Article Figures & Data
Figures
- Figure 1.
Trial sequence. Blocks began with instructions signaling the expectation and attention condition. In this example, the participant should expect target absence and ignore the concurrent visual search task (i.e., allocate full attention to Gabor detection). On each trial, a visual search target T was either absent (top) or present (bottom) with 50% probability. On each trial, a target Gabor was either present (top) or absent (bottom). The probability of Gabor presentation changes with experimental condition. Response cues followed the offset of the stimuli. Staircase trials were identical, except that there was no condition-specific instruction at the beginning and confidence reports were not requested.
- Figure 2.
Behavioral effects of expectation and attention on objective and subjective decision-making. A, Detection sensitivity d′ as a function of expectation and attention condition. No significant differences were found. B, Decision threshold c as a function of expectation and attention condition. Independently of attention, bias toward reporting “yes” (lower values of c) increased with the prior probability of Gabor presence. Error bars indicate within-subjects within-subject SEM.
- Figure 3.
Confidence as a function of attention, accuracy, report, and expectation. Each panel represents confidence as a function of expectation and yes/no report. Panels in the left column represent confidence for correct responses. Panels in the right column represent confidence for incorrect responses. Panels in the top row represent confidence under full attention. Panels in the bottom row represent confidence under divided attention. Independently of accuracy (column) and attention (row), confidence increases with expectation-response congruence. Within each panel, confidence for “yes” responses increases with increasing prior probability of presence (i.e., increasing congruence), and for “no” responses it decreases with increasing prior probability of presence (i.e., decreasing congruence). Error bars indicate within-subjects SEM.
- Figure 4.
Report prediction error. A, Results of contrast incongruent response > congruent response over whole brain. Only clusters surviving FDR cluster correction are shown. B, PEREPORT (incongruent − congruent), by region and perceptual report. PEREPORT for “yes”/”no” reports are presented in green/red, respectively. BOLD has been averaged over levels of attention. *p < 0.05, PEREPORT significantly different from zero. **p < 0.01, PEREPORT significantly different from zero. ***p < 0.001, PEREPORT significantly different from zero. Error bars indicate SEM.
- Figure 5.
The relationship between confidence and PEREPORT. A, BOLD as a function of confidence in each PEREPORT region. BOLD is significantly higher for guess than confident responses in rIFG only. B, rIFG BOLD as a function of attention, accuracy, and confidence. BOLD is higher for guess than confidence responses independently of attention and decision accuracy. C, Brain-behavior correlation. PEREPORT in rIFG is negatively correlated with the behavioral increase in confidence by expectation (ΔConfidence). *p < 0.05. Error bars indicate within-subject SEM.
- Figure 6.
PPI analysis. A, Clusters that exhibit congruence-dependent FC with rIFG that is correlated with ΔConfidence. The analysis revealed intracalcarine sulcus (blue) and bilateral OFC (red). The black cluster is our seed, rIFG. Black lines joining clusters indicate FC between regions. Scatterplots below represent the correlation between each region's congruence-dependent FC with rIFG and ΔConfidence. B, Significant quadratic modulation of left OFC BOLD by expectation. C, Right OFC BOLD as a function of attention and expectation. There is a significant effect of expectation under full but not diverted attention. D, Right OFC BOLD as a function of attention and confidence. Attention reverses the effect of confidence on BOLD. E, Marginally lower intracalcarine sulcus BOLD on Gabor absent than on Gabor present trials. Error bars indicate within-subject SEM.
- Figure 7.
Voxel-based morphometry results. WM density in right OFC negatively predicts the effect of expectation on perceptual decision.
- Figure 8.
