Parent-offspring conflict over food allocation can be modeled using two theoretical frameworks: passive (scramble competition) and active choice (signaling) resolution models. However, differentiating between these models empirically can be challenging. One possibility involves investigating details of decision-making by feeding parents. Different nestling traits, related to competitive prowess or signalling cryptic condition, may interact additively or non-additively as predictors of parental feeding responses. To explore this, we experimentally created even-sized, small broods of pied flycatchers and manipulated nestling cryptic quality, independently of size, by vitamin E supplementation. We explored how interactions between nestling cryptic condition, size, signals, and spatial location predicted food allocation and prey-testing by parents. Parents created the potential for spatial scramble competition between nestlings by feeding from and to a narrow range of nest locations. Heavier supplemented nestlings grew faster and were more likely to access profitable nest locations. However, the most profitable locations were not more contested, and nestling turnover did not vary in relation to spatial predictability or food supply. Postural begging was only predicted by nestling hunger and body mass, but parents did not favor heavier nestlings. This suggests that size-mediated and spatial competition in experimental broods was mild. Pied flycatcher fathers allocated food in response to nestling position and begging order, while mothers seemingly followed an active choice mechanism involving assessment of more complex traits, including postural intensity interacting with order, position, and treatment, and perhaps other stimuli when performing prey-testings. Differences in time constraints may underlie sex differences in food allocation rules.

Experimental setup

We monitored the nest boxes regularly to determine the exact start laying (±24 h) and hatching dates (±12 h; day 1 was considered the hatching day). Initially, we included 32 broods (174 nestlings) of monogamous pairs (Lundberg and Alatalo 1992) containing four to six nestlings that were five days old. Within each brood, we ranked the nestlings by mass and assigned them to either a vitamin E supplementation or control treatment, alternating the order between successive broods to ensure an unbiased sample with respect to hatching order and balanced across different brood sizes. As in a previous study conducted in the same population (Pérez-Rodríguez et al. 2019), supplemented nestlings were fed a dipteran larva (Calliphora sp.) soaked in a solution of vitamin E (DL-alpha-tocopherol acetate; Sigma-Aldrich ref. T3376) dissolved in organic coconut oil. The concentration of this vitamin E solution was adjusted across nestling development to provide a constant dose of 1.2 mg of vitamin E per kg per visit to the supplemented nestlings. Control nestlings were fed a larva soaked in coconut oil. Regardless of the treatment, nestlings received a new supplementation every other day, and we recorded their mass at each visit using an electronic balance (±0.01 g). There was no significant difference in initial (day 5) body mass between supplemented and control nestlings (β ± SE = -0.232 ± 0.155, t _{1, 133} = -1.500, P = 0.136).

Video recordings

We recorded video sequences of parent and chick behavior when the broods reached 7 days old. This age was selected since it corresponds to the peak of daily mass gain (Siikamäki 1996), and nestlings' feeding success still relies mostly on their begging behavior, before they can intercept food from adults at the nest entrance (Khayutin et al. 1988). On day 6, we attached a dummy camera to the roof of the nest-box to habituate the birds. In the morning of day 7, we fed the four nestlings with the most similar body masses (two from each treatment) to satiation with 1-5 larvae to equalize food deficit (Wright et al. 2002). We marked each nestling's head with a unique combination of blue adhesive kinesiology tape badges varying in shape (circle vs. triangle) and markings (closed vs. open dots) (Porkert and Špinka 2006). The badges were randomly assigned with respect to nestling treatment and relative mass. We used a Sony Go-Pro video camera to obtain video recordings for as long as the camera batteries allowed (about 2 h). We obtained video samples from 28 nests. Meanwhile, we kept the surplus chicks from each brood in heated containers, hand-fed them every 30 min, and returned them to their natal nest after recording. 

We played recorded video sequences at 1/4 speed to accurately measure the behavior of parents and chicks during each parental feeding visit. A trained observer, blind to the aim of the experiment, computed behavioral rates. During each parental visit, we distinguished between parental feeding and prey-testing behavior. We defined a feeding as a parent introducing a food item into the mouth of a nestling and the nestling swallowing it. Prey-testing involved a parent introducing its beak with prey into a nestling mouth but failing to release any food item (Slagsvold and Wiebe 2007). Both parents feed nestlings with prey of a similar size at this age, and nestlings can efficiently handle most prey (Slagsvold and Wiebe 2007; Wiebe and Slagsvold 2009a). However, we discarded testing events (four visits) where prey items were of an unusually large size, which could be explained by gape size constraints rather than parental choice (Wiebe and Slagsvold 2009a). We also recorded nestling behavior in a time window ranging from 10 seconds before the adult entered the nest box to 10 seconds after the adult left it, to record nestling behavior while adults remained poking at the nest box from the outside during each parental visit.

