Auditory Neuroscience: Temporal Anticipation Enhances Cortical Processing

Summary

A recent study shows that expectation about the timing of behaviorally-relevant sounds enhances the responses of neurons in the primary auditory cortex and improves the accuracy and speed with which animals respond to those sounds.

Scientists have traditionally viewed the auditory cortex, like other sensory cortices, as a passive detector of stimulus features. A number of studies have challenged this view, however, by showing that the responses of neurons in the primary auditory cortex (A1) can change with task demands [1] and learning 2, 3 and 4, and even register behaviorally relevant non-auditory events [5]. A recent study by Jaramillo and Zador [6] builds on this growing body of evidence by showing that the responses of rat A1 neurons are modulated by the expected timing of a target sound in ways that can account for improvements in the animals’ performance. Activity in the auditory cortex therefore represents not only the acoustic structure of a given sound, but also signals the cognitive functions that are carried out with it.

What Determines the Activity of Auditory Cortical Neurons?

Neurons in the auditory cortex are still commonly characterized by their pure-tone frequency and spatial tuning, but it is now accepted that — even in adulthood — these properties are constantly evolving. Ulanovsky et al. [7] showed that in anesthetized cats, the response to a pure tone of a particular frequency is highly dependent upon how rarely that frequency occurs within the animals’ acoustic environment. Similarly, consistencies in the temporal relations between pure tone frequencies can result in spike-timing-dependent changes in the tuning of A1 neurons in both awake and anesthetized ferrets [8]. While these effects are short lasting, passive exposure to certain frequency bands can induce a reorganization in the way different sound frequencies are represented in A1 that can last on the order of months [9].

The frequency tuning of A1 neurons also undergoes rapid changes in the actively listening animal [1]. Indeed, the tuning properties of the same neurons can change in different ways according to the nature of the sound detection or discrimination task the animal has been trained to perform. Together, these studies show that the responses of A1 neurons to particular sounds depend on the context in which they are presented and on their functional relevance to the animal.

Not only is the auditory cortex a plastic encoder of sound properties and their behavioral significance, but it can also represent non-auditory stimuli. For example, some neurons — even in A1 — receive visual inputs, which have been shown to enhance their sensitivity to stimulus location [10] or to communication calls [11]. Such influences appear to increase with learning as Brosch et al. [5] have shown that A1 neurons in monkeys that were extensively trained on an auditory categorization task can respond to task-related visual cues and when the animals grasp and release a touch bar in order to receive a reward. The auditory cortex is therefore responsive to non-auditory events that are relevant to the interpretation of sound, in addition to sound itself.

The study by Jaramillo and Zador [6] has extended our understanding of the cognitive auditory cortex by demonstrating that neurons in A1 are also sensitive to expectations about when a sound is to be presented. They trained rats to detect a frequency-modulated tone embedded in a train of pure tones, and to respond at a spout to the right or left depending on whether the modulated tone was high (31 kHz) or low (6.5 kHz) in frequency, respectively. To examine how temporal expectation affects perception, the authors added a novel dimension to the task by varying the probability that the frequency-modulated target occurred either early on or toward the end of each block of trials.

Jaramillo and Zador [6] found that their rats responded more accurately and quickly to early target sounds when these were expected — that is, when they were likely to occur early on in the block of trials — than when they were unexpected, particularly if the task was made more difficult by reducing the frequency modulation depth of the targets so that they became harder to distinguish from the pure tone distractors. They then used the GABA_A agonist muscimol to show that auditory cortical function is essential to rats’ performance on this task. While they did not measure the extent of the inactivation produced by muscimol application, this was nonetheless an important step, which established a direct link between the behavioral measurements and the subsequent cortical recordings.

In the final part of the experiment, Jaramillo and Zador [6] implanted tetrodes into A1 so that they could record spiking activity and local field potentials while rats performed the task. By comparing responses to tones and targets during the early period of the ‘expect-early’ blocks to the same period of the ‘expect-late’ blocks, they demonstrated an enhancement of the responses to sounds presented during times of high target expectation (Figure 1). Although this almost certainly reflects an attentional effect, as the animals could often safely ignore early sounds during the ‘expect-late’ condition, the enhancement was much more selective than might be predicted for general attentional modulation as it was limited to sounds near the preferred frequency of the neuron in question. This suggests that temporally-defined enhancement on this task might serve to sharpen the frequency tuning of the neurons, rather than globally amplifying auditory responses.

Figure 1.

