Shape Variability and Classification of Human Hair: A Worldwide Approach

Shape Variability and Classification of Human Hair: A Worldwide Approach

de la Mettrie, Roland

Abstract

Human hair has been commonly classified according to three conventional ethnic human subgroups, that is, African, Asian, and European. Such broad classification hardly accounts for the high complexity of human biological diversity, resulting from both multiple and past or recent mixed origins. The research reported here is intended to develop a more factual and scientific approach based on physical features of human hair. The aim of the study is dual: (1) to define hair types according to specific shape criteria through objective and simple measurements taken on hairs from 1,442 subjects from 18 different countries and (2) to define such hair types without referring to human ethnicity. The driving principle is simple: Because hair can be found in many different human subgroups, defining a straight or a curly hair should provide a more objective approach than a debatable ethnicity-based classification. The proposed method is simple to use and requires the measurement of only three easily accessible descriptors of hair shape: curve diameter (CD), curl index (i), and number of waves (iv). This method leads to a worldwide coherent classification of hair in eight well-defined categories. The new hair categories, as described, should be more appropriate and more reliable than conventional standards in cosmetic and forensic sciences. Furthermore, the classification can be useful for testing whether hair shape diversity follows the continuous geographic and historical pattern suggested for human genetic variation or presents major discontinuities between some large human subdivisions, as claimed by earlier classical anthropology.

KEY WORDS: HAIR CLASSIFICATION, ETHNICITY, HAIR SHAPE.

For many years cosmetic scientists have tried to measure physical features of human hair, such as its shape or color, because these features can be artificially modified by adequate and efficient cosmetic products. With regard to hair shape, previous anthropological studies have emphasized its variability within and between human ethnic groups. Many studies have broadly distinguished three ethnic human subgroups: African, Asian, and European, but this broad classification cannot account for the complexity of human biological diversity that arose from history, population differentiation, and admixture.

The appraisal of the wide variations of human hair in every part of the world is instinctive. Apart from hair’s length and its natural color, hair displays a tremendous morphologic diversity. With regard to individual hairs, intrinsic features-diameter, shape, physical properties, and so on-within various human populations have been the subject of numerous studies (Bernard 2003; Robbins 2002; Franbourg et al. 2003; Lindelof et al. 1988; Dekio and Jidoi 1990; Kreplak et al. 2001; Bouillon and Wilkinson 2005), the results of which are generally in fair agreement.

With regard to human populations, in which hair obviously shows various facets in shape, defining subtypes is highly difficult because, beyond the genetic uniqueness of Homo sapiens, the conventional division into African, Asian, and European subgroups has proven limits, overlaps, and uncertainty (Loussouarn 2001; McMichael 2003; Gray 2000). As an example, the European group includes individuals with highly different phenotypes with regard to hair, eye, or skin color, ranging from fair Scandinavians to dark-skinned Mediterranean types. In addition, terms such as ethnic subgroup and race remain vague (being possibly cultural, biological, geographic, etc.) and in extreme cases may be perceived as discriminatory.

The very word race, or racial, the use of which should be limited to firstsight categorization of the various human types that can be found in most mixed occidental huge cities, however, remains widely used by anthropologists or in official texts (Food and Drug Administration 2003; Rosenberg et al. 2002). Last, the frequent mixed-race human populations that have formed over the last 10 to 20 centuries have obviously produced many mixed phenotypes in an increasing number of countries (Goldstein and Chikhi 2002; Carvalho-Silva et al. 2001).

The description of hair shape with words ranges from the classic to the more sophisticated, using terms such as straight, wavy, curly, frizzy, kinky, woolly, and helical (Trotter 1938). Although these words clearly evoke a global appearance, such descriptors remain confusing because they are subjective, with ill-defined, overlapping limits. Objective descriptions are therefore required to more accurately define such verbal attributes. In this way measurements of average curvature, ratio of maximum to minimum curvature, crimp, ratio of natural to straight length, and so on are of interest to characterize human hair shape.

A major and impressive step forward, albeit at the individual level, was taken by Ogle et al. (1999). For forensic purposes Ogle and colleagues defined many objective parameters of an individual hair in an attempt to properly identify a hair found at a crime scene.

In the present study we mostly deal with the determination and type classification of hair shape from individuals of various origins. In no case was the natural color of the hair assessed, for the following reasons. First, there is no clear evidence of relationship with hair shape. Straight hair, although frequent in Asians, can be found in fair or black European hair as well. second, about 90% of inhabitants of the world have similar hair color, ranging from black to dark brown. As such, European hair color patterns (blond, red, auburn, etc.) are of a low statistical weight within world populations.

Could an approach that focuses on hair shape be applied to diverse human populations, and, if so, how? The purpose of the work reported here was to address this issue, keeping in mind that simple techniques should be used for easy handling and reproducibility by a great number of technicians.

The first step was to select the most accurate and simple quantitative characteristics of hair morphology. Then suitable conditions were determined to easily perform the measurements on large populations using well-defined protocols. Finally, collected data were processed using adequate statistical methods to identify various patterns from large samples.

Most of the variables used in this study are derived from the pioneering works of Hrdy (1973) andBailey and Schliebe (1985). Hrdy reported average curvature measurements, kinking, crimping level, and the ratio of normal to straight length for investigating quantitative hair shape variation. Bailey and Schliebe proposed the use of a numerical average curvature measurement for comparing curly hairs.

