We applied electrical stimulation to physiologically identified sites in macaque middle temporal area (MT) to examine its role in short-term storage of recently encoded information about stimulus motion. We used a behavioral task in which monkeys compared the directions of two moving random-dot stimuli, sample and test, separated by a 1.5-s delay. Four sample directions were used for each site, and the animals had to indicate whether the direction of motion in the sample was the same as or different to the direction of motion in the test. We found that the effect of stimulation of the same directional column in MT depended on the behavioral state of the animal. Although stimulation had strong effects when applied during the encoding and the storage components of the task, these effects were not equivalent. Stimulation applied during the presentation of the sample produced signals interpreted by the monkeys as directional motion. However, the same stimulation introduced during the period of storage no longer produced signals interpreted as unambiguous directional information. We conclude that the directional information used by the monkeys in the working memory task is likely to be provided by neurons in MT, and the use of this information appears to be dependent on the portion of the task during which stimulation was delivered. Finally, the disruptive effects of stimulation during the delay suggest that MT neurons not only participate in the encoding of visual motion information but also in its storage by either maintaining an active connection with the circuitry involved in storage or being an integral component of that circuitry.
The perception of visual motion depends on the activity of directionally selective cortical neurons (Newsome et al. 1989; Pasternak et al. 1985). Such neurons are found in a number of visual cortical areas, but are particularly common in the middle temporal area (MT), widely acknowledged as an important motion processing area (for reviews see Albright 1993; Andersen 1997). Single neurons in this area have been shown to perform integration of local motion signals (Movshon et al. 1985; Rodman and Albright 1989), and to discriminate motion direction in the presence of noise with nearly the same sensitivity as monkeys and humans (Britten et al. 1992). The role of MT neurons in the discrimination of stimulus direction has been demonstrated in microstimulation studies, which have shown that low current stimulation of directionally selective columns in MT can introduce a directional bias that affects monkey's choices in a direction discrimination task (Salzman et al. 1992).
While there is strong evidence supporting the role of MT neurons in the encoding of stimulus direction, little is known about their involvement in the temporary storage of this information. To address this issue we applied microstimulation to direction-selective columns in MT of monkeys performing a behavioral task that required them to briefly retain recently encoded information about stimulus direction. We compared the effects of microstimulation during encoding of visual motion with the effects of stimulation during its temporary storage. We found that when stimulation was applied during encoding, the monkeys generally behaved as if they were presented with the direction preferred by the stimulated neurons. The same stimulation applied during the delay also affected performance, but in a less predictable way. Thus the interpretation of signals produced by stimulation of MT appears to be dependent on the portion of the task during which stimulation was delivered. The effectiveness of stimulation during the delay suggests MT maintains an open line of communication with the circuitry that underlies the short-term retention of visual motion.
Two young adult male monkeys (Macaca nemestrina) were used in this study. On weekdays, water was restricted for a period of 22 h before testing, and their daily water ration, in the form of fruit juice, was provided during the behavioral testing. On weekends, the monkeys were not tested behaviorally and received 100 ml/kg water per day. Food was continually available in the home cage, and monkeys received supplements of fresh fruit and vitamins daily. Body weights were measured three times a week to ensure good health and normal growth. Experiments were carried out in accordance with the guidelines published in the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH publication no. 86-23, revised 1987).
After they had adapted to the apparatus and learned the basic behavioral task, both monkeys had scleral search coils and head restraint devices implanted for monitoring eye position (Remmel 1984). For all surgical procedures, the animals were initially anesthetized with ketamine hydrochloride (15 mg/kg im) and maintained with 3% isoflurane. All surgical procedures were performed under strict aseptic conditions. Craniotomies were made over parietal cortex and a commercial recording chamber (Crist Instrument, Hagerstown, MD) was implanted. The chamber was 20 mm diam and was attached to the skull by a ring of bone cement anchored to six to eight titanium screws evenly distributed around the craniotomy. The chamber was placed above the superior temporal sulcus (STS), allowing a dorsal approach to MT. The precise location and shape of the STS for each monkey was determined from T2-weighted magnetic resonance images (MRIs) obtained in 2T and 1.5T GE magnets with a small surface coil. This procedure has been described previously (Bisley and Pasternak 2000; Rudolph and Pasternak 1999). Briefly, the monkeys were anesthetized with pentobarbitol sodium (25 mg/kg iv), and placed in a specially constructed MRI-compatible stereotaxic frame. Coronal and horizontal scans were performed with the following parameters: TE/TR: 5000/90 or 3000/85; 1.5-mm-thick slices 0.2 mm apart; 256 by 256 array, field of view: 10 or 15 cm.
