Physiology & Behavior 80 (2003) 49–56

Involvement of the telencephalon in spaced-trial avoidance learning

in the goldfish (Carassius auratus)

Manuel Portavellaa,*, Cosme Salasa, Juan P. Vargasa,b, Mauricio R. Papinic

aLaboratorio de Psicobiologıa, Departamento de Psicologıa Experimental, Facultad de Psicologıa, Universidad de Sevilla,

c/ Camilo Jose Cela, s/n 41018, Sevilla, SpainbDepartment of Psychology, Bowling Green State University, Bowling Green, OH 43403, USA

cDepartment of Psychology, Texas Christian University, Box 298920, Fort Worth, TX 76129, USA

Received 22 January 2003; received in revised form 28 May 2003; accepted 19 June 2003

Abstract

Goldfish (Carassius auratus) received escape–avoidance training in a shuttle-response situation at a rate of a single trial per day. Widely

spaced training evaluates the ability of a discriminative stimulus to control an avoidance response in the absence of stimulus carry-over effects

from prior recent trials. In Experiment 1, master goldfish exhibited significantly faster avoidance learning than yoked controls. The results

suggest that the shuttle response was instrumentally acquired. Experiment 2 demonstrated a significant deficit in the acquisition of avoidance

behavior following ablation of the telencephalon. The implications of spaced-trial, telencephalon-dependent avoidance learning, as

demonstrated in these experiments for the first time, are discussed in the context of comparative research on instrumental learning in goldfish.

These results provide further support for the hypothesis that the fish telencephalon contains an emotional system that is critical for fear

conditioning.

D 2003 Elsevier Inc. All rights reserved.

Keywords: Avoidance learning; Emotion; Fear; Fish; Memory systems; Spaced-trial training; Telencephalon

1. Introduction

Learning theorists have traditionally assumed that medi-

ational states induced by aversive reinforcements such as

electric shock (i.e., fear) and the surprising omission of

appetitive reinforcers such as food (i.e., frustration) affect

behavior via similar mechanisms [1–4]. Gray [2] refers to

this as the fear = frustration hypothesis. For example, partial

food reinforcement and continuous food reinforcement

mixed with a partial shock punishment schedule both in-

crease resistance to extinction in rats relative to just contin-

uous food reinforcement [5]. In fact, many of the behavioral

(e.g., aggressive behavior) and physiological consequences

(e.g., pituitary–adrenal activation) of aversive events are

similar to those that occur following surprising reward

omissions (for a review, see Ref. [6]). Such similarities

suggest that at least some of the mechanisms underlying the

0031-9384/$ – see front matter D 2003 Elsevier Inc. All rights reserved.

doi:10.1016/S0031-9384(03)00208-7

* Corresponding author. Tel.: +34-954-55-44-01; fax: +34-954-55-17-

84.

E-mail address: portavel@us.es (M. Portavella).

conditioning of fear and frustration should covary in any

given species.

Teleost fish provide an opportunity to explore the

fear = frustration hypothesis in a comparative context. On

the one hand, there is good evidence that avoidance

learning is based on the acquisition of a mediational state

of fear in the goldfish. For example, prior Pavlovian

pairings of a warning stimulus (WS) with shock facilitate

subsequent avoidance acquisition relative to groups pre-

exposed to escape conditioning, shock only, and naive

controls [7]. Off-avoidance baseline, Pavlovian training

also endows the WS with the power to facilitate avoidance

responding trained separately [8]. Following Mowrer [3],

Zhuikov et al. [9] hypothesized that the Pavlovian contin-

gency (either in purely Pavlovian situations or that em-

bedded in the avoidance situation) leads to the acquisition

of an internal state of fear. Conditioned fear, in turn,

contributes to the development of the instrumental stimu-

lus–response association because of a reduction in fear

that follows the shuttle response. The model of Zhuikov et

al. [9] provides reasonably good quantitative predictions of

avoidance learning in goldfish under a variety of condi-

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–5650

tions. Furthermore, the complete ablation of the telenceph-

alon impairs avoidance learning in goldfish [8,10–14], a

result analogous to that observed in mammals with lesions

in various telencephalic structures, particularly the amyg-

dala [15–18]. However, the use of relatively massed

conditions of training (i.e., multiple trials per session) in

the cited experiments implies that the animal is responding

to the WS after having recently experienced electric shocks

and performed the shuttle response. Such salient events

may facilitate the acquisition of fear in their own right,

rendering ambiguous the theoretical claim that fear is

induced by the WS. Spaced-trial procedures (e.g., one trial

per day) can contribute to determine whether the WS can

elicit fear in the absence of carry-over effects from

previous trials.

