In species with biparental care, conflict arises over how much each parent provides to their offspring because both benefit from shifting the burden of care to the other. Here, we tested alternative models for how parents negotiate offspring care using a wild population of clownfish (Amphiprion percula). Using 60 breeding groups, we experimentally handicapped parents by fin-clipping the female in 20 groups, the male in 20 groups, and neither parent in 20 groups and measured changes in female, male, and pair combined effort in response to handicapping. First, we found that handicapping resulted in a decrease in the number of eggs laid by fin-clipped females and a decrease in the amount of parental care by fin-clipped males. Second, contrary to predictions, female effort did not change in response to the male being handicapped, or vice versa. Finally, the number of embryos that matured to hatching, an indicator of pair effort, was not influenced by the manipulation, suggesting that although the handicap was effective, clownfish do not face the predicted “cost to conflict” when one parent is handicapped. Together, these results question the generality of current theoretical predictions and expand our understanding of the diverse possible outcomes of parental conflict.

Study Population

Fieldwork was conducted on nine discrete platform reefs near Mahonia Na Dari Research and Education Center in Kimbe Bay, Papua New Guinea (5.1666667°S 150.5°E) using SCUBA. These reefs support a population of over 100 groups of A. percula living within magnificent sea anemones Heteractis magnifica. The study was conducted over two and a half lunar months and consisted of two weeks of preparation, one lunar month of data collection pre-manipulation (June 13 – July 11, 2018), and one lunar month of data collection post-manipulation (July 12 – August 11, 2018). Lunar months were used rather than calendar months because pairs breed on a lunar schedule in the wild (Buston & Elith 2011; Seymour et al. 2018). From June 1 to June 12, 2018, 69 breeding groups of A. percula were located and marked with numbered tags for identification. Over the course of the study period, 4-5 reefs were surveyed each day, such that all 69 groups were visited every other day.

Parental Size Measurements

To control for any effect of parental size on investment, the standard length of the female (FSL) and male (MSL) were measured. The female and male from each group were captured using hand nets before the start of the experiment (week of June 1, 2018) and following the end of the experiment (week of August 11, 2018). The color pattern of each fish was photographed to identify individuals and confirm that they survived (Nelson et al. 1998; Elliott & Mariscal 2001, Buston 2003). The standard length of each female and male was measured to the nearest 0.1 mm using calipers, and the average of these two standard length measurements provided an estimate of body size for each female and male within a group.

Fin Clipping Manipulation

To determine if pairs are negotiating their parental effort, parents were assigned one of three treatments: female handicapped (FH), male handicapped (MH), or control (C). To control for variation among groups that could potentially influence parental responses to manipulation, assignment to treatment was randomly stratified by distance from shore and reef ID (n = 24 pairs on 3 inshore reefs; n = 45 pairs on 6 offshore reefs) and the number of clutches laid in the first experimental month (n = 16 pairs with 1 clutch; n = 42 pairs with 2 clutches; n = 11 pairs with 3 clutches).

In the first week of the manipulation period (week of July 11), just after the new moon when groups are not breeding (Seymour et al. 2018), females and males in the FH and MH treatments, respectively, were captured using hand nets and slowly brought to the surface in plastic bags filled with seawater. Control fish were exposed to a sham capture to control for any effects of the catching process on the non-handicapped fish. Each of the experimental fish had a small portion of both their pectoral fins and caudal fin removed using precision scissors (Dr. Slick Co., Belgrade, MT). Each fin was measured before and after the portion was removed. On average, females had 5.2mm of each fin removed, and males had 4.5mm of each fin removed, such that both females and males had approximately 40% of each fin removed as a handicap. All individuals were returned to their home anemones and monitored until the handicapped individual began swimming within its anemone. All individuals appeared to resume normal swimming behavior within a short (<1 min) period.

Reproduction

To determine the effect of handicapping on reproduction, all groups were monitored throughout the study period for reproduction. Reproduction was detectable by the behavior of the male and the presence of eggs at the base of the anemone. The exact age of the eggs was determined by their color (day 1 = orange, day 2 = orange-brown, day 3 = brown, day 4 = black eyes, day 5 = silver eyes, day 6 = silver eyes with pupils, day 7 = gold eyes with pupils). Eggs hatched after 7 days. Photos were taken of every clutch observed with an Olympus Tough TG-870 camera (Olympus, Tokyo, Japan) early (day 1 or 2) and late (day 6 or 7) in development. The number of eggs were counted in each photograph using ImageJ (NIH, USA).

Parental Care

To determine the effect of handicapping on parental care, 29 minute videos were taken of parental behavior early (day 3 or 4) and late (day 6 or 7) in development by setting up a tripod with an underwater camera (Olympus Tough TG-870) centered on the location of the eggs. Videos were analyzed using JWatcher version 0.9 (Blumstein & Daniel 2007). The first 12 minutes of video served as an acclimation period, the following 15 minutes used for data collection, and the final 2 minutes were discarded from analysis to minimize any effect of approaching researchers at the end of the recordings. Parental care was recorded as time tending, i.e. the amount of time an individual spent within one body-length of the clutch, as tending behavior shows plasticity in both parents (Barbasch & Buston 2018). Behavior was scored separately for the female and male of each pair and the time spent tending served as measures of female and male parental care, respectively.

