A previous study using a rodent five-choice test of attention found poor choice accuracy and increased perseverative responding following medial prefrontal cortex (mPFC) lesions. As this rat cortical area includes at least two anatomically distinguishable subregions, the present study investigated their specific contributions to performance of this task. Rats were trained on the five-choice task prior to receiving excitotoxic lesions or sham surgery. In the first experiment, lesions of the dorsal mPFC (Zilles’s Cg1) resulted in poor accuracy, but no changes in perseverative responding. Introducing variable delays for stimulus presentation abolished these accuracy deficits, suggesting that Cg1-lesioned rats were impaired at using temporal cues to guide performance. In the second experiment, lesions of the ventral mPFC increased perseverative responding, but had only short-lasting effects on accuracy. Rats with complete mPFC lesions had both choice accuracy impairments and increased perseverative responding. Additional evidence of the functional dissociation of dorsal and ventral mPFC came from the analysis of the spatial and temporal distribution of the correct and incorrect responses. Only rats with ventral mPFC lesions showed delay-dependent deficits and bias towards a location that had recently been associated with reward. Taken together, these results suggest dissociable ‘executive’ functions of mPFC subregions. Circuits centred on Cg1 are critical for the temporal organization of behaviour, while networks involving the ventral mPFC are important for maintaining behavioural flexibility.
The cortical projection area of the mediodorsal thalamic nucleus (MD), in the rat, comprises several cytoarchitectonically distinct areas of the medial wall and the shoulder of the frontal lobes, including the pregenual part of Zilles’s Cg1, the prelimbic cortex (PRL, corresponding to Zilles’s Cg3) and a strip of Fr2. Ventral to PRL, the infralimbic area (IL) is also connected with MD, but not reciprocally (Groenewegen, 1988). Impairments on delayed response tasks following lesions of these cortical areas have been described for over a quarter of a century, in either mazes (Kolb et al., 1974; Granon et al., 1994; Kesner et al., 1996; Floresco et al., 1997; Porter et al., 2000), or in operant chambers (vanHaaren et al., 1985; Dunnett, 1990; Aggleton et al., 1995). Based on this finding, as well as on anatomical criteria (Preuss, 1995), it has been suggested that the medial prefrontal cortex (mPFC) of the rat is homologous to the dorsolateral PFC of humans and non-human primates (Kolb et al., 1974; Larsen and Divac, 1978).
Rats with mPFC lesions are also impaired on tasks that assess attentional and executive functions (Olton et al., 1988; Muir et al., 1996; Miner et al., 1997; Granon et al., 1998; Birrell and Brown, 2000). For example, a previous study from our laboratory found that excitotoxic lesions of the mPFC impaired the accuracy of detecting brief flashes of light presented unpredictably in one of five spatial locations (Muir et al., 1996). Specifically, mPFC (but not parietal) lesions reduced choice accuracy, lengthened the correct latencies and increased the number of perseverative responses. In man, the anterior cingulate cortex has been implicated in forms of attentional and executive functions (Pardo et al., 1990; Corbetta et al., 1991). Therefore it has also been argued, based on this and on anatomical evidence (Preuss, 1995), that the rat mPFC may be considered homologous to the anterior cingulate cortex of humans and non-human primates.
One possibility is that multiple functions of the rat mPFC, as evidenced by behavioural studies, are sustained by distinct sub-regions within this cortical area. Indeed, several authors have presented evidence of functional (Morgan and LeDoux, 1995; Seamans et al., 1995; Kesner et al., 1996; Fritts et al., 1998; Ragozzino et al., 1998) as well as anatomical (Sesack et al., 1989; vanEden et al., 1992) dissociation of dorsal and ventral regions of the rat mPFC. However, other authors have found no behavioural evidence of such dissociation (deBruin et al., 1994; Joel et al., 1997) and recent anatomical studies have demonstrated an extensive network of interconnection between these areas (Fisk and Wyss, 1999), suggesting that in the intact brain they may be working in close association. An alternative possibility is that deficits on tasks measuring attentional and executive functions may be related in some way to the impairments observed on delayed response tasks, following mPFC lesions. In other words, either a disruption of a common computational process normally sustained by this brain region may be responsible for the expression of both types of deficit or, alternatively, different deficits may be causally related.
Despite the wide range of deficits described in humans following PFC damage (Fuster, 1997), converging views of dorsolateral PFC function hold that all the cognitive and behavioural effects of its lesion must be related to a critical role of this brain region in maintaining memory representations in a highly active state, thereby allowing their guidance of behaviour under conditions of interference (Shallice, 1982; Goldman-Rakic, 1987, 1998; Duncan, 1995; Baddeley, 1996; Fuster, 1997; Robbins, 1998). According to this view, a disruption of neural circuits centred on the dorsolateral PFC would result in impairments in the retrieval of appropriate behaviour-guiding information (leading to deficits in delayed tasks) especially under conditions of interference (attentional deficits). Human and primate prefrontal studies also suggest that the appropriate selection of behaviour-guiding information requires the interaction of two functionally related circuits centred on two distinct dorsolateral PFC areas: the mid-dorsolateral and mid-dorsoventral PFC (Owen et al., 2000; Petrides, 2000).
Recent studies have suggested that areas of the rat mPFC may jointly play a critical role in the selection and activation of memory representations useful to guide the exploration of a radial arm maze (Seamans et al., 1995). Specifically, it was proposed that the ventral mPFC (PRL) might be particularly important to guarantee flexibility in the use of recently acquired information, while the dorsal mPFC (Cg1) may be critical in the context of well-learned task strategies during performance of non-automatic tasks. Several other studies employing maze tasks have suggested that lesions of the ventral mPFC lead to inflexible behaviour, guided by reward-based associations rather than by relevant mnemonic information, but they have provided only limited information regarding the specific functions of the dorsal mPFC (Ragozzino et al., 1998, 1999a,b). In the five-choice task of attention, while lesions of the whole mPFC resulted in accuracy and perseverative deficits (Muir et al., 1996), lesions sparing the dorsal mPFC mainly affected the number of per-severative responses, having only minor effects on other aspects of performance (Chudasama and Muir, 2001).
The aim of the present study was twofold. First, we aimed to provide further support for the hypothesis that distinct regions within the rat mPFC play distinct roles in processes allowing behavioural guidance by memory representations and, in particular, we aimed to investigate the role of the dorsal mPFC, which has so far received relatively little attention. Secondly, by using a behavioural paradigm designed to assess attention, but also requiring the animals sometimes to respond in the absence of guiding stimuli, we aimed to provide a common framework for the mnemonic (Seamans et al., 1995) and the attentional functions (Muir et al., 1996) of the rat mPFC. To this end, we tested rats with bilateral lesions to either the dorsal (experiment 1) or ventral mPFC (experiment 2) on the same five-choice task previously used by Muir and colleagues (Muir, 1996; Muir et al., 1996). By analysing the temporal and spatial distributions of responses characteristic of each lesion group, we provide evidence linking the mnemonic and the attentional/executive functions of these distinct mPFC sub-regions. A group of animals with complete lesions of the medial wall of the PFC was included in experiment 2, for comparison with previous experiments.