On the other hand, the fear = frustration hypothesis cannot

explain the results of a variety of appetitive experiments

involving surprising reward omissions. A wide range of

experiments with goldfish and other nonmammalian verte-

brates, trained under widely spaced conditions (e.g., one trial

per day), have thus far failed to provide any evidence that

their instrumental behavior can be modulated by the associa-

tive reinstatement of prior frustrative outcomes [19–21].

Instead, the results of such experiments can be understood

in terms of a simple learning rule according to which

reinforcement increases the strength of the instrumental

response, whereas nonreinforcement decreases it. It seems

possible, therefore, that just as conditioned frustration fails to

control widely spaced instrumental performance in goldfish,

and a similarly spaced protocol of avoidance training may fail

to induce a conditioned response of fear. By contrast, widely

spaced acquisition of avoidance behavior would suggest a

dissociation between the basic underlying mechanisms of

appetitive (i.e., frustration) and aversive (i.e., fear) learning in

the goldfish [1–4].

The following experiments were designed to answer the

following questions. First, can goldfish learn an avoidance

response in a spaced-trial situation? Second, is such spaced-

trial avoidance learning dependent on the integrity of the

telencephalon?

2. Experiment 1: Master–yoked acquisition

The goal of Experiment 1 was to validate the spaced-

training procedure and to demonstrate the instrumental

nature of avoidance learning under such conditions using

a master–yoked design. Master–yoked pairs of goldfish

received one daily trial (one trial per day, one trial per

session) of shuttle-avoidance training. A visual cue (the WS)

was turned on in the tanks of both master and yoked

animals. A shuttle response by the master animal (swim-

ming over a barrier) during the presentation of the WS

turned it off and prevented the shock from being adminis-

tered in both the master’s and yoked’s tanks. Failure to

respond by the master led to the administration of an electric

shock also in both tanks. A similar master–yoked study

involving multiple trials per session demonstrated that the

shuttle response is acquired to a higher level by master

animals experiencing an instrumental contingency than by

yoked animals exposed to purely Pavlovian contingencies

[14].

2.1. Materials and methods

2.1.1. Subjects

Sixteen experimentally naive goldfish (Carassius aura-

tus), purchased from a local aquarium in Seville, served as

subjects. Animals were 9–12 cm in body length and lived in

pairs in tanks measuring 42 cm in length, 26 cm in width,

and 15 cm in height. Living tanks were cleaned every other

day and the water was kept at a constant temperature of 21

jC. Water was continuously aerated and filtered. The

aquarium room was subject to a 14:10-h light:dark cycle

(light from 07:00 to 21:00 h). Three pellets of Tetra-pond

fish were provided daily, at least 60 min after the training

session ended.

2.1.2. Apparatus

Goldfish received avoidance training in conventional

shuttle tanks. Two pairs of identical tanks were used, built

entirely with glass plates and with identical internal dimen-

sions (50� 14� 25 cm). Each box was filled with continu-

ously aerated water to a depth of 9.5 cm. In the center of each

box, there was a trapezoidal barrier measuring 10 cm at the

top, 18 cm at its base, and 7.5 cm in height. A shuttle

response required that the fish swam over the barrier by

tilting its body on one side. This response was recorded by

photocells located at each side of the barrier and 1.5 cm

above the barrier. The end glass plates on each extreme were

translucent and allowed for the presentation of a green light

(10 W) used as the WS. The lateral walls of the tank were

covered with metallic plates connected to a conventional

stimulator. Electric shocks (0.39 V ac/cm, 50 Hz sinusoidal

stimulation) were delivered through these plates by a stim-

ulator. A computer controlled all events and recorded the

response latency, that is, the time from the turning on of the

warning visual stimulus to either the occurrence of the

shuttle response, or the termination of the trial.

2.1.3. Training procedure

Prior to the training period, there were three sessions of

habituation to the shuttle tank in which no WS or shocks

were delivered. In the first session, animals were placed in

the shuttle tank during 30 min with the water level set at 6 cm

above the barrier. The following two sessions were 5 min

long and the water level was dropped to 2 cm above the

barrier. Animals were then randomly assigned to one of two

groups, master or yoked (n = 8), and administered one

training trial per day (for 40 training days).

During training, animals were transferred in pairs from

their home tanks to the training tanks. Each particular shuttle

Fig. 1. Results of the first experiment in a master–yoked procedure. Data

are grouped in blocks of 10 trials training (one trial training per day). (A)

Means of avoidance response percentage F S.E.M. (B) Means of escape

response percentage F S.E.M. (C) Means of responseless percentage

(absence of shuttle response) F S.E.M. (D) Means of shuttle response

latencies F S.E.M. All variables were grouped in sets of 10 daily trials.

