vation in their tasks was related to sensory input, rather than motor output. Therefore, with regard to the activa- tion reported in this paper, we would suggest that the an- siform lobule activity is related to active movement of the ipsilateral arm and hand; the paramedian/biventer ac- tivity is likely to be concerned with integration of sen- sory (visual and proprioceptive) input into the visually guided movement, while the activity in the vermis is par- ticularly concerned with visual feedback control in our task.
However, it is important to note that the vermal site activated by visually guided hand tracking without ocu- lar tracking (Table 2, task A vs. D) was the same site most strongly activated by the eye-tracking movements. Comparison of eye tracking versus fixation in experi- ment 1 produced almost no statistically significant acti- vation (P>0.01); in experiment 2, we increased the pow- er of the contrast between visual fixation and visual tracking, and were then able to demonstrate activation of the vermis (Fig. 7). There was no arm movement re- quired in experiment 2 and so the vermal activation was clearly the result of the contrast in eye-movement condi- tions. This site is likely to lie within lobules VI and VII, the “oculomotor vermis”. It is interconnected to the ocu- lomotor nuclei via the fastigial nucleus and is clearly an important cerebellar region for oculomotor control (Carter and Zee 1997; Krauzlis and Miles 1998; Ohtsuka and Noda 1991; Takagi et al. 1998).
The third main discussion point is that, when looking at the areas more activated by the co-ordinated eye- and hand-tracking task in experiment 1, it was again the ansi- form lobule and the vermal site that stood out (Fig. 6). There are two interpretations: that the increase reflects some new activation in the same or similar vicinity that is only seen in the co-ordinated task and is not seen in other tasks, or that it is the same areas which are activat- ed by either eye or hand movement and that we have simply seen the combined activity when both were being used together. We would suggest that the former inter- pretation is more likely. First, the contrast used (–1, –1, +1, 0) tests for voxels with signal levels that are signifi- cantly greater than the sum of the signal level for the eye only and hand only conditions. This makes a less- stringent prediction than the conjoint analysis available in SPM, which would test for increases in activation level of equal magnitude from eye to eye-and-hand tracking as from hand to eye-and-hand tasks. Such a con- joint prediction is not justified for our data because of the greater activation observed in the hand task than in the eye task. However, our contrast could be confounded by an increase in activity in one task (e.g. hand) and a re- duction in the other (eye), as the sum would then be smaller than the maximum. Testing the individual con- trasts (hand vs. rest and eye vs. rest) showed that all the areas concerned were increased in activity in each condi- tion, and so the use of this comparison is valid. Hence, the areas indicated in Fig. 6 are significantly more acti- vated than expected from the linear sum of the activity in the individual eye or hand tracking conditions. Supra-
linear summation does not exclude a passive additive process. However, functional activation recorded with PET or fMRI is usually a declining function of motor performance (Dettmers et al. 1995; Jenkins et al. 1997), so one might predict sub-linear, rather than supra-linear summation if there was only a passive combination of the two activation levels. We suggest that there is indeed extra neural activation seen in the co-ordinated move- ment task, over and above that seen in the two individual motor tasks.
Second, there is the behavioural evidence that co- ordinated eye and hand tracking is more effective than independent eye or hand tracking alone (Abrams et al. 1990; Biguer et al. 1984; Koken and Erkelens 1992; Van Donkelaar 1997; Vercher et al. 1994). We did not see significantly better manual tracking performance in the co-ordinated tracking condition (Fig. 3), although the total mouse movement was somewhat smaller, suggest- ing smaller or fewer intermittent corrective movements (Miall et al. 1986, 1987). However, our measures of tracking performance were relatively coarse. Further- more, we deliberately chose tracking conditions to mini- mise differences in tracking performance, so that the re- sultant functional activity would not be simply con- founded by overt changes in performance. This choice had been successful, as there were no statistically signifi- cant differences between the mouse-tracking errors or mouse-movement distances in the two comparable con- ditions. It is possible, of course, that some other uncon- trolled difference between the two tracking modes used, compensatory and pursuit, may have affected the cere- bellar activation levels. Further experiments will be needed to resolve this point.
Third, there is the evidence that the cerebellum is a key site in the co-ordinative process (Van Donkelaar and Lee 1994; Vercher and Gauthier 1988). Clearly, if the in- crease in functional activation in the cerebellum were simply the result of the addition of two independent pro- cesses, there would be no reason to expect lesions in this structure to specifically effect co-ordinated eye and hand movement, rather than effect both non-specifically.
There are many lines of evidence that suggest the cer- ebellum – particularly the lateral hemispheres – may have roles other than control of movement. Possible in- fluences on functional activation in our tasks include at- tention (Allen et al. 1997; Coull and Nobre 1998), senso- ry processing (Gao et al. 1996; Jueptner et al. 1997a), er- ror detection (Flament et al. 1996; Inoue et al. 1998) or motor learning (Flament et al. 1996; Jueptner et al. 1997b; Shadmehr and Holcomb 1997). These roles are not necessarily exclusive – we do not imply that, because the cerebellum is concerned with motor control, it cannot also be concerned with other processes (Miall and Wolpert 1996). However, Allen et al. (1997) showed that the area activated by a motor task was different from that activated by a cognitive (attention shift) task: the atten- tion task activated most commonly the left superior pos- terior cerebellum, more lateral than the activation sites we observed in this study, while their motor task activat-