Early-life diet does not affect preference for fish in herring gulls (Larus argentatus)

Ethics

This study was approved by the University of Exeter, College of Life and Environmental Sciences, Penryn Ethics Committee (eCORN002962 9.1). GPS tagging (as part of a concurrent study and not included here) and marking of chicks with leg rings and non-toxic temporary dye were approved by the British Trust for Ornithology Special Methods Technical Panel on behalf of the Joint Nature Conservation Committee & Natural England (application licence numbers: 11962 and 11963). The total time that each chick was handled during ringing did not exceed 20 min, except for seven chicks that were GPS-tagged (GPS back-mounted thoracic weak-link harness), where handling time did not exceed 50 min. Avian specialist vets monitored the health and welfare of the gull chicks throughout the study and ascertained that chicks were fit for release. Chicks were kept in captivity until they demonstrated strong flying capabilities in the pens and were confirmed fit for release by the vet. Chicks that were included in this study were kept in captivity for the purpose of rescue and rearing to independence, and their inclusion in this experiment did not hinder or delay their release from captivity.

Chick husbandry during captivity and release

The herring gull chick rehabilitation facilities were at a private residence near Troon, Cornwall, which contained unimproved grassland areas on the property and was surrounded by improved grassland fields within 1 km.

All herring gull chicks brought into captivity for rearing and rehabilitation were first assessed by avian specialist veterinarians to check their condition. The chicks were apparent ‘orphans’ who could either not be reunited with their parents, placed back into their nest, or their exact origin was unidentifiable, thus they were unlikely to survive without intervention. All chicks were found and brought into the rehabilitation facilities within 24 h of rescue from towns across Cornwall, the majority from residential roofs. Over seventy percent of the county of Cornwall is agricultural land, while Cornwall also has 697 km of coastline. It therefore seems likely that most of the chicks would have been provisioned with a mixture of marine and terrestrial foods while still at the nest. However, other than their rescue location, we do not have any information regarding their early-life or dietary experiences in the wild before they entered the rehabilitation facilities.

Chick ages were estimated by ELI (an experienced ornithologist) on admission to captivity, according to their weight, feather development and other physiological features (e.g., presence or absence of an egg tooth). The mean chick age ± SD was estimated to be 10 days old ± 8 days (n = 27). All chicks included in our study were assessed by a vet to be healthy, with no major injuries or other health concerns. Any chicks brought in with treatable injuries or illnesses were not included in the study cohort. Some non-study chicks were present in the enclosures with the study chicks. Any chicks showing illness or distress during rehabilitation were immediately taken to a vet and received medical treatment. Two study chicks developed a respiratory disease whilst in care and were temporally separated and removed from the study until they had fully recovered and were confirmed healthy by a vet, and were returned to the study cohort. Both were from the Marine diet treatment group; one chick missed a 24 h test, and the other the day five test (see below for details of treatments and tests). As their experimental diets and apparent appetites were unchanged during this period, we decided not to remove them from the study sample. Chicks were weighed daily until ca. 30 days of age, then at least once weekly to check on their health and development. Tarsus length was also measured at least weekly to quantify skeletal growth. Handling times were ca. 5 min/day for small chicks and ca. 2 min/week for chicks >30 days old.

Chicks <5 days old were initially kept in an incubator together and fed with tweezers approximately every 2 h during the day, until they were reliably feeding independently. Chicks 5–25 days old were housed indoors overnight in groups of three in cages (0.6 m length × 0.5 m width × 0.5 m height) that were raised off the floor for safety reasons. From ca. 10 days old, they were housed in indoor ground pens (0.9 m × 0.9 m × 0.5 m height) during daylight hours (ca. 10 h a day), with up to 9 chicks in one pen with ad lib food, water and environmental enrichment (e.g., seaweed, cardboard boxes, water trays) until ca. 25 days old. At night, chicks were still housed in their groups of three in the raised overnight cages until ca. 25 days of age. Each chick was marked with a spot of non-toxic sheep marker-spray on downy feathers (approved by the British Trust for Ornithology (BTO) Special Methods Technical Panel (SMTP)) to allow for individual identification whilst <15 days old. Once the chicks were >15 days old, they were fitted with temporary poultry colour rings on their tarsus for ID purposes.

