Expectation and attention in hierarchical auditory prediction

Abstract

Hierarchical predictive coding suggests that attention in humans emerges from increased precision in probabilistic inference, whereas expectation biases attention in favor of contextually anticipated stimuli. We test these notions within auditory perception by independently manipulating top-down expectation and attentional precision alongside bottom-up stimulus predictability. Our findings support an integrative interpretation of commonly observed electrophysiological signatures of neurodynamics, namely mismatch negativity (MMN), P300, and contingent negative variation (CNV), as manifestations along successive levels of predictive complexity. Early first-level processing indexed by the MMN was sensitive to stimulus predictability: here, attentional precision enhanced early responses, but explicit top-down expectation diminished it. This pattern was in contrast to later, second-level processing indexed by the P300: although sensitive to the degree of predictability, responses at this level were contingent on attentional engagement and in fact sharpened by top-down expectation. At the highest level, the drift of the CNV was a fine-grained marker of top-down expectation itself. Source reconstruction of high-density EEG, supported by intracranial recordings, implicated temporal and frontal regions differentially active at early and late levels. The cortical generators of the CNV suggested that it might be involved in facilitating the consolidation of context-salient stimuli into conscious perception. These results provide convergent empirical support to promising recent accounts of attention and expectation in predictive coding.

Figure 1.

Experimental design. Auditory stimuli consisted of sequences of five tones of type A or B and were presented in experimental blocks of type X or Y. In X blocks, standard sequences (71.5%) were monaural repetitions of the same tone type, interspersed with rare deviant sequences with the fifth tone differing in either type (monaural; 14.25%) or laterality (interaural; 14.25%). Y blocks were similar, except that the standard sequences had a fifth tone differing in type. This effectively created an orthogonal contrast between temporally local versus global deviance in the pattern of tones. Additionally, interaural deviant sequences generated both local and global deviance. These were contrasted with control blocks consisting only of interaural deviant sequences.

Figure 2.

MRI and CT scans of an epileptic patient with intracranial electrodes. A and B depict MRI and CT scans of the epileptic patient after surgery to implant intracranial electrodes. In both panels, the three-dimensional location indicated is that of the electrode at the top right corner of the 8 × 8 grid covering frontal, central, and temporal regions of the right hemisphere.

Figure 3.

Classic local and global effects with and without attention. In each panel, the top half plots the spatial topography of the cluster at the time point of maximal difference between the pair of conditions. The green dot therein indicates the electrode at which this difference was obtained. The bottom half plots grand- average ERP time courses at this electrode in microvolts. The thick blue horizontal line indicates the temporal extent of the cluster, and the red dashed vertical line indicates the time point at which the topograph above is plotted. This time point is also specified in the title above, along with the Monte Carlo t and p values of the cluster. The vertical black dotted lines indicate the onset of the five tones in each sequence. A depicts the local MMN effect to monaural frequency deviants in the attend sequences condition. B plots the local effect in the interference condition, showing that it survives the absence of auditory attention. C depicts the global P300 effect to rare monaural sequences in the attend sequences condition. D depicts the significant difference between the global effects in the attend sequences and interference conditions, highlighting the fact that, in contrast to the local effect, it is contingent on the availability of auditory attention. This pattern of results replicates previous findings in a related study (Bekinschtein et al., 2009). glo. dev., Global deviations; std., standard.

Figure 4.

Level 1: local deviations (loc. dev.). Panels illustrate statistically significant spatiotemporal ERP clusters obtained in contrasts between pairs of conditions within 100–200 ms. A and C depict the clusters representing the larger MMN elicited by interaural versus monaural local deviance in attend sequences and interference conditions, respectively. D highlights the enhancement of the MMN by attention: a larger combined MMN was obtained in the attend sequences versus interference condition. B highlights the attenuation of the MMN attributable to top-down expectation: a smaller MMN was obtained in the attend tones condition than the attend sequences condition.

Figure 5.

Local deviations (loc. dev.) in source space. A and B show the smoothed, z-score normalized cortical activation maps at the peak of the interaural MMN effect in the attend sequences condition, highlighting stronger responses in TPJ and PFC, respectively. C and D depict activation time courses in picoampere-meters, highlighting statistically significant temporal clusters at both these ROIs in the attend sequences condition. Corresponding time courses in the interference condition (E, F) show that only the temporal cluster at TPJ survives the lack of attention.

Figure 6.

Local deviations (loc. dev.) in intracranial recordings. A and B indicate locations of intracranial ECoG electrodes on the patient's cortex at which LFPs were compared. Electrode numbers correspond to their entries in Table 2. Electrodes are colored by the statistical differences elicited by interaural and monaural deviants: no statistical differences were found at yellow electrodes. Only the attend sequences condition was significant at orange electrodes, and both attend sequences and interference conditions were significant at red electrodes. This coloring scheme shows that, although most electrodes over TPJ (A) registered a stronger response to interaural deviants with and without attention, electrodes over PFC (B) only responded when attention was engaged. Time courses and statistical comparisons at the electrodes encircled in green are shown in the panels below. C–E depict significant temporal LFP clusters obtained in contrasts between pairs of conditions within 100–200 ms. C and D depict stronger responses to interaural deviants at electrodes over both TPJ and dorsolateral PFC in the presence of attention. E and F show that this stronger response only survives in TPJ when attention is absent.

Figure 7.

Level 2: global deviations (glo. dev.). A and E depict the cluster representing the larger P300 within 200–600 ms elicited by interaural versus monaural global deviance in attend sequences and attend tones conditions. B depicts the cluster comparing the response to interaural deviants (dev.) in experimental versus control blocks in the attend sequences condition. C highlights the lack of a significant P300 to interaural or monaural global deviance in the interference condition. D (top) shows the spatially smoothed, z-score normalized cortical map of sources active at the time point of maximal difference between the P300s in A, and D (bottom) plots the activation time course at the prominent frontal source. F (top) plots the topography and significant spatial cluster of 50% area latency differences between P300s elicited by deviants in the attend sequences versus attend tones conditions. F (bottom) plots of grand-averaged ERPs at the electrode at which this difference was maximal, indicated by the green dot in the topograph. Colored regions in F indicate 50% of the area under the P300 between 200 and 600 ms. ctrl., Control.

Figure 8.

Level 3: top-down expectation. A describes the local CNV drift within the −600 to 0 ms window across all global standard sequences in the attend sequences versus interference conditions. B describes the global increase in CNV drift rate from early to late global standard sequences within the attend sequences condition. In both panels, the top half plots the topography and significant cluster of electrodes at which CNV drift was significantly different. The bottom half depicts the ERP time course at the electrode at which this difference was maximal, indicated by the green dot in the topograph. Colored dashed lines therein show the linear fits to the corresponding ERP time courses. C describes the local CNV drift across all global standard (glo. std.) sequences in the attend tones versus interference conditions. D highlights the lack of an increase in CNV drift rate from early to late global standard sequences in the attend tones condition. E shows the spatially smoothed z-score normalized cortical map of sources activated during the presentation of global standards in the attend sequences condition. F plots activations at the prominent frontal source for early and late global standards in attend sequences and attend tones conditions, along with linear fits to each time course.