The study of Wajima and Sawaguchi (2004)
, which produced the data that we present in this commentary, trained two macaque monkeys on an ODR task with 16 possible cue locations (8 polar angles by 2 eccentricities) and a control task in which the cue remained on the screen during the delay and go periods. These tasks were quite similar to those used in the previous neurochemical studies for the DLPFC (Sawaguchi & Goldman-Rakic, 1994
; Sawaguchi & Iba, 2001
). Briefly, in the ODR task, while the monkey fixated on a central spot (a white square, 0.2 × 0.2°) for 1 s, a visual cue (white square, 0.5 × 0.5°) appeared at one of 16 peripheral locations (8 directions separated by 45° with eccentricities of 10 and 20°) for 0.5 s, followed by a 4 s-delay period during which only the fixation spot was remained on and the monkeys kept fixation on it. After the delay period, the fixation spot turned off (‘go’ signal), which instructed the monkey to make a memory-guided saccade to the cued location. When the first saccade fell inside a circular window of 5° (diameter) around the cued location within 0.7 s, a white square (0.5 × 0.5°), which was the same as the cue, reappeared as a confirmation signal and a drop of water was delivered immediately as a reward. If the first saccade fell into one of the 15 incorrect target windows, the confirmation signal appeared at this spot without delivery of reward. In this case, even if the monkey made a second saccade to the correct target location within 0.7 s from the go signal, this corrective saccade was not rewarded. Thus, to get a reward, the monkey had to acquire the correct target location with the first try. The CON task was exactly the same as the ODR task except that the visual cue remained on during the “delay” and response periods, and the monkey made a visually guided saccade to the visible target.
After the monkeys performed several 5-min blocks of these tasks alternately (12 blocks = 60 min), the experimenters antagonized GABAA receptors in the DLPFC by means of unilateral local injection of bicuculline methiodide (BMI, 2.5–5.0µg/µl, 1µl), an antagonist of GABAA receptor. The postdrug period lasted at least 10 blocks (50 min) (but, in practice, often more than 100 min). The injected sites were located rostral to the frontal eye field, which was identified physiologically by intracortical microstimulation (22 cathodal pulses of 0.3 ms duration at 333 Hz, up to 100 µA). As such, they tested the effect of GABAA receptor blockade on the behavioral performance of the monkeys.
shows the comparison of error rate on the initial saccadic response between pre- and postdrug phases in a typical experimental session, during which BMI (2.5µg, 1µl) was injected into the left DLPFC. In the ODR task, the error rate was significantly increased after the drug injection, with errors coming almost entirely on trials associated with selective cue locations in the right visual field, which were contralateral to the injected site. shows two dimensional eye traces for one animal for trials with three cue locations, separately for pre- and postdrug periods. The number of misdirected saccades was clearly increased after the drug injection for trials with those three target locations. Importantly, the performance of the control task was not impaired by BMI injection (, right panel), which indicates that the deficit on the ODR task was not due to deficits in sensorimotor processes. Thus, the data considered up to this point replicate findings from previous studies of ODR performance after circumscribed, unilateral lesions of the principal sulcus of the DLPFC (Funahashi et al., 1993
; Sawaguchi & Goldman-Rakic, 1991
; Sawaguchi & Iba, 2001
). Where they offer new insight into the concept of the mnemonic scotoma
, however, is in the behavior that followed the misdirected initial saccade.
Figure 2 Results from Wajima and Sawaguchi (2004), illustrating the effects of BMI injection in left DLPFC on initial-saccade accuracy in the ODR and control (CON) tasks. Data are from 60 min before injection and from 100 min after injection. A. Percentage change (more ...)
Although the experimental procedure of Wajima and Sawaguchi (2004)
largely replicated that of previous ODR tasks, a critical difference was the inclusion in their analyses of data acquired after the registration of the initial response. Critically, when considering the time window following erroneous initial saccades, it was observed that on most of these trials the next saccade was made to the correct target location for that trial (). That is, the corrective saccade was made into the putative mnemonic scotoma. This corrective saccade usually occurred soon after the first saccade, although by design neither first, erroneous saccades nor second, corrective saccades were rewarded. The implication of this finding is that the errors on the initial saccades can not have been due to a failure of memory, per se. Rather, a more accurate characterization of behavior on error trials is that the monkey maintained an intact memory of the cued location, but, for its initial response, selected a target that was not the remembered target.
Figure 3 A. 2-dimensional trajectories of initial saccades in the ODR task, incorporating the same data from , but with the actual (solid circle) or potential (dashed circles) target windows. Saccades falling within the solid circle correspond to correct (more ...)
To explore why, on error trials, the initial saccade selected a target different from the remembered target, the authors investigated whether there was any systematicity to the direction of these initial saccades. The first observation was that erroneous saccades tended to select one of the 15 alternative target locations that had not been cued on that trial (). Thus, the initial saccade was to a learned target location, not just a haphazard movement. Secondly, and strikingly, the target selected by erroneous initial saccades tended to be related to a location that had been relevant on the previous trial, in that it had been the cued location and/or the target acquired by the initial saccade on the immediately preceding trial. This is illustrated for the three locations from in , and for all error trials in . To quantify this pattern, the discrepancy in degrees between the direction that would be required to acquire the current target (i.e., to respond correctly) and the direction of the actual saccadic response was computed for each error trial (gray bars in ), and this was compared with the discrepancy between the direction that had been required on the previous trial and that of the actual saccadic response (black bars in ). Interestingly, the value was smaller in the latter case, reflecting the fact that, on error trials, the monkey tended to choose the target that had been cued in the previous trial more often than the actual target location of the current trial (). This kind of error occurred more often when the previous trial’s target had been located in the visual field ipsilateral to the injected hemisphere (i.e., in the “good” visual field) than when it had been in the contralateral field ().
The effect of BMI was tested in a total of 47 injection sites. Of them, 36 sites yielded a significant deficit in the ODR task. In separate experimental sessions, saline was injected into 12 of the 36 effective sites, and it was confirmed that the saline injection had no significant effects either on the ODR or on the CON tasks relative to predrug performance.