From ca. 25 days of age, when chicks’ body feathers had grown sufficiently to make them waterproof, chicks were moved into partially sheltered outdoor aviaries (ca. 2 m × 4.6 m × 2.5 m) for a week to acclimatise and then into outdoor flight pens (ca. 7 m × 3.5 m × 2.5 m) with other chicks assigned to the same dietary treatment (up to 15 chicks per pen). Outdoor enclosures were on grass with netted roofs and included a paddling pool, perches, foraging enrichment (i.e., scattered dried mealworms (Tenebrio molitor larvae) and fresh seaweed), shelters, and ad lib food and water. Once the chicks were assessed by a vet and deemed fit for release at ca. 50 days old, they were ringed with a permanent BTO metal ring on one leg and an alpha-numeric coded blue plastic colour ring on the other. Prior to release, each chick was weighed, measured for minimal tarsus length (Caravaggi et al., 2021), head length (measured from the back of the skull to the tip of the upper mandible), bill length (from the tip of upper mandible to where the base of the upper mandible meets feathers) and bill depth (measured at the gonys angle on the bill) and scored for aggression (number of bites/pecks received by handler during ringing).

The first, older group of 15 chicks were soft-released on 26^th July at the average age (mean ± SD) of 69 ± 12 days old, by removing the netted tops of their enclosures to allow the gulls to leave by flight during the day. Video recordings showed that all the juveniles flew out of the pens within several hours of the netting being removed, and four individuals left within the first half an hour. However, the chicks remained at the rearing site and spent the night as a group in the surrounding field. Three of these chicks were predated by red foxes (Vulpes vulpes). A further two of the released chicks were found around the pens trying to enter them again so they were returned to captivity to be re-released at a later date, while a third was also seen but chose to fly away. The remaining nine gull chicks were not found onsite and are presumed to have been released successfully. The second group of 14 chicks (including the two returned to captivity) were released later at a local reservoir and nature reserve (Stithians Reservoir, Cornwall SW 7325 3691) on August 23^rd and 24^th, when their mean age (±SD) was 71 ± 12 days. One study chick was retained as it tended to approach humans and was released on 19^th September once it showed more appropriate wariness towards people.

Experimental dietary treatments

Chicks (n = 27) were assigned to one of two experimental dietary treatment groups; thirteen to the ‘Terrestrial’ group and fourteen to the ‘Marine’ group, where the dietary treatments resembled opposite extremes of gull parental provisioning strategies observed in the wild (van Donk et al., 2017; Sotillo et al., 2019; Serré et al., 2022). Chicks were provided with foods categorised as terrestrial or marine in an 80:20 ratio. All chicks were exposed to all foods offered, to avoid food neophobia (Rabinowitch, 1968). Each treatment diet also included high protein and low protein food, with the low protein food accounting for 20% of the wet mass of food provided (see Supplemental Materials Table S1 for the nutritional value of foods). The diet ratios were achieved by presenting the different food types (terrestrial and marine) for set amounts of time throughout daylight hours (ca. 12 h a day), with the order in which food types were presented changing daily, to achieve 80:20 ratios of the food types offered. All chicks were fed ad lib; the amount of food provided was proportional to the number of chicks present and their ages, food was always available during the day, and food was added to or exchanged four times a day for small chicks up to 15 days old, and three times a day for older chicks. The number of visits to older chicks was reduced to limit their exposure to people to minimize tameness, which could impact their survival upon release. All non-consumed foods were removed when a food type switch occurred. When all chicks reached ca. 25 days of age and were moved outside, they were provided with terrestrial and marine foods in a 50:50 ratio simultaneously. The latter change in the ratios of food offered was not initially planned, but we observed that chicks were reluctant to eat when not offered marine foods. The chicks prioritised consuming the marine foods when offered, which could have caused some less- competitive chicks to have a reduced food intake, because they refused to eat terrestrial foods.

