Two previous studies have shown that frontal–temporal disconnection in monkeys, produced by unilateral ablation of frontal cortex in one hemisphere and of visual inferior temporal cortex in the opposite hemisphere is entirely without effect on visual object–reward association learning in concurrent discrimination tasks. This is a surprising finding in light of the severe impairments that follow frontal–temporal disconnection in many other tests of visual learning and memory, including delayed matching-to-sample and several conditional learning tasks. To explore the limits of this preserved object-reward association learning, we trained monkeys on visual object discrimination learning set (DLS) prior to frontal–temporal disconnection. As a result of training with single object–reward associations, the monkeys acquired a proficient learning set, evidenced by the rapid learning of new single object–reward association problems. This rapid learning was not affected by unilateral ablations of either inferior temporal cortex alone or frontal cortex alone but was severely impaired after final surgery to complete the disconnection. Moreover, each individual monkey now learned single object–reward association problems at the slow rate at which that individual had learned such problems before the formation of learning set. This result shows that frontal–temporal disconnection abolishes visual learning set.
In monkeys, bilateral lesions of frontal cortical areas cause deficits on a large number of different memory tasks. Conditional learning, object recognition memory, reversal learning, visual discrimination learning, memory for multiple spatial locations, and the delayed response task have all been reported to be impaired by bilateral lesions to prefrontal cortical areas (Brush and others 1961; Butter 1969; Passingham 1985; Petrides 1985; Bachevalier and Mishkin 1986; Gaffan and Harrison 1989; Kowalska and others 1991; Meunier and others 1997; Rushworth and others 1997; Parker and Gaffan 1998a, 1998b; Levy and Goldman-Rakic 1999; Bussey and others 2001; Browning and others 2005). However, the performance of each of these visual tasks requires several potentially independent cognitive and motivational processes, and many studies describing deficits following bilateral frontal lesions do not indicate whether frontal cortical involvement in those tasks is with the visual processing of individual objects, or rather with some nonvisual aspects of motivation or executive control.
One way to attempt to resolve this issue is by testing the same tasks after disconnecting the frontal cortex from inferior temporal cortex. This disconnection can be achieved by removing frontal cortex in one hemisphere and removing inferior temporal cortex, which is known to be essential for visual object– identity processing, in the opposite hemisphere. The routes of corticocortical interaction between frontal cortex and inferior temporal cortex are predominantly ipsilateral; this surgical manipulation therefore severely reduces the interaction between processing in frontal cortex and in inferior temporal cortex. In principle, this disconnection therefore spares nonvisual aspects of frontal cortical processing because the remaining intact frontal lobe is free to interact with areas of the brain in support of other nonvisual cognitive functions. By this reasoning, investigation of frontal–temporal disconnection adds to the knowledge that can be gained by investigating bilateral frontal lesions.
Frontal–temporal disconnected monkeys are impaired at delayed matching-to-sample (Parker and Gaffan 1998a; Bussey and others 2002), visuomotor conditional learning (Bussey and others 2002), reward-visual conditional learning (Parker and Gaffan 1998b), a strategy implementation task in which different classes of objects require different patterns of responding (Gaffan and others 2002), and discrimination learning within complex scenes (Browning and others 2005). Two studies have shown, however, that monkeys with frontal–temporal disconnection learn object-reward associations at an entirely normal rate in 10-pair concurrent discrimination learning (CDL) (Parker and Gaffan 1998b; Gaffan and others 2002). This suggests a qualitative difference between frontal cortical involvement in object-reward association learning and its involvement other tasks.
In the present study, we explored the limits of preserved object-reward association learning in monkeys with frontal–temporal disconnection. We trained 4 monkeys on single-problem object-reward associations until a proficient discrimination learning set (DLS) had been established. We then assessed postoperative learning rates after crossed unilateral lesions of frontal and inferior temporal cortices. DLS formation by monkeys learning single-problem object-reward associations is shown by a gradual and substantial increase in the speed of within-problem learning (Harlow 1949). Such an increase in learning rate is not seen in monkeys learning object-reward associations concurrently (Murray and Gaffan 2006). That the concurrent learning of object-reward associations does not promote learning set formation opens up the possibility that object-reward association learning after DLS formation may be in some way qualitatively different from object-reward association learning in concurrent discrimination tasks.
