Decreasing parental task specialization promotes conditional cooperation

How much to invest in parental care and by who remain puzzling questions fomented by a sexual conflict between parents. Negotiation that facilitates coordinated parental behaviour may be key to ease this costly conflict. However, understanding cooperation requires that the temporal and sex-specific variation in parental care, as well as its multivariate nature is considered. Using a biparental bird species and repeated sampling of behavioural activities throughout a major part of reproduction, we show a clear division of tasks between males and females in provisioning, brooding and foraging. Such behavioural specializations fade with increasing nestling age, which stimulates the degree of alternated feeding visits, as a recently promoted form of conditional cooperation. However, such cooperation is thought to benefit offspring development, which is not supported by our data. Thus, from a proximate point of view, conditional cooperation via alternation critically depends on the division of parental tasks, while the ultimate benefits have yet to be shown.


Supplementary material 1 -Synchronization
Besides alternation, nest visit synchronization has been suggested as another form of parental cooperation [1][2][3] . Adaptive hypotheses for synchronization include improved exchange of information and signaling individual investment among care takers, equitable food distribution among offspring and reduced activity around the nest to minimize predator attention 2 . We here performed the same statistical analyses as described in the main document for alternation, but now assessing temporal variation in observed and expected levels of synchronization, as well as its impact on our four offspring fitness estimates.

Mathematical and statistical approach
Observed levels of synchronization (S) were calculated as twice the number of synchronous male and female visits (overlapping direct provisioning bouts, Sn) upon the total number of male and female nest visits: Similar as for alternation, an expected synchronization score could be calculated for each observation as the average between the theoretical maximum (Smax) and theoretical minimum (Smin) synchronization scores. This maximum depends on the number of visits by the least visiting parents (Tmin), for which all visit bouts ideally show overlap with its mate and is calculated as: The theoretical minimum synchronization score is 0 in all cases, representing an observation period in which direct offspring provisioning never overlaps.

Supplementary Figure 1: Temporal variation in population mean (± SE) nest visit synchronization as a proxy for parental cooperation.
The average between the theoretical maximal (white triangles pointing up) and minimal (white triangles pointing down) value was calculated for each observation and is considered the expected proportion of synchronized nest visits by chance (grey diamond symbols), given the observed parental nest visits. Observed synchronization scores (black dots, solid line) decreased with nestling age and were generally lower than the observed values. Daily Tukey corrected post-hoc differences between observed and expected samples are indicated as 'ns' (P>0.05), ** (0.01<P<0.001) and *** (P< 0.001).
Offspring growth, fledgling condition, fledgling body mass, NOx levels and nestling survival were not significantly affected by average synchronization levels (all P ≥ 0.16) and neither by changes in synchronization levels from day 3 until day 13 (all P ≥ 0.25). All other predictor variables had very similar effects as described for the analyses that included alternation (see main manuscript).

Conclusion
We found that the degree of synchronization strongly decreased with nestling age, which corresponds with the temporal pattern of female brooding. When females were on the nest and food was received by the male, females often directly transferred this received food to their offspring (A.I. personal observation). This frequently stimulated the males to jointly provide food directly to the offspring, explaining the highest synchronization levels during our early observations. The enhanced thermoregulation of the developing nestlings allowed females to allocate more time towards other behaviours, which may have reduced the trigger for males to jointly feed their nestlings. Synchronization in canaries can thus be summarized as a behavioural by-product of female brooding behaviour.
Similar as for the results on alternation, both overall and changing levels of synchronization have no effect on any of our offspring fitness estimates. The implications of these finding are discussed in the main article.

Statistical approach
The mixed models that explore temporal and sex-specific variation in the behavioural traits allowed to partition the between-individual from the within-individual variance through the random effect of BirdID 4 . This provides useful information about the temporal consistency of behavioural and cooperative parental strategies. We focused on the period of postnatal care and hence excluded the data collected during incubation. The repeatability of a given behavioural trait was then calculated as the between-individual variance upon the total variance, i.e. between-and within-individual variance 4 . The sim function (arm package) 5 was used to simulate values (10.000 iterations) of the posterior distribution for all model parameters and 95% credible intervals (CI).

Results
All six behavioural activities showed repeatable variation, with repeatability estimates being highest for foraging and resting, intermediate for offspring provisioning and brooding, and lowest for mate provisioning and singing behaviour (see Supplementary Table 1 Table   1).

Conclusion
Our three repeated measurements span a large part of the post-hatching period of parental care.
The observed repeatability was low (R < 0.3) for mate provisioning, song behaviour and both estimates for parental cooperation, moderate (0.3 ≤ R ≤ 0.5) for direct offspring provisioning, brooding and resting, and high (R > 0.5) for foraging behaviour. Thus the extent to which individuals are consistent in their behaviour, depends on the specific trait under consideration.
Furthermore, the observed repeatability values in the current study are generally lower than earlier reports for similar behavioural traits in our population 6,7 . This might be explained mainly by the longer period between the repeated measurements in the current study 8 . Given these behavioural repeatability estimates, combined with the described temporal patterns in the main article, we strongly recommend to include temporal dynamics of behaviour in future studies, and to carefully consider the time windows between repeated measurements 8 .