A Pathway in Primate Brain for Internal Monitoring of Movements

See allHide authors and affiliations

Science  24 May 2002:
Vol. 296, Issue 5572, pp. 1480-1482
DOI: 10.1126/science.1069590


It is essential to keep track of the movements we make, and one way to do that is to monitor correlates, or corollary discharges, of neuronal movement commands. We hypothesized that a previously identified pathway from brainstem to frontal cortex might carry corollary discharge signals. We found that neuronal activity in this pathway encodes upcoming eye movements and that inactivating the pathway impairs sequential eye movements consistent with loss of corollary discharge without affecting single eye movements. These results identify a pathway in the brain of the primate Macaca mulatta that conveys corollary discharge signals.

When the brain initiates a movement, it also generates internal information that is used by sensory systems to adjust for resultant changes to peripheral receptors and by motor planning systems to prepare subsequent movements (1–7). Information about an impending movement arises as a correlate, or corollary discharge, of the neuronal movement command (8). The concept of corollary discharge has been invaluable for understanding disparate behaviors such as the circling of insects, fish, and amphibians after visual field inversion (9, 10), electrolocation in fish (4), and song learning in birds (11). In humans, psychophysical studies have demonstrated that corollary discharge signals exist (3, 5,6) and lesion studies have emphasized that the thalamus and cerebral cortex are crucial for using corollary discharge information (2, 12–14). Some neurons in the cerebral cortex of nonhuman primates receive corollary discharge signals (15–17), but where these signals come from has remained unknown.

To identify neurons as conveying corollary discharge signals, one must show that they have movement-related activity and project upstream, away from motor neurons, instead of downstream, toward motor neurons. That is, their activity must transmit information about movement without causing movement. A promising system in which to look for such neurons is that for producing saccadic eye movements. An important node in this system is the superior colliculus—specifically its intermediate layer, which contains neurons that fire just before saccade generation (18). Some projections of the intermediate layer go downstream to saccade-generating circuits in the midbrain and pons (19), and some go upstream to mediodorsal thalamus (MD) relay neurons that project to a frontal lobe region known as the frontal eye field (20). Our hypothesis is that the ascending pathway carries corollary discharges of saccadic commands.

To test this hypothesis, first we recorded from 46 MD relay neurons in Macaca mulatta (21), all physiologically verified as receiving input from the superior colliculus and projecting to the frontal eye field (Fig. 1A). We studied their activity while monkeys made delayed saccades to visual targets (Fig. 1B), a procedure that facilitates determining whether the activity is related to vision or to movement (21). Most neurons (74%; 34/46) had increased activity just before saccades (Fig. 1B) that began on average 144 ms before saccade generation (SD, 106 ms; median, 101 ms; range, 23 to 392 ms). Because this activity began before movement, it could not have represented proprioception. Most of the saccade-related neurons (82%; 28/34) were spatially tuned, firing most strongly for saccades made within a restricted range of amplitudes and directions. For all tuned neurons, the best direction was contraversive.

Figure 1

Saccade-related activity in the ascending pathway from superior colliculus (SC) to frontal eye field (FEF) via MD. (A) We recorded from relay neurons in MD that were antidromically activated by electrical stimulation in FEF and orthodromically activated by stimulation in SC. (B) Activity of an MD relay neuron that exhibited a presaccadic burst. The task performed by the monkey is depicted above, and neuronal activity (rasters of individual action potentials and spike density curve of the averaged firing rate) is shown below, with scale at right (sp, spikes). (Left) After the monkey looked at a fixation spot (dot in circle), a target appeared in the periphery (right circle). (Middle) After a delay period of 500 to 1000 ms, the fixation spot disappeared (dotted circle), which was the cue to move. (Right) The monkey then made a saccade (arrow) to the target location.