Materials and Methods
The subjects were 4 female rhesus monkeys (Macaca mulatta) aged between 2 and 2.5 years at the start of training and weighing on average 3.06 kg at the time of their first surgery. Monkeys were housed either singly or in pairs in their home cages with water provided ad libitum. Two monkeys, A and B, were experimentally naive. The other monkeys, C and D, had taken part in an earlier experiment in which they had learned concurrent discriminations between objects presented in complex scenes (the object-in-place task: Gaffan 1994) and had received a unilateral ablation of inferior temporal cortex. The present experiment was the first experience of monkeys C and D in single-problem learning and in discrimination learning with objects that did not occupy a fixed position in a complex scene.
By the final postoperative testing phase, each monkey had been operated on twice, as outlined below, once to remove inferior temporal cortex in one hemisphere and once to remove the cortex of the frontal lobe (sparing primary motor cortex) in the contralateral hemisphere. Thus, after 2 surgeries, monkeys had crossed unilateral lesions of frontal and inferior temporal cortices. Two of the 4 monkeys, A and B, had the frontal ablation in their first surgery, and the remaining 2, C and D, had the inferior temporal cortex ablation in their first surgery.
The apparatus consisted of a computer-controlled touch-sensitive monitor (380 mm wide by 280 mm high) on which the stimulus material was presented. The monkey sat in a wheeled transport cage 150 mm from the screen and made choices between stimuli by reaching out through the bars of the cage. Touches were registered by the computer, and banana flavored pellets (190 mg supplied by Noyes Company Inc., Lancaster, NH) were delivered as rewards for correct responses to a food hopper placed centrally below the monitor. A single large food reward was delivered at the end of each session by the opening of a box set to one side of the centrally placed food hopper. The box contained proprietary monkey food, fruit, peanuts, and seeds, and this large reward provided the whole daily diet of the monkeys on testing days. The opening of the box with the large food reward, like all other aspects of the events and the experimental contingencies during any session of training, was under computer control.
The pool of available stimuli was 600 clipart “objects”. Stimuli were used from this pool without replacement to form the object pairs in each unique discrimination problem. The images were presented on a white background that occupied the whole touch screen. Each image was 128 by 128 pixels in resolution and was displayed on a screen that had a resolution of 800 pixels wide and 600 pixels high. The visible part of the images varied in shape because in many images some parts were the same as the white background color of the screen. An image, including any background parts, occupied a 60-mm square on the screen. All responses to images were followed by visual feedback that consisted of the image remaining on the screen for 1000 ms. In choice trials, the images were presented in pairs on the left and the right side of the screen; in this case the left–right position of the images was determined at random, and the edge of each image was 50 mm from the edge of the screen.
Monkeys A and B, which were experimentally naive (see Subjects), were first trained to touch objects presented on the touch screen to obtain food rewards.
In a daily session, monkeys worked through successive lists of trials, each list using a single discrimination problem. Each list of trials began with the presentation of the current S+ or S−, chosen randomly, in the center of the screen. A touch to this stimulus resulted in the delivery of reward if that stimulus was the S+ or no reward if that stimulus was the S−. Following this forced-choice trial, and after an intertrial interval of 5 s, both stimuli were presented as a discrimination problem for a further 10 trials, separated by 5-s intertrial intervals, after which the stimuli used for that problem were discarded and not used for the remainder of the experiment. Thus, each problem was presented for 11 trials with the first presentation being a forced-choice trial. Following these trials, a new pair of objects was chosen from the pool, and the sample phase began again, as described above, with the new pair of stimuli. The learning of single-discrimination problems continued in this way until the monkey reached the predetermined number of rewards for that session.
The number of rewards per session was 50 for monkeys A and B and 100 for monkeys C and D. (It is standard procedure in our laboratory to allow individual differences between monkeys in the extent of their willingness to work for food rewards. Because the main dependent measure in this experiment is the difference between the final performance tests preoperatively and after each stage of surgery, small differences between individual monkeys in the method of achieving that final performance level, in this instance differences in the number of rewards per day, do not vitiate the main dependent measure.) Monkeys completed sessions of object-reward association learning, as described above, until a clear learning set had been formed. This was assessed intuitively as a stabilization of the decline in errors per session. Following this, monkeys A and B then had their first surgery which was a unilateral right hemisphere frontal ablation. This was followed by a postoperative recovery period of 9–11 days. These monkeys were then tested for 8 sessions, the last 5 of which constituted the control performance test. Subsequently, these monkeys had their final surgery, unilateral left hemisphere inferior temporal cortex ablation, thus completing the disconnection of frontal cortex from inferior temporal cortex. Following a postoperative recovery period of 9 days, monkeys A and B were tested again for 8 sessions, the last 5 of which constituted the final postoperative performance measure.
Having acquired a DLS, monkeys C and D completed 8 further sessions of testing, the last 5 of which constituted their first performance measure. Thus, the control measure for all monkeys was 5 sessions completed with unilateral surgery. These monkeys then had a unilateral left hemisphere frontal lobe ablation thus completing the disconnection. Following the postoperative recovery period of 9 days, each monkey completed 8 sessions, the last 5 of which constituted the final postoperative performance measure.
Operations were performed under sterile conditions. The monkeys were anesthetized throughout surgery with barbiturate (thiopentone sodium, to effect) administered through an intravenous cannula. Monkeys were maintained in a state of deep anesthesia by monitoring by pulse rate, blood oxygenation, body temperature, and peripheral reflexes, consistent with United Kingdom Home Office regulations. The ablations were made by aspiration with the aid of an operating microscope. When the lesion was complete, overlying tissue was closed in layers. Monkeys received buprenorphine hydrochloride (0.01 mg/kg) as an analgesic and amoxicillin (8.75 mg/kg) to prevent infection, daily, for 1 week postoperatively.
Unilateral Frontal Lobe Ablation
The extent of the intended lesion is shown in Figure 1 and included the entire frontal lobe while sparing primary motor cortex. The cortex was removed in 3 sections, and in each case the cortical tissue and pia mater bounded by a continuous line of cautery were then removed by aspiration. On the lateral surface, the continuous line of cautery extended through the center of the precentral dimple from the most dorsal point on the lateral surface of the frontal lobe to the ventral surface, following a line approximately parallel to the central sulcus and 1–2 mm posterior to the descending limb of the arcuate sulcus. The line then extended around the boundary of the visible lateral surface of the frontal lobe, along the margin of the orbital surface and the lip of the interhemispheric fissure, until it rejoined the point at which it started. The second line of cautery began from the most ventral point of the previous line, and extended along the lip of the lateral sulcus on the orbital surface of the frontal lobe to the base of the olfactory tract and beyond it, to the boundary of the medial surface. It then extended along the orbital surface anteriorly to the frontal pole, joining the line of cautery from the first area of cortex that was removed. The final line of cautery extended from the origin of the first line of cautery ventrally to the corpus callosum, then followed the line of the corpus callosum anteriorly and ventrally, and then followed the boundary of the cortical surface until meeting the line of cautery made on the orbital surface. The cortical gray matter was removed from the surface of the brain and from sulci within the boundary shown. The white matter surrounding the corpus striatum and the striatum itself was left intact.
Unilateral Inferior Temporal Cortex Ablation
The arch of the zygoma was removed, and the temporal muscle was detached from the cranium and retracted. A bone flap was raised and extended with a rongeur, and then the dura mater was cut. The extent of the intended lesion is shown in Figure 1. The ablation extended from the fundus of the superior temporal sulcus to the fundus of the rhinal sulcus and posteriorly included both banks of the anterior part of the occipitotemporal sulcus. The posterior limit of the ablation was a line drawn perpendicular to the superior temporal sulcus, 5 mm anterior to the inferior occipital sulcus. The anterior limit of the ablation was the anterior tip of the superior temporal sulcus and a line drawn round the pole from that tip to the rhinal sulcus. Within these limits, all the cortices were removed including both banks of the anterior and posterior middle temporal sulci.
At the completion of all behavioral testing, subjects were deeply anesthetized and transcardially perfused with physiological saline followed by 10% formalin. The brains were extracted and cut on a freezing microtome in 50-μm sections in the coronal plane. Every fifth section was stained with cresyl violet, mounted on a slide, and coverslipped.
Figure 2 shows 4 actual sections and their corresponding reconstructions from subject A. Reconstructions from subjects B, C, and D at the same anterior–posterior level are also shown. The estimated extent of the lesion in each subject was plotted onto standard scanned sections and highlighted in red.
The reconstructions show that the frontal ablations were essentially intended although there was a tendency for the frontal lesions to be conservative. Tissue within posterior cingulate cortex was spared in monkey A and the cortex in and medial to the medial orbital sulcus was spared in monkey C. The frontal ablation in monkey C also spared cortex at the posterior extent of the intended lesion in and around the arcuate sulcus. The inferotemporal cortex ablation in all monkeys was as intended stretching from the rhinal sulcus medially to fundus of the superior temporal sulcus laterally.
DLS formation was achieved in 36, 44, 22, and 15 sessions of training by monkeys A–D, respectively, monkeys C and D being trained in sessions that were twice as long as the sessions for monkeys A and B (see Materials and Methods). Figure 3 shows the averaged learning curves for all monkeys over trials 2–11 of single-new problems at 3 stages of DLS formation. Stages 1, 2, and 3 represent performance in the first, middle, and last 10 problems of DLS training. The figure shows the substantial increase in the speed of within-problem learning rates over stages 1–3. Within-problem learning rate can also be simply summarized by the overall error rate in trials 2–11. The improvement in this measure over stages 1–3 is illustrated in Figure 4. A repeated-measures analysis of variance (ANOVA) on error scores from stages 1–3 revealed a significant effect of stage (F2,6 = 26.05, P = 0.002, Huynh–Feldt). Pairwise comparisons using the pooled error term showed a significant improvement in learning rates between stages 1 and 2 (t = 4.767, df = 6, P = 0.002) and between stages 2 and 3 (t = 2.309, df = 6, P = 0.030).
Monkeys C and D in the present experiment previously had unilateral inferotemporal ablation as part of a separate experiment. Figure 4 shows that this unilateral surgery had no effect on the ability of monkeys C and D to develop a learning set over stages 1–3 of training. Monkeys A and B were trained as intact monkeys. The mean percent error for monkeys A and B at stage 3 of preoperative learning set formation was 14.0 and 17.0 and after unilateral surgery 20.1 and 11.4 percent error, respectively. These 2 observations indicate that for all monkeys in the present experiment, unilateral surgery was without effect on the task.
Following frontal–temporal disconnection, monkeys made a mean of 3 times more errors, and a within-subject comparison of learning rates using a 1-way ANOVA showed that following crossed unilateral lesions of frontal cortex and inferior temporal cortex, monkeys made significantly more errors on the task (F1,3 = 36.46, P = 0.009). The impairment was stable over the 5 days of the final postoperative test, as shown by the absence of any linear effect of days upon performance (F1,3 < 1).
Figure 4 shows that the mean percent error score by each individual monkey postoperatively is similar to that individual's mean percent error score shown in stage 1 of DLS formation. This striking effect is shown over the wide range of error scores from individual monkeys. A repeated-measures ANOVA showed that the percent error scores by frontal–temporal disconnected monkeys were not different from the percent error scores by intact monkeys on stage 1, before the formation of DLS (F1,3=2.30, P = 0.226).
Figure 5 compares the learning curves of control and frontal–temporal disconnected monkeys learning object-reward associations in the present study and in 10-pair CDL (Parker and Gaffan 1998b; Gaffan and others 2002, 2001). Both intact and lesioned monkeys are known to score more errors on trials in which the first response to a novel problem is wrong, and because of this the CDL scores were adjusted to give equal weight to problems where trial 1 was correct and where trial 1 was wrong. DLS scores on trial 1 were at chance level because trial 1 was a randomized forced-choice trial (see Materials and Methods). The graph contrasts the learning rates on DLS and CDL shown by intact monkeys prior to complete frontal–temporal disconnection. The graph also compares learning rates in the 2 tasks following frontal–temporal disconnection.
Table 1 categorizes the error scores in the final 2 stages of the experiment: 2C columns represent the percent error in trials 3–11 on problems in which the response on the second trial of those problems was correct (trial 1 was a forced-choice trial, see Materials and Methods, Task Procedure), and 2W columns represent the percent error on those trials in which the response on the second trial of those problems was wrong. Such a breakdown has previously been employed to look for evidence of perseverative responding (Brush and others 1961; Browning and others 2005). An ANOVA on these data using stage and first response as factors revealed a significant effect of stage (F1,3 = 48.96, P = 0.006). There was no stage by first response interaction (F1,3 < 1), however, indicating that surgery did not differentially affect the 2 categories of error scores.
|Unilateral||FL × IT|
|Unilateral||FL × IT|
Note: Percent error scores in trials 3–11 as a function of the outcome of trial 2 in DLS (trial 1 was a forced-choice trial). Unilateral, monkeys with unilateral frontal or inferior temporal cortex ablation; FL × IT, monkeys with crossed unilateral ablation of frontal and inferior temporal cortices; 2C identifies problems where the monkey's choice on the first discrimination trial was correct; 2W identifies problems where the monkey's choice on the first discrimination trial was wrong.