The high-protein terrestrial food provided was tinned cat food (chicken, turkey, lamb and beef), all flavours in jelly and gravy (Tesco and Asda own-brand cat foods). Chicks <10 days old were provided with cat food chunks of ca. 0.5 cm × 1 cm × 1 cm, while older, larger chicks were given 2–4 cm³ chunks. The lower-protein terrestrial food was diced brown bread, sized ca. 1 cm³ when chicks were small, and 1 cm × 2 cm × 2 cm for larger chicks. We chose bread as the low-protein terrestrial food type as it is commonly consumed by gulls in urban areas from domestic garbage and from people feeding ducks at parks and lakes (Spaans, 1971; Pierotti & Annett, 1991; Scott, Duncan & Green, 2015; van Donk et al., 2017). We included bread in the terrestrial dietary treatment group to test whether early-life provisioning with bread would generate a preference for this food. In the marine treatment diet, the high-protein food was whole fish (European sprats Sprattus sprattus and Atlantic mackerel Scomber scombrus), cut into 1–2 cm³ pieces when chicks were small, while larger chicks were presented with whole sprats that were ca. 7 cm × 2 cm, and pieces of mackerel cut to similar sizes. The lower-protein marine food was pre-cooked mussels (Mytilus edulis, Tesco frozen cooked shell-less mussels, ca. 1 cm × 1.5 cm × 2 cm). To compensate for any potential nutritional differences between diets, we added thiamine (vitamin B1) and vitamin E supplements (Holland and Barrett 100 IU vitamin E capsules, oil removed from capsules, and Holland and Barrett 100 mg thiamine tables, quartered and crushed) at appropriate dosages (vitamin E: 100 IU kg⁻¹, thiamine: 25 mg kg⁻¹) per kg of fish, and a multi-vitamin and calcium powder supplement (Nutrobal^® Vetark) to bread, to ensure normal chick development. These dosages were prescribed by the veterinarian. It was not possible to overdose an individual chick as no one chick could consume all the fish presented in a feed, nor be able to monopolise the food, as it was presented in multiple bowls spaced through their enclosures. The supplements were also only added to the first feed of the day to avoid the potential for overdosing on vitamin E. However, it should be noted that other seabird species can tolerate large dosages exceeding the daily recommended allowance with no apparent side effects (Crissey, McGill & Simeone, 1998; Crissey, Slifka & McGill, 2002). Thiamine is water soluble and cannot cause overdoses as excess is excreted by the chicks naturally. The Nutrobal^® powder was added to the bread in sufficiently small doses that overdosing was not possible.

Experimental protocols

Immediately upon arrival to captivity, each chick was alternately and randomly assigned to either the Marine (n = 14) or the Terrestrial (n = 13) diet treatment group. For the experimental tests of chicks’ behaviour, chicks were individually placed and tested in ground pens adjacent to their overnight cages (0.9 m length × 0.9 m width × 0.4 m height) in which they were kept during the day, but in the absence of any group mates, objects or shelters. Chicks <15 days old were tested in either the ground pens or a clean overnight cage (0.6 m × 0.5 m × 0.5 m) if they were not yet familiar with the ground pen when < 5 days old. Once outside, chicks >25 days old were tested in an empty flight pen of the same dimensions as, and adjacent to, the flight pens they occupied (ca. 7 m × 3.5 m × 2.5 m). All tests were filmed using a Panasonic HC-V770 video recorder placed outside the enclosures. Chicks were tested at 24–72 h, 5, 10, 15 and 35 days in captivity. All experimental tests were conducted between 7–9:30 am, prior to weighing and ad lib feeding. The four food types (terrestrial high-and low-protein food, and marine high- and low-protein food) were offered in experimental food arrays comprising of four grey plastic cat food bowls (CatCentre^® 0.35 l bowls). Each bowl contained only 10 g of food to avoid chick satiation. The food bowls were the same as those used for the daily feeding of the chicks and were thus familiar to the test subjects. Some individuals missed some experimental tests for health reasons, but all chicks were tested at least four times.

A baseline experimental test for food preference, which was conducted 24–72 h after arrival into captivity, commenced once the chick did not show any apparent signs of distress, and tested for their baseline preference for marine or terrestrial high-protein foods (i.e., fish vs. cat food). After an initial 10 min to habituate to their test enclosure, two food bowls, each containing 10 g of one or the other food type, were presented equidistantly, at approximately 20 cm in front of the chick, for a maximum of 10 min. We measured the latency of the chick to approach and consume (i.e., food visibly eaten or pecked at three times) their first food choice and what that food category was. We also scored whether the chick approached the food/experimenter as the food bowls were being placed inside the cage or pen (yes or no) as a measure of tameness. Both foods were weighed before and after the test to calculate the amount consumed. The trial was stopped once all food had been consumed, or when 10 min had passed. The presentation side of the two foods (left or right hand side) was randomised. Chicks <3 days old that struggled to feed independently were not tested in this baseline trial.

