(With 3 Figures) (Acc. 15-IX-1994)

Transcription

1 FEMALE PREFERENCE FOR NESTS WITH EGGS IS BASED ON THE PRESENCE OF THE EGGS THEMSELVES by SARAH B.M. KRAAK and TON G.G. GROOTHUIS1) (Zoological Laboratory, University of Groningen, P.O. Box 14, 9750 AA Haren, the Netherlands) (With 3 Figures) (Acc. 15-IX-1994) Summary In many fish species of which males care for eggs in a nest, including Aidablenniusphynx, females prefer to mate with males that already guard eggs. In this paper we present two aquarium experiments with this Mediterranean blenny to determine whether the females actually use the presence of eggs as a cue or depend on male display behaviour signalling the presence of eggs. In experiment 1 the test-female was presented with two nests, one with eggs and one without eggs, and only one male in a closed container between the two nests. This male only served to stimulate the female. In all tests females preferred to lay eggs in the nest already containing eggs. This shows that females of this species can base their choice on the presence of the eggs alone. In experiment 2 the test-female was presented with two males in nests of which only one had eggs. Transparent partitions deprived the female of the opportunity to inspect the nest contents. The male with and the male without eggs did not differ in time spent displaying towards the female, and the female did not show a preference for either male, based on visual cues or odours. After removal of the partitions the female visited both nests. The nest she visited first was equally likely to be the nest without eggs as the nest with eggs. In contrast, the female finally deposited eggs preferentially in the nest with eggs. We conclude that in this experiment females did not base their preference for males with eggs on male behaviour, but on the nest contents after inspection. This does not exclude an influence of male behaviour under other conditions. Furthermore, females generally deposit their eggs adjacent to as many as possible of the eggs already present, suggesting that females can localize and possibly estimate the amount of eggs already present. 1) The collection of the fish in Corsica in 1993 was financed by the 'Dr. J.L. Dobberke Stichting voor Vergelijkende Psychologie' and the earlier fieldwork by KNAW (Royal Dutch Academy of Sciences), NWO (Dutch Organisation for Scientific Research), ASAB (Association for the Study of Animal Behaviour), and GUF (Groningen University Fund). The marine institute STARESO in Corsica provided good accomodations. We want to thank Jan NIJBOER for making all the partitions and containers as soon as we needed them, and Dick VISSER for drawing the figures. This paper has benefited from discussions with J.P. KRUIJT, G.P. BAERENDS, Gerrit DE Vos, Dave Vos and Albert Ros and from comments by Theo BAKKER. This work is part of a PhD thesis by Sarah B.M. KRAAK (University of Groningen).

2 190 Introduction Most of the recent studies on female mate choice address the adaptive significance of female preferences (HARVEY & BRADBURY, 1991). The mechanisms that females employ in their choice of mate, i.e. the question which cues trigger the females' decisions, have received far less attention. For several fish species with male care for eggs in a nest, it has been reported that females prefer to mate with males that already guard eggs (e.g. three-spined stickleback: RIDLEY & RECHTEN, 1981; GOLDSCHMIDT et al., 1993; river bullhead: MARCONATO & BISAZZA, 1986; fathead minnow: UNGER & SARGENT, 1988; fantail darter: KNAPP & SARGENT, 1989; sphynx blenny: KRAAK & VIDELER, 1991). Although this preference seems to be firmly established, it is still uncertain whether the females actually use the presence of eggs as a cue, or, instead or additionally, male behaviour (or other attributes) that correlate with the presence of eggs. In particular for the stickleback a controversy exists. RIDLEY & RECHTEN (1981) have found in their sequential choice experiments that the presence or absence of eggs did not influence the latency time of females following the male to the nest. The females made their decision to spawn or to refuse after they had put their snout into the nest. Hence, RIDLEY & RECHTEN concluded that females did not base their choice on differences in courtship level between the two males, but on the presence or absence of the eggs themselves. However, they have not excluded the possibility that the females were more stimulated to deposit eggs by more vigourous courtship behaviour of the male with eggs. JAMIESON & COLGAN (1989) criticized RIDLEY & RECHTEN's data on statistical grounds. Furthermore they had doubts about their interpretation that there is no influence of male courtship behaviour, because in sequential choice tests male-male competition is absent. Hence, RIDLEY & RECHTEN could not detect a differential female response depending on differences in male behaviour that arise during male-male competition. JAMIESON & CoLGAN (1989) showed that males that had recently spawned tended to court more intensely and were more successful at leading females to their nests. In their simultaneous choice tests this resulted in females being more likely to spawn in the nest with eggs than in the nest without, but there was no evidence of females actually choosing on the basis of the presence or absence of eggs. However, they did not test directly the possibility that