Control and calibration of eye position
Eye position was monitored with magnetic search coils. During behavioral testing, each monkey was placed in a magnetic field, generated by a 46-cm field coil, with its head held firmly by a head restraint device. Eye position was calibrated prior to each daily testing session by rewarding the monkey for positioning its gaze within an electronically defined 0.75–1° window centered on a fixation spot and maintaining it within this window for 700–1,000 ms. The spot was positioned in random locations on the display screen within the central 15°.
Recording from MT
Multi-unit activity in MT was recorded through Tungsten microelectrodes (0.5–2.0 MΩ; FHC) while the monkeys passively viewed a fixation stimulus (a small cross) for variable periods of time ranging from 3 to 7 s. The electrode was inserted through a guide tube positioned in a grid (Crist et al. 1988). We used criteria similar to those used by Salzman et al. (1992)to identify stimulation sites. Once a directionally selective site was identified, it was mapped with a coherently moving random dot field. The optimal speed, dot density, and size of the receptive field were then determined by selecting the parameters that produced maximal firing rates and the strongest directional selectivity. A direction selectivity profile was recorded with the optimal stimulus moving in one of the eight cardinal directions. The direction that produced the highest firing rate was taken as the “preferred” direction and was used as the basis for selecting the stimulus directions to be used in the behavioral task. The site's direction selectivity (ds) was calculated by
Stimuli and behavioral procedures
RANDOM DOT STIMULI.
The random-dot stimuli were identical to those used in previous studies (Bisley and Pasternak 2000; Rudolph and Pasternak 1999) and consisted of dots placed randomly within a circular aperture, the size of which was matched to the receptive field size of the recording site. Each dot was displaced by a constant step size (Δx) and temporal interval (Δt = 13 ms). The direction of motion for each dot was randomly chosen from a specified uniform distribution of directions. The lifetime of an individual dot was equal to the duration of the stimulus presentation (500 ms). The dots, viewed at a distance of 42 cm, were 0.03° diam, and their luminance was set about 3.5 log units above human detection threshold. The values for dot density and speeds used in each testing session were based on the optimal values determined for each stimulated site. The dot densities ranged from 0.74 to 7.56 dots/deg2 and the velocities ranged from 3.85°/s (Δx = 0.05°) to 40.0°/s (Δx = 0.52°).
The behavioral task was identical to that used previously and has been described in detail elsewhere (Bisley and Pasternak 2000; Rudolph and Pasternak 1999). The monkeys initiated each trial by fixating a small spot for 750–1,000 ms within a 1.5° window. During each trial two stimuli, the sample and test, were presented for 500 ms in the same retinal location (Fig.1). The stimuli were separated by a 1.5-s delay, and the animals were required to report whether the directions of motion in the two stimuli were the same or different by pressing the right or left response button, respectively. A correct response was rewarded with a drop of fruit juice. An incorrect response was followed by a 3- to 6-s tone and no reward. The animals were required to maintain fixation throughout the trial, and a break in fixation during the trial resulted in a brief tone and termination of the trial. The direction of motion in the sample stimulus was varied from trial to trial and was picked randomly from a set of four selected on the basis of the preferred direction of the site. The four directions included the preferred direction for a given site, the direction opposite to the preferred (the null direction), and the two directions orthogonal to the preferred. The sequence of presentation of same/different stimuli was also randomized from trial to trial, producing approximately equal numbers of trials for each condition.
The stimulation protocol was based on that of Salzman et al. (1992). Stimulation events lasted for 500 ms and consisted of 100 biphasic pulse-pairs at 200 Hz. Each pulse-pair consisted of a leading 0.2-ms 80-μA positive pulse separated from a 0.2-ms 80-μA negative pulse by a 0.1-ms delay (A.M.P.I. Master-8; FHC bp Isolator).