Bayesian signal detection model presented in Sherman et al. (2015) and possible roles for rIFG in this study. A, In the signal detection model presented in Sherman et al. (2015), the internal response to stimulus presence (red) or absence (blue) represents the posterior probability of that sensory event, given the evidence and the prior. Solid black line indicates the decision threshold, which is the posterior odds of presence to absence. The confidence thresholds (dashed gray lines) flank the decision threshold. Leftmost panel, Case where target absence is more likely than presence, and accordingly the peak of the stimulus absent distribution is higher. As the prior on presence increases (rightwards panels), the difference between these peaks changes accordingly. Similarly, while the decision threshold remains at the intersection of the two distributions, changes in the internal response drive the criteria to appear more liberal. This leads to apparent shifts in the accompanying confidence criteria, which are most liberal (have the largest area under the curve) when the congruent response is made. B, C, Alternative interpretations on the role of rIFG. B, rIFG may serve to compare the confidence threshold against the current evidence (green). C, rIFG may be involved in constructing the posterior probability distributions (bold).
Tables
- Table 1.
Significant results from four-way repeated-measures ANOVA on behavioral confidence
ANOVA Follow-up contrast Post-hoc t test F p η2 t p Mdiff (%) SEdiff (%) Response 9.32 0.007 0.341 No-yes 12.71 4.15 Response × accuracy 71.34 < 0.001 0.799 No-yes (correct) 0.83 0.416 −4.40 5.28 No-yes (incorrect) 7.85 <0.001 30.93 3.94 Response × attention 15.41 0.001 0.461 No-yes (full) 5.26 < 0.001 22.20 4.22 No-yes (diverted) 0.52 0.607 2.87 5.49 Expectation × attention 8.65 0.001 0.325 Quadratic trend 25% + 75% > 50% (full) 3.19 0.005 5.12 1.60 Quadratic trend 50% > 25% + 75% (diverted) 2.27 0.036 5.86 2.58 - Table 2.
Results of whole-brain analysis expectation-incongruent report > expectation-congruent reporta
Region BA Side Volume (cm3) Peak Z pFDR Peak MNI x y z MTG 21 R 2.29 4.78 0.007 54 −30 −2 SMG 9/10 R 4.15 4.54 < 0.001 12 58 32 IFG 47/48 R 2.70 4.45 0.004 56 12 −2 MOG 47/46 R 2.08 4.33 0.009 40 50 −6 AG 39 R 1.21 3.95 0.044 46 −64 36 IPL 40 R 1.21 3.91 0.044 58 −40 40 IFG 47 L 1.90 3.79 0.012 −38 26 −4 IPL 40/48 L 1.60 3.75 0.021 −54 −46 34 ↵aMTG, Middle temporal gyrus; SMG, superior medial gyrus; MOG, middle orbital gyrus; AG, angular gyrus; IPL, inferior parietal lobule.
Region Reported “no” Reported “yes” PEREPORT? F p η2 F p η2 MTG 8.82 0.008 3.29 5.83 0.006 0.245 Yes SMG 8.10 0.001 0.310 4.46 0.014 0.213 Yes rIFG 4.70 0.015 0.207 3.45 0.041 0.162 Yes MOG 1.95 0.157 0.098 3.42 0.044 0.160 No AG 3.52 0.040 0.164 4.07 0.025 0.185 Yes rIPL 4.71 0.044 0.207 7.17 0.015 0.285 Yes lIFG 5.62 0.008 0.238 2.87 0.070 0.137 No lIPL 5.39 0.032 0.230 6.04 0.005 0.251 Yes ↵aBoth effects should be significant for the region to be deemed a PEREPORT region. MTG, Middle temporal gyrus; SMG, superior medial gyrus; MOG, middle orbital gyrus; AG, angular gyrus; rIPL, right inferior parietal lobule; lIFG, left inferior frontal gyrus; lIPL, left inferior parietal lobule.
Bayes factorsb Region Confidence Accuracy Attention Confidence + others Null MTG 0.14 0.68 4.96 0.01−2.26 1.00 SMG 0.85 1.79 1.96 0.03−2.67 1.11 rIFG 13.62 0.88 0.71 < 0.01−2.07 3.96 AG 1.35 9.83 0.53 0.01−3.42 3.47 rIPL 2.64 0.84 2.42 < 0.01−1.13 10.07 lIPL 1.74 5.52 1.84 < 0.01−1.17 10.22
In this issue
