BEHAVIORAL AND NEURAL BIOLOGY 58, 171--179 (1992)
Strategy Selection in a Task with Spatial and Nonspatial Components: Effects of Fimbria-Fornix Lesions in Rats M. M'HARZI Laboratoire de Psychophysiologie, Universite Paris 7, 75251 Paris Cedex 05, France AND
LEONARD E. JARRARD 1 Department of Psychology, Washington and Lee University, Lexington, Virginia 24450
INTRODUCTION
The main purpose of the present research was to investigate the ability of rats to learn a 12-arm radial maze task that requires the concurrent utilization of both spatial and intramaze cue information. The task involves in a single trial both place and cue learning as well as reference memory (RM) and working memory (WM). Since the animal can choose place and cue arms in any order, the strategies employed to learn the task can be studied as well as the kinds of memory errors that are made. The results of Experiment 1 showed that the number of errors made on the place and cue components of the task did not differ, and that more RM than WM errors were made early during learning. As the task was learned, the animals tended to choose the place arms before choosing the intramaze cue arms, thus suggesting that a spatial strategy was employed first followed by a cue strategy. In Experiment 2 lesions of the fimbria-fornix resulted in temporary impairments in both RM and WM that were especially apparent on the spatial component of the task. The lesioned rats also switched from choosing mostly place arms early during the trial to choosing more cue arms. While fimbria-fornix lesioned rats recovered from the memory impairments with training, the change in response strategy persisted throughout postoperative testing. The procedure of combining both spatial and nonspatial components concurrently in the same task should prove of value in studying response strategies in animals.
The radial arm maze has proved to be an especially useful tool in the study of spatial m em o ry in rats (Buresova & Bures, 1982; Olton & Samuelson, 1976). In the radial maze experi m ent performance depends not only on the r a t r e m e m b e r i n g the general procedures of the t ask from day to day, b u t also which arms have been visited within a trial: the underl yi ng m e m o r y processes have been referred to as reference m e m o r y (RM) and working m e m o ry (WM), respectively (Honig, 1978). The procedure t h a t is used most often involves employing an 8arm radial maze, placing the reward at the end of each of the 8 arms, and leaving the r a t on the maze until all 8 reinforcements are obtained (see Olton & Samuelson, 1976). One problem in i nt er p re tin g impaired performance following drug or brain manipulations is t h a t repeated entries into arms already chosen on t h a t trial can reflect either a general m e m o r y dysfunction, a problem in remembering general procedures required to carry out the task (RM), or impaired ability to r e m e m b e r which arms have already been visited within t h a t trial
(WM). In an a t t e m p t to det erm i ne more precisely the n a t u r e of m e m o r y impairments, several investigators have employed the radial maze and a limited baiting procedure, e.g., fewer t h a n the total n u m b e r of arms are consistently baited over trials (Jarrard, 1978a; Olton & Papas, 1979). Using this procedure, RM errors are operationally defined as choices of arms t h a t are never baited, while WM errors are defined as repeated entries into baited and unbaited
© 1992 Academic Press, Inc.
l This research was carried out while M.M. was a visiting scientist at Washington and Lee University, Lexington, Virginia. The research was supported by National Science Foundation G r a n t s BNS 8507259 and 8809208 to L.E.J. Address r e p r i n t requests to M. M'Harzi, Centre de Recherches Roussel-UCLAF, Pharmacologie-Neurobiology, 111, route de Noisy, Nocard, RC, 93230 Romainville, France. 171
0163-1047/92 $5.00 Copyright © 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.
172
M'HARZI AND JARRARD
arms that had already been visited within the trial. WM errors can be divided further into repeated entries into baited arms (working memory correct or WM-C) and reentries into arms that have never been baited (working memory incorrect or WM-I). A further refinement in procedure involves employing the radial maze to study cue as well as spatial learning and memory (Nadel & MacDonald, 1980; Jarrard, 1983). In the experiment by J a r r a r d an 8-arm radial maze and a within-subjects design were used, together with a procedure that permitted determining two kinds of learning (place and cue) and two memory functions (WM and RM). In the place task the rat learned to choose 4 out of 8 similar arms, where all arms remained in the same spatial location from trial to trial. In the intramaze cue task, different textured floor inserts in the 8 arms were moved in a random order from trial to trial, and the rat was rewarded for choosing the same 4 cues. This more complicated testing procedure designed to compare place and cue learning has been employed in several experiments with some interesting differential effects resulting from brain manipulations (see Jarrard, 1983, 1986). In the research reported here, we describe an extension of the basic radial maze procedure that is designed to obtain information regarding the strategies used by rats in solving a combined spatial and cue radial maze task. Using a 12-arm maze, 6 of the arms were designated place arms and 6 were designated intramaze cue arms. Three of the 6 place arms were consistently baited from trial to trial thus permitting one to look at place WM and RM errors. The 6 cue arms contained different textured floor inserts that were moved in a random order from trial to trial; 3 of the 6 cues were consistently baited over trials permitting a determination of cue WM and RM errors. Since in performing the task the rat can choose to visit the arms in any order, it is possible to study the different strategies that were employed. Experiment 1 was carried out to see if rats can learn this more complicated version of the radial maze task, and to see whether the data indicate that different strategies are employed as the task is learned. It is well known that damage to different components of the hippocampal formation results in impaired acquisition of spatial radial mazes (Becker, Walker, & Olton, 1980; Bouffard & Jarrard, 1989; Jarrard, 1978a,b; J a r r a r d & Elmes, 1982; Olton, 1983; Olton, Becker, & Handelman, 1979). In order to determine the usefulness of the testing procedure described above, and to better understand the nature of the impairment found in rats following dam-
age to the septohippocampal system, rats that learned the task in Experiment 1 were retested in Experiment 2 following damage to the fimbria-fornix (FiFx). EXPERIMENT 1 The purpose of the first experiment was to see if the rats can learn the combined place-cue task described in the Introduction. Performance of the place component of the task should require the animals to use distal room cues to choose the 3 arms that are consistently baited and avoid the 3 arms that are never baited. The salient stimuli for the cue component of the task are the specific characteristics of the intramaze cues; however, the location of these cues within the 6 cue arms is changed over trials. We were interested in obtaining answers to several questions. If the task is learned, will the place and cue components be learned at different rates? Will the order in which the arms are chosen indicate a preference for either the place or cue arms?
Method Subjects. Fifteen male, albino, S p r a g u e - D a w l e y rats weighing 250-300 g were used. They were maintained under a 12:12 h light/dark cycle (07001900) at a temperature of 22°C (+_ 2 °) and were housed in individual cages for the duration of the experiment. Apparatus. The animals were trained on a 12arm radial maze, constructed of 0.019 m varnished wood, and elevated 0.65 m above the floor. Each arm (0.66 x 0.10 m) radiated from a circular central platform 0.76 m in diameter. Clear Plexiglas walls (6.5 cm high) extended the length of each arm. Small holes (1.2 cm in diameter and 0.5 cm deep) at the end of each arm served as wells for the sucrose solution. The apparatus was located in a room that was well lighted and sound attenuated. Many extramaze cues were present on the walls and around the room. Six removable inserts 0.6 x 0.09 m were constructed by gluing to a plywood backing different materials (carpet, cloth, chicken wire, ceiling tile, Plexiglas, screen). These intramaze cues were located in arms (1, 3, 5, 8, 10, and 12) and served to test for cue learning (the cue component) (see Fig. 1). The 6 inserts were moved pseudorandomly from trial to trial within the 6 arms that were designed for cue learning. The remaining 6 similar arms (2, 4, 6, 7, 9, and 11) served to test learning of the place task (the place component). Three of the 6 place arms and 3 of the 6 cues were baited on each trial.
STRATEGY SELECTION IN RATS
o]
FIG. 1. Schematic drawing of the 12-arm radial maze. The arms are numbered from 1 to 12. Place arms are identified by blank arms and cue arms are shown as containing different patterned stimuli.