For each selected site in MT, the effect of stimulation was measured during a testing session consisting of 400–800 trials separated by a 3-s inter-trial interval. During each testing session, stimulation was applied on 25% of trials chosen at random. It either began 5 ms after the onset of the sample stimulus or during the middle 500 ms of the 1.5-s delay (Fig. 1 B). In sites that were stimulated both during the sample and during the delay, trials in which stimulation occurred during the sample were presented in separate blocks from trials in which stimulation occurred during the delay. The blocks contained 50–60 trials and were interleaved 4–6 times for each site. As a rule, each daily session began with a block of trials in which stimulation was applied during the sample, followed by a block of trials in which stimulation was applied during the delay. Each stimulation session was concluded by trials during which stimulation was applied again during the sample. Data for a single paradigm were combined across blocks. In some sites the retinotopic specificity of microstimulation was studied by presenting the visual stimulus in the diagonally opposite visual quadrant to that of the receptive field. In these experiments stimulation only occurred during the sample period.
For most of the experiments, the dots in the sample stimulus moved in a broad range of directions. The reason for using these complex motion stimuli was that normal discrimination of their motion requires an intact MT (Bisley and Pasternak 2000; Rudolph and Pasternak 1999). The direction range in the sample stimulus was set to 180, 240, or 250°, the values approximately 1.3–1.7 times above range threshold for each monkey for a given eccentricity, target size, and speed. With these stimuli the monkeys performed the task at 85–100% correct. In one set of experiments, the sample was composed of stationary dots. Such stimuli were presented only on stimulation trials. In all experiments the test stimulus was composed of coherently moving dots (i.e., 0° direction range), which moved in the direction that was the same or opposite to the direction of motion in the sample. During all the microstimulation trials, the animals were rewarded only for correct responses. Thus to obtain the reward the animals would have to use the information from each of the two visual stimuli and ignore any effects of stimulation. In the trials in which stimulation was applied during the presentation of a sample stimulus containing no motion (stationary dots), animals were rewarded 50% of the time.
The effect of stimulation on performance was evaluated by comparing percent correct on stimulation and nonstimulation trials. In addition, the proportion of trials in which the direction of sample was reported to be the same as test direction was also computed and compared between paradigms. The validity of any difference was tested, for each direction within each session, with Fisher's Exact test.
Stimulation was applied to 67 sites in MT in 2 monkeys. Prior to the experiment, we mapped the retinotopic organization of MT in each monkey. The sites in MT were identified by the size of their receptive fields and their strong directional selectivities (0.73 ± 0.14, mean ± SD). The receptive field sizes ranged from 2 to 12° at eccentricities between 3 and 18°, consistent with published reports (Albright and Desimone 1987; Desimone and Ungerleider 1986).
Microstimulation during sample presentation
Microstimulation was applied to each site after mapping its receptive field, determining the optimal size and measuring its directional selectivity. The direction selectivity profile of activity of a stimulation site is shown in Fig.2 A. In this example, the preferred direction of motion was to the right (0°). The effect of stimulation on performance is shown in Fig. 2 B for each sample direction. On nonstimulation trials (gray columns) the monkey performed the task at levels approaching 100% correct for each of the four sample directions. The black columns show the animal's performance on stimulation trials. When the sample moved in the direction opposite to the preferred direction of the stimulated site (i.e., 180°), stimulation drastically reduced the performance from 100% to about 10% correct (P < 0.01, Fisher's Exact test). Stimulation also significantly affected the performance when the sample moved upward (P < 0.01) but had no detectable effect for the other two sample directions.
The nature of our task allowed us to not only evaluate the accuracy of an animal's performance in terms of percent correct, but also to gain an insight into the directional specificity of the motion signal induced by electrical stimulation (Fig. 2 C). This could be determined because on each trial the animal was required to judge the direction of the sample stimulus as the “same” or “different” to that of the test stimulus. Thus we computed the percentage of trials for which the monkey indicated that the stimulated sample direction was the same as the test direction.
