*2.1. Materials*

Forty pairs of modern human humeri are included in this study. All paired humeri were collected from archaeological sites with populations that varied in geographic location, chronological age, and subsistence pattern. Agricultural and nomadic/gathering populations are included because these lifestyles were the dominant subsistence patterns in pre-industrial East Asia. The subsistence patterns of these populations were determined by associated burial assemblages and relevant historical records. The populations will be referred to by their geographic locations, which are as follows: 1 Hubei population (HB): 9 males and 4 females collected from agricultural sites from Hubei Province, Central China spanning Qin-Han-Tang dynasties (221 BC ~ 907 AD). For some sites of this population, analyses of charred plant remains indicate that *Setaria italica* and *Panicum miliaceum* were the main food crops [47]. 2 Henan population (HN): 6 males and 5 females collected from an agricultural population from Junzicun cemetery, Henan Province, North China, which dates to Qing dynasty (1636 AD ~ 1912 AD). Historical records indicate that an agricultural economy was the dominant lifestyle of this population [48]. 3 Xinjiang population (XJ): 10 males and 6 females collected from nomadic populations attributed to Subeixi culture (1000 BC ~ 200 BC) from the Turpan Basin, Xinjiang Province, Northwest China. Burial assemblages such as bows, arrows, and stone artifacts for males and spinning wheels and potteries for females indicate a subsistence pattern of nomadism and gathering [49–51]. All individuals were adults. Their age and sex were determined according to cranial and pelvic osteological indicators. All humeral specimens were intact, well preserved, and showed no symptoms of osteoporosis or other pathologies.

## *2.2. Data Collecting and Processing*

All humeri were scanned by a 450 kV micro-CT scanner (designed by Institute of High Energy Physics, Chinese Academy of Sciences) located in Key Laboratory of Vertebrate Evolution and Human Origins, Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences. The scanning was performed under a voltage of 380 kV, current of 1.5 mA, 360◦ rotation with a step of 0.5◦, and an isometric voxel size of 160 μm. Raw data were virtually reconstructed and segmented in VGStudio Max 3.0. Volume renderings of all humeri were aligned to anatomical position using the standard protocol defined by Ruff [26]. To ensure that the humeri were consistently aligned and to avoid inter-observer error, all alignments were made by one author (Y.Z.). Paired humeri were always aligned synchronously. Three-dimensional meshes of each aligned humerus were generated and saved as PLY files in Avizo 8.1 for the following analyses.

## *2.3. Cross-Sectional Geometric Parameters Calculation*

Customized in-house scripts, mainly sourced from R package 'morphomap', were applied to calculate the CSG parameters [52]. For each humerus, the single-layer periosteum and endosteum surface meshes were firstly detached from the original humeral mesh. Second, 61 equidistant cross-sections were extracted from the surface meshes along the proximodistal diaphysis (between 20 and 80% of the biomechanical length). Third, 360 equiangular landmarks were placed along both the inner and outer contours on each cross-section. Finally, J values of the cross-sections at 35% and 50% of biomechanical length (J35 and J50), and SMA values of 360 directions on 61 cross-sections were calculated based on the landmark coordinates.

## *2.4. Bilateral Asymmetry Quantification*

Commonly used practices for assessing bilateral asymmetry are absolute asymmetry ([(max − min)/((max + min)/2)] × 100%) and directional asymmetry ([(right − left)/((right + left)/2)] × 100%). However, absolute asymmetry is not appropriate in this study, as the magnitude relationship between the left and right side is not consistent among different landmarks at humeral diaphysis. However, our study still focuses on absolute information of overall bilateral asymmetry, so directional asymmetry is also not suitable, because it does not eliminate the impact of handedness as well as behavioral laterality, which is not the issue this study attempts to investigate and may bring about bias to the conclusion. Therefore, bilateral asymmetry was quantified using dominant asymmetry ([(dominant − non-dominant)/((dominant + non-dominant)/2)] × 100%). The dominant side was decided according to the magnitude of J50, given that it is a valid indicator of handedness [5].
