Copyright 1990 by the American Psychological Association, Inc. 0735-7044/90/$00.75
Behavioral Neuroscience 1990, Vol. 104, No. 5, 655-665
Proactive Interference Effects on Short-Term Memory in Rats: I. Basic Parameters and Drug Effects
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Stephen B. Dunnett and Fran L. Martel University of Cambridge, Cambridge, United Kingdom An operant delayed-matching task was used to assess the role of proactive interference (PI) effects on short-term memory capacity of rats. Task performance was analyzed in terms of the influence of the sample positions and response choices on previous trials. PI was predominantly attributable to the influence of the immediately previous trial but not preceding trials and was abolished by increasing the intertrial intervals from 5 to 15s. Nicotine induced a decline in choice accuracy only on trials in which the previous response had been to the side opposite the current sample and correct response, suggesting an increased susceptibility to PI. Physostigmine induced a mild, relatively nonspecific decline in response accuracy. Clonidine induced delaydependent impairments irrespective of responses on previous trials. None of these drugs enhanced choice accuracy at any dose tested.
The particular advantage of delayed matching and other delayed response tasks in the study of short-term memory lies in the possibility of distinguishing mnemonic consequences of experimental manipulations from other non specific effects. Thus, the maintenance of accurate performance when the sample and choice responses are separated by only a short delay interval indicates that the animal is attending to the relevant stimuli and that the discrimination rule is well learned. The appearance of impairments in response accuracy only as delay intervals increase then suggests that the disruptive effects of the experimental treatment are attributable to a disturbance in the animals' working memory capacity rather than to nonspecific effects on attention, motivation, or sensory or motor capacities. Such a change in slope of the delay-performance function has been seen in aged animals (Bartus, Fleming, & Johnson, 1978; Dunnett, Evenden, & Iversen, 1988; Winocur, 1984), in elderly or mildly demented humans (Flicker, Ferris, Bartus, & Crook, 1984; Sahakian et al., 1988), and after experimental damage in the septo-hippocampal circuitry (Dunnett, 1985; Mishkin, 1982; Sahgal, 1984; Winocur, 1985; Zola-Morgan etal., 1989). Less clear is the nature of the underlying processes that lead to forgetting from working memory in such tasks. One potential process that has been found to influence delayed matching-to-sample performance in young animals is proactive interference. For example, in delayed matching- and nonmatching-to-sample tasks, there is substantial evidence that stimuli and responses on the previous trial (N - 1) can interfere with choice accuracy on the current trial (N) and that such interference diminishes as the intertrial interval (ITI) increases (Edhouse & White, 1988; Jarrard & Moise, 1971; Jarvik, Goldfarb, & Carley, 1969; Maki, Moe, & Bierley, 1977; Moise, 1976; Nelson & Wasserman, 1978; Roberts, 1980). Those studies were conducted in pigeons and monkeys, but proactive interference effects have also been observed in rats in radial and three-choice maze tasks (Roberts & Dale, 1981; Roitblatt & Harley, 1988) and in a serial nonmatching task (Pontecorvo, 1983). The present study has addressed five questions related to the role of proactive interference in rats' working memory.
Delayed matching-to-sample tasks have provided a powerful means of testing short-term memory in animals (Mackintosh, 1983). On each trial, the animal is first presented a sample stimulus and then, after a variable delay interval, is required to make a choice between the sample and one or several other stimuli; a response to the previous sample is reinforced, whereas other responses are not reinforced or are punished by "time-out." Such tasks generally reveal a decline in performance as the delay interval between the sample and choice is lengthened, and the slope of the delay-performance curve thereby provides an index of the animals' rate of forgetting from short-term (working) memory (Heise, 1975; Roitblatt, 1987). Delayed matching-to-sample and related tasks have been widely used in monkeys (D'Amato, 1973; Jarrard & Moise, 1971; Mishkin, 1982; Zola-Morgan, Squire, & Amaral, 1989) and in pigeons (Blough, 1959; Pontecorvo & Evans, 1985; Roberts, 1980) but have not until recently been used in rodent studies. This situation has changed with the introduction of delayed matching and nonmatching tasks for rats in both operant boxes (Dunnett, 1985; Etherington, Mittleman, & Robbins, 1987; Kesner, Bierley, & Pebbles, 1981; Ksir, 1974) and mazes (Aggleton, 1985; Alexinsky & Alliot, 1987; Haig, Rawlins, Olton, Mead, & Taylor, 1983; Rawlins & Olton, 1982; Roitblatt & Hayes, 1987). In particular, we have developed an operant delayed matching-to-position (DMTP) paradigm that has provided a powerful test for assessing the effects of aging, cholinergic and peptidergic drugs, lesions, and neural grafts on short-term memory capacities of rats (Dunnett, 1985; Dunnett, Badman, Rogers, Evenden, & Iversen, 1988; Dunnett, Evenden, & Iversen, 1988; Sahgal, 1987). This work was supported by a project grant from the Mental Health Foundation of the United Kingdom. Fran Martel is now at the Sub-Department of Animal Behaviour, University of Cambridge. Correspondence concerning this article should be addressed to Stephen B. Dunnett, Department of Experimental Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, United Kingdom.
