Experiment 1: do conspecifics affect oviposition decisions?

Group housing increased the probability that flies oviposited in small artificial hosts. When held alone, 7 of 39 females laid eggs in the artificial host provided to them. Given this percentage of oviposition in isolation (=18%), if there were no effect of social treatment, eggs should have been laid in only 45% of the cups holding 3 females (1 − probability that none of 3 females in a cup lay eggs = (1 − [1 − 0.18]^{3} = 0.45)). In fact, eggs were laid in 81% (29 of 36) of the cups holding 3 females (exact binomial probability < 0.0001|expected = 45%).

Using the percentage of groups in which no female laid eggs (=19%), the probability (

*f*) that a given female laid eggs when in the presence of 2 other females is:

*f* =

*P* (1 grouped female lays eggs) =

. Thus, we estimate that the probability that a female lays any eggs in a small artificial host increases from 0.18 to 0.42 when she is housed with conspecifics.

Although social environment affected the probability of ovipositing, it did not affect the number or size of clutches laid per female. We estimated an average 1.56 females oviposited in grouped treatments where eggs were found (see

APPENDIX). Given this estimate, each ovipositing female in grouped treatments (

*N* = 29 cups) laid an average of 15.69 (±1.80) eggs and an average of 4.1 (±0.50) clutches. Isolated females that laid eggs (

*N* = 7) laid an average of 11.00 (±3.22) eggs and an average of 2.86 (±0.86) clutches. These differences between ovipositing isolated and grouped females were not statistically significant (

*t*_{sqrt(eggs)} = 1.39, degrees of freedom [df] = 34,

*P* = 0.17;

*t*_{sqrt(clutches)}=1.20, df = 34,

*P* = 0.24). Furthermore, the size of individual clutches did not differ between treatments (grouped: 5.85 ± 0.50 eggs per clutch; isolated: 4.28 ± 0.79 eggs per clutch;

*t*_{eggs} = 1.42, df = 34,

*P* = 0.16).

Posttest dissections revealed that females housed in groups and females housed in isolation had similar egg loads at the end of the assay (isolated: average no. of mature oocytes = 24.7 ± 2.4.; grouped: average no. of mature oocytes per female: 24.6 ± 1.3; ANOVA: *F*_{1,69} < 0.01, *P* = 0.89). This indicates that egg load is unlikely to have been a mechanism driving differences in the probability of laying eggs. Because egg load was not affected by treatment, and 95% of all females contained eggs, we did not conduct dissections of females in experiments 2 and 3.

Experiment 2: does the effect of conspecifics depend on host quality?

The effect of social treatment was detected when small agar spheres were offered to females but not when they were offered large agar spheres. Controlling for social treatment, large spheres were more likely to contain eggs (85% contained eggs) than small spheres (52% contained eggs) (Mantel–Haenszel χ

^{2} = 26.2, df = 1,

*P* < 0.0001). As in experiment 1, only a small proportion of females held alone laid eggs in small spheres (11 of 45 = 0.24); the proportion of cups of grouped females in which eggs were laid into small spheres was greater than expected, based on oviposition by isolated females (expected proportion of spheres with eggs = 1 − [1 − 0.24]

^{3} = 0.56; observed: 0.81 [33 of 41]; (exact binomial probability = 0.0009|expected = 56%)). As in experiment 1, females housed in groups with small spheres laid eggs with an estimated probability of

.

When females were held in isolation with large spheres, a majority of females laid eggs (32 of 44 = 0.73). Given this high rate of acceptance, almost all large agar spheres housed with 3 females would be expected to contain eggs (expected prob.= 1 − [1 − 0.73]

^{3} = 0.98) regardless of any effect of conspecifics. This expectation was met (observed: 98% (41 of 42); binomial test

*P* = 0.8). Females housed in groups with large spheres laid eggs with an estimated probability

.

We estimated that in cups of grouped females where eggs were laid, an average of 1.56 and 2.19 females per cup laid eggs in small and large spheres, respectively. The number of eggs laid per ovipositing female (square root transformed) was influenced by a marginally significant interaction between the size of the sphere and the social treatment (*F*_{(size)1,113} = 4.87, *P* = 0.03, *F*_{(social)1,113} = 4.20, *P* = 0.04, *F*_{(social}_{ × size)1,113} = 3.50, *P* = 0.06). Grouped females laid significantly more eggs in small spheres per ovipositing female than isolated females (*t* = 2.07, df = 42, *P* = 0.04), whereas grouped and isolated females laid a similar number of eggs in large agar spheres (*t* = 0.18, df = 71, *P* = 0.86; ). We found no significant effects of sphere size or social treatment on the number of clutches laid per ovipositing female (*F*_{(size)1,113} = 0.88, *P* = 0.35, *F*_{(social)1,113} = 2.91, *P* = 0.09, *F*_{(social × size)1,113} = 2.04, *P* = 0.16; ). Clutch size was not affected by sphere size or social treatment (*F*_{(size)1,113} = 2.17, *P* = 0.14, *F*_{(social)1,113} = 0.02, *P* = 0.89, *F*_{(social × size)1,118} = 0.08, *P* = 0.78; ).