Materials and Methods
Male Lister hooded rats (Charles Rivers, Kent, UK) were housed in pairs in a temperature-controlled room (22°C), under a 12 h light/dark cycle (lights on at 8 p.m.). They were kept with food-restriction and water ad libitum. Behavioural testing was carried out every day between 10 a.m. and 1 p.m. (experiment 1) or between 1 and 5 p.m. (experiment 2). All experiments were completed under the terms of the UK Animals (Scientific Procedures) 1986 Act.
After training was completed and following acquisition of the behavioural baseline, animals were anaesthetized with an i.p. injection of Avertin (1 ml/100 g body wt) and placed in a Kopf stereotaxic frame fitted with atraumatic ear bars. In experiment 1, eight rats received bilateral injections of either 0.3 or 0.2 μl of 0.09 M quinolinic acid (Sigma, St Louis MO) in 0.1 M phosphate buffer, while five received infusions of the same volumes of 0.1 M phosphate buffer. Stereotaxic coordinates of the injection sites (millimetres from bregma or dura) and volumes infused were: AP +2.2, L ±0.7, DV –1.9 (0.3 μl); AP +2.7, L ±0.7, DV –1.9 (0.3 μl); AP +3.2, L ±0.7, DV –1.9 (0.2 μl). In experiment 2 there were three groups. Nine rats (denominated ‘mPFC’ henceforth) received bilateral injections of 0.09 M quinolinic acid at the following coordinates (millimetres from bregma or from dura): AP +2.6, L ±0.7, DV –1.5 (0.7 μl); AP +3.2, L ±0.7, DV –3.0 (0.5 μl) and –1.5 (0.5 μl); AP +3.8, L ±0.7, DV –1.5 (1 μl). Nine rats (denominated ‘PRL–IL’ henceforth) received bilateral infusions of 0.09 M quinolinic acid at AP +2.7, L ±0.7, DV –4.0 (0.5 μl) and at AP +4.0, L ±0.7, DV –3.5 (0.4 μl). As a control group, nine rats (‘CTR’) receiving sham surgery were injected with phosphate buffer at either the coordinates used for the mPFC rats (n = 5) or those used for the PRL–IL rats (n = 4). The incisor bar was kept at –3.3 mm relative to the interaural line, except for the PRL–IL surgery (sham or lesion), for which it was kept at +5.0 mm. Injections were made using two precision syringes, mounted in a Harvard infusion pump, connected by PP10 tubing to a 30 gauge stainless steel double cannula. At each of the injection sites, volumes were infused over a period of 2–3 min and then the cannula was left in place for an additional 2 min. Rats were singly caged following surgery and for the rest of the experiment.
Following completion of behavioural testing, animals were given an over-dose of barbiturate and then perfused with 0.01 M phosphate-buffered saline (PBS) followed by 4% PFA. The brains were then stored in 20% sucrose for dehydration before sections were cut at 60 μm thickness on a freezing microtome. Every third section was mounted on a glass slide for staining with cresyl violet.
The test apparatus for all the experiments were nine-hole boxes similar to those described previously (Carli et al., 1983). Each of these consisted of a 25 × 25 cm aluminium operant chamber equipped on the rear, concavely curved wall with five apertures, each 2.5 cm2, 4 cm deep and set 2 cm above floor level. Each hole could be illuminated by a standard 3 W bulb located at the rear of the hole. An infrared photocell beam placed at the entrance of each hole could be broken by the animal briefly inserting its nose in the aperture. Depending on the requirements of the different tasks, holes could be blocked by means of metal caps. A 3 W house light mounted on the roof provided illumination of each chamber. On the front wall of the chamber a magazine connected to a food dispenser allowed the automatic delivery of 45 mg food pellets (Noyes, NJ, USA) on successful trials. Animals obtained access to the food magazine by pushing a hinged Perspex panel monitored by a microswitch. Each chamber was housed in a wooden sound-attenuating cabinet where a fan provided ventilation, as well as low-level background noise. The apparatus and on-line data collection was controlled by means of an Acorn Computer system with software written in Arachnid (CeNeS plc, Cambridge, UK).
Training in the five-choice task has been described in earlier work (Carli et al., 1983). Briefly, training began with two 15 min sessions in which the house light and the tray light were on and 30 pellets were placed in the tray. In the first of these initial sessions, all the apertures were covered with metal caps and the magazine panel was partially open; in the second, the metal caps were removed from holes 1, 3, 5, 7 and 9 and the test schedule was implemented. A diagram of the task is represented in Figure 1. Each test session began with the illumination of the chamber by the house light and the delivery of a food pellet. The collection of this pellet by pushing the magazine panel started the first trial. After a fixed inter-trial interval of 5 s, the light at the rear of one of the response apertures was briefly illuminated. Responses in this aperture within a limited time of illumination of the hole (limited hold period) were recorded as correct responses and were rewarded by the delivery of a food pellet to the magazine. Responses in a non-illuminated hole were recorded as incorrect responses and were punished by a time-out period, during which the illumination of the chamber was extinguished. Additional responses in the apertures prior to food collection (perseverative responses) were punished with a time-out, as were failures to respond within the 5 s limited hold period (omissions) and responses in one of the apertures during the inter-trial interval (premature responses). Every new trial — after a correct response or after a time-out — was initiated by a response in the magazine panel. The tray light was illuminated following each correct response during early phases of training, when animals acquire the basic contingencies of the task; on later phases of training and during testing the use of the tray light was discontinued, in order to facilitate initiation of trials that followed a time-out period. The duration of the stimulus and the limited hold period were as long as 60 s in the first training sessions. Then, depending on the rats’ individual performances, these durations were progressively reduced to the final durations used during pre-surgical baseline and during testing. In experiment 1, the final stimulus duration was 0.35 s. This duration, which is shorter than that used in previous studies using the five-choice task (Muir et al., 1996) was selected in order to reduce the risk of ceiling effects and also to increase the number of trials in which responses were made in the absence of the visual target (see below). In experiment 2, the final stimulus duration was 0.40 s. This stimulus duration was selected because prior to surgery some of the rats in this experiment could not sustain stable performance with the 0.35 s length. In order to make the results comparable with those of experiment 1, it was chosen to keep the level of performance rather than the stimulus duration constant across experiments. The final limited hold period was 5 s in both experiments. Each daily session lasted until completion of 100 trials or for 30 min. Training in this procedure required between 4 and 5 weeks. When the animals had reached stable performance (accuracy >80% and omissions <15% over 3–5 sessions), they were assigned to either of two performance-matched groups and taken for surgery.
