Behaviour, LIX, 3-4

by M. DOW, A. W. EWING and I. SUTHERLAND (Department of Zoology, University of Edinburgh, Scotland)
(With 6 Figures) (Acc. 15-XI-1975)

INTRODUCTION Most functional systems in behaviour, such as courtships or aggressive encounters, are structured in time, that is, the frequencies of the components or elements comprising such systems will change during performance of the behaviour. While this is implicit in most descriptions of behaviour there have been rather few attempts to describe these temporal changes quantitatively. One exception, however, is that of courtship behaviour in glandulo-caudine fishes investigated by NF.LSON (1964) who demonstrated non-stationarities for some species, that is changes in behaviour with time, which were mainly of a cyclical nature. BALTHAZART (1974) also analysed the non-stationarities in fights between males of Tilapia macrochir using a method of dividing fights into quarters similar to that described in this paper. While large non-stationarities were found BALTHAZART concluded that these were mainly an artefact due to the experimental situation. The outcome of fights could be predicted by the end of the first quarter and the remaining period merely accentuated the differences in behaviour of dominant and subordinate fish. If sufficient data were available it might have been more instructive to have looked for non-stationarities within the first quarter. One possible reason for the paucity of studies concerned with long-term patterning of behaviour is that some of the more sophisticated methods of analysis which have been used, such as factor analysis, are not concerned with temporal organisation (WIEPKEMA, 1961, BALTHAZART, 1972), while stochastic or sequence analysis derived from matrices are only valid if the data show stationarity, that is, if the frequencies of behaviours and of transitions between them do not change with time (DELIUS, 1969; SLATER, !) We are grateful to Dr L. S. EWING, Professor A. MANNING, Professor J. MAYNARD SMITH and Dr P. SLATER for their helpful and constructive criticisms of the manuscript. Dr F. VON SCHILCHER kindly prepared the german summary.



1974; VAN DER KLOOT & MORSE, 1975). This is, for reasons already discussed, probably infrequent in complex behaviour although the courtship of one of the species of characid fish studied by NELSON (1964), Coryno-poma riisei, was stationary, as were the transition probabilities between song elements in Cardinal birds (LEMON & CHATFIELD, 1971). In a previous comparative study of aggressive behaviour in rivulins, evidence was presented which suggested that considerable non-stationarities existed for the individual elements comprising aggressive behaviour (EWING, 1975). In this paper we describe how the behaviour of Aphysemion striatum changes during the course of an aggressive encounter and suggest a mechanism which accounts for the pattern observed. MATERIALS AND METHODS The species of killifish used in this study was Aphyosemion striatum (Boulenger) form north-western Gabon. This species is relatively hardy and easy to breed. It is a small fish, males reaching a maximum total length of about 50 mm, females being somewhat smaller on average. Males are beautifully patterned with a bronze-green ground colour to the body along which are five rows of red spots. The caudal, dorsal and ventral fins are barred longitudinally with bands of red, yellow and blue. Wild caught fish were obtained and the fish used in this study were all the first and second generation progeny from these. Methods of maintaining stocks of fish and for recording aggressive behaviour have been described in detail previously (EWING & EVANS, 1973; EWING, 1975) and an abbreviated account is given here, except where the procedures differ in some major respect. Eighteen young, sexually mature males were removed from the stock tanks and kept visually isolated in paritioned aquaria 60 X 25 X 25 cm. Aggressive behaviour followed removal of the partition. The behaviour was recorded by voice on a tape recorder and subsequently transferred to paper along with a time scale. At the end of an encounter the fish were separated. Three series of fights were recorded of which the first, consisting of eight fights, was observed after a minimum period of eight weeks visual isolation. In the second series the fish were repaired so that winners fought winners and losers fought losers from the first series. The encounters were arranged between 28 and 32 days after the first fights. Finally after a further period of 24 - 26 days separation a third series of five fights was observed of which one was between fish which had won twice previously and one between fish which had lost twice. The remaining three fights were between fish which had both lost and won one fight. No significant differences were found in the behaviour performed in the three series. The object of the method of pairing used was to try and avoid one-sided encounters. The total number of fights observed was 21. AGGRESSIVE BEHAVIOUR The aggressive behaviour of a number of related species has been described by EWING & EVANS (1973) and EWING (1975) along with some evidence and speculations concerning the

