Parental provisioning behaviour plays a key role in linking

Parental provisioning behaviour plays a key role in linking personality
with reproductive success
Mutzel, A., Dingemanse, N. J., Araya-Ajoy, Y.G., Kempenaers, B.
Electronic Supplementary Material
Page
Contents
1
Text S1. Effects of song and model number on male aggressiveness
2
Text S2. Behavioural proxies for male aggressiveness
2
Text S3. Automated recording of visits to the nest
3
Text S4. Prior specifications for multivariate models
3
Table S1. Effects of nest stage on male aggressiveness
4
Table S2. Variance-covariance-correlation matrix for males
5
Table S3. Variance-covariance-correlation matrix for females
6
Table S4. Sources of variation in exploratory behaviour and feeding rate
7
Table S5. Sample sizes for exploratory behaviour and feeding rate
7
Figure S1. Experimental setup of the cage used for the exploration assay
8
Supplementary References
9
1
Text S1. Effects of song and model number on male aggressiveness
Likelihood ratio tests (LRT) were performed to test for effects of song and model number on male
aggressiveness using the restricted maximum likelihood method (REML). The test was constructed by
comparing a linear mixed-effect model with a Gaussian error structure containing nest stage and year as fixed
effect and random intercepts for song and model number (‘model 1’) with similar models where song number
(‘model 2’), or model number (‘model 3’), were respectively excluded. The LRT did not reveal a significantly
better fit of model 1 compared to model 2 (AICmodel 1= 356.2, log Lmodel 1= -170.1; AICmodel 2= 354.2, log Lmodel 2= 170.1, LRT= 0, P=1) and model 1 compared to model 3 (AICmodel 3= 354.2, log Lmodel 3= -170.1, LRT= 0, P=1). We
also ran this analysis using Bayesian methods (with similar model structure but using the MCMCglmm function
in R) [1], confirming that both song and model number had variance estimate that were close to zero (variance
explained by song number: 3.55×10-4 %, 95% CI: 2.32×10-5, 0.08; variance explained by model number: 1.43×103
%, 95% CI: 2.33×10-5, 0.20). Altogether these analyses imply that there was little to no variance explained by
song or model identity.
Text S2. Behavioural proxies for male aggressiveness
We recorded the following behavioural variables as proxies for male aggressiveness: latency to approach the
model within a 5 m radius (in sec), attack probability (a bivariate variable indicating whether a male attacked
the model), and latency to attack (in sec). We assumed that the latency and the probability of attack were both
good predictors of male aggressiveness in the context of territory defence [e.g. 2]. However, analysis based on
attack latency would have led to the exclusion of all subjects with missing values (i.e. non-attacking, nonaggressive birds; 55 out of 112 tested males) resulting in a highly biased dataset, whereas the bivariate variable
attack probability would not capture the total variation in aggressiveness of the population (see Dingemanse
and coworkers [3] for a similar discussion on the usage of latency variables). We therefore focused on
approach latency for the analyses presented in the main text; a continuous variable comprising all birds that
entered the 5 m radius around the intruder (80 out of 112 tested males). A mixed-effect model with attack
probability as the dependent variable, approach latency, nest stage as predictors, with individual identity as
random intercepts, and a binary error structure, showed that males that entered the 5 m radius more quickly,
were significantly more likely to attack the conspecific intruder (estimate: 0.27±0.10, p=0.008). This finding
suggests that approach latency was indeed a good predictor of male aggressiveness towards a conspecific
territorial intruder.
2
Text S3. Automated recording of visits to the nest
Our recording device consisted of an antenna (‘PIT-tag reader’) fitted around the entrance hole of the nest box,
a light barrier inside and another light barrier outside the nestbox hole, a power supply and a data logger
placed on the ground underneath the nestbox. The sequence of activation of the two light barriers indicated
the direction of the movement of a bird, allowing differentiation of entries and exits. Every time the bird
passed through the nestbox hole the PIT tag was read, thus determining the identity of the bird entering or
leaving the nestbox [4].