655
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
656
STEPHEN B. DUNNETT AND FRAN L. MARTEL
First, we have determined whether proactive interference influences performance in the standard version of the DMTP task. This was addressed by collecting data on response accuracy separately for trials in which the previous sample had been on the same side and for trials in which the previous sample had been on the opposite side. Reduction of choice accuracy on previous-sample-opposite trials below that observed on previous-sample-same trials would indicate proactive interference from the previous trial. Second, we have considered whether earlier trials (N — 2, N - 3, ...) before the previous trial N - 1 might also influence performance on the present trial N by considering whether runs of trials on the same side might provide further improvements in choice accuracy. Third, we have tested whether increasing the ITI beyond the 5 s involved in the standard task reduces the effects of proactive interference in choice accuracy. Fourth, in the search for the specific mechanisms involved in proactive interference effects, previous studies in pigeons and monkeys have indicated that the response and outcome on the previous trial influence performance on the current trial to as great an extent as the previous sample stimulus (Maki et al., 1977; Moise, 1976). We have therefore collected data separately not only for previous-sample-same and previous-sample-opposite trials but also according to whether the responses on the previous trials were correct or incorrect. Consequently, the data could be reclassified in terms of previous-response-same and previous-response-opposite trials. Finally, as part of a more general program investigating the role of aging in memory (Dunnett, Evenden, & Iversen, 1988), we have considered the effects of three drugs that have been reported to enhance learning or memory performance in young or aged animals and humans on proactive interference in the DMTP task: clonidine (e.g., Arnsten & Goldman-Rakic, 1985; Brozoski, Brown, Goldman, & Rosvold, 1979), physostigmine (e.g., Aigner & Mishkin, 1986; Bartus, Dean, & Beer, 1980; Bartus, Dean, Beer, & Lippa, 1982), and nicotine (e.g., Evangelista, Gattioni, & Izquierdo, 1970; Flood, Landry, & Jarvik, 1981; Sahakian, Jones, Levy, Gray, & Warburton, 1989). Method Subjects Twenty-two young adult female Sprague-Dawley rats (Olac, Bicester, Oxon, UK) were housed in groups of 5-6 rats per cage on a 12:12-hr light-dark cycle and with free access to water throughout. The rats were deprived food, except for 8-15 g standard laboratory chow per rat that was given at the end of each afternoon to maintain body weights at approximately 90% of free-feeding levels.
Apparatus Testing was conducted in six operant chambers (Campden Instruments, London, UK) under the online control of a Spider microprocessor (Paul Fray Computers, Cambridge, UK). Each chamber was fitted with two retractable levers situated 7.5 cm either side of a central food tray that had a hinged Perspex panel at which nose pokes were registered. A panel light was located on the outside of
the chamber above the food tray for back illumination of the Perspex panel, and a house light was located in the center of the ceiling. A food dispenser delivered 45-mg dustless reinforcement pellets (Bioserv, from Sandown Scientific, Esher, UK) to the food tray.
Delayed Matching-to-Position Schedule In the standard version of the DMTP schedule, each trial involved three stages: sample response, delay interval, and choice response. At the start of each trial, the side of the sample (left or right) and the delay interval (0-24 s or 0-32 s) were selected randomly. The sample lever was inserted into the chamber. As soon as the rat made a lever press response, the lever was retracted, the panel light was turned on, and the delay interval clock was started. Nose pokes at the panel were recorded throughout the delay but had no effect until the delay interval was completed. The first nose poke made by the rat after completion of the scheduled delay terminated the delay stage; the panel light was then turned off, and two levers were inserted together into the chamber (the choice). If the rat made a correct matching response (i.e., to the same lever as presented during the sample stage of the trial), then the levers were retracted, a food pellet was delivered to the food tray, and the panel light was switched on until a further nose poke indicated that the pellet had been collected. Alternatively, if the rat made an incorrect nonmatching response (i.e., to the other lever), then the levers were again retracted, but the house light was switched off for a 5-s time-out period, and no reinforcement was delivered. After collection of the food pellet on correct trials or time-out on the incorrect trials, a 5-s ITI with the house light switched on preceded the next trial. The purpose of requiring the rat to make a nose-poke response to the panel to terminate the delay stage was to prevent the adoption of a mediating response (e.g., waiting by the correct lever) to solve the DMTP task. Because the scheduled delay interval was unknown to the rat, the nose-poke response was maintained on a second-order variable-interval schedule that sustained a high rate of responding at the central panel (M = 1.11, SD = 0.42, responses per second) throughout all delay intervals.
Procedure The rats were tested in daily 40-min sessions. The procedures for initial habituation to the chambers, training to lever press, and training on the matching task in the absence of delay intervals, before training on the DMTP task itself, have been described fully elsewhere (Dunnett, 1985; Dunnett, Evenden, & Iversen, 1988). The rats received 30 days training on the task to establish efficient levels of performance in all rats before the present experiments began. Performance stabilized at 120-150 trials completed on each day (i.e., approximately 20 trials at each scheduled delay), and data were collapsed over several days (minimum of 4) for each experimental manipulation. The basic measure of performance was percentage correct responses at each delay interval. The rats were initially tested with scheduled delays randomly selected from the set: 0, 2, 4, 8, 12, 16, and 24 s. To detect proactive interference effects, the computer was programmed to collect data separately for trials in which the sample on the previous trial was on the same or on the opposite side to that on the current trial Variations of the schedule were conducted on probe test days. First, to determine whether samples before the previous sample would also interfere with performance on the current trial, trials were categorized according to whether the sample on the previous trial had been on the opposite side (i.e., the present sample comprises a run length of 1) or the number of preceding trials in which the sample had been on the same side as the present trial (run lengths of 2-4).
657
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
PROACTIVE INTERFERENCE IN RATS On those occasions that the four preceding trials had all been on the same side, the current sample was not selected randomly but was forced to the other side. Because the variation of delay intervals from trial to trial would seriously confound this analysis, the effect of run lengths was assessed on probe days in which all trials had the same delay of 0, 4, 12, or 24 s. Each of these four fixed delays was tested on 4 separate test days in cyclic order with 1 intervening baseline day between each, over 32 days. The side of the sample was determined randomly on each trial, and the effects of runs of up to four consecutive trials with the sample on the same side were recorded. Second, to assess whether proactive interference from the previous trial might be reduced by increasing the ITI, the rats were tested on the standard DMTP schedule with ITIs of 5,15, or 45 s. The sessions with longer ITIs were increased in duration (to 50 min for a 15-s ITI; to 60 min for a 45-s ITI) to enable similar numbers of trials to be completed on each day. The animals were tested four times on each of the alternate ITIs in cyclic order on consecutive days. Third, to determine whether the side of the animals' response on the previous trial may have a greater effect than the side of the sample lever per se, the programs were modified to collect not only the side of the previous sample but also to determine whether the response was correct. The task was also made more difficult by increasing the scheduled delays to 0, 4, 8, 12, 18, 24, or 32 s. Because the rats were performing at 90-95% correct overall by this stage, the proactive effects of the previous sample and previous response were generally concordant on the great majority of trials. To collect a sufficient number of the critical incorrect trials (in which the sample alone or the response alone would be on the side opposite to the present sample) for estimation of percentage correct performance for each rat at each delay, data were collapsed over a further 17 days of testing.