Four place-cue combinations were used with 4 squads of animals. Subjects in a given squad were always reinforced for choosing the same 3 places and the same 3 cues, even though the cues changed locations over trials.
Procedure. Before testing began, the subjects were given five successive sessions of handling (one per day). During this time, 5 to 10 ml of 25% sucrose (to be used as reward) was made available in the home cage of each rat. Amounts of food and water were limited to 12-15 g and 2 0 - 2 5 ml, respectively. These amounts were later adjusted for each rat so that body weight was maintained at 80-85% of the predeprivation level. Following handling, each rat was placed for 5 min each day, for 5 days on the central platform. During this time, sucrose was available in a small glass dish placed on the central platform and the animal could explore the maze. Food and water were given to each animal 1 h after being returned to the home cage. Behavioral testing took place during the light cycle (0800-1400). Rats were given 60 trials, 2 each day with an intertrial interval of approximately 2 h. On each trial, wells at the end of correct arms were baited with 0.5 ml of sucrose and the rat was placed on the central platform. A trial was continued until either all 6 reinforcements had been obtained (3 place and 3 cue), until the animal had made 24
173
choices (correct and incorrect) or until 8 min had elapsed. A correct choice was counted when the rat entered an arm that contained the reinforcement. Incorrect choices were recorded as follows: choices of arms or cues that were never baited were counted as reference memory (RM) errors; reentering previously baited arms or cues was considered as working memory correct (WM-C) errors; and reentering never baited arms or cues was recorded as working memory incorrect (WM-I) errors. In some analyses, WM (WM-C + WM-I) was used because too few WM-I errors were made for analysis. Response strategies employed by the rats in learning the task were analyzed by studying the order in which the place and cue arms were chosen. In addition to the first choice made (place (P) or cue (C)), and the first correct choice, the analysis included the number of trials where the 3 correct P arms were chosen before a cue arm ( P P P C . . . ) , and number of trials where the 3 correct C arms were chosen before a place arm ( C C C P . . . ) . A mixed strategy was operationally defined as a trial where an animal made a mixed sequence of P and C choices (e.g., PCCPPC etc., with the place component of the task being completed first; or CPCCPP etc., with the cue task being completed first). Statistical analysis was performed using mixed design analysis of variance (ANOVA) with subsequent pairwise comparisons being made with the N e w m a n - K e u l s test. Paired and unpaired t test were also employed.
Results and Discussion Figure 2 shows the mean number of RM, WM-C, and WM-I errors made by the rats (N = 15) on the place and cue components of the task during acquisition. A 2 x 3 x 6 mixed design ANOVA (Place-Cue x RM, WM-C, WM-I errors x Blocks of trials) was used to analyze these data. The overall analysis indicated that the total number of errors made on the place and cue components of the task did not differ, F < 1.0. However, there were significant differences in types of memory errors that were made, F(2, 28) = 200.75, p < .01. In addition, the interaction of Types of memory errors x Blocks of trials was significant, F(10, 140) = 27.10, p < .01. F u r t h e r analyses revealed that the animals made significantly more RM than WM-C errors through the first five blocks of trials (p < .01), more RM than WM-I errors through all six blocks (p < .01), and more WM-C than WM-I errors through the first three blocks of trials (p < .01). The interaction
174
M'HARZI AND JARRARD
25
20
15
o ~10 Q ©
=
5
2: ©
0
Blocks of 10 Trials FIG. 2. Mean n u m b e r of errors over blocks of trials during acquisition. Errors are identified as RM, reference memory; WMC, working memory correct; WM-I, working memory incorrect.
of P l a c e - C u e × Types of memory errors was also significant, F(2, 28) = 12.29, p < .01. Thus, the animals made significantly more WM-C errors on the place than the cue component of the task (p < .01), and conversely more RM errors on the cue than the place component of the task (p < .01). There was an overall decrease in number of errors over the six blocks of trials, F(5, 70) = 58.06, p < .01. The triple interaction of P l a c e - C u e x Types of memory errors × Blocks of trials was not significant. Although the animals made approximately the same number of total errors on the place and cue components of the task, the order in which the arms were chosen changed as the task was learned. The nature of these changes in preference can be seen in Fig. 3. Considering only the first choice that was made, the animals displayed an increased preference for place arms over blocks of trials beginning with 48.6% (a level expected by chance) on Block 1 to 93.3% on Block 6. The percentage of the trials on which the first correct arm was a place arm increased from 58.0% on Block 1 to 94.0% on Block 6. As a result, preference for the cue component decreased over blocks of trials (from 51.4 to 6.7%). As shown in Fig. 3, the animals also exhibited a decrease in adopting a mixed strategy, e.g., a mixed
sequence of place and cue choices (from 98.0 to 70.0%). Considering only the trials on which choices were mixed place and cue arms and the place component was resolved first, there was an increase from 60 to 86% on Blocks 1 and 6, respectively. There was an increase in completing all 3 "place before any cue" (PPPC) choice strategies (1.3 to 30.0%); however, choosing "cue before any place" (CCCP) choice strategy was almost never used by the animals (0.7% on Block 1 and 0.0% on Block 6). Even though the combined place and cue task is complex, the results clearly show that rats are able to learn to select the baited cue and place arms. If one looks only at total number of errors that were made during acquisition, the place and cue components appear to be of equal difficulty. However, a consideration of the order in which the arms were chosen shows that the rats did not respond to the place and cue components equally. Rather, as learning progressed there was a decided increase in choosing the place arms first, thus indicating a preference for the place arms. These results strongly suggest that the use of distal cues (the relevant stimuli for the place component) were easier for the rats than the proximal, intramaze cues (the relevant
Choice Stategy During Acquisition 100
80
.~ 60
,,"' q'~ o
I-O-P as 1st Correct Choice I "El"- PPPO... I-A- cccP...
0'"
g4o ~
2O
0
Q L)
0 1
2
3
4
5
6
Blocks of 10 Trials FIG. 3. Choice strategy d u r i n g acquisition. Abbreviations used: P, Place; P P P C . . . , the 3 correct place arms were visited before choosing a cue arm; C C C P . . . , the 3 correct cue arms were visited before choosing a place arm; Mixed, mixed place and cue arms.
STRATEGY SELECTION IN RATS stimuli for the cue component that changed location from trial to trial). Thus, as the task was learned the rats appeared to first employ a spatial strategy before then using a nonspatial, cue strategy. EXPERIMENT 2 Experiment 2 was designed to investigate the effects of lesions of the FiFx on retention of the task that had been learned in Experiment 1. In previous research the first author had found that damage limited to the most dorsal aspects of the fornix and fimbria both abolished hippocampal theta (M'Harzi & Monmaur, 1985) and resulted in impaired performance on a spatial task (M'Harzi, Palacios, Monmaur, Willig, Houcine, & Delacour, 1987). Similar small lesions of FiFx were used in Experiment 2 in order to see if performance of the spatial and cue components of the 12-arm radial maze task would be equally affected, and whether the predominant place strategy employed by the rats in performing the task would change.
