Size diversity in Swiss Bronze Age cattle

Abstract To date, osteometric data for Swiss Bronze Age cattle, particularly from Alpine sites, are scarce. In the present study, using a large dataset generated by combining preexisting data with recent data obtained from a large Alpine site, cattle size from the Late Neolithic to the Late Bronze Age (LBA) in populations from different sites and regions was evaluated using the logarithmic size index and other statistical analysis. Additionally, the finite mixture model and a meta‐analytic technique were used to observe possible effects of sex ratios on cattle size. Results indicated that sex ratios did not affect size distribution. Cattle populations did not differ over time, but the Alpine cattle were smaller than the Central Plateau cattle. There were two distinct sizes in the Alpine cattle populations. It is suggested that the different economic interrelationships between Alpine and other geographically related communities might have led to the emergence of size diversity in Swiss Bronze Age cattle. Further interdisciplinary studies with larger sample sizes are required to confirm these possibilities.

2016), introduction of new animal forms (Gaastra, 2014;MacKinnon, 2010), transalpine mobility, and migration of humans together with their livestock (Grupe, Hölzl, Mayr, & Söllner, 2017), or selection of specific sex, such as small female cattle (Manning, Timpson, Shennan, & Crema, 2015). Because body size and sex are strongly correlated (Davis et al., 2012) and sex ratio provides a hint of cattle exploitation, computation of sex ratios of cattle populations was considered necessary.
Recently, substantial Bronze Age osteometric data were acquired from an Alpine settlement at Savognin-Padnal (hereafter calls Padnal) (Bopp-Ito, 2012), which might have been near the copper mining site for bronze production (Della Casa, Naef, & Turck, 2016;Rageth, 1986). Thus, the aim of this study was to shed new light on the change and diversity of cattle size linked with husbandry practices across settlements during the Bronze Age. The analyses of cattle size and sex ratios as preconditions may provide evidence on possible influences of sex ratio on size diversity. Samples from the aforementioned different regional sites would address factors that might have had the most influence on cattle size diversity.

| Grouping of the studied samples by assemblages, regions, and periods
Combining the data on osteometric samples obtained from four assemblages (Horizont [synonymous with layer] B, C, D, E) in Padnal (Bopp-Ito, unpublished data) with preexisting data of samples obtained from 32 assemblages from LN (ca. 2800-2200 BC), EBA (2200-1600 BC), Middle Bronze Age (MBA; 1600-1300 BC), and LBA (1300-800 BC; Rychner, 1998b), sites of modern-day Switzerland and Liechtenstein, created a large sample size (total assemblages n = 36; Figure 1, Table 1, Table S1, and Appendices S1 and S2). Although cattle specimens were decreased at the Plateau sites, the ones at the Alpine sites were increased (Schibler, 2017; Figure S1); in particular, the senile adult cattle at the Alpine sites increased during the LBA (Bopp-Ito, 2012;Plüss, 2011;Stopp, 2015). In contrast, dominant senile Plateau cattle appeared between the LN and EBA (Hüster Plogmann & Schibler, 1997; Figure S2). The samples were grouped by (a) assemblage levels (a site that involved several settlements from different layers and a site that involved a single settlement) and (b) regional levels and by periods (Table S2 and Appendix S3). Sites were separated in different altitudes: Padnal (1,200 m), other Alpine sites (700-800 m), and Plateau sites (400 m).
The EBA and MBA Alpine samples were pooled into one period because some Alpine samples were from the transitional period between EBA and MBA (Rageth, 1986). Therefore, the Alpine samples were divided into two groups: (a) EBA and MBA and (b) LBA. Alpine LN samples were too scarce for this study due to insufficient number of settlements in this period, but Plateau LN samples were abundantly available. The Plateau samples were divided into three groups: (a) LN, (b) EBA, and (c) LBA. Plateau MBA sites have not been discovered thus far (Menotti, 2015).

| Osteometric samples obtainment and logarithmic size index
The osteometric samples from Padnal were identified as belonging to cattle (Bos taurus) (Bopp-Ito, unpublished data) following crossreferencing with published literature (e.g., Schmid, 1972)  This inclusion criterion (i.e., adult or subadult) was based on several previous studies (e.g., Becker, 1981;Becker & Johansson, 1981;Habermehl, 1975). Species identification and measurement of the originally reported data from the 32 assemblages were used following a description by Manning et al. (2015) and the references therein.
Calculation of the logarithmic size index (LSI = [log x − log m] = log (x/m), where x denotes the measurement to be judged and m denotes the measurement at the standard reference) is an effective method for estimating body size from specific osteometric parameters obtained FIGURE 1 Locations of the studied sites between Late Neolithic (square) and Bronze Age (dot) periods in Switzerland and Liechtenstein. Light gray (Green) dots (Nos. 1, 2, 3, and 4) correspond to the Alpine Bronze Age sites, dark gray (blue) dots (Nos. 5 and 6) to the Central Plateau Bronze Age sites, and dark gray (blue) squares (Nos. 6, 7, and 8) to the Central Plateau Late Neolithic sites. Numbers refer to the entries in Table 1: 1 = Savognin-Padnal; 2 = Cresta-Cazis; 3 = Schellenberg-Borscht; 4 = Ayent Le Château; 5 = Meilen-Obermeilen; 6 = Zürich sites; 7 = Pont-de-Thielle; 8 = Auvernier sites [Colour figure can be viewed at wileyonlinelibrary.com] from bone fragments, especially for generating a large sample size, and it is widely used in zooarchaeology to study geographical and chronological diversity of animal morphologies (Meadow, 1999;Simpson, Roe, & Lewontin, 1960). The LSI was applied to the osteometric data from Savognin-Padnal combined with the preexisting data of Swiss LN and Bronze Age cattle following a standardized protocol described by Breuer et al. (1999) and Schibler and Steppan (1999) and the references therein, because most of our samples provided data only on TABLE 1 Sample sources for the Late Neolithic and Bronze Age cattle excavated from four Alpine sites (18 assemblages) and nine Central Plateau sites (18 assemblages) in Switzerland and Liechtenstein. The map number (No.) corresponds to the locations shown in Figure 1. For details, see Appendix S1 width due to heavy fragmentation. The marginal measurements were excluded following aforementioned protocols and Duval et al. (2013). In total, 563 osteometric samples from postcranial skeletal elements based on eight measurement positions, which were selected to ensure adult or subadult specimens based on fusion of their epiphyses in later stages of age (e.g., Habermehl, 1975), were considered as parameters for the analysis (von den Driesch, 1976; Table 2). Standard reference values for LSI estimation were based on the bone measurements of a 13-year-old modern Hinterwälder breed female cow (Z-2431, withers height approximately 116.9 cm) from the collection of the Institute of Integrative Prehistory and Archaeological Science, University of Basel, as recorded by R. Ebersbach and G. Breuer (Breuer et al., 1999). LSI expresses the size of a cattle specimen relative to that of a reference specimen, with a specimen larger than the reference specimen (LSI value = 0) securing a positive value, and a specimen smaller than the reference specimen securing a negative value (Meadow, 1999). LSIs for each assemblage with at least over 10 observations were compared through notched box plots created using Wessa net software version 1.2.1 (Wessa, 2017), and histograms for Alpine assemblages were generated after Duval et al. (2013) using PAST software version 3.04 (Hammer, Harper, & Ryan, 2001).
The statistical differences between LSIs were examined according to assemblages and periods, following a standardized protocol described by Colominas, Schlumbaum, and Sana (2013) and Duval et al. (2013)

| Sex ratio
Differences in sex ratio possibly cause differences in body size distribution, as male cattle tend to be larger than female cattle. Our samples pooled the data on female, male, and castrated male cattle (Armitage, 1982;Grigson, 1982); however, sex information was not available from all the sites described in the references included in the current study. Therefore, to distinguish the sexes, the greatest distal breadth (Bd) of the metacarpus (see von den Driesch, 1976) as one of the most effective osteometric variable (Davis et al., 2012;Hüster, 1990) was extracted from the osteometric data and used for sex ratio determination (Table 1). Raw osteometric variables (not log-transformed) were used and analysed by fitting to a finite mixture model (McLachlan & Peel, 2000). The model assumes a mixture of two normal distributions with sex-specific means and standard deviations. From the model output, the percentage of female cattle can be identified by assuming that the lower mean value corresponds to female cattle. This method provided estimates of the percentage of females (referred to as female probability) and of the sex-specific means and standard deviations together with 95% confidence intervals. This model is fitted separately for each assemblage with at least six observations.
The variance in the true female probabilities based on the estimated female probabilities was estimated by using a meta-analytic technique (Russo, 2007

| Sex ratio diversity of cattle between assemblages and periods
The distribution of metacarpus Bd measurements of cattle bones from assemblages, with at least six observations, revealed a clear bimodal sex distribution ( Figure 2). Hence, it is possible to fit a finite mixture model successfully for each settlement. All estimates are reported in Table S3. No evidence for a variation in true sex ratios over time or within regions was found (tau = 0.0; p = .55), and the average female probability across assemblages was 60% (Figure 3). Furthermore, no evidence for an association between the mean LSI and female probability was found (p > .05).

| Size diversity of cattle between assemblages
The LSI values with over 10 observations revealed a highly significant difference in size between the cattle populations divided by assemblages and periods (χ 2 = 82.7, p < .0001). The size of Plateau cattle, except Zürich-Mythenschloss cattle, was homogenized. However, the Schellenberg-Borscht and Cresta-Cazis (hereafter calls Borscht and Cazis) Alpine cattle were larger than those at Padnal, which were the smallest (Figure 4). Pairwise comparisons confirmed that Padnal cattle were significantly smaller (Table 3). However, the histograms of all LSIs from Alpine assemblages showed that there was hardly any size reduction in Alpine cattle inside the settlements, that is, in Cazis and Padnal ( Figure 5). The distribution of mean LSI ( Figure 6) roughly grouped cattle

| Size diversity of cattle between regions
All pooled LSI data divided by regions and periods showed that size  Table S5.

