A Time to Remember: Neural Insights into Rapid Updating of Timed Behaviors

Processing temporal information is a critical brain function necessary for decision-making and cognition. Interval timing, or processing of time intervals on the scale of seconds to minutes, is essential for associative learning, sensorimotor processing, and optimal foraging [1] and many neuropsychiatric illnesses including Parkinson’s Disease, Huntington’s Disease, and schizophrenia involve impaired interval timing. Interval timing is considered distinct from timing at longer or shorter time scales, circadian rhythms, and sub-second timing respectively, and each of these timing processes appears to rely on distinct and largely independent neural circuits and mechanisms [2]. Though much recent attention has gone into studying brain circuits guiding interval timing, how the brain keeps track of time remains only partly understood.

Researchers have used a wide range of motor and sensory tasks to study timing in humans and animals. A common feature of timed behaviors is ‘scalar invariance’, or ‘timescale invariance’, in which the standard deviation of duration estimates is proportional to the interval being timed, a phenomenon analogous to Weber’s law for perceptual discrimination [2]. Despite some occasional counterexamples, scale invariance is typically found in timing tasks spanning sensory tasks (e.g. auditory tones must be judged as fitting in a long or short category) to motor tasks (e.g. an animal must reproduce a learned pattern or press a lever in anticipation of a learned interval). One widely used motor task that frequently produces scale-invariant behavior is the peak interval task, in which an animal must learn to respond in anticipation of a reward that is delivered after a fixed time interval elapses. Once this task is well-learned, animals tend to respond most frequently near the expected reward delivery time. When tested with probe trials lasting ~3 times longer than the fixed interval, responding eventually decreases, forming a peak response rate near the fixed interval. Consistent with scalar invariance, the width of response time distributions (reflective of error) typically scales with the duration being timed.

The use of these behavioral tasks has yielded exciting insights into brain circuits supporting timed behaviors. While early theories posited that a central ‘clock’ in the brain guides all time processes, little experimental evidence has supported this model. Rather, timed processes are distributed across multiple brain areas and may reflect intrinsic features of these areas: thus, a brain region dedicated to processing sound may also decode the temporal structure of these sounds. On the other hand, brain areas guiding motor behaviors may also compute the temporal structure of these movements [3]. This hypothesis is supported by many studies of a wide range of neuronal structures. For example, population activity in the striatum, a deep brain structure involved in action selection and reinforcement learning, predicts variation in timing judgment during a motor operant timing task [4]. In contrast, experiments using regularly timed visual cues produce population activity in the primary visual cortex that reflects learned intervals and predicts responding on a trial-by-trial basis [5]. A common feature of these studies is that the temporal profile of firing rates within these regions tends to scale with learned time intervals. When recordings in these areas are assessed trial by trial, they often predict time's passage and variability in timed responses.

Interval timing studies have relied heavily on tasks with fixed time intervals, but in nature, a truly ‘fixed’ interval is rare and animals must adapt to dynamic changes in time structures in the environment. Despite this, relatively little work has explored how animals respond to dynamic intervals, and almost nothing is known about how the brain updates behavior following changes in learned intervals. Previous work has shown that mice [6], rats [7], and humans [8] are all highly sensitive to changes in interval duration. In each of these studies, subjects have demonstrated rapid updating of timed behaviors that track dynamic interval durations. Building on this foundation, Xie et al. [9] in this edition of Neuroscience Bulletin developed a behavioral task in which mice experienced a fixed interval (10 s) before the delivery of a liquid reward. Mice developed an anticipatory licking behavior that peaked at the learned interval during probe trials, and this behavior was scale invariant across the many interval durations tested. Then, interspersed within this task were single-trial exposures of a trial of different duration (e.g. 8 s, 2 s shorter than the learned interval) just before a probe trial. Following this single exposure to intervals shorter than the learned interval, mice rapidly updated their readout of delivery expectation by peaking sooner during probe trials (Fig. 1). Interestingly, this modification was not bidirectional, as mice were more sensitive to trials that were shorter (rather than longer) than the trained intervals. Thus, this behavioral paradigm offers a unique platform for determining neural contributions to the rapid updating of timed behaviors.