Text S4. Prior specifications for multivariate models
For each analysis, we used 5.3 x 106 iterations with a burn-in phase of 300 000 and a thinning interval of 5000,
resulting in a sample of 1000 values for each estimate. We ran each model with 4 different priors: (i) inverse
Wishart (V=diag(n), nu=n), (ii) inverse Gamma (V=diag(n), nu=b.002), where b = n-1, (iii) flat covariances
(V=diag(n)*10-6, nu=c), where c = n+1, and (iv) parameter expanded (V=diag(n), nu=c). Using different priors
resulted only in minor changes in the model output, strongly suggesting that the presented results are not
influenced by our prior choice. The results presented in this paper are from models with an inverse-Wishart
prior.
3
Table S1. Fixed effects estimates derived from a univariate mixed-effect model where male aggressiveness was
fitted as the response variable. We implemented the model using maximum likelihood (ML) with a Gaussian
error structure, where nest stage (4 levels: empty box, nest building, egg laying and incubation) and year were
fitted as categorical fixed effects, and random intercepts were fitted for individual identity. The intercept gives
the estimate of the mean aggression score for reference category nest stage ‘empty box’ in the year 2009. 95%
confidence intervals (CI) were calculated with the sim-function of the R package ‘arm’. To test for an overall
effect of nest stage on male aggressiveness we compared the full model (‘model 1’) with a similar model but
without nest stage as fixed effect (‘model 2’). Model 1 fitted the data much better than model 2 as indicated by
the lower AIC value of model 1 (AICmodel 1= 388.5, AICmodel 2= 399.7). These results imply that nest stage
explained variation in male aggressiveness, with males being the most aggressive before the onset of nest
building (when competing for territories) and least aggressive during the incubation period.
β
95% CI
intercept
-5.67
-0.72, -0.94
nest building
-2.32
-3.86, -0.84
egg laying
-0.35
-2.19, 1.48
incubation
-5.24
-7.80, -2.36
year 2011
-3.06
-4.12, -1.97
4
Table S2. Correlations, covariances and variances between behaviours and fitness estimates for male blue tits. We give the point estimate
plus 95% credible intervals between brackets for each variance component. Variances are given on the diagonal (shaded in dark grey),
covariances on the lower off-diagonals (shaded in light grey), and correlations on the upper off-diagonals (no shading). Correlations and
covariances with non-overlapping credible intervals (implying significant covariance) are given in bold face.
♂ aggression
♂ exploration
♂ feed rate
♀ feed rate
lay date
brood size
fledgling
no.
fledgling mass
♂ aggression
6.69
(5.13, 11.23)
0.27
(0.01, 0.54)
-0.39
(-0.64, -0.06)
0.24
(-0.08, 0.53)
0.09
(-0.15, 0.46)
0.02
(-0.34, 0.26)
0.06
(-0.28, 0.27)
-0.20
(-0.37, 0.14)
♂ exploration
20.90
(-3.76, 68.81)
1967.71
(1229.78, 2602.70)
0.11
(-0.27, 0.36)
0.19
(-0.20, 0.39)
-0.12
(-0.44, 0.17)
0.06
(-0.18, 0.35)
0.09
(-0.08, 0.40)
-0.08
(-0.33, 0.16)
♂ feed rate
-10.65
(-26.07, -1.77)
27.88
(-128.68, 224.79)
139.19
(86.89, 218.03)
-0.28
(-0.63, -0.09)
-0.39
(-0.64, -0.07)
0.40
(0.19, 0.63)
0.27
(-0.08, 0.50)
-0.30
(-0.58, -0.03)
♀ feed rate
6.25
(-3.74, 16.79)
21.81
( -89.61, 208.00)
-40.87
(-94.78, -0.89)
98.77
(67.71, 167.63)
-0.25
(-0.53, 0.10)
0.44
(0.12, 0.59)
0.59
(0.40, 0.75)
0.12
(-0.38, 0.24)
lay date
1.88
(-3.04, 7.00)
-29.69
(-116.01, 38.14)
-22.42
(-50.40, -1.97)
-13.74
(-33.20, 8.38)
26.74
(20.44, 39.88)
-0.52
(-0.64, -0.25)
-0.41
(-0.59,-0.18)
0.18
(-0.09, 0.37)
brood size
0.14
(-2.20, 2.01)
11.49
(-20.39, 35.92)
13.26
(3.57, 20.60)
8.64
(2.20, 16.29)
-5.36
(-9.88, -2.34)
5.59
(3.95, 7.92)
0.76
(0.61, 0.83)
-0.55
(-0.70, -0.36)
fledgling no.