A protocol similar to the baseline preference test was used on days 5, 10, 15 and 35 to test for potential changes in chicks’ food preferences due to the dietary treatments. Chicks were given 10 min to habituate to the test area, after which they were presented with a 2 × 2 grid of food bowls with 10 g of one of the four foods (chopped fish, mussels, cat food and diced bread) in each, and left for 10 min (see Supplemental Material for video of a food preference test trial). The position of each food in the array was randomised in every trial to control for side biases and spatial learning. As before, the food offered was weighed before and after the trial to calculate the amount consumed. For each trial, we recorded the food that the chick consumed first (i.e., food visibly eaten or pecked at three times), the latency to consume the food (time between presentation of the food and it being consumed/pecked at for third time), the second food consumed, and if the chick approached the food as the food bowls were placed into the enclosure (yes or no). All 27 chicks were presented with up to four replicate trials of the food preference tests throughout their time in captivity, but not all chicks chose to participate and consume food during the trials (day 5: n = 25; day 10: n = 26; day 15: n = 24; day 35 (when outdoors and tested in adjacent outdoor test pen): n = 13).

All behaviour was scored from video recordings. Videos were scored by ELI and an independent observer blind to the hypothesis being tested to assess the inter-observer reliability of the behavioural scores. To test whether two independent observers agreed on the chicks’ food preferences, we calculated Intra Class Coefficients (ICCs). Both the latency for chicks to approach and consume food, and the first food eaten had very high repeatability between scorers (Latency: ICC = 0.917 (95% Cl [0.884–0.941]), P < 0.001; First food: ICC = 0.921 (95% Cl [0.887–0.945]), P < 0.001).

Statistical analyses

All analyses were conducted using R version 4.0.3 (R Core Team, 2020) and the packages lme4 (Bates et al., 2015) and MASS (Venables & Ripley, 2002). The error family and link function combination for each generalised linear mixed model (GLMM) conducted was tested to select for the lowest AIC value and smallest residuals.

To determine whether food location within the 2 × 2 arrays affected choice, we performed chi-squared contingency tests. We found that chicks were most likely to choose a bowl from the front of the array (i.e., nearest to them) first ( ${\chi }^2$ = 49.50, df = 1, n = 88 trials, P < 0.001). However, chicks did not appear to have side biases (left vs. right side comparison; ${\chi }^2$ = 0.44, df = 1, n = 111 trials, P = 0.51). Neither bowl position nor side influenced chicks’ second food choices (bowl at front or back of array: ${\chi }^2$ = 2.65, df = 1, n = 74 trials, P = 0.10; left vs. right side: ${\chi }^2$ = 0.22, df = 1, n = 74, P = 0.64). As food type position was randomised for each trial, bowl position and side were therefore not included in further analyses.

Individual chick weights and tarsus measurements taken over the duration of captivity were compared between the two diet treatment groups using Wilcoxon’s tests. Chick weights were compared when chicks were ca. 5 days old (n = 4 chicks per treatment group), 15 days (n = 8 from the marine group, 7 from the terrestrial) and when last weighed recorded prior to release (n = all 27 study chicks). For the final weighing, chick age ranged from 52–79 days due to differences in chick age when received into captivity. Chick tarsus measurements were compared when chicks were ca. 14–16 days old (n = 7 chicks from each treatment group). Tarsus measurements were not taken at younger ages due to time constraints. Chicks’ last tarsus measurement prior to release was taken when chick age ranged from 31–58 days old, when all chicks (n = 27) should have reached their growth asymptote (Supplemental Materials, Figs. S1 and S2 for chick weights and tarsus measurements, respectively).

To test whether chicks’ first, non-baseline, food choice for marine (fish or mussels) or terrestrial (cat food or bread) food types was influenced by their diet treatment group (Marine vs. Terrestrial), we used a GLMM of the binomial family (link = probit) with diet treatment group as a fixed effect. Included was test day (day 5, 10, 15 and 35) as a fixed effect, to test for the potential influence of increasing chick age and experience on food preferences. Chick ID was included as a random effect to account for repeated sampling. Test trials where no food was consumed were not included in the analysis.

We used Chi-squared tests to determine whether there was a significant preference for, or avoidance of, one of the four foods offered in the chicks’ first food choices for each test trial day (day 5, 10, 15 and 35) separately, to avoid pseudoreplication. To test for individual consistency in first and second food preferences, we performed two Intra-Class Coefficient tests (ICCs).