Yoked group did not perform avoid responses since the stimuli presentation

was not contingent to animal behavior. However, its shuttle responses were

classified as escape or avoidance responses in relation to its latency for

comparative analysis (Mann–Whitney test: *P< .01, **P < .05).

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–56 51

tank served as a master tank for half of pairs and as a yoked

tank for the rest. A trial started with an interval of variable

duration, ranging between 2 and 4 min in length, at the end of

which the WS was turned on for a maximum duration of 40

s. The trial finished with the same variable interval. When

the master animal did not respond within 20 s of WS onset,

the electric shock was turned on for a maximum of 20 s. The

tanks were yoked such that a shuttle response detected

during these 40 s in the master tank turned off the WS

and/or shock in both tanks. Failure to respond by the master

animal resulted in a maximum latency of 40 s assigned for

that particular trial. Therefore, master–yoked pairs of ani-

mals were matched in their exposure to both WS and shock

in terms of both their absolute duration and the degree

temporal overlap during each trial.

The main dependent variable was the latency to respond,

defined as the time since the onset of the WS to either the

occurrence of the shuttle response or the termination of the

trial if there was no response in the case of master animals

and until the master response for the yoked subjects. The

latency defines the type of response in each trial. The

responses were classified in three different categories:

avoidance responses (shuttle responses occurring before

shock onset; the latency was less than 20 s), escape

responses (shuttle responses occurring during the shock;

the latency was between 20 and 40 s), and responseless

(trials in which no shuttle response was recorded; the

latency was equal to 40 s. In this case, the percentage of

escape responses and nonresponse was recorded and ana-

lyzed to control for the emergence of other learning

phenomena (i.e., learned helplessness) and to ensure that

differences in avoidance were not related to a decrease in

activity. Nonparametric statistical tests were used to evalu-

ate the results. Friedman tests were used to evaluate

behavioral trends across blocks of sessions. Mann–Whitney

tests were used for pairwise comparisons of groups. The

yoked animals did not perform true avoidance responses

since the stimulus contingencies were independent of their

behavior. However, shuttle responses in these animals were

classified as ‘‘avoidance’’ or ‘‘escape’’ responses depending

on whether their respective latency was lower or higher than

20 s.

2.2. Results

For the purpose of statistical analyses, the data on

avoidance, escape, absence of shuttle responses, and laten-

cies were grouped in blocks of 10 training days (one trial per

day). Fig. 1A shows the percentage of trials with an

avoidance response for each group. An analysis of behav-

ioral trends indicates that group master showed a significant

increase in avoidance responding across trial blocks (Fried-

man test X(3) = 20.51, P < .00011), whereas the avoidance

performance of group yoked remained steady (X(3) = 1.35,

P>.71). A group comparison indicated that by the end of

training, the performance of group master was significantly

higher than that of the group yoked (Mann–Whitney test:

U = 0, P < .001, for the last trial block).

Fig. 1B and C show the results in terms of escape

responses and responseless, respectively. Statistical signifi-

cances are presented in Table 1. The decline of group

master’s escape performance in the fourth block is corre-

lated with the development of a high level of avoidance

behavior.

Finally, response latencies are plotted in Fig. 1D. Where

group master showed a significant reduction in latency

values across trial blocks (X(3) = 66.35, P < .0001), group

yoked’s performance remained stable (X(3) = 1.74, P>.62).

Response latencies were significantly higher in group

yoked than in group master only in the last trial block

(U = 0, P < .001).

2.3. Discussion

Overmier and Papini [14] found that it took approxi-

mately 20 trials for the master animals to start exhibiting

behavioral improvement above the level shown by the

yoked animals. Under the present conditions, the master

animals showed improvement in shuttle performance above

the yoked animals after about 30 trials. However, these

studies differ in a number of parameters, including the

Table 1

Experiment 1. Master–yoked acquisition

Group master versus yoked (Mann–Whitney test analyses)

Response kind Trial blocks

1–10 11–20 21–30 31–40

Escape (%) U= 5;

P=.004

U = 0;

P=.001

U= 7;

P=.007

U = 22.5;

P=.29

Responseless (%) U= 11;

P=.024

U = 1;

P=.001

U= 0.5;

P=.001

U = 0;

P=.001

Statistical analysis of differences of escape and responseless percentage

between group master and yoked.

