Supplementary material
Details of statistical models
Standard length (SL), body weight (BW), and the ten behaviors mentioned in the Material and Methods were compared between species using the generalized linear (mixed) model where species (T. rubripes (TR) and T. niphobles (TN)) was the fixed effect, and family (TR1–TR4 and TN1–TN4) and test year (2007 and 2009) were random effects. When observations were made on given behavior multiple times, the replications were also included in the model as a random effect. All the analyses were done using the R packages “Lme4” and “multcomp”. The results of model fitting for each trait were summarized at the end of this material.
1. Body size (SL and BW)
The data was assumed to be normally distributed, and the models were fit using maximum likelihood (ML). The formulae used and the results of the between species comparison are listed below. We compared the scores of this trait using glht function of “multcomp”; the two species were not significantly different for either trait (P = 0.751 and P = 0.0554 for SL and BW, respectively).
l Formula used: SL ~ Species + (1|Family) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 0.03333 / 0.10488 / 0.318 / 0.751
l Formula used: BW ~ Species + (1|Family) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 0.2094 / 0.1093 / 1.916 / 0.0554
2. Resume feeding (RF) from the acclimation test
We compared RF behavior between species by fitting the Poisson model with the Laplace approximation, and found a significant difference between species (P < 2x10-16).
l Formula used: RF ~ Species + (1|Family) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / -1.189 / 0.106 / -11.21 / 2x10-16
3. Face to observer (FT) from the feeding test
For this trait, behaviors were observed 20 times, and the replications were included as a random effect. We compared the score between species, by fitting the binomial model using the Laplace approximation. The scores was significantly different between species (P = 3.19x10-6).
l Formula used: FT ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison (Poisson)
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 6.403 / 1.375 / 4.658 / 3.19x10-6
4. Vertical movement (VT), horizontal movement (HR), swimming level (SW) and pausing (PS) behaviors from the swimming activity test
These behaviors were observed ten times each, and the replications were included as a random effect. The Poisson model with Laplace approximation was used for the traits VT and HR, and the Gaussian with ML for the traits SW and PS. The scores were significantly different between species for every trait (P < 2x10-16, P = 1.78x10-15, P < 2x10-16 and P < 2x10-16 for VT, HR, SW and PS, respectively).
l Formula used: VT ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 2.1413 / 0.1365 / 15.68 / <2x10-16
l Formula used: HR ~ Species + (1|Family) + (1| observation) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 3.4232 / 0.4306 / 7.949 / 1.78x10-15
l Formulas used for SW: SW ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 1.34105 / 0.03057 / 43.87 / <2x10-16
l Formula used: PS ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / -424.46 / 23.74 / -17.88 / <2x10-16
5. Ball nipping (BN) and ball pause (BP) from the novel object test
The Poisson model with Laplace approximation was used for BN and the Gaussian with ML for BP. These behaviors were observed two times each, and the replications were included as a random effect. The BN phenotype was found to be comparable between species (P = 0.156), while BP was significantly different between species (P < 2×10-16).
l Formulas used for BN: BN ~ Species + (1|family) + (1| replication) + (1|test year):
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 1.589 / 1.12 / 1.419 / 0.156
l Formula used: BP ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / -495.33 / 14.83 / -33.4 / <2x10-16
6. Swimming level after transfer (TL) and freezing after transfer (TF) from the open field test
For both traits, the Gaussian model was fit using ML. Both behaviors were observed two times each, and the replications were included as a random effect in the model. The scores were significantly different between species for both traits (P <2×10-16).
