Supplementary information

Measurement of nest visitation rate

Data loggers recorded the presence or absence of a PIT tag every sixteenth of a second. Since the antenna’s detection radius is relatively small, a bird is detected when entering or leaving the nestbox, but not whilst sitting on the nest. Nest visitation rate was taken as the number of minute periodsbetween 0630h and 1230h on day 15 in which a parent was present. Thus, for example, a parent present for 50 seconds in a given minute interval would get the same nest visitation rate (of 1) as a parent present for 5 seconds in the same interval. Loggers were set up on day 14, giving parents time to get used to equipment at the nestbox entrance before measuring nest visitation rate; although irregular visitation behaviour was often recorded shortly after logger set-up, by morning all nests showed a characteristic normal pattern of feeding behaviour, with parents making short (less than a minute long) visits to the nest.

Accuracy of Plasmodium species assignment by qPCR

As part of ongoing research into avian malaria at Wytham Woods, we used nested PCR (Waldenström et al. 2004) to diagnose haemosporidian cyt b lineage in blue tit samples taken in 2005-7, including all samples used in this study. This allowed us to compare Plasmodium species diagnosis by qPCR melting temperature and cyt b sequence. Agreement in species diagnosis by qPCR and sequence analysis was high; in 96% of infections for which nested PCR gave a single lineage diagnosis (n=302), the species assigned by cyt b sequence and qPCR product melting temperature was identical. In the twelve cases of disagreement, 10 involved assignment of P. circumflexum by qPCR where nested PCR assigned P. relictum, and 2 involved the reverse situation. Thus there is a small degree of error in species-assignment by this method (4% infections, assuming nested PCR assignment is 100% accurate). However, it has the benefit of allowing many samples for which a sequence is difficult to obtain because of low parasitaemia to be assigned to species.

Waldenström, J., Bensch, S., Hasselquist, D., & Ostman, O. (2004) A new nested polymerase chain reaction method very efficient in detecting Plasmodium and Haemoproteus infections from avian blood. Journal of Parasitology, 90, 191-194.

Table S1: Results of univariate analyses indicating that measures of reproductive timing,clutch size, Plasmodium infection status and parasitaemia did not differ with respect to manipulation (treated as a 3-level factor) before the experiment. For the binary response variable Plasmodium infection status, the test statistic is χ2, for all other responses the test statistic is F.Plasmodium infection status and parasitaemia relate to females only.

2006 / 2007
Response variable / n / Test statistic / p / n / Test statistic / p
Lay date / 142 / 0.11 / 0.900 / 163 / 0.20 / 0.819
Clutch size / 142 / 0.33 / 0.959 / 163 / 0.42 / 0.659
Hatch date / 142 / 0.04 / 0.718 / 161 / 0.07 / 0.935
Plasmodium parasitaemia / 64 / 0.67 / 0.517 / 75 / 0.35 / 0.707
Plasmodium infection status / 140 / 0.47 / 0.791 / 165 / 0.17 / 0.918

Table S2: Analyses confirming the effect of brood size manipulation on parental reproductive effort (as assessed by the number of fledglings produced and the total brood mass on day 14) did not differ significantly between years, or according to parental Plasmodium infection status (at nests where the infection status of both parents was known), and showed no evidence of non-linearity. Brood size manipulation is entered as a continuous variable (‘Manipulation’, -1, 0, 1). Effects in the minimal model are shown in bold; for non-significant terms, statistics are those at the point that factor left the model.

Effect / df / N fledglings (n=261) / Brood mass (n=241)
F / p / F / p
Manipulation / 2 / 34.19 / <0.001 / 70.09 / <0.001
Year / 1 / 9.73 / 0.002 / 5.48 / 0.020
Manipulation*Year / 2 / 1.92 / 0.148 / 1.62 / 0.200
(n=170) / (n=158)
F / p / F / p
Manipulation / 2 / 39.68 / <0.001 / 79.84 / <0.001
N parents infected / 1 / 0.04 / 0.838 / 0.31 / 0.579
N parents infected*Manipulation / 2 / 1.01 / 0.367 / 0.30 / 0.745
(n=261) / (n=241)
F / p / F / p
Manipulation / 1 / 64.66 / <0.001 / 137.37 / <0.001
Manipulation2 / 1 / 1.06 / 0.305 / 0.20 / 0.656

Table S3: Results of GLZs predicting the probability of Plasmodium infection among blue tit parents involved in a brood size manipulation experiment. All variables included in the starting model are listed and effects in the minimal model are in bold; for non-significant terms, statistics are those at the point that factor left the model. Effect size (Pearson’s r) is shown for all effects with a single degree of freedom.