-0.29
(-2.20, 1.93)
9.64
(-10.56, 45.58)
7.50
(-2.65, 16.68)
14.75
(7.72, 25.33)
-5.36
(-9.21, -1.96)
4.04
(2.60, 6.32)
6.92
(4.46, 8.74)
-0.14
(-0.38, 0.16)
fledgling mass
-0.48
(-1.28, 0.48)
-2.91
(-16.65, 8.42)
-3.75
(-8.80, 0.59)
-1.62
(-4.63, 3.26)
1.13
(-0.48, 2.58)
-1.32
(-2.25, -0.78)
-0.40
(-1.11, 0.50)
1.06
(0.82, 1.80)
5
Table S3. Correlations, covariances and variances between behaviours and fitness estimates for female blue tits. We give the point
estimate plus 95% credible intervals between brackets for each variance component. Variances are given on the diagonal (shaded in dark
grey), covariances on the lower off-diagonals (shaded in light grey), and correlations on the upper off-diagonals (no shading). Correlations
and covariances with non-overlapping credible intervals (implying significant covariance) are given in bold face.
♀ exploration
♀ feed rate
♂ feed rate
lay date
brood size
fledgling no.
fledgling mass
1740.28
(977.98, 2489.73)
0.41
(0.11, 0.70)
-0.28
(-0.57, 0.05)
-0.38
(-0.62, 0.02)
0.10
(-0.18, 0.33)
0.18
(-0.13, 0.35)
0.02
(-0.25, 0.27)
♀ feed rate
152.89
(21.05, 353.86)
102.94
(71.04, 168.47)
-0.31
(-0.58, -0.06)
-0.15
(-0.50, 0.12)
0.36
(0.06, 0.56)
0.58
(0.33, 0.74)
0.11
(-0.33, 0.33)
♂ feed rate
-142.96
(-318.17, 57.98)
-38.89
(-95.04, -6.27)
156.25
(95.17, 234.76)
-0.51
(-0.68, -0.13)
0.44
(0.29, 0.69)
0.31
(0.04, 0.57)
-0.43
(-0.61, -0.03)
lay date
-74.20
(-165.37, 11.88)
-8.66
(-30.84, 10.50)
-28.44
(-53.07, -2.54)
27.19
(21.60, 41.52)
-0.50
(-0.69, -0.34)
-0.45
(-0.61, -0.24)
0.18
(-0.06, 0.41)
brood size
8.16
(-20.05, 35.43)
8.95
(0.67, 16.13)
11.75
(5.73, 24.66)
-6.19
(-10.98, -3.51)
5.79
(3.91, 8.13)
0.77
(0.65, 0.85)
-0.47
(-0.65, -0.29)
fledgling no.
17.23
(-12.49, 45.13)
14.08
(7.35, 25.25)
8.79
(-0.78, 19.50)
-6.22
(-10.27, -2.49)
4.55
(3.08, 6.94)
6.65
(4.70, 9.44)
0.01
(-0.29, 0.24)
fledgling mass
1.75
(-13.74, 14.30)
0.44
(-3.92, 5.57)
-5.55
(-10.63, 0.06)
1.12
(-0.59, 2.81)
-1.41
(-2.27, -0.64)
0.03
(-0.96, 0.83)
1.32
(0.93, 2.20)
♀ exploration
6
Table S4. Sources of variation in exploratory behaviour and feeding rate based on repeated exploration tests in
2009 and 2011 and repeated measures of feeding rates in 2011. We used univariate mixed-effect models
(MCMCglmm) with a Gaussian errors structure with random intercepts fitted for individual. For the model on
exploratory behaviour we further fitted sex, test sequence (1st vs. 2nd) [see 5] and year (only when measures for
2009 and 2011 were included) as fixed effects, whereas for the feeding rate model only sex was included. We
give estimates of adjusted repeatability, i.e. the proportion of ‘phenotypic variance not explained by fixed
effects’ explained by differences between individuals [6]. Values are reported with 95% credible intervals (CI).