functions of the various elements. The qualitative aspect of the behaviour differ little between species and only a brief description is given here. Fin clamp (FC): All the fins are folded and held against the body. Sigmoid posture (SP) : The body is flexed so that the tail points towards, and the head away from, the opponent. The fins are usually folded. Full display (FD): All the fins are maxi



mally extended. Quiver (00) : The fins are partially folded and the posterior portion of the body makes high frequency, low amplitude movements directed at the opponent's head. Tail beat (TB) : Slow, powerful, large amplitude movements of the tail region with all fins spread maximally and directed at the opponent's flank. Attack (AT) : Attempts to bite directed at the flanks and head. Jaw lock (JL) : The consequence of mutual bites to the head. ANALYSIS OF AGGRESSIVE BEHAVIOUR The data from all fights were analysed initially using a computer programme which gave the grand total and the separate totals for each behaviour element and for the transitions between them for individual fish. Sequence diagrams derived from the computer analysis provide a visual representation of the data which is more comprehensible than a matrix. This is given in Fig. i where frequencies of elements and of transitions between them have each been transformed into percentages of the totals and the diagram drawn to scale. This provides a fairly complete summary of the data in that only transitions with frequencies of less than o.i per cent have been excluded (the total percentage of transitions rejected was 0.31).

Fig. 1. A sequence diagram to illustrate the elements of aggressive behaviour and the transitions between them. Percentage values are given for both and are proportional to the diameters of the circles and to the widths of the arrows. Only transitions with frequencies of less than 0.1 per cent have been excluded. FC: Fin Clamp, SP : Sigmoid Posture, 00: Quiver, TB: Tail Beat, AT: Attack, JL: Jaw lock, FD: Full Display. Number of transitions, 5913, Number of elements, 5941, Number of fights, 21.



Although Fig. i contains no information concerning non-stationarities or inter-individual variability, some initial conclusions concerning the patterning of the aggressive encounters can be drawn. Almost all elements within individuals occur in alternation with FD and the transitions FD ^ QQ, FD^TB and FD^AT account for 95.3 per cent of all transitions. The only diadic transitions not involving FD are, (i) QQ-> TB, which is a rare event, occurring when the recipient of the behaviour moves during quivering so that QQ becomes directed at the flank and not the head and therefore becomes transformed into TB. (ii) AT-> JL, and here the definition of both behaviours is such that JL can only follow from AT. (iii) FC ^ SP, the possible reasons for which will be discussed later. NON-STATION ARITY One source of non-stationarity which is obvious from direct observation of fighting is the way in which the tempo of fights increases with time. That is, the number of different behaviour elements, and therefore transitions between them per unit time increases during the progress of a fight. This

4 0

Quarters (by iime 1
Fig. 2. Percentages of each of the aggressive elements and for the total number of elements performed (plotted on a log scale) for quarters of fights divided on the basis of time.



can be demonstrated simply by dividing fights into quarters by time and calculating the number of elements occurring in each quarter. These are plotted on a percentage basis in Fig. 2 along with the values for each of the individual elements. Frequencies of SP and JL are too low to fit on this graph but, as could be expected from the associations of these elements seen in Fig. 1, the former follows closely the distribution in time of FC and the latter that of AT. The total percentage of behaviour elements performed in the first quarter is 12.7 compared with 35.2 in the last. For clarity no indication of inter-individual variability is given on this graph. However the non-stationarity in tempo can be shown to be significant: in 17/21 fights more elements were performed in the last quarter than in the first (p<o.025, Binomial test). Three out of the four fights in which more behaviour elements were to be found during the first quarter were the three shortest fights and the sample size in each case was therefore small. The graph also suggests that the non-stationarity is not merely in the tempo of the encounters but that the individual elements change their frequency of occurrence with time. QQ and FC become less frequent and the latter perhaps becomes more common again during the last quarter. AT increases in frequency at a faster rate than does the total for all behaviours. In order to eliminate the non-stationarity due to changes in tempo, fights were divided into quarters on the basis of the numbers of elements performed
2 0



