Abstract
The Inferior Colliculus (IC) has traditionally been regarded as an important relay in the auditory pathway, primarily involved in relaying auditory information from the brainstem to the thalamus. However, this study uncovers the multifaceted role of the IC in bridging auditory processing, sensory prediction, and reward prediction. Through extracellular recordings in monkeys engaged in a sound duration-based novelty detection task, we observed a "climbing effect" in neuronal firing rates, indicative of an enhanced response over sound sequences linked to sensory prediction rather than reward anticipation. Further exploration revealed a direct correlation between IC neuronal activity and behavioral choices, suggesting its involvement in decision-making processes. Additionally, our findings demonstrate reward prediction errors within the IC, highlighting its complex integration in auditory and reward processing. This research challenges conventional views of the IC, showcasing its integral role in cognitive and sensory processing and emphasizing its importance in integrated brain functions.
eLife assessment
This study presents a valuable finding on the role of the Inferior Colliculus in sensory prediction, cognitive decision-making, and reward prediction. The evidence supporting the claims of the authors is solid. The work will be of interest to neurobiologists working on auditory processing.
Significance of findings
valuable: Findings that have theoretical or practical implications for a subfield
- landmark
- fundamental
- important
- valuable
- useful
Strength of evidence
solid: Methods, data and analyses broadly support the claims with only minor weaknesses
- exceptional
- compelling
- convincing
- solid
- incomplete
- inadequate
During the peer-review process the editor and reviewers write an eLife assessment that summarises the significance of the findings reported in the article (on a scale ranging from landmark to useful) and the strength of the evidence (on a scale ranging from exceptional to inadequate). Learn more about eLife assessments
Introduction
The brain’s architecture is a complex hierarchy, where functions become increasingly sophisticated from lower to higher levels. The cerebral cortex, for instance, is pivotal for higher-order cognitive functions such as decision-making, motivation, attention, learning, memory, problem-solving, and conceptual thinking (Kremkow and Alonso, 2018; Miterko et al., 2018). Conversely, subcortical sensory systems are traditionally viewed as mere conduits for sensory information (Sherman, 2016), despite suggestions of thalamic structures playing roles in advanced perceptual tasks (Jones, 2012; Sherman, 2005). Yet, the involvement of sensory neurons below the thalamo-cortical level in cognitive behaviors remains an underexplored domain. Within the auditory hierarchy, spanning seven stages from the cochlea to the auditory cortex (AC) (Pickles, 2015), the inferior colliculus (IC) emerges as a crucial intermediary, instrumental in processing spatial representations (Cohen and Knudsen, 1999; Ono and Ito, 2015) and novelty detection (Malmierca et al., 2009; Perez-Gonzalez et al., 2005). Utilizing the oddball paradigm, researchers have identified stimulus-specific adaptation (SSA) within the inferior colliculus (IC) (Ayala et al., 2016; Duque and Malmierca, 2015; Malmierca et al., 2009; Valdés-Baizabal et al., 2021; Valdés-Baizabal et al., 2017), where neurons decrease their response to frequently occurring sounds while maintaining robust reactions to deviant ones. This phenomenon, indicative of predictive coding, underscores the IC’s relevance with sensory prediction. Exploration along the auditory pathway suggests SSA’s origination from the IC (Carbajal and Malmierca, 2018; Khouri and Nelken, 2015; Song et al., 2023). Intriguingly, the dynamic interplay between sensory prediction and sensory encoding, particularly within lower sensory systems charged with the faithful representation of external stimuli, remains an underexplored yet captivating area of inquiry.
Traditionally viewed as a key player in auditory processing, the inferior colliculus (IC) is situated in the auditory midbrain and serves as a vital conduit, channeling inputs from the brainstem’s auditory nuclei towards the thalamocortical auditory pathway. Intriguingly, the IC is endowed with dopamine receptors and exhibits nerve terminals positive for tyrosine hydroxylase, signifying a channel for dopaminergic inputs originating from the subparafascicular thalamic nucleus (Harris et al., 2021; Nevue et al., 2015; Nevue et al., 2016). This positions the IC as a crucial intersection where auditory processing and reward mechanisms converge. Recent research underscores the significant influence of dopamine on modulating auditory responses within the IC (Gittelman et al., 2013; Hoyt et al., 2019). This modulation suggests a complex interplay wherein dopamine has the potential to either suppress or enhance neural activity in the IC, depending on various factors (Gittelman et al., 2013; Hoyt et al., 2019). This amalgamation of sensory and reward coding within the IC paves new paths for research, urging a deeper inquiry into how this midbrain entity integrates auditory and reward information to mold behavioral processes, thereby challenging the traditional auditory-centric view of the IC and opening avenues for understanding its multifaceted role in cognitive and sensory modulation.
In this research, we embarked on a novelty detection task centered around sound duration with trained monkeys, performing extracellular recordings in the IC. Our observations unveiled a ‘climbing effect’—a progressive increase in firing rate after sound onset, not attributable to reward but seemingly linked to sensory experience such as sensory prediction. Moreover, we identified signals of reward prediction error and decision-making. These findings propose that the IC’s role in auditory processing extends into the realm of complex perceptual and cognitive tasks, challenging previous assumptions about its functionality.