l Formula used: TL ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / 0.71259 / 0.06388 / 11.15 / <2x10-16
l Formula used: TF ~ Species + (1|family) + (1| replication) + (1|test year)
Between species comparison
Difference / Estimated Std. Error / Z-value / P-valueTR / - / TN / -190.833 / 9.337 / -20.44 / <2x10-16
Summaries of model fitting
l SL
Linear mixed model fit by maximum likelihood
Formula used: SL ~ Species + (1 | Family) + (1 | Test year)
AIC / BIC / logLik / deviance / df.resid204.9 / 218.6 / -97.5 / 194.9 / 109
Scaled residuals:
Min / 1Q / Median / 3Q / Max-3.7801 / -0.4082 / 0.0086 / 0.4551 / 3.5428
Random effects:
Groups / Name / Variance / Std.Dev.Family / (Intercept) / 0 / 0
Test year / (Intercept) / 0.02898 / 0.1702
Residual / 0.31347 / 0.5599
Number of obs: 114, groups: Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / t value(Intercept) / 5.03916 / 0.14141 / 35.63
Species / 0.03333 / 0.10488 / 0.32
Correlation of Fixed Effects: Species–Intercept= -0.371
l BW
Linear mixed model fit by maximum likelihood
Formula used: RF ~ Species + (1 | Family) + (1 | Test year)
AIC / BIC / logLik / deviance / df.resid202.7 / 216.3 / -96.3 / 192.7 / 109
Scaled residuals:
Min / 1Q / Median / 3Q / Max-2.24404 / -0.62699 / -0.02141 / 0.56437 / 2.69818
Random effects:
Groups / Name / Variance / Std.Dev.Family / (Intercept) / 0.001311 / 0.0362
Test year / (Intercept) / 0.000286 / 0.01692
Residual / 0.315764 / 0.56193
Number of obs: 114, groups: Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / t value(Intercept) / 5.15335 / 0.07808 / 66
Species / 0.20937 / 0.10928 / 1.92
Correlation of Fixed Effects: Species–Intercept= -0.698
l RF
GLMM fit by maximum likelihood (Poisson /Laplace approximation)
Formula: RF ~ Species + (1 | Family) + (1 | Test year)
AIC / BIC / logLik / Deviance / df.resid436.1 / 447.1 / -214.1 / 428.1 / 110
Scaled residuals:
Min / 1Q / Median / 3Q / Max-1.7885 / -0.71081 / -0.04126 / 0.65561 / 2.11269
Random effects:
Groups / Name / Variance / Std.Dev.Family / (Intercept) / 0 / 0
Test year / (Intercept) / 0.0009015 / 0.03003
Number of obs: 114, groups: Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / z-value(Intercept) / 1.89862 / 0.05574 / 34.06
Species / -1.18921 / 0.10604 / -11.21
Correlation of Fixed Effects: Species–Intercept= -0.444
l FT
GLMM fit by maximum likelihood (Poisson /Laplace approximation)
Formula used: FT ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / Deviance / df.resid1610.9 / 1639.5 / -800.4 / 1600.9 / 2275
Scaled residuals:
Min / 1Q / Median / 3Q / Max-2.1785 / -0.2559 / -0.0374 / 0.5376 / 5.9968
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 0.1897 / 0.4355
Family / (Intercept) / 1.719 / 1.3111
Test year / (Intercept) / 2.5712 / 1.6035
Number of obs: 2280, groups: Replication, 20; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / Z-value(Intercept) / -5.336 / 1.603 / -3.328
Species / 6.403 / 1.375 / 4.658
Correlation of Fixed Effects: Species–Intercept= -0.606
l VT
GLMM fit by maximum likelihood (Poisson /Laplace approximation)
Formula used: VT ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / Deviance / df.resid60575.5 / 60600.7 / -30282.8 / 60565.5 / 1135
Scaled residuals:
Min / 1Q / Median / 3Q / Max-13.984 / -4.383 / -2.697 / 2.914 / 84.269
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 0.02474 / 0.1573
Family / (Intercept) / 0.037 / 0.1924
Test year / (Intercept) / 0 / 0
Number of obs: 1140, groups: Replication, 10; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / Z-value(Intercept) / 2.9815 / 0.1088 / 27.41
Species / 2.1413 / 0.1365 / 15.68
Correlation of Fixed Effects: Species–Intercept= -0.63