Method Following training in Experiment 1, the rats were divided into two groups that were matched using performance measures obtained during original acquisition. One group received the FiFx lesion (FiFx Group), and the other served as a control (Co Group). The rats were anesthetized using a mixture of sodium pentobarbitol and chloral hydrate. Lesions were produced electrolytically by passing a dc cathode current (2 mA for 20 s) through a stainless steel electrode. The lesion electrode was insulated except for the tip. The stereotaxic coordinates were as follows: 0.7 mm posterior to bregma; ML = + 1.0 mm, DV = - 4 . 1 mm; ML = +1.5 mm, DV = - 4 . 4 mm. The surface of the skull was used to determine DV coordinates. In operated control rats the electrode penetrations were similar to those in the FiFx Group except the electrode was lowered 1.0 mm above the target structure and no current was passed. Postoperative testing was begun 10 days after surgery. The behavioral procedures were similar to those used during acquisition in Experiment 1. The animals were given 2 trials on each of 10 days by which time performance had returned to the preoperative level. At the completion of the experiment the animals were deeply anesthetized and transcardially perfused with saline and 10% neutralized formalin. The brains were removed and stored in neutralized for-
175
malin for several days. The brains were then frozen, cut at 40 t~m in the coronal plane, and sections showing lesions and/or electrode tracts were stained with a cresyl violet stain. Preliminary analysis of the preoperative data for the FiFx and Co Groups showed that the groups did not differ in total errors or responding pattern prior to the surgery. In the analyses of the postoperative data WM-C and WM-I errors were summed to form WM errors since few reentries into never-baited arms or cues were made.
Results and Discussion Behavior. Animals with FiFx lesions were significantly impaired in performance early during postoperative testing but they rapidly recovered and testing was discontinued after 20 trials. The pattern of errors that were made is shown in Fig. 4. A 2 x 2 x 2 x 4 mixed design ANOVA was used to analyze the data with Groups as the between-subjects variable and Place-Cue, R M - W M , and 4 Blocks of 5 trials as within-subjects variables. The overall analysis indicated that animals with FiFx lesions made significantly more total errors than Co anireals, F(1, 33) = 6.15, p < .05. There was also a significant interaction of Groups x Blocks of trials, F(3, 33) = 3.58, p < .05, showing that performance of the groups was differentially affected over trials. A more detailed analysis revealed that FiFx rats made significantly more errors than Co rats on Blocks 1 and 2 (p < .01), but that the groups were similar by the third block of trials. The interaction of Groups x P l a c e - C u e was not significant; however, the triple interaction of Groups x P l a c e - C u e x Type of memory error was significant, F(1, 11) = 4.95, p < .05. FiFx rats made more RM and WM errors on the place component and more WM errors on the cue component (all p's < .05). The FiFx animals did not make more RM errors than controls on the cue task. Thus, damage to the FiFx resulted in a general impairment on the place task (both RM and WM) but only a WM impairment on the cue task. Finally, the interaction of Groups x Types of memory errors x Blocks of trials was significant, F(3, 33) = 3.00, p < .05. FiFx rats made more RM errors than Co rats on Blocks 1 and 2, and more WM errors on Blocks 1, 2, and 3 (all p's < .05). By the fourth block of trials the groups did not differ on any of the measures. Analysis of the order in which the place and cue arms were chosen indicated a change in preference that resulted from interrupting the axons in FiFx. There was general agreement among the various
176
M'HARZI AND JARRARD
measures; however, only the data for the mixed trials on which the place component was completed before the cue component are presented here (see Fig. 5). Specifically, the analysis indicated that Co animals completed the place component significantly more often than FiFx rats, F(1, 11) = 15.26, p < .01. There was a significant increase in the percentage of trials completed over blocks of trials, F(3, 33) = 3.9, p < 0.02, but the Groups × Blocks interaction was not significant, F < 1.0. These findings show that partial damage to the FiFx not only resulted in temporary impairments on both place and cue components of the task, but also that the lesions altered the choice strategies of the damaged animals.
~'~100 Postoperative Choice Strategy
~o 9 0 []
~80
~70 o 060
I
~50
Histology.
Pre
Histological examination of the brains indicated that in two rats the damage was not acceptable, and the data for these animals were not included in the final analyses. In all animals that were included, the lesions principally destroyed the dorsal aspect of the Fi and the dorsal Fx (see Fig. 6). The center of the lesion was located at the level shown in Plate 17 of the Paxinos and Watson (1982) atlas. In some animals the ventral part of the fimbria was slightly damaged but in none of the cases was the ventral hippocampal commissure in-
1
2
4
FIG. 5. Choice strategy (mixed trials with the place component being completed first) of FiFx and Co Groups during postoperative testing.
terrupted. In most rats the lesions extended rostrally to the septofimbrial nucleus and the nucleus of the lateral septum and caudally to the septal pole of the dorsal hippocampus. In some animals the lesions included a small amount of damage to the
PRM
8
3
Blocks of 5 Trials
PWM
6 4 2
[]
0
,
1 ~.
GLM FiFx (n = 7)
[]
Co
8" o
6"
!
(n = 6)
4 2; 2 []
0
, Pre
I
I
I
i
1 2 3 4 Blocks of Five Trials
~
I
I
l
I
Pre
1
2
3
4
FIG. 4. Mean number of memory errors on place and cue arms made by the FiFx and Co Groups on the last block of acquisition trials (Pre) and during the 4 blocks of postoperative testing. Errors are identified as CRM, cue reference memory; CWM, cue working memory; PRM, Place reference memory; PWM, Place working memory errors.
STRATEGY
SELECTION
177
IN RATS
•:..~...X,.:...~ ,~ ..,-..
k*~',.: "
"
" " ":','.'" - .
.
.
.
" ~'
? ":,'
;"
-
, ~'t
.~.
FIG.
6.
Photomicrograph
"-"-
-:"" :~C.--.~_."
~2-'~ , -
-'; ~' " 2 , "
-
£ "~," --_
with
violet from
,,.*.
-'~-,
of a coronal
section
stained
'ventral part of the overlying corpus callosum. As was intended, the damage to FiFx was similar in amount and location to that reported in previous research (M'Harzi & Monmaur, 1985; M'Harzi et al., 1987). GENERAL DISCUSSION The main purpose of the present research was to compare and contrast the utilization of concurrently presented spatial and nonspatial information in the rat. The results indicated that rats can learn a complex task that has a spatial component presumably requiring the use of distal extramaze cues and a separate nonspatial component requiring the use of intramaze cues. Since there were no differences in total number of errors that were made in learning the place and cue components of the task, one might be tempted to conclude that the two were treated similarily by the rats; however, analysis of the pattern of errors (RM, WM) and the order in which the arms were chosen shows that the animals did not respond to the place and cue components in the same way. Rather, as learning progressed, there was a decided preference for the place arms. This preference for place was demonstrated by choosing the place arms before choosing the arms containing the cues (see Fig. 3). Thus, as the animals learned the task, the sequence of choices within a trial changed from equal choices of place and cue arms to where
cresly
a representative
animal
in the FiFx
Group.
