Time Perception : Components of the Brian ’ s Clock

Our brains measure time continuously. We are aware of how long we have been doing a particular thing, how long it has been since we last slept, and how long it will be until lunch or dinner. We are ready, at any moment, to make complex movements requiring muscle coordination with microsecond accuracy, or to decode temporally complex auditory signals in the form of speech or music. Our timing abilities are impressive, diverse and worthy of investigation. But they are not very well understood. Many models of time perception have been put forward (for example, see [1–3]), collectively postulating a wide variety of different mechanisms. Regardless of their diversity, the models all agree that temporal information is processed in many ways: it is remembered, compared to other temporal information, combined with sensory information, and used in the production of motor outputs. The holy grail of timing research is to understand the ‘timedependent process’: a mechanism equivalent to a piezoelectric crystal in a man-made clock or the movement of a shadow on a sundial. This has proven an elusive goal, to the extent that ideas about how this mechanism might work remain near the level of conjecture. Researchers have had great difficulty in pinning timing-related activity in the brain to any specific type of function. This is largely because most time measurement tasks draw upon more than one process, making it difficult to tease the various components apart. In their recent study, Janssen and Shadlen [4] have shown how single unit recording can be used to partially bypass this issue. Janssen and Shadlen [4] recorded time-sensitive responses in the lateral inferior parietal (LIP) cortex of the macaque. They trained two monkeys to perform a visual delay task: the monkeys first fixated a light, then, in response to a ‘go’ signal, moved their eyes to a peripheral visual target as quickly as possible (Figure 1). The delay between target onset and ‘go’ signal varied according to two schedules: a bimodal schedule in which the ‘go’ cue could come early or late, but not between 0.75 and 1.75 seconds, and a unimodal schedule in which it came between 0.5 and 2 seconds. The schedules were presented in alternating blocks. The observed neural spike frequency in LIP correlated with the expectancy — ‘hazard function’ — of the ‘go’ cue for each of the two delay schedules. These results build upon the findings of an earlier paper from the same group [5] in which LIP neurons were recorded during a temporal discrimination task. Monkeys fixated during presentation of a standard time interval, as defined by a light, followed by a variable probe interval defined in the same manner. They then indicated that the probe was longer than the standard by saccading to a peripheral green target, or shorter by saccading to a peripheral red target (Figure 2). Recordings from LIP neurons showed that those with the short (green) light in their receptive field responded at a high frequency until the duration of the standard had elapsed, at which time their response gradually decreased. Neurons with the long (red) light in their receptive field gradually increased responding, such that their response rates eventually ‘crossed over’ and exceeded those of the green light neurons. Taken together, these datasets demonstrate that neurons in macaque LIP can respond to temporal information. The origin of that information and the purpose of such responses remain open for debate. If you are searching for the holy grail and you come across a shiny golden cup, it is natural to speculate that this is your object. Shadlen and colleagues have done so by suggesting that the neurons they observed measure time: “.. the monkey could base its judgement of time on the discharge of neurons with properties like the ones we observe” [5]. One of their arguments in favour of this interpretation is that the gradual change from high to low firing rates precludes input from an outside timer because the smooth shift in responding is inconsistent with information from a discreet decision. This pattern does not, however, exclude the involvement of input from a graded external timing signal [6]. In the more recent paper, Janssen and Shadlen [4] admit that they “cannot determine whether the timing-related anticipatory activity arises in area Time Perception: Components of the Brian’s Clock