Sensory cortex plasticity supports auditory social learning

The study investigates how social experience facilitates the acquisition of a novel auditory discrimination task and tests the hypothesis that auditory cortex (AC) plasticity that occurs during social exposure causally supports later task learning. Social learning, in which naïve animals observe conspecifics performing a behavior, accelerates acquisition across species and sensory modalities. Prior work shows gerbils can learn an auditory discrimination more rapidly after exposure to a performing conspecific, even without visual cues. Known neural substrates for observational learning include anterior cingulate cortex and amygdala (for social fear), motor circuits for imitation, and auditory forebrain regions in songbirds that store tutor memories used during later practice. Building on evidence that AC is involved in auditory learning, the authors predict: (1) inactivating AC during social exposure will diminish the social learning benefit; and (2) social exposure will enhance AC sensitivity to task-relevant auditory cues prior to active task practice.

The paper situates social learning across taxa (bumblebees, rodents, birds, bats, primates, humans) and modalities and highlights examples where observation alone modifies behavior (e.g., dietary preferences in rodents, tool use in primates, vocal learning in songbirds). Neural mechanisms for observational learning implicate anterior cingulate cortex and amygdala in social fear learning, and higher auditory cortex regions (e.g., caudomedial nidopallium) and anterior neostriatum in song learning, where tutor exposure creates auditory memories guiding later practice. In auditory perceptual learning, AC plasticity is often linked to active training and reinforcement. However, whether social exposure without direct reinforcement can drive AC plasticity that benefits later learning remains unclear. The authors build on prior gerbil work showing faster task acquisition after social exposure and pharmacological evidence implicating dopaminergic signaling in auditory social learning, motivating tests of AC’s necessity and plasticity during exposure.

Subjects: Mongolian gerbils (Meriones unguiculatus; total n ≈ 41 across protocols). Demonstrators (n = 18) were trained on a Go/Nogo amplitude modulation (AM) discrimination task until d' > 1.5. Observers were naïve animals paired with demonstrators for exposure sessions. Task: Go stimulus: 12 Hz AM frozen noise (100% depth); Nogo stimulus: 4 Hz AM. Go trials rewarded at food tray; Nogo trials followed by time-out. Performance quantified as d' = z(hit rate) − z(false alarm rate); computed when ≥15 Nogo trials occurred. Social exposure paradigm: For five daily sessions, a trained demonstrator performed the task while a naïve observer, separated by an opaque divider (no visual access), had access to auditory, olfactory, and other social cues. After exposure, observers practiced the task themselves in daily sessions. Non-social exposure control: A separate group received experimenter-triggered Go/Nogo stimuli and matched contingencies (hits, misses, correct rejects, false alarms, time-outs) without a demonstrator, for five sessions, then practiced the task. AC inactivation: To test necessity of AC during social exposure, observers were bilaterally implanted with AC cannulae and received muscimol (1 mg/mL; 0.5 µL/hemisphere at 0.1 µL/min) or saline infusions before each exposure session (no infusions during practice). Histology confirmed placements. Electrophysiology: Chronic multichannel electrode arrays (16 or 64 sites) were implanted in left AC of social observers and non-socially exposed animals. Recordings (spikes and LFPs) were obtained during exposure (early: days 1–2; late: days 3–4) and during practice. Spike sorting used UltraMegaSort 2000 or KiloSort; quality metrics identified single vs multi-units. Units recorded: social observers total 864 units (396 single); non-social exposure total 1019 units (373 single). A subset of putatively stable single units across early-to-late exposure was analyzed with waveform and spontaneous FR criteria (±2 Hz). Single-unit analyses: Computed firing rate (FR), coefficient of variation (CV), vector strength (phase locking; Rayleigh test p < 0.001). Neural discriminability (neural d') per unit was computed via a spike pattern classifier: leave-one-out Euclidean-distance template matching on 10 ms binned 1 s spike trains, repeated 250 times to minimize selection bias, yielding hit and false alarm rates and neural d'. Population decoding: For simultaneously recorded single units, spike counts (10 ms bins over first 1 s) formed response vectors. Linear SVMs (fitcsvm/predict, linear kernel) decoded Go vs Nogo; performance converted to population neural d'. Sessions with <5 simultaneously recorded single units were excluded. Manifold analysis: Mean-field geometric analysis estimated manifold capacity (α; separability per neuron), anchor radius (R), and anchor dimensionality (D) from population responses to assess coding efficiency and its changes across days and groups. LFP analyses: Evoked responses (FFT peak at modulation frequency) and induced power (Morlet wavelets) were computed in low (0–20 Hz) and high (20–80 Hz) bands. Induced power was examined in anticipatory (−0.5 to −0.1 s before stimulus onset) and reward-related (0–2 s from reward delivery) windows during demonstrator hit trials to test AC modulation by demonstrator behaviors (nose-poke initiation and reward acquisition). Correlations were tested between induced power and behavioral d' during practice. Statistics: Normality assessed by Shapiro–Wilk; parametric ANOVAs and mixed-model ANOVAs where appropriate; non-parametric Steel–Dwass comparisons and Wilcoxon rank-sum tests otherwise. Correlations with Pearson’s r. Criterion for practice success set at behavioral d' = 1.5.