Second, we reversibly inactivated MD by injecting muscimol unilaterally at the sites of previously recorded MD relay neurons (Fig. 2A) (21). Muscimol is a γ-aminobutyric acid type A (GABAA) agonist and inhibits neuron cell bodies, not axons (22), so it should suppress MD relay neurons without affecting transthalamic fibers passing nearby. We tested monkeys on a double-step task, in which they had to make successive saccades to two flashed targets (Fig. 2B, top) (21, 23, 24). Correct execution of the second saccade requires knowledge about the first saccade's metrics. Visual feedback regarding performance is not available, because the saccades begin after the targets disappear, and proprioception is unlikely to contribute because it plays little if any role in the online control of saccades (25–28). The ability to make a correct second saccade, therefore, is thought to rely critically on corollary discharge information about the first saccade (12–14). If inactivation totally disrupts corollary discharge (Fig. 2B, bottom), a monkey will be able to make a first saccade correctly but will have no internal information that the saccade was made. If the monkey then tries to complete the trial by making a saccade to the second target location, the second saccade will travel as if starting from the fixation point. The observed effect will be a contraversive shift of the second saccade end points.

Figure 2

Corollary discharge deficits during inactivation of the ascending pathway. (A) Muscimol was injected into MD to inactivate the relay neurons. (B) Monkeys performed a double-step task. (Top) After the monkey looked at a fixation spot (dot in blue circle), two targets were flashed sequentially (yellow circles, T1 and T2). The monkey then made sequential saccades (S1 and S2) to the target locations. Contra, contraversive direction. (Bottom) If corollary discharge (CD) remained intact (left), S1 would go rightward and S2 would go straight up before and during inactivation. If corollary discharge were completely impaired (right), S1 would go rightward and S2 would go upward before inactivation but would go at an angle during inactivation so that the S2 end points would shift contraversively. (C) Individual saccadic sequences from one experiment, before and during inactivation. (D) Means (and SDs) of initial fixation locations, first saccade end points, and second saccade end points for the same example. n.s.d., Not significantly different.

Results from an example injection are shown in Fig. 2C. Before inactivation, the monkey made saccadic sequences correctly. Because the saccades were made in total darkness, first saccades were shifted slightly upward (29). Second saccades went nearly straight up, which indicates that corollary discharge was intact. During inactivation, the second saccade end points shifted contraversively, which indicates that corollary discharge was impaired. Quantitatively (Fig. 2D) (21), the second saccade end points were indeed shifted horizontally [contraversive 2.5° shift; P < 0.001 (21)] but not vertically during the injection. Neither the initial fixation locations nor the first saccade end points were shifted significantly in either direction.

We performed a total of seven muscimol experiments in which there were a total of 22 cases of before versus during saccadic sequence pairs to analyze (21). In every case, the principle for identifying a corollary discharge deficit was the same as in Fig. 2. In 82% of the cases (18/22), there was a contraversive shift in second saccade end points (Fig. 3A), and the overall mean shift (1.12°) was significantly greater than zero. The contraversive shift in 11 of these cases was individually significant [and always reversible (30)]. First saccade end points did not exhibit a significant mean horizontal shift (Fig. 3B); neither did initial fixation locations (−0.09° shift; P > 0.025). In the vertical direction, there were no mean shifts in any of the data (31). For controls, we randomly interleaved trials in which targets appeared ipsilaterally. Identical target configurations were used but were flipped across the vertical meridian. In these trials, the first saccades were ipsiversive, a direction poorly represented by MD relay neurons. Accordingly, we found no corollary discharge deficits: the mean horizontal shift for second saccade end points was not significantly different from zero (−0.41°; P > 0.025).

