An integrative measure of cognitive performance, but not individual task performance, is linked to male reproductive output in budgerigars

Cognitive abilities such as learning and memory are key for survival and reproduction. Individuals with high cognitive abilities may be more successful at attracting mates and producing offspring. However, empirical tests of and evidence supporting this hypothesis remain scarce. We measured cognitive performance of male budgerigars in four tasks: problem solving, detour reaching, seed discrimination, and spatial memory. We then tested female choice for male cognition at three stages of the mating choice process: social pairing, extra-pair mating, and continued reproductive investment with a social mate. We also measured female reproductive output. We used an integrative measure of male cognitive performance that encapsulates performance across all tasks, the ‘composite cognitive score’ by summing performance on the four tasks. In the first stage, females did not choose their social mates based on any of the measures of male cognitive performance. In the second stage, however, males with higher composite cognitive scores sired and raised more offspring. In the third stage, females increased their reproductive investment after the first breeding attempt when paired with males with higher detour-reaching scores. These results suggest that female reproductive decisions may shape overall male cognitive performance.

individual was transferred from its housing cage to the test cage for testing and then transferred back to the housing cage once the test for that day ended.
Each male was also tested for exploratory behavior and neophobia before being tested in the battery of cognitive tests. Exploratory behavior was assessed as the proportion of regions marked on the test cage floor that the bird visited within 20 min.
We conducted this test of exploratory behavior the first day that birds were exposed to the test cage and again the following day. We tested individual response to 3 unfamiliar objects (problem-solving and seed discrimination device, and a small plastic statue of a red dragon next to a familiar food dish) and one novel food item (commercial papaya yogurt treats for parakeets). Neophobia was measured as the latency in seconds to first peck at the food or the object within a 20 min trial, which took place in the test cage. A Principal Components Analysis was conducted on all measures of exploration and neophobia and the first principal component, which explained 49.26% of the variance and reflected both exploratory and neophobia, was used as a personality score for subsequent analyses. Individuals with low values for this score were shy and little exploratory, whereas birds with high scores were bold and highly exploratory). For more details on personality tests in budgerigars and test correlations, see 1 .
All the birds were tested in the problem-solving task first. The order of testing for the other three tasks was not randomized, and it was rather dictated by logistics. However, we did not find a significant effect of testing order on performance in the spatial memory task (linear model (LM): t = -0.315, b = 1.073, P = 0.757), or the detour-reaching task (linear model (LM): t = -1.761, b = 1.238, P = 0.097), or the seed discrimination task (linear model (LM): t = -0.914, b = 64.064, P = 0.376).

Problem-solving task
Habituation: Before testing, individuals were habituated to the problem-solving device alone first and then with another male. During habituation, birds had open access to the food in the device (the clear plastic barrier was not present). An individual met the habituation criterion once it consumed food from the device in 3 individual habituation tests. Birds were habituated to the device for a maximum of 8 weeks. Only 3 individuals did not meet the habituation criterion and were not tested in the task.
Shaping: Habituation to the device was followed by a shaping phase, which consisted of progressive reduction of the degree of access to the reward by gradually sliding closed the clear plastic barrier. Each bird was shaped for a total of 20 5-min sessions over 2-4 days and then tested at 3 difficulty levels of the task (Figure 1c). Individuals were tested over 3 days in 3 blocks of 8 consecutive 5-minute trials, each block at a different difficulty level. The first trial of the daily session was preceded by a maximum of 3 min of free access to reward placed inside the device. If a bird consumed food before 3 min, then it was allowed to eat only for 10 s. Each individual was tested at all difficulty levels, from the lowest to the highest difficulty level.

Spatial memory task
Habituation: Males aged 11-17 months were habituated to the device until they consumed food for 2 days. Individuals that did not eat from the device in 5 days were not tested further (N = 9).

Detour-reaching task
Habituation and training: Birds were first habituated in the housing cages to an empty opaque cylinder of identical dimensions of the clear cylinder. Then birds were trained to eat from the opaque cylinder until they consumed millet from the inside of the cylinder 3 times (only one bird failed the training).

