Journal of Comparative Psychology Copynght 1984 by the

1984. Vol 98, No 4,421-431 American Psychological Association, Inc

Agonistic and Reproductive Interactions

in Betta splendens
Paul M. Bronstein
University of Michigan—Flint

Reproductive and agonistic behaviors in Siamese fighting fish were investi-

gated in eight experiments, and some consequences and determinants of these
sequences were isolated. First, fights and the formation of dominance-subor-
dinancy relations were studied. Second, it waB determined that large body size
as well as males' prior residency in a tank produced an agonistic advantage;
the magnitude of this advantage was positively related to the duration of
residency. Third, the prior-residency effect in Bettas was determined by males'
familiarity with visual and/or tactile cues in their home tanks. Fourth, domi-
nant males had greater access to living space and were more likely to display
at a mirror, build nests, and approach females than were subordinates. Finally,
it was discovered that chemical cues associated with presumedly inert plastic
tank dividers influence Bettas' social behavior.

It has been suggested that domesticated The current report extends the conten-
Siamese fighting fish (Betta splendens) tion that Bettas' social behavior is similar
possess an agonistic/reproductive strategy to that of other teleosts and that their
typical of many wild-living teleosts that learning can profitably be conceptualized
achieve fertilization external to the bodies as embedded in an ethogram. Specifically,
of females (Bronstein, 1980, 1981a, 1982). it was hypothesized that males residing in
Males build nests, defend those nest sites an aquarium, because they become familiar
against intrusions by adult conspecifics, with the stimuli of their environment,
and, after copulating, chase females from would dominate male intruders. This prior-
the nests; they then care for eggs and fry. residency effect, documented in other spe-
Bronstein (1981b, 1983a, 1983b) asserted cies (e.g., Braddock, 1949; DeBoer & Heuts,
that operant investigators, using Bettas as 1973; Figler & Einhorn, 1983), was con-
a vehicle to allege the existence of a "social- firmed. Furthermore, dominant males were
reinforcement process" (cf. Hogan & Ro- more successful than submissive opponents
per, 1978), have ignored and distorted many at completing several reproductive behav-
aspects of this strategy. Furthermore, the iors.
particular character of those distortions—
confusing elicited and reinforced behaviors
while interpreting species-specific acts as Experiment 1
indexes of learning—renders all claims for This study was designed to examine
a "social-reinforcement process" in Bettas fights between pairs of male Bettas and
ambiguous, if not invalid (Bronstein, determine whether a prior-residency effect
1981b). exists. In addition, the hypothesis (e.g.,
Barash, 1977; Dewsbury, 1982) that domi-
This research was supported by giants from the nant males court and mate with females
Faculty Development Committee of the University of more readily than submissive fish was con-
Michigan—Flint, a Rackham Grant from the Univer- firmed.
sity of Michigan—Ann Arbor, and by Grant MH
38792-01 from the National Institute of Mental
Health. Method
I am grateful to Larry Atherton, Jr. and William
Mykolajenko for their excellent technical aid and to Adult domesticated Siamese fighting fish, pur-
Colleen Pace for her secretarial assistance. chased from local suppliers, were used in all studies.
Requests for reprints should be sent to Paul M. Fish were red, blue-green, or purple and were main-
Bronstein, Department of Psychology, University of tained in individual clear glass jars containing 1-2
Michigan, Flint, Michigan 48502-2186. liters of water. (Unless otherwise indicated, water for

421
422 PAUL M. BRONSTEIN

all phases of all studies was tap water, aged at least 24 cm below the water surface, whereas the mound in
hr, in which no other fish had swum.) Males were each corner usually rose out of the water, often pro-
always kept in visual isolation when not being tested; viding a moist beach 1-2 cm higher than the water
females could see each other in adjacent containers. line. Each tank also had four sheets of white styrene
Fish were fed twice daily (morning and afternoon) 5 planted on edge in the gravel and radiating from the
days per week with dry food (Wardleys) or recently central mound toward each corner. This plastic, a
thawed adult brine shrimp {Artemia spp.). Lights went common material for dividing aquaria (manufactured
on at 0800 and off at 2200 hours, and space heaters by Penn Plax, Garden City, New York), was less than
maintained the colony at 28 °C ± 2 °C. Data collected 1 mm thick, with perforations (1 mm in diameter)
from approximately 1,100 different individuals con- uniformly distributed (five holes/cm 2 ); each sheet was
tributed to these experiments, with a minority of fish 10 cm long and extended 1-2 cm above the water. This
participating in more than one study. Animals were arrangement of gravel and plastic permitted fish to
acclimated to the laboratory for at least 1 week before swim around each of the barriers but also provided
being used for the first time, and fish were maintained adequate space and visual cover for one male fleeing
for at least 3 weeks between studies. Animals that ' from another. With this apparatus design, animals
sustained marked injuries, including severely torn fins, rarely were injured after becoming trapped by an
were never reused, and each male in a combat test had attacker.
had the same recent social experience as its opponent. In a first replication, nine pairs of males were
Pre-fight testing. Aggression in male Bettas is created, with residents placed in their aquaria for 3
highly variable (e.g., Simpson, 1968). Furthermore, the days prior to the introduction by net of an intruder.
intensity of male-male aggression during a prolonged Dyads were then watched continually for 2 hr, and
fight appears positively correlated with the intensity tanks were also observed for 20 min at 4, 6.5, and 24
of their agonistic display during brief encounters with hr after pairing. During the first 2 hr of observation,
either their mirror image or a live conspecific (Bron- a detailed catalogue of each encounter was kept; this
6tein, 1981a, Meliska, Meliska, Hoyenga, Hoyenga, & record included information about the changing posi-
Ward, 1975). Animals that escalate their aggression tion of each fish (especially, relative to nest sites),
most quickly are also the most persistent fighters chases and escapes by each animal, the number of
(Bronstein, 1981a). Consequently, in an attempt to seconds a resident spent at its nest, and the number
remove some of the agonistic variability within pairs of seconds the pair spent fighting (biting and/or using
of combatants, males were pretested with a mirror 5 lateral and frontal displays within one body length of
days prior to the start of a dominance test. Pairs were each other). The identity of the animal that delivered
matched on their pretest reactions to a mirror and the first bite and the latency of that first bite were
mismatched on their color in order to permit the easy also recorded. During subsequent, brief examinations
discrimination of individuals by visual inspection. of the tanks, bites, chases, gill-cover erections, and
The pretest tank was an all-glass aquarium (40.4 x escapes were noted, along with incidental information
20.3 cm), having undyed gravel flooring and filled to about the coloration, position, and nest-building activ-
a depth of 15 cm. The water for pretests and postests ity of the animals.
was not changed between subjects. Individual fish were The second replication involved 16 resident/in-
netted, put into the tank, and, after a 5-min delay, truder pairs. Also, in order to determine whether the
exposed to a mirror that was placed outside the tank strength of the prior-residency effect was directly re-
over one short wall of this aquarium. Vertical lines lated to the duration of an animal's occupancy, two
had been drawn 7.6 cm from the mirror wall on the independent groups were created. Seven males were
sides adjacent to the mirror wall. The number of in residence for 1 day, and 9 fish swam in their test
seconds an animal spent in agonistic display between aquaria for 4 days prior to the introduction of intrud-
the mirror and these lines (i.e., within about two body ers. Observations of the fish lasted approximately 20
lengths of the mirror) was recorded during the 15-min min and were made approximately 2, 4, 6, 9.5, and 24
tests. For the pretests, display was defined as the hr after pairs had been formed.
composite of any agonistic movements—biting, side- As many as six experienced observers, each acting
to-side tail beating, and gill-cover erection. Fish were independently, watched the fish in any given study.
netted and returned to their home containers imme- Because the attainment of dominance in Bettas was
diately following the pretest. From a large pool of found to be a discrete and unambiguous event, inter-
potential subjects, fish were reliably matched with an judge reliability approached perfection in every study,
opponent, with all males included in the study exhib- and data were used only if no discrepancies were found
iting some display. In each replication of Experiment among the observers.
1, the partner-to-partner linear correlations of display Post-fight testing. The first replication was ex-
duration was .99 (ps < .01). tended in two ways. First, an adult female was added
Prior-residency test. All-glass aquaria (51 x 26 to each aquarium 24 hr after males had been paired.
cm), with undyed gravel flooring, plastic-and-glass Three-minute observations of each tank were made
covers, and aged water (9 cm deep), were used for immediately following the females' introduction and
dominance testing in all studies except Experiment 4. 6, 21, 24, and 30 hr later. Fish were then returned to
Tanks always had brown paper covering their two their individual containers. In addition to the cata-
short sides, with one long side against a blank wall logue of male-male interactions noted above, ap-
and the other long side facing into the center of the proaches to the female by each male were now re-
laboratory, approximately 2 m from any other tank. corded, as were copulations and the presence of eggs.
Undyed gravel covered the entire floor, but it was piled Second, 4 days after the breakup of the triads, the
in the center of each tank and was especially elevated 15-min pretest was repeated on all socially experienced
in each of the corners. The central hill was about 3 males except for 1 animal that died in the interim.
 BETTA SPLENDENS 423