(a) Previous sample 100 I
90
80
O — O previous sample same • — • previous sample opposite
12
16
(b) Runs of samples on same side 100 T
90-
80-
• 0 V----V A— A O O
run length run length run length run length
1 2 3 4
Drug Effects 70
The effects of physostigmine salicylate (Addenbrookes Hospital, Cambridge, UK; 0.0, 0.03, 0.06, or 0.1 mg/kg ip), nicotine sulfate (Sigma Chemical Co., Poole, Dorset, UK; 0.0, 0.03, 0.1, or 0.3 mg/ kg sc), and clonidine hydrochloride (Sigma; 0.0, 0.003, 0.01, or 0.1 mg/kg ip) on the DMTP schedule were assessed using the set of longer (0-32 s) delays. Each drug was diluted or dissolved in isotonic saline and administered 5 min before the test in a series of increasing doses on consecutive days. The series of injections was repeated four times for each drug, with 2 or 3 baseline days between each cycle.
12
(c) Increasing intertrial intervals 100T
90--
0 — O 5s ITI previous same
Results All animals learned to perform the DMTP task efficiently, although 2 animals were removed from later tests because of respiratory infections.
Side of Previous Sample The effect of the previous sample on the percentage correct at each delay over Test Days 30-40 is shown in Figure la. As described previously, the rats showed a decline in performance at longer delays, indicating a delay-dependent for-
Figure 1. a: Proactive interference in the delayed matching-to-position task. (Performance accuracy [percentage correct responses] at each trial delay separately for trials in which the sample on the previous trial was on the same or the opposite side to the sample
24
•— • 80--•
A— A A— A V— V T— V
"~~—-_.0
5s ITI previous opposite 15s ITI previous same 15s ITI previous opposite 45s ITI previous same 45s ITI previous opposite
I
70
12
18
24
32
delay (sec)
on the current trial, b: Performance accuracy at four trial delays, tested on separate sessions, for trials in which the previous trial was on the opposite side [run length = 1], or trials in which the most recent sample on the opposite side was two, three, or four trials previously [run lengths = 2-4, respectively], c: Effect of intertrial interval on proactive interference. Performance accuracy at each trial delay separately for trials in which the sample on the previous trial was on the same or the opposite side to the sample on the current trial, and for sessions with 5-, 15-, or 45-s intertrial intervals. Vertical bars indicate two SEs derived from the relevant interaction terms of the analyses of variance.)
658
STEPHEN B. DUNNETT AND FRAN L. MARTEL
getting from short-term memory. Moreover, performance at the longer delays was worse when the previous sample had been on the opposite side than when it had been on the same side as the current trial, indicating a proactive interference effect. All effects were highly significant: Delay, F(6, 126) = 35.33; Previous Trial, F(l, 21) = 16.86; Delay x Previous Trial, F(6, 126) = 4.38; allps < .001.
(a) Previous sample and response
100 T
w
80-
•— • O—O D— D •—•
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
60 ••
SAMPLE
RESPONSE
same same opposite opposite
same opposite same opposite
Runs of Preceding Samples O
40
0
4
8
12
18
24
32
(b) Previous response
100
Effect oflntertrial Intervals
90--
80-
—* previous response same — 4 previous response opposite
70 0
8
12
18
24
32
(c) Previous sample 100 T
As shown in Figure Ib, rats performed worse at all longer delays on trials in which the previous sample had been on the opposite side (defined as a run length of 1) than on trials in which the previous sample was on the same side. However, preceding trials had no further influence: The trials in which the most recent opposite sample was 2, 3, or 4 trials previously (run lengths = 2, 3, or 4, respectively) did not differ. All effects were significant: run length, F(3, 63) = 12.62, p < .001; delay, F(3, 63) = 106.09, p < .001; Run Length x Delay, F(9, 189) = 2.12, p < .05.
•
As shown in Figure Ic, proactive interference from the previous trial was abolished with the longer 15- and 45-s ITIs. In particular, performance was disrupted only in the case in which the previous sample had been on the opposite side and the ITI was short. Performance on trials with short ITIs in which the previous sample had been the same did not differ from performance on either type of trial on the test days with longer ITIs. All main effects were significant: ITI, F(2, 42) = 9.56, p < .001; previous trial, F(l, 21) = 4.96, p < .05; delay, F(6,126) = 68.35,p < .001; the only significant interaction was ITI x Previous Trial, F(2, 42) = 6.16, p < .001. Effect of Previous Response
90-
80
• — • previous sample same • — • previous sample opposite
70
12
18
24
32
delay (sec) Figure 2. Proactive effects analyzed in terms of the previous sample and the previous response, (a: Trials subdivided into whether the previous sample [determined randomly] had been on the same or opposite side and in terms of whether the response was correct [previous sample and response both on the same or on the opposite sides] or was incorrect [previous sample same and previous response opposite, or vice versa]. Note that correct trials occurred 8-10 times more frequently than incorrect trials, b: Data of Panel a collapsed in terms of the side of the previous response, c: Data of Panel a collapsed in terms of the side of the previous sample. Note that proactive interference from the previous trial is more clearly ap-
The separate effects of previous sample and response are shown in Figure 2a. Consideration of the trials in which a correct matching response had been made on the previous trial again revealed a proactive interference effect (i.e., performance was better on trials when both the previous sample and the previous response had been on the same lever than when they had both been on the opposite lever). However, the greatest disruption of performance was on trials in which an erroneous response had been made on the previous trial to the opposite lever after presentation of the sample on the same side as on the present trial, resulting in a significant main effect of previous sample, F(l, 21) = 8.45, p < .01, that is, in the direction opposite from that observed in previous analyses in which all trials had equal weight. The main effect of previous response was also highly significant, F(l,
parent when the data are analyzed in terms of the response rather than the sample on the previous trial. The vertical bar indicates two SEs derived from the three-way interaction of the analysis of variance.)
