- Email: [email protected]
Effects of telencephalic ablation on short-term memory and attention in goldfish
Behavioural Brain Research 86 (1997) 191 – 199
Effects of telencephalic ablation on short-term memory and attention in goldfish Ken Ohnishi * Department of Physiology, Nara Medical Uni6ersity, Kashihara, Nara 634, Japan Accepted 26 November 1996
Abstract Two hypotheses regarding the functions of the teleost telencephalon (short-term memory and nonspecific arousal hypotheses) were examined by using a Y-maze training paradigm. A delayed reinforcement method, which allowed the separation of choice process (a process in which choice responses are evoked) and reward process (a process in which choice responses are reinforced), showed that normal fish can acquire clear learned responses to choice stimuli under different stimulus conditions between the choice process and the reward process, while telencephalon-ablated fish showed greatly impaired learning performance. Neither normal nor telencephalon-ablated fish could acquire learned responses to choice stimuli under neutral stimulus conditions in the reward process with respect to choice stimuli in the choice process. These results suggest that the telencephalon facilitates extratelencephalic short-term memory function essential for memory retention of choice stimuli and evoked choice responses until reinforcement, and support the supplementary function of the telencephalon suggested previously [Ohnishi, K., Telencephalic function implicated in food-reinforced colour discrimination learning in the goldfish, Physiol. Beha6., 46 (1989) 707 – 712 and Savage, G.E., Temporal factors in avoidance learning in normal and forebrainless goldfish (Carassius auratus), Nature, 218 (1968) 1168–1169]. In addition, it was shown that cue information in the reward process is very important for the fixation of short-term memory of choice stimuli and choice responses. Furthermore, telencephalon-ablated fish also showed clear visual aspect selection, as did normal fish, when they were reinforced to a visual compound stimulus containing heterogeneous aspects (pattern and colour). This result shows that the telencephalon-related arousal or attentional function is not critical for aspect selection in goldfish. It seems that visual aspect selection in goldfish is performed without paying telencephalon-related ‘selective attention’ to an aspect. © 1997 Elsevier Science B.V. Keywords: Teleost; Goldfish; Telencephalon; Short-term memory; Attention; Visual discrimination; Aspect selection; Learning
1. Introduction The teleost telencephalon provides a good opportunity for studying the function of the vertebrate limbic system, because it is a simple structure homologous to the limbic structure of higher vertebrates [13,26] and can be easily ablated without serious surgical damage to other components of the nervous system due to an independent hemispheric form. Many studies (for a review see [3,4,6,16,24]) have been conduced to examine the various hypotheses proposed for the function of the * Corresponding author. Fax: +81 74 4257657 0166-4328/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S 0 1 6 6 - 4 3 2 8 ( 9 6 ) 0 2 2 6 5 - 6
teleost telencephalon in, (1) short-term memory, (2) nonspecific arousal, (3) inhibition of dominant responses, and (4) utilization of secondary reinforcement. The hypothesis of short-term memory was proposed based on the results of a series of experiments [22,25], in which reinforcement delivery was delayed a few seconds. Telencephalon-ablated fish were impaired in instrumental learning under such a delayed reinforcement process. The telencephalon appears to participate in retention of memory of choice stimuli and an evoked choice response until reinforcement is delivered. This hypothesis seems to be suitable for instrumental learning but not for classical conditioning. Overmier and
192
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
Savage [18] reported that there were no significant differences between normal and telencephalon-ablated fish in a trace classical conditioning experiment using heart rate in which a temporal delay was interposed between presentation of conditioning and unconditioning stimuli. Thus, at present, it seems that short-term memory is required for association of instrumental responses and reinforcement. From the results of pioneering experiments [22,25], however, doubt has remained that the resultant impairment of learning in telencephalon-ablated fish may actually be attributable to temporal factors such as delayed reinforcement. In those experiments, the reinforcement for evoked responses were given to fish at a location apart from the initial location where choice stimuli were presented. As a result, the stimulus conditions at reinforcement delivery were not equivalent to those at the time of choice response. The effect of such spatial separation on learning was not discussed in the early short-term memory hypothesis. Thereafter, considering this issue, Overmier and Patten [17] trained fish under two different stimulus conditions, i.e. the same stimulus conditions during both the choice and reward processes, and neutral stimulus conditions during the reward process with respect to the choice process. They demonstrated that both normal and telencephalon-ablated fish showed similar acquisition of learned responses under the same stimulus conditions despite the interposition of delay time, while both groups showed greatly impaired learning under neutral stimulus conditions; in this case, the latter showed a greater impairment in learning. These results suggest that stimulus condition under which evoked choice responses are reinforced is more critical for association of evoked choice response with reinforcement rather than interposition of a temporal delay between the response and reinforcement. This suggestion demands reconsideration of the short-term memory hypothesis. The primary purpose of the present study was to re-examine the short-term memory hypothesis, giving further insight into stimulus conditions at reinforcement delivery. The second purpose of the present study was to test whether the visual aspect selection reported previously in goldfish [15] depends on telencephalon-related arousal or attentional mechanisms. According to the nonspecific arousal hypothesis, the telencephalon functions as an arousal or attentional centre which controls general arousal level not specific to any sensory modality. Spontaneous activity level is reduced after telencephalic ablation [8]. Electrical stimulation of the telencephalon produces arousal or an attentional-like posture in goldfish [23] and increased neural activity in the preoptic area in sunfish [5]. If this telencephalon-related arousal or attentional mechanism is important for visual aspect selection, telencephalon-ablated fish would
have some difficulty in attending to surrounding visual stimuli and processing only one aspect of discriminative stimuli in visual aspect selection. In higher vertebrates, it is generally thought that an attentional mechanism, ‘selective attention’, plays an important role in visual aspect selection, based upon the proposal that an animal attends to an aspect of stimulus when its response is under the control of the stimulus [9,11,10,20]. However, to date, it has not been examined whether this is the case with aspect selection in teleosts.
2. Materials and methods
2.1. Animals and apparatus Common goldfish (Carassius auratus), 10 –13 cm in body length, which were maintained in aerated home tanks at 20–23°C, were used throughout the present study. A Y-maze (Fig. 1) slightly modified from that described by Overmier and Patten [17] was used for discrimination training. The apparatus, which was made of transparent acrylic, consisted of two parallel swim-ways subdivided by sliding doors into a starting chamber (5 cm wide at the stem, 10.5 cm wide at the larger end, 20 cm long, and 9 cm in height) and choice, waiting and goal chambers (each 5 cm wide, 13 cm long, and 9 cm in height). It was filled with water to a depth of 8 cm. The water was kept at 20–23°C during experiments and replaced daily. Varied discriminative stimuli and neutral stimuli were attached to the lateral inner walls of choice and goal chambers and sliding door, according to experimental schedule, and neutral stimuli were attached to those of start and waiting chambers. These coloured stimuli were illuminated by fluorescent lamps (National FL205, EDL.42; general
Fig. 1. Y-maze training apparatus. CC, choice chamber; GC, goal chamber; SC, starting chamber; WC, waiting chamber.
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
193
shown in Fig. 2. The reflectance peaks (lmax) of the blue and the green papers were about 480 nm and 520 nm, respectively. These peaks correspond approximately to the sensitivity maxima (lmax) at 470 and 540 nm [12] or 460 and 520 nm of goldfish [1]. The subjective brightness of the coloured papers was adjusted to be equivalent for trained fish according to the following equation [15]: R=
&
700
P(l)rb(l)S(l) dl/
400
&
700
P(l)rw(l)S(l) dl
400
where rb(l)= the reflectance spectra of a blue paper; rw(l)= the reflectance spectra of perfect white with r(l)= 100 at all wavelengths. For a green paper, rb(l) was replaced with rg(l). Table 1 shows the relative values of the subjective brightness of each coloured paper for the goldfish. The coloured papers were presented randomly in the following combinations from trial to trial, (1) light blue vs. light green, (2) light blue vs. dark green, (3) dark blue vs. light green, or (4) dark blue vs. dark green to prevent the fish from discriminating the coloured papers on the basis of brightness. The reflectance spectra of the grey paper used for neutral stimuli were about 30% at all wavelengths.
2.3. Training procedure Fig. 2. Relative reflectance r(l) of the light blue (L.B.) and green (L.G.) coloured papers, and the dark blue (D.B.) and green (D.G.) coloured papers (A); the spectral sensitivity S(l) of goldfish (from [12]) (B); and the relative radiant power P(l) of the light source (C). The relative values of the goldfishes’ subjective brightness of the blue and green coloured papers were obtained from (A), (B) and (C).
colour rendering index, Ra= 98, colour temperature= 5000 K) with spectral characteristics similar to sunlight positioned above the apparatus. The intensity of illumination, measured at the position of the lateral inner walls, was approximately 1000 lux.