| DISCUSSION
It is evident that Alpine cattle tended to be smaller than Plateau cattle, but there was no evidence for any time trends. In general, a substantial varia-  (Hüster Plogmann & Schibler, 1997). The correlation of sex ratios with mean LSI from a corresponding metaregression also did not hint at an influence of sex ratios on size distribution. Therefore, our hypothesis that the intensification of dairy production based on the increase in senile cattle at the Alpine sites during the LBA was not supported.
The Alpine cattle, especially Padnal cattle, could have been exploited for meat and as working power in agriculture and bronze production (Bopp-Ito, 2012;Plüss, 2011;Stopp, 2015). However, a larger sample size is necessary to draw definite conclusions regarding the role of sex ratios in determining size diversity.
Although there was no time trend, significant trends were found between the two regions and 12 assemblages. Cattle size distributions seemed to be related to the altitude for large Plateau cattle in the flat Plateau, middle Borscht and Cazis cattle in the lower Alps, and small Padnal cattle in the middle of the Alps. These habitats were formed by various types of fodder in soil, temperature, or agropastoral land use (Davis, 1981;Duval et al., 2013;Jacomet, 1998;Knockaert et al., 2017;Reitmaier, 2012), and the quality of diet could also have affected cattle size (Breuer et al., 1999). However, the lack of data based on various proxies, for example, pollen or tree rings, that enable dating and climate reconstruction, makes it difficult to explore the climatic effects on the samples from Alpine sites (Schibler, 2017).
As to the reason why small Padnal cattle suddenly appeared in the Alps, there is a possibility that they could be an introduced allochthonous population (Gaastra, 2014;MacKinnon, 2010), but it is difficult to discuss this due to the lack of the osteometric samples from Neolithic Alpine cattle. There is an alternative possibility that they could be a migrated population accompanied by migrants (Grupe et al., 2017) from areas outside the Swiss Alpine region, such as from the southern parts of the Alps (northern Italy and Tyrolian Austria), because the size of cattle in the southern parts of the Alps over 1,000 m above the sea level (e.g., Riedel & Tecchiati, 1998) is smaller than that in the northern parts (Riedel & Tecchiati, 2001;Stopp, 2015;Trixl et al., 2017), but homogeneous with that of Padnal cattle (Bopp-Ito, unpublished data), indicating a similar environment and close economic relationship between Padnal and the southern communities.
The location of Padnal beside the road that crosses over the Julier pass towards northern Italy, and excavation of outstanding foreign objects, particularly those with attributes of northern Italy (Rageth, 1986), suggest a deep relationship with north Italian communities through traffic and trade networks (Della Casa, 2007;Jecker, 2015;Jennings, 2015). The southern Alpine people who might have had a technology of bronze production (Marti-Grädel, Stopp, Deschler-Erb, Hüster Plogmann, & Schibler, 2012) perhaps migrated into the northern parts of the Alps with their livestock and selected the location at Padnal to build their settlement for bronze production (Della Casa et al., 2016;Rageth, 1986 Reitmaier et al., 2017) evidence to reconstruct phylogenetic relationships and places of origin and migration.
The middle-sized Alpine cattle at Borscht and Cazis might have migrated from the southern but more so from the northern communities, such as Plateau or South Germany, and these cattle might have been phylogenetically related because these sites possessed objects with northern attributes much more than Padnal during the EBA (Jecker, 2015). As they were located along the river Rhine (cf. Figure 1), contact with the northern communities through the river and lakes was more probable and frequent than with Padnal, which

| CONCLUSION
Osteometric data of Swiss LN and Bronze Age cattle from different regions and assemblages across periods were used to evaluate the differences in the size of cattle using the LSI and to explore the effect of sex ratio on size, based on a finite mixture model and a meta-analytic technique using the measurements of the greatest Bd of the metacarpus. Sex ratio did not correlate with mean LSI. Our hypothesis that the increase in female cattle for dairy production in the LBA Alpine sites was not proved. There was no evidence for time trends but there were differences in size between cattle populations that were presumably caused by varied nutrition in the diet linked to the altitude, but this is open to debate. The smallest Padnal cattle might have been migrated populations accompanying migrants mainly from the southern parts of the Alps, and they could have been exploited for meat and as working power for bronze production, mining, and agriculture.
Middle-sized Borscht and Cazis cattle might be more phylogenetically related to large Plateau cattle because these sites might have been economically interrelated more frequently with northern communities.
Further investigations with larger sample sizes and interdisciplinary studies, including genetics and geometric morphometrics to show a possible link between cattle populations, isotopic analyses to show diet variations, and strontium analyses to show the cattle and human migration, are required to confirm this possibility.