3 191 male courtship and presence of eggs both influenced the females' decisions. In response to this discussion, GOLDSCHMIDT et al. (1993) reported an experiment in which females were subject to a simultaneous choice test. In their experiment a female was introduced in a separate tank placed in front of two tanks each containing one male, of which only one had eggs in his nest. They measured the duration of the female's head-up posture directed towards the male with eggs and the one without. The duration of the female's head-up posture towards a male was assumed to reflect female preference, since this behaviour positively correlates with the probability of spawning with that male (McLENNAN & MCPHAIL, 1990). The females showed no significant difference in this behaviour towards the two males. Apparently the females did not express any preference based on seeing the males performing courtship display. GOLDSCHMIDT et al. did not provide data that tell whether the males with and without eggs differed in courtship rates, therefore it is not known whether the females actually had a cue to base their preference on. GOLDSCHMIDT et al. concluded from their study that the female preference must be a direct consequence of the presence of eggs themselves. However, in their experiment a complete interaction between female and male was not possible, and it was not tested whether the female was actually more likely to spawn in the nest containing eggs. In this paper we present two experiments to determine whether females Aidablennius sphynx base their preference for nests with eggs on male behaviour signalling nest contents or on the nest contents themselves. The first experiment was designed to investigate whether females are able to choose a nest with eggs over a nest without eggs in a test situation that excludes the possible influence of differences in male behaviour. We did this by providing the female with the choice between two nests, one with eggs and one without, both nests being without a male. We used only one male in a separate container to stimulate the female to lay eggs. The second experiment was designed to investigate whether females prefer males with eggs without having the opportunity to inspect nest contents, i.e. whether females base their choice on male traits correlating with the presence of eggs. This experiment is basically similar to GOLDSCHMIDT et al.'s experiment: a female is presented with two males of which only one has eggs in his nest, the female being deprived of the opportunity to

4 192 inspect the nests inside. However, we added some relevant procedures. Firstly, we measured the time that both males displayed (in addition to the measurement of female behaviour). Secondly, we allowed the female to subsequently visit the nests and deposit eggs, in order to find whether the female actually preferred to spawn in the nest with eggs. Thirdly, we wanted to distinguish between the situation in which the female could only use visual cues of the males, and the situation in which odours (from the males and the eggs) could also be used. We did this by separating the female from the males by a non-perforated partition and a perforated partition respectively. Pheromones are known to play a role in fish reproductive behaviour (review HARA, 1993). Evidence suggests that in Blennius pavo males use pheromones secreted from special glands on the anal fin rays to attract females (LAUMEN et al., 1974). With respect to the adaptive significance of female preference for nests that already contain eggs, several suggestions have been made (see KRAAK & VIDELER, 1991). ROHWER (1978) proposed that, if males daily consume a certain amount of eggs in order to remain in sufficient condition for guarding the rest of the brood, it pays females to prefer nests with many eggs, since thereby they dilute the risk of their own eggs being cannibalized. We hypothesized that it would be advantageous for females to position their eggs in nests not in a random way but in such a spatial relation to the eggs already present that the risk of being eaten or infected is minimized. Therefore we investigated the location of eggs deposited by females in relation to the eggs already present. Methods Species and housing. We chose the Mediterranean fish Aidablenniusphynx as study species. For this species it is known that in the field females prefer males guarding >100 eggs (KRAAK& VIDELER, 1991). It is also known that paternal males in the field cannibalize on average 36 eggs per day from their brood, resulting in a significant decrease of the risk per egg of being cannibalized with increasing brood size. This probably constitutes the adaptive significance of the female preference (KRAAK, 1994a; KRAAK & WEISSING, 1994). A. sphynx is slightly sexually dimorphic: males are larger than females (adult males range from mm total length, whereas adult females range from mm total length), and, although males and females have the same patterns of coloration, males are more brightly coloured. During the breeding season males defend crevices in the rock as a nest, in which they continuously care for up to 7000 eggs. Females generally lay several hundreds of eggs every two to three days in these nests. Both males and females mate promiscuously. When a female approaches a nest, the resident male may perform courtship display, which consists of coming partly out of the nest with the dorsal fin erect and is often accompanied