Simultaneously, 10 males that had been pretested but

ring symmetrically—each fish attacking
not given the social experience of the prior-residency
the other. Finally, after some period of com-
test were readrainistered the 15-min mirror test. These
bat, one male escaped its opponent and did
animals had been denied exposure to live conspecifics
not return, and the other fish continued
because their colors and/or pretest display scores,
which were all greater than 0 s, did not permit their
periodically to approach and exhibit gill-
being well matched with a male partner. Thus, the
cover displays. (The few exceptions to this
existence of these inexperienced males permitted an
discrete, unambiguous dominance-subor-
assessment of the test-re test reliability of the 15-min
dinancy relation were excluded from analy-
mirror-display procedure, with the two tests separated
sis and are discussed below in General Dis-
by 11 days. These inexperienced males also served as
a control group for assessing the effect on mirror-
cussion.)
elicited display of males having become dominant or
submissive. A male was always described as dominant
if, after 24 hr of combat, it exhibited gill-
cover displays while its opponent escaped
Results and Discussion and never responded with bites or gill-cover
displays. The escapes of submissive males
Agonistic and sexual behaviors. Taken took the form of their rapidly swimming
together, both replications confirmed the away from opponents, with many fish wrig-
existence of a prior-residency effect. Of the gling out of the water onto the piles of
25 pairs run in both replications, the resi- gravel in the corners of the tanks. These
dents became dominant in 18 dyads within semiterrestrial refugees were capable of liv-
9 hr after first encountering their partners ing totally out of the water (albeit in a very
(p < .04, binomial test). Judgments made humid environment) for at least tens of
24 hr after the start of a fight were identical minutes, taking oxygen from the air. Dom-
to those made 9 hr after encounters began, inant animals were never observed to follow
and all but four of the contests had been submissive fish out of the water and fre-
settled 6.5 hr after intruders were added to quently were seen extending their nests
the tanks. Finally, there was no suggestion following the defeat of an opponent. A male
from the second replication that 4 days of also was considered dominant if its partner
residency produced a stronger prior-resi- was found dead (a rare event which did not
dency effect than did 1 day of occupancy, 2 occur in Experiment 1) or if the partner
X 2 x 2 (l, N = 16) = 0.08. had beached itself even though the actual
Also, the extended observations of fight- escape had not been witnessed.
ing during the first replication supported The intensity of aggression was highly
and extended prior observations of agonis- variable, a result confirming several pre-
tic and reproductive activities in Bettas. All vious findings (e.g., Bronstein, 1981a;
residents had constructed nests prior to the Simpson, 1968). The median latency to a
start of an encounter, and the presence of first bite was 706 s (range: 211-5,697 s),
an intruder always elicited an alternation and the fish to bite first was both the resi-
between fighting and nest fixation on the dent and the eventual victor in six of the
part of residents. During the initial 30 min nine pairs. The median duration of fighting
of pairing, each resident issued frontal, gill- was 67 s (range: 2-4,978 s); three pairs
cover displays to the intruder and also es- fought for more than 3,500 s, and three
caped from its opponent to assume a posi- fought for 60 s or less. The median time
tion near its nest. Intruders typically fol- that residents spent at their nests was 1,768
lowed residents to the nests where agonistic s (range: 112-3,911 s). In spite of pairs
encounters continued. The subsequent being well matched on their mirror-display
fighting occurred in all parts of the aquaria. pretest, there was great variability in the
As seen previously in different situations details of the fights. However, the overall
(Bronstein, 1981a, 1981b, 1983a, 1983b), style of aggression was the same in every
opponents approached each other and dis- pair—mutual escalation followed by one
played for a few seconds, and one or both animal's escaping the conflict.
fish then escaped. However, after repeated The addition of a female to the nine
proximal and mutual stimulation the es- tanks of the first replication always re-
capes of both partners were in most cases sulted in the early stages of courtship, that
eliminated, with injurious fighting occur- is, females were approached by at least one
424 PAUL M. BRONSTEIN

displaying male. In all tanks, the dominant

male had the most consistent, unobstructed
access to the female, and in six cases the
submissive male was never observed to ap-
proach the female. Copulation by 4 domi- 1000
nant males was noted, with eggs resulting
in three cases. Copulation was observed
only in those tanks where submissive males
appeared never to pursue the female. In the
three tanks where a submissive male did
approach the female intermittently, the soo
dominant male exhibited proximate gill-
cover displays and drove off the submissive
courter.
It appears, then, that dominant status
permits a male access to some amount of
space. By repelling male competitors with
P H POST P K POST PRt POST
their relatively persistent displays (cf.
DOM SUB CONTROLS
Bronstein, 1981a, Experiment 2), dominant N =9 N-a NOO
males also gain the opportunity to approach GROUPS
females; however, given males' access to
receptive females, breeding depends upon Figure 1. Duration of mirror-elicited display by dom-
inant (DOM) and subordinate (SUB) males during
relative privacy. Bronstein (1982) showed pretests and posttests of Experiment 1. (Controls were
that visual cues from a second male inter- administered the two 15-min mirror exposures but
fere with the reproductive success of court- lacked intervening social experience.)
ing Bettas. Consistent with those findings,
the current results show that when one
male, although dominant, was not able to These data extend earlier findings showing
produce the relatively permanent escape of an elevation in agonistic behaviors among
another male, the proximity of the submis- dominant males (Baenninger, 1970; Mel-
sive male disrupted breeding of the domi- iska et al., 1975; Simpson, 1968). However,
nant fish. Males appear to require nearly because of enormous within-subjects vari-
exclusive access to females in order to breed ability, the analysis of variance showed a
successfully. An approaching male, espe- reliable effect only of group identity, F(2,
cially if displaying, releases aggression that 24) = 4.97, p < .05. There was no significant
typically competes with and overwhelms effect of repeated testing and the Groups X
the sexual behavior of the dominant fish. Tests interaction was also unreliable (Fs <
Mirror-elicited display. The pre- and 1).
posttest display times of the inexperienced It is attractive to hypothesize that
control group were linearly correlated, r(8) achieving dominance instigated the height-
= .84, p < .01, a result indicating that the ened mirror-elicited display. However, the
mirror test was reliable over 11 days of current study does not permit a clear un-
isolated maintenance. Display times for all derstanding of how the mirror-elicited dis-
groups were also subjected to a 3 X 2 mixed play of the dominant males increased.
analysis of variance. The three treatment Those fish were exposed to males, females,
groups (dominant, submissive, and inex- and four changes of water between tests.
perienced controls) identified one inde-
pendent variable, and the two identical Experiment 2
tests were a within-subjects variable.
As noted in Figure 1, there was an in- The purpose of this study was to examine
crease in the display duration of dominant the possible importance of body length in
males following their social experience, t(8) determining dominance among males.
= 4.74, p < .01, but the display of the two Prior work with other teleost species (e.g.,
other groups did not vary between tests. Baerends & Baerends-Van Roon, 1950;
 BETTA SPLENDENS 425