We divided the nest cup into 45° circular sectors, numbered clockwise from 1 to 8, to record parent and nestling behavior during feedings (Khayutin et al. 1988). We noted the sector from which each parent fed nestlings and the sector occupied by each nestling when parents entered the nest box. During each feeding visit, we recorded its duration, the spatial location of parents, the number of feedings and prey-testings received, the order in which nestlings begged (1 = first) and the maximum begging postural intensity of each nestling. We used a five-level ordinal scale to score the degree of body stretching as an indicator of begging postural intensity, following Redondo and Castro (1992). Additionally, we computed the rate of individual nestling movement as the average number of sector changes between parental visits. To calculate time of food deprivation for each nestling during a given visit, we measured the time elapsed since the previous visit when it was fed. Since initial deprivation times varied between nests, we computed deprivation times for each nestling as a fraction of the maximum value of deprivation time for each brood in the entire recording session.

Nestling mouth coloration

On day 7, we photographed the mouth coloration of 72 vitamin E and 81 control nestlings from 32 broods. We gently opened each nestling's mouth in a standardized position and captured images using a Nikon D80 digital camera with a Sigma AF 50/2.8 objective and consistent illumination. To ensure accurate color representation, we included a ColorChecker Passport Photo 2 (X-Rite, Regensdorf, Switzerland) color reference in a standard position in each image. Images in RAW format were processed using SpotEgg software (Gómez and Liñán-Cembrano 2017) to normalize and linearize them. We then used Adobe Photoshop CS6 v13.0 (Adobe Systems Incorporated, San Jose, CA) to quantify the color descriptors of the areas of interest. Color descriptors from digital pictures are a reliable and efficient method to assess nestling mouth coloration (Saino et al. 2003; Dugas and McGraw 2011; Dugas and Dillow 2013), allowing us to obtain average color estimates of whole areas with minimal manipulation time, thus reducing interference on study subjects. However, this method does not capture ultraviolet reflectance. In pied flycatcher nestlings, the orange palates are bordered by white-yellowish fleshy flanges. To quantify nestling mouth coloration, we recorded the saturation value of the nestling's palate as an indicator of carotenoid pigmentation (Saino et al. 2003; Dugas and McGraw 2011) and the lightness (total reflectance) of the inner and outer parts of the mouth flanges as an indicator of conspicuousness (Kilner and Davies 1998; Wiebe and Slagsvold 2009b). To assess the repeatability of color variables, we measured a subset of 24 individuals twice. The results showed high repeatability for palate saturation (r = 0.69, F _{22, 25} = 5.41, P < 0.001), inner flange lightness (r = 0.79, F _{22, 25} = 8.44, P < 0.001), and outer flange lightness (r = 0.89, F _{22, 25} = 16.97, P < 0.001).

Statistical analyses

We used linear mixed effects models with restricted maximum likelihood to perform statistical analyses (Zuur et al. 2009). We employed the “nlme” package (Pinheiro et al. 2020) in R 4.0.2 (R Development Core Team 2020) to fit the models (See Table S1 for a description of all saturated models). To improve the stability of the models, likelihood of model convergence, and accuracy of parameter estimates (Harrison et al. 2018), we Z-transformed (mean centered with SD of 1) all numerical independent variables. Standardized beta coefficients were used to estimate standardized effect size for comparisons between fixed effects (Schielzeth 2010). In all models, we included original brood size at day 7 and hatching date (1 = June 8th) as covariates among fixed effects because the number of nestmates and differences in resource availability and parental condition as the season progresses could affect adult and nestling behavior. Validation of model assumptions were performed by inspecting residual plots generated by the “performance” package (Lüdecke et al. 2021) for linear mixed models and the “DHARMa” package for generalized mixed models (Hartig 2022). Predicted model values were plotted using the package “ggeffects” (Lüdecke 2018).

To confirm that the treatment had the intended effect on nestling mass gain, we compared differences in nestling mass between the vitamin E and control groups on days 7 and 9, while controlling for initial mass at day 5 as a covariate. To analyze the effect of treatment on nestling flange conspicuousness, we conducted a principal component analysis (PCA) using the lightness values of the inner and outer areas of the flange with the prcomp() function from the “stats” package. We used individual scores on the first principal component (PC1), which explained 63.2% of the total variance, as a measure of overall lightness, as they were highly correlated with lightness values of both the inner and outer parts of the flange (PC1 loading value = 0.707).

