Application of Forensic Discriminant Functions to a Spanish Cranial Sample, Douglas H. Ubelaker
July 2002 - Volume 4 - Number 3
Research and Technology
Application of Forensic Discriminant Functions to a Spanish Cranial Sample
Douglas H. Ubelaker
Department of Anthropology
National Museum of Natural History
Ann H. Ross
C. A. Pound Human Identification Laboratory
University of Florida
Sally M. Graver
Department of Anthropology
National Museum of Natural History
Since the 1960s, forensic scientists have been aware of the importance of discriminant function analysis of measurements to assist in the estimation of sex and ancestry of human remains (Giles 1964; Giles and Elliot 1962; Hanihara et al. 1964; Steel 1962). Such functions offer powerful classification approaches once investigators have learned to record measurements and use the functions properly.
The forensic applications of discriminant function analysis were augmented considerably in recent years with the development of FORDISC 2.0 (Ousley and Jantz 1996). This interactive computer program offers custom discriminant functions for up to 21 cranial measurements. Unlike most previously published functions, this program allows classification, even with incomplete remains for which only a limited number of measurements is possible.
The reference samples used in FORDISC 2.0 are based on data recorded in the Forensic Data Bank (Jantz and Moore-Jansen 1988; Moore-Jansen et al. 1994). These data were collected from identified forensic cases, making them ideally suited for forensic applications. FORDISC 2.0 uses over 1,400 cases from the Forensic Data Bank, offering more relevant information than many of the previously published discriminant functions based largely on museum collections. FORDISC 2.0 also classifies individuals based on the Howells database (Howells 1973; Howells 1989) of cranial measurements taken on museum collections of archeologically recovered remains from around the world.
The discriminant functions derived from these two databases offer different, somewhat contrasting approaches to ancestry classification. The Forensic Data Bank includes categories of American black males and females, American Indian males and females, American white males and females, Chinese males, Japanese males and females, Vietnamese males, and Hispanic males (from the United States, Mexico, and Central America, but mostly representing Mexican Americans). Unknown remains are classified into the racial categories that are represented within the Forensic Data Bank. The race categories are then based on known information about the identified individuals within the database. Ousley and Jantz (1996: 20-24) provide an excellent discussion of these social race categories.
The Howells groups are those used in his 1973 and 1989 publications and largely reflect the names assigned to the archeological samples examined. Since these are older samples and more geographically diverse than those in the forensic database, they offer a different perspective.
Discriminant function classification assumes that the unknown originates from one of the reference samples within the database. Thus, the utility of the software in a forensic context depends largely on the similarity of the samples in the database with the population of the unknown (Birkby 1966). Although it would be surprising if an unknown forensic case originated from any of the specific Howells samples, the option could offer some potentially useful insight into the likely ancestral origins of the unknown individual.
FORDISC 2.0 has emerged as a powerful tool in the forensic analysis that is routinely employed in most North American forensic laboratories. Because the Forensic Data Bank was developed largely from North American forensic cases, its obvious use would be in applications to North American cases (Ubelaker 1998). However, Ousley and Jantz (1996:19) caution against applying it to individuals whose "race or ethnic group is not represented in the reference sample." The caution becomes more apropos as the international use of the program increases. With the increase in today's world travel, a representative of almost any population in the world could end up being a forensic case in almost any place in the world.
The Departamento de Biología Animal of the Universidad Complutense of Madrid, Spain, curates a skeletal collection from Wamba, near the Spanish towns of Villanubla and Valladolid in northwestern Spain. The collection, likely dating from the 16th to 17th century, originates from a large secondary ossuary deposit in the Church of Santa Maria and represents individuals affiliated with that church (Lopez de los Bueis 1998; Perez de Barradas 1952). Through the courtesy of curator Maria Dolores Garralda of the Universidad Complutense, this sample was made available to the authors for measurement during the summer of 2001. The 20 measurements listed in Table 1, as defined by Moore-Jansen et al. (1994), were used in the Forensic Data Bank and this study. This large Spanish sample offers an excellent opportunity to examine the application of FORDISC 2.0 to a diverse sample not represented in the database.
Even though this large sample represents a specific group at a particular time, the authors, during their examination of the remains, were impressed with the morphological heterogeneity within the group. Measurements were recorded on 95 individuals in this sample. Of these, 58 were assessed to be male and 37 female using gross morphological indicators (Bass 1995; Ubelaker 1999). The measurements for each cranium were entered into FORDISC 2.0 and classified using both the Forensic Data Bank and Howells samples. Each cranium was then entered into FORDISC 2.0 and analyzed twice, once using reference groups of both sexes and once using reference groups of the assigned sex.
As shown in Table 2, all of the 37 crania that the authors estimated to represent females were also classified as females by both the Howells and Forensic Data Bank options. In contrast, of the 58 crania the authors estimated to represent males using morphological indicators, 33 (57 percent) were classified as females using the Forensic Data Bank, and 30 (52 percent) were classified as females using the Howells database. Because the Spanish crania tend to be comparatively small, the sex discrepancy likely represents FORDISC's classification solely by measurement without considering non-metric observations.
All of the crania in this Spanish sample originated from a 16th to 17th-century community associated with a church in northwestern Spain. Although specific information on the ancestry of any particular individual in the sample is not available, generally all individuals should be considered examples of this group.
Table 3 presents the race classification of all individuals in this sample using the Forensic Data Bank option. Of the 95 individuals, 42 (44 percent) were classified as white, 35 percent as black, 9 percent as Hispanic, 4 percent as Japanese, 4 percent as American Indian, and the remaining three individuals as Chinese and Vietnamese. Using the sex estimates made by the authors at the time of examination as Sex Known and classifying the crania by race within the specific sex category using the Forensic Data Bank option, the greatest percentage classified this time into the black category. However, the distribution of the other groups was similar to the race, Sex Unknown classification. The comparison of Sex Known racial classifications with Sex Unknown racial classifications reveals that group classification changed for 17 estimated males and 3 females. Note in Table 3 that the Forensic Data Bank samples of Hispanic, Vietnamese, and Chinese consist only of males.
Table 4 illustrates that, using the Howells database, the 95 individuals were classified into 21 different groups. The most common group was Egyptian, followed by a medieval sample from Austria. The 21 different groups represent samples from 15 different countries of Australia, Asia, Africa, Europe, and the Americas. Note that Table 4 includes only the classified groups of the Spanish sample.
Using the Sex Unknown option, group classifications using the Howells database were different from those using Sex Known (estimates of sex made by the authors) in 40 percent of males and 11 percent of females. The Egyptian group was the most frequently classified in both Sex Unknown and Sex Known attempts. Fifty-eight percent of crania from the Spanish sample were classified into European or North African groups using the Sex Unknown search and 55 percent using the Sex Known information.
Table 4 also presents the typicality probabilities for the group classifications of the Spanish sample. This statistic offers information about the likelihood that a particular skull actually belongs to the group under discussion. Generally values less than .05 are considered atypical and those greater than .05 typical.
Crania classified into European or North African groups were considered typical in 63 percent of classifications using Sex Known and 69 percent using Sex Unknown. Typicality values were slightly greater for crania classified into other groups (non-European or North African) with values of 74 percent for Sex Known and 73 percent for Sex Unknown.
Of the 65 crania of Sex Known with typicality values greater than .05, 51 percent were classified as European or North African and the remaining 49 percent as others. Of the 68 crania of Sex Unknown with typicality values greater than .05, 56 percent were classified as European or North African and 44 percent into other groups.
The attempt to classify the Spanish sample in terms of sex and race produced a variety of results. The sample displayed considerable heterogeneity and was not directly represented in either of the databases used in FORDISC 2.0. Differences between sex estimates made morphologically by the authors and those generated by FORDISC 2.0 can be explained largely by size factors within this Spanish sample. The results demonstrate why all morphological indicators should also be consulted in forensic analysis.