In the present experiment, monkeys were trained on successive single-pair object-reward associations until they had developed a clear learning set. After final surgery to complete frontal–temporal disconnection, all 4 monkeys were severely impaired at learning single object–reward associations, and the group scored approximately 3 times as many errors overall. As shown in Figure 5, the impairment in DLS is a task-specific impairment because frontal–temporal disconnection, surgically produced in the same manner as in the present experiment, left monkeys clearly unimpaired in 10-pair concurrent object-reward association learning, tested in the same apparatus, and using the same general methods as in the present experiment (Parker and Gaffan 1998b; Gaffan and others 2002). Not only was the group impaired as a whole but also each individual monkey's final postoperative learning rate was similar to the rate which that individual monkey had shown in stage 1 of DLS training before learning set was formed (Fig. 4). Frontal–temporal disconnection thus selectively abolished each monkey's learning set.
Table 1 shows that this impairment cannot easily be attributed to perseveration. A deficit of inhibitory processing would be expected to produce a lesser impairment or even a facilitation in learning those problems on which the monkey's first choice was correct (2C) compared with those in which the first choice was wrong (2W). A similar analysis, with similar results, has been carried out on concurrent object-in-place learning following frontal–temporal disconnection (Browning and others 2005). The present results also suggest that difficulty per se is not an important factor in determining whether frontal cortical processing of object representations is necessary for normal learning. Figure 5 shows that the learning rates of control monkeys in DLS are faster than those in CDL, and yet DLS severely impaired by disconnection of frontal cortex from inferior temporal cortex.
The present impairment indicates that, as hypothesized in the Introduction, the learning of object-reward associations following the formation of DLS is qualitatively different from the learning of object-reward associations in concurrent tasks. More specifically, the comparison between stage 1 DLS learning rates and final postoperative learning rates, shown in Figure 4, shows that this difference is wholly dependent upon the interaction of frontal cortex and inferior temporal cortex.
Both DLS and CDL tasks require monkeys to associate in memory visual objects with food reward. In CDL, monkeys may simply adopt an innate or default strategy of approaching items that are associated in memory with reward. In contrast, the gradual and substantial increase in learning rates associated with the formation of a learning set is indicative of an acquired learning strategy that develops over many problems. The learning curves of frontal–temporal disconnected monkeys on DLS and CDL, shown in Figure 5, suggest that such an additional capacity is absent in operated monkeys and that these monkeys may revert to the default strategy described above, of simply approaching items that are associated in memory with reward. One description of the behavioral strategy acquired during the formation of learning set could be in terms of a “win-stay, lose-shift” rule. This description is problematic, however, because it cannot distinguish between associative learning in CDL and in DLS. The present results show that this distinction is necessary for an explanation of the function of frontal–temporal interactions.
Murray and Gaffan (2006) have proposed that in learning set formation monkeys learn to lay down a prospective memory, on each trial of within-problem learning, of the correct choice to be made at the next trial with that problem. In support of this proposal, they showed that monkeys fail to form a learning set when many new problems are presented for learning in the same fashion as in the conventional learning-set procedure, but in concurrently presented sets of problems rather than in single problems. In concurrent learning, they reasoned a monkey cannot easily form a representation of the temporally extended event that is comprised by, for example, choosing object A on trial “i” and receiving reward and then choosing object A again on trial “i + 1” and receiving a reward because in concurrent learning there is a long time interval (on average 10 trials) between the successive trials with any individual problem. Prospective memory, by its nature, relies on the representation of this type of trial-by-trial temporally extended event. Prospective memory is a learned anticipation, on trial “i” with an individual problem, of what the correct choice will be on trial “i + 1” with that problem (Murray and Gaffan 2006). A monkey will only learn to anticipate the next problem if he is rewarded for doing so. What is rewarded during learning set formation therefore is an extended representation of at least 2 response outcome events; trial “i” and trial “i + 1”. Extending the interval between successive trials with the same problem, as in CDL, provides the monkey with little incentive to use this extended representation to learn to anticipate successive trials with the same problem because the reward for doing so is many trials in the future. We propose that it is this representation that is abolished by frontal–temporal disconnection.
The analysis of learning set formation by Murray and Gaffan (2006), in conjunction with the present results and the previously reported results from concurrent learning (Parker and Gaffan 1998b; Gaffan and others 2002), suggests that frontal–temporal disconnection specifically impairs the process of representing temporally extended events, which is important in learning set formation and not in CDL. The loss of this capacity may also be responsible for all the deficits reported in monkeys with frontal–temporal disconnection, including delayed matching-to-sample and conditional learning (Eacott and Gaffan 1992; Parker and Gaffan 1998a, 1998b; Bussey and others 2002; Gaffan and others 2002; Browning and others 2005). In summary, several recent results, including those presently reported, suggest a role for frontal–temporal interaction in the integration of visual information across time. Future experiments will test this hypothesis further.
This research was supported by the UK Medical Research Council. Conflict of Interest: None declared.