659
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
PROACTIVE INTERFERENCE IN RATS
21) = 51.88, p < .001 (i.e., the animals performed better on trials in which the previous response had been to the same lever), and the Previous Sample x Response interaction was also significant, F(l, 21) = 34.57, p < .001. The Previous Response x Delay interaction, but not the Previous Sample x Delay interaction, was also significant, F(6, 126) = 5.92, p < .001, and F(6, 126) = 2.09, p = .059, respectively. Because the animals achieved 85-95% correct choices overall, the side of the previous sample and the previous response acted in concert on the great majority of trials. However, the pattern of results observed when these two factors were considered independently suggests that proactive effects from the previous trial are better described in terms of interference from the previous response than from the previous sample. To illustrate this, the same data were reanalyzed separately in terms of trials in which the previous sample was on the same or opposite side as the present sample (as previously) or in which the previous response was on the same or the opposite side. Proactive effects, as indicated by separation of performance on previous-same and previous-opposite trials, were much greater when considered in terms of the previous response (see Figure 2b), previous response, F ( l , 21) = 44.92; Previous Response x Delay, F(6, 126) = 6.43; bothps < .001, than when considered in terms of previous sample (see Figure 2c), previous sample, F ( l , 21) = 19.21, p < .001; Previous Sample x Delay, F(6, 126) = 2.03, p = 0.066. Inspection of trials in which the previous response was made to the opposite side indicated that performance was higher when the previous responses was correct to an opposite sample than when it was in error to a same sample. This suggests that the previous sample per se has no direct influence on proactive interference. Consequently, subsequent analyses were conducted in terms of the influence of the response, rather than of the sample, on the previous trial. Effect of Physostigmine The effects of injections of saline or three doses of physostigmine are shown in Figure 3. There was a significant main effect of drug dose, F(3, 63) = 5.56, p < .01, which was attributable to performance under the two higher doses of physostigmine that was moderately (but significantly) below that under saline: Newman-Keuls tests, saline versus 0.1 mg/kg, t(4, 63) = 4.90, p < .01; saline versus 0.06 mg/kg, t(3,63) = 3.54,p < .05. There were no significant interactions between drug dose and previous response or delay. Thus, at the doses used, physostigmine appears to induce a mild generalized disturbance of performance in DMTP. Effect of Nicotine The effects of injections of saline or three doses of nicotine are shown in Figure 4. There was a significant main effect of drug dose, F(3, 63) = 206.77, p < .001, which was attributable to performance under the two higher doses of nicotine that was substantially below that under saline: NewmanKeuls tests, saline versus 0.3 mg/kg, t(4, 63) = 31.04; saline versus 0.1 mg/kg, t(3, 63) = 13.20; bothps < .01. The dis-
ruption of performance induced by nicotine was most marked on trials in which the previous response was on the side opposite to the present sample, whereas even high doses had little effect on trials in which the previous response was on the same side (see Figure 4). All drug interactions were highly significant: Drug x Previous Response, F(3, 63) = 30.03; Drug x Delay, F( 18,378) = 9.23; Drug x Previous Response x Delay, F(IS, 378) = 2.87; allps < .001. Thus, at the doses used, nicotine appears to induce a marked disturbance of DMTP performance, in particular by increasing the effects of proactive interference from the response made on the previous trial. Effect
ofClonidine
The effects of injections of saline or three doses of clonidine are shown in Figure 5. There was a significant main effect of drug dose, F(3, 57) = 80.13, p < .001, which was attributable to a disruption of performance under the highest dose of clonidine alone: Newman-Keuls tests, saline versus 0.1 mg/ kg, t(4, 57) = 17.38, p < .01. The Dose x Delay interaction was highly significant, F(18, 342) = 10.60, p < .001, but the Dose x Previous Response and the three-way interactions were not significant. This suggests that the highest dose of clonidine has no effect on the proactive interference from the previous response per se but induced a marked, delay-dependent impairment of performance irrespective of responses on previous trials. Drug Effects on Response Rate The mean numbers of trials completed in each 40-min test session were compared at each dose of the three drugs to provide an indication of nonspecific effects of these treatments. As shown in Figure 6, the higher doses of all three compounds induced a significant decrease in the number of trials completed: physostigmine, F(3, 63) = 26.08; nicotine, F(3, 63) = 56.72; clonidine, F(3, 57) = 177.49; all/?s < .001. Other response rate measures, such as the rate of nose poking at the central panel during the delay intervals, showed a similar decline in the cases of the higher doses of all three drugs (data not shown).
Discussion The present observations provide clear evidence that the response made on the previous trial can interfere with choice accuracy of rats trained in the DMTP task. The results are compatible in several respects with previous descriptions of proactive interference effects in animals' short-term memory performance, which has been studied most systematically in pigeons trained on delayed matching- and nonmatching-tosample tasks. Thus, for example, the effects were most apparent at short ITIs, affected trials with long but not short delays, and carried over from the previous trial but did not accumulate over longer series of trials (Edhouse & White, 1988; Maki et al., 1977; Moise, 1976; Nelson & Wasserman, 1978; Roberts, 1980; Roitblat & Harley, 1988). These observations are based on comparisons of previous-sample-
660
STEPHEN B. DUNNETT AND FRAN L. MARTEL
(a) Saline
(b) Physostigmine 0.03 mg/kg
lOO-i
100, g
«. " O. ,
90o o
*?
O - - - O Previous response same
„
0
4
8
12
18
24
32
0
4
8
12
18
(c) Physostigmine 0.06 mg/kg
(d) Physostigmine 0.10 mg/kg
100
100
T
a.
"°
o
24
32
24
32
o..