2.2. Discriminati6e stimuli Five types of discriminative stimuli were presented: (1) compound stimuli, vertical patterns (7 mm interval space between black bars, bar width =2 mm) on green background (VG) vs. horizontal patterns (the same conditions as vertical patterns) on blue background (HB); (2) compound stimuli, VG vs. vertical patterns on blue background (VB); (3) simple stimuli, vertical patterns on white background (VW) vs. horizontal patterns on white background (HW); (4) simple stimuli, green vs. blue; and (5) simple stimuli, black vs. white. The reflectance spectra r(l) of the coloured papers used for these discriminative stimuli, which were measured with a Hitachi Colour Analyzer M-307, are
Goldfish were placed individually in the apparatus for at least 30 min daily for adaptation purposes, and 3–4 days later pretraining trials and training trials were started. A training trial was performed in the following manner. After setting the discriminative stimuli to the lateral inner walls of the choice and goal chambers, individual fish were placed in the start chamber and about 10 s later the first sliding door was opened, allowing fish to swim into one of the two choice chambers. After they remained in the choice chamber for about 5 s, the second sliding door was opened and they were guided into the waiting chamber where they remained for about 5 s. Finally, the third door was opened and fish were guided into the goal chamber where dry food (Kyorin Food Ind.) was delivered (in the case of immediate reward delivery after correct response, the food was delivered in the choice chamber Table 1 Relative values of the goldfishes’ subjective brightness of the colored papers used in the present study Color Light Dark Light Dark Grey
Relative brightness (%) blue blue green green
29.6 24.8 28.2 22.6 30.3
194
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
and the fish were guided into the waiting chamber, and then after about 1 min they were returned to a waiting tank). The fish were moved from the goal chamber to the waiting chamber to reduce the effects of handling on the memory fixation in the goal chamber. During only pretraining trials, food was given to fish regardless of the choice chamber which they entered. After remaining in the goal chamber for about 1 min, they were guided back into the waiting chamber and then returned to the waiting tank. Such trials were performed 6 times a day with 20 min intertrial intervals for 6 successive days (pretraining sessions) and then for the following 12–40 successive days (training sessions). Twelve trials made up one session. When fish remained in the correct chamber for at least 5 s, the choice response was regarded as correct. When fish entered an incorrect chamber, they were allowed to swim into the correct chamber. Such a choice response in which fish remained in the incorrect chamber for over 5 s was regarded as incorrect. Nonpreferred stimuli during the pretraining sessions were used as reinforced stimuli in the training sessions. Most fish showed preference to blue colour and horizontal pattern [15]. A pair of discriminative stimuli was presented randomly according to a Gellerman series [2] so that fish could not acquire position learning. To check for the occurrence of aspect selection, the test trials (10 trials a day) without reinforcements were performed in the compound stimuli (VG vs. HB) discrimination by using the compound stimuli and the constituent colour (green and blue) and pattern aspects (VW and HW). More detailed explanations of the training procedure have been described elsewhere [15].
Fig. 3. Representative goldfish brain. (A) Bilaterally telencephalonablated brain. (B) Normal brain. Cb, cerebellum; OB, olfactory bulb; OT, optic tectum; Tel, telencephalon; VL, vagal lobe. Scale bar is 2 mm.
2.5. Statistical analysis The percentages of correct responses per session (during training trials) were analyzed between normal and telencephalon-ablated groups by three-way analysis of variance (ANOVA). The percentages of correct choices obtained from the test trials were analyzed by paired t-test.
3. Results
3.1. Compound stimuli (VG 6s. HB) discrimination under stimulus conditions equi6alent between the choice and reward processes
2.4. Operation and 6erification Telencephalic ablation was performed as described previously [14]. Briefly, fish anaesthetized with 0.1% tricaine methanesulfonate (MS 222) were positioned in a stereotaxic apparatus which was slightly modified from that of Peter and Gill [19], the dorsal part of the skull was removed with a dental drill and the exposed telencephalic lobes and olfactory bulbs were aspirated. The skull wound caused by drilling was closed with cotton wetted with saline solution, and covered with dental cement. Operated fish remained in their home tanks for at least 2 days after the operation. After training, the brains were removed from the skulls and frontal sections 50 mm thick were cut according to the procedure of Ohnishi [14] to check the extent of aspiration. The aspiration procedure was very reliable, producing complete removal of the telencephalic lobes without injuring the optic tectum or preoptic area in all operated fish. Representative normal and telencephalon-ablated goldfish brains are shown in Fig. 3.
Normal fish (n= 6) were trained to discriminate between the compound stimuli (VG vs. HB), which were presented in the choice and goal chambers. As shown in Fig. 4A, the fish acquired clear learned responses (about 90% correct responses) after nine sessions. To check which aspect of the compound stimuli was selectively learned in this discrimination, the test trials (Fig. 4B) were performed by using the constituent pattern (vertical and horizontal patterns) and colour aspects (green and blue). When tested with the pattern aspect, the fish showed very high percentages (about 90%) of correct choice responses after 13 sessions. In contrast, fish tested with the colour aspect showed very low percentages (about 40% of correct choice responses after 11 sessions) and the responses were not incremental (about 45%; t5 = 0.84, P\ 0.4) despite the high rate training trials (23 sessions). Such low percentages of correct choice responses to the colour aspect were similar to those of correct responses (about 40%) during the pretraining period. These results indicate that
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
195
the fish trained with VG vs. HB showed aspect selection in that they acquired learned responses on the basis of pattern aspect. This group’s colour discrimination ability was likely normal because a previous report [15] showed that a group trained with compound stimuli (VG vs. VB) different only in colour aspect could discriminate between them.
3.2. Simple stimuli (black 6s. white) discrimination under stimulus conditions equi6alent between choice and reward processes Normal (n= 6) and telencephalon-ablated fish (n= 7) were trained to discriminate between the simple stimuli (black vs. white), which were presented in the choice and the goal chambers. As shown in Fig. 5, both groups similarly acquired clear learned responses (about 80% correct responses) at the 9th session (Groups, F1,11 =2.77, P \0.05; Sessions, F5,55 = 20.61, PB 0.01; Groups ×Sessions, F5,55 =1.46, P \ 0.05, as described in Section 2, the percentages of correct responses per session during pretraining trials, three sessions, were not included in ANOVA). This result indicates that the black and white stimuli used under stimulus conditions which differed between choice and reward processes (described below) could be discriminated for normal and telencephalon-ablated fish.
Fig. 4. Compound stimuli (VG vs. HB) discrimination in normal fish under stimulus conditions equivalent between the choice and reward processes. (A) Learning curve. c and p indicate the points at which colour and pattern tests were performed, respectively. The arrowhead indicates the start of the training trials. During the pretraining trials, the fish were rewarded regardless of which choice stimulus they chose, but during the training trials, they were rewarded only when they chose the stimulus preferred less frequently during the pretraining trials; most fish showed a preference for HB. (B) Percentages of correct choices. The open and stippled columns in the colour test represent the percentages of correct choices after 11 and 23 sessions, respectively. Vertical bars, 9 S.D.
Fig. 5. Simple stimuli (black vs. white) discrimination in normal () and telencephalon-ablated fish () under stimulus conditions equivalent between the choice and reward processes. See Fig. 4 for further explanation.
