A M . ZOOLOGIST, 6:629-641 (1966).
Operant and Classically-Conditioned Aggressive Behavior
in Siamese Fighting Fish
TRAVIS THOMPSON
Department of Psycliiatry and Neurology
University of Aiimiesota Medical School, Minneapolis, Minnesota
SYNOPSIS. Male Betta splendens were conditioned to emit an operant response sequence
reinforced by presentation of a model of another male Betta in aggressive display (Thompson, 1963). The color of the model with respect to the color of the subject co-varied
with the rate of operant response (Thompson and Sturm, 1965). In another experiment,
unconditioned aggressive display behavior elicited by the mirror image of a male Betta
was brought under the control of a previously ineffective stimulus by classical conditioning. Relative rates of acquisition of four components of the display were compared. A
discriminative conditioning procedure, using two different colors of light as the CS,
revealed that the response was elicited specifically by the CS (Thompson and Sturm, 1966).
The dramatically aggressive behavior of
Siamese fighting fish (Betta splendens) has
long interested behavioral scientists. Detailed observations and investigations by
Lissmann (1932), Smith (1937), Braddock
and Braddock (1958), Foselius (1957), and
others, have elucidated major classes of
variables controlling unconditioned aggressive display by male Bettas. One of the
various factors known to affect the occurrence of aggressive display is the presentation of a visual image of anouier male
Betta. This stimulus has been uniformly
demonstrated to be a releaser for aggressive
display. If a male Betta that has been exposed to the visual image of another male
along one wall of an aquarium is observed
for some time following exposure, one will
see the subject return repeatedly to the area
where the "intruding" male Betta was encountered, and, as the subject periodically
swims along the aquarium wall where the
intruder was presented, the fish may display
aggressively. Such observations suggested
that presentation of this releasing stimulus
(i.e., the visual image of another male
Betta) might act as a reinforcing consequence for an operant response. And further, it seemed that the display behavior
This research was supported in part by research
grants GB-1912 and GB-4S18 from the National
Science Foundation, and has been reported elsewhere
by Thompson (1963) and Thompson and Sturm
(1965o, 19656).
Experiments II and III reported here were done
in collaboration with Thomas Sturm.
(629)
(behavior occurring in the absence of the
appropriate unconditioned releasing stimulus), might come under the control of other
stimuli which had been associated with the
releaser. These considerations led to a
series of experiments in which the visual
image of a male Betta was conceived of
both as a reinforcer for an operant response
and as an unconditioned eliciting stimulus
for aggressive display. In addition, the
tendency for various components of the
display to become conditioned to other environmental stimuli via classical conditioning was also explored.
Male Betta splendens of domestic stock,
6 to 10 cm long, purchased from local
aquarium supply stores, served as subjects
in all the following experiments. The fish
were fed a diet of tubifex worms and/or
frozen brine shrimp once a day throughout
the course of the experiments. When not
in the experimental test chamber, each fish
was maintained in a 2-liter pyrex beaker
filled with conditioned tap water at 26° to
29°C.
EXPERIMENT I
The test tank was 30 by 20 by 25 cm and
had transparent Lucite walls (Fig. 1). A
121/^-cm Lucite ring with a 71/^-cm circular
aperture was suspended from a translucent
Lucite tank lid, and a beam of light, focused
across the aperture of the ring, fell on a
photoelectric cell imbedded in the plastic
of the opposite side of the ring. A response
630
TRAVIS THOMPSON
was recorded when the fish swam all the
way through the ring, first breaking the
beam of light, then allowing the light beam
to fall back on the photoelectric cell when
the subject's transit of the ring was complete. The back wall of the tank was used
for presentation of visual stimuli.
Three stimuli were examined sequentially for reinforcing properties. The first
was a one-way mirror attached to the back
wall of the tank. The mirror was presented
for 20 sec by shutting off a 75-watt lamp
behind it. Subsequently, a moving model
of a male B. splendens in aggressive display
was tested as a reinforcing stimulus and,
finally, a stationary model of a male B.
splendens in aggressive display was tested.
The moving model was presented across
the wall of the tank (30 cm) for 20 sec, and,
after conditioning and extinction with this
stimulus as a reinforcer, the same model
was presented in a stationary position for
20 sec.
Figure 2 presents the composite curves
of the number of responses per 24-hr session
for the four fish throughout the 60 days of
the first part of this experiment. The operant level was determined on the first two
days (10 to 25 responses per session), and
the acquisition of the response reinforced
by mirror presentation after the first two
days may be seen in the figure. This behavior approached stabilization after 12 to
14 days at 600 to 700 responses per session.
Extinction of the response reinforced by
mirror presentation began on day 17. Im-
A.
10"
10"
a. 6 watt, IIOv.a.c. lamp
fat response ring
Hue-green
c one-way glass
d. transluscent
e. 75 watt,
red
lucite
IIOv.a.c. lamp
f. 45° front surface mirror
^
g. 1/2"
lucite
tubing
h. cl-404 glass photocell
FIG. 1. Apparatus used for studying the reinforcing
effects of visual stimuli in Betta splendens. All dimensions are in inches. (A) Side view of test tank
illustrating the relationship of the one-way mirror
(c) and lamp (e) to the response ring (b). (B)
Front view of test tank facing the one-way mirror.
The beam of light from the 6-watt bulb (a) is directed by a 45° mirror (f) across the aperture of the
response ring (b) falling upon the photo cell (h).
(C) Model of male B. splendens used as reinforcer.
631
AGGRESSIVE BEHAVIOR IN FISH
ACQUISITION
MIRROR
UGHT
EXT
MIRROR CHANGE
MOVING
MODEL
EXT.
MMODEL
EXT.
ST. MODEL
STATIONARY
MODEL
700
700 "
g
in
UlU
600
500
•600
500
LU
400
400LU
'I
300
Q_
to
•300
200"
III
100
- T
200
II
100
T
1
1
1
10
20
30
40
SUCCESSIVE 24 HOUR SESSIONS
1
50
6O
FIG 2. Total number of responses per 24-hour session for four fish, expressed as a range. After the
first two sessions of operant level, the sequence of
procedures indicated at the top of the graph is described in the text,
mediately after extinction began, the mean
response rate began to drop (from 650 responses on day 16 to 125 responses on day
21). For the next five days (day 22 to day
26), the light was shut off after a response,
but the mirror was removed. Thus, a
change in illumination alone was substituted for the illumination change plus mirror presentation. The response rate increased slightly on day 22 but returned to
25 to 100 responses for the four succeeding
days, indicating that the mirror presentation, not the change in illumination, was
responsible for maintaining behavior.
Beginning with day 27, each response
was followed by presentation of a model
of a male B. splendens in aggressive display
that moved across the wall of the tank.
The number of responses for this reinforcer
increased sharply over 10 days, stabilizing
at approximately 575 responses per session.
Extinction began on day 28, and reached
a level of 10 to 50 responses within six days.
The final manipulation involved presentation of the same model in a stationary
position along the wall of the tank for 20
sec. The model was then removed until
another response was made. Responding
for this reinforcer reached an asymptotic
rate in ten sessions (150 to 200 responses)
and was extinguished to the operant level
of 10 to 25 responses.
These findings parallel Forselius' detailed description (1957) of the use of models and mirror to release a variety of agonistic behaviors in anabantid fishes, including Betta. These findings are consistent
with Forselius' observations on the ability
of these stimuli to act as releasers, and they
suggest that the relative positive-reinforcing
properties of visual stimuli, similar to the
visual properties of a male fish in aggressive
display, co-vary with the degree to which
stimuli will evoke unlearned display.