Results
Response dynamics of IC neurons in duration novelty detection
Neurons within IC (Supplementary Fig.1) are distinguished by their ability to sustain firing, illustrating a consistent encoding of sound duration through tonic firing throughout the entire auditory length in one example neuron (Fig. 1A). An analysis of 104 neurons (53 from Monkey B and 51 from Monkey J) demonstrated that these neurons maintained sustained firing responses to sounds of varying durations—100, 500, and 1000 ms (Fig. 1B). This sustained firing behavior suggests the IC’s role in processing temporal aspects of auditory perception, highlighting its potential involvement in functions related to duration perception.
Despite traditionally being viewed as a lower-level nucleus within the auditory pathway in comparison to the auditory cortex, and rarely explored from a perceptual standpoint, this study breaks new ground by examining the involvement of IC neurons in duration perception. Monkeys were trained to perform a novelty detection task focusing on duration. Figures 1C and 1D outline the behavioral task, presenting repetitive sounds in blocks with oddball stimuli. Each block consisted of 3 to 6 repeated white-noise bursts of 300-ms duration at 600-ms interstimulus intervals, concluding with either a standard sound (control condition) or a deviant sound (Fig. 1D). Monkeys were required to press a button within 600 ms after the offset of the deviant stimulus to receive a water reward. Correct rejections occurred when the monkey refrained from pressing the button within 600 ms following a standard sound in control trials, where the deviant duration equaled the standard duration. The task difficulty was modulated by setting deviant durations at five levels: 300 (control), 357, 425, 506, and 600 ms (Fig. 1D), with the pressing ratio approaching zero in the control condition and increasing with the deviant duration, reaching 1 at the 600 ms deviant duration in one example session (Fig. 1E).
The initial analysis focused on the deviant response during correct trials, revealing that an example IC neuron exhibited continuous firing throughout the sound presentation (Fig. 2A). Following the peak response in the initial 0-60 ms window, the neuron’s firing rate began to increase approximately 100 ms after the sound’s onset, continuing until the sound’s conclusion. This phenomenon, illustrated by aligning the responses to the sound’s offset (Fig. 2B), indicated a proportional increase in firing rate with longer sound durations, suggesting a duration-dependent "climbing effect." To quantitatively assess this dynamic, the neuronal firing rate was examined across different sound durations within two distinct temporal windows: the "onset window" (0-60 ms post-sound onset) and the "late window" (−100-0 ms relative to sound offset). Notably, the firing rate of the example neuron in the late window escalated with increasing duration (p = 6.51e-23, ANOVA with post hoc test; red in Fig. 2C), while remaining constant in the onset window (p = 0.69, ANOVA with post-hoc test; black in Fig. 2C), contrasting sharply with the non-behavior condition (Supplementary Fig. 1) which showed no climbing effect.
To assess if the "climbing effect" was contingent on behavioral context, we compared responses across behavior and non-behavior conditions. Sound durations were analyzed at three levels (100, 500, and 1000 ms for the non-behavior condition as shown in Fig. 1A, B) and five levels (300, 357, 425, 506, and 600 ms for the behavior condition, illustrated in Fig. 1C, D). In the same neuronal population (n=99), the firing rate increased after the peak onset response for the behavior situation (red line of Supplementary Fig. 2A), while the firing rate stayed constant for the non-behavior condition from 100 ms to the end of the sound (blue line of Supplementary Fig. 2A). The "Response Dynamic Index" (RDI) was introduced to quantify the climbing effect, calculated as the normalized difference between responses in the late window (−100-0 ms relative to the sound’s offset) and the after-peak window (100-200 ms relative to the sound’s onset). This comparative analysis indicated a significantly higher RDI in the behavior condition (p = 4.40e-19, paired t-test, Supplementary Fig. 2B), indicating the climbing effect’s reliance on behavioral context. Additionally, RDIs across various deviant sounds showed strong correlations (p < 0.001, Pearson Correlation Analysis, Supplementary Fig. 3) and similar values across differing deviants (Supplementary Fig. 3). Detailed insights into the tonotopic distribution of the climbing effect within the IC are provided in Supplementary Fig. 4.
Exploration into whether the climbing effect manifested in responses to preceding standard stimuli showed that the same example neuron exhibited the climbing effect post-peak response to standard stimuli, which intensified as successive standard sounds progressed from the second to the seventh stimulation (Fig. 2D). The firing rate markedly decreased from the first to the second sound in both the late window (p = 3.33e-11, paired t-test, red in Fig. 2E) and onset window (p = 2.02e-40, paired t-test, black in Fig. 2E). The rate then increased in the late window (p = 3.05e-5, ANOVA with post hoc test: second versus fourth, p = 0.0398; fourth versus seventh, p = 1.00; red in Fig. 2E) but remained consistent in the onset window (p = 0.963, ANOVA with post hoc test, black in Fig. 2E) from the second to the seventh sound. Interestingly, no climbing effect was noted in the late window for the first stimulus, and the effect amplified with each subsequent standard presentation.