Figure 2 C shows the results of this type of analysis for the example site. In each testing session the sample and test moved in the same direction on approximately half of the trials. Thus perfect performance would require the animal to report that the two stimuli moved in the same direction on approximately 50% of trials. The “same as test” choices for nonstimulation trials are shown for each of the four test directions by the gray columns, while the “same as test” choices made on stimulation trials are represented by the black columns. The data show that whenever the test moved in the preferred direction (rightward, 0°), the animal always equated the direction of the stimulated sample with that direction, irrespective of the actual direction of the sample. Furthermore, the monkey almost never indicated that the sample moved in the direction opposite to the preferred. When the test moved in the orthogonal directions, the monkey more often reported that the stimulated sample was the same as motion downward (90°) than the same as upward motion.
For each site we calculated an index that showed the proportional change in performance for each direction. The index (I) of the stimulation effect was calculated by and gives an indication of the strength of the effect with an index of 0 representing no effect and indexes >0.5 representing strong effects. Note that an index of 0.5 generally represents performance at chance levels, whereas indexes >0.5 represent performances that are worse than chance. In the example shown in Fig. 2 B,the indexes were −0.11, −0.05, 0.91, and 0.46 for the directions 0, 90, 180, and 270°, respectively.
EFFECTS OF STIMULATION ON PERFORMANCE.
Figure 3 illustrates the effects of stimulation applied during the presentation of the sample for all of the sites in the study. The histograms in Fig. 3 A show the distribution of indexes for all sites for each of the three classes of sample directions: preferred, null, and orthogonal to preferred. Every site in the population had at least one sample direction of motion that was significantly affected by stimulation (P < 0.01, Fisher's Exact test performed on each direction from each session independently). In the majority of sites, there was little measurable change in performance when the sample moved in the preferred direction (Fig. 3 A, top), as indicated by the low indexes. On the other hand, when the sample moved in the direction opposite to the preferred (Fig. 3 A, middle) the indexes were generally large, indicating strong effects of stimulation. Thus for the majority of sites the effects of stimulation were similar to the example shown in Fig. 2.
The effects of stimulation when the sample moved in directions orthogonal to the preferred were less consistent, although a number of sites were strongly affected by stimulation (Fig. 3 A,bottom).
ANALYSIS OF THE “SAME AS TEST” RESPONSES.
The analysis illustrated in Fig. 2 C was applied to the data for each stimulation site. We calculated the percentage of stimulated trials on which the animal equated the direction of the sample with that of the test. The data were analyzed separately for the three categories of test directions defined with respect to the directional selectivity of each site: preferred, null, and orthogonal to the preferred (Fig. 3 B). Columns on the left of the histograms (toward 0%) represent sites in which the monkeys frequently judged the direction of sample to be different from that of the test. Columns on the right (toward 100%) represent sites for which the monkeys consistently reported that the direction of the sample was the same as that of the test. The distribution of the “same as test” choices during nonstimulation trials (light gray columns in Fig. 3 B), was centered on 50%. This distribution represents performance with few errors, as the number of trials in which the sample and test moved in the same direction was approximately equal to the number of trials in which the sample and test were different. The data from the stimulation trials are shown in the outlined dark and hollow columns. We compared the percent of “same as test” choices for stimulation and nonstimulation trials for each test direction within a session with Fisher's Exact test. The dark gray columns represent sites in which the responses were significantly different (P < 0.05). The hollow columns represent sites for which there was no significant difference between the two types of trials.
The data show that in well over half of the sites, stimulation resulted in the animals equating the direction of the sample with the preferred direction of a given site (Fig. 3 B, top), irrespective of whether the actual direction of motion was in the preferred direction or opposite to it. Furthermore, in the majority of sites, the animals acted as if the direction of the stimulated sample was not the same as the null direction (Fig. 3 B, middle). This suggests that the animals generally based their choices on a strong directional signal in the preferred direction of the stimulated site.
There were a small number of sites (5/44) for which the monkeys reported that the direction of motion in the stimulated sample was not in the preferred direction (Fig. 3 B, top, columns toward the left). It must be noted that for the same sites the animals equated the direction of motion in the stimulated sample with the null direction suggesting that stimulation of these sites produced a strong effect in the null direction.
In about 20% of sites, the monkeys equated the direction of sample with one of the two directions orthogonal to the preferred (bottom). When this occurred, they almost never equated the direction of the stimulated sample with the other orthogonal direction.