EXPERIMENT II
The above findings led us to wonder
exactly what visual properties of a male
Betta in aggressive display make it an effective reinforcer for an operant response.
Therefore, in the next experiment, the
color of the visual reinforcer was varied
and the relative reinforcing properties of
these stimuli compared. Subjects were three
male B. splendens: Fish A was red and 6
632
1.
2.
3.
4
5.
6.
7.
8.
9.
10
II.
12
13
14
TRAVIS THOMPSON
6 Volt Light Bulb
Focusing Lens
Reflecting Mirror
Photoelectric Cell
Gate (o.b.and c)
Plexiglass Partition
Model
Fin Illumination Light Bulb
Model Electric Train Engine
Car Carrying Model
Track
Aquarium Tank
Opaque Screen
Live Male Betta (Subject)
FIG. 3. Side and top views of experimental apparatus for operant conditioning of Siamese fighting
fish. The model electric train (9 and 10) used to
present the reinforcing model (7) can be seen in the
stopped position. The subject (14) is about to pass
through gate Sc which will cause the model to move
quickly to position A, where it will stop for 4 sec,
then continue slowly to position B, where it will
increase speed as it circles around the tank on the
track (11) stopping again until the next response
sequence (5a -> 5b-» 5c) has been completed. All
dimensions are in cm.
cm long, Fish B was blue and 6 cm long,
and Fish C was blue with a green dorsal
fin and was 6.5 cm long. Each fish was
tested for one hour at approximately the
same time each day, and the one-hour
session began with the first reinforced
response.
A fixed sequence of responses was required of the fish so that they would be in
a proper orientation vis-a-vis the visual
stimulus when presentation occurred. Thus,
the fish had to swim through three gates in
an underwater maze in a prescribed order.
(The gate positions are shown in Fig. 3.)
At each gate, a beam of light focused into
the tank on a small mirror ( ^ by ?4 c m )
was reflected back out onto a photoelectric
cell. The mirrors were small enough not
to elicit display from the fish to its image.
By passing through a gate, the fish interrupted the light beam and closed a relay,
and, when all three relays were closed in
proper order, the visual reinforcer was
presented. One of six colored models of
a male Betta in aggressive display was
mounted on an electric train in such a way
that the model, but not the train, could be
seen by the fish. The model appeared in
lateral display on the left side of the tank,
stopped in front of the third gate for four
seconds (which allowed the fish enough
time to display aggressively), and then
moved slowly across the front of the tank
from left to right. The model traveled
across the front of the tank in four seconds,
allowing the fish to follow and to continue
presenting to the stimulus. After moving
beyond the end of the tank, the train
changed speed, circled back, and stopped
at the left side of the tank, out of the subject's sight. In order to see the model
again, the fish was required to renegotiate
the maze.
The reinforcing stimuli were six models
made of painted balsa wood (body) and
colored cellophane (fins). Scales, fin spikes,
633
AGGRESSIVE BEHAVIOR IN FISH
120
2
10
PERCENTAGE OF OPERANT RESPONSES FOLLOWED
BY AGGRESIVE DISPLAY
Bella A
(Red)
, Green Model
Model
Sessions
°
Betto A
(Red)
Betta B
(Blue)
Bet to C
(Blue-Green)
O
Si
80
LJ
CC
bj
>
<
60
I
40
20
4
6
8
160
10
1-5
Red
6-10
1-5
Blue
6-10
Green
1 - 5 6-10
96
100
97
97
100
97
100
98
100
100
100
100
99
100
95
95
92
85
Red Model
Bella C
(Blue-Green)
SESSIONS
O
Si 100
100
C
80
Blue Model
60
60
rr
g
40
l 20
4
6
8
SESSIONS
4
6
8
SESSIONS
FIG. 4. Cumulative records of complete response
sequences leading to presentation of the red, green,
and blue models for each fish. The table indicates
the percentage of reinforced operant responses which
were followed by aggressive display for each fish,
and fin rays were drawn on the model with
black ink, based on Lissmann's description
of the lateral view of male Bettas (Lissmann, 1932). A 6-v lamp was illuminated
behind the model as it passed in front of
the tank to increase the luminescence of
the fins, and the ceiling light over the tank
illuminated the body. The colors of the
models were matched with colors in Munsell's Color System Atlas, and all models
showed the full aggressive display. Models
were run for two blocks of five sessions
each to determine their reinforcing properties. The following order was employed
to control for a possible order effect—blue,
led, green, green, red blue—and two models
of each color were used to control for slight
differences in body shape due to hand carving; one model of each color was used in
the first block and the second model of each
color was used in the second five-session
block.