To monitor the dynamic of the climbing effect across repetitive sound presentations, we analyzed responses to the 300 ms sound across the entire oddball block in population. We normalized the firing rate relative to the response to the first stimulus in the block, considering both the onset and late windows (Fig. 3A), to account for variability across the neuronal population. In the onset window, responses remained relatively stable from the second to the seventh sound (black line, Fig. 3A). Conversely, in the late window, responses ascended from the second to the fourth stimulus (ANOVA with post-hoc test: second versus fourth, p = 8.29e-4, red line, Fig. 3A) and maintained a stable level from the fourth to the seventh stimulus (p = 0.653, ANOVA with post hoc test, red line, Fig. 3A). This pattern was mirrored in the peri-stimulus time histograms (PSTHs), which consolidated responses from the second to the seventh stimulus in one scale (left ordinate in Fig.3B) and the first stimulus in another scale (right ordinate in Fig.3B) for comparative purposes. Onset responses to the second through seventh stimuli completely overlapped, but responses in the late window diverged. The response profile remained flat in the late window for the first stimulus (black in Fig.3B), indicating a minimal climbing effect. Remarkably, starting from the second stimulus, the climbing effect began to manifest and progressively intensified, with a notable increase observed from the second to the fourth stimulus. This effect appeared to stabilize and overlap in responses from the fourth to the seventh sound (Fig. 3B), illustrating the nuanced modulation of neuronal firing rates across successive sound presentations within the oddball sequence.
Reward effect on neuronal responses of IC neurons
The accumulation of the climbing effect alongside repetitive sound presentations suggests a potential linkage to reward prediction, mirroring an increase in the probability of receiving a reward as the sound sequence extends. A specialized experimental setup was utilized to determine if the anticipation of a reward drives the climbing effect. In this setup, a sequence of 1000-ms white noise bursts was played across 150 trials (Fig. 4A), segmented into an initial 50-trial phase without reward (early no-reward block), a subsequent 50-trial phase with a water reward dispensed 500 ms after the sound offset (reward block), and a concluding 50-trial phase without reward (late no-reward block). An exemplary IC neuron exhibited sustained firing throughout the 1000-ms sound across all trials (Fig. 4B), with a spike cluster coinciding with the reward delivery, indicating a reward-responsive behavior within the IC neuron (Fig. 4B). The PSTHs for these conditions revealed nearly identical responses during the sound period (p = 0.098, ANOVA with post hoc test, Fig. 4C), suggesting the absence of a climbing effect following the peak response.
Out of fifty-nine IC neurons tested using this protocol, twenty-four demonstrated significant reward responses (p < 0.05, paired t-test, Fig. 4D), while thirty-five showed no significant reward response (Fig. 4E). Across both neuron types, PSTHs following the onset peak response maintained a consistently flat profile (Fig. 4D, E). However, the RDI during the reward condition, calculated from the normalized difference between responses in the late window and the after-peak window, indicated a behavior-dependent climbing effect, with most values lying above the unitary line (p = 9.75e-11, paired t-test, Fig. 4F). Furthermore, an additional set of twenty-two neurons was analyzed using 500-ms sounds within the reward protocol to ensure consistency in sound duration. This analysis showed that the PSTH firing rate increased in the behavioral condition but remained flat in the reward condition following the peak response (100-500 ms) (Supplementary Fig. 5A). The majority of these neurons’ RDIs exceeded the unitary line in pairwise comparisons (p = 3.02e-5, paired t-test, Supplementary Fig. 5B), reinforcing the notion that the climbing effect is contingent upon the behavioral context rather than the reward.
Reward prediction error in IC neuronal response
Notably, the reward response of the showcased neuron in Fig. 4B was more pronounced at the start of the reward block. To explore this, we compared the first 15 trials to the last 15 trials within the reward block (Fig. 5A), revealing that the reward response (0-200 ms relative to the reward onset) was significantly stronger during the initial trials (p = 1.50e-2, ANOVA, Fig. 5A). Twenty-eight neurons exhibiting significant reward responses were compiled; auditory responses were consistent across initial and final trials (p = 0.534, ANOVA, Fig. 5B), yet the reward responses in the initial trials were markedly higher (p = 7.00e-3, ANOVA, Fig. 5B). A comparison on an individual neuronal basis showed most points below the unity line (p = 3.20e-3, paired t-test, Fig. 5C), indicating a variance in reward responsiveness.
For another example neuron without a reward response (Fig. 5D), auditory and reward responses between the first and last trials were indistinguishable for both phases (p = 0.378 for auditory response, ANOVA, Fig. 5D; p = 0.637 for reward response period, ANOVA, Fig. 5D). A collection of thirty-five neurons without reward responses exhibited similar firing rates for both auditory and reward response periods across trials (p = 0.506 for auditory response, ANOVA, Fig. 5E; p = 0.585 for reward response period, ANOVA, Fig. 5E), with the majority of pairwise comparisons aligning closely with the unity line (p = 0.62, paired t-test, Fig. 5F).