3.3. Simple stimuli (black 6s. white) discrimination under stimulus conditions differing between choice and reward processes Normal fish (n=6) were trained with the simple stimuli (black vs. white) and the compound stimuli (VG vs. HB), which were presented in the choice and the goal chambers, respectively. This group acquired clear learned responses (about 90% correct responses) to the choice stimuli (presented in the choice chambers) after 12 sessions, but the acquisition rate to the choice stimuli was very late (Fig. 6A) compared with that in the discrimination under stimulus conditions equivalent between choice and reward processes (Fig. 5). Interestingly, the test trials (Fig. 6B) showed that prior to acquisition of the learned responses to the choice stimuli, the learned responses to the reward stimuli (presented in the goal chambers) had already been acquired. The acquisition rate of learned responses to the rewarded stimuli was similar to that to the choice stimuli (VG vs. HB) under the same stimulus conditions (Fig. 4): in the test trials, the percentages of correct choice responses to the reward stimuli were about 95% after eight sessions and significantly different from those obtained prior to training (t5 = 9.69, PB0.001). When tested with the pattern aspect, the fish showed very high percentages (about 95%) of correct choice responses after nine sessions. This high percentages of correct choice responses were not due to preference to a pattern, since nonpreferred pattern (vertical) was used as correct stimulus (most fish preferred horizontal pattern in the present study and a previous study [15]). In contrast, when tested with the colour aspect, the fish showed very low percentages (about 45%) of correct choice responses after ten sessions. No increments in correct responses (about 55%) were observed with more training trials (14 sessions; t5 = 1.73, P\0.1). This
196
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
correct choice responses level was similar to that (about 45% correct responses) during the pretraining period. These results indicate that the fish trained with black vs. white (choice stimuli) and VG s. HB (reward stimuli) could discriminate between both types of discriminative stimuli in each process, and that aspect selection occurred in the reward process in parallel with the acquisition of learned responses to the reward stimuli. Telencephalon-ablated fish (n =8) were similarly trained with the simple stimuli (black vs. white) and the compound stimuli (VG vs. HB), which were presented in the choice and the goal chambers, respectively. Fig. 7A shows that the learning rate to the choice stimuli in this group was very late (about 60% correct responses in later sessions: the fish acquired learned responses, Sessions, F10,70 =3.38, P B0.01) compared with that in the normal group (about 90% correct responses in later sessions: Groups, F1,12 =4.77, P B0.05; Sessions, F10,120 = 12.22, P B 0.01; Groups × Sessions, F10,120 = 7.23, PB0.01). This poor discrimination ability of telencephalon-ablated fish does not result from difficulty in discrimination between the black and white stimuli because they could clearly discriminate between them, as already shown in Fig. 5. The telencephalon-ablated
Fig. 6. Simple stimuli (black vs. white) discrimination in normal fish under stimulus conditions which differed between choice (black and white) and reward processes (VG and HB). The discriminative stimuli were presented with the combination of black-HB and white-VG. (A) Learning curve; co, c and p indicate the points at which compound, colour and pattern tests were performed, respectively. (B) Percentages of correct choices. The open and closed columns in compound tests represent the percentages of correct choices during pretraining (after 3 sessions) and training (after 8 sessions), respectively. The open and dotted columns in colour tests represent the percentages of correct choices after 10 and 14 sessions, respectively. * PB 0.001 (compared with the pretraining percentage). See Fig. 4 for further explanation.
Fig. 7. Simple stimuli (black vs. white) discrimination in telencephalon-ablated fish under stimulus conditions which differed between choice (black and white) and reward processes (VG and HB). (A) Learning curve. (B) The percentages of correct choices. * PB 0.001 (compared with the pretraining percentage). See Figs. 4 and 6 for further explanation.
fish also showed aspect selection to the compound stimuli presented in the goal chambers, as did normal fish. The test trials (Fig. 7B) performed after the acquisition of learned responses to the reward stimuli (compound stimuli test) clearly showed the occurrence of aspect selection (t7 = 12.98, PB 0.001). When tested with the pattern aspect, the fish showed high percentages of correct choice responses (about 85% after 9 sessions). Since nonpreferred pattern (vertical) was used as a reinforced stimuli, this high percentages of correct choice responses were not due to preference to a pattern. In contrast, when tested with the colour aspect, the fish showed low percentages [about 60% after 10 sessions and about 45% after 14 sessions; there was no significant difference between the percentages of correct choice response (t7 = 0.45, P\ 0.6)]. In this discrimination, it seems that the colour vision of the telencephalon-ablated fish was not defective. When other telencephalon-ablated fish (n=5) were trained to discriminate between the compound stimuli (VG vs. VB) presented in both the choice and the goal chambers, they showed high percentages of correct responses (about 80%) in later sessions (after 12 sessions), as shown in Fig. 8. Since these compound stimuli have no cue information concerning pattern, this result suggests that the telencephalon-ablated fish discriminated between the compound stimuli on the basis of colour information. The colour test trials (inset of Fig. 8), showing high percentages of correct choice responses (about 80%) to the constituent colours, confirmed this suggestion.
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
197
Fig. 8. Compound stimuli (VG vs. VB) discrimination in telencephalon-ablated fish under stimulus conditions equivalent between choice and reward processes. The inset shows the percentages of correct choices in colour test. See Fig. 4 for further explanation.
3.4. Compound stimuli (VG 6s. HB) discrimination under neutral stimulus conditions in the reward process Normal (n=5) and telencephalon-ablated fish (n=8) were trained with the compound stimuli (VG vs. HB) and the neutral stimuli (grey coloured papers) presented in the choice and the goal chambers, respectively. As shown in Fig. 9, unlike the results described above, in this discrimination, both groups showed no learned responses to the choice stimuli (Groups, F1,11 = 5.12, PB 0.05; Sessions, F11,121 =1.30, P \0.05; Groups× Sessions, F11,121 =0.69, P \ 0.05). No increments in correct responses were observed in later sessions (about 50%). It seems that these fish had difficulty in discriminating between the choice stimuli under such neutral stimulus conditions in the reward process. These results were very different from those reported by Overmier and Patten [17]. They trained fish to discriminate between black and white under similar neutral stimulus conditions in the reward process, and showed that the fish acquired clear learned responses to the choice
Fig. 9. Compound stimuli (VG vs. HB) discrimination in normal () and telencephalon-ablated fish () under neutral stimulus conditions (grey coloured papers) in the reward process. Note no improvement in rate of correct responses in later sessions. See Fig. 4 for further explanation.
Fig. 10. Compound stimuli (VG vs. HB) discrimination in normal () and telencephalon-ablated fish () with reward delivery immediately after choice responses. (A) Learning curve. (B) Percentages of correct choices. The open and stippled columns in colour tests represent the percentages of correct choices after 11 and 15 sessions, respectively.
stimuli. The reason for this discrepancy between the results of the present study and those reported by Overmier and Patten [17] is unknown, but differences in procedure between the two studies [e.g. type of choice trial (free-choice or forced-choice trials), period of training days and/or strength of reinforcement] may be responsible.
3.5. Compound stimuli (VG 6s. HB) discrimination in which reward was deli6ered immediately after a correct response To determine whether unknown outside factors [previously presented visual information (black and white) in the choice process and/or delayed task-related factors] affect the aspect selection observed in delayed reinforcement discrimination, normal (n= 8) and telencephalon-ablated fish (n=8) were rewarded immediately in the choice chambers after correct choice responses to the compound stimuli (VG vs. HB) presented in the choice chambers. As shown in Fig. 10A, both groups showed clear learned responses in later sessions, and the learning ability of the telencephalonablated fish was poorer than that of the normal fish (Groups, F1,14 = 24.17, PB 0.01; Sessions, F11,154 = 54.13, PB 0.01; Groups×Sessions, F11,154 =2.23, PB 0.05). The test trials (Fig. 10B) showed that both groups obtained the learned responses to pattern aspect but not colour aspect [there were no significant differences in the percentages of correct choices between each
198
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
colour test (t7 =1.00, P \0.3 in normal, t7 =1.23, P\ 0.2 in telencephalon-ablated)]. These results indicate that the aspect selection observed in the delayed reinforcement discrimination was not affected by the above described outside factors in both of normal and telencephalon-ablated groups.