Neural enhancement due to temporal expectation in auditory cortex.

(A) Stimulus paradigm used by Jaramillo and Zador [6]. Targets (frequency modulated tones) were presented within a train of roving pure tone frequencies. In separate testing blocks, targets had a high probability of being presented either early (top panel) or late in the train (bottom panel).

(B) Cartoon of frequency tuning curves in rat auditory cortex during the early phase of “expect-early” (red) or “expect-late” (blue) testing blocks. Spike rate responses to both the preceding tones and the target sounds were enhanced during periods of high temporal expectancy, but only if the stimulus frequency was close to the neuron’s preferred frequency. Adapted from Jaramillo and Zador [6].

What Is the Role of Temporal Expectation Enhancement?

While an improvement in cortical frequency selectivity should support better target detection, Jaramillo and Zador [6] found no correlation between the strength of neural response enhancement and the accuracy of the rats’ judgments on the two-alternative choice task. On the other hand, they did find that A1 enhancement was negatively correlated with the animals’ reaction time. As slow responses were often incorrect, they argue that this indirectly relates improved auditory perception to the neural enhancement exhibited by the cortical neurons. It is possible, however, that this finding instead indicates that the observed changes in neuronal responses reflect the animals’ level of attention — whether the animal is likely to miss a target — rather than perceptual acuity for frequency modulation. As both these factors are correlated in the present study, it is difficult to distinguish between them.

Irrespective of whether the enhanced cortical responses actually reflect an improvement in auditory perception, this study [6] illustrates that A1 neurons can carry information about when behaviorally-relevant stimuli are likely to occur. Previous recording studies have reported that anticipation of the timing of task-related sensory events can also modulate neuronal activity in area V4 of the visual cortex [12] and in motor-related cortical areas 13 and 14. This is therefore likely to be a widespread phenomenon within the brain, but the finding that temporal expectation can influence neuronal responses in A1 shows that even the earliest stages of the cortical hierarchy are engaged in the predictive processing that helps to make sense of the world.

Basis for Dynamic Coding in the Auditory Cortex

The modulation of auditory cortical responses brought about by temporal expectation differs from the attentional effects described by Fritz et al. [1] in behaving ferrets. In the latter study, the receptive fields of neurons in A1 change in a task-dependent fashion to enhance responses to the target sound, whereas Jaramillo and Zador [6] observed a frequency-specific enhancement of responses to the targets and the non-target tones that preceded them, with no change in the neurons’ preferred sound frequency. An influence of behavioral performance on the responses of cortical neurons to identical sounds has also been found by Bizley et al. [15], who showed that the amplitude of cortical local field potentials recorded while ferrets perform a two-alternative pitch discrimination task is more highly correlated on a trial-to-trial basis with the choice made by the animals than with the pitch of the stimuli presented. Thus, while common attentional mechanisms may well facilitate the dynamic processing observed across these studies, the nature of the resulting changes in cortical responses may only be understood in light of the particular task demands, reinforcement procedures and stimulus types used in each case.

Increasing receptive field sizes and temporal integration windows within the ascending auditory pathway have led to a concept of hierarchical organization in auditory processing. However, it has become evident that there are many inputs to A1 other than ascending ones [16]. These originate primarily from other regions of the auditory cortex, but also from other sensory cortices, prefrontal cortex, and the basal forebrain. The mechanisms of task-dependent plasticity in auditory cortex are largely unknown, but microstimulation studies suggest that at least some forms of tonotopic map plasticity in A1 are mediated by neuromodulatory influences, including cholinergic inputs from the nucleus basalis in the basal forebrain [17]. In addition, bottom-up thalamocortical projections serve to modulate corticocortical communication [18], and this may also play a role in adaptive cortical processing during behavior.

How Do We Study a Plastic Auditory Cortex?

If even the most basic response properties and the tonotopic organization of auditory cortex are dynamic on the scale of milliseconds to months, how are we to study the physiology of these structures? The elegant study of Jaramillo and Zador [6] emphasizes that we need to move away from the idea that auditory cortical responses are simply a fixed function of the acoustical properties of a stimulus, and instead study these responses in light of their cognitive context. Along these lines, it is important to combine behavioral and physiological measurements within the same subjects, as these authors have done, in order to appreciate what features cortical neurons actually represent. Finally, as an increasing number of studies illustrate the important of top-down feedback on auditory processing, further investigation into the functional anatomy of neuromodulatory projections from regions associated with learning and attention will be critical in understanding auditory cortical function.