The heightened responses during the initial trials of the reward block suggest the presence of a reward prediction error, marked by an unexpected reward at the start, which gradually becomes predictable towards the block’s end. This phenomenon was further investigated by examining the responses in the duration novelty detection task. In the control condition, where the reward was unpredictable, a distinct response was observed, contrasting with the predictable reward scenario in the deviant condition. In Fig. 6, two example IC neurons responded to the deviant sounds and subsequent rewards, each exhibiting a climbing effect post-peak response. The firing rate of these neurons returned to baseline post-sound for all durations, with a noticeable cluster of neuronal responses following the control condition sound presentation, absent in all four deviant conditions (top row of Fig. 6). When responses were aligned to reward delivery (the bottom row of Fig. 6), notable reward responses were only observed in the control condition (p < 0.001, ANOVA), underscoring the auditory IC neurons’ capability to encode reward prediction errors.
Decision related signal of IC neurons in duration novelty detection
We have shown the neuronal dynamic of IC neurons in the duration novelty detection task for correct trails. Next, we will explore whether the IC, a lower-level nucleus of the auditory pathway, plays a role in decision making of a duration-related task. During a session where a monkey excelled in performing the task (Fig. 7A), trials were categorized based on the monkey’s decisions—whether to press a button or not. Particularly in instances involving two specific deviant durations (357 ms and 425 ms), where the likelihood of pressing the button hovered around 50%, a sufficient number of trials were available to assess the correlation between neuronal activities and the monkey’s decisions. This focus aims to delve deeper into how neuronal responses might inform or influence behavioral outcomes in these scenarios.
Analysis did not show the deviant response being influenced by the sequence of preceding standard sounds (Supplementary Fig. 6), leading to an aggregation of conditions for evaluation. In an illustrative neuron responding to the 357-ms deviant sound, the activation was markedly stronger when the monkey identified the sound as deviant and opted to press the button (indicated by the red color in Fig. 7B), in stark contrast to when the monkey did not perceive the sound as deviant, thus choosing not to press the button (illustrated in black color of Fig. 7B). Aligning this observation with the moment of button press further accentuated the response disparity (p = 1.32e-8, ANOVA, right panel in Fig. 7B).
To quantitatively assess the decision related signal, an analysis incorporating Receiver Operating Characteristic (ROC) was conducted (Fig. 7C), revealing a pronounced difference in firing rates aligned with the act of pressing the button (p = 2.58e-6, ANOVA, Fig. 7C). The area under the ROC curve (AUC), signifying detection probability, stood at 0.86 (p < 0.001, permutation test), indicating a high likelihood that an ideal observer could discern between the two choices based solely on the neuronal firing rate.
Across 69 recorded sessions showcasing reliable behavioral outcomes (Fig. 7D), population-level PSTHs highlighted a tendency for heightened responses during trials where the button was pressed (p = 3.08e-4, ANOVA, left panel of Fig. 7E), with a notable difference preceding the button press (p = 7.93e-7, ANOVA, right panel of Fig. 7E). The average detection probability across these sessions was 0.56, surpassing the chance level of 0.5 (p = 2.57e-4, t-test, Fig. 7F), with fifteen out of sixty-nine neurons showing significant detection probabilities. Specifically for the 425-ms sound trials, eight out of sixty-eight neurons displayed a significant detection probability, with an average value of 0.54, also above the chance level (p = 4.30e-2, t-test, Supplementary Fig. 7).
Discussion
The study unveils novel insights into the inferior colliculus’s (IC) role in auditory processing, highlighting its involvement beyond traditional sensory encoding. Neurons within the IC exhibit a sustained firing pattern, indicating a robust processing of sound duration, which is crucial for temporal perception (Fig. 1). A behavior-dependent "climbing effect" in neuronal firing rate suggests that IC’s response to auditory stimuli is modulated by the context of the behavior, with a significant increase in response as the sound sequence progresses, implying a link to sensory experience (Fig. 2, 3). This effect was further supported by experiments that distinguished between behavior and reward conditions, showing that the climbing effect is contingent upon the behavioral context rather than mere reward anticipation (Fig. 4, 5). Additionally, IC neurons encoded reward prediction errors, demonstrating a nuanced capability to adjust responses based on reward predictability (Fig. 6). Lastly, evidence of decision signals in IC neurons correlated with the monkey’s decisions in a duration novelty detection task, indicating a direct involvement in decision-making processes related to auditory perception (Fig. 7). Overall, our results strongly suggest that the inferior colliculus is actively engaged in sensory experience, decision making and reward prediction, shedding light on its intricate functions in these processes.
Sensory experience and response dynamics in IC
The IC neurons showed a rising firing rate after the onset peak (Fig. 2A-D). The climbing effect was strongly modulated by perception (Supplementary Fig.2), echoing results from previous research (Metzger et al., 2006). This change was not due to reward, as shown by an experiment controlling for reward effects (Fig. 4). Reward only caused minor changes in the response profile (Fig. 4D). Therefore, the climbing effect reflects perception rather than reward. Another intriguing aspect of the profile change is its modulation along successive sound presentations in the oddball paradigm. The rising profile becomes steeper from the first to the fourth stimulation and then remains relatively stable from the fourth to the seventh stimulation (Fig. 3B). This modulation by preceding sensory experiences challenges the traditional view of the IC as merely a relay station, suggesting a more complex role in auditory processing influenced by both ascending and descending neural pathways. The IC’s extensive descending network of connections, including those from the auditory cortex (Adams, 1980; Andersen et al., 1980; Bajo and Moore, 2005; Coleman and Clerici, 1987; Diamond et al., 1969; FitzPatrick and Imig, 1978; Winer et al., 2002), thalamic nuclei (Adams, 1980; Kuwabara and Zook, 2000; Winer et al., 2002), and limbic system areas (Adams, 1980; Coleman and Clerici, 1987; Larue et al., 2005; Marsh et al., 2002), underscores its integration within broader sensory and cognitive processes, potentially underpinning the observed climbing effect.
The accumulation of the climbing effect across successive sound presentations appears to be intrinsically linked to the structure of oddball stimulation. Within the oddball paradigm, both sensory and reward predictions intensify alongside the recurrence of standard sounds, suggesting that the strength of these predictions could significantly influence neuronal responses. Our experimentation with rewards has effectively dismissed the role of reward prediction (Figs. 4 and 5), highlighting the potential significance of sensory prediction in molding the climbing effect. Historically, oddball stimuli have been a focal point for investigating novelty detection within the auditory cortex (Carbajal and Malmierca, 2018; Malmierca et al., 2015; Rui et al., 2018; Zhai et al., 2019), though many studies lacked incorporation of a relevant behavioral task (Khouri and Nelken, 2015; Song et al., 2023). Sensory prediction was indirectly assessed by contrasting predictive and non-predictive stimuli in control experiments, hence not addressing sensory prediction in a direct, real-time manner (Carbajal and Malmierca, 2018; Fishman and Steinschneider, 2012; Parras et al., 2017). The progression of the climbing effect with successive sound presentations could potentially offer a direct and real-time perspective on the impact of sensory prediction. Notably, while the onset response remained consistent from the second through the seventh sound, significant dynamic changes in the late response window suggest a prolonged period is required for sensory prediction to exert its influence. Nonetheless, a comprehensive understanding of this dynamic change necessitates further detailed exploration.
Decision related signal in IC
Having established how behavioral modulation affects responses in IC, we now delve into the relationship between IC responses and behavioral choices in a trial-by-trial analysis within a duration novelty detection task. This approach provides insight into the potential role of sensory neurons in perceptual tasks by comparing neural responses and perceptual decisions concurrently in the same subject. While numerous studies have explored this dynamic for cortical neurons (Tsunada et al., 2016), the contribution of subcortical neurons to decision-making processes remains less understood. To our knowledge, this study offers the direct correlation between IC neuron activity and perceptual choice, observed in real-time during experiments with animals (Fig. 1). These correlations are crucial for understanding the mechanisms by which population encoding and decoding influence and direct behavior. Focusing on the novelty detection task—a ubiquitous aspect of sensory systems and an ongoing phenomenon in daily life—we employed an oddball paradigm with randomized presentations of standard sound numbers (Fig. 1) to simulate natural occurrences of novel stimuli. This approach allowed us to closely examine both the behavioral responses of monkeys and the corresponding neuronal correlates in the IC.
When neural responses were grouped based on behavioral choices, the PSTHs diverged (Fig. 7B), indicating a strong correlation between neural firing variations and choice variations. Notably, within the overall population, the two PSTHs, sorted by the monkeys’ choices, diverged before the end of the sound, converged after its end, and diverged once more prior to the movement decision (Fig. 7E). This pattern during the sensory encoding phase underscores a potential critical role for IC in managing duration-sensitive behaviors. Traditionally, decision signals are thought to be predominantly driven by top-down inputs, as evidenced by previous studies (Cumming and Nienborg, 2016; Nienborg et al., 2012; Nienborg and Cumming, 2009). While the decision-related signal in the IC may derive from these top-down mechanisms, our findings also entertain the possibility that variations in IC neuronal activity could directly modulate higher brain areas, influencing behavioral outcomes. In either scenario, our findings clearly demonstrate the IC’s significant involvement in cognitive processing, extending well beyond its traditional role as a mere auditory relay station.
Furthermore, neuronal responses to sound duration typically become more transient throughout the auditory pathway with neurons in the thalamus and cortex primarily showing onset responses (Anderson and Linden, 2011; deCharms and Merzenich, 1996; DeWeese et al., 2003; He, 2001; Metzger et al., 2006; Nevue et al., 2015; Rauschecker et al., 1995; Xia et al., 2000; Yasui et al., 1991; Zhai et al., 2019). However, exceptions exist, as sustained neurons have been identified in the medial geniculate body (MGB) (Allon et al., 1981; Bartlett and Wang, 2011) and the auditory cortex (AC) (Lu et al., 2001; Malone et al., 2002; Rauschecker et al., 1995; Wang et al., 2005). In stark contrast, the majority of Inferior Colliculus (IC) neurons exhibit sustained responses to prolonged sounds across various species, including frogs (Ratnam and Feng, 1998), bats (Luo et al., 2008), and rodents (Pérez-González et al., 2006; Xia et al., 2000; Yin et al., 2008). Our examination of IC neurons in macaque monkeys revealed a similar sustained firing pattern in response to long-duration sounds (Fig. 1A-B), aligning with observations across diverse species. This evolutionary conservation of sustained response characteristics in the IC suggests its capacity to encode sound duration. However, the implications of this capability for behavior have remained uncertain. The present study offers direct evidence highlighting the significant role of the IC in decision-making processes related to duration perception (Fig.7).
Reward effect and reward prediction error in IC
In addition to the role in sensory processing, another important finding in our study is the existence of a reward effect and a reward prediction error in the IC. The connection between the IC and the reward system is not very clear, but some studies have suggested that dopamine may play a role in modulating auditory processing in the inferior colliculus (Gittelman et al., 2013; Hoyt et al., 2019; Nevue et al., 2015). Dopamine is a neurotransmitter that is involved in the reward system and other functions (Hoyt et al., 2019). Dopamine may have heterogeneous effects on the responses of many neurons in the IC, such as suppressing or enhancing neural activity, depending on various factors (Hoyt et al., 2019; Nevue et al., 2015). However, the exact mechanisms and functions of dopamine modulation in the inferior colliculus are still not fully understood. In this study, we first noticed that some auditory neurons responded to the reward (Fig. 4B, D), but the reward modulation effect on auditory responses is slim in the awake monkey (Fig. 4B-E).
Like the response of the dopaminergic neurons in the reward system (Hollerman and Schultz, 1998), the reward response of IC neurons displays prediction error signals, which were uncovered in two experiments. Firstly, during the experiment in which sound was followed by reward, the reward responses were significantly stronger in the initial 15 trials than in the final 15 trials (Fig. 5A-C), as the reward is a surprise at the beginning while becoming a routine at the end. Secondly, during the novelty detection task, as the number of standard sounds was randomly chosen and the monkeys were not informed about the end of the stimulation in the control trials, the reward in such trials was not expected, but then a strong response was detected (Fig. 6). By contrast, during the deviant trials, the monkey pressed the button, expected a reward, and no reward response was detected (Fig. 6). Collectively, these findings affirm the presence of reward prediction error signals in the IC, underscoring its sophisticated engagement in reward-related auditory processing dynamics.
Methods
Subjects and Apparatus
Experiments were conducted in a sound-proof room. Acoustic stimuli were digitally generated using a computer-controlled Auditory Workstation (Tucker-Davis Technologies, TDT, Alachua, FL) and delivered via a loudspeaker (LS50, KEF, UK). The sound pressure was calibrated with a ¼″condenser microphone (Brüel & Kjær 4954, Nærum, Denmark) and a PHOTON/RT analyzer (Brüel & Kjær). All stimuli were presented contralateral to the recording site.
Data were collected from two male rhesus monkeys (M. mulatta), which were chronically implanted with a plastic head-restraint ring and a recording grid, as described in detail in previous publications (Gong et al., 2024; Yu et al., 2012; Yu et al., 2015). All procedures were approved by the Animal Care and Use Committee of Zhejiang University and were performed according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
We recorded extracellular neural activity using epoxy-coated tungsten microelectrodes (FHC, 2–4 MΩ). Electrodes were inserted into the inferior colliculus based on MRI structure (SF. 1a) through 26-gauge transdural guide tubes and were advanced by a remote-controlled microdrive (FHC). Raw neural activity was amplified and filtered (300 Hz to 3000 Hz). The spike train of the neuron was analyzed off-line. Single units were identified based on waveform shape, latency, and amplitude. Only well-isolated neurons were included in the analyses.
Auditory Stimulation
Tones (100 ms; 5-ms rise-fall time) with multiple combinations of frequencies and intensities were presented to determine the frequency response areas (FRAs). Tones were presented randomly with 5 repetitions at each frequency (0.5–48 kHz in 26 logarithmic steps) and intensity (0–70 dB SPL in 10 dB steps) and interstimulus intervals of 300 ms. We used the FRAs to determine the CF.
To assess the ability of sustained firing in the IC, we presented white noise bursts with three kinds of duration (100 ms, 500 ms, 1000 ms) at 2-s interstimulus intervals. The three sound durations were randomly presented at 60 dB SPL, each for 20 trials.
To explore the reward effect on the neuronal firing of IC neurons, 1000-ms white noise was presented 150 times at 60 dB SPL at 4-s interstimulus interval (Fig. 5A). The initial 50 trials consisted of no reward (early no-reward block), the middle 50 trials consisted of a reward of water 500 ms after the offset of each sound (reward block); and the final 50 trials again consisted of no reward (late no-reward block). (At the beginning of the experiment, we presented stimuli only 100 times, consisting of early no-reward block and reward block and collected 4 IC neurons.)
Behavioral Protocol
Animals were trained to perform a novelty detection task in two steps: 1) train the monkeys to press the button after the sound; 2) train the monkeys to press the button only after the deviant sound is presented. At both steps, the monkeys received a water reward when they made the correct response.
