SUPPLEMENTARY MATERIAL: METHODS

DATA COLLECTION

The pup-escort relationship, pup begging and escort feeding

To quantify associations between pups and helpers, we carried out two hours of scan observations each day once new litters started foraging. Every five minutes, for each pup we recorded distance to (±10 cm) and identity of the nearest helper. At the end of each session, I classified a helper as an escort if the same pup was within 2 m for ≥ 40% of scans. When pup habituation allowed, we conducted focal watches on pups, following each pup for ≥ 20 min, counting begging calls using a hand-held clicker and recording ad libitum food items provided by helpers. At the end of every minute, we recorded the number of begging calls during that minute and the identity of every helper who was within 2 m of the pup for ≥ 20 seconds. I restricted analyses to pups observed for ≥ 3 observation sessions, to give a total of 572 focal watches on 37 female and 44 male pups from 16 litters in five packs, with an average litter size of 4.8 (range 1-17) and ranging in age from 28-81 days.Some focal watches were excluded in some analyses because data were missing for certain variables (due to rain, equipment failure or hostile megafauna).

Pup deprivation experiment

On a control morning, we carried out focal watches on 21 pups when the pack started foraging. The following morning, we removed these pups (either singly or in pairs) when packs emerged from the den. We picked up pups by hand, placed them in a covered trap and stored them in a cool shed. They were provided with ad lib water, but no food. Removals took 5 min and some removed pups gave distress calls, causing packs to search for them for 2-5 min, before foraging as normal. We released the removed pups at the start of afternoon foraging (c. 17:00), and waited for 10 min to allow them to find their escorts and for the pack to acclimatise to their return, before carrying out focal watches on the deprived pups. When two pups were released at the same time, two observers carried out simultaneous watches. Pups were accepted back without aggression and returned immediately to their escorts. The average age of removed pups was 47.4 days (range 38-55). They spent an average of 8.9 hours out of the group (±0.1 hours S.E.), with no discernible effects on development, survival or habituation (unpub. data).

Ethical Note

Pups of this age are frequently separated from their packs for up to 11 hours – they remain behind in the den with a babysitter (usually a young male not involved in escorting; Cant 2003). This means that the period of deprivation experienced by these pups was well within the natural range. However, removing pups and keeping them in traps is likely to be considerably more stressful. To minimise this stress, we did the following:

  1. Captured pups in pairs. A companion seemed to reduce distress, as indicated by frequency of distress calls and agitated movement.
  2. Placed traps in a cool, darkened shed, covered by a blanket. Darkness calmed the pups, and they usually spent most of the removal period asleep.
  3. Provided an ad-lib source of water from a hamster water bottle.
  4. Checked pups at least once an hour to monitor for signs of extreme distress or injury.
  5. Ensured that the room in which pups were stored was left undisturbed and was secure from ants, predators and toxins.

Pups lost an average of 8.5 g (± 0.2 S.E.) during removal (an average of 3.1 % ± 0.02 S.E. of body weight). This is in comparison to an average gain of 22.4 g (± 0.76 S.E.) across a similar time period on unmanipulated days (an average of 8.6 % ± 0.26 S.E.) when pups foraged with their packs. Weight lost during removal was not significantly different to weight lost on days when pups were naturally separated from their packs (paired t-test, t8=0.13, p=0.903).

Removal had no discernible effect on pup behaviour after release – they showed no increase in wariness toward observers and allowed themselves to handled and weighed as before removal. In many cases, pups would voluntarily re-enter a trap within minutes of release, indicating that they had not formed a negative association with the trap.

STATISTICAL ANALYSIS

Model construction

I carried out simple parametric tests in Minitab (all tests two-tailed), and constructed linear models using Genstat 8.1 (Lawes Agricultural Trust, Rothamsted, Harpenden, UK). Where analysis involved repeated sampling from within the same litter, I used a linear mixed model (LMMs), which allows fixed and random terms to be included. In these cases, the variance components were estimated using the Restricted Maximum Likelihood (REML) method, and the random term ‘Litter’ was retained in the model unless the variance component was found to be unmeasurable (<10-5). I sequentially dropped all potential explanatory terms until only terms explaining significant variation remained. I then tested each dropped term in the minimal model to obtain its level of non-significance. I tested all two-way interactions, but only present those explaining significant variation. I present the effect sizes of all significant terms – these are parameter estimates from the models and can be interpreted as the change in y per unit change in x. For categorical variables, such as sex, one level of the factor is set at 0, and the effect is relative to that factor level.

Pup deprivation experiment, changes in escort feeding

To investigate how pup and escort characteristics influenced the extent to which escorts increased their feeding rates in response to an increase in pup begging rates, I calculated the percentage change in escort feeding rate between control and experimental sessions, and constructed an LMM with percentagechange in feeding rate as the response variable and Litter as a random factor (see Table 1 for all potential explanatory variables tested). Litter accounted for significant variability (estimated variance component= 0.588 ± 0.265 S.E.).

Pup deprivation experiment, changes in pup begging

To investigate how pup and escort characteristics influenced the change in pup begging after deprivation, I calculated the percentage change in pup begging rate between control and experimental sessions, and constructed a linear model with percentage change in begging rate as the response variable (see Table 2 for all potential explanatory variables tested). Analysis was conducted on 20 pups with 20 different escorts (10 male pups with 4 female and 6 male escorts; 10 female pups with 6 female and 4 male escorts) from 11 litters in 5 packs (1 pup excluded because it was not possible to weigh its escort to obtain a measure of escort condition). I initially analysed the data using an LMM with Litter as a random factor, but Litter did not account for measurable variability (estimated variance component <10-6), and the results were not qualitatively different.

I used percentage change in begging rate as the measure of the change in pup begging effort because there wereconsistent individual differences in average begging rate during unmanipulated observation sessions: some pups begged at consistently higher rates than others. In an LMM investigating variables affecting natural variation in pup begging rates, individual identity accounted for significant variation after controlling for other variables, including pup sex and escort condition (LMM: χ2=121.35, P<0.0001, from 572 focal watches carried out on 81 pups from 16 litters in 5 packs, Individual identity entered as a fixed factor). This means that the same absolute change in begging rate represents a substantially different change in the average begging effort for different pups. Similarly, there were consistent individual differences in provisioning rates between escorts, after controlling for pup begging rate (LMM χ2=83.28, p=0.004, Escort identity entered as a fixed factor).

However, the results of both models the results are qualitatively unchanged if the analysis is repeated using absolute changes in begging and feeding rates: LMM on absolute change in escort provisioning rate: significant interaction between absolute change in begging rate*escort condition: χ2=13.95, p<0.001; significant interaction between absolute change in begging rate*pup sex: χ2=5.59, p=0.018. Linear model on absolute change in pup begging rate: significant effect of escort condition F1,19=17.86, p<0.001; significant effect of pup sex F1,19=5.04, p=0.04.

1