Eight–ten days following surgery, rats were tested across 13 sessions on the standard schedule of the task. Subsequently, a series of manipulations to the basic test schedule was instituted. First, the stimulus duration was reduced to half of the final training duration during one test session. Secondly, in order to assess the effect of stimulus unpredictability, animals were exposed to variations in the inter-trial interval (ITI) duration: 1 day of a short ITI schedule (0.5, 1.5, 3.0 and 4.5 s), followed by 1 day of a long ITI schedule (4.5, 6.0, 7.5 and 9.0 s). These manipulations were followed by a number of other behavioural manipulations and pharmacological challenges, including, for example, pre-feeding, reducing the brightness of the stimuli, interpolating distracting bursts of white noise, rendering the stimulus presentations spatially predictable — one-choice task (Dalley et al., 2002) — and a ‘rat-paced’ version of the five-choice task. In the interest of the economy of presentation, the effects of these manipulations are not reported or discussed here. The effects of pharmacological challenges are published elsewhere (Passetti et al., 2002). Each of the manipulations was preceded by one or two baseline sessions.
The behavioural measures recorded in the five-choice task have been described elsewhere (Carli et al., 1983) and include: correct and incorrect responses; percentage correct (correct responses divided by the sum of correct plus incorrect responses); omissions; total number of completed trials (correct + incorrect + omission trials); premature responses; perseverative responses; perseverative panel pushes; and three measures of speed — correct latency, incorrect latency and magazine latency. Earlier studies (Robbins et al., 1993; Muir, 1996) give descriptions of the behavioural interpretation of changes in these measures of performance.
In addition to the usual measures of performance, trial-by-trial data were also collected. These data were used to investigate how correct and incorrect nose pokes were distributed in the limited hold period (i.e. the time allowed for responding from the stimulus onset) and across the five spatial locations. In the five-choice task, rats are required to schedule their behaviour accurately for successful performance, given that after collection of a food reward they have a limited time to turn around and begin scanning the apertures in search of the visual stimuli (Passetti et al., 2000). In addition, on some trials the animals have to respond in the absence of guiding behavioural stimuli. Thus, on these trials the animals have to hold ‘on-line’ mental representation of prospective plans/ locations for responding, while inhibiting responding to irrelevant, distracting stimuli. The following measures were introduced to enable the analysis of behaviour scheduling, the accuracy decrement profile following stimulus offset (i.e. when rats are required to respond in the absence of guiding external stimuli) and bias toward a specific location or toward a location previously associated with reward.
Analysis of the Distribution of the Responses in the Limited Hold Period
The total numbers of (correct and incorrect) nose poke responses made by each rat in the final six baseline sessions were divided according to the time in the limited hold period at which they were made and expressed as a percentage of the total. Each response was allocated to one of five time bins (0–35, 36–80, 81–130, 131–250, 251–500 cs, in experiment 1 and 0–40, 41–80, 81–130, 131–250, 251–500 cs, in experiment 2). If any of the lesions affected behaviour scheduling, then it would be expected to find differences in the way responses were distributed in the limited hold period in lesioned or control rats. Measures of choice accuracy for the same time bins in the limited hold period were also calculated. If any of the lesions affected the ability to hold ‘on-line’ information relative to the response to be performed, then accuracy deficits should be particularly evident after the stimulus offset (i.e. in the second or in the third bin), but not while the stimulus is still on (first bin). If, however, the lesions resulted in impairments of the ability to discriminate the stimuli, then accuracy deficits should be evident also while the stimulus is still on. Thus, this analysis enables us to divide accuracy deficits into delay-dependent (i.e. mainly apparent when the rat responds after stimulus offset) and delay-independent (i.e. apparent regardless of stimulus offset).
Two measures of spatial bias were calculated. The first of these measures gave the value of bias towards responding in one of the spatial locations and was obtained by calculating what proportion of correct and incorrect nose poke responses were made by each rat in each of the five holes in the final six sessions. The second measure estimated the tendency to respond in a hole, the choice of which had been previously reinforced (win–stay strategy) and was calculated by counting the incorrect responses made in the location reinforced in the last preceding correct trial and expressed as a percentage of the total. On any given trial, the probability of making an incorrect response if responding randomly is 0.8. Thus, in each group the proportion of incorrect responses made in a previously reinforced hole was tested against a calculated probability of 0.2 (which is equivalent to 0.8 × 0.25). The incorrect responses made in a hole adjacent to the hole in which the stimulus was presented on that trial were also counted (calculated probability = 0.4). This kind of error reflects the degree to which the rat’s behaviour is under the control of the task contingencies, rather than being guided by bias or by win–stay strategy.
Data for each variable were subjected to analysis of variance (ANOVA) using the SPSS statistical package. Prior to ANOVA, data were plotted and exploratory analyses using ‘box plots’, tests of skewness and tests of homogeneity of variance were performed. Based on these preliminary analyses, outliers were identified and skewed data, which violate the assumption of normality required by analysis of variance (ANOVA), were subjected to arcsine, square root, or logarithmic transformation as recommended by Howell (Howell, 1997). ANOVA included the between-subjects factor ‘lesion’ and the within-subjects factor ‘day’ (or stimulus duration, ITI, bin, etc.). When significant interactions between factors in the analyses were found, post hoc analysis was carried out to clarify the source of the interaction, using the Student–Newman–Keuls test (S–N–K). The Greenhouse–Geisser epsilon correction for degrees of freedom was used whenever appropriate (significant Mauchly’s ‘sphericity’ test). Response bias toward a specific location was analysed for each individual rat using the χ2 test. One-sample t-tests against the calculated random probability were used to test the probability of responding at a previously reinforced location or in next-to-correct locations.
In experiment 1, the histological examination revealed that one of the dorsal mPFC lesions was unilateral and in another case the lesion was incomplete. These animals were thus discarded. In all other cases, the area of destruction was centred on a region including Zilles’s Cg1 between the genu and ~+4.0 from bregma. In experiment 2, two of the mPFC lesions were incomplete and two of the ventral mPFC lesions were revealed to have encroached Zilles’s Cg1. Data from these animals were discarded, leaving n = 7 for the group with mPFC lesions (‘mPFC’ group) and n = 7 for the ventral mPFC (‘PRL–IL’ group). In all other cases, the area of destruction was centred on PRL, IL and Zilles’s Cg1 between the genu and ~+4.0 from bregma in the mPFC group and on PRL and IL, sparing Cg1, in the PRL–IL group. In mPFC lesions, damage to Fr2 was infrequent and in only two of the PRL–IL lesions was there minor damage posterior to the genu. In order to preserve matching of pre-operative performance, data from one of the sham-operated, control animals of experiment 2 were also discarded from the following analyses (final n = 8, ‘CTR’ group). Figure 2 shows the extent of the lesions. A photomicrograph of typical lesions is shown in Figure 3.
Effects of Dorsal, Ventral or Complete mPFC Lesions on Performance of the Five-choice Task
Eight to ten days after surgery, animals were reintroduced to the baseline schedule and tested for 13 days (data from the sessions of days 4–13 were used for the following analyses). Figures 4 and 5 show a summary of the main results. Preliminary analyses comparing the sham-operated rats of the two experiments revealed significant differences involving the factor ‘day’. These were on the number of perseverative responses and perseverative panel pushes. Thus, data from the two experiments were analysed separately, but presented together in order to facilitate comparison of the effects of dorsal and ventral lesions.