All Plasmodium (n=388) / P. relictum (n=383) / P. circumflexum (n=383)
Predictor / df / χ2 / p / r / χ2 / p / r / χ2 / p / r
Year (2006) / 1 / 0.59 / 0.441 / -0.039 / 5.53 / 0.019 / 0.120 / 3.43 / 0.064 / 0.095
Woodland sector / 4 / 6.95 / 0.139 / 15.66 / 0.004 / 34.97 / <0.001
Age (yearling) / 1 / 0.96 / 0.327 / -0.050 / 0.01 / 0.925 / 0.005 / 1.75 / 0.186 / 0.068
Sex (female) / 1 / 0.57 / 0.449 / -0.038 / 0.07 / 0.793 / 0.106 / 0.51 / 0.477 / 0.036
Clutch size / 1 / 2.27 / 0.132 / 0.077 / 0.92 / 0.336 / 0.049 / 0.70 / 0.402 / 0.043
Single parenthood (single) / 1 / 2.25 / 0.114 / 0.076 / 1.33 / 0.249 / 0.059 / 0.18 / 0.667 / 0.022
Manipulation / 1 / 0.01 / 0.941 / 0.004 / 2.60 / 0.107 / 0.082 / 1.16 / 0.282 / 0.055
Manipulation*Year / 1 / 0.06 / 0.811 / 0.012 / 1.22 / 0.270 / 0.056 / 0.16 / 0.689 / 0.021
Manipulation*Age / 1 / 0.91 / 0.341 / 0.048 / 0.47 / 0.494 / 0.035 / 0.79 / 0.373 / 0.046
Manipulation*Sex / 1 / 2.35 / 0.125 / 0.078 / 2.88 / 0.090 / 0.087 / 0.19 / 0.662 / 0.022

Table S4: Results of GLZs predicting probability of Plasmodium infection one year after the brood size manipulation experiment. Data are presented for models both (a) excluding, and (b) including infection status in the year of manipulation as a covariate. Effects in the minimal model are in bold; for non-significant terms, statistics are those at the point that factor left the model. Sample sizes are slightly lower for species-specific analyses as not all infections could be assigned to species.

All Plasmodium (n=92) / P. relictum (n=84) / P. circumflexum (n=84)
df / χ2 / p / χ2 / p / χ2 / p
(a)
Year / 1 / 1.30 / 0.255 / 5.58 / 0.018 / 0.24 / 0.622
Manipulation / 1 / 3.55 / 0.060 / 0.07 / 0.793 / 3.31 / 0.069
Manipulation*Year / 1 / 0.51 / 0.477 / 0.55 / 0.458 / 0.00 / 0.975
(b)
Year / 1 / 4.11 / 0.043 / 7.17 / 0.007 / 1.02 / 0.313
Manipulation / 1 / 1.94 / 0.163 / 0.13 / 0.716 / 0.86 / 0.353
Manipulation*Year / 1 / 0.31 / 0.575 / 1.20 / 0.274 / 1.50 / 0.221
Infection status / 1 / 28.11 / <0.001 / 6.72 / 0.010 / 32.63 / <0.001

Year differences in pre-experimental conditions

Conditions differed between the two years of study in several relevant respects. Blue tits in Wytham bred considerably later in 2006 than 2007, with a mean lay date of 26.4(where 1=1st April) compared to 16.5 in 2007, a difference clearly observed among nests involved in our experiment (year: F1,307=585.02, p<0.001); mean lay date in 2006 was the latest recorded in this population over the nine years for which data are available (2001-9). Prior to the experiment, females were significantly lighter in 2006 than in 2007 (year F1,303=10.84, p<0.001, controlling for time of day: F1,303=2.05, p=0.154). Among females for which tarsus measurements were available (those caught feeding nestlings) this difference in initial mass remained when tarsus was controlled for (year F1,217=4.36, p=0.038, tarsus F1,217=18.38, p<0.001, time of day F1,217=3.66, p=0.057), suggesting females were in poorer condition before the experiment in 2006. This appears to be a within-female effect, since among females present in both years, their body mass was significantly lower in 2006 (GLM for body mass with individual identity as random: year F1,64=4.63, p=0.039). Among non-manipulated individuals,Plasmodium parasitaemia was also generally higher in 2006 than 2007 (year F1,88=4.11, p=0.046), although there were not sufficient data to explore whether this was a within-individual effect. Hence, by several measures, female blue tits appear to have been physiologically more stressed in the first of the two years (2006) in which the experiment was carried out.

Analyses of within-female changes in parasitaemia

As a supplement to the analyses presented in Table 2, we also tested for within-individual changes in parasitaemia with respect to manipulation and single parenthood, among infected females. To do this we modelled parasitaemia using a mixed modelwith REML estimation, including female I.D. as a random factor. Sampling occasion (pre- or post-manipulation), manipulation, and their interaction term were fitted as fixed effects. The interaction terms tests for within-female changes in parasitaemia with respect to manipulation. Contrary to our expectation that brood enlargement would cause an increase in parasitaemia in 2006 (see Table 2), manipulation did not predict changes in parasitaemia in either year or years combined (manipulation*occasion, 2006: F1,67=0.59, p=0.448, 2007: F1,79=0.35, p=0.558, 2006-7: F1,151=0.00, p=0.998). Inclusion of females that were Plasmodium-positive at either occasion, but not both, did not change these results qualitatively. The apparent discrepancy between this result and the finding that manipulation is associated with a higher post-manipulation parasitaemia, may be due to a lack of power, since this analysis included approximately half the number of females as that in Table 2 (n=36 in 2006 and n=42 in 2007), and did not control for covariates known to predict parasitaemia (e.g. single parenthood). In a mixed model testing for the effect of single parenthood on changes in parasitaemia, single parenthood tended to increase parasitaemia among infected females in 2006 but not 2007 (SP*occasion, 2006: F1,67=3.29, p=0.079; 2007: F1,79=0.24 p=0.624).

1