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–5652

maximum length of the WS (10 vs. 40 s in the present

experiment) and, of course, the administration of multiple

trials per session versus a single trial per session. In a single-

trial situation, the behavior is relatively uncontaminated by

carry-over effects from prior trials, including the WS, the

electric shock, response-elicited cues, and more extensive

exposure to the training context. Some, or even all, of these

cues may facilitate memory retrieval during the course of

acquisition and thus account for relatively better perfor-

mance [22]. The evidence presented here suggests that

goldfish can reactivate whatever information is necessary

for performing the avoidance response (e.g., mediational

state of fear [8]) essentially on the basis of the discrimina-

tive stimulus.

3. Experiment 2: Bilateral telencephalic ablation

If this spaced-trial instance of avoidance learning was

dependent on the same mechanisms that operate under

massed-trial conditions, then it should be vulnerable to the

bilateral ablation of the telencephalon [12]. In Experiment 2,

the space-trial performance of goldfish with telencephalic

lesions was compared to that of sham-operated and intact

controls. In addition, a group exposed only to the WS

provided a nonshocked baseline condition to determine the

extent to which the lesions impaired spontaneous levels of

shuttling behavior.

3.1. Materials and method

3.1.1. Subject and apparatus

Twenty-three experimentally naive goldfish served as

subjects in the present experiment. These animals were

between 9 and 12 cm in body length. They were obtained

from the same local source and maintained as described in the

previous experiment. The same living tanks and shuttle

avoidance tanks described before were also used in the

present experiment.

3.1.2. Surgical and histological procedures

Animals were anesthetized by immersion in a solution of

tricaine methanosulfonate (1:20,000). When movement had

ceased, the animal was placed in the surgical chamber where

it remained partially bathed in water with a constant

concentration of anesthetic. Lateral holders that maintained

the body still during surgery fixed the animal into position.

An adjustable tube was inserted into the animal’s mouth and

connected to a pump that provided a constant flow of water.

The concentration of the anesthetic in the water was kept at

1:20,000 during surgery.

The dorsal skin and skull were carefully removed and the

underlying fatty tissue removed by aspiration. The exposed

telencephalon was aspirated with a micropipette connected

to a manual vacuum system. Surgery was performed under

visual inspection by means of a binocular microscope.

Following the ablation, the piece of skull was replaced in

its original position and fixed with cyanocrilate glue. The

fish was returned to its home tank for a recovery period of 5

days. Sham operations were performed exactly as described,

except that the nervous tissue was not injured.

At the end of the experiment, fish with telencephalic

ablations and sham operations were deeply anesthetized

(1:5000) and perfused with 50 ml of 0.9% saline solution,

followed by 125 ml of fixative solution (10% formalin in

phosphate buffer 0.1 M, pH 7.4). The brains were removed

from the skull, inspected for the preliminary evaluation of

the ablation, and cut in 50 Am thick transversal sections for

histological analysis.

3.1.3. Training procedure

There were four groups in the present experiment. The

animals in the group ablation (n = 8) received a bilateral

ablation of the telencephalon before the start of the training,

whereas those in group sham (n = 4) received a sham

operation. The goldfish assigned to group intact (n = 4) were

not subjected to any surgery. Finally, the goldfish assigned to

group S-only (n = 8) were also intact but received different

training These animals were exposed to sessions that were

similar in every respect to the avoidance sessions of the other

goldfish, except no shock was ever delivered. This condition

was introduced to assess the extent to which the change in

illumination per se was reinforcing to these animals, as well

as to provide a no shock baseline to evaluate the extent of the

lesion effect. All training parameters were the same as in

Experiment 1. All animals were trained for 30 days (one trial

per day). The same dependent variables described in Exper-

iment 1 were recorded. Friedman tests were computed to

analyze behavioral trends across trial blocks. At each trial

block, overall group comparisons were based on the Krus-

kal–Wallis test and pairwise comparisons on the Mann–

Whitney test.

3.2. Results

One goldfish in group ablation died in the course of the

experiment, leaving the group with an n = 7. Optical inspec-

tion of the ablated brains indicated that telencephalic abla-

tions were complete (Fig. 2). The perfusion and brain

extraction were carried out 40 days after surgery (5 postsur-

Fig. 2. Nissl stained of sections of sham (A) and telencephalic (C). Level of brain sections brains (B). Abbreviations: Cer, cerebellum; Dc, dorsocentral telen-

cephalon; Dd, dorsodorsal telencephalon; Dl, dorsolateral telencephalon; Dm, dorsomedial telencephalon; Olf T, olfactory tracts; ON, optic nerves; Op tec, optic

tectum; PO, preoptic nucleus; PP, periventricular preoptic nucleus; Tel, telencephalon; VN, vagal nucleus; LH, lateral hypothalamus; V, ventral telencephalon.