In experiment 1, lesion of Cg1 pre-genu impaired the accuracy of performance, as revealed by a reduction in the percentage of correct responses (Fig. 4A). ANOVA found a significant main effect of ‘lesion’ [F(1,9) = 9.77, P < 0.02], no main effect of ‘day’ [F(9,81) = 1.26, n.s.] and no significant ‘lesion’ × ‘day’ interaction [F(9,81) = 1.26, n.s.]. These accuracy deficits of Cg1-lesioned rats were due to a reduction in the number of correct responses [F(1,9) = 7.42, P < 0.05], rather than an increase in the number of incorrect responses [F(1,9) < 1, n.s.]. In experiment 2, both lesion groups were impaired at reporting the presence of the visual stimuli (Fig. 5A). There was a significant effect of ‘lesion’ on accuracy [F(2,19) = 21.18, P < 0.01] and post hoc comparisons showed this to be due to PRL–IL rats being less accurate than sham-operated animals (P < 0.05) and to mPFC animals being less accurate than both PRL–IL animals (P < 0.05) and CTR animals (P < 0.01). There was also a significant effect of ‘day’ [F(9,171) = 2.38, P < 0.02], but no ‘lesion’ × ‘day’ interaction [F(18, 171) = 1.16, n.s.], suggesting similar improvements in all groups across testing days. In order to establish whether these accuracy deficits were due to a reduction in the number of correct responses or to increased incorrect responses, these measures were also subjected to ANOVA. This revealed that animals with mPFC lesions made significantly fewer correct responses than the animals in the other groups [F(2,19) = 29.79, P < 0.01; S–N–K, P < 0.05], while PRL–IL animals made significantly more incorrect responses than CTR animals [F(2,19) = 8.47, P < 0.01; S–N–K, P < 0.05].
Speed of Responding and Omissions
In experiment 1, rats with Cg1 lesions took longer than controls to respond correctly to the targets [F(1,9) = 4.92, P = 0.054; Fig. 4B]. A main effect of ‘lesion’ was also found on the latency to collect earned food rewards [F(1,9) = 11.25, P < 0.01] and on the number of omissions [F(1,9) = 9.59, P < 0.02; Fig. 4C]. These deficits of Cg1-lesioned rats were present during the baseline post-operative testing, but disappeared during the manipulations. There was no effect of Cg1 lesion on the latency to make an incorrect response [F(1,9) = 1.55, n.s.]. In experiment 2, rats with mPFC lesions made more omissions [F(2,19) = 31.90, P < 0.01; Fig. 5C] and were slower to respond correctly [F(2,19) = 18.87, P < 0.01; Fig. 5B] or to collect earned food rewards [F(2,18) = 18.04, P < 0.01; S–N–K, P < 0.05], compared with PRL–IL or CTR rats. mPFC-lesioned rats were also slower to make an incorrect response [F(2,19) = 11.01, P < 0.01]. No significant difference was found between CTR and PRL–IL rats on any of these measures of performance (S–N–K, n.s.). The latency to collect a pellet could not be calculated for one of the mPFC-lesioned rats due to a technical failure.
Premature and Perseverative Responses
In experiment 1, lesioned and control subjects did not differ in the number of premature responses [F(1,9) < 1, n.s.], perseverative responses [F(1,9) < 1, n.s.; Fig. 4D] or perseverative panel presses [F(1,9) < 1, n.s.]. In contrast, in experiment 2 there was a significant effects of the lesions on perseverative responding [F(2,19) = 13.08, P < 0.01; Fig. 5D]. Post hoc analysis revealed that both mPFC and PRL–IL animals made significantly more perseverative responses than sham-operated animals (P < 0.05), but were not significantly different from each other (S–N–K, n.s.). Rats with complete mPFC lesions also made more premature responses than CTR or PRL–IL animals [F(2,19) = 12.93, P < 0.01]. Neither of the lesions of experiment 2 affected the number of perseverative panel presses [F(2,19) = 2.76, n.s.]).
Temporal Distribution of Correct and Incorrect Responses
As shown in Figures 6 and 7 (left panels) control animals (black columns) made the majority of nose poke responses early in the limited hold period or immediately after the stimulus offset (0–80 cs). About 80% of the responses were made in the first 80 cs of the limited hold period, with the remaining 20% more or less equally distributed between the other bins. In experiment 1, rats with Cg1 lesions made only ~67% of the responses in the first 80 cs and ~13% in the last 250 cs of the limited hold period, i.e. about twice as many as the control rats in this time bin (5.6%). ANOVA of the proportion of responses found a significant effect of ‘bin’ [F(2,21) = 39.35, P < 0.01] and a significant interaction between ‘lesion’ and ‘bin’ [F(2,21) = 3.58, P < 0.05]. Post hoc analysis revealed that rats with Cg1 lesions made relatively fewer responses in the first time bin and relatively more responses in the last time bin, compared to sham-operated animals (P < 0.05; Fig. 6, left). In experiment 2 (Fig. 7, left), PRL–IL rats made only ~22% of the responses in the first 40 cs and >30% between 80 cs and the end of the limited hold period. Rats with mPFC lesions showed an almost flat distribution of responses. ANOVA of the proportion of responses found a significant effect of ‘bin’ [F(2,38) = 42.97, P < 0.01] and a significant ‘lesion’ × ‘bin’ interaction [F(4,38) = 12.14, P < 0.01]. Post hoc analysis revealed that control animals made relatively more responses than PRL–IL animals in the first bin and relatively fewer in the last bin, i.e. between 250–500 cs from stimulus onset (P < 0.05). Rats with mPFC lesions were different from controls in all bins (P < 0.05).
In terms of the distribution of accuracy, there was no significant difference between control rats and rats with Cg1 lesions (Fig. 6, right). All rats of experiment 1 almost always responded correctly when the stimulus was still on (0–35 cs). Then choice accuracy decreased gradually to ~90% correct in the second bin (35–80 cs) and to 50% correct in the third (80–130 cs). ANOVA of the percentage of correct responses found a significant effect of ‘bin’ [F(4,36) = 73.75, P < 0.01]), but no significant ‘lesion’ × ‘bin’ interaction [F(4,36) < 1, n.s.]. In contrast, in experiment 2 control rats differed from the lesioned subjects (Fig. 7, right). While in control rats (black columns) there was a similar pattern to that of the rats of experiment 1, the performance of rats with PRL–IL lesions dropped to near-chance level earlier than the other groups, with only ~35% correct in the third bin (80–130 cs). In the group of mPFC rats, the peak of accuracy was shifted towards longer latencies than in the other groups, suggesting that a marked slowing of responding contaminated the data in these animals. ANOVA of the percentage of correct responses found a significant effect of ‘bin’ [F(4,68) = 166.70, P < 0.01] and a significant ‘lesion’ × ‘bin’ interaction [F(8,68) = 11.59, P < 0.01]. Post hoc analysis revealed that mPFC animals were significantly less accurate than sham-operated rats or PRL–IL rats (P < 0.05) in the first time bin (0–40 cs), whereas PRL–IL rats were selectively impaired shortly after stimulus offset, i.e. between 80 and 130 cs from the beginning of the limited hold period (P < 0.05). In addition, rats with mPFC lesions had significantly higher accuracy scores in the third bin (P < 0.05). Taken together, these results suggest that: (i) the accuracy decrement of rats with Cg1 lesions was comparable to that of their controls; (ii) PRL–IL rats were not impaired when guiding stimuli were present in the environment, but were less accurate than controls when the stimulus had disappeared; and (iii) mPFC rats had the greatest level of accuracy in the second and third time bins, indicating attentional and motor deficits, also suggested by their overall slowed speed of responding.