and the percentage values for each element was recalculated. The result is plotted in Fig. 3 and shows the trends for the different behaviours. FD, as it alternates with the other elements and has an average frequency of almost 50 per cent, remains near its expected value of 12.5 per cent in each quarter, QQ and FC decrease with time as does TB while AT increases. It therefore seems possible that there are non-stationarities in all the behavioural elements except for FD. These were investigated, where possible, by calculating regressions for each element against time for individual fish. The following regressions were calculated: QQ duration against time, inter-QQ interval against time, inter-TB and inter-AT intervals against time. Changes in the duration of the latter two elements were not tested as each lasts for about one second and could not be measured accurately with the method of recording behaviour used. TB and AT were therefore considered to have fixed durations. The frequencies of JL, SP and FC were not sufficiently high to enable regressions to be carried out and, furthermore, from Fig. 3 it seems possible that FC is not linearly distributed. The Friedman two-way analysis (SIEGEL, 1956) was applied to the distribution of FC, SP and JL by quarters and the results are given in Table 1. All show significant non-stationarity, JL increasing in frequency each quarter and the two former elements decreasing in frequency over the first three quarters and then showing an increase during the last one. TABLE 1 Results o f Friedman two-way analysis o f variance by ranks demonstrating that the distribution o f FC, SP and J L is significantly different in the four quarters o f fights, the fights being divided on the basis o f the number o f behaviour elements performed as in Figure 3 X2 FC SP JL 16.0 12.8 11.3 df p 3 3 3 < 0.01 < 0.02 < 0.02

JL was present in only 7/21 fights and the analysis was carried out using only the data from these fights. The regression analyses were calculated for individual fish and thus the maximum number of possible regressions is 42. However in some short fights, where n<3 for an element, regressions could not be calculated. The results are given in Table 2 from which it can be seen that both the duration and the frequency of QQ decreases with time while the frequency of AT increases. The situation with regard to TB is more equivocal and there appear to be two opposing trends with inter-TB interval becoming shorter



TABLE 2 The distribution o f the four major elements o f behaviour with respect to time Positive Negative Regression QQ duration/time inter-QQ interval/time inter-TB interval/time inter-AT 16 19 5 35 0.029 2'3 — 0.0040 slopes 2 slopes 15 6 Number of significant slopes 0 3

Overall p N for any trend 0.004 1 0.012 2

Weighted 4 mean slope — 0.0038 0.0138



1 Tested by the distribution of positive and negative slopes 2 Tested by the distribution of significant slopes 3 Of the 5 significant slopes, 2 were positive and 3 negative 1 4 Weighting = -------- =var. b in some animals and longer in others as the fights progress. However this is explicable on the following basis. In those individuals showing high frequencies of attack, inter-TB interval decreases with time and, conversely, where AT frequency is low, TB increases. This can be shown by plotting inter-TB regressions against AT frequency where a significant positive regression is obtained (see Figure 4). FC and SP have a bimodal distribution with time, being found at the beginning and end of encounters while the frequency of JL increases in parallel with that of its associated element, AT. Thus all the elements of behaviour except for FD show changes during the course of a fight.
0-2 c

< v 0-1 -


< U
~ c m t< -0-2 D _ " c -03 -0-1



20 30 percentage AT

Fig. 4. A plot of inter-TB interval regression against AT percentage for 34/35 individuals for which the former was obtained. One aberrant regression was rejected after applying an outlier test (SNEDECOR & COCHRAN, 1967). The regression is significant (p<o.oi) : y = —0.1611 + o.6oo6x.



The data so far presented are consistent with the idea that the individual elements of behaviour reach maximum expression at increasing levels of aggression throughout the course of a fight. This is corroborated by the fact that overt attacking, representing the highest level of aggression, increases in frequency during fighting. The sequence QQ, TB, AT, almost certainly incurs increasing energy expenditure and each successive element is potentially more damaging to the opponent. The increase in tempo which we have demonstrated could also be a reflection of the same trend. JL does not occur at a sufficiently high frequency to enable one to fit it into a hierarchical schema. However, in some related species it may indicate a higher level of aggression than AT as, in these, it is prolonged and usually terminates a fight (E WING, 1975). FC could be a signal of unwillingness to fight. Evidence for this is that in 18/21 fights the last element performed by the losing fish was FC while no winner showed this behaviour at this time. SP appears to represent an ambivalent state between aggression and flight. Thus in 4/18 fights SP preceded FC as the penultimate element of a loser's behaviour and, in one case SP terminated a fight. Further, as can be seen in Fig. 1 SP is often intermediate between FC and FD. The latter posture is a clear visual indication of readiness to fight and contrasts strongly with FC. Thus FC (and SP) are found at the beginning of encounters when levels of aggression are low, almost disappear later when aggression is high and then reappear in losers only at the very ends of fights. FD alternates throughout a fight with the other elements of behaviour but it is also seen in the absence of an opponent. Isolated males will often spend long periods of time motionless in FD. When another male is introduced to them, they usually swim slowly towards it mainly using movements of the pectoral fins, while FD becomes accentuated and the colours darker and brighter. FD is also found during courtship. For this reason it is safest to consider that FD is indicative of arousal. There are features of the above descriptive model that require further explanation. Why, for example, does the level of aggression continue rising throughout the encounter and why, at the end, does it appear to fall precipitously? That the end is abrupt can be demonstrated by looking again at the ends of fights in detail. For winners, in 12/21 cases the last behaviour performed was AT. For losers the element preceeding FC (or SP -> FC) was also AT in 13/21 fights. Thus the eventual winner did not gradually gain the ascendancy over its opponent: levels of aggression remained high in both individuals until the last seconds of a fight when FC from one animal signalled the end.