90 -
•-0
90-
80-
80-
70
70
8
12
18
24
32
4
8
12
18
(e) Previous response opposite
(f) Previous response opposite
1UO-
10
?*
Behavioral Neuroscience 1990, Vol. 104, No. 5, 655-665
Proactive Interference Effects on Short-Term Memory in Rats: I. Basic Parameters and Drug Effects
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Stephen B. Dunnett and Fran L. Martel University of Cambridge, Cambridge, United Kingdom An operant delayed-matching task was used to assess the role of proactive interference (PI) effects on short-term memory capacity of rats. Task performance was analyzed in terms of the influence of the sample positions and response choices on previous trials. PI was predominantly attributable to the influence of the immediately previous trial but not preceding trials and was abolished by increasing the intertrial intervals from 5 to 15s. Nicotine induced a decline in choice accuracy only on trials in which the previous response had been to the side opposite the current sample and correct response, suggesting an increased susceptibility to PI. Physostigmine induced a mild, relatively nonspecific decline in response accuracy. Clonidine induced delaydependent impairments irrespective of responses on previous trials. None of these drugs enhanced choice accuracy at any dose tested.
The particular advantage of delayed matching and other delayed response tasks in the study of short-term memory lies in the possibility of distinguishing mnemonic consequences of experimental manipulations from other non specific effects. Thus, the maintenance of accurate performance when the sample and choice responses are separated by only a short delay interval indicates that the animal is attending to the relevant stimuli and that the discrimination rule is well learned. The appearance of impairments in response accuracy only as delay intervals increase then suggests that the disruptive effects of the experimental treatment are attributable to a disturbance in the animals' working memory capacity rather than to nonspecific effects on attention, motivation, or sensory or motor capacities. Such a change in slope of the delay-performance function has been seen in aged animals (Bartus, Fleming, & Johnson, 1978; Dunnett, Evenden, & Iversen, 1988; Winocur, 1984), in elderly or mildly demented humans (Flicker, Ferris, Bartus, & Crook, 1984; Sahakian et al., 1988), and after experimental damage in the septo-hippocampal circuitry (Dunnett, 1985; Mishkin, 1982; Sahgal, 1984; Winocur, 1985; Zola-Morgan etal., 1989). Less clear is the nature of the underlying processes that lead to forgetting from working memory in such tasks. One potential process that has been found to influence delayed matching-to-sample performance in young animals is proactive interference. For example, in delayed matching- and nonmatching-to-sample tasks, there is substantial evidence that stimuli and responses on the previous trial (N - 1) can interfere with choice accuracy on the current trial (N) and that such interference diminishes as the intertrial interval (ITI) increases (Edhouse & White, 1988; Jarrard & Moise, 1971; Jarvik, Goldfarb, & Carley, 1969; Maki, Moe, & Bierley, 1977; Moise, 1976; Nelson & Wasserman, 1978; Roberts, 1980). Those studies were conducted in pigeons and monkeys, but proactive interference effects have also been observed in rats in radial and three-choice maze tasks (Roberts & Dale, 1981; Roitblatt & Harley, 1988) and in a serial nonmatching task (Pontecorvo, 1983). The present study has addressed five questions related to the role of proactive interference in rats' working memory.
Delayed matching-to-sample tasks have provided a powerful means of testing short-term memory in animals (Mackintosh, 1983). On each trial, the animal is first presented a sample stimulus and then, after a variable delay interval, is required to make a choice between the sample and one or several other stimuli; a response to the previous sample is reinforced, whereas other responses are not reinforced or are punished by "time-out." Such tasks generally reveal a decline in performance as the delay interval between the sample and choice is lengthened, and the slope of the delay-performance curve thereby provides an index of the animals' rate of forgetting from short-term (working) memory (Heise, 1975; Roitblatt, 1987). Delayed matching-to-sample and related tasks have been widely used in monkeys (D'Amato, 1973; Jarrard & Moise, 1971; Mishkin, 1982; Zola-Morgan, Squire, & Amaral, 1989) and in pigeons (Blough, 1959; Pontecorvo & Evans, 1985; Roberts, 1980) but have not until recently been used in rodent studies. This situation has changed with the introduction of delayed matching and nonmatching tasks for rats in both operant boxes (Dunnett, 1985; Etherington, Mittleman, & Robbins, 1987; Kesner, Bierley, & Pebbles, 1981; Ksir, 1974) and mazes (Aggleton, 1985; Alexinsky & Alliot, 1987; Haig, Rawlins, Olton, Mead, & Taylor, 1983; Rawlins & Olton, 1982; Roitblatt & Hayes, 1987). In particular, we have developed an operant delayed matching-to-position (DMTP) paradigm that has provided a powerful test for assessing the effects of aging, cholinergic and peptidergic drugs, lesions, and neural grafts on short-term memory capacities of rats (Dunnett, 1985; Dunnett, Badman, Rogers, Evenden, & Iversen, 1988; Dunnett, Evenden, & Iversen, 1988; Sahgal, 1987). This work was supported by a project grant from the Mental Health Foundation of the United Kingdom. Fran Martel is now at the Sub-Department of Animal Behaviour, University of Cambridge. Correspondence concerning this article should be addressed to Stephen B. Dunnett, Department of Experimental Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, United Kingdom.