Variation in racial classification represents the lack of a Spanish sample within the FORDISC 2.0 databases as well as the human variation inherent within them. Individual crania were classified according to the best fit with the existing samples of the database, but the samples clearly were inadequate to elucidate the specific geographical origin of the overall Spanish sample. Although the actual ancestries of the individuals within the Spanish sample are unknown, it is possible, especially considering migration patterns across that part of Europe, that some ancestral representatives of the groups suggested by FORDISC 0.2 may be represented. However, it is more likely that the Spanish sample simply is quite different from samples in the existing databases. A majority of the crania were classified into European or North African groups. However, some crania were classified into groups with no clear geographic or ancestral relationship with the Spanish sample.
The authors agree with Ousley and Jantz (1996) that caution is called for when applying FORDISC 2.0, as well as other similar discriminant function approaches, to samples that are not well represented in the relevant databases. Investigators should remember that such approaches should complement, not displace, overall evaluation of remains. The authors also agree that additional and more complete samples from different geographical regions and groups are needed to augment the existing databases. Such additions would improve an already useful forensic tool and make it more applicable to international forensic cases.
Bass, W. M. Human Osteology, A Laboratory and Field Manual. 4th ed., Missouri Archaeological Society, Columbia, Missouri, 1995.
Birkby, W. H. An evaluation of race and sex identification from cranial measurements, American Journal of Physical Anthropology (1966) 24(1):21-28.
Giles, E. Sex determination by discriminant function analysis of the mandible, American Journal of Physical Anthropology (1964) 22(2):129-135.
Giles, E. and Elliot, O. Race identification from cranial measurements, Journal of Forensic Sciences (1962) 7(2):147-157.
Hanihara, K., Kimura, K., and Minamidate, T. The sexing of Japanese skeleton by means of discriminant function, Nihon Hôigaku Zassi [Japanese Journal of Forensic Medicine] (1964) 18:107-114.
Howells, W. W. Cranial variation in man. In: Papers of the Peabody Museum. Volume 67, Peabody Museum of Archeology and Ethnology, Harvard University, Cambridge, Massachusetts, 1973.
Howells, W. W. Skull shapes and the map. In: Papers of the Peabody Museum. Volume 78, Peabody Museum of Archeology and Ethnology, Harvard University, Cambridge, Massachusetts, 1989.
Jantz, R. L. and Moore-Jansen, P. H. A database for forensic anthropology: Structure, content and analysis. In: Report of Investigations No. 47, Department of Anthropology, University of Tennessee, Knoxville, Tennessee, 1988.
Lopez de los Bueis, M. I. Indicadores de presión ambientál y dimorfismo sexual en una población española (Wamba, Valladolid). Doctoral thesis, Universidad Complutense de Madrid, Spain, 1998.
Moore-Jansen, P. H., Ousely, S. D., and Jantz, R. L. Data Collection Procedures for Forensic Skeletal Material. 3rd ed., University of Tennessee Forensic Anthropology Series, Knoxville, Tennessee, 1994.
Ousley, S. D. and Jantz, R. L. FORDISC 2.0: Personal Computer Forensic Discriminant Functions. University of Tennessee, Knoxville, Tennessee, 1996.
Perez de Barradas, J. La Huesera de wamba, Publicaciones de Revista de Antropologia y Etnologia (1952) 7:529-531.
Steel, F. L. D. The sexing of long bones, with reference to the St. Bride's series of identified skeletons, Journal of the Royal Anthropological Institute (1962) 92: 212-222.
Ubelaker, D. H. FORDISC 2.0: Personal computer forensic discriminant functions, International Journal of Osteoarchaeology (1998) 8:128-133.
Ubelaker, D. H. Human Skeletal Remains, Excavation, Analysis, Interpretation. 3rd ed., Taraxacum, Washington, 1999.