Mapping Frequency-Specific Tone Predictions in the Human Auditory Cortex at High Spatial Resolution

Abstract

Auditory inputs reaching our ears are often incomplete, but our brains nevertheless transform them into rich and complete perceptual phenomena such as meaningful conversations or pleasurable music. It has been hypothesized that our brains extract regularities in inputs, which enables us to predict the upcoming stimuli, leading to efficient sensory processing. However, it is unclear whether tone predictions are encoded with similar specificity as perceived signals. Here, we used high-field fMRI to investigate whether human auditory regions encode one of the most defining characteristics of auditory perception: the frequency of predicted tones. Two pairs of tone sequences were presented in ascending or descending directions, with the last tone omitted in half of the trials. Every pair of incomplete sequences contained identical sounds, but was associated with different expectations about the last tone (a high- or low-frequency target). This allowed us to disambiguate predictive signaling from sensory-driven processing. We recorded fMRI responses from eight female participants during passive listening to complete and incomplete sequences. Inspection of specificity and spatial patterns of responses revealed that target frequencies were encoded similarly during their presentations, as well as during omissions, suggesting frequency-specific encoding of predicted tones in the auditory cortex (AC). Importantly, frequency specificity of predictive signaling was observed already at the earliest levels of auditory cortical hierarchy: in the primary AC. Our findings provide evidence for content-specific predictive processing starting at the earliest cortical levels.SIGNIFICANCE STATEMENT Given the abundance of sensory information around us in any given moment, it has been proposed that our brain uses contextual information to prioritize and form predictions about incoming signals. However, there remains a surprising lack of understanding of the specificity and content of such prediction signaling; for example, whether a predicted tone is encoded with similar specificity as a perceived tone. Here, we show that early auditory regions encode the frequency of a tone that is predicted yet omitted. Our findings contribute to the understanding of how expectations shape sound processing in the human auditory cortex and provide further insights into how contextual information influences computations in neuronal circuits.

Keywords: auditory cortex; fMRI; predictions; predictive processing.

Figure 1.

Sequences with regularity in both frequency and temporal domain are behaviorally most predictable. A, Sequences in the psychophysical paradigm. Frequency expectations were modulated by presenting tones in an orderly (left) or scrambled (right) way with the target frequency always remaining as the final, fourth tone (blue circle). Temporal expectations were manipulated by inserting either regular (top) or irregular (bottom) ISIs between tones. The target could either be the lowest or the highest frequency in the presented sequence and the participants' task was to distinguish between the two cases and indicate whether the target was the highest or the lowest tone as quickly as possible. B, Ordered sequences were associated with significantly faster reaction times (∼40 ms) than scrambled sequences (F_(1,14) = 51.76, p = 3.031e⁻⁹). Temporally regular sequences produced, on average, 20 ms faster responses than sequences with irregular ISIs (F_(1,14) = 3.60, p = 0.078). There was no interaction between spectral and temporal regularities (F_(1,14) = 0.40, p = 0.54). Because ordered sequences with regular ISIs were responded to the fastest, we chose ascending and descending sequences with constant ISIs as stimuli for the neuroimaging experiment. **Significance below p < 0.01.

Figure 2.

Sequences in the neuroimaging experiment. One pair of sequences was in the low-frequency range (first and second column from the left) and the second pair was in the high-frequency range (third and fourth column from the left). For half of the trials, the sequences were complete (A), whereas for the other half, the targets were omitted (B). The three tones preceding the targets were identical for both sequence pairs; therefore, the omission trials contained identical tones for every sequence pair. The third and target tones of the scrambled condition were the same as for the descending sequence in high-frequency range (i.e., HSLT sequence), whereas the first and second tones were reversed in order (C).

Figure 3.

Regions of interest (ROIs). A, Anatomically defined ROIs: HG (blue), PT (green), PP (red), and STG (yellow). B, Functionally defined PAC based on the tonotopic gradient. C, Predictive responses for regular and scrambled incomplete sequences. Responses in voxels encoding the HSLT target frequency were stronger when the omitted tone was preceded by a regular sequence compared with when the tones were scrambled. *Significance below p < 0.05; **significance below p < 0.01.

Figure 4.

Strength of responses is preserved for targets in incomplete sequences. A, Best-sequence map to presentation of complete sequences in a representative subject. Each voxel was tagged with the sequence that elicited its highest response and the color was chosen correspondingly. B, Strength of responses to incomplete sequence in voxels that respond strongest to the HSLT complete sequences in A across ROIs in a representative subject and group (C). In both cases, the preference is preserved during omissions. D, Strength of responses to incomplete TOs and NTOs. Voxels respond more strongly to presentation of the incomplete part of their complete preferred sequence than a different incomplete sequence. **significance below p < 0.01.

Figure 5.

Spatial correspondence for sequences with the same predicted or perceived target. A, Schematic depiction of the performed analyses. We compared patterns of voxels activation (i.e., β values) for complete and incomplete sequences with the same target (i.e., blue arrows) or across different targets (e.g., complete HSLT and incomplete HSHT target; red arrows). B, Sequences with the same target demonstrated higher spatial correlation than sequences with different targets in the PAC, HG, and PP. *Significance below p < 0.05.

Figure 6.

Tonotopic maps reconstructed from the activity to sequence presentation (Complete Seq, Omissions) and localizer frequencies (Localizer) for all subjects. Sequence preference maps were reconstructed by color coding the voxels independently of the complete and omission sequences that evoked the highest responses. Color codes were assigned depending on the frequency of the target in the sequence. The best-frequency maps were obtained by color coding each voxel depending on which of the four localizer frequencies elicited the highest voxel activation. Obtained surface maps are highly similar and reflect the characteristic tonotopic gradients in the AC.