The Motor System and Duration

Because the durations used in movement, for instance, in muscle phasing and coordination, fall within the subsecond range, it has been suggested that the timers used to measure these intervals may be located within the motor system. One candidate structure for such involvement is the cerebellum. Observations that the cerebellum is frequently active in tasks involving measurement of subsecond intervals (Belin et al., 2002; Coull et al., 2000; Coull and Nobre, 1998; Jancke et al., 2000b; Jueptner et al., 1995, 1996; Kawashima et al., 2000; Lutz et al., 2000; Maquet et al., 1996; Parsons, 2001; Penhune and Doyon, 2002; Penhune et al., 1998; Rao et al., 1997; Roland et al., 1981; Schubert et al., 1998; Schubotz et al., 2000; Schubotz and von Cramon, 2001) and that cerebellar lesions lead to deficits in this type of movement-related timing (Ivry et al., 1988) have led to the idea that this structure may contain subsecond specific timers (see Diedrichsen et al., this volume; Hazeltine et al., 1997; Ivry, 1996). Further, network models of the cerebellum have shown that the structure could feasibly measure subsecond intervals in a number of different ways (De Zeeuw et al., 1998; Guigon et al., 1994; Medina et al., 2000; Perrett et al., 1993). However, the idea that the cerebellum is exclusively involved in movement-related timing, or for that matter, in the measurement of subsecond intervals, has been rejected due to evidence showing cerebellar involvement both in perceptual (i.e., nonmotor) timing (Casini and Ivry, 1999; Ivry and Keele, 1989; Nichelli et al., 1996) and in timing of intervals as long as 21 sec (Malapani et al., 1998; Nichelli et al., 1996).

Other regions of the motor system, for instance, the premotor cortex, could also be involved in time measurement. One possible mechanism (Lewis and Miall, 2002) for such involvement is the predictable activity of buildup cells, which has been shown by others (Matsuzaka et al., 1992) to increase or decrease during movement preparation. Central pattern generators (CPGs) offer another possibility. They are known to produce rhythmic activity with periods ranging from under 60 msec to several seconds (Arshavsky et al., 1997) for all manner of rhythmic motor activity, especially locomotor, respiratory, and chewing actions. Brain stem and spinal cord CPGs are modulated by top-down control (Armstrong, 1988) and have projections to cerebral regions (Arshavsky et al., 1978). They therefore have the potential to elicit fMRI-measurable activity in the cortex and cerebellum; cortical pattern generators are also a possibility.

21.2.2 Hypothesis: Two Systems for Time Measurement The Automatic Timing System

We propose that if an interval is measured again and again without change or interruption (as in self-paced finger tapping or perception of an isochronous rhythm), the temporal measurement can be performed by an automatic circuit, which does not require overt attention. This idea is in keeping with a loose interpretation of the motor program concept (Schmidt, 1982), which suggests that all of the information needed for an overlearned movement can be stored in such a way that, once selected and initiated, the movement is essentially performed automatically. Hence, it might be necessary to attend the first cycle or two of temporal production or perception in order to select the appropriate timing mechanism and set it running, but after that, attention should be required only when there is a mismatch between interval and expectation.

Studies of overlearned movement support this model because they have shown that explicit attention is not required for performance of these "automatic" movement tasks (Passingham, 1996). If attention is not required for the movement, then it cannot be required for the related temporal measurements. We therefore propose that a timing system exists for the measurement of brief intervals that are produced continuously and via movement, as in paced finger tapping or execution of other overlearned motor programs. This system likely recruits timing circuits within the motor system that can act without attentional modulation; we will therefore refer to it as the automatic timing system. CPGs would provide an ideal mechanism for the automatic system because they are characterized by continuous rhythmic output. The proposed timing mechanisms of the cerebellum would be similarly appropriate to measurement of intervals in automatic movement, as the cerebellum seems to have an important role in automated actions (Nixon and Passingham, 2000). The Cognitively Controlled Timing System

Although the automatic system may be very handy for the nonattended measurement of time under certain very predictable conditions, it is unlikely to serve in all circumstances. For a start, automatic timing may only be possible when the interval in question is repeated over and over without stopping because unpredictable breaks in the sequence may mean that attention is required to restart or reset the timer for each new epoch. Furthermore, there may be limitations on the maximum duration length that the timers used by this system can conveniently measure (De Zeeuw et al., 1998; Guigon et al., 1994; Medina et al., 2000; Perrett et al., 1993). Finally, if the timers of the automatic system lie within the motor cortex or cerebellum, then they may be preferentially used for measurement of intervals that are part of a movement. We suggest, therefore, that intervals longer than a second or so, measured as discrete events rather than as part of a predictable sequence, and not defined by movement, are not appropriate for the automatic system and must draw instead upon a directly attended framework, which we will refer to as the cognitively controlled timing system.