a place strategy was adopted on early choices before then using a cue strategy. In a previous experiment by the second author, place and cue learning were tested separately in the same animals using an 8-arm radial maze with different textured floor inserts being placed in the arms during cue testing (and cues changing locations over trials) and with the inserts being removed for the place task (Jarrard, Okaichi, Steward, & Goldschmidt, 1984). While in that study unoperated animals made more total errors in learning the cue task as compared to the place task, RM and WM errors were equally distributed. In the present more direct comparison of place and cue learning, total errors did not differ for place and cue but there were differential effects in terms of RM and WM errors. Specifically, the rats made more RM errors on the cue component than the place component, e.g., there were significantly more choices of never baited cues. While it is possible that the greater number of cue RM errors could reflect increased interference resulting from a tendency for the rats to return to specific arms (places) that had been reinforced on previous trials rather than being influenced by the intramaze cues, inspection of the data failed to support this view. In contrast with the greater number of cue RM errors made by the rats in learning the task, more WM errors were made on the place component, e.g., the rats repeatedly entered more correct, previously
178
M'HARZI AND JARRARD
baited place arms that had already been visited during a trial. Since the cues remained on the same arms during a trial, both cues and places could be used as discriminative stimuli for the working memory component. However, the fact that fewer WM errors were made on the cue than the place component makes it less likely that the rats were using spatial location in performing the WM component of the cue task. The underlying explanation for the above differential effects is not readily apparent, but if one assumes that number of errors reflects task difficulty, the results suggest that the rats had more trouble remembering correct from incorrect cues over days, and which place arms had already been visited during a trial. While there were these differential effects for choice accuracy, the sequence of choices indicated that as learning progressed the rats tended to choose the place arms before the cure arms. The small lesions of FiFx resulted in a temporary impairment on both place and cue components of the task. In terms of the types of memory errors that were made, the FiFx animals exhibited a general impairment on the place task with significant increases in both RM and WM errors, while only WM was affected on the cue task. This pattern of results is similar to that found in the study described above with place and cue learning and large lesions of the FiFx (see Jarrard et al., 1984). In that experiment the FiFx was completely interrupted and the impairments persisted throughout postoperative testing, while in the present study performance of rats with the more limited lesions did recover. Thus, by using more limited damage to FiFx we were able to look at preferences for place as compared to cue arms when the groups did not differ in terms of overall choice accuracy. As pointed out in the Introduction, it had been established in previous research that limited lesions similar to those used in the present experiment served to disrupt hippocampal theta and impair original acquisition of a complex spatial task (M'Harzi & Monmaur, 1985; M'Harzi et ~1, 1987). Of special interest in the present study was the change in the order that the place and cue arms were visited following damage to the FiFx. Compared to control animals and performance before the operations, the lesioned rats switched from visiting primarily place arms on early choices within a trial to choosing more cue arms. Such a change in strategy following damage to the septohippocampal system would follow from the "spatial mapping" theory of hippocampal function proposed by O'Keefe and Nadel (1978). If the hippocampus does form the
neural basis for spatial mapping as these authors propose, damage to the axons in FiFx should result in the animals experiencing more difficulty with the spatial arms. Even though with practice performance recovered in terms of choice accuracy, the FiFx animals continued to show a decreased preference for the place arms. A number of investigators have pointed out the similarities in the effects on behavior that result from damage to the septohippocampal system and blocking the cholinergic systems with drugs (see Hagan & Morris, 1987). Morris, Tweedie, Schenk, and Jarrard (1990) recently reported that in the Morris swimming task rats with the hippocampus removed swim rapidly in large circles often around the edge of the pool; similar effects have been described by Whishaw and Petrie (1988) following cholinergic blockade with atropine. Of more direct relevance for the present research are reports that following cholinergic blockade rats are more impaired in acquiring locale strategies than taxon strategies (Buresova, Bolhuis, & Bures, 1986; Hagan, Tweedie, & Morris, 1986; Whishaw, 1985). The motor response requirements for performing the place and cue components of the 12-arm radial maze task would appear to be similar; thus, one would not think that the changes in strategy that were found following FiFx lesions would be due to differences in motor response requirements, e.g., the sequential organization of responses, or selection of motor response strategies. Rather, the present results suggest that the changes in strategy after damage to the septohippocampal system must be due to changes in other more central processes. By combining both spatial and nonspatial components concurrently in the same task, the present research shows that one can obtain information regarding the strategies that rats use in performing this kind of task. A similar approach should prove of value in other research designed to investigate the effects on behavior of brain damage and drug manipulations. REFERENCES Becker, J. T., Walker, J. A., & Olton, D. S. (1980). Neuroanatomical bases of spatial memory. Brain Research, 200, 307320. Bouffard, J-P, & Jarrard, L. E. (1988). Acquisition of a complex place task in rats with selective ibotenate lesions of hippocampal formation: Combined lesions of subiculum and entorhinal cortex versus hippocampus. Behavioral Neuroscience, 102, 828-834. Buresova, O., & Bures, J. (1982). Radial maze as a tool for assessing the effects of drugs on the working memory of rats. Psychopharmacology, 77, 268-271.
STRATEGY SELECTION IN RATS Buresova, O., Bolhuis, J. J., & Bures, J. (1986). Differential effects of cholinergic blockade on performance of rats in the water tank navigation task and in a radial water maze. Behavioral Neuroscience, 100, 476-482. Hagan, J. J., & Morris, R. G. M. (1987). The cholinergic hypothesis of memory: A review of animal experiments. In L. L. Iversen, S. D. Iversen, & S. Snyder (Eds.), Handbook ofpsychopharmacology, Vol. 20, pp. 237-323. New York: Plenum Press. Hagan, J. J., Tweedie, F., & Morris, R. G. M. (1986). Lack of task specificity and absence of posttraining effects of atropine on learning. Behaworal Neuroscience, 100, 483-493. Honig, W. K. (1978). Studies of working memory in the pigeon. In S. A. Hulse, H. Fowler & W. K. Honig (Eds.), Cognitive processes in animal behavior, pp. 211-248. Hillsdale, NJ: Erlbaum. Jarrard, L. E. (1978a). Selective hippocampal lesions and spatial discrimination in the rat. Society for Neuroscience Abstracts, 3, 222. Jarrard, L. E. (1978b). Selective hippocampal lesions: Differential effects on performance by rats of a spatial task with preoperative versus postoperative training. Journal of Comparative and Physiological Psychology, 96, 1119-1127. Jarrard, L. E. (1983). Selective hippocampal lesions and behavior: Effects of kainic acid lesions on performance of place and cue tasks. Behavioral Neuroscience, 97, 873-889. Jarrard, L. E. (1986). Selective hippocampal lesions and behavior: Implications for current research and theorizing. In R. L. Isaacson & K. H. Pribram (Eds.), The hippocampus, Vol. 4, pp. 93-126. New York: Plenum. Jarrard, L. E., & Elmes, D. G. (1982). Role of retroactive interference in the spatial memory of normal rats and rats with hippocampal lesions. Journal of Comparative and Physiological Psychology, 96, 699-711. Jarrard, L. E., Okaichi, H., Steward, O., & Goldschmidt, R. B. (1984). On the role of hippocampal connections in the performance of place and cue tasks: Comparisons with damage to hippocampus. Behavioral Neuroscience, 98, 946-954.
179
M'Harzi, M., & Monmaur, P. (1985). Selective lesions of the fimbria and the fornix in the rat: Differential effects on CA1 and dentate theta. Experimental Neurology, 89, 361-371. M'Harzi, M., Palacios, A., Monmaur, P., Willig, F., Houcine, O., Delacour, J. (1987). Effects of selective lesions of fimbriafornix on learning set in the rat. Physiology and Behavior, 40, 181-188. Morris, R. G. M., Tweedie, F., Schenk, F., & Jarrard, L. E. (1990). Dissociation between components of spatial memory after ibotenate lesions of the hippocampus. European Journal of Neuroscience, 2, 1016-1028. Nadel, L., & MacDonald, L. (1980). Hippocampus: Cognitive map or working memory? Behavioral and Neural Biology, 29, 405-409. O'Keefe, J., & Nadal, L. (1978). The hippocampus as a cognitive map. London: Oxford Univ. Press, (Clarendon). Olton, D. S. (1983). Memory functions and the hippocampus. In W. Seifert (Ed.), Neurobiology of the hippocampus, pp. 335373. New York: Academic Press. Olton, D. S., Becker, J. T., & Handelman, G. E. (1979). Hippocampus, space and memory. Behavioral Brain Sciences, 2, 313-365. Olton, D. S., & Papas, B. C. (1979). Spatial memory and hippocampal function. Neuropsychologia, 17, 669-682. Olton, D. S., & Samuelson, R. J. (1976). Remembrance of places passed: Spatial memory in rats. Journal of Experimental Psychology: Animal Behavior Processes, 2, 97-116. Paxinos, G., & Watson, C. (1982). The rat brain in stereotaxic coordinates. New York: Academic Press. Whishaw, I. Q. (1985). Cholinergic receptor blockade in the rat impairs locale but not taxon strategies for place navigation in a swimming pool. Behavioral Neuroscience, 99, 979-1005. Whishaw, I. Q., & Petrie, B. F. (1988). Cholinergic blockade in the rat impairs strategy selection but not learning and retention of nonspatial visual discrimination problems in a swimming pool. Behavioral Neuroscience, 102, 662-667.