DeNight & Ward, 1982) has shown that decide about the imminent dominance po-
large body size confers an agonistic advan- tential of an opponent on the basis of visual
tage on fishes, which suggests a similar cues.
relation in Bettas. Six of the females either were approached
equally by both male companions or es-
caped from all males by beaching them-
Method selves repeatedly on the gravel mounds. In
Thirty-four pairs of males were created according the remaining 28 aquaria, the previously
to three criteria. First, the animals had to be easily dominant male had nearly exclusive access
differentiable on the basis of coloration. Second, 2 to the potential mate. One dominant male
days before the start of the study they had to exhibit copulated with a female. Thus, defeated
frontal display when a mirror was presented for 5 min males typically are chased from females by
on the outside of their home containers. Third, four
experienced, independent observers had to agree that the dominant males.
one member of each dyad was larger than its partner.
It was decided to have decisions about size depend
upon high interjudge reliability rather than risk injury Experiment 3
to the subjects by netting and measuring the animals.
Intermittent post hoc measurements showed that the
large member of each dyad had a standard body length A further implication of the previous
of about 4 cm; their partners were typically between 5 study is that the prior-residency effect
and 10 mm smaller. noted in Experiment 1 might have been
Standard residency tanks were divided into two contaminated by not matching partners on
equal-sized compartments by the use of gravel-and- body size. Consequently, I repeated those
styrene barriers rising 3-4 cm above the water surface,
running parallel to the short walls of the tanks, and male-male encounters after first matching
through or over which partners could not visually partners on their body length. In addition,
communicate. Chemical communication between I tested the hypothesis, confirmed for other
partners was possible, however, Two green plastic teleosts (e.g., Figler & Einhorn, 1983), that
plants were located centrally on both sides of each
tank. Partners were placed in abutting sections of the magnitude of the prior-residency effect
each aquarium for 2 hr, and social interactions began is directly related to the duration of resi-
with the destruction of the gravel-and-plastic barriers. dency.
This procedure took about 1 min and involved creating
a channel between compartments at least 15 cm wide
by sweeping gravel gently toward all edges of the tanks.
In this and all subsequent studies except Experiment Method
4, a dommance/subordinancy decision was reached by
watching dyads for 30 min following 24 hr of combat. In this and in all subsequent studies, fish were
The behavioral criteria of Experiment 1 remained in matched as in Experiment 2, that is, both partners
effect. were of the same size, but of obviously different color-
Also, a female was netted and added to each aquar- ation, and recently had shown some display to a mirror
ium immediately after a dominance decision was image. Eighty-four pairs of males were used, with 48
reached. Triads were observed for 1 hr over the next males in residence for 3 days and 36 males in residence
day, with the same behaviors noted in Experiment 1 for 10 days prior to testing. The apparatus and pro-
again recorded. One 20-min observation was made 2, cedures were identical to those of the first study.
5, and 24 hr after the females were added

Results
Results
Residents dominated intruders in 70.24%
Twenty-eight of the relatively large fish of the pairs (p = .0002), with the effect
(82%) dominated their smaller partners. being strongest among 10-day residents.
This result is highly reliable (z = 3.60, p < Thus, after size is controlled for, a robust
.0006) and confirms the importance of body prior-residency effect occurs in male Bet-
size in determining the social status of male tas. In the 3-day group, 63% (30/48) of
Bettas. Also, several of the small-sized op- residents became dominant (p = .08), and
ponents responded to their larger partners 81% (29/36) of the 10-day residents were
by escaping immediately, that is, at the victorious (p = .0004). The difference be-
start of testing—even prior to any biting. tween groups approached significance, 2 x
This observation suggests that Bettas can 2 x 2 (l, N = 84) = 2.40, p < .10.
426 PAUL M. BRONSTEIN

Experiment 4 It
-J U

The main purpose of this study was to

further test the observation in Experiment Z <«
1 that, following combat, dominant males < Q

are more likely to construct nests than are

subordinate fish. In addition, the generality
of prior results was assessed by, first, con-
ducting several days of observations and, ?4
second, using an aquarium much larger
than those previously employed.
Z) I