The most satisfactory model which accounts for the overall patterning of aggressive encounters is to consider that a positive feedback system is in operation such that the aggressive postures of one fish act to increase the level of aggression in the other. Such a system will account for three aspects of the data: the increase in tempo of the fight with more frequent transitions between elements occurring as the fight progresses, the sequential replacement of behaviours representing higher levels of aggression and the sudden termination of aggression due to overloading. In addition there is a prediction that can be made from the positive feedback model which is that during the course of a fight there should be matching in the levels of aggression shown by the two participants. That is, while the levels of aggression may be very different at the beginning of a fight they should be closer at the end because the more aggressive actions of the more aggressive fish will have a greater stimulatory effect on its less aggressive partner than the in reverse situation. We will examine this point in more detail not only because the prediction is an important one for the model but it provides clues concerning the adaptive reasons for the patterns of aggressive behaviour in this species of fish. INTER-INDIVIDUAL DIFFERENCES So far we have only shown that the behaviour of winners differs from that of losers in that, at the end of a fight, the latter frequently show FC or SP-^ FC. In order to examine further any possible differences between the behaviour of these two classes of individual a modified discriminant function analysis was carried out. The modification was required because of the dependence of the winner — loser classification within each fight and was as follows. In each fight, the frequency of each element of behaviour in the loser was subtracted from that of the winner and the matrix so derived was transformed into a set of simultaneous equations by equating the behavioural differences within each fight to an arbitary constant of + i. A least squares solution was derived for the coefficients of the behavioural differences and then the 'discriminant' score for each fight was calculated from the solution. If there were no differences in the behaviour of winners and losers, these scores should be symmetrically distributed about a mean of zero. A x 2 test was used to determine whether this was the case. On this basis it is possible to separate the behaviour of winners and losers (p<o.05). The equation was, 0.5894 FD — 0.1044 QQ — °-5775 TB — 0.7402 FC — 0.1694 AT — 0.0362 SP ^ o. Our positive feedback model predicts that these differences will be


greatest at the beginning of fights and we therefore tested for this by repeating the above test using the data from first and last quarters of fights. Differences could still be predicted in both quarters. The major differences in the first quarter were in the frequencies of QQ and AT, both being higher in eventual winners (-0.1095 FD + 0.3638 QQ — 0.0566 TB + 0.0533 + 1.2329 AT + 0.0592 SP i= o). That winners show both more QQ and AT is at first sight anomalous in that the former is a less aggressive element. However an examination of the raw data shows that differences in QQ frequency only occur in those fights in which no attacks occur in the first quarter. In other words, these are fights with a slower rise time in the level of aggression and in this situation winners are initially still more aggressive than losers. A similar situation is seen where, in those fights with positive regressions for inter-TB interval, AT frequency is low and these are fights with a slower tempo (see Fig 4). The major difference between winners and losers in the last quarter was that the latter show more FC for reasons already discussed (-0.4111 FD + 0.4185 QQ + 0.1220 TB — 0.8784 FC + 0.1284 AT — 0.0592 SP