Strategy Selection in a Task with Spatial and Nonspatial Components: Effects of Fimbria-Fornix Lesions in Rats M. M'HARZI Laboratoire de Psychophysiologie, Universite Paris 7, 75251 Paris Cedex 05, France AND
LEONARD E. JARRARD 1 Department of Psychology, Washington and Lee University, Lexington, Virginia 24450
INTRODUCTION
The main purpose of the present research was to investigate the ability of rats to learn a 12-arm radial maze task that requires the concurrent utilization of both spatial and intramaze cue information. The task involves in a single trial both place and cue learning as well as reference memory (RM) and working memory (WM). Since the animal can choose place and cue arms in any order, the strategies employed to learn the task can be studied as well as the kinds of memory errors that are made. The results of Experiment 1 showed that the number of errors made on the place and cue components of the task did not differ, and that more RM than WM errors were made early during learning. As the task was learned, the animals tended to choose the place arms before choosing the intramaze cue arms, thus suggesting that a spatial strategy was employed first followed by a cue strategy. In Experiment 2 lesions of the fimbria-fornix resulted in temporary impairments in both RM and WM that were especially apparent on the spatial component of the task. The lesioned rats also switched from choosing mostly place arms early during the trial to choosing more cue arms. While fimbria-fornix lesioned rats recovered from the memory impairments with training, the change in response strategy persisted throughout postoperative testing. The procedure of combining both spatial and nonspatial components concurrently in the same task should prove of value in studying response strategies in animals.
The radial arm maze has proved to be an especially useful tool in the study of spatial m em o ry in rats (Buresova & Bures, 1982; Olton & Samuelson, 1976). In the radial maze experi m ent performance depends not only on the r a t r e m e m b e r i n g the general procedures of the t ask from day to day, b u t also which arms have been visited within a trial: the underl yi ng m e m o r y processes have been referred to as reference m e m o r y (RM) and working m e m o ry (WM), respectively (Honig, 1978). The procedure t h a t is used most often involves employing an 8arm radial maze, placing the reward at the end of each of the 8 arms, and leaving the r a t on the maze until all 8 reinforcements are obtained (see Olton & Samuelson, 1976). One problem in i nt er p re tin g impaired performance following drug or brain manipulations is t h a t repeated entries into arms already chosen on t h a t trial can reflect either a general m e m o r y dysfunction, a problem in remembering general procedures required to carry out the task (RM), or impaired ability to r e m e m b e r which arms have already been visited within t h a t trial
(WM). In an a t t e m p t to det erm i ne more precisely the n a t u r e of m e m o r y impairments, several investigators have employed the radial maze and a limited baiting procedure, e.g., fewer t h a n the total n u m b e r of arms are consistently baited over trials (Jarrard, 1978a; Olton & Papas, 1979). Using this procedure, RM errors are operationally defined as choices of arms t h a t are never baited, while WM errors are defined as repeated entries into baited and unbaited
© 1992 Academic Press, Inc.
l This research was carried out while M.M. was a visiting scientist at Washington and Lee University, Lexington, Virginia. The research was supported by National Science Foundation G r a n t s BNS 8507259 and 8809208 to L.E.J. Address r e p r i n t requests to M. M'Harzi, Centre de Recherches Roussel-UCLAF, Pharmacologie-Neurobiology, 111, route de Noisy, Nocard, RC, 93230 Romainville, France. 171
0163-1047/92 $5.00 Copyright © 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.
172
M'HARZI AND JARRARD
arms that had already been visited within the trial. WM errors can be divided further into repeated entries into baited arms (working memory correct or WM-C) and reentries into arms that have never been baited (working memory incorrect or WM-I). A further refinement in procedure involves employing the radial maze to study cue as well as spatial learning and memory (Nadel & MacDonald, 1980; Jarrard, 1983). In the experiment by J a r r a r d an 8-arm radial maze and a within-subjects design were used, together with a procedure that permitted determining two kinds of learning (place and cue) and two memory functions (WM and RM). In the place task the rat learned to choose 4 out of 8 similar arms, where all arms remained in the same spatial location from trial to trial. In the intramaze cue task, different textured floor inserts in the 8 arms were moved in a random order from trial to trial, and the rat was rewarded for choosing the same 4 cues. This more complicated testing procedure designed to compare place and cue learning has been employed in several experiments with some interesting differential effects resulting from brain manipulations (see Jarrard, 1983, 1986). In the research reported here, we describe an extension of the basic radial maze procedure that is designed to obtain information regarding the strategies used by rats in solving a combined spatial and cue radial maze task. Using a 12-arm maze, 6 of the arms were designated place arms and 6 were designated intramaze cue arms. Three of the 6 place arms were consistently baited from trial to trial thus permitting one to look at place WM and RM errors. The 6 cue arms contained different textured floor inserts that were moved in a random order from trial to trial; 3 of the 6 cues were consistently baited over trials permitting a determination of cue WM and RM errors. Since in performing the task the rat can choose to visit the arms in any order, it is possible to study the different strategies that were employed. Experiment 1 was carried out to see if rats can learn this more complicated version of the radial maze task, and to see whether the data indicate that different strategies are employed as the task is learned. It is well known that damage to different components of the hippocampal formation results in impaired acquisition of spatial radial mazes (Becker, Walker, & Olton, 1980; Bouffard & Jarrard, 1989; Jarrard, 1978a,b; J a r r a r d & Elmes, 1982; Olton, 1983; Olton, Becker, & Handelman, 1979). In order to determine the usefulness of the testing procedure described above, and to better understand the nature of the impairment found in rats following dam-
age to the septohippocampal system, rats that learned the task in Experiment 1 were retested in Experiment 2 following damage to the fimbria-fornix (FiFx). EXPERIMENT 1 The purpose of the first experiment was to see if the rats can learn the combined place-cue task described in the Introduction. Performance of the place component of the task should require the animals to use distal room cues to choose the 3 arms that are consistently baited and avoid the 3 arms that are never baited. The salient stimuli for the cue component of the task are the specific characteristics of the intramaze cues; however, the location of these cues within the 6 cue arms is changed over trials. We were interested in obtaining answers to several questions. If the task is learned, will the place and cue components be learned at different rates? Will the order in which the arms are chosen indicate a preference for either the place or cue arms?
Method Subjects. Fifteen male, albino, S p r a g u e - D a w l e y rats weighing 250-300 g were used. They were maintained under a 12:12 h light/dark cycle (07001900) at a temperature of 22°C (+_ 2 °) and were housed in individual cages for the duration of the experiment. Apparatus. The animals were trained on a 12arm radial maze, constructed of 0.019 m varnished wood, and elevated 0.65 m above the floor. Each arm (0.66 x 0.10 m) radiated from a circular central platform 0.76 m in diameter. Clear Plexiglas walls (6.5 cm high) extended the length of each arm. Small holes (1.2 cm in diameter and 0.5 cm deep) at the end of each arm served as wells for the sucrose solution. The apparatus was located in a room that was well lighted and sound attenuated. Many extramaze cues were present on the walls and around the room. Six removable inserts 0.6 x 0.09 m were constructed by gluing to a plywood backing different materials (carpet, cloth, chicken wire, ceiling tile, Plexiglas, screen). These intramaze cues were located in arms (1, 3, 5, 8, 10, and 12) and served to test for cue learning (the cue component) (see Fig. 1). The 6 inserts were moved pseudorandomly from trial to trial within the 6 arms that were designed for cue learning. The remaining 6 similar arms (2, 4, 6, 7, 9, and 11) served to test learning of the place task (the place component). Three of the 6 place arms and 3 of the 6 cues were baited on each trial.
STRATEGY SELECTION IN RATS
o]
FIG. 1. Schematic drawing of the 12-arm radial maze. The arms are numbered from 1 to 12. Place arms are identified by blank arms and cue arms are shown as containing different patterned stimuli.