Figure 4 shows the cumulative number
of responses by the three fish that led to
presentation of each model. The separation between session 5 and session 6 on
the graph indicates that the last five sessions were not run continuously with the
first five. As the graph indicates, the red
fish (A) responded at the highest rate when
the green model was the reinforcing stimulus, at an intermediate rate when the blue
model was used, and at the lowest rate when
the red model was used. The blue fish (B)
634
TRAVIS THOMPSON
Fish A (Red)
FishS(Bkie)
FishC(Blue-Green)
0 L = 008RSP/l0 minutes
(Calculated from 4-lhf Sessions)
OL'O25HSP/IOmim,les
(Calculated from 4-lhr Sessions]
0 L - 016 RSP/IO minutes
(Calculated Irom 3-1 nr Sessions)
Q 25
j
o
5
20
30
40
50
40
60
50
20
30
40
50
60
10 Minute Intervols
FIG. S. Total responses during successive 10-min
periods for each model and all three subjects. Oper-
ant levels are indicated on each set of graphs for
each subject (O.L.).
responded most with the red model, intermediately with the green model, and least
with the blue model. The blue and green
fish (C) responded in approximately the
same order as Fish B, except that the differential rate between the red and green models was greater than between the green
and blue models. For Fish B, the differential rates were approximately equal among
the three model colors.
A more detailed presentation of response
distribution for each fish during the 60minute sessions is shown in Figure 5. Most
responding occurred during the first 10
minutes of each session, and response rates
tapered off to rates of from four to ten
times the operant level throughout the
remainder of each session. The overall rate
differences in responses leading to the three
different colored models may be more
clearly seen in the individual curves. The
greatest consistent differences were between
the reinforcer that maintained the highest
rate and the two weaker reinforcers. However, this was less evident with Fish B's performance, whose three response curves are
much closer together, than the response
curves produced by the other two subjects.
Each fish emitted the lowest number of
responses to the model which had the color
most like its own, and Fish B and C emitted
the highest rate of response to the color
least like their own. The rate of respond-
ing to the sets of same-color models (used
as controls in the second block of five sessions) was similar to the rate of responding
to the set of models used in the first block
of five sessions, as can be seen in the data
presented earlier.
The table in Figure 5 indicates the percentage of time that the fish displayed aggressively in the presence of the model. All
displayed at more than 95% of the reinforcement presentations, except for the
green-blue fish (Fish C), which did not
display at all in the presence of either of
the green models.
REMARKS
In the two foregoing experiments, the
highest levels of response were maintained
by reinforcement with the mirror image
of a male Betta. Intermediate levels of: responding were maintained by a moving
model of a male Betta in aggressive display,
and the lowest levels of responding were
maintained by the same model in a stationary position. These data suggested that
the subject's coloration with respect to
the model's coloration might be an important variable determining operant reinforcing strength. In the second experiment, it
appeared that a fish would respond least
when the coloration was like its own.