Method
Seventeen pairs of males were created, and the fish
of each dyad were introduced simultaneously into
rectangular, all-glass aquaria (45.7 X 152.4 cm) Tanks
were filled to a depth of 9 cm, were covered with clear DOM*1
Q O
glass, had their short sides covered with brown paper,
SUB
and had one long side facing a blank wall. The undyed
gravel floors of these vessels were without any major
hills or depressions, and in each corner of these tanks
stood an opaque plastic bottle. Bottles were placed 3 1-T-
DAYS
cm from each adjacent wall, had square cross-sections
(6.4 cm on a side), and extended above the water
Figure 2 Day-by-day frequency of frontal, gill-cover
surface. Also in each corner, planted in a 90° arc
displays (top panel and the frequency of mutual dis-
around each bottle, six green plastic plants were em-
bedded in the gravel; the radius of these arcs was plays (center panel) for dyads of Experiment 4; in the
approximately 15 cm, and this artificial vegetation lower panel, daily display ratios for dominant and
extended to the water surface. Replicas of 13 plant subordinate males. (Vertical bars indicate SD. DOM
species were used. Thus, a visually isolated area, typ- = dominant; SUB = submissive.)
ically favored as nest sites by male Bettas (Bronstein,
1981a), was created in each corner of each tank.
There was no particular criterion of dom-
Fish were netted and placed into their tanks at 0900
inance held to be valid a priori; however, as
of Day 1. They were observed for a total of 30 min
daily—10 min per observation at 1000,1200, and 1500
shown in the bottom panel of Figure 2,
hours—for 4 consecutive days. The following measures
members of all pairs clearly segregated
were taken: frontal displays, bites, escapes (one male
themselves into the more or the less domi-
fleeing the locale of another following a display or bite
nant member from the very start of testing.
by its partner), nest building, and nest fixation (an
animal remaining within one body length of its bubble
With the exception of three dyads wherein
nest).
partners had identical display ratios for the
first day, no decision about dominance
made on Day 1 was altered during the final
Results and Discussion 3 days of testing. As the study proceeded,
Display ratios (displays/displays + es- partners became increasingly different.
capes) were computed daily for each sub- The submissive males became more likely
ject; means and standard deviations are to retreat and less likely to display at a
shown in the bottom panel of Figure 2. This displaying opponent; hence, mutual dis-
measure is similar to a ratio used by Barlow plays (both fish displaying simultaneously)
and Ballin (1976) in describing midas cich- were noted mostly on Day 1 (see the middle
lids. A score of 0.0 represents a totally panel of Figure 2). Within both subgroups,
submissive male that always reacts to the but especially among the subordinates,
displays of its opponent with escape and there was a progressive reduction in dis-
never itself initiates displays. A maximum play-ratio variability during the 4 days of
score (1.0) denotes a male that always an- testing. On Day 4, 11 of the submissive
swers its partner's displays in kind and may males were never observed to display at
also display at a nondisplaying opponent. displaying partners, and the most aggres-
 BETTA SPLENDENS 427

sive of the subordinates had a display ratio aquarium, and both fish remained undisturbed for 10
only half as great as its opponent's. One days after which a 24-hr combat test was initiated. In
one half of the tank were two pieces of white perfo-
male was severely bitten and killed by its rated styrene, placed as in Experiment 1. The other
partner on Day 2; data from this dyad are part of each tank included two centrally located plastic
omitted from Figure 2. Finally, the upper plants. Also, after both 3 and 7 days of residency, 150
panel of Figure 2 shows that there was no ml of water was slowly siphoned from both sides of
day-to-day decrease in the number of dis- each tank and gently poured into the opposite side of
the same tank.
plays per dyad. However, as shown in the Testing was initiated by, first, removing the famil-
two lower panels, there was a shift in the iar visual/tactile cue (either the white styrene or the
style of agonistic encounters during the 4- artificial plants) from one side of each tank and, next,
day test. Each subject initially attacked its destroying the central barrier, as in Experiment 2.
partner, and, as in the smaller tanks, com- Thus, unlike in Experiments 1 and 3, both members
of each pair had lived in the same-sized environments,
bat occurred throughout the aquarium. had had the same experiences being transferred be-
However, by the end of Day 1 one fish in tween tanks, and had been exposed equally to water-
each pair was consistently escaping its op- borne traces of their partner during the period of
ponent. These findings indicate that the residency. However, only one member of each dyad
scoring techniques adopted in Experiments encountered its opponent in the presence of familiar
visual/tactile cues, and those males were hypothesized
1-3 are valid predictors for several succes- to become dominant in a high proportion of encoun-
sive days at least. ters.
Observations of nest building were as
predicted and confirmed the findings of
Experiment 1. Nests always abutted cor- Results and Discussion
ners, outcroppings, and artificial plants,
and dominant males had exclusive access Males fighting in the presence of familiar
to nest sites. In nine tanks nests were ob- stimuli did not have any agonistic advan-
served being tended only by the dominant tage; such fish became dominant in only
male on Days 2-4. In a tenth tank one male 39% (15/38) of the contests. It thus appears
began to build a nest on Day 1; however, that the prior-residency effects noted in
this subject was dominated by its opponent, Experiments 1 and 3 were not the result of
and its nest disappeared on Day 2, with the residents' becoming familiar with visual
now dominant male constructing a nest at and/or tactile aspects of their tanks. It
a location different from the submissive might have been, for instance, that intrud-
male's prior nest site. Dominant males gain ers were at a disadvantage because of their
access to nest sites and often construct having recently been netted and removed
nests. Subordinate animals lack these be- from a body of water. The current proce-
haviors. dures removed this distinction and conse-
quently may have eliminated the prior-res-
idency effect. On the other hand, the spe-
Experiment 5 cific details of testing in the current study
may have overwhelmed the potent expres-
In some teleost species, the prior-resi- sion of residents' familiarization with intra-
dency effect is based on resident animals' tank cues; that is, learning may have oc-
being more familiar with the visual and/or curred, yet performance could have been
tactile qualities of their environment than impaired.
intruders (e.g., DeBoer & Heuts, 1973). The The next three experiments were contin-
purpose of the present study was to exam- ued attempts to analyze the prior-residency
ine this hypothesis for Bettas. effect found in Experiment 3. In proceeding
further and, in part, because of evidence
that prior-residency effects in other teleost
Method species are multiply determined (Figler &
Einhorn, 1983), I considered the null find-
Thirty-eight pairs of males were used, and one ings of Experiment 5 as but a weak sugges-
member of each pair was placed in half of a bifurcated
tank, as used in Experiment 2. The size-matched tion that visual/tactile cues cannot control
partners were placed on opposite halves of the same Bettas' prior-residency effect.
428 PAUL M. BRONSTEIN