Four place-cue combinations were used with 4 squads of animals. Subjects in a given squad were always reinforced for choosing the same 3 places and the same 3 cues, even though the cues changed locations over trials.
Procedure. Before testing began, the subjects were given five successive sessions of handling (one per day). During this time, 5 to 10 ml of 25% sucrose (to be used as reward) was made available in the home cage of each rat. Amounts of food and water were limited to 12-15 g and 2 0 - 2 5 ml, respectively. These amounts were later adjusted for each rat so that body weight was maintained at 80-85% of the predeprivation level. Following handling, each rat was placed for 5 min each day, for 5 days on the central platform. During this time, sucrose was available in a small glass dish placed on the central platform and the animal could explore the maze. Food and water were given to each animal 1 h after being returned to the home cage. Behavioral testing took place during the light cycle (0800-1400). Rats were given 60 trials, 2 each day with an intertrial interval of approximately 2 h. On each trial, wells at the end of correct arms were baited with 0.5 ml of sucrose and the rat was placed on the central platform. A trial was continued until either all 6 reinforcements had been obtained (3 place and 3 cue), until the animal had made 24
173
choices (correct and incorrect) or until 8 min had elapsed. A correct choice was counted when the rat entered an arm that contained the reinforcement. Incorrect choices were recorded as follows: choices of arms or cues that were never baited were counted as reference memory (RM) errors; reentering previously baited arms or cues was considered as working memory correct (WM-C) errors; and reentering never baited arms or cues was recorded as working memory incorrect (WM-I) errors. In some analyses, WM (WM-C + WM-I) was used because too few WM-I errors were made for analysis. Response strategies employed by the rats in learning the task were analyzed by studying the order in which the place and cue arms were chosen. In addition to the first choice made (place (P) or cue (C)), and the first correct choice, the analysis included the number of trials where the 3 correct P arms were chosen before a cue arm ( P P P C . . . ) , and number of trials where the 3 correct C arms were chosen before a place arm ( C C C P . . . ) . A mixed strategy was operationally defined as a trial where an animal made a mixed sequence of P and C choices (e.g., PCCPPC etc., with the place component of the task being completed first; or CPCCPP etc., with the cue task being completed first). Statistical analysis was performed using mixed design analysis of variance (ANOVA) with subsequent pairwise comparisons being made with the N e w m a n - K e u l s test. Paired and unpaired t test were also employed.
Results and Discussion Figure 2 shows the mean number of RM, WM-C, and WM-I errors made by the rats (N = 15) on the place and cue components of the task during acquisition. A 2 x 3 x 6 mixed design ANOVA (Place-Cue x RM, WM-C, WM-I errors x Blocks of trials) was used to analyze these data. The overall analysis indicated that the total number of errors made on the place and cue components of the task did not differ, F < 1.0. However, there were significant differences in types of memory errors that were made, F(2, 28) = 200.75, p < .01. In addition, the interaction of Types of memory errors x Blocks of trials was significant, F(10, 140) = 27.10, p < .01. F u r t h e r analyses revealed that the animals made significantly more RM than WM-C errors through the first five blocks of trials (p < .01), more RM than WM-I errors through all six blocks (p < .01), and more WM-C than WM-I errors through the first three blocks of trials (p < .01). The interaction
174
M'HARZI AND JARRARD
25
20
15
o ~10 Q ©
=
5
2: ©
0
Blocks of 10 Trials FIG. 2. Mean n u m b e r of errors over blocks of trials during acquisition. Errors are identified as RM, reference memory; WMC, working memory correct; WM-I, working memory incorrect.
of P l a c e - C u e × Types of memory errors was also significant, F(2, 28) = 12.29, p < .01. Thus, the animals made significantly more WM-C errors on the place than the cue component of the task (p < .01), and conversely more RM errors on the cue than the place component of the task (p < .01). There was an overall decrease in number of errors over the six blocks of trials, F(5, 70) = 58.06, p < .01. The triple interaction of P l a c e - C u e x Types of memory errors × Blocks of trials was not significant. Although the animals made approximately the same number of total errors on the place and cue components of the task, the order in which the arms were chosen changed as the task was learned. The nature of these changes in preference can be seen in Fig. 3. Considering only the first choice that was made, the animals displayed an increased preference for place arms over blocks of trials beginning with 48.6% (a level expected by chance) on Block 1 to 93.3% on Block 6. The percentage of the trials on which the first correct arm was a place arm increased from 58.0% on Block 1 to 94.0% on Block 6. As a result, preference for the cue component decreased over blocks of trials (from 51.4 to 6.7%). As shown in Fig. 3, the animals also exhibited a decrease in adopting a mixed strategy, e.g., a mixed
sequence of place and cue choices (from 98.0 to 70.0%). Considering only the trials on which choices were mixed place and cue arms and the place component was resolved first, there was an increase from 60 to 86% on Blocks 1 and 6, respectively. There was an increase in completing all 3 "place before any cue" (PPPC) choice strategies (1.3 to 30.0%); however, choosing "cue before any place" (CCCP) choice strategy was almost never used by the animals (0.7% on Block 1 and 0.0% on Block 6). Even though the combined place and cue task is complex, the results clearly show that rats are able to learn to select the baited cue and place arms. If one looks only at total number of errors that were made during acquisition, the place and cue components appear to be of equal difficulty. However, a consideration of the order in which the arms were chosen shows that the rats did not respond to the place and cue components equally. Rather, as learning progressed there was a decided increase in choosing the place arms first, thus indicating a preference for the place arms. These results strongly suggest that the use of distal cues (the relevant stimuli for the place component) were easier for the rats than the proximal, intramaze cues (the relevant
Choice Stategy During Acquisition 100
80
.~ 60
,,"' q'~ o
I-O-P as 1st Correct Choice I "El"- PPPO... I-A- cccP...
0'"
g4o ~
2O
0
Q L)
0 1
2
3
4
5
6
Blocks of 10 Trials FIG. 3. Choice strategy d u r i n g acquisition. Abbreviations used: P, Place; P P P C . . . , the 3 correct place arms were visited before choosing a cue arm; C C C P . . . , the 3 correct cue arms were visited before choosing a place arm; Mixed, mixed place and cue arms.
STRATEGY SELECTION IN RATS stimuli for the cue component that changed location from trial to trial). Thus, as the task was learned the rats appeared to first employ a spatial strategy before then using a nonspatial, cue strategy. EXPERIMENT 2 Experiment 2 was designed to investigate the effects of lesions of the FiFx on retention of the task that had been learned in Experiment 1. In previous research the first author had found that damage limited to the most dorsal aspects of the fornix and fimbria both abolished hippocampal theta (M'Harzi & Monmaur, 1985) and resulted in impaired performance on a spatial task (M'Harzi, Palacios, Monmaur, Willig, Houcine, & Delacour, 1987). Similar small lesions of FiFx were used in Experiment 2 in order to see if performance of the spatial and cue components of the 12-arm radial maze task would be equally affected, and whether the predominant place strategy employed by the rats in performing the task would change.
Method Following training in Experiment 1, the rats were divided into two groups that were matched using performance measures obtained during original acquisition. One group received the FiFx lesion (FiFx Group), and the other served as a control (Co Group). The rats were anesthetized using a mixture of sodium pentobarbitol and chloral hydrate. Lesions were produced electrolytically by passing a dc cathode current (2 mA for 20 s) through a stainless steel electrode. The lesion electrode was insulated except for the tip. The stereotaxic coordinates were as follows: 0.7 mm posterior to bregma; ML = + 1.0 mm, DV = - 4 . 1 mm; ML = +1.5 mm, DV = - 4 . 4 mm. The surface of the skull was used to determine DV coordinates. In operated control rats the electrode penetrations were similar to those in the FiFx Group except the electrode was lowered 1.0 mm above the target structure and no current was passed. Postoperative testing was begun 10 days after surgery. The behavioral procedures were similar to those used during acquisition in Experiment 1. The animals were given 2 trials on each of 10 days by which time performance had returned to the preoperative level. At the completion of the experiment the animals were deeply anesthetized and transcardially perfused with saline and 10% neutralized formalin. The brains were removed and stored in neutralized for-
175
malin for several days. The brains were then frozen, cut at 40 t~m in the coronal plane, and sections showing lesions and/or electrode tracts were stained with a cresyl violet stain. Preliminary analysis of the preoperative data for the FiFx and Co Groups showed that the groups did not differ in total errors or responding pattern prior to the surgery. In the analyses of the postoperative data WM-C and WM-I errors were summed to form WM errors since few reentries into never-baited arms or cues were made.