There was no difference between controls and Cg1-lesioned animals on any of the measures of response bias (data not shown). Three of the controls and four of the Cg1-lesioned rats had bias toward one of the locations (χ2, P < 0.05). The proportion of incorrect responses made in a previously rewarded location was not significantly greater than chance [t(10) = 1.27, n.s.], nor significantly different in sham and Cg1-lesioned rats [F(1,9) < 1, n.s.]. Each animal made approximately half of the incorrect responses in a location next to the correct one [significantly >0.4; t(10) = 4.96, P < 0.01] and no significant difference between lesioned and sham-operated animals was found on this measure [F(1,9) < 1, n.s.]. In experiment 2, instead, all the rats with PRL–IL lesions and all but one of the rats with mPFC lesions had a bias toward one of the spatial locations (χ2, P < 0.05). As in experiment 1, only about half of the controls preferred one of the locations to the others, suggesting that lesions of PRL–IL (or larger mPFC lesions including Cg1 as well as PRL and IL) increased the bias to respond at a specific location. On any incorrect trial, rats with ventral or complete mPFC lesions, but not control animals, were also more likely to choose a location that had been rewarded on the last preceding correct trial. The proportion of incorrect responses made in a previously reinforced location was greater than the calculated probability (0.2) in PRL–IL [t(6) = 3.75, P = 0.01] or in mPFC rats [t(6) = 3.99, P < 0.01], but not in sham-operated controls [t(7) < 1, n.s.]). Overall, approximately half of the incorrect responses were made in a hole next to the correct one, which is significantly greater than the calculated probability [0.4; t(21) = 4.59, P < 0.01]. The proportion of incorrect responses made in a hole next to the correct one was significantly different from the calculated probability in controls [t(7) = 5.32, P < 0.01] and PRL–IL animals [t(6) = 2.71, P < 0.05], but not in mPFC-lesioned rats [t(6), n.s.].
Manipulations of the Basic Five-choice Schedule
Reducing the Stimulus Duration
This impaired the accuracy of discrimination in lesioned and control subjects in both experiments (Figs 8A and 9A). In both experiments, ANOVA of the percentage of correct responses found a significant effect of stimulus duration [F(1,9) = 36.31, P < 0.01, in experiment 1; F(1,18) = 70.68, P < 0.01, in experiment 2], but no significant ‘lesion’ × ‘stimulus duration’ interaction [F(1,9) < 1.0, n.s., in experiment 1; F(2,18) = 1.62, n.s., in experiment 2]. Rats with PRL–IL lesions tended to be impaired by this manipulation more than animals of other groups (mean ± SEM difference scores: experiment 1, 14 ± 4% correct for Cg1-lesioned and 13 ± 3% correct for control rats; experiment 2, 22 ± 2% correct for PRL–IL, 13 ± 4% correct for mPFC and 13 ± 3% correct for CTR rats), but this effect did not reach statistical significance. In experiment 1, reducing the stimulus duration also increased correct response latencies in Cg1-lesioned and sham-operated rats [F(1,9) = 9.91, P < 0.05; Fig. 8A, right], suggesting speed/accuracy trade-off. In experiment 2, reducing the stimulus duration increased perseverative responding in all groups [F(1,18) = 6.80, P < 0.02; ‘lesion’ × ‘stimulus duration’, F(2,18) = 1.34, n.s.]. There was no significant effect of stimulus duration on magazine latency or on the number of omissions in either experiment. Data from one of the PRL–IL animals were lost and are thus not included in these analyses.
Variable Short ITI
In experiment 1, the introduction of short variable ITI resulted in poorer choice accuracy compared to the fixed ITI condition [F(1,9) = 16.12, P < 0.01], although rats with Cg1 lesions tended to be less affected than sham-operated subjects (nearly significant ‘lesion’ × ‘ITI’ interaction, F(1,9) = 4.75, P = 0.057; main effect of ‘lesion’, F(1,9) = 6.66, P < 0.05; Fig. 8B, left]. There was no significant effect of this manipulation on correct latency [F(1,9) = 3.0, n.s.; Fig. 8B, right]. In experiment 2, the introduction of short ITI resulted in poorer performance in all groups, compared to the fixed ITI condition. ANOVA found a significant effect of the ITI on accuracy [F(1,19) = 34.87, P < 0.01; Fig. 9B, left] and on correct latency [F(1,19) = 31.22, P < 0.01; Fig. 9B, right], suggesting speed/accuracy trade-off. mPFC rats were affected less by this manipulation than were PRL–IL or CTR rats. ANOVA of the correct latencies found a significant ‘lesion’ × ‘ITI’ interaction [F(2,19) = 3.73, P < 0.05; Fig. 9B, right], indicating that mPFC rats were slowed less by variable short ITI than were PRL–IL or CTR rats. There was no significant interaction between ‘lesion’ and ‘ITI’ on accuracy [F(2,19) = 2.88, n.s.]. In both experiments, rats made more omissions [experiment 1, F(1,9) = 109.01, P < 0.01; experiment 2, F(1,19) = 156.63, P < 0.01] and fewer premature responses (experiment 1, F(1,9) = 8.68, P < 0.05; experiment 2, F(1,19) = 43.05, P < 0.01] with variable than with fixed ITI.
Variable Long ITI
Overall, the introduction of long variable ITI did not affect accuracy in experiment 1 [F(1,9) < 1, n.s.]. However, ANOVA found a significant ‘lesion’ × ‘ITI’ interaction [F(1,9) = 17.86, P < 0.01], indicating that this condition tended to impair performance of control animals, but improved target discrimination by Cg1-lesioned rats (Fig. 8C, left). Qualitatively similar findings emerged from the analysis of the data relative to mPFC rats in experiment 2. Overall, there was no effect of introducing long variable ITI on discriminative accuracy [F(1,19) = 3.26, n.s.]. However, ANOVA revealed a significant ‘lesion’ × ‘ITI’ interaction [F(2,19) = 5.01, P < 0.02], indicating that rats with mPFC lesions tended to be more accurate in the variable than in the fixed ITI condition (S–N–K, P = 0.06), whereas the other animals were impaired by this manipulation (S–N–K, P < 0.01). In terms of the other measures of performance, in experiment 1 the introduction of long variable ITI also resulted in an increase in the number of omissions [F(1,9) = 16.09, P < 0.01], but did not affect correct latency [F(1,9) = 1.1, n.s.]. In experiment 2 there was no effect of introducing long variable ITI on omissions [F(1,19) = 2.67, n.s.], but correct latencies were shortened in all groups [F(1,19) = 11.62, P < 0.01; Fig. 9C, right]. In both experiments, the introduction of long variable ITI resulted in increased anticipatory responding [experiment 1, F(1,9) = 65.61, P < 0.01; experiment 2, F(1,19) = 147.63, P < 0.01]. In experiment 2 there was also a significant ‘lesion’ × ‘ITI’ interaction, suggesting that larger increases occurred in the groups of CTR and PRL–IL animals than in the mPFC rats, possibly due to a ceiling effect in this latter group.