To test whether chicks were preferentially choosing higher protein food (fish or cat food) over lower protein options available (bread or mussels), we used exact binomial tests for each test trial day, excluding the baseline trial, to determine whether chicks chose a higher protein food first (yes or no) more often than would be expected by chance (0.5). We also used exact binomial tests for each test trial day to determine whether chicks were significantly more likely than chance to choose high protein foods consecutively in both their first and second choice (probability of occurring by chance was 0.83).

Does dietary treatment affect chick growth?

Chicks in the terrestrial group were significantly lighter at their last weight measurement before release as compared to the marine group (median difference = 110 g, 95% CL [10–215] g; Wilcoxon test, W = 138, n = 14 Marine diet chicks, 13 Terrestrial, P = 0.02). This weight difference measured when 52–79 days old was not apparent when chicks were ≤15 days old (Wilcoxon tests at 5 and 15 days old: P value > 0.05). Similarly, when chicks were 31–58 days old, the tarsi of those in the terrestrial group were significantly smaller than those in the marine group (median difference = 3.5 mm, 95% CL [1.7–4.9] mm; Wilcoxon test, W = 152, P = 0.004). This difference between the diet groups was not quite significant when chicks were 14–16 days old (Wilcoxon test, W = 40, n = 8 Marine chicks, 7 Terrestrial chicks, P = 0.06).

Do herring gull chicks have food preferences, and does the rearing diet influence individual food preferences?

Twenty-three of the twenty-seven chicks participated in the baseline food preference test. Chicks showed no preference for either food type (exact binomial test: number of chicks choosing fish = 13/23; P = 0.68). In the later test trials with all four treatment diet foods presented in a 2 × 2 grid, chicks’ first food choices did differ from chance (Day 5 trial: ${\chi }^2$ = 26.68, df = 3, n = 25, P < 0.001; Day 10 trial: ${\chi }^2$ = 12.46, df = 3, n = 26, P < 0.001; Day 15 trial: ${\chi }^2$ = 14.33, df = 3, n = 24, P < 0.001; Day 35 trial: ${\chi }^2$ = 10.69, df = 3, n = 13, P = 0.01). As shown in Table 1 and Fig. 1, chicks preferred to consume fish first and appeared to avoid bread. Although mussels and fish seemed to be preferred over cat food, and both cat food and mussels appeared to be preferred over bread, these patterns were not significant. We found no influence of the diet treatment group nor the test day on chicks’ preferences for marine or terrestrial food types (Table 2; GLMM: P > 0.05). Chick ID explained no variation when included as a random effect in this GLMM model, indicating that the variation between individuals in initial food preferences was undetectable (Table 2). This might be due to most chicks consistently choosing marine and higher protein foods first in the great majority of trials (Fig. 2A). For chicks’ second food choices, fish and mussels were preferred over cat food, while bread remained a rare choice (Fig. 2B).

Chicks’ first and second food choices appeared to broadly reflect the amount of each food type that they consumed (Fig. 3). Bread consumption was very low, and only occurred after chicks had consumed all the food in the other bowls. Fish consumption showed the opposite pattern, with most chicks consuming all of the fish available. However, cat food consumption was quite high considering that cat food was less frequently the first or second choice of the chicks, compared to the marine foods. The amount of cat food consumed appeared to decrease with increasing time spent in captivity, while mussel consumption varied widely between individuals and across test days (Fig. 3).We found no significant differences between the chicks in their first and second food choices, nor significant within-individual consistency between test trial days 5, 10 and 15; day 35 could not be included in this analysis due to reduced chick participation (Table S2 in the Supplemental information; all ICCs: P > 0.05).

Do chicks prefer higher protein foods over lower protein foods?

Higher protein foods were generally more frequently consumed as chicks’ first choice (Fig. 4A) and second choice (Fig. 4B) during the food preference trials. However, exact binomial tests of a high protein food being chosen first (yes or no, probability by chance = 0.5) showed that only in the day 5 trials chicks showed a significant preference for high protein foods for the first food consumed (Table 3). Chicks generally ate higher percentages (>50%) of high protein food first in trials on days 10, 15 and 35, but this was not significantly different from the 50% expected by chance. Within the first two food choices made during trials, almost all chicks chose a high protein food (Table 3). Exact binomial tests of the frequencies of choice for higher protein foods within the first two choices suggest that chicks preferentially chose higher protein foods on days 5 and 10 (P = 0.02). Yet, their first two choices for high-protein foods on days 15 and 35 did not differ significantly from chance (P = 0.11 and P = 0.71 respectively; Table 3).