Exploratory behaviour
(both years combined)
Exploratory behaviour
(2011 alone)
Feeding rate
(2011 alone)
β (95% CI)
β (95% CI)
β (95% CI)
96.85 (88.24, 105.02)
114.15 (105.31, 123.02)
21.02 (18.04, 24.06)
1.89 (-7.77, 12.57)
-5.31 (-19.06, 7.34)
-2.00 (-6.25, 2.00)
sequence
30.30 (20.63, 39.55)
30.45 (15.71, 44.54)
-
year
15.62 (7.13, 24.53)
-
-
r (95% CI)
r (95% CI)
r (95% CI)
0.66 (0.44,0.76)
0.60 (0.13,0.79)
0.78 (0.64,0.83)
Fixed effects
intercept
sex
Repeatability
Table S5. Sample sizes for exploratory behaviour and feeding rates for 2009 and 2011. Time intervals between
repeated measures of exploratory behaviour ranged between 13-69 days within year and between 297-402
days between years. The time interval between repeated measures of feeding rate was 4 days.
Exploratory behaviour
Feeding rate
2009
2011
combined
2011
N individuals
108
128
236
96
N total
132
152
284
192
7
Figure S1. Experimental setup of the cage used for the exploration assay. The cage consisted of a solid plastic
box with one mesh side (122L x 50W x 50H cm, Joko-Systemtechnik) and was fitted with 6 wooden perches
covered with plastic ivy to increase habitat complexity. For video analysis both the mesh side and the ground
were divided into 4 sections, resulting in a total of 14 different potential positions (including the 6 perches). A
sliding door on the right wall connected a small holding box (11L x 12W x 11H cm) with the experimental
chamber. The right side of the holding box was closed with a transparent sliding door, which was covered by a
piece of cloth. The bird was transferred to the dark holding box right after capture and was allowed to recover
from handling stress for two minutes. The subject was then released in the experimental chamber by opening
the left sliding door and at the same time moving the piece of cloth [see Ref. 3]. This movement induced all
birds to promptly enter the experimental chamber without any further handling.
8
Supplementary References
1. Hadfield J.D. 2010 MCMC methods for multi-response generalized linear mixed models: the MCMCglmm R
package. Journal of Statistical Software 33(2), 1-22. (http://www.jstatsoft.org/v33/i02/).
2. Garamszegi L.Z., Rosivall B., Hegyi G., Szollosi E., Torok J., Eens M. 2006 Determinants of male territorial
behavior in a Hungarian collared flycatcher population: plumage traits of residents and challengers. Behav Ecol
Sociobiol 60(5), 663-671.
3. Dingemanse N.J., Both C., Drent P.J., Van Oers K., Van Noordwijk A.J. 2002 Repeatability and heritability of
exploratory behaviour in great tits from the wild. Anim Behav 64, 929-938.
4. Mutzel A., Blom M., Spagopoulou F., Wright J., Dingemanse N.J., Kempenaers B. In Press. Temporal tradeoffs between nestling provisioning and defence against nest predators in blue tits. Anim Behav.
5. Dingemanse N.J., Bouwman K.M., van de Pol M., van Overveld T., Patrick S.C., Matthysen E., Quinn J.L. 2012
Variation in personality and behavioural plasticity across four populations of the great tit Parus major. J Anim
Ecol 81(1), 116-126. (DOI 10.1111/j.1365-2656.2011.01877.x).
6. Nakagawa S., Schielzeth H. 2010 Repeatability for Gaussian and non-Gaussian data: a practical guide for
biologists. Biol Rev 85(4), 935-956.
9