During the recording session, the sound was presented in blocks (Fig. 1A-B). In each block, 3 to 6 repeated white noise bursts (standard sound) were presented at 600-ms interstimulus interval, and the last stimulus was either the same sound (the control condition) or a deviant sound (Fig. 1A-B). The monkey needed to press the button immediately within 600 ms after the offset of the deviant stimulus to get the reward of a drop of water. The task reaction time setting precluded the possibility that the monkey made the decision just because it was the last stimulus in the block; in the control condition, the monkey did not need to press the button within 600 ms after the onset of the last stimulus to get the reward. The number of standard stimuli was randomly selected from 3 to 6 to avoid that the monkey made the decision based on counting the number of sounds and avoided that the monkey guessed when the deviant would be presented, so that the monkey had to press the button based on the detection of the current deviant stimulus. In each recording session, the standard duration was kept the same (300 ms), but the deviant duration was systematically changed to control the difficulty of the task according to the following formula: Durationdeviant = Duration standard * λn (n=0,1,2,3,4) (λ=1.19), where Durationdeviant and Durationstandard were deviant and standard duration, respectively. Thus, in each session, 5 conditions were presented. Each condition was usually presented 30-40 times. During the recording, all sounds were presented at 60 dB SPL with 5 ms rise-fall time in the profile.
We assessed the behavior of the monkey according to two criteria: 1) the monkey made the correct rejection. In the control condition, the ratio of pressing a button had to be less than 0.1; 2) The monkey made the correct hit. In the most differential condition, the ratio of pressing the button had to be greater than 0.85. Only the behavior meeting these two criteria was kept for further analysis.
Data analysis
To characterize the climbing effect, we defined a factor termed the response dynamic index (RDI). RDI is calculated according to the following formula:
and were the neuronal firing rates within the late window (−100-0 ms relative to the offset of the sound) and the after-peak window (100-200 ms relative to the onset of the sound), respectively.
To define a reward-responsive neuron, we compared the response within 100 ms after the reward and the response within 200 ms before the reward was offered in the reward blocks (paired t-test). A neuron was considered reward-responsive only if the difference reached significance (p < 0.05).
Behavioral performance was quantified by plotting the proportion of “pressing button” choices as a function of duration difference between the deviant and standard sounds in ratio. Psychometric data were fit with a Gaussian integral function:
In this expression, p is the proportion of pressing button choices, r is the ratio between deviant and standard duration.
Acknowledgements
We are grateful to Prof. Liping Wang for their invaluable comments on the early version of the manuscript, as well as to Xiaokai Kou and Fujin Gao for their assistance with the experiments. This work was supported by STI2030-Major Projects (2022ZD0204800 and 2022ZD0204600) to XY, the National Natural Science Foundation of China (32171044 to XY and 32100827 to YZ).
Declaration of interests
The authors declare no competing interests.
Figure Legends
References
- 1.Crossed and descending projections to the inferior colliculusNeurosci Lett 19:1–5
- 2.Responses of single cells in the medial geniculate body of awake squirrel monkeysExp Brain Res 41:222–232
- 3.The topographic organization of corticocollicular projections from physiologically identified loci in the AI, AII, and anterior auditory cortical fields of the catJ Comp Neurol 191:479–494
- 4.Physiological differences between histologically defined subdivisions in the mouse auditory thalamusHear Res 274:48–60
- 5.Stimulus-specific adaptation in the inferior colliculus: The role of excitatory, inhibitory and modulatory inputsBiol Psychol 116:10–22
- 6.Descending projections from the auditory cortex to the inferior colliculus in the gerbil, Meriones unguiculatusJ Comp Neurol 486:101–116
- 7.Correlation of neural response properties with auditory thalamus subdivisions in the awake marmosetJ Neurophysiol 105:2647–2667
- 8.The Neuronal Basis of Predictive Coding Along the Auditory Pathway: From the Subcortical Roots to Cortical Deviance DetectionTrends Hear 22
- 9.Maps versus clusters: different representations of auditory space in the midbrain and forebrainTrends in neurosciences 22:128–135
- 10.Sources of projections to subdivisions of the inferior colliculus in the ratJ Comp Neurol 262:215–226
- 11.Feedforward and feedback sources of choice probability in neural population responsesCurr Opin Neurobiol 37:126–132
- 12.Primary cortical representation of sounds by the coordination of action-potential timingNature 381:610–613
- 13.Binary spiking in auditory cortexJ Neurosci 23:7940–7949
- 14.The projection of the auditory cortex upon the diencephalon and brain stem in the catBrain Res 15:305–340
- 15.Stimulus-specific adaptation in the inferior colliculus of the mouse: anesthesia and spontaneous activity effectsBrain Struct Funct 220:3385–3398
- 16.Searching for the mismatch negativity in primary auditory cortex of the awake monkey: deviance detection or stimulus specific adaptation?J Neurosci 32:15747–15758