Table 2

Experiment 2. Bilateral telencephalic ablation

Animals Avoidance (%)

Trials: 1–10 Trials: 11–20 Trials: 21–30

Sham 1 30 80 80

Sham 2 50 80 80

Sham 3 10 70 90

Sham 4 30 40 100

Intact 1 50 90 90

Intact 2 60 70 100

Intact 3 30 10 100

Intact 4 50 80 70

Individual data of avoidance percentage from sham and intact animals of

group control.

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–56 53

gery days, 30 training days, and 4–6 posttraining days). The

ablated brains presented the aspect of a swelled cylinder on

the optic nerves. This cylinder composed principally by

diencephalon is the product of a probable process of regen-

eration and fusion of two hemispheres. The good perfusion of

this brain part is difficult.

The histological analysis showed no damage to the

preoptic area and optic tracts. The optic tectum was also

spared. None of the sham animals exhibited any evidence of

damage to the telencephalon or optic tectum.

In this experiment, also, the percentage of escape re-

sponses and nonresponse were recorded and analyzed to

control the effects of the telencephalon ablation on the motor

capabilities of operated animals and to guarantee that the

differences in avoidance levels were not related with this

possible interference.

No statistically significant differences were observed

between sham and intact animals in responseless (Usz 3,

Ps=.13), escape (Usz 4, Ps>.23), and avoidance responses

(Mann–Whitney test: Usz 3, Ps>.13). Individual data of

avoidance percentages are showed in Table 2. Differences

were found only in the second block of latency values (first:

U = 2, P>.08; second: U = 0, P < .021; third: U = 2, P>.08).

The sham and intact groups showed similar learning levels

in avoidance, escape, and responseless; therefore, sham and

intact data are pooled into group control (n = 8) for statistical

analyses.

The main results of this experiment are presented in Fig.

3. Analyses of the behavioral trends exhibited by each of

these groups in each dependent variable yielded the follow-

ing results. The avoidance performance of these groups is

presented in Fig. 3A. Group control showed a significant

and progressive increase in avoidance behavior across trial

blocks (Friedman test: X(2) = 10.18, P < .007). In contrast,

groups ablation and S-only showed a low level of avoidance

responses in all trial blocks, but whereas the former group

was steady (X(2) = 0.5, P>.77), the latter exhibited a signif-

icant increase across trial blocks (X(2) = 6.44, P < .041).

Groups were also compared in terms of their avoidance

performance at the level of each trial block. Significant

group differences were found in all trial blocks (Kruskal–

Wallis tests: Xs(2)z 9.28, Ps < .01). Pairwise comparisons

indicated that the performance of group control was signif-

icantly higher than that of group ablation (Mann–Whitney

test: UsV 6, Ps < .01) and that of group S-only at each trial

block (Mann–Whitney test: UsV 10.5, Ps < .05). There

were no differences between groups ablation and S-only

(Mann–Whitney test: Usz 18.5, Ps>.2).

The escape performance of all the groups is plotted in

Fig. 3B and the trials without response in Fig. 3C. Statistical

significances are shown in Table 3.

Finally, the average response latencies for each group are

presented in Fig. 3D. In terms of behavioral trends, group

Table 3

Experiment 2. Bilateral telencephalic ablation

Response kind Trial blocks

1–10 11–20 21–30

All groups (Kruskal–Wallis test analyses)

Escape (%) X(2) = 15.52;

P=.0

X(2) = 9.46;

P=.007

X(2) = 12.9;

P=.002

Responseless (%) X(2) = 15.42;

P=.0

X(2) = 16.51;

P=.0

X(2) = 18.76;

P=.0

Group S-only versus control (Mann–Whitney test analyses)

Escape (%) U = 0.5;

P=.001

U = 29;

P=.746

U= 19.5;

P=.178

Responseless (%) U = 0;

P=.001

U = 0;

P=.001

U= 0;

P=.001

Group S-only versus telencephalic (Mann–Whitney test analyses)

Escape (%) U = 1.5;

P=.001

U = 4;

P=.005

U= 5;

P=.007

Responseless (%) U = 0;

P=.001

U = 0;

P=.001

U= 0;

P=.001

Group control versus telencephalic (Mann–Whitney test analyses)

Escape (%) U = 14.5;

P=.108

U = 5.5;

P=.008

U= 0.5;

P=.001

Responseless (%) U = 26.5;

P=.856

U = 24.5;

P=.629

U= 20;