Microstimulation during the delay
Stimulation applied during the delay did not produce the same pattern of results as stimulation during the sample. The data in Fig.4 show the effects of stimulation during delay for the same site as that shown in Fig. 2. The performance dropped to chance in trials with the sample moving in the nonpreferred direction (180°), but it never reached the below chance levels seen when the stimulation was applied during the sample (compare Figs.2 B and 4 A). The analysis of the “same as test” decisions (Fig. 4 B) shows that stimulation during the delay did not induce the animal to consistently report that the stimulus moved in one particular direction, although the monkey was more likely to report that the sample moved in the direction preferred by the stimulated site (see left black column, Fig.4 B).
EFFECTS OF MICROSTIMULATION ON PERFORMANCE.
For about 84% of the sites, stimulation significantly affected the performance on at least one sample direction. When the sample moved in the preferred direction (Fig. 5 A,top), there was little measurable change in performance for most of the sites. This distribution of indexes was not significantly different from the distribution shown in Fig. 3 A (P = 0.36, Wilcoxon signed-ranks test), although it appeared that fewer sites were strongly effected by stimulation during the delay. On the other hand, a comparison of the distributions of the effects measured for sample stimulation versus delay stimulation for the null direction revealed significant differences (P < 0.01). When stimulation was applied during the delay, fewer sites had large index values, indicative of the introduction of a directional signal, but more sites had indexes around 0.5, indicative of performance at chance levels (compare middle panels in Figs. 3 A and5 A).
The effects of stimulation applied during motion in directions orthogonal to the preferred was not significantly different from the effects of stimulation applied during the delay in the same conditions (P = 0.056). However, there were two clear trends (compare bottom panels in Figs. 3 A and5 A). First, there were more sites that were not affected by stimulation during the delay. Second, for sites affected by stimulation during the delay, the indexes were more evenly spread between values of 0.5 and 1.0.
ANALYSIS OF THE “SAME AS TEST” RESPONSES.
Stimulation during the delay produced reports of motion in the preferred direction in only about 40% of sites (Fig. 5 B,dark gray columns, top), and when the monkeys did indicate that motion was in the preferred direction, they did so less often than in the sample stimulation condition (such as in the example in Fig.4 B). Furthermore, in only about 30% of sites stimulated during the delay did the animals indicate that the remembered sample was not in the null direction (Fig. 5 B, middle). This suggests that stimulation during the delay was less effective at convincing the animals that motion was not in the null direction. It is interesting to note that unlike the effects of stimulation during sample, we found that for only 16% of the sites the animals equated the sample with the test moving in the preferred direction (Fig.5 A, top, dark gray columns) and also did not equate the sample with the test moving in the null direction (Fig. 5 B, middle, dark gray columns). This suggests that stimulation during the delay did not produce unambiguous directional information, as was the case with stimulation during the sample.
The data in the bottom panel of Fig. 5 B show that stimulation of half of the sites also affected responses to the sample moving in the orthogonal directions. Responses to such stimuli ranged from always equating the direction of the sample with an orthogonal direction to never equating the direction of the sample with an orthogonal direction.
In summary, although the effect of stimulation during the delay was quite strong, it was less consistent with respect to the preferred direction than stimulation during the sample. These results suggest that the behavioral effects of stimulation depended on the portion of the task during which stimulation was delivered.
Effects of stimulation during the presentation of a stationary sample
In the experiments involving stimulation during the sample, the animals appeared to use the information provided by stimulation rather than by the visual stimulus. We examined the robustness of this observation by stimulating during a sample that consisted of stationary dots and thus contained no motion information. On nonstimulation trials the sample stimulus was moving and was identical to the motion stimuli used in other experiments. Data from six sites were recorded and analyzed by assessing how often the animals judged the direction of motion in the stimulated sample as the same as the direction of the test stimulus.
Figure 6 A illustrates the direction selectivity of a site in MT stimulated in one of these experiments, while Fig. 6 B shows a comparison of the effects of stimulation of this site during moving (■) and stationary (▧) samples. In this example, the preferred direction was leftward (180°). On stimulation trials, the monkey always equated the sample with the leftward and not the rightward moving test, irrespective of whether the sample was moving or not. The same was true for the two orthogonal directions. The comparison of the “same as test” reports for stationary and moving sample stimuli for all six stimulation sites is shown in Fig. 6 C. Each point represents trials for a single direction of the test stimulus from each site. The solid circles show the “same as test” reports from sites significantly affected by stimulation for both the stationary and moving samples (P < 0.05, Fisher's Exact test). Open circles represent sites for which the percentage of the “same as test” reports on stimulated trials was not significantly different from the percentage of reports on nonstimulated trials.