Persistence of the Deficits
Approximately 2 months post-surgery, Cg1-lesioned rats were still significantly less accurate than the sham-operated animals [F(1,9) = 9.58, P < 0.02; mean ± SEM — control, 88 ± 2%; Cg1, 77 ± 2%). By contrast, there was no difference in the number of omissions [F(1,9) = 1.79, n.s.], the latency to respond correctly [F(1,9) < 1, n.s.] or the latency to collect food rewards [F(1,9) = 1.79, n.s.]. In the last session, Cg1-lesioned rats also made more perseverative panel pushes [F(1,9) = 14.12, P < 0.01; mean ± SEM — control, 87 ± 13; Cg1, 167 ± 18] and more perseverative nose-poke responses [F(1,9) = 6.14, P < 0.05; mean ± SEM — control, 49 ± 14; Cg1, 100 ± 17) than controls. These latter two effects of Cg1 lesion were not present during the first 13 post-operative baseline sessions, but emerged towards the end of the behavioural challenges. In experiment 2, ~2 months post-surgery the performance of mPFC rats was still significantly poorer than that of sham-operated rats in terms of accuracy [F(2,19) = 4.17, P < 0.05; S–N–K, P < 0.05; mean ± SEM — control, 84 ± 2%; mPFC, 70 ± 2%; PRL–IL, 77 ± 6%), correct latency [F(2,19) = 11.43, P < 0.01; S–N–K, P < 0.01], omissions [F(2,19) = 27.38, P < 0.01; S–N–K, P < 0.01] and perseverative responses [F(2,19) = 6.43, P < 0.05; mean ± SEM — control, 14 ± 4; mPFC, 65 ± 8; PRL–IL, 95 ± 39). PRL–IL rats still made more perseverative responses than CTR rats (P < 0.05), but were comparable to these animals in terms of choice accuracy (S–N–K, n.s.). In the last session, rats with PRL–IL or mPFC lesions also made more perseverative panel presses than CTR rats [F(2,19) = 4.79, P < 0.05; mean ± SEM — control, 70 ± 9; mPFC, 164 ± 34; PRL–IL, 157 ± 36), an effect which was not present during baseline testing, but emerged during the course of the behavioural challenges.
The main result of the study is that lesions of two distinct mPFC subregions induced dissociable cognitive deficits. Cg1 lesions reduced choice accuracy and increased omissions, but did not affect the number of perseverative responses. Lesions of PRL–IL increased perseverative responding, but had only transient effects on accuracy, which disappeared following re-training. In addition, rats with Cg1 lesions were relatively improved by variable ITI, and the temporal distribution of their responses was compatible with a deficit in the temporal organization of behaviour. Rats with PRL–IL lesions were impaired at responding specifically after stimulus offset and often guided their responses based on bias toward a location that had been recently associated with reward. Taken together, these results suggest that the dorsal (Cg1) and ventral (PRL–IL) subregions of the mPFC may have distinct and dissociable executive functions, as discussed in the following sections.
Dorsal mPFC Lesions
Lesions of the dorsal mPFC (Zilles’s Cg1) impaired the ability of the animals to respond accurately to brief flashes of light presented at spatially unpredictable locations. This effect, which was robust and long-lasting, suggests that the animals were impaired at either attending to the stimuli or at selecting and organizing temporally their responses in the experimental chamber (Passetti et al., 2000). Dorsal mPFC lesions also increased the number of omissions and lengthened the magazine latency, which may suggest that these rats were not motivated to perform the task (Carli and Samanin, 1992; Muir, 1996). However, at the end of the experiment rats with Cg1 lesions were still impaired in terms of choice accuracy, but no longer had greater omission rates or magazine latencies than control rats, suggesting possibly a trade-off between omissions and incorrect responses during earlier phases of testing. In addition, although full data following this manipulation are not presented here, when the rats were tested following 1 h of free access to food, there was an effect of pre-feeding on performance of all animals, but no significant interaction between the effects of the lesion and pre-feeding, suggesting that Cg1-lesioned rats were not less motivated than controls to perform the task.
The behavioural challenges further clarified the nature of the impairments resulting from lesion of Cg1. Reducing the duration of the stimuli did not differentially affect performance of Cg1-lesioned rats and control subjects, suggesting that discrimination of the visual stimuli was not a critical determinant of the accuracy deficits associated with the lesion. This finding is in keeping with the findings of other experiments, in which the accuracy impairments resulting from disconnection lesions disrupting the mPFC–dorsal-striatal circuit were not differentially sensitive to reductions of the stimulus duration (Christakou et al., 2001). In contrast, rendering the stimulus presentations unpredictable with variable ITI selectively impaired the choice accuracy of sham-operated rats, but did not affect performance of rats with dorsal mPFC lesions, i.e. there was an improvement of these latter subjects relative to controls. Thus, the introduction of variable ITI revealed that controls, but not Cg1-lesioned rats, had been using temporal cues to perform the standard version of the task. This finding is also consistent with previous experiments with mPFC–dorsal-striatum disconnection lesions (Christakou et al., 2001) and suggests that rats with Cg1 lesions were unable to utilize temporal cues to plan their responses. It should be emphasized that the relative improvement of Cg1-lesioned rats cannot be explained in terms of these animals being faster or slower than controls to turn around after collecting the food reward. In fact, similar effects were found with both short and long variable ITI (albeit more pronounced in the latter case) and a careful analysis of the data (not presented here for brevity) revealed that between-subjects variability in the Cg1 lesion group (such that some rats might have been slower than average to turn around and some other faster) was not sufficient to explain this finding.
Towards the end of behavioural testing, two deficits emerged which were not present immediately after surgery: the number of perseverative responses and perseverative panel pushes increased in Cg1-lesioned rats. These may be ‘secondary’ effects of the lesion. For example, exaggerated perseverative responding might have emerged as a consequence of poor accuracy, possibly related to the lower number of rewards per session received by the lesioned animals after surgery. The present data do not permit any strong conclusions regarding these effects of the lesion. However, it is clear that the accuracy deficits were not secondary to perseverative responding or increased panel pressing, as the latter deficits only appeared in a few late sessions, while the former were present throughout post-surgical testing.