Parental Effort

The reproduction and egg care data were used to quantify parental effort. The number of eggs laid per clutch, the number of eggs laid per month, and the female time spent tending were used as indicators of female parental effort. The number of eggs laid was considered a measure of female effort as egg production is thought to be costly for females (Williams 1966) and because egg production is related to female size (Barbasch et al. in review). Male time spent tending and embryo survival (the proportion of eggs laid per clutch that survived to day 6 or 7) were used as indicators of male parental effort. Embryo survival was used as an indicator of male effort because the proportion of embryos surviving in a clutch has been shown to be associated with male parental care (Barbasch et al. in review). Finally, the number of embryos maturing to day 6 or 7 per clutch and per month were used as indicators of the pair’s combined effort, because they have been shown to be associated with the number of eggs laid, female and male time tending, and embryo survival (Buston & Elith 2011; Barbasch et al. 2020). These metrics were used to determine first, the effect of the manipulation on the parental effort of the handicapped individual, second, the effect of the manipulation on the parental effort of the handicapped individual’s partner, and third, the effect of the manipulation and of the resulting changes in parental effort on pair effort.

Punishment

To assess how the manipulation influenced punishment (Table 1; H3), the videos of egg care also provided observations of aggression. In each video, the number of aggressive behaviors (bites, chases) and the identity of the aggressor were recorded (Wong et al. 2013). The number of aggressive behaviors by each individual per month was used to quantify degree of punishment, and compared across treatment groups.

Statistical Analysis

Models for Female Effort

First, to determine the effect of the manipulation on female parental effort, three models were fit with the following response variables representing female parental effort: 1) number of eggs laid per clutch, 2) number of eggs laid per month, and 3) female time tending. The number of eggs laid per clutch and per month were fit with a Poisson error distribution, controlling for overdispersion using an observation level random effect. Time tending was fit with Gaussian error distribution. Individual ID was fit as a random effect in all models, to control for repeated measures with the same individual. Treatment (FH, MH, Control), Month (Pre-manipulation, Post-manipulation), and their interaction were fit as predictors/independent variables. Because response variables were measured before and after the manipulation, the interaction captures how each metric of female effort changes in response to the handicapping of the female or male relative to the control.

For the models of number of eggs laid, we included FSL to control for effects of female size on egg laying (Buston & Elith 2011; Barbasch et al. 2020). FSL was included because it cannot be considered a response to manipulation. For the model of female time tending we included clutch age (early/young or late/old) to control for effect of clutch age on parental care (Barbasch & Buston 2018). The number of eggs laid was also included as a covariate to determine whether there is an effect of the manipulation on female tending, above and beyond that accounted for by number of eggs laid. [The model of female time tending was assessed without number of eggs laid to determine if there was an overall effect of manipulation on female care, without controlling for the effects of number of eggs laid (see Appendix)].

Models for Male Effort

Second, to determine the effect of the manipulation on male parental effort, models were fit with two response variables representing male parental effort: 1) male time tending, and 2) embryo survival (proportion of embryos surviving per clutch). Time tending was fit with a Gaussian error distribution and embryo survival, a proportion, was fit with a binomial error distribution. Individual ID was included as a random effect. Treatment (FH, MH, Control), Month (Pre-manipulation, Post-manipulation), and their interaction were fit as predictors/independent variables.

For the model of male time tending, clutch age (early/young or late/old) was included as a covariate. Additionally, the model of male time tending was assessed with the number of eggs laid, as male time tending is strongly associated with the number of eggs laid (Barbasch et al. 2020). When the number of eggs is included, the model asks whether there is an effect of the manipulation on male tending above and beyond that accounted for by number of eggs laid. Similarly, the model for embryo survival was run with male tending as a covariate, as male tending has been shown to be associated with embryo survival (Barbasch et al. in review). When this covariate is included, the model assesses whether there is an effect of the manipulation on embryo survival above and beyond the effects of male time tending to the eggs. [The models for male tending and embryo survival were also run without number of eggs laid and male tending as covariates, respectively].

Models for Pair Effort

Third, to determine the effect of the manipulation on indicators of pair effort, models were fit with two response variables: 1) number of embryos maturing per clutch, and 2) number of embryos maturing per month. The number of embryos maturing per clutch and per month were fit with Poisson error distributions. The models were fit with an observation level random effect to control for overdispersion. Individual ID was also included as a random effect. Treatment (FH, MH, Control), Month (Pre-manipulation, Post-manipulation), and their interaction were fit as predictors/independent variables. Both models were assessed with covariates number of eggs laid and male time tending to determine whether there is an effect of the manipulation on pair effort, above and beyond the effects of female egg laying and male time tending for the eggs. [The models were also fit without these covariates to determine the overall effect of the manipulation on pair effort].