5 193 by nodding. When the female approaches closer she may present her white swollen belly to the male. This female display appears to reflect female motivation to mate (KRAAK& VIDELER, 1991). Irrespective of whether the resident male displays or not the female may subsequently enter the nest. Females may leave the nest without laying eggs, or they may spend 15 minutes to several hours laying eggs, which are fertilized by the male, and then leave the eggs to the care of the male. The fishes were caught in the wild in May 1993 at the marine biology institute STARESO, Corsica, France. The experiments were conducted from June to August When not tested, the fishes were held in 50 or 60 litre tanks, usually 2 to 4 individuals per tank, males and females together. The tanks contained artificial nests in order to allow the animals to perform reproductive behaviour. Artificial nests consisted of small bricks with a hole containing a glass test tube. Eggs were laid in one layer against the tube wall. The number and the location of eggs in the nests could be monitored by drawing them on sheets of transparent plastic folded around the tubes. Water temperature was held between 20 C and 25 C. The day-night regime was 15h-9h. The fishes were fed on six days per week with frozen artemia or mysids. Experiment 1 Methods. The purpose of this experiment is to investigate whether females are able to choose a nest with eggs over a nest without eggs in a test situation that excludes the possible influence of differences in male behaviour. Figure I shows part of the experimental tank viewed from above. In the 60 litre tank a triangular transparent 3 litre container, completely closed at all sides and bottom, was placed, containing a nest and one male. On each side of the triangular container a nest was placed, in such a way that the entrances of the three nests were very close to each other. For each test, one of the outer tubes was removed and replaced by a tube with eggs which was taken from a male in another tank. The side where the tube with eggs was placed was alternated per test. The number and position of the eggs attached to the tube wall were recorded. The number of eggs varied between 130 and >L000. In each test one female was introduced in the aquarium. We checked for the presence of new eggs at intervals of 1-2 hours during daytime. The test ended when inspection revealed that the female had laid eggs in one or both tubes (after several hours to maximally two days). In this experimental setup the female did not have access to the tube occupied by the male in the triangular container. The male had no information on where the eggs were (no odours of eggs could diffuse into his closed container), thus he could not signal such by his behaviour. The male in his container was able to perform courtship behaviour when the female approached, but of course he could not go inside the nest with the female and fertilize eggs. Since there was only one male, there could not be any influence of differences in male behaviour or other male attributes on nest choice. Although the experiment did not exclude the influence of behavioural interaction between male and female on the female's response, the only choice available was between an empty nest and a nest containing eggs. Thirteen tests were performed with I I different females (two females were used twice). The same male was used in all tests. Before testing, the females had been laying eggs in other tanks with other males; however, during three or four days just before the test they were held in a tank with females only, just after having laid eggs. By this procedure females were expected to have ripe eggs on the day of testing, since females usually laid eggs every two to three days. The male had been receiving eggs from these and other females before, but during the entire period of testing he was in the experimental tank as described above, without eggs in his tube.

6 194 Fig. 1. Part of the experimental tank of experiment I viewed from above, with the transparent triangular container, the three nests, of which one on the side contains eggs, and the male and the female. A and B are the positions of the water in-flow and out-flow pipes respectively. Incoming water flows over the edge of the triangular container; a net prevents the male from leaving his container. Results. I shows that in all but two cases the female laid all her eggs in the tube that already contained eggs. In the remaining two cases one and two Table were laid in the empty tube, while 28 and I 10 respeceggs respectively to tively were laid in the tube with eggs. Hence, all females preferred = 11 the females tested in the nest with N spawn (of eggs (binomial test, twice only the first test is used), p = ). In seven cases some of the initially present had been eaten by the female before laying, leaving open the possibility that the female preferred to sit in the laid tube with eggs because of the presence of food, and then incidentally her eggs there. But even the result of six remaining choices (laying in the eggs that were tube with eggs before or without having from chance (binomial test: p = ). differs eaten eggs) statistically This result shows that females

7 195: TABLE 1. Some results of experiment l; female choice between nests with and without eggs a in these two tests the same female was used. b in these two tests the same female was used. prefer to spawn in nests containing eggs, and that they are able to base their decision on the presence of the eggs themselves. There is no correlation between the numbers of eggs initially present and the number of eggs laid (Spearman rank correlation test, p = 0.38). Furthermore, there does not seem to be a positive influence of the numbers of eggs present on the strength of the preference, since the two cases in which a few eggs were laid in the empty tube involve large initial numbers of eggs (Table 1). Finally, there was no correlation between the numbers of eggs present and the latency time before laying (Spearman rank correlation test, = p 0.32). This latency time might have been influenced by the behaviour of the male. However, later we found that females will even deposit eggs in nests in the complete absence of any male, meaning that stimulation by a male is not indispensable to get females to lay eggs. Experiment 2 Methods. In this experiment it was tested whether males with and without eggs behave differently, and whether females base their preference for males with eggs on differential male behav-