All six sites are equally represented by the solid points, which are fitted by the regression line (slope, 0.71 and R = 0.832, P < 0.001). The data show that the behavior in the presence of the moving and the stationary stimuli was similar. If the monkey equated the direction of the moving sample with a given test direction, he was also likely to do so when the sample contained no visual motion. This was true even in cases where the sample moved in an orthogonal direction (40% of the solid points). This result suggests that signals produced by stimulation during the sample were strongly directional and were used by the monkeys in place of the visual stimuli.
Retinotopic specificity of the stimulation effect
Although the effect of stimulation during the sample was directionally specific and was usually related to the directional selectivity of the stimulated site, we wanted to determine whether this effect was also spatially selective. Such spatial selectivity would provide additional evidence that the directional signals used by the monkeys on stimulation trials were provided by a retinopically organized region, such as area MT. We compared the performance when stimuli were placed in the stimulated portion of the visual field representation in MT to the performance when stimuli were placed at equal eccentricity in the visual quadrant diagonally opposite to the site being stimulated (Fig.7 A). The directional selectivity profile of a strongly directional site in MT and the effects of stimulation, both at the site of behavioral testing and in the remote location, are shown in Fig. 7 B. The data demonstrate that while stimulation in the location of the behavioral task had a strong directional effect, stimulation of the same site had no effect when the stimuli were presented in the remote location. This experiment was performed with four directions for each of the 10 MT sites, and we found that for all directions but one the performance during stimulation in the remote location was not significantly different from the performance measured during the nonstimulation trials (P > 0.05, Fisher's Exact test). These results are in accordance with those of Salzman et al. (1992)and demonstrate that the directionally selective effects of microstimulation during the presentation of the sample are spatially localized to the portion of the visual field engaged in a behavioral task.
We found pronounced effects of microstimulation during the performance of the match-to-sample task, and the nature of these effects depended on the portion of the task during which stimulation was delivered. When stimulation was applied during the encoding phase of the task, the animals appeared to use directional signals produced by electrical stimulation rather than by the visual stimulus. Stimulation during the retention period of the task also produced a strong effect on performance. However, the nature of this disruption suggests that stimulation did not produce a clear directional signal that the monkey used to complete the task.
Microstimulation during the sample
The behavioral task used in this study required the monkeys to compare two moving stimuli and to judge their directions as the “same” or “different.” This provided us with information about the identity of motion signals used by the monkeys on stimulation trials. On the majority of such stimulation trials, the monkeys reported that the direction of the sample stimulus was the same as the preferred direction of the stimulated site in MT. Thus it is likely that neurons activated by microstimulation were the source of signals used by the monkeys to make the discrimination.
In the majority of cases, the effect of stimulation could be predicted from the directional selectivity of a given site since the animals behaved as if stimulation during the sample produced a percept of motion in the preferred direction. In a small number of cases, the monkeys behaved as if stimulation produced a percept in the null direction (see Fig. 3 B, middle, columns on theright). These cases may be explained by inadvertent shifts in the position of the stimulating electrode, after the directional selectivity of a specific site had been determined and prior to the microstimulation experiments. There is also the possibility that the precise location of the tip of the stimulating electrode within the column might have an influence on the behavioral effects.
Romo et al. (1998) and Salzman et al. (1992) have suggested that electrical microstimulation in sensory areas may produce signals that animals cannot differentiate from sensory stimuli. In our experiments, the monkeys also behaved as if they did not distinguish between the electrical and visual stimulus. This was particularly apparent in the control experiment involving the stationary sample. There was no indication that the animals were experiencing anything unusual since we saw no obvious changes in response latencies, eye movements, or other changes in behavior commonly observed when the monkeys are exposed to anything out of the ordinary, even in the first stimulation trials the animal experienced. Furthermore, the same was true when stimulation was applied during the delay; the monkeys' overt behavior was the same on stimulation and on nonstimulation trials.