Experiment 6 These data show that the prior-residency

effect in Bettas survives the residents' being
The intruders of Experiment 3 differed netted and transferred between environ-
from residents in that the former lacked ments. Second, the agonistic superiority of
recent exposure to the prominent features residents in Group 2 shows that different
of residents' tanks (the white styrene bar- concentrations of personal chemosignals
riers). However, intruders also were unique played no discernible role in creating the
in that they had recently been netted and prior-residency effect. The failure of the
introduced to water that may have con- residents in Group 3 to dominate intruders
tained novel chemical cues (e.g., nests) pro- may mean that the prior-residency effect is
duced by their opponent. The current study based on familiarity with visual/tactile
examined the possible roles of these factors cues. However, because there is no cer-
in controlling the prior-residency effect. tainty that the white styrene and the green
plastic plants create the same chemical en-
Method vironment for the fish, it is possible that
differences in water-borne cues play some
Sixty-nine pairs of males were used. Fish were role in changing the agonistic behavior of
divided into three groups, and all resident animals Bettas.
spent 10 days in tanks described in Experiment 1.
Group 1 was designed to control for the possible
disruption of intruders due to their being netted and Experiment 7
removed from water for a few seconds just prior to
testing. Intruders were treated as before, but resident The influence of being netted just prior
males were netted and returned to their own tanks at to being tested on the formation of a dom-
the start of a 24-hr contest. Group 2 was a control for
the potential influence of chemosignals identified with inance-subordinancy relation appears min-
a particular fish. Both residents and intruders were imal. However, there exist several other
treated identically to those in Group 1 during the 10 extraneous and confounded differences be-
days preceding the dominance test However, combat tween residents and intruders that might
occurred in tanks that were arranged identically to have contributed to the prior-residency ef-
residents' aquaria except that no fish had swum there.
Thus, residents continued to be exposed to a familiar fects in Experiments 1, 3, and 6. First, for
visual/tactile environment, but in novel water. Group the 10 days prior to fighting, residents were
3 was treated like Group 2 except that dominance tests exposed to larger tanks than were intrud-
occurred in tanks identical to the residents' aquaria ers. Second, residents had their water
but planted, not with white styrene, but with two changed 10 days before testing (when they
centrally located clumps of artifical vegetation. Group
3 residents confronted the intruders in environments were rehoused in the aquaria); intruders
that differed from their home tanks on visual and remained in their 1-2-liter jars and in the
tactile dimensions. same water where they had been main-
In this study and in both subsequent experiments, tained. Third, residents had been swim-
both fish in a dyad were introduced into the test ming above gravel, whereas intruders
environments by net and simultaneously.
lacked this experience. Finally, residents,
but not intruders, were exposed to visual,
Results tactile, and, perhaps, chemical cues associ-
ated with the white styrene in their tanks.
Reliable prior-residency effects of a mag- The current study primarily was an at-
nitude similar to that seen in Experiment tempt to determine whether any of these
3 were obtained for Groups 1 and 2, with factors actually play a role in creating the
79% (15/19) and 72% (22/29), respectively, prior-residency effect. In addition, the as-
of residents becoming dominant (ps < .02). sociation of nest-building activity with
However, only 43% (8/21) of residents were dominance was again assessed. Experi-
victorious in Group 3; the difference be- ments 1 and 4 showed that subordinate
tween Group 2 and Group 3 approached males were unlikely to construct bubble
significance, 2 X 2 x*(l, N = 50) = 4.32, p nests. However, both of those results could
< .10, and Group 3 was reliably different have been brought about, in part, by the
from Groups 1 and 2, taken together, 2 x 2 dominants harassing the subordinate fish.
X2(l,iV = 69) = 6.23,p<.05. The current design eliminated this partic-
 BETTA SPLENDENS 429

ular influence so as to determine whether 8 wo

dominant and subordinate animals, when
maintained in isolation, show different
nest-building tendencies.