8 196 iour. Pairs of tests were done with the same trios of individuals: male A and male B (difl?eringmaximally 3 mm in length), and a female. In test I male A guarded a nest with eggs and male B guarded an empty nest. In test 2 male A guarded an empty nest and male B guarded eggs. The female had not been held with the males before (they might have met before in the field where they were caught). Just before testing the females had been held for 3 days in a tank without males, just after having laid eggs. By this procedure females were expected to have ripe eggs on the day of testing, as in experiment 1. The males had been held together without other males in a tank with two nests for at least one week before testing, so that each of them occupied and defended one of the nests. Before test I the males were held with a female (which was not used as test-female with these males). After this female had laid eggs in the nest of one of the males she was removed, the day after which test I took place. Note that in test I the male with eggs had been chosen by the egg laying female, suggesting that this male might have been the more attractive one; this situation was reversed in test 2 where the other male was given the tube with eggs. The day after test I the test-female was returned to the tank without males until she was used for test 2 with the same males three days later. Between test I and test 2 the males were held with another female, which sometimes laid additional eggs. On the day before test 2 this female was removed and the tubes were exchanged between the two males, so that the male that had eggs in his nest in test I had no eggs in test 2, and viceversa.just before test I as well as test 2 it was checked whether the nest with eggs still contained eggs; if this was not the case, the tube was replaced by a tube with eggs from another male which was not being tested (there were no indications that unrelated eggs were less well guarded than own eggs). The eggs in the tube were counted and their locations recorded (range , median 181 eggs). The tests started in the morning, 1-5 hours after the light was switched on. The males with their nests were placed in the test aquarium (Fig. 2) in the same positions as they were Fig. 2. The experimental tank of experiment 2, with the males in their artificial nests and the female, viewed from above.the partition between the males is transparent and perforated. Of the two transparent partitions between the female and the males one is perforated and the other is not. The non-perforated screen is removed for the second part of the test.

9 197 in their home tank, separated from each other by a perforated transparent partition. Perforations were too small to allow the fishes to swim through. The test-female was introduced into the test tank, separated from the males by two adjoining transparent partitions, of which one was perforated and the other not. In this way the female could not visit the nests to inspect their contents. The water circulation was stopped in order to prevent diffusion of water along the sides of the partitions. In this situation males and female could see each other, but no odours from the males or the eggs could diffuse into the female's compartment (this was checked once by adding methylene blue as a dye into the male compartments). A video camera positioned in front of the aquarium started recording. After a few minutes of acclimatization the video recorder ran for half an hour. Then the non-perforated partition was removed. The water circulation was restored in such a way that water-with chemical information from either the eggs or the males or bothfrom the left and right male compartments flowed through the perforations into the female compartment at the left and the right side respectively (which was checked by adding methylene blue to the male compartments). Again after a few minutes of acclimatization the video recorder ran for half an hour. Finally all the partitions were removed, and the video recorder was slowed down to one sixth of the normal speed for the rest of the day. Later the video tape was analyzed. Of the first two half hours the following behaviours were recorded: for females a. time present in each half of her compartment, b. time spent belly presenting to each male, for males c. time spent displaying. Time was recorded in seconds. Both female behaviours may be used to measure female preference. After removal of the partitions it was recorded which nest the female entered first and how much time the female spent in either nest. The following day it was checked whether the female had laid eggs in any of the nests, by comparing the number and the locations of eggs in the tubes to those recorded the day before. Data analyses. In total, ten different females were tested in test I, with eight pairs of males (two sets of males were used twice with a different female). Only eight of these females were tested in test 2 (with seven pairs of males). Therefore seven independent pairs of test I and test 2 were conducted. Firstly, we investigated whether female or male behaviour was influenced by the presence of eggs. Therefore we defined male A as the male that guarded eggs in test I and male B as the male that guarded eggs in test 2. Ratios were calculated as follows. The "time present ratio" is the proportion of total time that the female spent in the half of her compartment in front of male A. The "belly ratio" is time that the female presented her belly to male A divided by the total time that the female presented her belly to male A and male B. Higher ratios indicate a stronger preference for male A. The "display ratio" is the time that male A displayed divided by the sum of the times that male A and male B displayed. A higher ratio indicates a higher display rate of male A. If the females' or the males' behaviours were influenced by the presence of eggs, the ratios in test I are expected to be higher than those in test 2, since in test I only male A had eggs and in test 2 only male B had eggs. This was tested by paired t-tests with the seven pairs of tests, for the situation with and without the non-perforated partition separately, and also for these situations taken together. Secondly, since male A had aquired his eggs by having been chosen by a non-test female (before test I) whereas male B received the eggs by the exchange of tubes (before test 2), male A might have been more attractive or active than male B. Therefore, it was tested by t-tests whether the above mentioned ratios deviated from 0.5. This was done for the situations with and without the non-perforated partition separately, and for these situations