4. Discussion Savage’s observation that interposition of a delay between choice response and reinforcement produced impaired learning in telencephalon-ablated but not in normal fish in instrumental training experiments [22,25] in which reinforcement was delivered under a surrounding environment different from that at choice response, provided the short-term memory hypothesis as one of telencephalic functions. Overmier and Patten [17], thereafter, emphasized the necessity of the reconsideration of the short-term memory hypothesis on the basis of the observation that telencephalon-ablated fish normally acquired learned responses regardless of interposition of a delay between choice response and reinforcement in a Y-maze paradigm when stimulus condition at choice response was the same as that at reinforcement delivery. This observation suggested that the causal factor of the impaired learning is not the interposed delay but the change in stimulus condition, and thus led Overmier and Patten to be critical of the short-term memory hypothesis. However, they did not discuss their other important observation that telencephalon-ablated fish showed more impaired learning than normal controls when stimulus condition at reinforcement delivery was neutral with respect to that at choice response. If the telencephalon was not important for short-term memory as they proposed, impaired learning would not have been observed in telencephalon-ablated fish under such stimulus conditions. From this, the possibility still remains that the telencephalon contributes to retention of short-term memory until delivery of reinforcement. Therefore, it seemed necessary to test the short-term memory hypothesis further. In the present study, I trained fish under varied stimulus conditions and showed that the difference in stimulus conditions between choice response and reinforcement delivery has a considerable influence on learning performance. No learned responses to choice stimuli were observed in either normal or telencephalon-ablated fish when they were trained with reward stimuli neutral with respect to choice stimuli. The percentages of correct responses in both groups did not increase and remained at the level of chance (about 50% correct responses) throughout all sessions. In contrast, clear learned responses to choice stimuli were observed in normal fish when they were trained with
reward stimuli including discriminable visual cue information which were different from choice stimuli. In this case, the fish first acquired learned responses to the reward stimuli and then associated the reward stimuli with choice stimuli because they showed high percentages of correct choices to the reward stimuli in test trials before acquiring learned responses to choice stimuli. These results suggest that cue information of discriminative stimuli at reinforcement delivery is critical for the acquisition of learned responses and it largely serves to retain the short-term memory which contributes to the association of reward stimuli with choice stimuli. This association process may be a cause of the difference in the rate of learning between discrimination trials under the same and under different stimulus conditions. Telencephalon-ablated fish showed greatly impaired learning performance to choice stimuli under conditions which differed between choice response and reinforcement delivery. This observation indicates that telencephalon-ablated fish have great difficulty in associating reward stimuli with choice stimuli and thus the telencephalon plays an important role in short-term memory. However, slightly increased correct responses were observed in later sessions. This indicates that unknown extratelencephalic attentional systems might exist. The present study did not control the delay time between the presentation of choice and rewarded stimuli. Further experiments without the intervening delay would support the short-term memory hypothesis. On the other hand, in previous studies [14,21] which examined other telencephalic functions, telencephalon-ablated fish showed impaired learning at a training rate sufficient for normal fish to learn colour but it was improved with more training trials. This result suggested that the telencephalon is supplementary in that it facilitates integration of neural events in extratelencephalic areas which are necessary for an instrumental process such as the utilization of changes in conditioned motivational reactions. In the present study, such telencephalic facilitation in utilization of secondary reinforcement should also be considered. The slightly impaired learning performances in telencephalon-ablated fish in the black vs. white discrimination under the same stimulus conditions and the compound stimuli discrimination with immediate reward delivery were probably the result of the loss of this facilitation function of the telencephalon. Considering this point, the greatly impaired learning performances in telencephalon-ablated fish in that discrimination under the different stimulus conditions probably resulted from the binal loss of the facilitation functions in not only utilization of secondary reinforcement but also short-term memory. Facilitation may be a fundamental function mediated by the teleost telencephalon for higher neural events.
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
It is generally thought that the attentional mechanism, ‘selective attention’, is closely related to visual aspect selection in higher vertebrates. If this is the case with goldfish, telencephalon-ablated fish will show some affected behaviours in visual aspect selection reflecting the hypothesis that the telencephalon is an arousal or attentional centre. Contrary to this assumption, the results of the present study demonstrated that telencephalon-ablated fish showed no affected aspect selection behaviour. They selectively processed only one aspect in the compound stimuli discrimination as did normal fish. (It has been reported that normal goldfish selectively process only more easily discriminated aspects in a Y-maze instrumental conditioning regardless of a high rate of training when they were trained with visual compound stimuli composed of a more easily discriminated aspect and a more difficult aspect [15]. In the present study, pattern might be more discriminable than colour. Similar easiness to discriminate pattern has been reported by Ingle [7]). Thus, the telencephalon-related attentional function appears not to be critical for aspect selection in goldfish. This result supports the suggestion that aspect selection in goldfish is performed without paying attention to a specific aspect but is automatic in accordance with the intensity or salience of constituent aspects [15]. However, some unknown extratelencephalic attentional system(s) may also participate in aspect selection in teleosts.
Acknowledgements I thank Dr K.-I. Takahashi and Y. Ogawa for their useful comments, and Professors Y. Enoki and F. Motokszama and co-workers for their great encouragement.
References [1] Beauchamp, R.D. and Rowe, J.S., Goldfish spectral sensitivity: a conditioned heart-rate measure in restrained or curarized fish, Vision Res., 17 (1977) 617–624. [2] Fellows, B.J., Chance sequences for discrimination tasks, Psychol. Bull., 67 (1967) 87–92. [3] Flood, N.C., Overmier, J.B. and Savage, G.E., Teleost telencephalon and learning: an interpretive review of data and hypotheses, Physiol. Beha6., 16 (1976) 783–798. [4] Flood, N.B. and Overmier, J.B., Learning in teleost fish: role of the telencephalon. In P. Laming (Ed.), Brain Mechanisms of Beha6iour in Lower Vertebrates, Cambridge University Press, Cambridge, 1981, pp. 259–279. [5] Hallowitz, R.A., Woodward, D.J. and Demski, L.S., Forebrain activation of single units in preoptic area of sunfish, Comp. Biochem. Physiol., 40A (1971) 733–741.
199
[6] Hollis, K.L. and Overmier, J.B., The function of the teleost telencephalon in behaviour: a reinforcement mediator. In D.I. Mostofsky (Ed.), The Beha6iour of Fish and other Aquatic Animals, Academic Press, New York, 1978, pp. 137–195. [7] Ingle, DJ., Interocular transfer in goldfish: colour easier than pattern, Science, 149 (1965) 1000 – 1002. [8] Janzen, W., Untersuchungen uber Grosshirnfunktionen des Goldfishes (Carassius auratus), Zool. Jahrb. Abt. Allgem. Zool. Physiol. Tiere, 52 (1933) 592 – 628. [9] Krechevsky, I., ‘Hypotheses’ in rats, Psychol. Re6., 39 (1932) 516 – 532. [10] Lashley, K.S., Conditional reactions in the rat, J. Psychol., 6 (1938) 311 – 324. [11] Lea, S.E.G., Complex general process learning in mammalian vertebrates. In P. Marler and H.S. Terrace (Eds.), The Biology of Learning, Springer-Verlag, Berlin, 1984, pp. 373–398. [12] Neumeyer, C., On spectral sensitivity in the goldfish: evidence for neutral interactions between different ‘cone mechanisms’, Vision Res., 24 (1984) 1223 – 1231. [13] Northcutt, R.G. and Bradford, M.R., New observations on the organization and evolution of the telencephalon of actinopterygian fishes. In S.O.E. Ebbesson (Ed.), Comparati6e Neurology of the Telencephalon, Plenum Press, New York, 1980, pp. 41 – 98. [14] Ohnishi, K., Telencephalic function implicated in food-reinforced colour discrimination learning in the goldfish, Physiol. Beha6., 46 (1989) 707 – 712. [15] Ohnishi, K., Goldfish’s visual information processing patterns in food-reinforced discrimination learning between compound visual stimuli, J. Comp. Physiol. A, 168 (1991) 581–589. [16] Overmier, J.B. and Hollis, K.L., The teleostean telencephalon in learning. In R.E. Davis and R.G. Northcutt (Eds.), Fish Neurobiology, Vol. 2, The University of Michigan, Ann Arbor, 1983, pp. 265 – 284. [17] Overmier, J.B. and Patten, R.L., Teleost telencephalon and memory for delayed reinforces, Physiol. Psychol., 10 (1982) 74 – 78. [18] Overmier, J.B. and Savage, G.E., Effects of telencephalic ablation on trace classical conditioning of heart rate in goldfish, Exp. Neurol., 42 (1974) 339 – 346. [19] Peter, R.E. and Gill, V.E., A stereotaxic atlas and technique for forebrain nuclei of the goldfish, Carassius auratus., J. Comp. Neurol., 159 (1975) 69 – 102. [20] Reynolds, G.S., Attention in the pigeon, J. Exp. Anal. Beha6., 4 (1961) 203 – 208. [21] Savage, G.E., Temporal factors in avoidance learning in normal and forebrainless goldfish (Carassius auratus), Nature, 218 (1968) 1168 – 1169. [22] Savage, G.E., Some preliminary observations on the role of the telencephalon in food-reinforced behaviour in the goldfish, Carassius auratus, Anim. Beha6., 17 (1969) 760 – 772. [23] Savage, G.E., Behavioural effects of electrical stimulation of the telencephalon of the goldfish, Carassius auratus, Anim. Beha6., 19 (1971) 661 – 668. [24] Savage, G.E., The fish telencephalon and its relation to learning. In S.O.E. Ebbesson (Ed.), Comparati6e Neurology of Telencephalon, Plenum Press, New York, 1980, pp. 129–174. [25] Savage, G.E. and Swingland, I.R., Positively reinforced behaviour and the forebrain in goldfish, Nature, 221 (1969) 878– 879. [26] Schroeder, D.M., The telencephalon of teleosts. In S.O.E. Ebbesson (Ed.), Comparati6e Neurology of Telencephalon, Plenum Press, New York, 1980, pp. 99 – 115.