These findings ostensibly contradict results
635
AGGRESSIVE BEHAVIOR IN FISH
Q-Wooden comportment
b-IOO wott light bulb
c-Omon-skin paper
d-Red cellophane
e - 7 '/z wott light bulb
f - Aquarium tank
g - One way mirror
cally conditioned, and we first sought to
establish the feasibility of a light as a conditioned stimulus. A discriminative conditioning procedure was used in this experiment in order to demonstrate that the four
aggressive-display components were specifically elicited by the conditioned stimulus.
This procedure eliminated the possibility
that our results were the product of sensitization.
Part 1
FIG. 6. Observer's view of apparatus for classically
conditioning aggressive display in Betta splendens.
obtained in the first experiment, since the
color of the mirror image was the same as
that of the subject and should, therefore,
have been associated with the lowest rate
of responding. However, the mirror image
is a highly complex stimulus, composed of
all the movements and color changes which
occur during the aggressive-display sequence, and, because of the complex nature
of the mirror image, the second experiment
dealt solely with one aspect of the visual
reinforcing stimulus, i.e., coloration.
E X P E R I M E N T III
In the second experiment, it was found
that male Bettas exhibited aggressive display 95% of the time in the presence of a
visual reinforcer. We noticed that some
subjects began to display following the
emission of the operant, even before the
reinforcing model was in view. This suggested that the display was elicited by stimuli which had been repeatedly paired with
model presentation.
Adler and Hogan (1963) made a similar
observation when they classically conditioned gill-cover erection, one component
of this complex display. They used a weak
electric shock as the conditioned stimulus
and a mirror as the unconditioned stimulus.
Our third experiment was therefore designed to study the relative rates at which
four components of the unconditioned
aggressive-display sequence could be classi-
A red stimulus-light was presented for
10 sec before and at the beginning of a
15-sec mirror-presentation along one wall
of an aquarium tank. Two compartments,
each enclosing two 100-watt light bulbs,
were placed 14 cm apart, allowing just
enough space for the aquarium tank (12
by 25 by 23 cm) to rest between the compartments (Fig. 6). The compartment walls
adjacent to the aquarium were made of
translucent onionskin paper, one of which
was covered by a one-way mirror. Turning
off the light behind the one-way mirror
and turning on the light at the opposite
side of the tank made the mirror highly
reflective. Reversing the illumination terminated the mirror effect. A 71/^-watt bulb
behind a sheet of red cellophane stretched
across the end of the tank (12 by 12 cm)
illuminated the interior of the tank with
bright-red light during presentation of the
conditioned stimulus. The subject's behavior was observed through the opposite
end of the tank. Red-light and mirroipresentation were programmed by automatic electronic switching and timing devices. Data were recorded by observing the
four display components during red-light
presentation.
Each fish was tested daily for 10 trials
programmed at random intervals varying
from 30 to 240 seconds, with a mean intertrial time of 120 sec. This aperiodic intertrial procedure was used to minimize
development of temporal discrimination.
Because aggressive display occurred occasionally, immediately after the home tank
was moved, in the absence of an eliciting
stimulus, the home tank was transferred to
636
TRAVIS THOMPSON
A
FRONTAL APPROACH
UNDULATING MOVEMENT 2
g
UNDULATING MOVFMFNT
^P®""
MAXIMUM DISPLAY
FIG. 7. Four photographic views of a single male
Betta splendens illustrating four components of the
complex aggressive display seauence from different
observational positions. A. Frontal approach, viewed
head-on, revealing fin erection and gill erection (GE).
B and C. Two successive
rapidly undulating body
mum display in frontal
Notice that all fins and
the test apparatus 10 min before each session. The observer sat at the end of the
tank and recorded the occurrence or nonoccurrence of the display for each trial on
the standard score sheets. The four following display components could be distinguished: (1) frontal approach to the mirror
and/or to the red stimulus light, (2) fin
erection, (3) gill-cover erection, and (4)
undulating movements (described in detail
by Forselius, 1957).
Figure 7 presents four photographic views
of a male Betta that illustrate the four
aggressive-display components. The upper
left photograph shows the subject in frontal
approach with the gill covers erected (GE).