- 17.Projections of auditory cortex upon the thalamus and midbrain in the owl monkeyJ Comp Neurol 177:573–555
- 18.Dopamine modulates auditory responses in the inferior colliculus in a heterogeneous mannerJ Assoc Res Otolaryngol 14:719–729
- 19.Neural correlates of novelty detection in the primary auditory cortex of behaving monkeysCell reports 43
- 20.Dopamine in Auditory Nuclei and Lemniscal Projections is Poised to Influence Acoustic Integration in the Inferior ColliculusFront Neural Circuits 15
- 21.On and off pathways segregated at the auditory thalamus of the guinea pigJ Neurosci 21:8672–8679
- 22.Dopamine neurons report an error in the temporal prediction of reward during learningNat Neurosci 1:304–309
- 23.Dopamine Acts via D2-Like Receptors to Modulate Auditory Responses in the Inferior ColliculuseNeuro 6
- 24.The thalamus
- 25.Detecting the unexpectedCurr Opin Neurobiol 35:142–147
- 26.Thalamocortical Circuits and Functional ArchitectureAnnu Rev Vis Sci 4:263–285
- 27.Geniculo-collicular descending projections in the gerbilBrain Res 878:79–87
- 28.The central gray and the inferior colliculus: an auditory-limbic interfaceAssoc Res Otolaryngol Abstr
- 29.Temporal and rate representations of time-varying signals in the auditory cortex of awake primatesNat Neurosci 4:1131–1138
- 30.Duration-sensitive neurons in the inferior colliculus of horseshoe bats: adaptations for using CF-FM echolocation pulsesJ Neurophysiol 99:284–296
- 31.The cortical modulation of stimulus-specific adaptation in the auditory midbrain and thalamus: a potential neuronal correlate for predictive codingFront Syst Neurosci 9
- 32.Stimulus-specific adaptation in the inferior colliculus of the anesthetized ratJ Neurosci 29:5483–5493
- 33.Context-dependent adaptive coding of interaural phase disparity in the auditory cortex of awake macaquesJ Neurosci 22:4625–4638
- 34.Projection to the inferior colliculus from the basal nucleus of the amygdalaJ Neurosci 22:10449–10460
- 35.Effects of reward and behavioral context on neural activity in the primate inferior colliculusJ Neurosci 26:7468–7476
- 36.Shaping Diversity Into the Brain’s Form and FunctionFront Neural Circuits 12
- 37.Dopaminergic Input to the Inferior Colliculus in MiceFront Neuroanat 9
- 38.Dopaminergic projections of the subparafascicular thalamic nucleus to the auditory brainstemHear Res 341:202–209
- 39.Decision-related activity in sensory neurons: correlations among neurons and with behaviorAnnu Rev Neurosci 35:463–483
- 40.Decision-related activity in sensory neurons reflects more than a neuron’s causal effectNature 459:89–92
- 41.Functional organization of the mammalian auditory midbrainJ Physiol Sci 65:499–506
- 42.Neurons along the auditory pathway exhibit a hierarchical organization of prediction errorNat Commun 8
- 43.Novelty detector neurons in the mammalian auditory midbrainEur J Neurosci 22:2879–2885
- 44.Duration selective neurons in the inferior colliculus of the rat: topographic distribution and relation of duration sensitivity to other response propertiesJ Neurophysiol 95:823–836
- 45.Auditory pathways: anatomy and physiologyHandb Clin Neurol 129:3–25
- 46.Detection of auditory signals by frog inferior collicular neurons in the presence of spatially separated noiseJ Neurophysiol 80:2848–2859
- 47.Processing of complex sounds in the macaque nonprimary auditory cortexScience 268:111–114
- 48.Frequency-dependent stimulus-specific adaptation and regularity sensitivity in the rat auditory thalamusNeuroscience 392
- 49.Thalamic relays and cortical functioningProg Brain Res 149:107–126
- 50.Thalamus plays a central role in ongoing cortical functioningNat Neurosci 19:533–541
- 51.Stimulus-specific adaptation (SSA) in the auditory system: Functional relevance and underlying mechanismsNeurosci Biobehav Rev 149
- 52.Causal contribution of primate auditory cortex to auditory perceptual decision-makingNat Neurosci 19:135–142
- 53.In vivo whole-cell recordings of stimulus-specific adaptation in the inferior colliculusHear Res 399
- 54.Endocannabinoid Modulation of Stimulus-Specific Adaptation in Inferior Colliculus Neurons of the RatSci Rep 7
- 55.Sustained firing in auditory cortex evoked by preferred stimuliNature 435:341–346
- 56.Descending projections to the inferior colliculus from the posterior thalamus and the auditory cortex in rat, cat, and monkeyHear Res 168:181–195
- 57.Neural representation of sound duration in the inferior colliculus of the mouseActa Otolaryngol 120:638–643
- 58.Non-dopaminergic projections from the substantia nigra pars lateralis to the inferior colliculus in the ratBrain research 559:139–144
- 59.Local inhibition shapes duration tuning in the inferior colliculus of guinea pigsHear Res 237:32–48
- 60.Detection thresholds of macaque otolith afferentsJ Neurosci 32:8306–8316
- 61.Neuronal thresholds and choice-related activity of otolith afferent fibers during heading perceptionProc Natl Acad Sci U S A 112:6467–6472
- 62.Integrative stimulus-specific adaptation of the natural sounds in the auditory cortex of the awake ratBrain Struct Funct 224:1753–1766
Article and author information
Version history
- Sent for peer review:
- Preprint posted:
- Reviewed Preprint version 1:
Copyright
© 2024, Du et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
- views
- 217
- download
- 1
- citations
- 0
Views, downloads and citations are aggregated across all versions of this paper published by eLife.