o). The difference in AT frequency found in the first quarter

has almost disappeared by the last and, as this is the most frequent element performed during the latter portion of encounters, provides additional evidence for matching. The fish progressively synchronise or match their behaviour throughout the course of a fight. Thus, although there is considerable variability between fights, one would anticipate some positive correlation of behaviour within fights even although initial levels of aggression can be different. One way in which this can be demonstrated is to calculate the correlation coefficient for inter-AT interval regressions between losers and winners of fights. This is r = 0.905 (df = 13, p<o.Oi), indicating a high degree of matching. However good this correlation is, it should be closer at the end than at the beginning of fights and a comparison of AT interval regressions for winners and losers is shown in Fig. 5. This provides an idealised picture but the differences in slopes are significant as are their intercepts (p<o.05 for both), with losers starting with higher inter-AT intervals which decrease more rapidly than those of winners. In other words, in the terms of the positive feedback model, the effect of the higher level of aggression in the eventual winner is to make the loser more aggressive while, conversely, the less aggressive behaviour of the loser has a smaller stimulatory effect on the winner and thus matching occurs. Figs 6a and 6b show samples of actual fights, one short and the other long. The short, six minute fight, is given in its entirety while five, one












Fig. 5. Mean regressions of inter-AT intervals against time for winners and losers of fights. The difference between the slopes is significant, p<o.os. Winners, y = 356 —0.37X, losers, y = 684 —0.9OX.

---------- 9-9---------- 9 ---------------------------------<J—o ------3-3-3 • 9 9—95 90 105 !2(


-tt' a

O Quiver 9 Tail Beat

V Fin Clamp Altack (~JL—


Jaw Lock
Full Display

Fig. 6a. Diagrammatic representation of a complete fight. W: winner, L: loser.

w a-o 600

----- g --- j

W — • ------- •-* --------- »900 915 930

Fig. 6b. As for 6a, but a longer fight with samples of the first, 6th, nth, 16th and last minutes. See text for a further description.


minute samples are given for the longer, twenty-five minute one. The following points are clearly illustrated: the increasing tempo of the behaviour, the progression from QQ through TB to AT, initial alternation of behaviour followed by synchrony and the final FC of the losing fish which terminates the encounter. In neither of these fights was FC found at the beginning. The synchrony illustrated in these figures, particularly for AT, is almost certainly not perfect and 'both AT and TB are probably initiated by one fish closely followed by the other. Our method of recording does not, unfortunately, enable us to record this. Many of the fights observed are prolonged (up to 30 minutes) and during this period the combatants do not apparently respond to major disturbances in their environment and this is clearly inadaptive where there are possible predators. It has previously been suggested that fights need to be long to enable physiological changes to occur which would consolidate the relative status of the individual fish (EWING & EVANS, 1973). We can now provide an alternative and more attractive hypothesis. Winning aggressive encounters usually confers selective advantages; territory, a mate or a food source. Unfortunately almost nothing is known about the behavioural ecology of Aphyosemion species but, under aquarium conditions, it can easily be demonstrated that dominant males are more successful in mating than subordinates (E WING, unpublished). Long fights constitute an adaptive strategy only if the fish which start at an initial disadvantage due to a lower level of aggression have some chance of winning. Examining the data from the first quarters of fights we find that out of ten fights in which AT occurred during this period, in one fight the animal which had a lower frequency of AT finally won and, in the remaining eleven fights, two animals showing lower QQ frequency won. These three fights were in no other way outstanding and were of approximately average length. Thus in 3/21 fights the outcome was against statistical expectation. The positive feedback between individuals effectively minimises any chance differences in initial levels of aggressive motivation and possibly helps to ensure that the fittest individual will win. If the benefits of winning are sufficiently great it is of advantage to a potential loser to continue fighting as long as possible and this will result in extended fights. DISCUSSION Control theory has been used frequently in physiology and, to a lesser extent, in behaviour. Models so produced have normally been concerned with negative feedback, a concept particularly useful in describing the maintenance of steady state systems and regulatory mechanisms. Examples are the