. .
Effects of telencephalic ablation on short-term memory and attention in goldfish Ken Ohnishi * Department of Physiology, Nara Medical Uni6ersity, Kashihara, Nara 634, Japan Accepted 26 November 1996
Abstract Two hypotheses regarding the functions of the teleost telencephalon (short-term memory and nonspecific arousal hypotheses) were examined by using a Y-maze training paradigm. A delayed reinforcement method, which allowed the separation of choice process (a process in which choice responses are evoked) and reward process (a process in which choice responses are reinforced), showed that normal fish can acquire clear learned responses to choice stimuli under different stimulus conditions between the choice process and the reward process, while telencephalon-ablated fish showed greatly impaired learning performance. Neither normal nor telencephalon-ablated fish could acquire learned responses to choice stimuli under neutral stimulus conditions in the reward process with respect to choice stimuli in the choice process. These results suggest that the telencephalon facilitates extratelencephalic short-term memory function essential for memory retention of choice stimuli and evoked choice responses until reinforcement, and support the supplementary function of the telencephalon suggested previously [Ohnishi, K., Telencephalic function implicated in food-reinforced colour discrimination learning in the goldfish, Physiol. Beha6., 46 (1989) 707 – 712 and Savage, G.E., Temporal factors in avoidance learning in normal and forebrainless goldfish (Carassius auratus), Nature, 218 (1968) 1168–1169]. In addition, it was shown that cue information in the reward process is very important for the fixation of short-term memory of choice stimuli and choice responses. Furthermore, telencephalon-ablated fish also showed clear visual aspect selection, as did normal fish, when they were reinforced to a visual compound stimulus containing heterogeneous aspects (pattern and colour). This result shows that the telencephalon-related arousal or attentional function is not critical for aspect selection in goldfish. It seems that visual aspect selection in goldfish is performed without paying telencephalon-related ‘selective attention’ to an aspect. © 1997 Elsevier Science B.V. Keywords: Teleost; Goldfish; Telencephalon; Short-term memory; Attention; Visual discrimination; Aspect selection; Learning
1. Introduction The teleost telencephalon provides a good opportunity for studying the function of the vertebrate limbic system, because it is a simple structure homologous to the limbic structure of higher vertebrates [13,26] and can be easily ablated without serious surgical damage to other components of the nervous system due to an independent hemispheric form. Many studies (for a review see [3,4,6,16,24]) have been conduced to examine the various hypotheses proposed for the function of the * Corresponding author. Fax: +81 74 4257657 0166-4328/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S 0 1 6 6 - 4 3 2 8 ( 9 6 ) 0 2 2 6 5 - 6
teleost telencephalon in, (1) short-term memory, (2) nonspecific arousal, (3) inhibition of dominant responses, and (4) utilization of secondary reinforcement. The hypothesis of short-term memory was proposed based on the results of a series of experiments [22,25], in which reinforcement delivery was delayed a few seconds. Telencephalon-ablated fish were impaired in instrumental learning under such a delayed reinforcement process. The telencephalon appears to participate in retention of memory of choice stimuli and an evoked choice response until reinforcement is delivered. This hypothesis seems to be suitable for instrumental learning but not for classical conditioning. Overmier and
192
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
Savage [18] reported that there were no significant differences between normal and telencephalon-ablated fish in a trace classical conditioning experiment using heart rate in which a temporal delay was interposed between presentation of conditioning and unconditioning stimuli. Thus, at present, it seems that short-term memory is required for association of instrumental responses and reinforcement. From the results of pioneering experiments [22,25], however, doubt has remained that the resultant impairment of learning in telencephalon-ablated fish may actually be attributable to temporal factors such as delayed reinforcement. In those experiments, the reinforcement for evoked responses were given to fish at a location apart from the initial location where choice stimuli were presented. As a result, the stimulus conditions at reinforcement delivery were not equivalent to those at the time of choice response. The effect of such spatial separation on learning was not discussed in the early short-term memory hypothesis. Thereafter, considering this issue, Overmier and Patten [17] trained fish under two different stimulus conditions, i.e. the same stimulus conditions during both the choice and reward processes, and neutral stimulus conditions during the reward process with respect to the choice process. They demonstrated that both normal and telencephalon-ablated fish showed similar acquisition of learned responses under the same stimulus conditions despite the interposition of delay time, while both groups showed greatly impaired learning under neutral stimulus conditions; in this case, the latter showed a greater impairment in learning. These results suggest that stimulus condition under which evoked choice responses are reinforced is more critical for association of evoked choice response with reinforcement rather than interposition of a temporal delay between the response and reinforcement. This suggestion demands reconsideration of the short-term memory hypothesis. The primary purpose of the present study was to re-examine the short-term memory hypothesis, giving further insight into stimulus conditions at reinforcement delivery. The second purpose of the present study was to test whether the visual aspect selection reported previously in goldfish [15] depends on telencephalon-related arousal or attentional mechanisms. According to the nonspecific arousal hypothesis, the telencephalon functions as an arousal or attentional centre which controls general arousal level not specific to any sensory modality. Spontaneous activity level is reduced after telencephalic ablation [8]. Electrical stimulation of the telencephalon produces arousal or an attentional-like posture in goldfish [23] and increased neural activity in the preoptic area in sunfish [5]. If this telencephalon-related arousal or attentional mechanism is important for visual aspect selection, telencephalon-ablated fish would
have some difficulty in attending to surrounding visual stimuli and processing only one aspect of discriminative stimuli in visual aspect selection. In higher vertebrates, it is generally thought that an attentional mechanism, ‘selective attention’, plays an important role in visual aspect selection, based upon the proposal that an animal attends to an aspect of stimulus when its response is under the control of the stimulus [9,11,10,20]. However, to date, it has not been examined whether this is the case with aspect selection in teleosts.
2. Materials and methods
2.1. Animals and apparatus Common goldfish (Carassius auratus), 10 –13 cm in body length, which were maintained in aerated home tanks at 20–23°C, were used throughout the present study. A Y-maze (Fig. 1) slightly modified from that described by Overmier and Patten [17] was used for discrimination training. The apparatus, which was made of transparent acrylic, consisted of two parallel swim-ways subdivided by sliding doors into a starting chamber (5 cm wide at the stem, 10.5 cm wide at the larger end, 20 cm long, and 9 cm in height) and choice, waiting and goal chambers (each 5 cm wide, 13 cm long, and 9 cm in height). It was filled with water to a depth of 8 cm. The water was kept at 20–23°C during experiments and replaced daily. Varied discriminative stimuli and neutral stimuli were attached to the lateral inner walls of choice and goal chambers and sliding door, according to experimental schedule, and neutral stimuli were attached to those of start and waiting chambers. These coloured stimuli were illuminated by fluorescent lamps (National FL205, EDL.42; general
Fig. 1. Y-maze training apparatus. CC, choice chamber; GC, goal chamber; SC, starting chamber; WC, waiting chamber.
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
193
shown in Fig. 2. The reflectance peaks (lmax) of the blue and the green papers were about 480 nm and 520 nm, respectively. These peaks correspond approximately to the sensitivity maxima (lmax) at 470 and 540 nm [12] or 460 and 520 nm of goldfish [1]. The subjective brightness of the coloured papers was adjusted to be equivalent for trained fish according to the following equation [15]: R=
&
700
P(l)rb(l)S(l) dl/
400
&
700
P(l)rw(l)S(l) dl
400
where rb(l)= the reflectance spectra of a blue paper; rw(l)= the reflectance spectra of perfect white with r(l)= 100 at all wavelengths. For a green paper, rb(l) was replaced with rg(l). Table 1 shows the relative values of the subjective brightness of each coloured paper for the goldfish. The coloured papers were presented randomly in the following combinations from trial to trial, (1) light blue vs. light green, (2) light blue vs. dark green, (3) dark blue vs. light green, or (4) dark blue vs. dark green to prevent the fish from discriminating the coloured papers on the basis of brightness. The reflectance spectra of the grey paper used for neutral stimuli were about 30% at all wavelengths.
2.3. Training procedure Fig. 2. Relative reflectance r(l) of the light blue (L.B.) and green (L.G.) coloured papers, and the dark blue (D.B.) and green (D.G.) coloured papers (A); the spectral sensitivity S(l) of goldfish (from [12]) (B); and the relative radiant power P(l) of the light source (C). The relative values of the goldfishes’ subjective brightness of the blue and green coloured papers were obtained from (A), (B) and (C).
colour rendering index, Ra= 98, colour temperature= 5000 K) with spectral characteristics similar to sunlight positioned above the apparatus. The intensity of illumination, measured at the position of the lateral inner walls, was approximately 1000 lux.
