1 DATA SUPPLEMENT 2 3 The allometry between secondary sexual traits and 4 body size is non-linear among cervids 5 6 Lemaître, J.F*., Vanpé, C., Plard, F. & Gaillard, J.M. 7 8 Université de Lyon, F-69000, Lyon; Université Lyon 1; CNRS, UMR5558, Laboratoire de 9 Biométrie et Biologie Evolutive, F-69622, Villeurbanne, France. 10 11 * Corresponding author: jeff.lemaitre@gmail.com; jean-francois.lemaitre@univ-lyon1.fr 12 13 14 1 15 Table S1: Dataset used in the analyses. Species Male Male body antler mass (Kg)1 length (cm)2 Male References antler (antler weight) weight (g) Alces alces Moose 482.5 144 11000 Geist & Bayer 19883 Axis axis Chital 89.5 84.5 2300 Geist & Bayer 1988 Axis porcinus Hog deer 41 39.9 Blastocerus dichotomus Marsh deer 130 60 Capreolus capreolus Roe deer 28 23.35 300 Geist & Bayer 1988 Capreolus pygargus Eastern roe deer 42 30.45 Cervus albirostris White-lipped deer 204 115 2950 Leslie 20094 Cervus canadensis Wapiti 350 133.7 12500 Geist & Bayer 1988 Cervus duvaucelii Barasingha 236 81.3 7700 Geist & Bayer 1988 Cervus elaphus Red deer 250 93.6 6300 Geist & Bayer 1988 Cervus eldii Eld's deer 105 97.15 3400 Geist & Bayer 1988 Cervus nippon Sika deer 52 48 2520 Geist & Bayer 1988 Cervus timorensis Timor deer 95.5 67.5 2500 Cranbrook 19915 Cervus unicolor Sambar 192 104.9 8200 Geist & Bayer 1988 Dama dama Fallow deer 67 61.5 3500 Geist & Bayer 1988 Elaphodus cephalophus Tufted deer 18 2.5 Elaphurus davidianus Père David's deer 214 73.7 7900 Geist & Bayer 1988 Hippocamelus bisulcus Chilean guemal 95 30 Mazama americana Red brocket 24.5 11.5 56 Geist & Bayer 1988 Mazama gouazoupira Gray brocket 18 9.25 Megamuntiacus vuquangensis Giant muntjac 45 22.75 Muntiacus crinifrons Black muntjak 23 3.55 Muntiacus gongshanensis Gongshan muntjak 21 7.17 Muntiacus muntjak Muntjac 19 14.2 Muntiacus putaoensis Leaf deer 12 3.25 Muntiacus reevesi Chinese muntjak 13.5 11.4 70 Geist & Bayer 1988 Odocoileus hemionus Mule deer 112.5 88.5 1650 Geist & Bayer 1988 Odocoileus virginianus White-tailed deer 154.5 65.6 1500 Geist & Bayer 1988 Ozotoceros bezoarticus Pampas deer 33.5 21 257 Ungerfeld et al. 20116 Pudu puda South pudu deer 13 8.5 13 Geist & Bayer 1988 Rangifer tarandus Reindeer 106.5 91 12000 Geist & Bayer 1988 16 17 1 All body mass data are extracted from : Plard et al. 2011. Oikos 120: 601-606 All antler length data are extracted from : Plard et al. 2011. Oikos 120: 601-606. 3 Geist & Bayer 1988. Journal of Zoology 214: 45-53. 4 Leslie 2009. Mammalian Species 42(849): 7-18. 5 Cranbook 1991 Mammals of South-east Asia. New York, NY: Oxford University Press. 6 Ungerfeld et al. 2011. North Western Journal of Zoology 7(2):208-212. 2 2 18 19 20 21 22 23 24 25 26 27 Table S2: Analysis of the relationship between antler weight (log-transformed) and body mass (log-transformed) across 20 cervid species. We compared models based on AICc and AICcw (see materials and methods section). In the threshold model with 1 slope, antler length increased linearly (one slope) up to 75.4 kg and remained constant from this body mass while with the model including 2 slopes, there is 1 slope before and 1 slope after the threshold value. ΔAICc is the difference of Akaike’s criteria corrected for small sample size between the candidate model and the best model and k is the number of parameters in the model. These results are qualitatively similar to the ones obtained with antler length since the threshold model (1 slope) is very close to the quadratic model (see Table S4). Retained model is in