- Necessity of AC during exposure for social learning: Bilateral muscimol in AC during social exposure significantly delayed subsequent task acquisition relative to saline controls (Steel–Dwass nonparametric comparison, two-sided, p = 0.033). - Behavioral benefit of social versus non-social exposure: Electrode-implanted social observers (n = 6) reached criterion d' = 1.5 in 5.0 ± 0.5 practice days, whereas non-socially exposed animals (n = 5) required 9.6 ± 0.5 days; social observers learned significantly faster (Steel–Dwass, p = 0.037). - Single-unit response properties: Across groups, FR increased and CV decreased from early to late exposure; vector strength improved from early to late exposure, with additional late-exposure to practice gains in non-social animals. Thus, basic AC responsiveness to task cues increased with exposure in both groups. - Single-unit neural discriminability (neural d'): Social observers showed significant increases in neural d' from early to late exposure; non-social exposure animals did not (classifier-based analyses). For putatively held units, a larger fraction improved with exposure in social observers (62%) than in non-social animals (29%); group distribution difference significant (Likelihood ratio Chi-square, p = 0.0006). Changes in median neural d' during exposure correlated with days to reach behavioral criterion (all animals: Pearson p = 0.001; group differences in neural d' change: X²(1) = 6.56, p = 0.011; behavioral days difference: Wilcoxon W = 679, p = 0.009). Within-group correlations: social observers p = 0.02; non-social p = 0.14. - Population decoding: AC population neural d' improved across exposure days for social observers but not for non-social animals; during practice, both groups improved. Mixed-model ANOVA for population neural d': group p < 0.0001; days p < 0.001; interaction p = 0.014. - Manifold geometry: Capacity (α) increased with days more in social observers (group p < 0.0001; days p < 0.0001; interaction p = 0.0008). Anchor radius decreased with exposure/practice in social observers but not in non-social animals (group p < 0.001; days p < 0.0001; interaction p = 0.0008). Anchor dimensionality showed no significant effects (group p = 0.277; days p = 0.425). These changes indicate more efficient, more linearly separable population codes during social exposure. - LFP markers of social cue encoding: During late exposure, induced power in anticipatory windows (around demonstrator nose-poke) increased for social observers but not for non-social animals; during reward delivery, low-frequency power decreased in social observers (not in non-social animals). During practice, in social observers, low-frequency induced power during nose-poke correlated with behavioral d' (r = 0.31, p = 0.037); no significant high-frequency correlation (r = 0.09, p = 0.563). Only induced LFP during nose-poke predicted practice performance for social observers. - Evoked LFP FFT: Peak amplitude at stimulus modulation frequencies (12 Hz Go; 4 Hz Nogo) increased from early to late exposure in both groups (mixed-model ANOVA F(1,36) = 39.9, p < 0.0001); no overall group effect (F(1,36) = 3.3, p = 0.078) but a significant group-by-exposure interaction (F(1,36) = 4.5, p = 0.039), with larger increases for social observers.

The findings support the hypothesis that AC plasticity occurring during social exposure, before any active task engagement by observers, contributes causally to faster subsequent learning. Inactivating AC during exposure abolished the social learning advantage, demonstrating that AC activity is necessary during exposure to encode task-relevant information. Social exposure enhanced AC single-neuron and population discriminability for task cues during exposure, whereas non-social exposure did not, indicating that social context specifically primes the auditory system. Neural improvements during exposure predicted behavioral learning rates, linking early sensory plasticity to later performance. Population analyses show increased decoding capacity and reduced manifold radius in social observers, consistent with more efficient and separable representations that could be exploited by downstream decision circuits (e.g., striatum, parietal cortex). LFP induced power changes time-locked to demonstrator behaviors (anticipatory nose-pokes and rewards) indicate AC encodes social cues during exposure, and anticipatory signals correlate with later performance, suggesting engagement of reward prediction or arousal-related neuromodulatory systems during live social exposure. These results extend prior work that associated AC plasticity primarily with active training by showing that incidental, socially mediated exposure can selectively enhance AC representations. The data align with broader literature in songbirds and humans indicating that live social contexts potentiate auditory learning, potentially via neuromodulators (e.g., dopamine) recruited by demonstrator-associated sounds or behaviors.

This study demonstrates that auditory cortex activity during social exposure is required for the accelerated acquisition of an auditory discrimination task and that social exposure alone enhances AC neuron and population sensitivity to task cues before practice. The magnitude of exposure-induced AC plasticity predicts subsequent learning speed, and AC encodes demonstrator behavioral cues during exposure. Collectively, these findings indicate that social experience shapes sensory cortical representations in ways that facilitate later learning. Future directions include identifying the specific social signals (e.g., vocalizations, movement sounds) that drive AC plasticity, determining the contribution of neuromodulatory systems (e.g., dopaminergic pathways) and downstream regions (striatum, parietal cortex) to socially facilitated learning, and testing generalization across tasks and sensory modalities.

The study does not identify which specific demonstrator-derived social signals (e.g., vocalizations, movement-generated sounds, chewing) drive the observer’s AC plasticity during exposure. Although dopaminergic and other neuromodulatory contributions are implicated by prior work, their causal roles were not directly tested here. The behavioral paradigm focuses on Mongolian gerbils and an AM discrimination task without visual access; generalizability to other species, tasks, or sensory modalities requires further study. The authors also note they cannot exclude that mechanisms observed may contribute to acquisition of both social and non-social tasks.