Figure 3

Results from all the inactivations. (A) Histogram of the horizontal shift in second saccade end points for all cases (n = 22). Mean shift is indicated. Cases that were individually significant are shown in black. Ipsi, ipsiversive; Contra, contraversive. (B) Histogram of the horizontal shift in first saccade end points. (C) Scatter of second saccade end points. Standard deviations (SD) of end-point clusters are plotted during (ordinate) and before (abscissa) inactivation. Horizontal and vertical SDs are plotted separately. Filled symbols represent significant differences during inactivation (F test, P < 0.025 criterion). SD increased during inactivation in only one case (filled circle above dashed unity line). (D) Severity of deficits for the 11 individual cases in which there was a significant horizontal, contraversive shift in second saccade end points. See text for calculation of percent deficit.

We also considered whether inactivation might have degraded a monkey's ability to see the second target or to remember its location. If visual or memory deficits occurred, there should have been greater scatter of the second saccade end points during inactivation because of greater uncertainty about the second target location. This did not occur, however (Fig. 3C). If there were subtle visual or memory deficits, they did not appear to affect performance in our task.

Although we consistently observed effects indicative of impaired corollary discharge, we never found as large a shift in second saccade end points as expected from a total deficit. In Fig. 2D, for example, second saccade end points shifted 2.5° horizontally instead of the 10° expected (cf. Fig. 2B); hence, in this case there was a 25% deficit. On average, there was a 19% deficit (Fig. 3D). We see three possible reasons for the partial deficit. (i) Other pathways may also contribute to oculomotor corollary discharge (20). (ii) Our injections may have been too small, so that we failed to inactivate all the MD relay neurons. (iii) The monkeys may have been able to exploit proprioceptive input after losing corollary discharge signals during inactivation.

In principle, the monkeys could have made preplanned sequences of saccades (32). For example, the target flashes shown in Fig. 2B could have triggered a saccadic program to “look right then look up.” With this strategy, corollary discharge signals could be ignored. We discouraged this by randomizing the target configurations across trials and modifying them between experiments (21). Inactivation never caused deficits consistent with disruption of preplanned sequences (such as generation of errant sequences or random scattering of first and second saccade end points).

Finally, we examined whether MD inactivation impaired the general ability to execute saccades. Notably, recall that inactivation did not impair first saccades in the double-step task (Fig. 3B). To test this in more detail, with four muscimol injections we also had monkeys make single saccades to visual or remembered targets at several eccentricities and directions (21). An example is shown inFig. 4A; in this experiment, there were no significant changes in single saccade accuracy during inactivation, although there were significant deficits in the double-step task (similar to the deficits shown in Fig. 2). Overall, significant changes in the accuracy (and reaction time) of single saccades were infrequent, small, and dissimilar between experiments. To examine saccadic dynamics, we plotted peak speed versus amplitude (Fig. 4B) with data from two experiments in which we elicited a broad range of amplitudes of single-step saccades. There were no clear impairments during inactivation, and logarithmic fits did not change significantly. In sum, the inactivations negligibly affected single saccades.

Figure 4

Single saccade controls. (A) Vector diagram showing mean (and SD) of single saccade end points for one experiment. All saccades started at the center and were made to one of eight possible visual targets. (B) Graphs summarizing the dynamics of contraversive and ipsiversive saccades. Curves show logarithmic fits.

These results support our hypothesis that the pathway from superior colliculus to frontal eye field via MD conveys corollary discharge information. Neurons in the pathway have activity appropriate for representing corollary discharge; interrupting this activity causes deficits consistent with loss of corollary discharge, even as the ability to make single saccades remains intact. Signals in this pathway carry information about movement but do not appear to be involved in generating movement, matching the definition of corollary discharge.

Corollary discharge signals are used not only for planning sequential movements but also for maintaining a stable visual percept despite the sudden retinal shifts caused by saccades (1,5, 6). Additional work is needed to determine whether the corollary discharge signals described here are used for such a sensory function. In particular, it would be intriguing to test whether these signals cause the presaccadic shifts in visual receptive fields of cerebral cortical neurons that are thought to help stabilize perception across saccades (15, 16,33).

  • * To whom correspondence should be addressed. E-mail: mas{at}


View Abstract

Navigate This Article