A likely explanation of our effects is that the electrically produced directional signals were stronger than the visual stimulus and thus were more likely to be used by the monkeys (Salzman and Newsome 1994). This “winner takes all” mechanism would predict stimulation effects to be largely independent of the direction of the visual stimulus. This hypothesis is supported by our data collected from trials in which stimulation was applied during the presentation of a stationary sample. On these trials, the monkey treated the stationary stimulus in the same way it treated the moving stimulus and consistently equated both with the preferred direction of the stimulated site. The similarity of the effects obtained with stationary and moving sample stimuli argue against the possibility that the information provided by the visual stimulus was used in combination with the directional signal produced by electrical stimulation. Rather, the monkeys appeared to use signals produced by stimulation in place of the visual stimuli.
AMPLITUDE OF STIMULATING PULSES.
We performed eight additional experiments in which we determined whether the same effects observed with 80-μA pulses could be elicited at lower currents. The current was varied in blocks of interleaved trials, and we found that in one case the same effects could be produced with currents as low as 20 μA. Generally, pulses of 40 μA were more effective, producing reliable effects in half of the sites, while pulses of 60 μA elicited the effects in five of the sites. In the remaining three sites, only 80-μA pulses produced significant effects. From these observations it appears that the amplitude of stimulation is only one of the factors contributing to the size of behavioral effects we observed. Additional factors may include the precise positioning of the stimulating electrode and the selection of a specific site since some sites were affected more and some less by the same amplitude of stimulation.
COMPARISON WITH OTHER STUDIES.
Newsome and his colleagues were the first to use microstimulation to study the role of directionally selective neurons in direction discrimination (Celebrini and Newsome 1995;Murasugi et al. 1993a,b; Salzman et al. 1992). They showed that low current stimulation of directionally selective columns in areas MT and MST (medial superior temporal) introduced a directional bias that affected monkey's choices in a direction discrimination task. In the present study we also measured the effect of stimulation of directional columns in MT on direction discrimination. However, in this study we used higher current stimulation and a more complex behavioral task in which the monkey was required not only to identify stimulus direction but also to briefly store this information. Despite these procedural differences, there were a number of similarities in both sets of studies: stimulation during the presentation of the visual stimulus affected the behavior of the monkeys; these effects reflected the preferred direction of the stimulated directional columns, and they were spatially localized to the portion of the visual field affected by stimulation.
There were also differences between the effects measured in the present study and those reported by Newsome's group. For example,Murasugi et al. (1993a) found that in contrast to low-amplitude stimulation that introduced directional bias, high-amplitude stimulation (similar to that used in this study) resulted in performance dropping to chance levels, an effect consistent with the introduction of noise into the system. On the other hand, we found that the same high current levels resulted in apparently unambiguous signals that were used by the monkeys in a very consistent manner. The explanation for this difference may lie in the difference between the behavioral tasks. For instance, Murasugi et al. (1993a) stimulated for a full second during which time the animal not only had to extract information about stimulus direction but also plan a saccadic response. In our task, stimulation during the sample occurred during the encoding portion of the task and before the response could be planned.
The second difference is that, unlike Murasugi et al. (1993b), we observed a strong directional effect of stimulation during the presentation of a stationary dot pattern. One possibility is that in combination with the difference in the behavioral task, the greater current used in the present study may have produced more robust effects.
Stimulation during the delay
One of the more striking findings of our study was that stimulation applied during the delay, when no visual stimulus was present, had a profound effect on the animals' behavior. Generally, unlike stimulation during the sample, the animals did not behave as if we had injected a clear directional signal. Rather, a large proportion of sites were strongly affected, with performance often degrading to chance levels for at least one sample direction and the animals only rarely making consistent decisions about the direction of the remembered stimulus.
There is a possibility that stimulation during the delay produced an explicit visual percept and the behavioral response was a product of matching the test direction to this percept. However, in that case we would expect one of two possible behavioral outcomes. First, we expect that the monkeys would learn to ignore the effect of stimulation since it often brought their performance down from 95 to 50% correct. The fact that we did not see a diminution of the effect over the course of the study argues against this possibility. Second, the monkeys could have used the electrical signal delivered during the delay to complete the task. In this case, we would expect the performance to be the same as the performance with stimulation applied during the sample. This was not the case. We found that stimulation during the sample that was strong enough for the monkeys to behave as if motion was present when it was absent did not produce the same pattern of behavioral choices when it was applied during the delay. Thus it is unlikely that the animals were explicitly matching an induced percept of motion with the test direction.