Ventral mPFC Lesions
Lesions of the ventral mPFC (including PRL and IL) selectively increased the number of perseverative responses, with little effect on other measures of performance. This finding is largely compatible with the results of previous studies with selective lesions of PRL (Chudasama and Muir, 2001) and highlights the importance of this region in the inhibitory control of perseverative, inappropriate responses. The only notable exception was related to the effect of the lesions on omissions, which were increased following PRL lesions in an earlier study (Chudasama and Muir, 2001), but not in experiment 2 of the present work. As discussed previously, a trade-off between incorrect responses and omissions, possibly related to the slightly different training and testing conditions used (e.g. shorter stimulus duration), might explain these discrepancies. Thus, the increased omissions of the study by Chudasama and Muir (Chudasama and Muir, 2001) following PRL lesions might be equivalent to the small choice accuracy deficits found in the present experiment.
Given that IL was lesioned together with PRL, the present results do not exclude the small effects of the PRL–IL lesions on accuracy being due to damage to IL and the perseverative deficits to damage of PRL (Chudasama and Muir, 2001). However, this seems unlikely on the following grounds. First, complete mPFC lesions resulted in similar accuracy deficits, regardless of whether IL was spared (Muir et al., 1996) or included in the lesion (mPFC rats of experiment 2). Secondly, recent experiments found intact choice accuracy following selective IL lesions (Y. Chudasama and T.W. Robbins, manuscript in preparation). Thirdly, recent anatomical studies indicate that IL is extensively interconnected with PRL, but not with any other cortical structure (Fisk and Wyss, 1999) and that there is a great deal of overlap in the connections between either region and the hippocampus, amygdala, striatum and ventral tegmental area (Groenewegen et al., 1990; Jay and Witter, 1991; Condé et al., 1995; Carr and Sesack, 1996; Hedou et al., 2000). Thus, PRL and IL may be operationally as well as anatomically tightly coupled and the accuracy deficits of PRL–IL rats might have been the result of damage to this functional unit.
Analysis of the Temporal and Spatial Distribution of the Responses
The analysis of the temporal and spatial distribution of the responses proved to be useful for further clarifying the effects of lesions on the attentional control of instrumental performance. In both experiments, control rats made the large majority of the responses very early in the holding period. This suggests that they were very consistent, across trials, in turning around immediately after food collection and responding as soon as the stimulus appeared. In terms of accuracy, this was, for control subjects, nearly 100% correct while the stimulus was still on and it was still >90% correct up to ~0.4 s after its offset. This time window (0–80 cs) includes all the responses, which were either completed or at least initiated while the stimulus was still on [see Passetti et al. (Passetti et al., 2000), for an estimate of average rat reaction and motor times]. Subsequently, choice accuracy gradually decreased, being still well above 40% correct until ~0.9 s after stimulus offset. This indicates that naïve, well-trained animals could hold ‘on-line’ mental representations of planned responses well after the stimulus had disappeared. It also suggests that the period between 0.4 and 0.9 s after stimulus offset is when the contribution of circuits holding ‘on-line’ behaviour-guiding information is most critical for performance. Thus, this analysis suggests that five-choice task performance depends at least in part upon the rats’ temporal organization of behaviour (as responses initiated late are more likely to be incorrect) and at least in part upon working memory (as on a minority of trials responses are made after stimulus offset). In terms of the spatial distribution of the incorrect responses, these were often in a location adjacent to the correct one and rarely predictable based on the outcome of previous trials, indicating that the behaviour of well-trained naïve rats is generally under control of the instrumental contingencies of the task and not guided by spatial bias.
Frontal damage profoundly disrupted this pattern, with dissociable effects of dorsal (Cg1) or ventral (PRL–IL) mPFC lesions. The temporal distribution of the responses of rats with dorsal mPFC lesions was closer to random than that of controls, suggesting that these rats were ‘disorganized’ and often missed the presentation of the stimulus. Indeed, Cg1-lesioned rats had an increased omission rate, poor accuracy (on some of the trials in which they missed the stimulus presentation they presumably ‘guessed’ which was the correct location by a random response) and were differentially affected by the introduction of variable ITI. However, these rats were as good as controls at bridging the gap between the stimulus offset and their nose poke response, and their performance was not especially disrupted by previous experience of rewarded responses. In contrast, the accuracy of rats with ventral mPFC lesions sharply declined after stimulus offset and their incorrect responses tended to cluster in locations that had recently been associated with reward. Thus, these results strengthen the dissociation of dorsal and ventral mPFC functions. Circuits centred on the dorsal mPFC appear to be involved in the integration of temporal and visuo-spatial information necessary for the organization of responses required by task performance. In contrast, ventral mPFC-centred circuits are relevant for the mnemonic aspects of the task and may also be critical for ensuring behavioural flexibility.
The finding of stimulus-dependent accuracy deficits in rats with PRL–IL lesions might explain the fact that their overall accuracy impairments were comparatively less robust and longer lasting than those associated with Cg1 lesions (which also resulted in increased omissions). In fact, like animals in all groups, PRL–IL-lesioned rats completed the majority of the responses early in the trial, i.e. when useful behaviour-guiding stimuli could be accessed in the environment. Thus, their accuracy deficits were transient, tended to be increased by reductions of the stimulus duration and presumably could be compensated for by improving consistency in the temporal organization of the responses required by the task.
While there was a clear dissociation between Cg1 and PRL–IL lesions in terms of the delay-dependent accuracy deficits (which were found following PRL–IL, but not following Cg1 lesions), both lesions resulted in a shift of the temporal distribution of the responses toward longer latencies. This pattern of results might suggest a functional hierarchy, within the mPFC, with information flowing from PRL–IL to Cg1. This notion would be analogous to the suggestion that in monkeys and humans the mid-ventrolateral PFC is involved in the formation and ‘on-line’ maintenance of internal representations, while the mid-dorsolateral PFC mediates processes such as monitoring and manipulating such information (Owen et al., 2000; Petrides, 2000).
Comparison with Previous Studies
Cg1 and PRL–IL lesions had complementary effects in determining the behavioural changes associated with complete mPFC lesions (Muir et al., 1996). The effects of mPFC lesions on accuracy were mainly reproduced by dorsal lesions, including Cg1 but sparing PRL and IL, while the effects of mPFC lesions on perseverative responding were reproduced by ventral, but not dorsal lesions (Table 1). PRL–IL lesions also resulted in accuracy deficits, although these were short lasting. The sum of the accuracy deficits of Cg1 and PRL–IL animals was roughly equivalent to the accuracy deficit of rats with complete mPFC lesions. In addition, accuracy deficits were determined by a reduced number of correct responses in Cg1-lesioned rats and by an increased number of incorrect responses in rats with PRL–IL lesions. Together with the results of the analyses of bias and the temporal distribution of the responses discussed above, these findings suggest that Cg1 and PRL–IL lesions affected accuracy by disrupting distinct mechanisms and their effects were additive.