Housing and monitoring during the test of female choice for a social mate (First stage)
Each group was housed in a separate closed room (3.4 x 1.7 m) in a wrought iron flight cage (79 x 52 cm, and 135 cm high) under a 14:10 light:dark cycle (lights on at 0600, and off at 2000 hours). The room temperature was maintained at 24 ± 2 ºC under natural spectrum lighting. Humidity was maintained on average between 32-90%. All cages were equipped with multiple wooden perches of different thickness, toys, and parakeet seed mix, cuttlebone, and vitamin water were available ad libitum. Birds were fed ad libitum with a commercial breeding formula for parakeets with high protein and fat contents. In order to stimulate pair formation, 4 nest boxes were attached to the cage with free access for the birds.
All individuals in each group were video recorded each day for 15 min (males) or 5 min (females) in randomized order, preceded by a 5 min acclimation period with the experimenters in the room. 2 experimenters sat with a video camera approximately 3 m away from the cage. One experimenter filmed the focal individual to ensure that it remained in the field of view during the entire sampling time. The other experimenter took notes of the interactions observed and the leg bands of the interacting birds to ensure that each individual was correctly identified in case that the leg bands were not visible in the video recording. Video-recording sessions were conducted between 0800 and 1000 hours, 2-5 days a week. A pair was considered as stable if: 1) A male and female were seen copulating during 2 consecutive observation periods, or 2) a female engaged in allopreening with or was allofed by the same male during 2 consecutive observation days.
The validity of these criteria for pair stability was confirmed since pairs identified using such criteria remained together for the remainder of the study. All groups were allowed to form stable pairs for a maximum of 10 weeks.
Aggressiveness was quantified based on the interactions mediated by agonistic behaviors; whereas sociability was quantified based on interactions mediated by affiliative behaviors. Affiliative (courtship displays, allofeeding, allopreening, copulation solicitation, and copulation) and agonistic behaviors (displacements, attacks, bill thrusts, etc.) observed during the pairing experiment were classified from the video recordings using the behavioral descriptions in 2,3 and 4 . Only interactions that occurred one week before the first stable pair was detected (pre-pairing) were used to quantify both aggressiveness and sociability, since these interactions were the ones that were likely to play a role in female mate choice. The widely used social network metric, degree centrality 5 , was calculated to estimate the levels of both aggressiveness and sociability of each individual. Degree centrality is the product of the number of individuals that a focal bird interacts with, and the average weight (number of interactions) with these individuals, adjusted by the tuning parameter a. We set a to 0.5 and the degree centrality was divided by the total observation time for each individual, which varied between individuals from different groups. We were not able to obtain measures of aggressiveness or sociability for the 6 males in one group replicate because the first stable pair was observed within the first week that the pairing experiment was initiated. Social network metrics were calculated with the R package 'tnet' version 3.0.14 6 . Male body mass (to the nearest 0.1 g) and tarsus length (to the nearest 0.01 mm) were measured for all individuals at the beginning of their pairing experiment and used to calculate a body condition index (BCI) by dividing body mass by (tarsus length) 3 , and then multiplying by 10 4 , due to small values, as done in previous studies 7, 8 .

Breeding and nest monitoring
Birds were kept under a 14:10 light:dark cycle, under standard fluorescent lighting, at a room temperature of 24 ± 2 ºC. Their diet consisted of parakeet seed mix supplemented with parakeet breeding formula with high fat and protein content. The supplemented seed mix was provided in six different feeders, in close proximity, all in the same location of the aviary. These feeders were filled with seed mix, and seed husks were removed to prevent the blockage of the feeding trays at least every other day. Some feeders were filled less frequently than others because birds favored the location of some feeders over others. Additional small open seed containers were also offered to the birds in periods of high food demand (i.e. when pairs had nestlings or fledglings). Birds were also introduced to foraging toys throughout the breeding phase of the study, placed in different locations of the aviary. The foraging toys were filled with millet only, a highly preferred food item by this species. Birds were also provided with parakeet toys, cuttlebone, and water available ad libitum.
The contents of each nest box were checked and recorded three times a week immediately after the conclusion of the morning observation session. Each egg found in the nest boxes was marked with a number with a soft lead pencil in numerical sequence.
Eggs were candled in order to estimate the laying date based on the size of the air sac.
Whenever an egg hatched, we estimated the hatching date based on the last time that the nest box was monitored and the size of the nestling. A non-toxic permanent marker was used to mark each nestling on the right leg with a unique color that indicated its hatching order. This color-coding was consistent across nest boxes. Nestlings were banded with the same system employed for adults approximately at 30 days of age (before fledging).
We monitored from beginning to end a total of 29 nests (i.e. nesting attempts) and collected DNA from 83 nestlings throughout the 6-month breeding period.