P=.117

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–5654

control demonstrated a significant decrease in response

latency across trial blocks (Friedman test: X(2) = 49.17,

P < .0001). Group ablation also exhibited a significance

decrease in latency scores (X(2) = 6.26, P < .05), whereas

group S-only showed and minor but significant decrease in

response latency in the last block of trials (X(2) = 10.225,

P < .01). The analyses of latency performance indicated

significant group differences at each trial block (Kruskal–

Wallis tests: X(2)z 73.66, Ps < .001). Pairwise comparisons

showed that groups control and ablation exhibited differences

in the second and third trial blocks (first: U = 14, P>.1;

second: U = 0, P < .001; third: U = 0, P < .001), their perfor-

mance was different that of S-only group in all trial blocks

(control vs. S-only: UsV 1, Ps < .01; ablation vs. S-only:

UsV 3, Ps < .01).

3.3. Discussion

The impairment of avoidance learning in telencephalon-

ablated goldfish puts them at similar level to that of

goldfish exposed to S-only training. Moreover, previous

experiments showed that telencephalon-ablated goldfish

can be conditioned in a food-reinforced color discrimina-

tion task [23–25]. Therefore, such an impairment cannot

Results from statistical analysis of the escape and responseless percentage

between group control, ablation, and S-only.

Fig. 3. Results from the second experiment. Data are grouped in blocks of

10 trials training (one trial training per day). (A) Means of avoidance

response percentage F S.E.M. (B) Means of escape response percentage

F S.E.M. (C) Means of responseless percentage (absence of shuttle

response) F S.E.M. (D) Means of shuttle response latencies F S.E.M. All

variables were grouped in sets of 10 daily trials. S-only group did not

perform avoid and escape responses since unconditioned stimulus was not

present. However, its shuttle responses were classified as escape or

avoidance responses in relation to its latency for comparative analysis

(Kruskal–Wallis test: * *P < .01).

be attributed to an interference in the goldfish’s visual

capacities.

Ablation of the telencephalon impairs acquisition of the

spaced-trial shuttle avoidance response just as it does in the

massed trial situation [12]. This similarity is consistent with

the hypothesis that these two training protocols engage

similar telencephalic mechanisms.

4. General discussion

These experiments demonstrate robust, telencephalon-

dependent, spaced-trial avoidance learning in goldfish. To

the best of our knowledge, this is the first demonstration of

spaced-trial (i.e., one trial per day) avoidance learning in any

species. This result has implications for the fear = frustration

hypothesis [2] when integrated with the results of appetitive

spaced-trial experiments. In mammals, such spaced-trial

experiments with food reinforcers involving the surprising

omission of rewards, or the surprising reduction in reward

magnitude or quality, lead to emotional responses collective-

ly referred to as frustration [1,2]. Goldfish, as well as other

species of bony fish, amphibians, turtles, and pigeons, have

thus far provided no evidence that requires the assumption of

an emotional mediational state of frustration [19–21]. For

example, goldfish trained with large rewards acquire a

swimming alleyway response faster than goldfish trained

with small rewards; however, when shifted from the large to

the small reward magnitude, they show little or no change in

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–56 55

behavior [26]. That is, they provide no evidence of succes-

sive negative contrast—a deterioration of behavior that

follows a downward shift in reward magnitude in rats [27].

Such frustration-related instrumental effects in rats and other

mammals can be reduced by treatment with benzodiazepine

anxiolytics [28,29], induce pituitary–adrenal activation

[30–32], and support escape conditioning [33]. These are

properties shared by frustration- and fear-inducing situations

that suggest some degree of overlap in underlying mecha-

nisms. The fact that spaced-trial avoidance learning is readily

obtained in a species—the goldfish—that shows no spaced-

trial contrast and related phenomena suggests a fundamental

asymmetry in the evolution of the neural systems involved in

the associative reinstatement of fear and frustration. The

potential dissociation between fear and frustration in the

goldfish could be approached in at least two directions. One

possibility is that there is a dissociation of the basic mech-

anisms subserving appetitive and aversive learning in gold-

fish. The other possibility is that the experiments with

goldfish involving reward omission manipulation were not

particularly sensitive to detect the effects of frustration on

instrumental behavior.