10 198 taken together, all within test I as well as within test 2. For the female behaviours N = 10 in test I and N = 8 in test 2, and for the male behaviour N = 8 in test I and N = 7 in test 2 (here the arcsine square root transformed ratios of the pairs of males that were used twice were averaged). Thirdly, to investigate a possible effect of odour of the males or the eggs on the female, it was tested by paircd t-tests, within test I (N = 10) as well as within test 2 (N = 8), whether the female behaviours differed between the situations with and without the non-perforated screen present. Proportions or ratios were arcsine square root transformed for the t-tests. For all t-tests two-tailed p-values are given. Finally, by binomial tests it was investigated whether after removal of the partitions the female was equally likely to visit first the nest with eggs as the nest without eggs, and whether the female was equally likely to deposit eggs in the nest with eggs as in the nest without eggs. This was done within test I and test 2 separately. Results. There was no significant difference in female behaviour between the situations with and without the non-perforated screen (proportion of time spent in front of male with eggs, test 1 t = 1.84, p = 0.10 and test 2 t = , p = 0.81; time spent belly presenting to male with eggs, test I t = 1.67, p = 0.13 and test 2 t = 2.08, p = 0.08; time spent belly presenting to male without eggs, test I t = 1.76, p = 0.11 and test 2 t = 1.06, p = 0.32). This indicates that the females did not behave differently when olfactory cues were available than when they were not. Therefore it seems reason- able to analyse the two situations taken together. Table 2 presents the results of both half hours, with and without the non-perforated screen, taken together. On average the females spent 44% of the time in the half of the compartment in front of the male with eggs (Table 2). This means that the average "time present ratio" was 0.44 in test I and 0.56 in test 2. Note that for testing whether the ratios differed between test 1 and test 2 (N = 7), case 5 and case 8 were left out since the same males had been used in case I and case 10 respectively (Table 2). The "time present ratios" did not differ significantly between test I and test 2 (with non-perforated screen t = 0.17, = p 0.87; without nonperforated screen t = -1.19, p = 0.28; together t = -0.68, = p 0.52). Time spent presenting belly varied considerably (range seconds per hour, see Table 2). Also the "belly ratios" did not differ significantly between test I and test 2 (with non-perforated screen t = 1.25, p = 0.26; without non-perforated screen t = -0.98, p = 0.36; together t = 0.32, p = 0.76). These results indicate that female preference was not influenced by whether the males guarded eggs or not. The males did not spend much

11 199 TABLE 2. Some results of experiment 2; female choice between males with and without eggs *a (most of) the initial eggs had disappeared (< 14 eggs remaining, whereas in all other cases > 64 eggs remained). *b after removal of the partitions, the male with eggs frequently took over the nest without eggs. *c after removal of the partitions, the male without eggs frequently took over the nest with eggs. In case I and 5 the same set of males was used; in case 8 and 10 the same set of males was used. time displaying (range 0-58 seconds per hour, see Table 2). The "display ratios" did not differ significantly between test I and test 2 (with nonperforated screen t = -0.86, p = 0.42; without non-perforated screen t = 1.39, p = 0.21; together t = 0.09, p = 0.93), indicating that male display rates were not influenced by whether the males guarded eggs or not.