regulation of food and water intake (MCFARLAND, 1974), orientation mechanisms in insects (MITTELSTAEDT, 1964) the optokinetic responses of Crustacea (HORRIDGE, 1966), the control of flight in insects (e.g. LAND & COLLETT, 1974) and, one of the earliest and best known examples, prey capture in mantids (MITTELSTAEDT, 1957). These are readily described in terms of control theory and all involve negative feedback. Negative feedback, by definition, acts decrementally to promote stability: positive feedback, by contrast, is incremental in its effect and may result in an unstable or fluctuating output. The former is therefore more appropriate to homeostatic mechanisms within biological systems. However not all such systems are, at least in the short term, stable and it may be that the role of positive feedback has been partially overlooked. ROF.DER (1976) draws attention to the possible role of positive feedback in sustaining certain types of behaviour and neural processes. WILSON (1966) by means of computer simulation studies, demonstrated that positive feedback between neurons or populations of neurons might be used in the control of insect locomotion. The output from such a system shows increasing frequency of firing leading to run-away until fatique or accumulated refactoriness terminates firing. Such systems also tend to show synchrony. Another example from neurophysiology is in the control of swimming in the nudibranch Tritonia where the motor units innervating the dorsal longitudinal muscles interact through a positive feedback system to provide accelerating bursts of synchronous firing followed by a silent period (WILLOWS, 1967). WALDRON (1967) provides an example of positive feedback in the flight systems of locusts and quotes further examples from both invertebrates and vertebrates. The role of positive feedback in behaviour is less well documented. A model employing such feedback has been proposed to explain aspects of patterning in blackbird (Turdus meruda) songs (TODT, 1971). The probability of certain song phrases occurring is facilitated by the bird having heard a phrase containing similar starting notes within a prior critical period. Two birds can thus influence each others song so that some matching will occur and they produce songs of similar overlapping phrases. The situation has at least superficial similarities to that found in A . striatum aggression particularly towards the ends of fights where tail beats and attacks are almost synchronous. In one of the species of glandulocaudine fish (Pseudo-corynopoma doriae) studied by NELSON (1964), a saw-tooth pattern of male courtship activity was found suggesting the action of a positive feedback mechanism with cycles of increasing excitation followed by overload. It is therefore possible that the model proposed for the organisation of


aggressive behaviour in A . striatum may have a more general application. Elements of positive feedback could contribute to the patterning in time of some courtship and aggressive sequences. Stimulus — response chains have been used to account for the progression of behaviour sequences, however, such chains are seldom rigidly determined (MORRIS, 1958). Feedback mechanisms could, in such cases, ensure overall synchrony in the behaviour. In the well studied case of the Siamese Fighting Fish aggressive encounters often progress from display elements such as facing and gill cover erection through biting to jaw locking indicating an escalation of aggression. Termination of encounters in this species appears to be abrupt and thus similar to the situation in A . striatum. (BRADDOCK & BRADDOCK, 1955; SIMPSON, 1968). A consequence of the positive feedback between males of A . striatum during fighting is that matching occurs in levels of aggression. A similar consequence arises from the head-down threat posture of territory holding males of Haplochromis burtoni where the orientation of the eye-bar acts to raise the level of aggression in the recipient thus making him more prepared to withstand the attack of his opponent (HEILIGENBERG et al., 1972). A similar situation may exist in the cockroach, Nauphoeta cinerea. When females are introduced into a stable heirarchy of males the most dominant males start courting first and release the pheromone 'seducin'. One effect of seducin is to sexually stimulate females. Another is to raise the levels of aggression in males and following the renewed fighting some of the previously low-ranking males may suceed in mating (EWING, 1972 and per. comm.). The adaptive value of aggression is generally accepted but until recently there has been little attempt to relate specific patterns of aggression to evolutionary strategies. Two models based on games theory have been proposed to describe optimum fighting strategies from the standpoint of individual selection (MAYNARD SMITH & PRICE, 1973; MAYNARD SMITH, 1974; PARKER, 1974). MAYNARD SMITH (1974) divides fights into 'tournaments' and 'displays', and his model is concerned with the latter which he defines as a contest in which no physical contact takes place and the winner is the contestant which continues displaying longest. A tournament involves contact, a test of 'strength' and possible damage. A . striatum fights do not fit into the former category as they almost always escalate from displays into tournaments. The feature of Parker's model is that it requires the combatants to assess their relative Resource Holding Power (RHP) which can be defined as absolute fighting ability. It also requires that these estimates are made following bouts which cause injury and thus change fitness and