655
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
656
STEPHEN B. DUNNETT AND FRAN L. MARTEL
First, we have determined whether proactive interference influences performance in the standard version of the DMTP task. This was addressed by collecting data on response accuracy separately for trials in which the previous sample had been on the same side and for trials in which the previous sample had been on the opposite side. Reduction of choice accuracy on previous-sample-opposite trials below that observed on previous-sample-same trials would indicate proactive interference from the previous trial. Second, we have considered whether earlier trials (N — 2, N - 3, ...) before the previous trial N - 1 might also influence performance on the present trial N by considering whether runs of trials on the same side might provide further improvements in choice accuracy. Third, we have tested whether increasing the ITI beyond the 5 s involved in the standard task reduces the effects of proactive interference in choice accuracy. Fourth, in the search for the specific mechanisms involved in proactive interference effects, previous studies in pigeons and monkeys have indicated that the response and outcome on the previous trial influence performance on the current trial to as great an extent as the previous sample stimulus (Maki et al., 1977; Moise, 1976). We have therefore collected data separately not only for previous-sample-same and previous-sample-opposite trials but also according to whether the responses on the previous trials were correct or incorrect. Consequently, the data could be reclassified in terms of previous-response-same and previous-response-opposite trials. Finally, as part of a more general program investigating the role of aging in memory (Dunnett, Evenden, & Iversen, 1988), we have considered the effects of three drugs that have been reported to enhance learning or memory performance in young or aged animals and humans on proactive interference in the DMTP task: clonidine (e.g., Arnsten & Goldman-Rakic, 1985; Brozoski, Brown, Goldman, & Rosvold, 1979), physostigmine (e.g., Aigner & Mishkin, 1986; Bartus, Dean, & Beer, 1980; Bartus, Dean, Beer, & Lippa, 1982), and nicotine (e.g., Evangelista, Gattioni, & Izquierdo, 1970; Flood, Landry, & Jarvik, 1981; Sahakian, Jones, Levy, Gray, & Warburton, 1989). Method Subjects Twenty-two young adult female Sprague-Dawley rats (Olac, Bicester, Oxon, UK) were housed in groups of 5-6 rats per cage on a 12:12-hr light-dark cycle and with free access to water throughout. The rats were deprived food, except for 8-15 g standard laboratory chow per rat that was given at the end of each afternoon to maintain body weights at approximately 90% of free-feeding levels.
Apparatus Testing was conducted in six operant chambers (Campden Instruments, London, UK) under the online control of a Spider microprocessor (Paul Fray Computers, Cambridge, UK). Each chamber was fitted with two retractable levers situated 7.5 cm either side of a central food tray that had a hinged Perspex panel at which nose pokes were registered. A panel light was located on the outside of
the chamber above the food tray for back illumination of the Perspex panel, and a house light was located in the center of the ceiling. A food dispenser delivered 45-mg dustless reinforcement pellets (Bioserv, from Sandown Scientific, Esher, UK) to the food tray.
Delayed Matching-to-Position Schedule In the standard version of the DMTP schedule, each trial involved three stages: sample response, delay interval, and choice response. At the start of each trial, the side of the sample (left or right) and the delay interval (0-24 s or 0-32 s) were selected randomly. The sample lever was inserted into the chamber. As soon as the rat made a lever press response, the lever was retracted, the panel light was turned on, and the delay interval clock was started. Nose pokes at the panel were recorded throughout the delay but had no effect until the delay interval was completed. The first nose poke made by the rat after completion of the scheduled delay terminated the delay stage; the panel light was then turned off, and two levers were inserted together into the chamber (the choice). If the rat made a correct matching response (i.e., to the same lever as presented during the sample stage of the trial), then the levers were retracted, a food pellet was delivered to the food tray, and the panel light was switched on until a further nose poke indicated that the pellet had been collected. Alternatively, if the rat made an incorrect nonmatching response (i.e., to the other lever), then the levers were again retracted, but the house light was switched off for a 5-s time-out period, and no reinforcement was delivered. After collection of the food pellet on correct trials or time-out on the incorrect trials, a 5-s ITI with the house light switched on preceded the next trial. The purpose of requiring the rat to make a nose-poke response to the panel to terminate the delay stage was to prevent the adoption of a mediating response (e.g., waiting by the correct lever) to solve the DMTP task. Because the scheduled delay interval was unknown to the rat, the nose-poke response was maintained on a second-order variable-interval schedule that sustained a high rate of responding at the central panel (M = 1.11, SD = 0.42, responses per second) throughout all delay intervals.
Procedure The rats were tested in daily 40-min sessions. The procedures for initial habituation to the chambers, training to lever press, and training on the matching task in the absence of delay intervals, before training on the DMTP task itself, have been described fully elsewhere (Dunnett, 1985; Dunnett, Evenden, & Iversen, 1988). The rats received 30 days training on the task to establish efficient levels of performance in all rats before the present experiments began. Performance stabilized at 120-150 trials completed on each day (i.e., approximately 20 trials at each scheduled delay), and data were collapsed over several days (minimum of 4) for each experimental manipulation. The basic measure of performance was percentage correct responses at each delay interval. The rats were initially tested with scheduled delays randomly selected from the set: 0, 2, 4, 8, 12, 16, and 24 s. To detect proactive interference effects, the computer was programmed to collect data separately for trials in which the sample on the previous trial was on the same or on the opposite side to that on the current trial Variations of the schedule were conducted on probe test days. First, to determine whether samples before the previous sample would also interfere with performance on the current trial, trials were categorized according to whether the sample on the previous trial had been on the opposite side (i.e., the present sample comprises a run length of 1) or the number of preceding trials in which the sample had been on the same side as the present trial (run lengths of 2-4).
657
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
PROACTIVE INTERFERENCE IN RATS On those occasions that the four preceding trials had all been on the same side, the current sample was not selected randomly but was forced to the other side. Because the variation of delay intervals from trial to trial would seriously confound this analysis, the effect of run lengths was assessed on probe days in which all trials had the same delay of 0, 4, 12, or 24 s. Each of these four fixed delays was tested on 4 separate test days in cyclic order with 1 intervening baseline day between each, over 32 days. The side of the sample was determined randomly on each trial, and the effects of runs of up to four consecutive trials with the sample on the same side were recorded. Second, to assess whether proactive interference from the previous trial might be reduced by increasing the ITI, the rats were tested on the standard DMTP schedule with ITIs of 5,15, or 45 s. The sessions with longer ITIs were increased in duration (to 50 min for a 15-s ITI; to 60 min for a 45-s ITI) to enable similar numbers of trials to be completed on each day. The animals were tested four times on each of the alternate ITIs in cyclic order on consecutive days. Third, to determine whether the side of the animals' response on the previous trial may have a greater effect than the side of the sample lever per se, the programs were modified to collect not only the side of the previous sample but also to determine whether the response was correct. The task was also made more difficult by increasing the scheduled delays to 0, 4, 8, 12, 18, 24, or 32 s. Because the rats were performing at 90-95% correct overall by this stage, the proactive effects of the previous sample and previous response were generally concordant on the great majority of trials. To collect a sufficient number of the critical incorrect trials (in which the sample alone or the response alone would be on the side opposite to the present sample) for estimation of percentage correct performance for each rat at each delay, data were collapsed over a further 17 days of testing.