12 200 None of the ratios deviated significantly from 0.5 ("time present ratio" test I with non-perforated screen t = 0.21, p = 0.84; without nonperforated screen t = -1.39, = p 0.20; together t = -0.73, p = 0.49; test 2 with non-perforated screen t = 0.68, = p 0.52; without non-perforated screen t = 0.63, = p 0.55; together t = 0.63, = p 0.55; "belly ratio" test 1 with non-perforated screen t = 1.60, p = 0.14; without non-perforated screen t = -0.55, = p 0.59; together t = 0.67, = p 0.52; test 2 with nonperforated screen t = 0.65, = p 0.54; without non-perforated screen t = 0.80, p = 0.45; together t = 0.82, = p 0.44; "display ratio" test I with non-perforated screen t = -0.10, p = 0.92; without non-perforated screen t = 2.18, p = 0.07; together t = 0.61, p = 0.56; test 2 with non-perforated screen t = 1.59, p = 0.16; without non-perforated screen t = 0.95, p = 0.38; together t = 1.31, p = 0.24). These results indicate that the females had no significant preference for male A or male B, and that male A and B did not differ significantly in display rate. Because in many cases the males did not spend much time displaying, it might be interesting to consider the subset of cases in which at least one of the males displayed more than 10 seconds. There are 8 of such cases (cases 2, 8 and 10 of test 1, and cases 1, 2, 4, 6 and 10 of test 2, Table 2). In 6 of these cases the male with eggs displayed more than the male without, and in 2 cases it was the other way round (binomial test, = p 0.14). Only in 4 of the 8 cases did the female present her belly more time to the male that displayed more (binomial test, p = 0.64), and once this was the male without eggs. Hence, there is no significant relation between diplay and presence of eggs, nor between display and female response. The nest that was visited first after removal of the partitions was equally likely to be the nest with eggs as the nest without eggs (binomial test: test 1 p = 0.62, test 2 p = 0.36). This suggests that until this moment the females exhibited no preference for the nest with eggs. In all tests the female had visited both nests before spawning in one of them. In test I the female laid eggs in the nest already containing eggs in 7 out of 10 cases; in the remaining three cases no eggs were laid. Hence, if the female mated, her preference to mate with the male with eggs was significant (binomial test, p = ). In test 2 the test-female laid eggs in only 3 out of 8 cases, twice in the nest with eggs and once in the nest without eggs, which does not constitute a significant preference to mate with the male with eggs (binomial test, p = 0.5).

13 201 In some cases, after removal of the partitions one male frequently took over the other nest, trying to monopolize both nests (Table 2). If this happened no eggs were laid by the test-female, and in most of these cases most of the initial eggs had disappeared between the start of the experi- ment and the time of checking the tubes the next day (Table 2). The eggs might have been eaten by the males that took over the other nest. These males behaved dominantly, and were always the males that had guarded a nest with eggs in test 1, i.e. the ones that had been chosen by the nontest-female before test 1. This suggests that traits of dominance coincide with traits of attractiveness. In three other cases (without nest take-overs) the initial eggs had disappeared. In the one case of test 1, the tube contained only 2 newly laid eggs the next day; it is not known whether the initial eggs had disappeared before or after these eggs were laid, nor whether in fact a clutch of more than 2 eggs had been laid of which all but 2 had disappeared. In summary, the results indicate that male display rate did not depend on the presence of eggs in the nest and that the females showed no preference for males with eggs based on visual or chemical cues. Only after having entered both nests, females preferred to lay eggs in the nest with eggs (only in test 1 ). Patterns Methods. of egg laying We wanted to investigate whether females lay their eggs in a nest preferentially in certain locations. During four seasons ( 1986, 1987, 1989 and 1992) of field study on A. sphynx at the marine research institute STARESO, Corsica, France, data were collected as follows. In the habitat of the fish artificial nests were placed, consisting of blocks of concrete with a hole containing a glass test tube. Males settled in these nests and females deposited eggs in them throughout the breeding season. Broods were monitored daily, by folding sheets of transparent plastic around the tubes on which the outlines of the areas occupied by eggs were drawn. By placing drawings of subsequent days of the same nest on top of each other, the areas of eggs that had been added or that had disappeared during the 24h interval could be determined. Over the four seasons thousands of such drawings have been made and analyzed (KRAAK& VIDELER, 1991; KRAAK, 1994a). We did not analyze the locations of the newly laid eggs systematically, but a small random sample of drawings will be presented. Results. In the thousands of drawings that have been made, the patterns of where newly laid eggs were located was very consistent and obvious. Therefore

14 202 we decided to present here only a small random sample of drawings to show these patterns. In Figure 3 some typical patterns of egg laying are shown. In nests that started out empty, eggs were laid singly or in groups of 2 to about 40 eggs (Fig. 3a and 3b). In nests that already contained many eggs it can be seen that the areas occupied by these eggs formed large patches (Fig. 3c, 3d, 3e). The eggs that were added were usually laid Fig. 3. Five examples of patterns of egg deposition in nests. The grey areas represent the surfaces of the nest wall occupied by eggs on day x; the black areas represent the surfaces of the nest wall where new eggs had been deposited between day x and day x+1. Dots represent single eggs. The outlines of the areas and the dots are copied (about 75% of true size) from the transparent sheets on which the outlines of the areas occupied by eggs and single eggs were drawn. In a. and b. the nests were empty on day x (no grey areas present).