bold. k 2 3 4 4 5 Constant Linear Quadratic Threshold (1 slope) Threshold (2 slopes) Deviance 82.92 49.20 40.50 42.84 38.14 AICc 87.62 56.70 51.17 53.51 52.42 ΔAICc 36.46 5.53 0.00 2.34 1.25 AICcw 0.00 0.03 0.52 0.16 0.28 28 29 30 31 32 33 34 35 36 37 Table S3: Analysis of the relationship between ‘Independent contrasts of antlers length’ and ‘Independent contrasts of body mass’. We compared models based on AICc and AICcw (see materials and methods section). ΔAICc is the difference of Akaike’s criteria corrected for small sample size between the candidate model and the best model, and k is the number of parameters in the model. Retained models are in bold. Here it is important to note that the number of contrasts in the analyses is relatively low (n = 23) compared to the number of species (n = 31) in the analyses performed on antler length (see also Figure S2). This might explain why in this case, the threshold model does not perform better than the linear model. 38 Constant Linear Quadratic Threshold (1 slope) Threshold (2 slopes) k 2 3 4 4 5 Deviance 22.68 5.02 4.26 3.68 3.44 AICc 27.28 12.28 14.48 13.90 16.98 ΔAICc 15.00 0.00 2.20 1.62 4.70 AICcw 0.00 0.53 0.18 0.24 0.05 39 40 41 3 42 43 44 45 46 47 48 49 Table S4: Analysis of the relationship between antler length (log-transformed) and body mass (log-transformed) across the 31 cervid species. We compared models based on AICc and AICcw (see materials and methods section). In the threshold model with 1 slope, antler length increased linearly (one slope) up to 113 kg and remained constant from this body mass while with the model including 2 slopes, there is 1 slope before and 1 slope after the threshold value. ΔAICc is the difference of Akaike’s criteria corrected for small sample size between the candidate model and the best model and k is the number of parameters in the model. Retained models are in bold. Constant Linear Quadratic Threshold (1 slope) Threshold (2 slopes) k 2 3 4 4 5 Deviance 97.20 46.08 38.54 39.10 38.00 AICc 101.63 52.96 48.09 48.64 50.41 ΔAICc 53.54 4.87 0.00 0.55 2.32 AICcw 0.00 0.04 0.46 0.35 0.14 50 51 52 53 54 55 56 57 58 59 Table S5: Analysis of the relationship between antler length (log-transformed) and body mass (log-transformed) across species with male body mass inferior to 113 Kg (n = 22). We compared models based on AICc and AICcw (see materials and methods section). In the threshold model with 1 slope, antler length is fixed constant after the computed threshold value of male body mass (78 Kg) while with the model including 2 slopes, there is one slope before and on slope after the threshold value. ΔAICc is the difference of Akaike’s criteria corrected for small sample size between the candidate model and the best model (in bold) and k is the number of parameters in the model. Retained model is in bold. 60 Constant Linear Quadratic Threshold (1 slope) Threshold (2 slopes) k 2 3 4 4 5 Deviance 67.02 32.48 32.44 32.08 32.26 AICc 71.65 39.82 42.79 44.44 44.01 ΔAICc 31.84 0.00 2.98 4.62 4.19 AICcw 0.00 0.69 0.16 0.07 0.08 61 4 62 63 64 65 Figure S1: Relationship between antler weight and body mass (on a log-log scale) across 20 cervids. The quadratic (full line) and threshold (with one slope, dashed line) models best described the relationship. 66 67 68 69 70 Figure S2: Relationship between ‘Independent contrasts of antlers length’ and ‘Independent contrasts of body mass’. The linear (full line) and threshold (with one slope, dashed line) models best described the relationship. 71 5