The present results are also important to understand the

deficit in avoidance learning induced by telencephalic

lesions. The spacing of the trials leaves little doubt that

goldfish can retrieve the appropriate information to perform

the shuttle response purely on the basis of the WS. By

contrast, in relatively mass conditions of training, there are

additional sources of control over the behavior, including

the recent presentation of shocks and the performance of the

shuttle response. Recent powerful events like these may

enhance retrieval and allow the information to carry-over

across trials. The present results suggest that telencephalic

lesions affect avoidance learning by impairing the retrieval

of an anticipatory fear response by the WS or by interfering

with the ability of an internal state of fear to induce an

avoidance response. In light of the success of the present

procedure, it would be interesting to determine if transfer of

control from a Pavlovian signal for shock to the shuttle

avoidance response would occur in ablated goldfish trained

under spaced conditions. A previous experiment showed

that lesioned animals shuttled in the presence on the

Pavlovian signal trained off-baseline just as much as normal

controls [8]. However, testing of the Pavlovian signal was

carried out after five avoidance conditioning trials involving

both shuttling responses and exposure to the electric shock.

The spaced-trial procedure used in the present experi-

ments can contribute to clarify the role of the telencephalon

in avoidance learning. The conditioned emotional response

in mammals is acquired in a few trials. The spaced-trial

avoidance learning procedure used in the present experi-

ments shows a similar effectiveness. For example, 10 trials

were enough to acquire the avoidance response in Experi-

ment 2. Additional experimental evidence shows that the

total or partial ablation of the telencephalon in goldfish

produces specific impairments in spatial learning

[23,25,34–36]. The specificity of these effects is similar

to what has been described in mammals with hippocampal

lesions [17,18,37]. These results suggest the presence of a

telencephalon-based spatial learning system in the goldfish.

Recent work provided evidence of the presence of two

memory systems in the teleost telencephalon based on

discrete pallial regions: dorsomedial (emotional system)

and dorsolateral (spatial system) [38,39]. The absence of

carry-over effects, the rapid acquisition of avoidance

responses, and the impairment exhibited by such avoidance

behavior in animals with telencephalic ablation provide

further support for the hypothesis that the fish telencepha-

lon contains an emotional system that is critical for fear

conditioning.

Acknowledgements

This research was supported by grants from the Spanish

D.G.E.S. (PB.96-1334) and Junta de Andalucıa. The

participation of M.R.P. in the project was supported by

grants from Junta de Andalucıa, Spain. We thank G.

Labrador and M. T. Gutierrez for technical assistance.

References

[1] Amsel A. Frustration theory. Cambridge (UK): Cambridge Univ

Press; 1992.

[2] Gray JA. The psychology of fear and stress. Cambridge (UK): Cam-

bridge Univ Press; 1987.

[3] Mowrer OH. On the dual nature of learning—a re-interpretation of

‘‘conditioning’’ and ‘‘problem solving’’. Harvard Educ Rev 1947;17:

102–48.

[4] Wagner AR. Frustrative non-reward: a variety of punishment? In:

Campbell BA, Church RM, editors. Punishment and aversive behav-

ior. New York: Appleton-Century-Crofts; 1969. p. 157–81.

[5] Brown RT, Wagner AR. Resistance to punishment and extinction

following training with shock or non-reinforcement. J Exp Psychol

1964;68:503–7.

[6] Papini MR, Dudley RT. Consequences of surprising reward omis-

sions. Rev Gen Psychol 1997;1:175–97.

[7] Gallon RL. Effects of pre-training with fear and escape conditioning

on shuttle-box avoidance acquisition by goldfish. Psychol Rep 1972;

31:919–24.

[8] Overmier JB, Starkman N. Transfer of control of avoidance behavior

in normal and telencephalon ablated goldfish (Carassius auratus).

Physiol Behav 1974;12:605–8.

[9] Zhuikov AY, Couvillon PA, Bitterman ME. Quantitative two-process

analysis of avoidance conditioning in goldfish. J Exp Psychol, Anim

Behav Processes 1994;20:32–43.

[10] Flood NB, Overmier JB, Savage GE. The teleost telencephalon and

learning: an interpretative review of data and hypotheses. Physiol

Behav 1976;16:783–98.

[11] Flood NB, Overmier JB. Learning in teleost fish: role of the telence-

phalon. In: Laming PR, editor. Brain mechanisms of behavior in lower

vertebrates. Cambridge: Cambridge Univ. Press; 1981. p. 259–79.

[12] Hainsworth FR, Overmier JB, Snowdon CT. Specific and permanent

deficits in instrumental avoidance learning following forebrain abla-

tion in the goldfish. J Comp Physiol Psychol 1967;63:111–6.

[13] Overmier JB, Hollis KL. Fish in the think tank: learning, memory and

M. Portavella et al. / Physiology & Behavior 80 (2003) 49–5656

integrated behavior. In: Kesner RP, Olton DS, editors. Neurobiology

of comparative cognition. Hillsdale (NJ): Erlbaum; 1990. p. 204–36.

[14] Overmier JB, Papini MR. Factors modulating the effects of teleost

telencephalon ablation on retention, relearning, and extinction of in-

strumental avoidance behavior. Behav Neurol 1986;100:190–9.