The upper right and lower left photographs
illustrate the undulating swimming movements, in which the fish first orients its
head toward the opponent with its body
curved slightly away then bends its body
in the opposite direction with a rapid thrust
of the tail. The lower left photograph also
shows the gill covers erected, revealing the
bright-red gill membranes. The maximum
fin erection, as well as the gill erection, can
be seen in the lower right photograph.
Two adaptation sessions of 10 trials each
were run in which the red conditionedstimulus light, but not the mirror, was presented to establish that the light alone did
not elicit display behavior. Conditioning
was continued on this schedule, using a
15-sec mirror-presentation until the four
views of the fish exhibiting
movements. D. The maxiapproach viewed laterally.
the gill covers are erected.
637
AGGRESSIVE BEHAVIOR IN F I S H
EXPERIMENT I
UWUUTWG MOJEN€NTS
.
/
~
"
M l COVER K C T C N
C U COVER EFECTDJ
FRONTAL APPROACH
1
TRIALS IN BLOCKS OF TEN
FIG. 8. The number of occurrences of each of the
four aggressive display components during each successive block of 10 trials exhibited by Fish 1 and
2 during Experiment I. Extinction began at the
arrow (Ext).
components occurred on each of 10 consecutive trials in the presence of the conditioned stimulus. In extinction, the oneway mirror was removed, and the illumination changes occurred as before. Since illumination changes accompanied mirror-presentation, it would not have been adequate
to discontinue mirror-presentation by no
longer switching the lights on during ex-
tinction. Illumination changes may have
become a conditioned stimulus for the conditioned response. Thus, in extinction the
conditioned stimulus (red light) was presented and followed by illumination
changes, but not by mirror-presentation.
In order to control light intensity, the mirror was replaced by additional onionskin
sheets until the light intensities of the two
638
TRAVIS THOMPSON
•/> 120
8
% 100 -
Fish I - R e d Light CS
Fish 2-Red Light CS.
°— °
o— o
*— +
o-—o
" ^ " Fin Erection
° — ° Unduloting Movements
+ — * Gill Cover Erection
° — <• Frontol Approach
Fin Erection
Unduloting Movements
Gill Cover Erection
Frontol
I 80
/ // y
// /
60 r-
40 |^
20
60
80
100
120
140
160
180
Trials
0
20
40
60
80
100
120
Trials
FIG. 9. The cumulative number of displays exhibited
by Fish 1 and 2 during Experiment I. Extinction be-
gan at the arrow (Ext).
sides of the tank, as measured by a light
meter, were equal. This procedure was continued until 10 consecutive trials without
any occurrences of the four components had
elapsed.
Figure 8 shows the number of occurrences
of each of the four aggressive-display components during each block of 10 trials.
During the adaptation sessions, neither fish
displayed aggressively in the presence of
the red light. Fin erection was clearly the
most rapidly acquired response, while
frontal approach required the greatest number of trials to reach the criterion level.
There appeared to be little difference between the rates at which Fish 1 acquired
undulating movements and gill-cover erection, but Fish 2 acquired the former more
rapidly than the latter. Extinction of all
four responses required 30 to 40 trials for
each fish.
The regularity of the accelerated acquisition curves is much more apparent in the
cumulated data, and comparison among
the four components is facilitated, as indicated in Figure 9. The marked difference
in the rate at which Fish 1 acquired fin
erection versus frontal approach may be
seen in relation to the very similar difference in this subject's acquisition of undulating movements versus gill-cover erection.
By contrast, the four acquisition curves for
Fish 2 are more clearly separated. Extinc-
tion of all four responses required from 30
to 40 trials for each fish, but the form of
extinction curves is clarified when expressed cumulatively.
