leads observers to respond that they see objects more frequently (i.e. false alarm rate increases). As discussed above (see
Figure 6), foveal load appeared to affect sensitivity on the peripheral task, implying that the fixed resource model of attention holds for peripheral color judgment. Thus, foveal load would also be expected to influence response bias C. To the contrary, our results for C indicate that foveal load did not consistently modulate C: the moderate foveal load produced the smallest response bias and the highest foveal load produced the largest. It also remains unclear why the effect of presentation time on C was limited to the larger colored area. We speculate that the modulation of bias by attention changed qualitatively based on the type of visual field, perifoveal versus far peripheral. We return to this effect in the Discussion section.