[15] Aggleton JP. The amygdala: neurobiological aspects of emotion,

memory, and mental dysfunction. New York: Wiley-Liss; 1992.

[16] Aggleton JP. The amygdala: a functional analysis. Oxford: Oxford

Univ. Press; 2001.

[17] Eichenbaum H, Otto T, Cohen NJ. The hippocampus. What does it

do? Behav Neural Biol 1992;57:2–36.

[18] O’Keefe J, Nadel L. The hippocampus as a cognitive map. Oxford:

Clarendon Press; 1978.

[19] Bitterman ME. The comparative analysis of learning. Science 1975

(May 16);188:699–709.

[20] Papini MR. Role of reinforcement in spaced-trial operant learning in

pigeons (Columba livia). J Comp Psychol 1997;111:175–97.

[21] Papini MR, Muzio RN, Segura ET. Instrumental learning in toads

(Bufo arenarum): reinforcer magnitude and medial pallium. Brain

Behav Evol 1995;46:61–71.

[22] Miller RR. Effects of inter-trial reinstatement of training stimuli on

complex maze learning in rats: evidence that ‘‘acquisition’’ curves

reflect more than acquisition. J Exp Psychol, Anim Behav Processes

1982;8:86–109.

[23] Lopez JC, Bingman VP, Rodrıguez F, Gomez Y, Salas C. Dissociation

of place and cue learning by telencephalic ablation in goldfish. Behav

Neurosci 2000;117:687–99.

[24] Ohnishi K. Telencephalic function implicated in food-reinforced col-

our discrimination learning in goldfish. Physiol Behav 1989;46:

707–12.

[25] Salas C, Broglio CR, Rodrıguez F, Lopez JC, Portavella M, Torres B.

Telencephalic ablation in goldfish impairs performance in a ‘‘spatial

constancy’’ problem but not in a cued one. Behav Brain Res 1996;79:

193–200.

[26] Lowes G, Bitterman ME. Reward and learning in the goldfish.

Science 1967 (July 28);157:455–7.

[27] Crespi LP. Quantitative variation of incentive and performance in the

white rat. Am J Psychol 1942;55:467–517.

[28] Feldon J, Gray JA. The partial reinforcement extinction effect after

treatment with chlordiazepoxide. Psychopharmacology 1981;73:

269–75.

[29] McNaughton N. Effects of anxyolitic drugs on the partial reinforce-

ment effects in runway and Skinner box. Q J Exp Psychol 1984;36B:

319–30.

[30] Coe CL, Stanton ME, Levine S. Adrenal responses to reinforcement

and extinction: role of expectancy versus instrumental responding.

Behav Neurosci 1983;97:654–7.

[31] Dantzer R, Arnone M, Mormede P. Effects of frustration on behaviour

and plasma corticosteroid levels in pigs. Physiol Behav 1980;24:1–4.

[32] Osborne B, Silverhart T, Markgraf C, Seggie J. Effects of fornix

transection and pituitary–adrenal modulation on extinction behavior.

Behav Neurosci 1987;101:504–12.

[33] Daly HB. Reinforcing properties of escape from frustration aroused in

various learning situations. Psychol Learn Motiv 1974;8:187–231.

[34] Lopez JC, Broglio CR, Rodrıguez F, Thinus-Blanc C, Salas C. Re-

versal learning deficit in a spatial task but not in a cued one after

telencephalic ablation in goldfish. Behav Brain Res 2000;109:91–8.

[35] Rodrıguez F, Lopez JC, Vargas JP, Gomez Y, Broglio C, Salas C.

Conservation of spatial memory function in the pallial forebrain of

reptiles and ray-finned fishes. J Neurosci 2002;22(7):2894–903.

[36] Salas C, Rodrıguez F, Vargas JP, Duran E, Torres B. Spatial learning

and memory deficits after telencephalic oblation in goldfish trained in

place and turn maze procedures. Behav Neurosci 1996;110:965–80.

[37] Olton DS, Meck WH, Church RM. Separation of hippocampal and

amygdaloid involvement ion temporal memory dysfunction. Brain

Res 1987;404 [180B188].

[38] Portavella M. Implication of different forebrain areas in two-way

active avoidance conditioning in goldfish (Carassius auratus). Un-

published PhD dissertation. Sevilla: University of Sevilla; 2000.

[39] Portavella M, Vargas JP, Torres B, Salas C. The effects of telence-

phalic pallial lesions on spatial, temporal, and emotional learning in

goldfish. Brain Res Bull 2002;57(3/4):397–9.