Part 2
The results of Part 1 of this experiment
indicate that classical conditioning techniques may bring four components of the
unconditioned aggressive display by male
Bettas under the control of a previously
neutral stimulus light. The possibility remained, however, that these conditioned
responses might be partly the product of
sensitization. Through repeated presentation of the unconditioned stimulus (the
mirror) within this environment, novel or
similar simuli might have become capable
of eliciting the aggressive-display behaviors.
To eliminate this possibility, a positive
conditioned-stimulus light was repeatedly
paired with mirror-presentation, while a
negative conditioned-stimulus light was repeatedly presented alone. If the aggressivedisplay behaviors occurred in the presence
of the positive conditioned stimulus and
not in the presence of the negative conditioned stimulus, the results must reflect
discriminative classical conditioning rather
than sensitization.
Two experimentally naive male Bettas
(Fish 3 and 4), measuring 5.0 and 6.0 cm,
639
AGGRESSIVE BEHAVIOR IN F I S H
respectively, were used, and the equipment
was the same as that employed in Part 1
of this experiment. In addition, a piece
of green cellophane was placed above the
red cellophane, and a second 7i/£-watt bulb
was situated behind the green cellophane
window. Either light could be presented
to illuminate the interior of the tank with
one or the other color.
Fish 3 and 4 were tested daily for 20 trials
each. These trials comprised 10 green-light
presentations followed by mirror-presentation, and 10 red-light presentations not
followed by niirror-presentation. Thus, the
green light was the positive and the red
light the negative conditioned stimulus.
The inter-trial interval varied from 30 to
240 sec with a mean inter-trial time of
EXPERIMENT II
UNDULATING M&flvtNTS
(1I
z
UJ
I
a!
tti
C l U CWER ERECTION
COVER ERECTCH
Q
>
8
FRONTAL AffHMCH
FRONTAL APPflQtH
TRIALS IN BLOCKS OF TEN
FIG. 10. The number of occurrences of each of the
four aggressive display components during each
block of 10 trials by Fish 3 and 4 in Experiment II.
The upper dark line is the record of displays in the
presence of the green light (the positive CS) and
the lower lighter line indicates the number of displays elicited by the red light (negative CS). ExUnction began at the arrow (Ext).
640
TRAVIS THOMPSON
Fish 4 -
Fish 3-
1
Fin Erection
Undulating Movements
• Gill Cover Erection
Frontal Approoct)
' Fin Erection
1
Unduloting Movements
• Gill Cover Erection
Frontol Approach
*. 4 0 •
.Q
6
3
0
0
20
40
60
60
100
120
' 140
160
180
Trials
FIG. 11. The cumulative number of displays exhibited by Fish 3 and 4 during Experiment II.
120 sec. The randomized presentation of
the two stimulus lights was arranged according to the Gellerman series (1933).
As in Part 1, the home tank was transferred to the test apparatus 10 min before
the first trial began, and responses were
recorded on standardized score sheets. Extinction was conducted by presenting the
green light for 10 sec, but no longer following it by mirror-presentation. During
this procedure, the one-way glass was left
in place. This contrasted with the extinction procedure of Part 1 of this experiment,
in which the illumination changes associated with mirror-presentation also occurred
during extinction. The procedure in Part
2 controlled for the possibility that the
mirror might have produced, during conditioned-stimulus presentation, a slight
reflection capable of eliciting the unconditioned response. The criteria for acquisition and extinction were the same as those
used in Part 1.
Figure 10 presents the number of occurrences of the four aggressive-display components in the presence of the positive
conditioned stimulus and the negative conditioned stimulus during each block of 10
trials. With the exception of one fin
erection by Fish 4 in the presence of both
the positive and negative conditioned stimuli, no responses occurred during the 20
40
60
80
100
120
140
18
Trials
Extinction began at the arrow (Ext).
adaptive trials. Fin erection and undulating movements were the first two components to be acquired by both fish, whereas
gill-cover erection and frontal approach
were the last two to be acquired. In the
last 10 trials, in which all four components
were elicited by every green-light presentation (the conditioning criterion), no responses were elicited by the red light. From
20 to 50 trials were required for the four
components to reach the extinction criterion of 10 consecutive trials without any
occurrence of display.