Method so
Two hundred nineteen pairs of males were used;
these dyads were separated into six groups and were
differentiated by the treatment of the intruders. Group
1 was a direct replication of the first group in Exper-
iment 6. Intruders were maintained in jars containing
1-2 liters of water, that is, the same water in which
the fish had been residing prior to their being paired
with a resident male. Intruders of Group 2 also resided
Groups
in jars for 10 days; however, these fish were rehoused Figure 3 Percentage of resident males that achieved
in aged fresh water at the same time their partners dominance over intruders for the six groups of Exper-
were placed in the standard resident tanks Intruders iment 7.
of Group 3 were maintained identically to fish in
Group 2 except that 2-3 cm of undyed gravel was
added to their jars Group 4 intruders were treated
identically to males in Group 3 except that a piece of nated for Groups 4 and 6, with intruders
white styrene, identical to the material in residents' housed with a buried piece of styrene (ps
tanks, was buried under the gravel. In Group 5, the > .27). The percentage of dominant resi-
intruders lived for 10 days in tanks identical in size to dents was lower in Group 4 than in Groups
residents' aquaria. These tanks contained 9 cm of 1, 2, and 5, 2 X 2 x 2 (l. 92 > N > 73) > 3.25,
water and a flat layer of undyed gravel (2-3 cm deep)
as flooring material. Group 6 was treated like Group ps < .10. No other intergroup differences
5 except that a piece of styrene was buried beneath approached significance. Also, taken to-
the gravel. All intruders were visually isolated prior to gether, Groups 4 and 6, with 56.5% of their
testing. residents becoming dominant, were reliably
Dominance-subordinancy tests were conducted in different from the combination of Groups
the residents' tanks, and, in order to equalize the 1, 2, 4, and 5, with 74.6% of resident males
disruption due to capture, handling, and transport,
resident males were always netted and returned to victorious, 2 x 2 x 2 (l, N = 219) = 7.02, p <
their own aquaria just as the intruders were added .02. It appears, therefore, that the styrene
Furthermore, immediately following the 24-hr test, commonly used to divide laboratory
111 males were placed in individual goldfish bowls aquaria is biologically active. This plastic
containing 2-3 liters of fresh water and with 2 cm of emits some chemical trace(s) affecting
undyed gravel as flooring. Fish were visually isolated,
and the number of males building nests was recorded agonistic behavior in Bettas so as to elimi-
three times per day during the next 68 hr. Nests were nate the prior-residence effect.
defined, a priori, as any contiguous cluster of surface Furthermore, these findings provide a
bubbles measuring at least 2 cm across and 4 mm deep.
This criterion served to differentiate obvious nest plausible explanation of why males of Ex-
builders from fish that had a few isolated bubbles in periment 5 that fought in the presence of
their bowl. In fact, no marginal cases existed; males familiar visual and tactile cues did not dom-
defined as nest builders all produced obvious clumps inate males fighting in more novel environ-
of bubbles.
ments. Both groups of males had been ex-
posed to chemical traces of styrene during
Results the 10 days preceding combat, and that
experience, as demonstrated in the current
As seen in Figure 3, Groups 1, 2, 3, and study, can take precedence over males' be-
5 all showed reliable prior-residency effects coming familiar with visual/tactile stimuli.
(ps < .01); there were no reliable differ- Finally, 48% (27/56) of dominant males
ences among these groups. That is to say, constructed nests, whereas just 13% (7/55)
the dominance advantage of residents was of the subordinate males did so. The fre-
not diminished by having intruders housed quency of nest building among subordi-
with gravel, in fresh water, or in a tank of nates was lower than would be expected by
the same size as the vessel in which resi- chance, x 2 (l, N = 55) = 15.28, p < .01, and
dents lived. However, the effect was elimi- the difference between groups was also re-
430 PAUL M. BRONSTEIN

liable, 2 X 2 X 2 U, N = 111) = 14.82, p < and immobile after a day together. Another
.01. Thus, even in the absence of ongoing small group of males (11 pairs) did not
visual, tactile, or chemical interference by achieve a clear outcome to their encounters
a victorious male, subordinates are less ef- because their combat resolved itself into
fective nest builders than dominant males. sexual behavior including one or more full
copulatory clasps. Male-male sexual behav-
ior in Bettas has been reported previously
Experiment 8 (Figler, Klein, & Thompson, 1975; Klein,
Figler, & Peeke, 1976; D. Gubernick, per-
The goal of this study was to determine sonal communication, June 1980; E. Wyers,
whether resident Bettas become dominant personal communication, June 1983; how-
because of their familiarity with intratank ever, some studies have failed to detect this
visual/tactile cues. activity (Figler & Klauenberg, 1980). From
the current experiments, I suggest that sex-
Method ual activity among males is sufficiently rare
that its presence or absence in prior studies
Sixty-six pairs of males were used. One member of could have been determined, in part, by
each pair spent 10 or 11 days in the standard resident
aquarium. The partners of those males were simulta- sampling error. Deaths of individuals were
neously housed in tanks of the same size, but with the also unusual, with only 27 dyads (4.3%)
four pieces of white styrene placed in each of the having one member killed. Bronstein
corners rather than radiating toward the corners from (1981a) made the incidental observation
the centers of the tanks. On the day of testing, both that a small minority of males became im-
members of a pair were netted and placed in a third
tank, also of the same size, containing four pieces of mobile and then died when first encounter-
styrene radiating toward the corners, along with water ing another male. This "sudden-death"
in which no fish had swum. phenomenon was never seen in the current
Thus, all fish had the same exposure to conspecific set of experiments.
chemosignals as well as to chemical cues associated
with the styrene. However, only one member of each These studies support most earlier de-
dyad fought in the presence of visual and tactile cues scriptions of male-male competition in Si-
with which they recently had been familiarized. It was amese fighting fish. Males first display to
hypothesized that these fish would have an agonistic
advantage. opponents, with encounters typically esca-
lating into injurious combat. Further, prior
to the onset of intense fighting, residents
Results retreat from opponents, approach their
nests, and engage in nest-building behavior.
This suggestion was confirmed. Sixty- Also as anticipated, agonistic behavior was
five percent (43/66) of the males fighting highly variable; some males did not escalate
in the presence of familiar cues were vic- their fighting but rapidly became submis-
torious (p = .02). Consequently, it may be sive when presented with an aggressive op-
concluded that residents' learning about a ponent (cf. Bronstein, 1981a).
pattern of spatial details in their environ- Several new observations concerning the
ment is sufficient to produce the prior- outcome of fighting and the prediction of
residency effect in male Bettas. dominance in Bettas also emerged. First,
large males are almost always victorious
General Discussion over small opponents. Second, males kept
in residence from 3 to 10 days are typically
The current studies involved 625 pairs of dominant over recently introduced chal-
males of which 552 pairs were included as lengers, with the magnitude of the resi-
producing acceptable data. Seventy-three dents' agonistic advantage being directly
pairs (11.7%) were excluded because it was proportional to the duration of residency.
not possible to reliably assess their domi- Third, males that become dominant show
nance-subordinancy relation after 24 hr of a subsequent elevation in their mirror-elic-
combat. Most of these null-outcome con- ited display and also are more likely to build
tests occurred because both combatants nests and approach females than are sub-
were either actively fighting or were passive ordinates.
 BETTA SPLENDENS 431