PCR reactions, microsatellite design, and genotyping
Parental and offspring genomic DNA was extracted from the cards following the manufacturer's protocol with minor modifications (Whatman, Maidstone, UK). Genomic Samples were processed for fragment analysis on an ABI 3130XL using LIZ-600 size standard at the University of Texas El Paso's Genomic Analysis Core Facility. Genotypes were scored visually using Geneious 9.0.5 (http://www.geneious.com) 9 .
We searched for microsatellite regions by scanning the M. undulatus genome downloaded from https://www.ncbi.nlm.nih.gov/genome/?term=melopsittacus. We used the software Geneious version 9.0.5 (http://www.geneious.com) 9 to find these microsatellite regions in the budgerigar genome. Primers for 27 microsatellite regions were designed with the software Primer3 0.4.0 10, 11 with PCR product lengths between 150 and 400 base pairs. Only 6 from the 27 regions tested, amplified satisfactorily and also showed actual microsatellite markers after confirmation via sequencing.
We tested for Hardy-Weinberg equilibrium using Genepop version 4.2 12 . Presence of null alleles was tested with Cervus 3.0.7 13 . From the 8 microsatellite markers amplified, 7 were in the Hardy-Weinberg equilibrium and their maximum frequency of null alleles was 6% (Table S8). The marker BGMSAT14 was excluded from paternity analyses because it was not in the Hardy-Weinberg equilibrium, suggesting null alleles (Table   S10). The mean expected heterozygosity across the remaining 7 loci was 0.61 and the observed heterozygosity was 0.
We used Cervus 3.0.7 to assign paternity to the most likely father under both relaxed (80%) and strict (95%) confidence levels by calculating likelihood ratio scores (LODs).
Critical values for these scores were estimated with simulations using the following custom parameters: 30 candidate fathers and 0.99 proportion of loci typed. We simulated 10,000 offspring genotypes using default parameters in Cervus. We ran the 'paternity analysis' option since the known mothers were confirmed by our observations, except for 12 offspring for which the mother was unknown because two females laid eggs and raised these offspring in the same nest box. Paternity analyses were carried out in two steps: in the first step, we included as candidate fathers all 30 males. We excluded 8 offspring in this step because they were typed at fewer than 4 loci. We assigned a father to 61 offspring in the first step. The combined non-exclusion probability for this first round was 0.063 for the first parent and 0.009 for the second parent when the first was known. The nestlings that: 1) were not assigned a father with either 80% or 95% of confidence, and 2) had a known mother, were then selected for the second step (N = 14).
In this step, the pool of candidate fathers was then reduced to the female's mate and any other males that were seen copulating with the female during our daily observations.
Copulations were a common event during these observations and females allowed copulation only with few males. The total number of nestlings that met the 80% or 95% confidence criterion after both the first and second steps was 72. This final subset of nestlings was then employed in further statistical analyses to assess the effects of male cognition on number of nestlings sired and extrapair nestlings.

Supplementary Figures
Supplementary Figure 1. Effect of male cognitive performance on eggs laid in the first and subsequent nesting attempts. Measures of male cognitive performance shown are the composite cognitive score, problem-solving score, average number of errors in the spatial memory task, % maximum discrimination in the seed discrimination task, and detour-reaching score. The number of eggs that the female laid in the first nesting attempt correspond to the red line, and the number of eggs laid in subsequent nesting attempts correspond to the blue line.

Supplementary Tables
Supplementary Table 1