15 203 along the edges of existing egg patches such that patches of eggs grew larger and eventually touched and became one bigger patch (Fig. 3c). If there was empty space in the middle of a large patch, new eggs were often laid along the edges of the empty space such that this empty space was filled up with eggs (Fig. 3d). When, by hatching or cannibalism, single eggs had disappeared somewhere in a big patch, resulting in empty spaces of the size of just one egg, such spaces were used to deposit new single eggs (Fig. 3e). Discussion With experiment I we have shown that females A. sphynx have a strong preference for nests with eggs over nests without eggs, based on the presence or absence of eggs alone. In experiment 2 no differences in display behaviour were detected depending on whether males guarded eggs or not. Females deprived of the opportunity to inspect the nests inside, did not show a preference for the male with eggs. In contrast, after the females had inspected the nests, they preferentially spawned in the nest with eggs. This experiment indicates that females based their preference for nests with eggs directly on the nest contents. Males did not provide information on nest contents by their behaviour or pheromones, or, if information were available, females did not use it, at least not in the experimental setup used here. However, it must be kept in mind that in test 2 of experiment 2 the male had not fertilized the eggs himself. In sticklebacks the presence of eggs influences a male's behaviour differently if the male has fertilized them than if he has not (VAN IERSEL, 1953). Therefore the results of test 2 might be less relevant. Furthermore, in experiment 2, while the partitions were present, the females could not approach the male to a smaller distance than about 10 cm. VVe cannot exclude the possibility that behaviour elements that naturally occur when females come closer, correlate with nest contents. For example, male "shivering" occurs only when females are at less than 5 cm distance (KRAAK, 1994b). As RIDLEY & RECHTEN (1981) already pointed out, males that signal in any way that they have no eggs would be selected against. Hence, cheating males are expected to evolve. In fact, RoHwER (1978) explained the phenomenon of "displacement fanning" in sticklebacks (VAN IERSEL,

16 , SEVENSTER, 1961), i.e. fanning without having eggs, as a male strategy to deceive females. Therefore, the actual presence of eggs is a more reliable cue than any male signal that might be deceptive. Hence, it seems to be functional that females can discriminate between nests on the basis of the presence or absence of the eggs themselves, in particular in species, such as A. sphynx, that deposit eggs in one layer on a hard substrate nest wall, making the number of eggs present easily assessable. The lack of a correlation between the strength of the preference and the number of eggs present in experiment I was expected, since KRAAK & VIDELER (1991) showed that female preference does not gradually increase with the number of eggs present, but instead rises steeply when nests contain about one hundred eggs; with up to several thousands of eggs the preference stays at more or less the same level. Since the probability of an egg being cannibalized decreases with brood size at a diminishing rate (KRAAK, 1994a; KRAAK & WEISSING, 1994), the marginal advantage of laying in larger broods becomes smaller. Therefore it is probably functional that females use a fixed threshold criterion. The same reasoning may explain the lack of correlation between the number of eggs initially present on the one hand and the number of eggs laid or the latency time to laying on the other hand. It is likely that the number of ripe eggs that a female has available and the readiness to spawn them are strongly under the influence of internal factors. Furthermore, the male's behaviour was probably not constant and this might have influenced the female's readiness to spawn. The presence of eggs is probably not the only factor that predicts the survival chances of a female's offspring. In addition the male's physical condition, vigour of male courtship display, male coloration, or male dominance may signal parental ability, "good genes" or predict the current level of filial cannibalism of the male at hand. KRAAK & VIDELER (1991) already showed that in A. sphynx not all variation in male mating success could be explained by the number of eggs present, suggesting that male attributes play an additional role in female mate choice. The fact that in experiment 2 no eggs were laid if the male with eggs was dominated by the other, suggests that the attractiveness of a nest with eggs and the subordination of the "owner" are confounding stimuli to the female, or that females are unwilling to lay eggs if males frequently change nests. Since in most of these cases the initial eggs had disappeared, the females