2.2. Discriminati6e stimuli Five types of discriminative stimuli were presented: (1) compound stimuli, vertical patterns (7 mm interval space between black bars, bar width =2 mm) on green background (VG) vs. horizontal patterns (the same conditions as vertical patterns) on blue background (HB); (2) compound stimuli, VG vs. vertical patterns on blue background (VB); (3) simple stimuli, vertical patterns on white background (VW) vs. horizontal patterns on white background (HW); (4) simple stimuli, green vs. blue; and (5) simple stimuli, black vs. white. The reflectance spectra r(l) of the coloured papers used for these discriminative stimuli, which were measured with a Hitachi Colour Analyzer M-307, are
Goldfish were placed individually in the apparatus for at least 30 min daily for adaptation purposes, and 3–4 days later pretraining trials and training trials were started. A training trial was performed in the following manner. After setting the discriminative stimuli to the lateral inner walls of the choice and goal chambers, individual fish were placed in the start chamber and about 10 s later the first sliding door was opened, allowing fish to swim into one of the two choice chambers. After they remained in the choice chamber for about 5 s, the second sliding door was opened and they were guided into the waiting chamber where they remained for about 5 s. Finally, the third door was opened and fish were guided into the goal chamber where dry food (Kyorin Food Ind.) was delivered (in the case of immediate reward delivery after correct response, the food was delivered in the choice chamber Table 1 Relative values of the goldfishes’ subjective brightness of the colored papers used in the present study Color Light Dark Light Dark Grey
Relative brightness (%) blue blue green green
29.6 24.8 28.2 22.6 30.3
194
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
and the fish were guided into the waiting chamber, and then after about 1 min they were returned to a waiting tank). The fish were moved from the goal chamber to the waiting chamber to reduce the effects of handling on the memory fixation in the goal chamber. During only pretraining trials, food was given to fish regardless of the choice chamber which they entered. After remaining in the goal chamber for about 1 min, they were guided back into the waiting chamber and then returned to the waiting tank. Such trials were performed 6 times a day with 20 min intertrial intervals for 6 successive days (pretraining sessions) and then for the following 12–40 successive days (training sessions). Twelve trials made up one session. When fish remained in the correct chamber for at least 5 s, the choice response was regarded as correct. When fish entered an incorrect chamber, they were allowed to swim into the correct chamber. Such a choice response in which fish remained in the incorrect chamber for over 5 s was regarded as incorrect. Nonpreferred stimuli during the pretraining sessions were used as reinforced stimuli in the training sessions. Most fish showed preference to blue colour and horizontal pattern [15]. A pair of discriminative stimuli was presented randomly according to a Gellerman series [2] so that fish could not acquire position learning. To check for the occurrence of aspect selection, the test trials (10 trials a day) without reinforcements were performed in the compound stimuli (VG vs. HB) discrimination by using the compound stimuli and the constituent colour (green and blue) and pattern aspects (VW and HW). More detailed explanations of the training procedure have been described elsewhere [15].
Fig. 3. Representative goldfish brain. (A) Bilaterally telencephalonablated brain. (B) Normal brain. Cb, cerebellum; OB, olfactory bulb; OT, optic tectum; Tel, telencephalon; VL, vagal lobe. Scale bar is 2 mm.
2.5. Statistical analysis The percentages of correct responses per session (during training trials) were analyzed between normal and telencephalon-ablated groups by three-way analysis of variance (ANOVA). The percentages of correct choices obtained from the test trials were analyzed by paired t-test.
3. Results
3.1. Compound stimuli (VG 6s. HB) discrimination under stimulus conditions equi6alent between the choice and reward processes
2.4. Operation and 6erification Telencephalic ablation was performed as described previously [14]. Briefly, fish anaesthetized with 0.1% tricaine methanesulfonate (MS 222) were positioned in a stereotaxic apparatus which was slightly modified from that of Peter and Gill [19], the dorsal part of the skull was removed with a dental drill and the exposed telencephalic lobes and olfactory bulbs were aspirated. The skull wound caused by drilling was closed with cotton wetted with saline solution, and covered with dental cement. Operated fish remained in their home tanks for at least 2 days after the operation. After training, the brains were removed from the skulls and frontal sections 50 mm thick were cut according to the procedure of Ohnishi [14] to check the extent of aspiration. The aspiration procedure was very reliable, producing complete removal of the telencephalic lobes without injuring the optic tectum or preoptic area in all operated fish. Representative normal and telencephalon-ablated goldfish brains are shown in Fig. 3.
Normal fish (n= 6) were trained to discriminate between the compound stimuli (VG vs. HB), which were presented in the choice and goal chambers. As shown in Fig. 4A, the fish acquired clear learned responses (about 90% correct responses) after nine sessions. To check which aspect of the compound stimuli was selectively learned in this discrimination, the test trials (Fig. 4B) were performed by using the constituent pattern (vertical and horizontal patterns) and colour aspects (green and blue). When tested with the pattern aspect, the fish showed very high percentages (about 90%) of correct choice responses after 13 sessions. In contrast, fish tested with the colour aspect showed very low percentages (about 40% of correct choice responses after 11 sessions) and the responses were not incremental (about 45%; t5 = 0.84, P\ 0.4) despite the high rate training trials (23 sessions). Such low percentages of correct choice responses to the colour aspect were similar to those of correct responses (about 40%) during the pretraining period. These results indicate that
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
195
the fish trained with VG vs. HB showed aspect selection in that they acquired learned responses on the basis of pattern aspect. This group’s colour discrimination ability was likely normal because a previous report [15] showed that a group trained with compound stimuli (VG vs. VB) different only in colour aspect could discriminate between them.
3.2. Simple stimuli (black 6s. white) discrimination under stimulus conditions equi6alent between choice and reward processes Normal (n= 6) and telencephalon-ablated fish (n= 7) were trained to discriminate between the simple stimuli (black vs. white), which were presented in the choice and the goal chambers. As shown in Fig. 5, both groups similarly acquired clear learned responses (about 80% correct responses) at the 9th session (Groups, F1,11 =2.77, P \0.05; Sessions, F5,55 = 20.61, PB 0.01; Groups ×Sessions, F5,55 =1.46, P \ 0.05, as described in Section 2, the percentages of correct responses per session during pretraining trials, three sessions, were not included in ANOVA). This result indicates that the black and white stimuli used under stimulus conditions which differed between choice and reward processes (described below) could be discriminated for normal and telencephalon-ablated fish.
Fig. 4. Compound stimuli (VG vs. HB) discrimination in normal fish under stimulus conditions equivalent between the choice and reward processes. (A) Learning curve. c and p indicate the points at which colour and pattern tests were performed, respectively. The arrowhead indicates the start of the training trials. During the pretraining trials, the fish were rewarded regardless of which choice stimulus they chose, but during the training trials, they were rewarded only when they chose the stimulus preferred less frequently during the pretraining trials; most fish showed a preference for HB. (B) Percentages of correct choices. The open and stippled columns in the colour test represent the percentages of correct choices after 11 and 23 sessions, respectively. Vertical bars, 9 S.D.
Fig. 5. Simple stimuli (black vs. white) discrimination in normal () and telencephalon-ablated fish () under stimulus conditions equivalent between the choice and reward processes. See Fig. 4 for further explanation.