Results and Discussion Behavior. Animals with FiFx lesions were significantly impaired in performance early during postoperative testing but they rapidly recovered and testing was discontinued after 20 trials. The pattern of errors that were made is shown in Fig. 4. A 2 x 2 x 2 x 4 mixed design ANOVA was used to analyze the data with Groups as the between-subjects variable and Place-Cue, R M - W M , and 4 Blocks of 5 trials as within-subjects variables. The overall analysis indicated that animals with FiFx lesions made significantly more total errors than Co anireals, F(1, 33) = 6.15, p < .05. There was also a significant interaction of Groups x Blocks of trials, F(3, 33) = 3.58, p < .05, showing that performance of the groups was differentially affected over trials. A more detailed analysis revealed that FiFx rats made significantly more errors than Co rats on Blocks 1 and 2 (p < .01), but that the groups were similar by the third block of trials. The interaction of Groups x P l a c e - C u e was not significant; however, the triple interaction of Groups x P l a c e - C u e x Type of memory error was significant, F(1, 11) = 4.95, p < .05. FiFx rats made more RM and WM errors on the place component and more WM errors on the cue component (all p's < .05). The FiFx animals did not make more RM errors than controls on the cue task. Thus, damage to the FiFx resulted in a general impairment on the place task (both RM and WM) but only a WM impairment on the cue task. Finally, the interaction of Groups x Types of memory errors x Blocks of trials was significant, F(3, 33) = 3.00, p < .05. FiFx rats made more RM errors than Co rats on Blocks 1 and 2, and more WM errors on Blocks 1, 2, and 3 (all p's < .05). By the fourth block of trials the groups did not differ on any of the measures. Analysis of the order in which the place and cue arms were chosen indicated a change in preference that resulted from interrupting the axons in FiFx. There was general agreement among the various
176
M'HARZI AND JARRARD
measures; however, only the data for the mixed trials on which the place component was completed before the cue component are presented here (see Fig. 5). Specifically, the analysis indicated that Co animals completed the place component significantly more often than FiFx rats, F(1, 11) = 15.26, p < .01. There was a significant increase in the percentage of trials completed over blocks of trials, F(3, 33) = 3.9, p < 0.02, but the Groups × Blocks interaction was not significant, F < 1.0. These findings show that partial damage to the FiFx not only resulted in temporary impairments on both place and cue components of the task, but also that the lesions altered the choice strategies of the damaged animals.
~'~100 Postoperative Choice Strategy
~o 9 0 []
~80
~70 o 060
I
~50
Histology.
Pre
Histological examination of the brains indicated that in two rats the damage was not acceptable, and the data for these animals were not included in the final analyses. In all animals that were included, the lesions principally destroyed the dorsal aspect of the Fi and the dorsal Fx (see Fig. 6). The center of the lesion was located at the level shown in Plate 17 of the Paxinos and Watson (1982) atlas. In some animals the ventral part of the fimbria was slightly damaged but in none of the cases was the ventral hippocampal commissure in-
1
2
4
FIG. 5. Choice strategy (mixed trials with the place component being completed first) of FiFx and Co Groups during postoperative testing.
terrupted. In most rats the lesions extended rostrally to the septofimbrial nucleus and the nucleus of the lateral septum and caudally to the septal pole of the dorsal hippocampus. In some animals the lesions included a small amount of damage to the
PRM
8
3
Blocks of 5 Trials
PWM
6 4 2
[]
0
,
1 ~.
GLM FiFx (n = 7)
[]
Co
8" o
6"
!
(n = 6)
4 2; 2 []
0
, Pre
I
I
I
i
1 2 3 4 Blocks of Five Trials
~
I
I
l
I
Pre
1
2
3
4
FIG. 4. Mean number of memory errors on place and cue arms made by the FiFx and Co Groups on the last block of acquisition trials (Pre) and during the 4 blocks of postoperative testing. Errors are identified as CRM, cue reference memory; CWM, cue working memory; PRM, Place reference memory; PWM, Place working memory errors.
STRATEGY
SELECTION
177
IN RATS
•:..~...X,.:...~ ,~ ..,-..
k*~',.: "
"
" " ":','.'" - .
.
.
.
" ~'
? ":,'
;"
-
, ~'t
.~.
FIG.
6.
Photomicrograph
"-"-
-:"" :~C.--.~_."
~2-'~ , -
-'; ~' " 2 , "
-
£ "~," --_
with
violet from
,,.*.
-'~-,
of a coronal
section
stained
'ventral part of the overlying corpus callosum. As was intended, the damage to FiFx was similar in amount and location to that reported in previous research (M'Harzi & Monmaur, 1985; M'Harzi et al., 1987). GENERAL DISCUSSION The main purpose of the present research was to compare and contrast the utilization of concurrently presented spatial and nonspatial information in the rat. The results indicated that rats can learn a complex task that has a spatial component presumably requiring the use of distal extramaze cues and a separate nonspatial component requiring the use of intramaze cues. Since there were no differences in total number of errors that were made in learning the place and cue components of the task, one might be tempted to conclude that the two were treated similarily by the rats; however, analysis of the pattern of errors (RM, WM) and the order in which the arms were chosen shows that the animals did not respond to the place and cue components in the same way. Rather, as learning progressed, there was a decided preference for the place arms. This preference for place was demonstrated by choosing the place arms before choosing the arms containing the cues (see Fig. 3). Thus, as the animals learned the task, the sequence of choices within a trial changed from equal choices of place and cue arms to where
cresly
a representative
animal
in the FiFx
Group.