17 205 might also have been unwilling to lay eggs because neither of the tubes contained any eggs. The initial eggs were generally older in test 2 than in test 1. This might be an additional explanation for the tendency of females being less likely to spawn in test 2, though it has not been investigated whether the age of the eggs influences female decisions. The fact that females are able to deposit their eggs in such a non- random way, as shown in Fig. 3c, 3d and 3e, suggests that females acquire precise information as to which part of the nest wall is occupied by eggs. It is quite possible that, if females can sense where the eggs are located, they can make an assessment of the quantity of eggs present in the nest at the same time. Females might do this either by visual, chemical or tactile sense. However visual sense is not a very probable candidate, since not much light enters the nest through its narrow entrance. Also olfactory sense is not very likely to provide the female with a precise picture of the location of the eggs. Females have a genital papilla which they use when they deposit eggs. The first author of this paper once observed a female depositing eggs in a tube which was placed in a dark hollow container with a hole through which the observer could look inside. The female was seen to move her papilla along the wall of the tube until it encountered an egg patch, and then she deposited an egg. It is appealing to think that females use their genital papilla to provide them with tactile information about the location of eggs, and at the same time acquire knowledge about the size of the area occupied by eggs. Females that visit a nest usually spend some time (several tens of seconds up to several minutes) in the nest before rejecting or starting oviposition (unpublished results, KRAAK). This suggests that females assess nest contents before deciding to leave or spawn. The deposition of eggs along the edges and in the middle of patches of other eggs is likely to be adaptive if it results in spreading the risk that eggs are cannibalized or die by other causes, such as infection by fungi. In order to confirm whether this is the case, the probabilities of eggs being cannibalized or being infected should be investigated in relation to their location respective to other eggs. References GOLDSCHMIDT, T., BAKKER, T.C.M. & FEUTH-DE BRUIJN, E. (1993). Selective copying in mate choice of female sticklebacks. - Anim. Behav. 45, p

18 206 HARA, T.J. (1993). Role of olfaction in fish behaviour. - In: Behaviour of teleost fishes. 2nd edition (T.J. PITCHER, ed.). Chapman & Hall, London, p HARVEY, P.H. & BRADBURY, J.W. (1991). Sexual selection. - In: Behavioural ecology: An evolutionary approach. 3rd edition (J.R. KREBS & N.B. DAVIES, eds). Blackwell Scientific Publications, London, p VAN IERSEL, J.J.A. (1953). Ananalysis of parental behaviour of the male three-spined stickleback (Casterosteus aculeatus L.). - Behaviour, suppl. 3, 159 pp. JAMIESON, I.G. & COLGAN, P.W. (1989). Eggs in the nests of males and their effect on mate choice in the three-spined stickleback. - Anim. Behav. 38, p KNAPP, R.A. & SARGENT, R.C. (1989). Egg mimicry as a mating strategy in the fantail darter, Ethiostoma flabellare: females prefer males with eggs. - Behav. Ecol. Sociobiol. 25, p KRAAK, S.B.M. (1994a). Female preference and filial cannibalism in Aidablennius sphynx (Teleostei, Blenniidae); a combined field and laboratory study. - In: Female mate choice in Aidablennius sphynx, a fish with paternal care for eggs in a nest (S.B.M. KRAAK). PhD thesis, University of Groningen. - - (1994b). A description of the reproductive behaviour of the Mediterranean blenny Aidablenniusphynx (Teleostei, Blenniidae) in its natural habitat. - In: Female mate choice in Aidablennius sphynx, a fish with paternal care for eggs in a nest (S.B.M. KRAAK). PhD thesis, University of Groningen. - - & VIDELER, J.J. (1991). Mate choice in Aidablennius sphynx (Teleostei, Blenniidae); females prefer nests containing more eggs. - Behaviour 119, p & WEISSING, F.J. (1994). Costs and benefits of female preference for nests with many eggs: a simulation study. - In: Female mate choice in Aidablenniusphynx, a fish with paternal care for eggs in a nest (S.B.M. KRAAK). PhD thesis, University of Groningen. LAUMEN, J., PERN, U. & BLUM, V. (1974). Investigations and hormonal regulation of the anal appendices in Blennius pavo (Risso).-J. exp. Zool. 190, p MARCONATO, A. & BISAZZA, A. (1986). Males whose nests contain eggs are preferred by female Cottus gobio L. (Pisces, Cottidae). - Anim. Behav. 34, p MCLENNAN, D.A. & MCPHAIL, J.D. (1990). Experimental investigations of the evolutionary significance of sexually dimorphic nuptial colouration in Gasterosteus aculeatus (L.): the relationship between male colour and female behaviour. - Can. J. Zool. 68, p RIDLEY, M. & RECHTEN, C. (1981). Female sticklebacks prefer to spawn with males whose nests contain eggs. - Behaviour 76, p ROHWER, S. (1978). Parent cannibalism of offspring and egg raiding as a courtship strategy. - Am. Nat. 112, p SEVENSTER, P. (1961). A causal analysis of a displacement activity (fanning in Gasterosteus aculeatus L.). - Behaviour, suppl. 9, 170 pp. UNGER, L.M. & SARGENT, R.C. (1988). Allopaternal care in the fathead minnow, Pimephales promelas: females prefer males with eggs. - Behav. Ecol. Sociobiol. 23, p