l HR
GLMM fit by maximum likelihood (Poisson /Laplace approximation)
Formula used: HR ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / deviance / df.resid25942.2 / 25967.4 / -12966.1 / 25932.2 / 1135
Scaled residuals:
Min / 1Q / Median / 3Q / Max-8.381 / -1.956 / -1.249 / 1.184 / 79.19
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 0.0124 / 0.111229
Family / (Intercept) / 0.367 / 0.605707
Test year / (Intercept) / 3.89x10-7 / 0.000623
Number of obs: 1140, groups: Replication, 10; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / Z-value(Intercept) / 0.4933 / 0.3083 / 1.6
Species / 3.4232 / 0.4306 / 7.949
Correlation of Fixed Effects: Species–Intercept= -0.706
l SW
Linear mixed model fit by maximum likelihood
Formula used: SW ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / deviance / df.resid1746.5 / 1776.8 / -867.3 / 1734.5 / 1134
Scaled residuals:
Min / 1Q / Median / 3Q / Max-3.0819 / -0.5463 / -0.1539 / 0.4191 / 3.4171
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 0.000211 / 0.01451
Family / (Intercept) / 0 / 0
Test year / (Intercept) / 0.014077 / 0.11865
Residual / 0.266318 / 0.51606
Number of obs: 1140, groups: Replication, 10; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / t value(Intercept) / 1.35912 / 0.08676 / 15.67
Species / 1.34105 / 0.03057 / 43.87
Correlation of Fixed Effects: Species–Intercept= -0.176
l PS
Linear mixed model fit by maximum likelihood
Formula used: PS ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / deviance / df.resid14840.4 / 14870.7 / -7414.2 / 14828.4 / 1134
Scaled residuals:
Min / 1Q / Median / 3Q / Max-2.97372 / -0.01795 / 0.01028 / 0.69136 / 1.82813
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 0 / 0
Family / (Intercept) / 865.4 / 29.42
Test year / (Intercept) / 351.9 / 18.76
Residual / 25764 / 160.51
Number of obs: 1140, groups: Replication, 10; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / t value(Intercept) / 421.52 / 21.51 / 19.59
Species / -424.46 / 23.74 / -17.88
Correlation of Fixed Effects: Species–Intercept= -0.533
l BN
GLMM fit by maximum likelihood (Poisson /Laplace approximation)
Formula used: BN ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / deviance / df.resid328.1 / 345.2 / -159 / 318.1 / 223
Scaled residuals:
Min / 1Q / Median / 3Q / Max-1.1294 / -0.3811 / -0.055 / -0.0186 / 23.6617
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 1.433 / 1.197
Family / (Intercept) / 17.146 / 4.141
Test year / (Intercept) / 1.117 / 1.057
Number of obs: 228, groups: Replication, 10; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / Z-value(Intercept) / -6.694 / 4.413 / -1.517
Species / 1.589 / 1.12 / 1.419
Correlation of Fixed Effects: Species–Intercept= -0.131
l BP
Linear mixed model fit by maximum likelihood
Formula used: BP ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / deviance / df.resid2810.5 / 2831.1 / -1399.2 / 2798.5 / 222
Scaled residuals:
Min / 1Q / Median / 3Q / Max-4.6626 / -0.2382 / -0.2382 / 0.6967 / 3.2901
Random effects:
Groups / Name / Variance / Std.Dev.Replication / (Intercept) / 0 / 0
Family / (Intercept) / 0 / 0
Test year / (Intercept) / 0 / 0
Residual / 12534 / 112
Number of obs: 228, groups: Replication, 10; Family, 8; Test year, 2
Fixed effect:
Estimate / Std. Error / t value(Intercept) / 522 / 10.49 / 49.78
Species / -495.33 / 14.83 / -33.4
Correlation of Fixed Effects: Species–Intercept= -0.707
l TL
Linear mixed model fit by maximum likelihood
Formula used: TL ~ Species + (1|family) + (1| replication) + (1|test year)
AIC / BIC / logLik / deviance / df.resid279.8 / 300.3 / -133.9 / 267.8 / 222
Scaled residuals: