Forensic anthropologists assign sex and population group (race) to individuals on the basis of skeletal remains. While the most useful bones for these determinations are cranial and pelvic, these are not always available. The purpose of this paper is to provide models for classification using metacarpals and hand phalanges. Four samples of 40 individuals each (black and white males and females) form the dataset. Measurements include lengths and radioulnar and dorsopalmar widths of the 19 bones of each hand. The large number of total variables necessitated separate models for metacarpal and phalangeal categories; due to the considerable number of significant differences between corresponding right and left hand variables, separate models were created for right and left sides. A stepwise discriminant procedure was used to select variables, with some highly correlated (r > 0.85) variables subsequently removed. The model for left hand metacarpals has the greatest power of discrimination (89.4%); that for right hand middle phalanges, the least (71.7%). Metacarpals assign approximately 87–89%, proximal phalanges 76–79%, middle phalanges 72–79%, and distal phalanges 81–83% of individuals to their correct sex and population groups. Models exchanging variables selected from one side for corresponding variables on the other show discriminating power ranging from 72.3 to 85.6%. Thus roughly 70–90% of individuals are correctly classified by these models; more conservative “jackknife” estimates yield a success rate of approximately 67–82%. When these models are used for classification of sex alone, 89.9–94.4% (“jackknife” range, 88.7–94.4%) of cases are correctly classified; for race alone, 80.5–98.1% (“jackknife” range, 77.4–96.9%).

Data from the Child Research Council (Denver, CO) were analyzed to model longitudinal growth changes in the humerus, radius, femur, and tibia in 31 boys and 36 girls between 3 and 10 years of age. Multilevel modeling of growth changes allowed efficient estimates of bone size and bone growth variation to be obtained as well as comparisons of growth patterns within and between limbs. The long bones displayed decelerating growth through time, with greater velocities for the larger lower limb (vs. smaller upper limb) bones and the larger proximal (vs. smaller distal) elements within limbs. Coordination for bone size and growth velocity is good both within and between limbs, suggesting a common growth control mechanism that should make growth prediction possible. Adjusted for size, the tibia appears to be the most variable of these four long bones, which may be due to a combination of environmental effects and flexible growth potential.

Although cranial and pelvic bones are the preferred skeletal material used by forensic anthropologists to assign unknown individuals to their most probable sex and population (racial) groups, these remains may be unavailable. This paper presents models for classification using metatarsals, proximal pedal phalanges, and the first distal phalanx of the foot. Measurements include lengths and mediolateral and dorsoplantar widths of these foot bones. Four samples of 40 individuals each (black and white males and females) comprise the dataset. Models were developed separately for right and left sides. Three models are provided for each side: a metatarsal model, a proximal phalangeal model, and a combination model involving selected metatarsal and phalangeal measurements. A stepwise discriminant procedure was used for variable selection, with some highly correlated (r > 0.85) variables subsequently removed. The metatarsal models correctly assign approximately 77–84% of individuals to their correct sex and population groups; proximal phalangeal models yield correct assignments in 70–72% of cases, and the combination models give correct classifications in 87% of cases. Models exchanging variables selected from one side for corresponding variables on the other show discriminating power ranging from approximately 67–86%. More conservative “jackknife” estimates give correct assignments in 64–82% of cases. When these models are used for classification of sex alone, 86.2–93.7% (“jackknife” range, 84.3–91.2%) of cases are correctly classified; for race alone, 78.6–96.2% (“jackknife” range, 75.5–92.4%).

Data from the Child Research Council (Denver, CO) were utilized to model longitudinal adolescent growth of the humerus, radius, femur, and tibia for 36 girls (10-16 years) and 33 boys (10-17 years). Multilevel modeling procedures were used to estimate variation, covariation, and the polynomial parameters necessary for generating growth curves. At age 10, long bone lengths for girls and boys are similar; by age 16, each of the boys' arm bones is about 20 mm longer and each of their leg bones is about 30 mm longer. Due to the earlier maturation of girls, the models show the length of each of their long bones exceeding that of boys to some degree during some period of adolescence. Peak velocities for leg bones are attained earlier than those for arm bones; in both sexes, age at humeral peak velocity coincides with age at peak height velocity (PHV). At age 13, correlations among lengths and among velocities of bones are strong to moderate, and girls consistently display higher variation than do boys for both long bone length and growth velocity. Considered relative to the average velocity of each bone's growth, the tibia is the most variable of the four long bones at age 13 years.

Estimation of stature in adult forensic cases with available long bones of the limbs is routine, but such estimation is less common in subadult cases. Long bones from subadult cases are often used to estimate age, but in some instances stature may be helpful or even critical for identification. Few published regression equations exist for consultation in such cases. Data from the longitudinal growth study conducted by the Child Research Council in Denver in the mid-1900s are utilized to produce dual-sex and single-sex regression equations for the six long bones of the limbs (humerus, radius, ulna, femur, tibia, and fibula) and for the combined femur+tibia length. All measurements are from radiographs and are of diaphyseal length. Examples show that similar results can be obtained using a two-step process of "ballpark" estimation from published tables of the Denver data, but these new regressions allow a one-step standard error estimate for the means. Regressions are further compared with those previously published by Finnish researchers, which are generally broadly comparable. More routine stature estimation in subadult cases is encouraged both as an aid to possible identification and as a test of the available regression equations.