Statistical learning of recurring sound patterns encodes auditory objects in songbird forebrain

Abstract

Auditory neurophysiology has demonstrated how basic acoustic features are mapped in the brain, but it is still not clear how multiple sound components are integrated over time and recognized as an object. We investigated the role of statistical learning in encoding the sequential features of complex sounds by recording neuronal responses bilaterally in the auditory forebrain of awake songbirds that were passively exposed to long sound streams. These streams contained sequential regularities, and were similar to streams used in human infants to demonstrate statistical learning for speech sounds. For stimulus patterns with contiguous transitions and with nonadjacent elements, single and multiunit responses reflected neuronal discrimination of the familiar patterns from novel patterns. In addition, discrimination of nonadjacent patterns was stronger in the right hemisphere than in the left, and may reflect an effect of top-down modulation that is lateralized. Responses to recurring patterns showed stimulus-specific adaptation, a sparsening of neural activity that may contribute to encoding invariants in the sound stream and that appears to increase coding efficiency for the familiar stimuli across the population of neurons recorded. As auditory information about the world must be received serially over time, recognition of complex auditory objects may depend on this type of mnemonic process to create and differentiate representations of recently heard sounds.

Keywords: electrophysiology; memory; multielectrode; novelty; single-unit.

Fig. 1.

Experimental design, stimuli, and responses in stimulus stream experiments. (A) Ascending auditory pathway in songbirds. Auditory nuclei of avian hindbrain innervate nucleus mesencephalicus lateralis, pars dorsalis (MLd; inferior colliculus homolog). MLd innervates nucleus ovoidalis (OV; medial geniculate homolog). OV projects to forebrain field L2 (analog of A1, layer IV). Field L2 innervates L1 and L3, which in turn project to NCM and the caudal mesopallium (CM). (Modified with permission from ref. .) (B) Example of a synthesized zebra finch song syllable. (C) A segment of the long word stream made up of syllables (like the one in B) with different fundamental frequencies. The red bracket indicates the start and the end of one artificial six-syllable word. (D) The order of syllables in words, nonwords, and part-words. Syllables within a word are shown in the same color. Each syllable is labeled with a number indicating its order in pitch (low to high). Nonword sequences were made from the same syllables as words, but in a changed order. Part-words consist of the three final syllables of one word (red box) and the first three syllables of the next word (blue box). (E) Differences in single-unit responses between nonwords and part-words vs. words. Differences were significant for both nonwords (blue) and part-words (red) in NCM/CLM of birds exposed to the word stream, but not in field L (black) or in control birds (green) without preexposure. (F) The overall pattern of effects seen for single-unit data (E) was also seen for multiunit recordings; nonwords also showed a significantly larger effect than part-words. (G) Cumulative frequency distributions of differences in responses between words and part-words were symmetrically distributed around zero in NCM/CLM of control birds (green) and in field L (black) neurons. (H) NCM and CLM showed significant differences for nonwords vs. words in single-unit data, but NCM showed larger differences than CLM only in multiunit data. Error bars show ± SEM.

Fig. 2.

Comparisons of response amplitudes and correlation coefficients between naïve birds and birds exposed to the stimulus stream. (A) Mean response amplitude to all stimuli in the test phase. Response amplitude was significantly lower in birds that had heard the stimulus stream (solid bar) in the part-word experiment than in naïve birds (open bar) in a control experiment. (B) Representative PSTHs of responses to 12 stimuli (six words and six part-words) in one neuron recorded from a naïve bird. For most stimuli, PSTHs showed clear peaks in response to each syllable onset (black vertical ticks) and to final offset (red tick). (C) Representative PSTHs of responses to the same 12 stimuli as in A in a neuron recorded from a bird exposed to the stimulus stream. Response peaks at syllable onsets showed heterogeneous response patterns for different stimuli. (D) Mean of correlation coefficients between response PSTH waveforms in birds exposed to the stimulus stream and in naïve birds. Correlation coefficients were significantly lower in birds that had heard the stimulus stream. Error bars show ± SEM.

Fig. 3.

Nonadjacent stimuli and neural responses in experiment 5. (A) Example of triplet stimuli. Triplets are made from syllables like those in Fig. 1 B and C. The two triplets shown share the same first and last syllable, but the middle syllable is variable. Numbered bars above the first triplet show the timing of three response windows (see text and E). (B) Structure of the stimulus set. One version of triplets was heard during passive exposure, and both versions were played in the testing phase. The two versions share the same sounds in different combinations. Letters with same color indicate triplets that share the first and second syllable. Boxes with same color indicate triplets that share the second and third syllable. (C) Differences in single-unit responses between novel and familiar triplets were significant only in the right hemisphere. (D) Differences in multiunit responses between novel and familiar triplets in each hemisphere were only significant in the right hemisphere. (E) Differences in responses between novel and familiar triplets in the three response windows. No significant difference was found in window 1 for either hemisphere. Significant differences were found in window 2 for both hemispheres. In window 3, significant differences were seen only in the right hemisphere. Error bars show ± SEM.