The results of experiment 2 were largely consistent with the results of previous experiments using the five-choice task (Muir et al., 1996; Chudasama and Muir, 2001). The main exceptions were restricted to the effects of complete mPFC lesions during the manipulation phase. Muir and colleagues (Muir, 1996; Muir et al., 1996) found that after recovery, the accuracy deficits of mPFC-lesioned rats could be partly reinstated by reducing the stimulus duration or by introducing a variable ITI. In contrast, in the present experiments mPFC rats never recovered from the accuracy deficits and the manipulations did not worsen the impairments of mPFC-lesioned rats (if anything, rats with mPFC lesions showed a tendency for a relative improvement during the sessions with variable ITI). The reason for these discrepancies is unclear and can only be speculated upon. However, it should be noted that rats with PRL–IL lesions in the present study did show a tendency to be differentially affected by reducing the stimulus duration, suggesting that minor differences in the focus of the mPFC lesions might have contributed to curtail the effect of this manipulation in the present study, in which the effect of the challenges on performance of rats with complete mPFC lesions resembled rather more that of rats with dorsal than ventral mPFC lesions. In addition, control rats were affected by the introduction of variable ITI in the present experiments, but not in Muir et al. (Muir et al., 1996), suggesting that only in the present experiment had the animals learnt to use temporal cues in order to guide performance prior to surgery. Thus, a number of factors, including the different focus of the lesions, the shorter stimulus duration and perhaps the more extensive training (suggested by the importance of temporal cues and also by similar pre-surgical performance with shorter stimulus duration), might be possible sources of the minor discrepancies between experiment 2 and the experiments of Muir and colleagues (Muir et al., 1996).
The findings of the present experiments are not incompatible with previous studies using maze tasks and they provide a common framework for the mnemonic and attentional deficits of mPFC-lesioned rats. For example, our results are not incompatible with previous findings that ventral mPFC lesions result in delay-dependent accuracy deficits (Ragozzino et al., 1998) or behavioural inflexibility (Ragozzino et al., 1999a,b) in maze tasks; and they are also not incompatible with studies showing no effect of Cg1 lesions on performance of the same maze tasks (Ragozzino et al., 1998, 1999a). If, as suggested by Seamans et al. (Seamans et al., 1995), Cg1 is mainly involved in the organization of behaviour in well-learned task strategies, damage to this region would not be expected to affect tasks involving learning or placing little demand on response selection and planning. In contrast, Cg1 lesions would be expected to result in delay-independent deficits on pre-surgically trained tasks emphasizing egocentric strategies for response selection, either in mazes (Kesner et al., 1996), or in operant chambers (present study).
Nature of the Deficits
In the five-choice task, while control animals used temporal cues to guide performance, Cg1-lesioned rats appeared to have lost this ability, as revealed by the effect of introducing temporally unpredictable stimulus presentations and by the analysis of the temporal distribution of their responses. Thus, circuits centred on Cg1 might be relevant for the formation of appropriate response ‘sets’ when a long sequence of behaviours needs to be organized over time, or when performance depends upon the rapid integration of information regarding the context and the spatial and temporal features of the actions to be performed. Given its extensive interconnection on one side with PRL–IL and on the other with Fr2 and the retrosplenial cortices, Cg1 may be the site at which short-term memories, preparatory motor set and information regarding the temporal ordering of the responses to be made are put together in order to guide behaviour (Fuster, 1997). In keeping with this notion, the dorsal striatum, which is the main striatal projection area of Cg1, has been implicated in the formation of the response ‘sets’ required by performance of attentional tasks (Brown and Robbins, 1991).
Lesions of PRL and IL resulted in delay-dependent accuracy deficits, in increased perseverative responses and in bias towards a location previously associated with reward. An intriguing possibility is that circuits centred on the ventral mPFC may contribute to performance by maintaining task-relevant mental representations in an active state — a process that would prevent, by competition, prepotent, stimulus-triggered inappropriate responses taking control over behaviour. The neural circuitry of PRL–IL would support this suggestion. PRL–IL receives direct and indirect projections from association cortices and other prefrontal regions and is also mutually interconnected with the hippocampus, the amygdala and the ventral tegmental area. Through these, PRL–IL may access real-time information regarding processes that are becoming active and it may also regulate such activation (Floresco et al., 1997). Dopaminergic projections, which are densest in the PRL and IL areas of the mPFC, may be particularly relevant for this process. In fact, it has been suggested that dopamine release in the mPFC may affect the balance between cortical and subcortical influences in the control of action (Yang and Seamans, 1996) and it may help to stabilize mental representations needed to guide behaviour (Durstewitz et al., 1999, 2000). Finally, the main striatal projection area of PRL–IL includes the nucleus accumbens, which is known to be important for the influence on behaviour of reward-associated stimuli (Robbins and Everitt, 1992).
Relevance to Human Studies: Dissociating the ‘Central Executive’
Recent views of the human PFC hold that it has ‘executive’ functions and that these depend, at least in part, upon its property of maintaining memory representations in a highly active state and allowing their guidance of behaviour under conditions of interference (Shallice, 1982, 1988; Goldman-Rakic, 1987, 1998; Duncan, 1995; Baddeley, 1996; Fuster, 1997; Kesner, 1998; Robbins, 1998; Shallice and Burgess, 1998). Although most authors acknowledge this general notion, the precise nature and organization of executive functions is unclear. These may either be emergent properties resulting from the interaction between specialized subsystems or they may be associated with the operation of a single central executive, which may in turn have dissociable components (Robbins, 1998). For example, Robbins and colleagues have found evidence for dissociable impairments of planning and response inhibition, following frontal damage (Robbins, 1998).
The present results suggest that subregions of the rat mPFC may also have dissociable executive functions. Indeed, the dissociation of response selection (Cg1) and perseveration deficits (PRL–IL) resembles the dissociation of planning and response inhibition of humans. This has two important implications. First, it suggests possible general organizational principles for the mammalian brain. Rats and primates share a similar anatomical organization of parallel segregated cortical–striatal loops (Groenewegen et al., 1991) and they may also share a similar functional organization of systems for selecting, updating, monitoring and utilizing task-relevant mental representations (as suggested by the present results). The second important implication is the validation of rodent models of frontal lobe dysfunction, which may prove invaluable for studying the interactions between the PFC and other structures of the brain. Thus, while species differences are such as to suggest caution in drawing homologies between the prefrontal cortices of rats and primates (Preuss, 1995), rodent models can be applied to the task of further dissociating executive mechanisms of response selection.
This work was supported by the Wellcome Trust and completed within the MRC Cooperative in Brain, Behaviour and Neuropsychiatry. F.P. was supported by CeNeS plc.
|Key: –, reduced; +, increased; 0, no effect; ‘improved’, relative recovery compared to standard task.|
|Effect of reducing the stimulus duration||0||0||0|
|Effect of variable ITI||improved||improved||0|
|Temporal distribution of the responses||altered||altered||altered|
|Key: –, reduced; +, increased; 0, no effect; ‘improved’, relative recovery compared to standard task.|
|Effect of reducing the stimulus duration||0||0||0|
|Effect of variable ITI||improved||improved||0|
|Temporal distribution of the responses||altered||altered||altered|