(a) Previous sample 100 I
90
80
O — O previous sample same • — • previous sample opposite
12
16
(b) Runs of samples on same side 100 T
90-
80-
• 0 V----V A— A O O
run length run length run length run length
1 2 3 4
Drug Effects 70
The effects of physostigmine salicylate (Addenbrookes Hospital, Cambridge, UK; 0.0, 0.03, 0.06, or 0.1 mg/kg ip), nicotine sulfate (Sigma Chemical Co., Poole, Dorset, UK; 0.0, 0.03, 0.1, or 0.3 mg/ kg sc), and clonidine hydrochloride (Sigma; 0.0, 0.003, 0.01, or 0.1 mg/kg ip) on the DMTP schedule were assessed using the set of longer (0-32 s) delays. Each drug was diluted or dissolved in isotonic saline and administered 5 min before the test in a series of increasing doses on consecutive days. The series of injections was repeated four times for each drug, with 2 or 3 baseline days between each cycle.
12
(c) Increasing intertrial intervals 100T
90--
0 — O 5s ITI previous same
Results All animals learned to perform the DMTP task efficiently, although 2 animals were removed from later tests because of respiratory infections.
Side of Previous Sample The effect of the previous sample on the percentage correct at each delay over Test Days 30-40 is shown in Figure la. As described previously, the rats showed a decline in performance at longer delays, indicating a delay-dependent for-
Figure 1. a: Proactive interference in the delayed matching-to-position task. (Performance accuracy [percentage correct responses] at each trial delay separately for trials in which the sample on the previous trial was on the same or the opposite side to the sample
24
•— • 80--•
A— A A— A V— V T— V
"~~—-_.0
5s ITI previous opposite 15s ITI previous same 15s ITI previous opposite 45s ITI previous same 45s ITI previous opposite
I
70
12
18
24
32
delay (sec)
on the current trial, b: Performance accuracy at four trial delays, tested on separate sessions, for trials in which the previous trial was on the opposite side [run length = 1], or trials in which the most recent sample on the opposite side was two, three, or four trials previously [run lengths = 2-4, respectively], c: Effect of intertrial interval on proactive interference. Performance accuracy at each trial delay separately for trials in which the sample on the previous trial was on the same or the opposite side to the sample on the current trial, and for sessions with 5-, 15-, or 45-s intertrial intervals. Vertical bars indicate two SEs derived from the relevant interaction terms of the analyses of variance.)
658
STEPHEN B. DUNNETT AND FRAN L. MARTEL
getting from short-term memory. Moreover, performance at the longer delays was worse when the previous sample had been on the opposite side than when it had been on the same side as the current trial, indicating a proactive interference effect. All effects were highly significant: Delay, F(6, 126) = 35.33; Previous Trial, F(l, 21) = 16.86; Delay x Previous Trial, F(6, 126) = 4.38; allps < .001.
(a) Previous sample and response
100 T
w
80-
•— • O—O D— D •—•
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
60 ••
SAMPLE
RESPONSE
same same opposite opposite
same opposite same opposite
Runs of Preceding Samples O
40
0
4
8
12
18
24
32
(b) Previous response
100
Effect oflntertrial Intervals
90--
80-
—* previous response same — 4 previous response opposite
70 0
8
12
18
24
32
(c) Previous sample 100 T
As shown in Figure Ib, rats performed worse at all longer delays on trials in which the previous sample had been on the opposite side (defined as a run length of 1) than on trials in which the previous sample was on the same side. However, preceding trials had no further influence: The trials in which the most recent opposite sample was 2, 3, or 4 trials previously (run lengths = 2, 3, or 4, respectively) did not differ. All effects were significant: run length, F(3, 63) = 12.62, p < .001; delay, F(3, 63) = 106.09, p < .001; Run Length x Delay, F(9, 189) = 2.12, p < .05.
•
As shown in Figure Ic, proactive interference from the previous trial was abolished with the longer 15- and 45-s ITIs. In particular, performance was disrupted only in the case in which the previous sample had been on the opposite side and the ITI was short. Performance on trials with short ITIs in which the previous sample had been the same did not differ from performance on either type of trial on the test days with longer ITIs. All main effects were significant: ITI, F(2, 42) = 9.56, p < .001; previous trial, F(l, 21) = 4.96, p < .05; delay, F(6,126) = 68.35,p < .001; the only significant interaction was ITI x Previous Trial, F(2, 42) = 6.16, p < .001. Effect of Previous Response
90-
80
• — • previous sample same • — • previous sample opposite
70
12
18
24
32
delay (sec) Figure 2. Proactive effects analyzed in terms of the previous sample and the previous response, (a: Trials subdivided into whether the previous sample [determined randomly] had been on the same or opposite side and in terms of whether the response was correct [previous sample and response both on the same or on the opposite sides] or was incorrect [previous sample same and previous response opposite, or vice versa]. Note that correct trials occurred 8-10 times more frequently than incorrect trials, b: Data of Panel a collapsed in terms of the side of the previous response, c: Data of Panel a collapsed in terms of the side of the previous sample. Note that proactive interference from the previous trial is more clearly ap-
The separate effects of previous sample and response are shown in Figure 2a. Consideration of the trials in which a correct matching response had been made on the previous trial again revealed a proactive interference effect (i.e., performance was better on trials when both the previous sample and the previous response had been on the same lever than when they had both been on the opposite lever). However, the greatest disruption of performance was on trials in which an erroneous response had been made on the previous trial to the opposite lever after presentation of the sample on the same side as on the present trial, resulting in a significant main effect of previous sample, F(l, 21) = 8.45, p < .01, that is, in the direction opposite from that observed in previous analyses in which all trials had equal weight. The main effect of previous response was also highly significant, F(l,
parent when the data are analyzed in terms of the response rather than the sample on the previous trial. The vertical bar indicates two SEs derived from the three-way interaction of the analysis of variance.)