Seidemann et al. (1998) found that electrical stimulation applied after the presentation of the visual stimulus in some cases produced effects similar to the effects of stimulation applied during the visual stimulus, with a strength that weakened over time. When stimulation was applied within 1 s after the visual stimulus, two of their four animals showed no effect; the other two showed effects equal to those found with stimulation during the visual stimulus. The authors were able to train the two affected monkeys to better ignore the stimulation and proposed the existence of a mechanism, downstream from MT, which could gate out motion information that was not relevant to the current behavioral task.
The present study differs from that of Seidemann et al. (1998) in one important respect. In our study, stimulation was applied while the monkey was remembering the sample stimulus and awaiting the presentation of the test. Thus stimulation occurred well before a perceptual decision could be made and the response planned.Seidemann et al. (1998), on the other hand, applied stimulation after the monkey had seen the stimulus, identified its direction and most likely had completed planning of the response saccade. Thus in the two studies, stimulation was delivered at different stages of the behavioral task. Despite these differences, there is a striking parallel between the two studies. Just likeSeidemann et al. (1998), we found that the effect of stimulation of the same directional column in MT depends on the behavioral state of the animal. Thus the mechanism underlying the gating proposed by Seidemann et al. (1998) might also play a role in explaining our results. During the presentation of the sample stimulus, a signal coming from MT was readily interpreted as a directional signal. On the other hand, the same signal introduced during the period of storage was no longer interpreted by the circuitry as unambiguous directional information. Rather, it appeared to introduce some confusion. This confusion could be due either to the introduction of noise or the introduction of a weak directional signal. Our results do not give us a clear indication of which of these two possibilities is more likely. In either case, the different effects of stimulation during the sample and during the delay suggest that the interpretation of the electrical signal appears to be dependent on the component of the task during which stimulation was applied.
The fact that stimulation during the delay had a strong, albeit not clearly directional, effect suggests that after MT neurons complete the process of encoding of the visual stimulus they are unlikely to stop participating in the execution of the task. Rather, our results suggest that they maintain an active connection with the circuitry involved in storage of this information or they may be an integral part of that circuitry.
A number of recent findings strengthen the notion that MT neurons may be involved in the performance of our task. First, microstimulation and lesion experiments have shown that MT neurons play an important role in the direction discrimination component of the present task (Pasternak and Merigan 1994; Rudolph and Pasternak 1999; Salzman et al. 1992). Second, a recent lesion study has demonstrated that normal retention of complex motion signals requires the presence of MT (Bisley and Pasternak 2000). Finally, we have recently found that some MT neurons show a transient increase in activity during the delay while monkeys perform a task identical to that used in the present study (Droll et al. 2000).
The observation that a cortical area, primarily sensory in nature, is involved in the retention of specific sensory information, provides direct support for the proposal put forward by Fuster (1997) that the same systems involved in sensory processing also participate in retaining sensory information. The evidence supporting this view comes from imaging studies in humans (Courtney et al. 1997), as well as from neurophysiological experiments in monkeys in other sensory systems (Zhou and Fuster 1996, 1997). The present result strongly supports this idea by showing that neurons in area MT, which represent an important stage in the analysis of motion information, are not only involved in processing of complex motion, but are also in active communication with areas involved in the temporary storage of this information.
We thank M. Shadlen, L. Abbott, and B. Merigan for insightful comments on the manuscript. We also thank J. Nichols and B. Newsome for advice on microstimulation and A. Bahirwani, M. Schear, and B. Singer for software development.
This work was supported by National Eye Institute Grants R01 EY-05911 (T. Pasternak) and (in part) P30 EY-01319 (Center for Visual Science).
Present address of J. W. Bisley: Laboratory of Sensorimotor Research, National Eye Institute, 49 Convent Dr., Bldg. 49, Bethesda, MD 20892.
Address for reprint requests: T. Pasternak, Dept. of Neurobiology and Anatomy, Box 603, University of Rochester, Rochester, NY 14642 (E-mail:).
- Copyright © 2001 The American Physiological Society