3.3. Simple stimuli (black 6s. white) discrimination under stimulus conditions differing between choice and reward processes Normal fish (n=6) were trained with the simple stimuli (black vs. white) and the compound stimuli (VG vs. HB), which were presented in the choice and the goal chambers, respectively. This group acquired clear learned responses (about 90% correct responses) to the choice stimuli (presented in the choice chambers) after 12 sessions, but the acquisition rate to the choice stimuli was very late (Fig. 6A) compared with that in the discrimination under stimulus conditions equivalent between choice and reward processes (Fig. 5). Interestingly, the test trials (Fig. 6B) showed that prior to acquisition of the learned responses to the choice stimuli, the learned responses to the reward stimuli (presented in the goal chambers) had already been acquired. The acquisition rate of learned responses to the rewarded stimuli was similar to that to the choice stimuli (VG vs. HB) under the same stimulus conditions (Fig. 4): in the test trials, the percentages of correct choice responses to the reward stimuli were about 95% after eight sessions and significantly different from those obtained prior to training (t5 = 9.69, PB0.001). When tested with the pattern aspect, the fish showed very high percentages (about 95%) of correct choice responses after nine sessions. This high percentages of correct choice responses were not due to preference to a pattern, since nonpreferred pattern (vertical) was used as correct stimulus (most fish preferred horizontal pattern in the present study and a previous study [15]). In contrast, when tested with the colour aspect, the fish showed very low percentages (about 45%) of correct choice responses after ten sessions. No increments in correct responses (about 55%) were observed with more training trials (14 sessions; t5 = 1.73, P\0.1). This
196
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
correct choice responses level was similar to that (about 45% correct responses) during the pretraining period. These results indicate that the fish trained with black vs. white (choice stimuli) and VG s. HB (reward stimuli) could discriminate between both types of discriminative stimuli in each process, and that aspect selection occurred in the reward process in parallel with the acquisition of learned responses to the reward stimuli. Telencephalon-ablated fish (n =8) were similarly trained with the simple stimuli (black vs. white) and the compound stimuli (VG vs. HB), which were presented in the choice and the goal chambers, respectively. Fig. 7A shows that the learning rate to the choice stimuli in this group was very late (about 60% correct responses in later sessions: the fish acquired learned responses, Sessions, F10,70 =3.38, P B0.01) compared with that in the normal group (about 90% correct responses in later sessions: Groups, F1,12 =4.77, P B0.05; Sessions, F10,120 = 12.22, P B 0.01; Groups × Sessions, F10,120 = 7.23, PB0.01). This poor discrimination ability of telencephalon-ablated fish does not result from difficulty in discrimination between the black and white stimuli because they could clearly discriminate between them, as already shown in Fig. 5. The telencephalon-ablated
Fig. 6. Simple stimuli (black vs. white) discrimination in normal fish under stimulus conditions which differed between choice (black and white) and reward processes (VG and HB). The discriminative stimuli were presented with the combination of black-HB and white-VG. (A) Learning curve; co, c and p indicate the points at which compound, colour and pattern tests were performed, respectively. (B) Percentages of correct choices. The open and closed columns in compound tests represent the percentages of correct choices during pretraining (after 3 sessions) and training (after 8 sessions), respectively. The open and dotted columns in colour tests represent the percentages of correct choices after 10 and 14 sessions, respectively. * PB 0.001 (compared with the pretraining percentage). See Fig. 4 for further explanation.
Fig. 7. Simple stimuli (black vs. white) discrimination in telencephalon-ablated fish under stimulus conditions which differed between choice (black and white) and reward processes (VG and HB). (A) Learning curve. (B) The percentages of correct choices. * PB 0.001 (compared with the pretraining percentage). See Figs. 4 and 6 for further explanation.
fish also showed aspect selection to the compound stimuli presented in the goal chambers, as did normal fish. The test trials (Fig. 7B) performed after the acquisition of learned responses to the reward stimuli (compound stimuli test) clearly showed the occurrence of aspect selection (t7 = 12.98, PB 0.001). When tested with the pattern aspect, the fish showed high percentages of correct choice responses (about 85% after 9 sessions). Since nonpreferred pattern (vertical) was used as a reinforced stimuli, this high percentages of correct choice responses were not due to preference to a pattern. In contrast, when tested with the colour aspect, the fish showed low percentages [about 60% after 10 sessions and about 45% after 14 sessions; there was no significant difference between the percentages of correct choice response (t7 = 0.45, P\ 0.6)]. In this discrimination, it seems that the colour vision of the telencephalon-ablated fish was not defective. When other telencephalon-ablated fish (n=5) were trained to discriminate between the compound stimuli (VG vs. VB) presented in both the choice and the goal chambers, they showed high percentages of correct responses (about 80%) in later sessions (after 12 sessions), as shown in Fig. 8. Since these compound stimuli have no cue information concerning pattern, this result suggests that the telencephalon-ablated fish discriminated between the compound stimuli on the basis of colour information. The colour test trials (inset of Fig. 8), showing high percentages of correct choice responses (about 80%) to the constituent colours, confirmed this suggestion.
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
197
Fig. 8. Compound stimuli (VG vs. VB) discrimination in telencephalon-ablated fish under stimulus conditions equivalent between choice and reward processes. The inset shows the percentages of correct choices in colour test. See Fig. 4 for further explanation.
3.4. Compound stimuli (VG 6s. HB) discrimination under neutral stimulus conditions in the reward process Normal (n=5) and telencephalon-ablated fish (n=8) were trained with the compound stimuli (VG vs. HB) and the neutral stimuli (grey coloured papers) presented in the choice and the goal chambers, respectively. As shown in Fig. 9, unlike the results described above, in this discrimination, both groups showed no learned responses to the choice stimuli (Groups, F1,11 = 5.12, PB 0.05; Sessions, F11,121 =1.30, P \0.05; Groups× Sessions, F11,121 =0.69, P \ 0.05). No increments in correct responses were observed in later sessions (about 50%). It seems that these fish had difficulty in discriminating between the choice stimuli under such neutral stimulus conditions in the reward process. These results were very different from those reported by Overmier and Patten [17]. They trained fish to discriminate between black and white under similar neutral stimulus conditions in the reward process, and showed that the fish acquired clear learned responses to the choice
Fig. 9. Compound stimuli (VG vs. HB) discrimination in normal () and telencephalon-ablated fish () under neutral stimulus conditions (grey coloured papers) in the reward process. Note no improvement in rate of correct responses in later sessions. See Fig. 4 for further explanation.
Fig. 10. Compound stimuli (VG vs. HB) discrimination in normal () and telencephalon-ablated fish () with reward delivery immediately after choice responses. (A) Learning curve. (B) Percentages of correct choices. The open and stippled columns in colour tests represent the percentages of correct choices after 11 and 15 sessions, respectively.
stimuli. The reason for this discrepancy between the results of the present study and those reported by Overmier and Patten [17] is unknown, but differences in procedure between the two studies [e.g. type of choice trial (free-choice or forced-choice trials), period of training days and/or strength of reinforcement] may be responsible.
3.5. Compound stimuli (VG 6s. HB) discrimination in which reward was deli6ered immediately after a correct response To determine whether unknown outside factors [previously presented visual information (black and white) in the choice process and/or delayed task-related factors] affect the aspect selection observed in delayed reinforcement discrimination, normal (n= 8) and telencephalon-ablated fish (n=8) were rewarded immediately in the choice chambers after correct choice responses to the compound stimuli (VG vs. HB) presented in the choice chambers. As shown in Fig. 10A, both groups showed clear learned responses in later sessions, and the learning ability of the telencephalonablated fish was poorer than that of the normal fish (Groups, F1,14 = 24.17, PB 0.01; Sessions, F11,154 = 54.13, PB 0.01; Groups×Sessions, F11,154 =2.23, PB 0.05). The test trials (Fig. 10B) showed that both groups obtained the learned responses to pattern aspect but not colour aspect [there were no significant differences in the percentages of correct choices between each
198
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
colour test (t7 =1.00, P \0.3 in normal, t7 =1.23, P\ 0.2 in telencephalon-ablated)]. These results indicate that the aspect selection observed in the delayed reinforcement discrimination was not affected by the above described outside factors in both of normal and telencephalon-ablated groups.