659
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
PROACTIVE INTERFERENCE IN RATS
21) = 51.88, p < .001 (i.e., the animals performed better on trials in which the previous response had been to the same lever), and the Previous Sample x Response interaction was also significant, F(l, 21) = 34.57, p < .001. The Previous Response x Delay interaction, but not the Previous Sample x Delay interaction, was also significant, F(6, 126) = 5.92, p < .001, and F(6, 126) = 2.09, p = .059, respectively. Because the animals achieved 85-95% correct choices overall, the side of the previous sample and the previous response acted in concert on the great majority of trials. However, the pattern of results observed when these two factors were considered independently suggests that proactive effects from the previous trial are better described in terms of interference from the previous response than from the previous sample. To illustrate this, the same data were reanalyzed separately in terms of trials in which the previous sample was on the same or opposite side as the present sample (as previously) or in which the previous response was on the same or the opposite side. Proactive effects, as indicated by separation of performance on previous-same and previous-opposite trials, were much greater when considered in terms of the previous response (see Figure 2b), previous response, F ( l , 21) = 44.92; Previous Response x Delay, F(6, 126) = 6.43; bothps < .001, than when considered in terms of previous sample (see Figure 2c), previous sample, F ( l , 21) = 19.21, p < .001; Previous Sample x Delay, F(6, 126) = 2.03, p = 0.066. Inspection of trials in which the previous response was made to the opposite side indicated that performance was higher when the previous responses was correct to an opposite sample than when it was in error to a same sample. This suggests that the previous sample per se has no direct influence on proactive interference. Consequently, subsequent analyses were conducted in terms of the influence of the response, rather than of the sample, on the previous trial. Effect of Physostigmine The effects of injections of saline or three doses of physostigmine are shown in Figure 3. There was a significant main effect of drug dose, F(3, 63) = 5.56, p < .01, which was attributable to performance under the two higher doses of physostigmine that was moderately (but significantly) below that under saline: Newman-Keuls tests, saline versus 0.1 mg/kg, t(4, 63) = 4.90, p < .01; saline versus 0.06 mg/kg, t(3,63) = 3.54,p < .05. There were no significant interactions between drug dose and previous response or delay. Thus, at the doses used, physostigmine appears to induce a mild generalized disturbance of performance in DMTP. Effect of Nicotine The effects of injections of saline or three doses of nicotine are shown in Figure 4. There was a significant main effect of drug dose, F(3, 63) = 206.77, p < .001, which was attributable to performance under the two higher doses of nicotine that was substantially below that under saline: NewmanKeuls tests, saline versus 0.3 mg/kg, t(4, 63) = 31.04; saline versus 0.1 mg/kg, t(3, 63) = 13.20; bothps < .01. The dis-
ruption of performance induced by nicotine was most marked on trials in which the previous response was on the side opposite to the present sample, whereas even high doses had little effect on trials in which the previous response was on the same side (see Figure 4). All drug interactions were highly significant: Drug x Previous Response, F(3, 63) = 30.03; Drug x Delay, F( 18,378) = 9.23; Drug x Previous Response x Delay, F(IS, 378) = 2.87; allps < .001. Thus, at the doses used, nicotine appears to induce a marked disturbance of DMTP performance, in particular by increasing the effects of proactive interference from the response made on the previous trial. Effect
ofClonidine
The effects of injections of saline or three doses of clonidine are shown in Figure 5. There was a significant main effect of drug dose, F(3, 57) = 80.13, p < .001, which was attributable to a disruption of performance under the highest dose of clonidine alone: Newman-Keuls tests, saline versus 0.1 mg/ kg, t(4, 57) = 17.38, p < .01. The Dose x Delay interaction was highly significant, F(18, 342) = 10.60, p < .001, but the Dose x Previous Response and the three-way interactions were not significant. This suggests that the highest dose of clonidine has no effect on the proactive interference from the previous response per se but induced a marked, delay-dependent impairment of performance irrespective of responses on previous trials. Drug Effects on Response Rate The mean numbers of trials completed in each 40-min test session were compared at each dose of the three drugs to provide an indication of nonspecific effects of these treatments. As shown in Figure 6, the higher doses of all three compounds induced a significant decrease in the number of trials completed: physostigmine, F(3, 63) = 26.08; nicotine, F(3, 63) = 56.72; clonidine, F(3, 57) = 177.49; all/?s < .001. Other response rate measures, such as the rate of nose poking at the central panel during the delay intervals, showed a similar decline in the cases of the higher doses of all three drugs (data not shown).
Discussion The present observations provide clear evidence that the response made on the previous trial can interfere with choice accuracy of rats trained in the DMTP task. The results are compatible in several respects with previous descriptions of proactive interference effects in animals' short-term memory performance, which has been studied most systematically in pigeons trained on delayed matching- and nonmatching-tosample tasks. Thus, for example, the effects were most apparent at short ITIs, affected trials with long but not short delays, and carried over from the previous trial but did not accumulate over longer series of trials (Edhouse & White, 1988; Maki et al., 1977; Moise, 1976; Nelson & Wasserman, 1978; Roberts, 1980; Roitblat & Harley, 1988). These observations are based on comparisons of previous-sample-
660
STEPHEN B. DUNNETT AND FRAN L. MARTEL
(a) Saline
(b) Physostigmine 0.03 mg/kg
lOO-i
100, g
«. " O. ,
90o o
*?
O - - - O Previous response same
„
0
4
8
12
18
24
32
0
4
8
12
18
(c) Physostigmine 0.06 mg/kg
(d) Physostigmine 0.10 mg/kg
100
100
T
a.
"°
o
24
32
24
32
o..
90 -
•-0
90-
80-
80-
70
70
8
12
18
24
32
4
8
12
18
(e) Previous response opposite
(f) Previous response opposite
1UO-
10
?*