Analogous to the overlapping use of the motor system for motor control and timing, we imagine that the cognitively controlled system may use neural circuits that are typically invoked for other cognitive operations, but can be recruited, when appropriate, for storing and processing information for temporal processes. Hence, we envisage that the cognitively controlled timing system draws on flexible, multipurpose cognitive modules within the prefrontal and parietal cortex, and thus shows overlap in functional imaging experiments with many other cognitive tasks. Following from the conclusions of Rammsayer (1999) and Mitriani et al. (1977) that cognitively controlled timing draws on active working memory and attention, we might therefore predict the involvement of the premotor cortex (PMC) or dorsolateral prefrontal cortex (DLPFC), both of which are known for working memory processing (Petrides, 1994; Smith and Jonides, 1999), and of some portion of the attentional system, currently thought to comprise the parietal, anterior cingulate, and frontal cortex (for a review, see Coull, 1998).

21.2.3 Supporting Evidence from the Neuroimaging Literature

If our hypothesis is correct and activity in the automatic and cognitively controlled systems can be measured using neuroimaging techniques, then an analysis of the existing neuroimaging literature should show dissociation in the brain areas activated by time measurement tasks with different characteristics. We have recently undertaken such an analysis, including all neuroimaging studies of primate time measurement known to us (Belin et al., 2002; Brunia and de Jong, 2000; Coull et al., 2000; Coull and Nobre, 1998; Gruber et al., 2000; Jancke et al., 2000a; Jueptner et al., 1995, 1996; Kawashima et al., 1999, 2000; Larasson et al., 1996; Lejeune et al., 1997; Lewis and Miall, 2002, submitted; Lutz et al., 2000; Macar et al., 2002; Maquet et al., 1996; Onoe et al., 2001; Penhune et al., 1998; Rao et al., 1997, 2001; Roland et al., 1981; Rubia et al., 1998, 2000; Sakai et al., 1999; Schubotz et al., 2000; Schubotz and von Cramon, 2001; Tracy et al., 2000). Two relevant studies were excluded because the complete results were not reported (Parsons, 2001; Schubert et al., 1998), two because they examined learning specific activities (Ramnani and Passingham, 2001; Penhune and Doyon, 2002), and one because it dealt with non-control subjects (Volz et al., 1999).

To test our hypothesis, it is necessary to examine how the pattern of activity observed in each study relates to the characteristics of the task performed. Accordingly, we have categorized the studies in three ways: (1) according to whether a duration greater than 1 sec was measured, (2) according to whether the measured duration was defined by movement, and (3) according to whether timing was continuous or occurred in discrete episodes. We listed all brain areas that were activated by these studies and recorded which studies showed activity in each. To be inclusive, we used the most lenient subtraction presented (for instance, test vs. rest) rather than a more rigorous control condition, as in Coull and Nobre (1998). In papers presenting multiple data sets, each independent set was included as a distinct study (Coull and

Nobre, 1998; Jancke et al., 2000b; Lewis and Miall, in preparation; Rao et al., 1997; Rubia et al., 1998, 2000; Sakai et al., 1999). Finally, we performed a meta-analysis, using all of this information to determine the percentage of studies with certain task characteristics that showed activity in any given area.

The results of the meta-analysis are shown in Table 21.1. Brain areas are listed across the top row, with the laterality of each area listed just below. To reduce the complexity of this table, only those areas that were active in at least 40% of the studies in one of our categories are shown; thus many areas reported to be active in a minority of studies are not included. Different combinations of studies are dealt with in rows 1 to 9, with the relevant category of task characteristics indicated to the left of each row. Thus, row 1 deals with all studies in the review, while row 2 deals only with studies in which any two out of three task characteristics are associated with the cognitively controlled timing system. Rows 3 to 5 deal with pairings of task characteristics associated with the cognitively controlled system. Rows 6 to 9 follow a similar model, but deal with studies in which task characteristics are associated with the automatic system. The remainder of the table shows the percentage of the studies in each category (row) that report activity in each brain area, with more commonly activated regions shaded more darkly. Significance of the Meta-Analysis

The first row of Table 21.1 shows no strong consensus regarding the areas involved in time measurement. Only the bilateral supplementary motor area (SMA) and cerebellum are active in more than 50% of studies, and no area is active in more than 61% of studies. The remainder of the table, however, shows clearly that a different set of areas is active during tasks associated with the cognitively controlled timing system from that active during tasks associated with the automatic timing system.

Tasks associated with the automatic timing system most commonly elicit activity in the bilateral SMA and sensorimotor cortex. The right hemispheric cerebellum, PMC, superior temporal gyrus, and to some extent, the left hemispheric basal ganglia and thalamus are also frequently activated in these tasks, though they do not appear so commonly if intervals longer than 1 sec are measured (see row 9). Activity associated with these tasks is also observed in the occipital cortex under some conditions and to a lesser extent, in the right interior parietal cortex. Interestingly, the DLPFC and the remainder of the parietal cortex rarely activate in tasks associated with automatic timing. In tasks associated with cognitively controlled timing, however, the right hemispheric DLPFC activates more commonly than any other area. The left hemispheric cerebellum and PMC, and right hemispheric intraparietal sulcus (IPS) are also very frequently active during all combinations of cognitively controlled tasks. A number of other areas activate commonly in more specific conditions: the bilateral SMA, left IPS, and right PMC so long as the interval measured is longer than one second (see row 4), the right insula, ventrolateral prefrontal cortex (VLPFC), inferior parietal, and left DLPFC so long as timing occurs in discrete epochs (see row 2), and the right basal ganglia so long as timing does not rely upon movement (see row 3).

0 0

Post a comment