Furthermore, subordinate males engage Bronstein, P. M. (1981b). Social reinforcement in

in two types of behavior that reduce or Betta splendens: A reconsideration. Journal of Com-
parative and Physiological Psychology, 95,943-950.
terminate the attacks of dominant animals. Bronstein, P. M. (1982). Breeding, paternal behavior,
These behaviors also keep the subordinates and their interruption in Betta splendens. Animal
away from females. First, they escape from Learning & Behavior, 10, 145-151.
their attackers, at times launching them- Bronstein, P. M. (1983a). Agonistic sequences and the
selves out of the water. In addition, many assessment of opponents in male Betta splendens.
American Journal of Psychology, 96,163-177.
males adopt a pattern of high-contrast ver- Bronstein, P. M. (1983b). Onset of combat in male
tical bars, coupled with low levels of motor Betta splendens. Journal of Comparative Psychology,
activity. These tactics are identical to those 97, 135-139.
used by female Bettas to protect themselves DeBoer, J. N., & Heuts, B. A. (1973). Prior exposure
from males at the start of courtship (Bron- to visual cues affecting dominance in the jewel fish,
Hemichromis bimaculatus Gill 1862 (Pisces, Cichli-
stein, 1982; Robertson & Sale, 1975), and dae). Behaviour, 44, 299-321.
they have been noted previously in male- DeNight, M. L., & Ward, J A. (1982). Relationship of
male encounters (Wallen, personal com- chin-spot size to dominance in the black-chinned
munication, November 1982). mouthbreeding cichlid fish, Sarotherodon melanoth-
eron. Animal Behaviour, 30, 1099-1104.
These behavioral details suggest that Dewsbury, D. A. (1982). Dominance rank, copulatory
Bettas' agonistic and reproductive activities behavior, and differential reproduction. Quarterly
can be understood as an organized and hol- Review of Biology, 57, 135-159.
istic social strategy. Successful aggression Figler, M. H., & Einhorn, D. M. (1983). The territorial
prior residence effect in convict cichlids (Cichlasoma
by males has the benefits of permitting nigrofasciatum Gunther): Temporal aspects of es-
access to females and nest sites and simul- tablishment and retention, and proximate mecha-
taneously eliminating the sexual competi- nisms. Behaviour, 85, 157-183.
tion from defeated males for several days, Figler, M. H., & Klauenberg, B. J. (1980). Pentobar-
that is, for at least one breeding cycle. bital sodium and attack behavior in male Siamese
fighting fish. Psychopharmacology, 69, 207-208.
Furthermore, a period of residency in an Figler, M. H., Klein, R. M., & Thompson, C. S. (1975).
area enables males to more successfully Chlordiazepoxide (Librium)-induced changes in
defend the reproductively useful resources intraspecific attack and selected nonagonistic be-
(e.g., nest sites and food) in that locale. havior in male Siamese fighting fish. Psychophar-
macologia, 42, 139-145.
Hogan, J. A., & Roper, T. J. (1978). A comparison of
References the properties of different reinforcers. In J. S. Ro-
senblatt, R. A. Hinde, C. Beer, and M.-C. Busnel
Baenninger, R. (1970). Visual reinforcement, habit- (Eds.), Advances in the study of behavior (Vol. 8, pp.
uation, and prior social experience of Siamese fight- 155-255). New York: Academic Press.
ing fish. Journal of Comparative and Physiological Klein, R. M., Figler, M. H., & Peeke, H. V. S. (1976).
Psychology, 71,1-5. Modification of consummately (attack) behavior
Baerends, G. P., & Baerends-Van Roon, J. M. (1950). resulting from prior habituation of appetitive
An introduction to the study of the ethology of (threat) components of the agonistic sequence in
cichlid fishes. Behaviour, Suppl. 1, 1-242. male Betta splendens (Pisces: Belontiidae). Behav-
Barash, D. P. (1977). Socwbwlogy and behavior. New iour, 58, 1-25.
York: American Elsevier. Mehska, J. A., Meliska, C. J., Hoyenga, K. T., Hoy-
Barlow, G. W., & Ballin, P. J. (1976). Predicting and enga, K B., & Ward, E. F. (1975). Approach tend-
assessing dominance from size and coloration in the ency and threat display as related to social status of
polychromatic unidas cichlid. Animal Behaviour, 24, Siamese fighting fish, Betta splendens Animal
793-813. Learning & Behavior, 3, 135-139.
Braddock, J. C. (1949). The effect of prior residence Robertson, C. M., & Sale, P. F. (1975). Sexual discrim-
upon dominance in the fish Platypoecilus maculatus ination in the Siamese fighting fish (Betta splendens
Physiological Zoology, 22,161-169. Regan). Behaviour, 54,1-26.
Bronstein, P. M. (1980). Betta splendens: A territorial Simpson, M. J. A. (1968). The display of Siamese
note. Bulletin of the Psychonomic Society, 16, 484- fighting fish. Animal Behaviour Monographs, 1, 1-
485. 73.
Bronstein, P. M. (1981a). Commitments to aggression
and nest sites in male Betta splendens Journal of
Comparative and Physiological Psychology, 95, 436- Received January 24, 1984
449. Revision received May 27,1984