These relationships are clearer when presented as the cumulative number of displays for each component, as in Figure 11.
The difference in rate of acquisition of fin
erection vs. undulating movements, as opposed to the difference in rate of acquisition of gill-cover erection vs. frontal approach, is marked. A comparison of the
slopes of the cumulative extinction curves
reveals considerable differences in the rates
of extinction, which is not apparent from
examining the number of displays per block
of 10 trials. Gill-cover erection and frontal
approach extinguished most rapidly, as revealed by their lesser slopes. The cumulative curves for all components paired with
the green light were positively accelerated,
while the curves for responses elicited by
the red light were negatively accelerated.
641
AGGRESSIVE BEHAVIOR IN FISH
Discussion
The three experiments reported here suggest several interesting relations between
the "releaser" concept used by ethologists
and the term "reinforcer" employed by
experimental psychologists. A releaser can
act as a positive reinforcer for an operant
response, since its contingent presentation
increases the frequency with which the
operant occurs. Thus, a releaser has the
same functional properties as other stimuli
which are capable of (1) acting as unconditioned stimuli in a classical-conditioning
paradigm and (2) functioning as reinforcers
for operants (e.g., food). If there is anything unusual about this reinforcer (the
visual image of another male Betta in aggressive display), it is that the consequences
in a natural setting would almost surely be
damaging to the organism. Usually, stimuli
which are associated with damaging consequences do not function as positive reinforcers, but are avoided or removed. A
possible interpretation of these data is that
the reinforcing consequence of presenting
the visual image of another male Betta is
not that the visual stimulus is inherently
reinforcing, but that it elicits behavior
which is more probable than the operant
under consideration. It has been shown
that operants can be reinforced when the
only consequence is the opportunity to engage in more probable behavior (Premack,
1959). Certainly, in Betta splendens, aggressive display is more probable than the
swimming operant (operant level = 3-10/
hr).
Another implication of these findings is
that display behavior occurring in the absence of the usual releaser is not necessarily
"vacuum activity" (i.e., a fixed action pattern occurring under conditions of high
motivation in the absence of the appropriate releaser; Verplanck, 1957). In these
experiments, four components of the aggressive display were brought under the
control of a previously ineffective stimulus
through classical conditioning. This obviously does not suggest that all behavior
occurring in the absence of the usual releaser is conditioned, but that this is one
possible mechanism by which such behavior
may come to occur.
REFERENCES
Adler, N., and J. A. Hogan. 1963. Classical conditioning and punishment of an instinctive response
in Betta splendens. Animal Behaviour 11:351-354.
Braddock, J. C, and Z. I. Braddock. 1959. Effects
of isolation and social contact upon the development of aggressive behaviour in the Siamese fighting fish, (Betta splendens). Animal Behaviour
7:222-232.
Forselius, S. 1957. Studies of anabantid fishes. I.
Zool. Bidrag. 32:93-302.
Lissmann, H. W. 1933. Die Umwelt des Kampfisches.
Z. Vergl. Physiol. 18:65-111.
Premack, D. 1959. Toward empirical behavior laws:
I. Positive reinforcement. Psychol. Rev. 66:219233.
Smith, H. M. 1937. The fighting fish of Siam. Nat.
Hist. 39:265.
Thompson, T. 1963. Visual reinforcement in Siamese fighting fish. Science 141:55-57.
Thompson, T., and T. Sturm. 1965a. Visual reinforcer color and operant behavior in Siamese
fighting fish. J. Exptl. Anal. Behavior 8:341-344.
Thompson, T., and T. Sturm. 19656. Classical conditioning of Siamese fighting fish. J. Exptl. Anal.
Behavior 8:397-403.