Fig. 1.

Schematic of the fixed interval task. The mouse is positioned in front of a drinking spout and sucrose rewards are delivered every 10 s. Intermittently, a 30-s probe trial is introduced in which licking behavior peaks and declines (top). Lick rate data are shown for probe trials following a single 12-s interval, note licking peaks at 10 s (middle). On the other hand, following a single trial with an 8-s interval, mice peak earlier in the following probe trial, suggesting rapid updating of timed behaviors. Image Credit: Charlie Maddox. Image adapted from Xie et al. [9].

If brain timers are distributed and intrinsic to specialized brain regions [3], then the brain area responsible for motor control of licking may also guide the timing of licking behaviors. Accordingly, the authors next perturbed an area of the M2 motor cortex in the anterior lateral motor cortex (ALM), which is partially responsible for executing licking actions. When the ALM was optogenetically inhibited on the trial preceding a probe trial, mice tended to peak their licking behavior earlier during probes despite never experiencing any deviation in the target interval (Fig. 2). Thus, by disrupting the ALM during the trial before the performance, the authors suggest they disrupted normal memory formation during the previous trial, resulting in early responding during probes. In contrast, inhibiting other prefrontal cortex regions failed to modulate timed behaviors in the following probe trial, directly supporting intrinsic timing models. Thus, this work provides novel causative evidence that the ALM can modulate updating of learned intervals.

Fig. 2.

Schematic of the experimental design. Channelrhodopsin-2 was expressed in inhibitory neurons in the anterior lateral motor cortex (ALM) and fiber optics delivering laser light targeted this region. Following 10 s fixed interval training, the ALM was stimulated in the trial before a probe trial (upper). This manipulation resulted in earlier peaking of licking in the probe trial (red, lower) relative to the control trials without stimulation (black, lower). This result suggests that inhibition of the ALM can rapidly modify learned timed behaviors. Image Credit: Charlie Maddox. Image adapted from Xie et al. [9].

A characteristic feature of brain regions that appear to encode time is that a subset of neurons display stereotypic activity profiles (e.g. ramping) that scale with the interval being timed, and activity from these areas predicts variation in timed responding [4, 7, 10]. Therefore, the authors next set out to determine the firing activity of ALM neurons during probe trials following trials in which rewards appeared earlier or later than anticipated. They found activity profiles that scaled to the interval experienced in the previous trial, and interestingly, only the early portion of probe trials (when mice were actively timing) scaled. Activity profiles were then assessed on a trial-by-trial level. Neuronal activity was predictive of objective time passage, and this prediction was stronger in neurons that scaled versus those that did not. Therefore, the activity of time-sensitive ALM neurons early in probe trials that followed a temporal deviation scaled with the previous trial length and predicted the actual reward time, strongly supporting an intrinsic timing function in this region.

This work adds to a growing body of literature suggesting that organisms are capable of rapidly updating timed behaviors, even immediately after a single exposure to deviations in learned intervals [6–8]. This hints at neuronal mechanisms that are capable of rapid modification of timed processes. Moreover, this study demonstrates a causative relationship between ALM activity and timed licking behaviors, a finding supporting intrinsic models of timing [3]. Thus, if similar rapid updating is noted in alternative timing tasks relying on distinct sensory and motor modalities, it seems likely that the regions encoding that learned behavior may also be responsible for keeping track of duration history. Future work should explore this possibility. Moreover, it is interesting that only earlier-than-expected trials resulted in rapid modification of timed behaviors. While speculative, it is possible that sensitivity to earlier trials could be attributed to positive prediction error signals conveyed by dopamine neurons [11]. Due to the role of dopamine in interval timing [12], it would also be worth investigating how dopamine modulates this learning process. Finally, studies exploring the rapid updating of timed responses may inform models of interval timing, supporting models capable of updating following single-trial exposures to new intervals [8]. Understanding how the brain not only represents but how it learns about novel time information will be essential for learning how the brain times intervals. It may also shed light on neuronal processes gone awry in neuropsychiatric illnesses characterized by timing deficits.