4. Discussion Savage’s observation that interposition of a delay between choice response and reinforcement produced impaired learning in telencephalon-ablated but not in normal fish in instrumental training experiments [22,25] in which reinforcement was delivered under a surrounding environment different from that at choice response, provided the short-term memory hypothesis as one of telencephalic functions. Overmier and Patten [17], thereafter, emphasized the necessity of the reconsideration of the short-term memory hypothesis on the basis of the observation that telencephalon-ablated fish normally acquired learned responses regardless of interposition of a delay between choice response and reinforcement in a Y-maze paradigm when stimulus condition at choice response was the same as that at reinforcement delivery. This observation suggested that the causal factor of the impaired learning is not the interposed delay but the change in stimulus condition, and thus led Overmier and Patten to be critical of the short-term memory hypothesis. However, they did not discuss their other important observation that telencephalon-ablated fish showed more impaired learning than normal controls when stimulus condition at reinforcement delivery was neutral with respect to that at choice response. If the telencephalon was not important for short-term memory as they proposed, impaired learning would not have been observed in telencephalon-ablated fish under such stimulus conditions. From this, the possibility still remains that the telencephalon contributes to retention of short-term memory until delivery of reinforcement. Therefore, it seemed necessary to test the short-term memory hypothesis further. In the present study, I trained fish under varied stimulus conditions and showed that the difference in stimulus conditions between choice response and reinforcement delivery has a considerable influence on learning performance. No learned responses to choice stimuli were observed in either normal or telencephalon-ablated fish when they were trained with reward stimuli neutral with respect to choice stimuli. The percentages of correct responses in both groups did not increase and remained at the level of chance (about 50% correct responses) throughout all sessions. In contrast, clear learned responses to choice stimuli were observed in normal fish when they were trained with
reward stimuli including discriminable visual cue information which were different from choice stimuli. In this case, the fish first acquired learned responses to the reward stimuli and then associated the reward stimuli with choice stimuli because they showed high percentages of correct choices to the reward stimuli in test trials before acquiring learned responses to choice stimuli. These results suggest that cue information of discriminative stimuli at reinforcement delivery is critical for the acquisition of learned responses and it largely serves to retain the short-term memory which contributes to the association of reward stimuli with choice stimuli. This association process may be a cause of the difference in the rate of learning between discrimination trials under the same and under different stimulus conditions. Telencephalon-ablated fish showed greatly impaired learning performance to choice stimuli under conditions which differed between choice response and reinforcement delivery. This observation indicates that telencephalon-ablated fish have great difficulty in associating reward stimuli with choice stimuli and thus the telencephalon plays an important role in short-term memory. However, slightly increased correct responses were observed in later sessions. This indicates that unknown extratelencephalic attentional systems might exist. The present study did not control the delay time between the presentation of choice and rewarded stimuli. Further experiments without the intervening delay would support the short-term memory hypothesis. On the other hand, in previous studies [14,21] which examined other telencephalic functions, telencephalon-ablated fish showed impaired learning at a training rate sufficient for normal fish to learn colour but it was improved with more training trials. This result suggested that the telencephalon is supplementary in that it facilitates integration of neural events in extratelencephalic areas which are necessary for an instrumental process such as the utilization of changes in conditioned motivational reactions. In the present study, such telencephalic facilitation in utilization of secondary reinforcement should also be considered. The slightly impaired learning performances in telencephalon-ablated fish in the black vs. white discrimination under the same stimulus conditions and the compound stimuli discrimination with immediate reward delivery were probably the result of the loss of this facilitation function of the telencephalon. Considering this point, the greatly impaired learning performances in telencephalon-ablated fish in that discrimination under the different stimulus conditions probably resulted from the binal loss of the facilitation functions in not only utilization of secondary reinforcement but also short-term memory. Facilitation may be a fundamental function mediated by the teleost telencephalon for higher neural events.
K. Ohnishi / Beha6ioural Brain Research 86 (1997) 191–199
It is generally thought that the attentional mechanism, ‘selective attention’, is closely related to visual aspect selection in higher vertebrates. If this is the case with goldfish, telencephalon-ablated fish will show some affected behaviours in visual aspect selection reflecting the hypothesis that the telencephalon is an arousal or attentional centre. Contrary to this assumption, the results of the present study demonstrated that telencephalon-ablated fish showed no affected aspect selection behaviour. They selectively processed only one aspect in the compound stimuli discrimination as did normal fish. (It has been reported that normal goldfish selectively process only more easily discriminated aspects in a Y-maze instrumental conditioning regardless of a high rate of training when they were trained with visual compound stimuli composed of a more easily discriminated aspect and a more difficult aspect [15]. In the present study, pattern might be more discriminable than colour. Similar easiness to discriminate pattern has been reported by Ingle [7]). Thus, the telencephalon-related attentional function appears not to be critical for aspect selection in goldfish. This result supports the suggestion that aspect selection in goldfish is performed without paying attention to a specific aspect but is automatic in accordance with the intensity or salience of constituent aspects [15]. However, some unknown extratelencephalic attentional system(s) may also participate in aspect selection in teleosts.
Acknowledgements I thank Dr K.-I. Takahashi and Y. Ogawa for their useful comments, and Professors Y. Enoki and F. Motokszama and co-workers for their great encouragement.
References [1] Beauchamp, R.D. and Rowe, J.S., Goldfish spectral sensitivity: a conditioned heart-rate measure in restrained or curarized fish, Vision Res., 17 (1977) 617–624. [2] Fellows, B.J., Chance sequences for discrimination tasks, Psychol. Bull., 67 (1967) 87–92. [3] Flood, N.C., Overmier, J.B. and Savage, G.E., Teleost telencephalon and learning: an interpretive review of data and hypotheses, Physiol. Beha6., 16 (1976) 783–798. [4] Flood, N.B. and Overmier, J.B., Learning in teleost fish: role of the telencephalon. In P. Laming (Ed.), Brain Mechanisms of Beha6iour in Lower Vertebrates, Cambridge University Press, Cambridge, 1981, pp. 259–279. [5] Hallowitz, R.A., Woodward, D.J. and Demski, L.S., Forebrain activation of single units in preoptic area of sunfish, Comp. Biochem. Physiol., 40A (1971) 733–741.
199
[6] Hollis, K.L. and Overmier, J.B., The function of the teleost telencephalon in behaviour: a reinforcement mediator. In D.I. Mostofsky (Ed.), The Beha6iour of Fish and other Aquatic Animals, Academic Press, New York, 1978, pp. 137–195. [7] Ingle, DJ., Interocular transfer in goldfish: colour easier than pattern, Science, 149 (1965) 1000 – 1002. [8] Janzen, W., Untersuchungen uber Grosshirnfunktionen des Goldfishes (Carassius auratus), Zool. Jahrb. Abt. Allgem. Zool. Physiol. Tiere, 52 (1933) 592 – 628. [9] Krechevsky, I., ‘Hypotheses’ in rats, Psychol. Re6., 39 (1932) 516 – 532. [10] Lashley, K.S., Conditional reactions in the rat, J. Psychol., 6 (1938) 311 – 324. [11] Lea, S.E.G., Complex general process learning in mammalian vertebrates. In P. Marler and H.S. Terrace (Eds.), The Biology of Learning, Springer-Verlag, Berlin, 1984, pp. 373–398. [12] Neumeyer, C., On spectral sensitivity in the goldfish: evidence for neutral interactions between different ‘cone mechanisms’, Vision Res., 24 (1984) 1223 – 1231. [13] Northcutt, R.G. and Bradford, M.R., New observations on the organization and evolution of the telencephalon of actinopterygian fishes. In S.O.E. Ebbesson (Ed.), Comparati6e Neurology of the Telencephalon, Plenum Press, New York, 1980, pp. 41 – 98. [14] Ohnishi, K., Telencephalic function implicated in food-reinforced colour discrimination learning in the goldfish, Physiol. Beha6., 46 (1989) 707 – 712. [15] Ohnishi, K., Goldfish’s visual information processing patterns in food-reinforced discrimination learning between compound visual stimuli, J. Comp. Physiol. A, 168 (1991) 581–589. [16] Overmier, J.B. and Hollis, K.L., The teleostean telencephalon in learning. In R.E. Davis and R.G. Northcutt (Eds.), Fish Neurobiology, Vol. 2, The University of Michigan, Ann Arbor, 1983, pp. 265 – 284. [17] Overmier, J.B. and Patten, R.L., Teleost telencephalon and memory for delayed reinforces, Physiol. Psychol., 10 (1982) 74 – 78. [18] Overmier, J.B. and Savage, G.E., Effects of telencephalic ablation on trace classical conditioning of heart rate in goldfish, Exp. Neurol., 42 (1974) 339 – 346. [19] Peter, R.E. and Gill, V.E., A stereotaxic atlas and technique for forebrain nuclei of the goldfish, Carassius auratus., J. Comp. Neurol., 159 (1975) 69 – 102. [20] Reynolds, G.S., Attention in the pigeon, J. Exp. Anal. Beha6., 4 (1961) 203 – 208. [21] Savage, G.E., Temporal factors in avoidance learning in normal and forebrainless goldfish (Carassius auratus), Nature, 218 (1968) 1168 – 1169. [22] Savage, G.E., Some preliminary observations on the role of the telencephalon in food-reinforced behaviour in the goldfish, Carassius auratus, Anim. Beha6., 17 (1969) 760 – 772. [23] Savage, G.E., Behavioural effects of electrical stimulation of the telencephalon of the goldfish, Carassius auratus, Anim. Beha6., 19 (1971) 661 – 668. [24] Savage, G.E., The fish telencephalon and its relation to learning. In S.O.E. Ebbesson (Ed.), Comparati6e Neurology of Telencephalon, Plenum Press, New York, 1980, pp. 129–174. [25] Savage, G.E. and Swingland, I.R., Positively reinforced behaviour and the forebrain in goldfish, Nature, 221 (1969) 878– 879. [26] Schroeder, D.M., The telencephalon of teleosts. In S.O.E. Ebbesson (Ed.), Comparati6e Neurology of Telencephalon, Plenum Press, New York, 1980, pp. 99 – 115.
. .