In summary, the effect of conspecifics on oviposition was not independent of sphere size; the effect was only detectable when flies were provided with small spheres. This result may indicate that the presence of conspecifics decreased choosiness, but we are faced with the possibility that the lack of an observed effect in the large sphere treatment was due to a ceiling effect. However, if grouped females were more likely to lay eggs than isolated females in the large sphere treatment (i.e., *f* > 0.73), then our estimate of the number of females ovipositing per cup (2.19) would be an underestimate. We would expect in turn that our estimate for the number of eggs per ovipositing female would be higher in groups than for isolated flies. This was not the case. In fact, our estimates of eggs laid per ovipositing female were nearly identical in isolated and grouped females with large spheres and for grouped females with small spheres (). If the probability of laying eggs in large spheres was any greater when flies were in groups, this would mean that grouped females laid fewer eggs per capita in large spheres than in small spheres. Therefore, it seems unlikely that the lack of an observed effect of social treatment on propensity to oviposit was due to a ceiling effect on the number of eggs each fly could lay.

Experiment 3: is the effect of conspecifics restricted to the resource item on which conspecifics reside?

The results of experiment 3 indicated the facilitating effect of conspecifics was not restricted to artificial hosts near those conspecifics. The first trial of this experiment indicated that the presence of females in one cup influenced the oviposition behavior of focal females in the arena. Only 2 of 10 females in arenas without conspecifics present laid any eggs. In contrast, 6 of 10 females in arenas with conspecifics present laid eggs. The difference is marginally significant (Fisher's Exact test, *P* = 0.08). The presence of females in one of the cups resulted in an increase in the percentage of scans during which the focal female was seen on either sphere (with flies: 2.37 (± 0.74) of scans; without flies: 0.97 (±0.31) scans; Mann–Whitney *U* test: *U* = 22, *N*_{1} = *N*_{2} = 10, *P* < 0.02). Given that only 2 isolated females laid eggs, we cannot statistically compare the clutch number or size, however, there was a trend for females in arenas with conspecifics to lay more and larger clutches (mean no. of clutches_{isolated} = 3.5; mean no. of clutches_{with conspecifics} = 8.17; mean clutch size_{isolated} = 1.8, mean clutch size_{with conspecifics} = 3.4). Among the 10 pairs of cups that contained flies in one of the cups, no strong bias was observed toward or away from the cup containing the female cues. Data from these 10 pairs of cups were analyzed along with data from the 2 blocks of the second trial of this experiment to increase our statistical power to detect any bias toward or away from cup-containing conspecifics.

Fifty females across 3 blocks were tested for a tendency to spend time and/or lay their eggs near conspecifics. Females were observed more often on the sphere set away from conspecifics (*t* = 2.309, df = 49, *P* = 0.025). However, in general, focal females demonstrated no strong bias toward or away from conspecifics (). Females did not spend more or less time on the side of the cage with containing conspecifics (*t* = 0.122, df = 49, *P* > 0.91). Similarly, there was no difference in the number of clutches laid in either sphere (*t* = .379, df = 49, *P* = 0.76) or in the size of clutches laid on either side (*t* = 0.715, df = 12, *P* = 0.49; ).

Experiment 4: does previous experience with conspecifics increase oviposition response?

Rearing flies in groups increased their propensity to lay eggs and decreased their level of aggressive behaviors. The probability that females attempted oviposition was higher for those reared with other females (χ^{2} =5.28, df = 1, *P* < 0.025). Similarly, the presence of a resident female on the host during testing, increased the probability that a female attempted oviposition (χ^{2} = 5.28, df = 1, *P* < 0.025) (). There were marginally significant trends in same direction when analyzing the proportion of females that successfully oviposited (rearing treatment: χ^{2} = 3.52, df = 1, *P* < 0.06; resident presence: χ^{2} = 3.52, df = 1, *P* < 0.06) (). There was no significant interaction between treatment factors on either attempted or successful ovipositions. Although nearly all the successful ovipositions were in the puncture provided (20 of 21 ovipositions when no conspecific was present; 31 of 33 when conspecific was present), oviposition attempts were not more frequent on the side of the fruit containing the puncture (pooled across treatments, proportion of attempts on puncture side = 0.51, *t*_{one-sample} = 0.373, df = 80, *P* = 0.78).

When residents were present, the frequency of attacks by a focal female on a resident, measured in terms of lunges, head butts and foreleg kicks, depended on social history treatment, as well as whether or not females attempted oviposition (). In both rearing treatments, females that attempted oviposition engaged in more attacks than females that did not attempt oviposition (reared alone: Mann–Whitney *U* = 172.0, *N*_{attempt} = 22, *N*_{no attempt} = 31, *P* = 0.001; reared socially: Mann–Whitney *U* = 279.0, *N*_{attempt} = 32, *N*_{no attempt} = 24, *P* = 0.042). Among females that attempted oviposition, those reared alone engaged in a markedly greater number of attacks on the residents than females reared in groups (Mann–Whitney *U* = 240.0, *N*_{social} = 32, *N*_{isolated} = 24, *P* = 0.038) ().

The effect of social history on egg laying and aggression was not due to an effect of rearing conditions on egg maturation. Among individuals used in the analysis (i.e., individuals with egg load >0), females held alone carried 27.62 (±1.57) mature oocytes on average (*N* = 75), whereas females held in groups carried 25.19 (±1.49) mature oocytes on average (*N* = 83). As in experiment 1, the difference in egg load is not statistically significant (*t*_{156} = 1.12, *P* = 0.26). Females from different social history treatments also did not differ significantly either in age or in wing vein length, a proxy for body size (*t*-tests, *P* > 0.26).