therefore the probability of winning the fight. This seems too simple a model to account for the behaviour of A . striatum which is not bouted and any assessment would have to be continual. Further, it is not clear in the fights of A , striatum what information fighting individuals could use in assessing relative RHP as the only consistent differences found were in levels of aggression which are not necessarily related to RHP. In a related species A . bivittatum, EWING & EVANS (1973) could find no correlation between characteristics which would provide information on RHP, such as body size, colour or fin area, and winners and losers of fights and the same appears true of A . striatum. A . striatum males will fight in an escalating fashion even when the probable outcome is clear from early in an encounter. This may be a consequence of staged fights in an aquarium or of domestication and such fights may not be found in the wild. If this is so then, while it is still valid to investigate mechanisms underlying the patterning, it is not possible to attempt adaptive explanations. Without field observations this question cannot be resolved. However the fish used in these experiments were only the first and second generation progeny of wild caught parents and are therefore unlike to be significantly different genetically from wild fish. All Aphyosemion species inhabit small bodies of water and the size of the aquaria do not appear to place unnatural constraints on the fish. One might expect that, under natural conditions, long fights would be less common due, perhaps, to habitation and encounters between grossly mismatched fish. Again, in the absence of ecological data this is not certain. We feel that the potentialities for escalating fights exist and therefore such fights are likely to occur under natural conditions. We have already suggested that escalating fights leading to matching would minimise differences in the initial level of aggression. Such differences could be due factors unrelated to reproductive fitness such as fluctuations in arousal levels, prior activities such as feeding and courtship or attack by predators. There still remains the problem as to why a fish will continue fighting even when the possibility of it losing is considerable. It could be that, unless the initial discrepancy in levels of aggression is very great, the fish are unable to assess their relative RHP and the only strategy in this situation would be to fight as long as possible. This would be analogous to the situation proposed by MAYNARD SMITH (1974) for displays which contain no information concerning RHP. Another possibility is that the consequences of losing are so severe, that is, the probability of reproducing is greatly reduced, that it is worth while escalating even if the probability of winning is low. Field studies are required to resolve this problem.



SUMMARY The fighting behaviour of Aphyosemion striatum is described. Particular attention was paid to the changes that occurred in the pattern of aggression during the course of an encounter. It was shown that the tempo of fights increased, that is, as the fight progressed the number of elements of behaviour per unit time increased. The three most frequent elements were Quivering, Tail Beating and Attacking and it was shown that they succeeded each other, in maximum frequency of expression, in the order given above, throughout a fight. Fin Clamp, occurring at the beginning of fights and, in losers, at the very end, signalled unwillingness to fight. Almost all the elements alternated with Full Display which was indicative of arousal. The increasing level of aggression followed by sudden defeat of one fish could be accounted for in terms of a positive feedback interaction between participants followed by overloading in one animal. A major prediction of this model is that progressive matching in levels of aggression will occur during a fight and this was shown to happen. An adaptive reason for this pattern of aggressive behaviour is proposed, which is that it can minimise the effects of factors unrelated to reproductive fitness in deciding the final outcome of a fight. REFERENCES macrochir. (Boulenger 1912). — Behaviour 46, p. 37-72. ---- (1974). Non-stationarite et aspects fonctionnels du comportement agonistique chez Tilapia macrochir (Boulenger, 1912) (Pisces: Cichlidae). — Acta Zoologica et Pathologica Antverpiensia, 58, p. 41-50. BRADDOCK, J. C. & BRADDOCK, Z. I. (1955). Aggressive behavior among females of the Siamese fighting fish, Betta splendens. — Physiol. Zool. 28, p. 142-172. DELIUS, J. D. (1969). A stochastic analysis of the maintenance behaviour of skylarks. — Behaviour 33, p. 137-178. EWING, A. W. (1975). Studies on the behaviour of Cyprinodont fish, II. The evolution of aggressive behaviour in Old World rivulins. — Behaviour 52, p. 172-195. -----& EVANS, V. (1973). Studies on the behaviour of Cyprinodont fish, I. The agonistic and sexual behaviour of Aphyosemion bivittatum (Lonnberg 1895). — Behaviour 46, p. 264-278. EWING, L. S. (1972). Hierarchy and its relation to territory in the cockroach Nauphoeta cinerea. — Behaviour, 42, p. 152-174. HEILIGENBERG, W., KRAMER, U. & SCHULZ, V. (1972). The angular orientation of the black eye-bar in Haplochromis burtoni (Cichlidae, Pisces) and its relevance to aggressivity. — Z. vergl. Physiol. 76, p. 168-176. HORRIDGE, G. A. (1966). The optokinetic response; a study of a system. — Symp. Soc. exp. Biol. 20, p. 179-198. KI.OOT, W. van der & MORSE, M. J. (1975). A stochastic analysis of the display behavior of the Red-breasted merganser (Mergus serrator). — Behaviour 54, p. 181-216. LAND, M. F. & COLLETT, T. S. (1974). Chasing behaviour of house-flies (Fannia canicularis). A description and analysis. — J. comp. Physiol. 89, p. 331-357. LEMON, R. E. & CHATFIELD, C. (1971). Organisation of song in cardinals. Anim. Behav. 19, p. 1-17. MAYNARD SMITH, J. & PRICE, G. R. (1973). The logic of animal conflict. — Nature (London) 246, p. 15-18. ---- (1974). The theory of games and the evolution of animal conflict. — J. theor. Biol. 47, p. 209-221. MCFARLAND, D. J. (1971). Feedback mechanisms in animal behaviour. — Academic Press, Lopdon and New York.