a place strategy was adopted on early choices before then using a cue strategy. In a previous experiment by the second author, place and cue learning were tested separately in the same animals using an 8-arm radial maze with different textured floor inserts being placed in the arms during cue testing (and cues changing locations over trials) and with the inserts being removed for the place task (Jarrard, Okaichi, Steward, & Goldschmidt, 1984). While in that study unoperated animals made more total errors in learning the cue task as compared to the place task, RM and WM errors were equally distributed. In the present more direct comparison of place and cue learning, total errors did not differ for place and cue but there were differential effects in terms of RM and WM errors. Specifically, the rats made more RM errors on the cue component than the place component, e.g., there were significantly more choices of never baited cues. While it is possible that the greater number of cue RM errors could reflect increased interference resulting from a tendency for the rats to return to specific arms (places) that had been reinforced on previous trials rather than being influenced by the intramaze cues, inspection of the data failed to support this view. In contrast with the greater number of cue RM errors made by the rats in learning the task, more WM errors were made on the place component, e.g., the rats repeatedly entered more correct, previously
178
M'HARZI AND JARRARD
baited place arms that had already been visited during a trial. Since the cues remained on the same arms during a trial, both cues and places could be used as discriminative stimuli for the working memory component. However, the fact that fewer WM errors were made on the cue than the place component makes it less likely that the rats were using spatial location in performing the WM component of the cue task. The underlying explanation for the above differential effects is not readily apparent, but if one assumes that number of errors reflects task difficulty, the results suggest that the rats had more trouble remembering correct from incorrect cues over days, and which place arms had already been visited during a trial. While there were these differential effects for choice accuracy, the sequence of choices indicated that as learning progressed the rats tended to choose the place arms before the cure arms. The small lesions of FiFx resulted in a temporary impairment on both place and cue components of the task. In terms of the types of memory errors that were made, the FiFx animals exhibited a general impairment on the place task with significant increases in both RM and WM errors, while only WM was affected on the cue task. This pattern of results is similar to that found in the study described above with place and cue learning and large lesions of the FiFx (see Jarrard et al., 1984). In that experiment the FiFx was completely interrupted and the impairments persisted throughout postoperative testing, while in the present study performance of rats with the more limited lesions did recover. Thus, by using more limited damage to FiFx we were able to look at preferences for place as compared to cue arms when the groups did not differ in terms of overall choice accuracy. As pointed out in the Introduction, it had been established in previous research that limited lesions similar to those used in the present experiment served to disrupt hippocampal theta and impair original acquisition of a complex spatial task (M'Harzi & Monmaur, 1985; M'Harzi et ~1, 1987). Of special interest in the present study was the change in the order that the place and cue arms were visited following damage to the FiFx. Compared to control animals and performance before the operations, the lesioned rats switched from visiting primarily place arms on early choices within a trial to choosing more cue arms. Such a change in strategy following damage to the septohippocampal system would follow from the "spatial mapping" theory of hippocampal function proposed by O'Keefe and Nadel (1978). If the hippocampus does form the
neural basis for spatial mapping as these authors propose, damage to the axons in FiFx should result in the animals experiencing more difficulty with the spatial arms. Even though with practice performance recovered in terms of choice accuracy, the FiFx animals continued to show a decreased preference for the place arms. A number of investigators have pointed out the similarities in the effects on behavior that result from damage to the septohippocampal system and blocking the cholinergic systems with drugs (see Hagan & Morris, 1987). Morris, Tweedie, Schenk, and Jarrard (1990) recently reported that in the Morris swimming task rats with the hippocampus removed swim rapidly in large circles often around the edge of the pool; similar effects have been described by Whishaw and Petrie (1988) following cholinergic blockade with atropine. Of more direct relevance for the present research are reports that following cholinergic blockade rats are more impaired in acquiring locale strategies than taxon strategies (Buresova, Bolhuis, & Bures, 1986; Hagan, Tweedie, & Morris, 1986; Whishaw, 1985). The motor response requirements for performing the place and cue components of the 12-arm radial maze task would appear to be similar; thus, one would not think that the changes in strategy that were found following FiFx lesions would be due to differences in motor response requirements, e.g., the sequential organization of responses, or selection of motor response strategies. Rather, the present results suggest that the changes in strategy after damage to the septohippocampal system must be due to changes in other more central processes. By combining both spatial and nonspatial components concurrently in the same task, the present research shows that one can obtain information regarding the strategies that rats use in performing this kind of task. A similar approach should prove of value in other research designed to investigate the effects on behavior of brain damage and drug manipulations. REFERENCES Becker, J. T., Walker, J. A., & Olton, D. S. (1980). Neuroanatomical bases of spatial memory. Brain Research, 200, 307320. Bouffard, J-P, & Jarrard, L. E. (1988). Acquisition of a complex place task in rats with selective ibotenate lesions of hippocampal formation: Combined lesions of subiculum and entorhinal cortex versus hippocampus. Behavioral Neuroscience, 102, 828-834. Buresova, O., & Bures, J. (1982). Radial maze as a tool for assessing the effects of drugs on the working memory of rats. Psychopharmacology, 77, 268-271.
STRATEGY SELECTION IN RATS Buresova, O., Bolhuis, J. J., & Bures, J. (1986). Differential effects of cholinergic blockade on performance of rats in the water tank navigation task and in a radial water maze. Behavioral Neuroscience, 100, 476-482. Hagan, J. J., & Morris, R. G. M. (1987). The cholinergic hypothesis of memory: A review of animal experiments. In L. L. Iversen, S. D. Iversen, & S. Snyder (Eds.), Handbook ofpsychopharmacology, Vol. 20, pp. 237-323. New York: Plenum Press. Hagan, J. J., Tweedie, F., & Morris, R. G. M. (1986). Lack of task specificity and absence of posttraining effects of atropine on learning. Behaworal Neuroscience, 100, 483-493. Honig, W. K. (1978). Studies of working memory in the pigeon. In S. A. Hulse, H. Fowler & W. K. Honig (Eds.), Cognitive processes in animal behavior, pp. 211-248. Hillsdale, NJ: Erlbaum. Jarrard, L. E. (1978a). Selective hippocampal lesions and spatial discrimination in the rat. Society for Neuroscience Abstracts, 3, 222. Jarrard, L. E. (1978b). Selective hippocampal lesions: Differential effects on performance by rats of a spatial task with preoperative versus postoperative training. Journal of Comparative and Physiological Psychology, 96, 1119-1127. Jarrard, L. E. (1983). Selective hippocampal lesions and behavior: Effects of kainic acid lesions on performance of place and cue tasks. Behavioral Neuroscience, 97, 873-889. Jarrard, L. E. (1986). Selective hippocampal lesions and behavior: Implications for current research and theorizing. In R. L. Isaacson & K. H. Pribram (Eds.), The hippocampus, Vol. 4, pp. 93-126. New York: Plenum. Jarrard, L. E., & Elmes, D. G. (1982). Role of retroactive interference in the spatial memory of normal rats and rats with hippocampal lesions. Journal of Comparative and Physiological Psychology, 96, 699-711. Jarrard, L. E., Okaichi, H., Steward, O., & Goldschmidt, R. B. (1984). On the role of hippocampal connections in the performance of place and cue tasks: Comparisons with damage to hippocampus. Behavioral Neuroscience, 98, 946-954.
179
M'Harzi, M., & Monmaur, P. (1985). Selective lesions of the fimbria and the fornix in the rat: Differential effects on CA1 and dentate theta. Experimental Neurology, 89, 361-371. M'Harzi, M., Palacios, A., Monmaur, P., Willig, F., Houcine, O., Delacour, J. (1987). Effects of selective lesions of fimbriafornix on learning set in the rat. Physiology and Behavior, 40, 181-188. Morris, R. G. M., Tweedie, F., Schenk, F., & Jarrard, L. E. (1990). Dissociation between components of spatial memory after ibotenate lesions of the hippocampus. European Journal of Neuroscience, 2, 1016-1028. Nadel, L., & MacDonald, L. (1980). Hippocampus: Cognitive map or working memory? Behavioral and Neural Biology, 29, 405-409. O'Keefe, J., & Nadal, L. (1978). The hippocampus as a cognitive map. London: Oxford Univ. Press, (Clarendon). Olton, D. S. (1983). Memory functions and the hippocampus. In W. Seifert (Ed.), Neurobiology of the hippocampus, pp. 335373. New York: Academic Press. Olton, D. S., Becker, J. T., & Handelman, G. E. (1979). Hippocampus, space and memory. Behavioral Brain Sciences, 2, 313-365. Olton, D. S., & Papas, B. C. (1979). Spatial memory and hippocampal function. Neuropsychologia, 17, 669-682. Olton, D. S., & Samuelson, R. J. (1976). Remembrance of places passed: Spatial memory in rats. Journal of Experimental Psychology: Animal Behavior Processes, 2, 97-116. Paxinos, G., & Watson, C. (1982). The rat brain in stereotaxic coordinates. New York: Academic Press. Whishaw, I. Q. (1985). Cholinergic receptor blockade in the rat impairs locale but not taxon strategies for place navigation in a swimming pool. Behavioral Neuroscience, 99, 979-1005. Whishaw, I. Q., & Petrie, B. F. (1988). Cholinergic blockade in the rat impairs strategy selection but not learning and retention of nonspatial visual discrimination problems in a swimming pool. Behavioral Neuroscience, 102, 662-667.