To investigate the effect of vitamin E supplementation on nestling behavior, we built separate models for various dependent variables, including the probability of nestling begging (gaping or not), begging order, postural intensity, profitability of a nestling spatial location, inter-feeding intervals, and the rate of change between nest sectors. We included the sex of the feeding adult entering the nest box as a fixed effect (Wetzel et al. 2020), since nestlings may behave differently according to the parent's sex. To control for intra-brood differences in nestling body mass that could reflect differences in relative size or age, we used the mass difference of each nestling relative to the average of its brood rather than its absolute body mass. The probability of nestling begging was analyzed using a generalized linear mixed model with a binomial error distribution. Although begging order and postural intensity were measured as ordinal variables, they were treated as numeric in linear mixed models. We verified that conclusions did not differ from those from an ordinal regression (Christensen 2022).

Spatial profitability was computed as the fraction of total feedings given by each parent to the sector occupied by each nestling, relative to the parent's location in a given visit. Profitability values were transformed as logits prior to analysis. We also calculated the rate of spatial change between nest sectors for each nestling by dividing the number of changes between sectors by the number of feeding visits. To test whether nestling mobility was related to hunger levels, we computed the mean time of food deprivation for each chick and the average rate of change between sectors per brood by dividing the mean number of position changes by nestlings for the entire brood by the number of feeding visits (McRae et al. 1993). We included the number of caring parents and the degree of overlap between parent sexes in the use of nest sectors as fixed effects in this model to investigate whether the predictability of parental feeding position influenced nestling movement (Kölliker and Richner 2004). We calculated the degree of overlap between parent sexes in the use of nest sectors as the difference between the number of shared nest sectors multiplied by the number of feedings given from these sectors minus the number of feedings given from non-overlapping sectors, divided by the total number of feeding events. This resulted in an index ranging from 0 (no overlap) to 1 (maximum overlap). We assumed that uniparental broods had a maximum (1) overlap.

We analyzed parental feeding preferences using generalized linear mixed models with a binomial error distribution and a binary response variable (1 = fed, 0 = not fed). Since allocation patterns may vary by parental sex (Kilner 2002, Ryser et al. 2016), we conducted separate analyses for females and males. To avoid convergence issues caused by a high number of predictors, we proceeded in two steps. First, we built models with all nestling traits, including relative mass, begging behavior (order and postural intensity), coloration (flange lightness and palate saturation), and spatial profitability, interacting with treatment and hatching date. Second, we built a second group of models considering only significant predictors from the first step (begging behavior, position, and treatment), including all pairwise interactions among them as well as with relative body mass (Table S1). To equalize the predictive potential and precision of spatial profitability with the other predictors, we transformed its raw values into 4-level ranks using the rank() function from the “base” package. We analyzed sex differences in the feeding rate of parents and the duration of each feeding visit to investigate whether differences between females and males in their degree of familiarity with nestlings could explain sex differences in allocation rules (Gottlander 1987; Lucass et al. 2016) (see Table S1).

We explored the prey-testing behavior of parents in three steps. First, we looked for differences between visits where testing occurred and those where it did not. We compared the total number of begging nestlings and the intra-brood coefficient of variation in time of food deprivation, begging order, postural intensity and spatial profitability between visits where testing occurred and those where it did not, including the occurrence of testing (yes or no) as a categorical predictor (Table S1). Second, in a given visit, we compared the behavior of nestlings that were either tested, fed or neither fed nor tested. We built separate models for nestling behavior, relative body mass, and time of food deprivation as response variables. We also included treatment and a new three-level categorical predictor (“Testing”) depending on whether a nestling had been tested, fed, or neither fed nor tested in a given visit (Table S1). Third, we investigated whether the probability of a nestling being fed after having been tested varied according to its relative body mass, treatment, and time of food deprivation.

We determined the optimal random- (intercepts and slopes) and fixed-effect structure for each model by comparing nested models with likelihood ratio tests, using a top-down strategy (Zuur et al. 2009) (Table S1). To find out the optimal random structure, we started with the most complex fixed and random structure (Barr et al. 2013) (or no fixed effects in the case of generalized linear mixed models), and progressively removed random effects one by one, comparing pairs of models with the anova() function from the “stats” package. Next, to determine the optimal fixed structure, we began with the most complex fixed-effect structure and the optimal random structure selected in the previous step. We then progressively removed fixed effects with lower β coefficients one by one, beginning with the interactions, and compared both models by likelihood ratio tests using the anova() function. 

R 4.0.2 (R Development Core Team 2020) software was used.