BALTHAZART, J. (1972). Analyse factorielle du comportement agonistique chez Tilapia



(1957). Prey capture in mantids. — In: Recent Advances in Invertebrate Physiology, Ed. B. T. SCHEER, University of Oregon Publications, Eugene. ---- (1964). Basic control patterns of orientational homeostasis. — Symp. Soc. exp. Biol. 18, p. 365-386. MORRIS, D. (1958). The reproductive behaviour of the ten-spined stickleback (Pygosteus pungitius L.). — Behaviour Suppl. 6, p. 1-154. PARKER, G. A. (1974). Assessment strategy and the evolution of fighting behaviour. — J. theor. Biol. 47, p. 223-243. ROEDER, K. D. (1976). Feedback, spontaneous activity, and behaviour. — In: Essays on Function and Evolution in Behaviour, Ed. BAERENDS, G. P., BEER, C. & MANNING, A. p. 55-70. Clarendon Press, Oxford. SIEGEL, S. (1956). Nonparametric statistics for the behavioral sciences. — McGraw-Hill, New York. SIMPSON, M. J. A. (1968). The display of the Siamese fighting fish, Betta splendens. — Anim. Behav. Monograph 1, p. 1-73. SLATER, P. J. B. (1973). Describing sequences of behaviour. — In: Perspectives in Ethology, Ed. P. P. G. BATESON & P. H. RLOPFER, p. 131-153. Plenum, New York. SNEDECOR, G. W. & COCHRAN, W. G. (1967). Statistical Methods, (sixth edition). — Iowa State University Press. TODT, D. (1971). Interactions between periodicity and positive feedback controlling the vocalization of alternative motor-patterns in blackbird song. — Biokybernetik 3, p. 168-172. WALDRON, I. (1067). Mechanisms for the production of the motor output pattern in flying locusts. — J. exp. Biol. 47, p. 201-212. WIEPKEMA, P. R. (1961). An ethological analysis of the reproductive behaviour of the bitterling. — Archs neerl. Zool. 14, p. 103-199. WILSON, D. M. (1966). Central nervous mechanisms for the generation of rhythmic behaviour in arthropods. — Symp. Soc. exp. Biol. 20, p. 109-228. ZUSAMMENFASSUNG Das Aggressionsverhalten von Aphysemion striatum wird beschrieben. Frequenz-anderungen der einzelnen Verhaltenselemente wahrend eines Kampfes wurden besonders beachtet und es zeigte sich, daB die relative Haufigkeit mit der die einzelnen Elemente auftraten, zunahm. Die drei haufigsten Elemente waren Quivering (Schwanzzittern), Tailbeating (Schwanzschlagen) und Attack (direkter Angriff). Sie erreichten ihre Maximalfrequenzen in der angegebenen Reihenfolge. Finclamp (Zusammenklemmen der Flossen) wurde oft am Anfang eines Kampfes beobachtet, auQer bei Verlierern, die hiermit am Ende eines Kampfes ihren Unwillen den Kampf fortzusetzen signalisierten. Full display (Prasentieren der Breitseite mit aufgestellten Flossen) deutete auf all-gemeine aggressive Erregung hin und wechselte mit alien anderen Verhaltenselementen ab. Die wahrend eines Kampfes ansteigende Erregung konnte durch ein positives Riick-koppelungsmodell gedeutet werden, wobei im unterlegenen Tier tlberladungserscheinun-gen zur Kampfaufgabe fuhren. Eine Bestatigung dieses Modells wurde in der wahrend eines Kampfes zunehmenden Ubereinstimmung der Aggressionsintensitat der beiden Partner gefunden. Die evolutionsbiologische Bedeutung dieses Phanomens wird darin gesehen, daB es den EinfluB derjenigen Faktoren, die fur die allgemeine „Fitness" eine untergeordnete Rolle spielen, auf das Endresultat eines Kampfes auf ein Minimum reduziert.


Sign up to vote on this title
UsefulNot useful