We chose to use four hair characteristics. Two of the characteristics are related to hair curliness: (1) curve diameter (CD), that is, the smallest curve diameter of hair cut at a given stretched length, referred to as L^sub 2^; and (2) curl index (i), that is, the ratio of the stretched length (L^sub 2^) to the greatest length (L^sub 1^) occupied by the same piece of hair at rest (no stress applied). The other two characteristics are related to the kinking of hair (Hrdy 1973): (3) number of twists or kinks (i), that is, a sudden constriction and twisting of the hair shaft along its main axis, expressed as number of coils per hair sample; and (4) number of waves (w) per hair sample.

Materials and Methods

Hair Samples. Hair was sampled from 1,442 human volunteers (72% women, 28% men) from 18 different countries. The volunteers were recruited at random through newspaper ads in large cities (Table 1). They were informed about both the purpose and conditions of the study and gave written informed consent. Natural, colored, bleached, and nonchemically straightened hairs (e.g., use of styling products or heat tongs) were included because they had not been involved in a process that had permanently altered hair morphology. However, hair treated with a permanent waving or a hair straightening product within the preceding nine months was not included.

Methods. From each volunteer single hairs were selected at random from three different areas of the head: vertex, temples, and nape. The following measurements were performed in real-life, noncontrolled conditions (relative humidity and temperature), thus reflecting the casual state of the hair sample on the day of collection.

Curl Index (i). Hair is cut at a fully extended length of 6 cm from its root. This length is referred to as the L^sub 2^ value. The hair is immersed for 3 min in 100 ml of a 1% solution (in water) of a basic nonconditioning shampoo and then thoroughly rinsed using tap water and placed on a paper towel. The hair is carefully blotted. After allowing the hair to air dry, the hair is placed for a minimum of 5 min on a dry paper towel.

The hair is then carefully laid on a glass plate without applying any mechanical stress, in order to allow it to maintain its natural shape. A second glass plate is gently placed onto the hair, carefully avoiding any side shifting or sliding. The distance between the two farthest points of the hair is then measured and referred to as L^sub 1^. The curl index i is defined as the ratio L^sub 2^/L^sub 1^.

Curve Diameter (CD). When hair is placed between the two glass slides, the radius of the curvature of each hair curve can be determined by superimposing a transparent template made up of circles of known radius (expressedin centimeters; see Figure 1) and matching the curvature of the hair with the closest circle of the template. The smallest or the tightest curvature is selected as the value of the curve diameter (CD).

Number of Wave s (w). The upper glass plate is removed and the two ends of the hair fiber are taped on the bottom glass plate. Each tape must cover only 0.5 cm of hair, with the distance between the two tapes being at least 4 cm. Between the tapes hair takes a sigmoid form whose number of peaks and number of valleys can be counted. (Figure 2). The higher number is referred to as w (number of waves).

Number of Twists (t). When curly or frizzy hair is fully extended, thicker and thinner parts or constrictions appear (Barbarat et al. 2005). The number of twists (i) equals the number of natural constrictions along the hair axis (Figure 3).

Measurements of CD, i, t, and w were carried out on each collected piece of hair, and at least three hairs (one from each area of the scalp, i.e., vertex, temples, and nape) were characterized for each subject. The mean value of each parameter was used as a descriptor of the hair of a given subject.

For statistical analysis the mean values of t and w were used. However, values of CD and i were expressed as logarithms [Neperian logarithms of (CD + 1) and (i + 1)] because they vary within a large range (CD from 0.1 to 350 and i from 1 to 26).

Statistical Analysis and Results. Principal components analysis, followed by hierarchical ascendant classification, was used to identify homogeneous groups of hair. A primary objective was to find key variables for assigning group membership. A further objective was to establish simple rules for predicting group membership of new subjects. To this end, a segmentation tree analysis was performed.

SPAD, version 5.6 (Decisia, Paris, France), SAS, version 8.2 (SAS Institute Inc., Cary, North Carolina), and SPSS Answer Tree, version 3.1 (SPS Inc., Chicago, Illinois), were used to analyze the collected data.

Principal Components Analysis. Principal components analysis (Benzécri 1984) reduces multivariate data by transforming it into orthogonal components that are linear combinations of the original variables. Principal components are calculated in such a way that the first principal component explains the larger part of the data variance and the subsequent principal components account for progressively decreasing independent parts of the remaining variance. Variables and objects are characterized by their coordinates relative to these components. Principal components plots can conveniently summarize the data and highlight relation patterns between the objects.

Table 2 shows the variance percentage explained by each principal component. It can be seen that two principal components account for more than 93% of the variance.

Hierarchical Ascendant Classification. Hierarchical ascendant classification (Diday et al. 1981; Lebart et al. 1982; Breiman et al. 1984) is a clustering algorithm method. Objects are first projected into principal component space; hierarchical ascendant classification involves iterative clustering of objects that are the closest in the space according to Ward’s criterion (Lebart et al. 1995). Ward’s approach consists in merging the two clusters with the lowest increase in intracluster variance and the lowest decrease in intercluster variance of the partition. When intracluster variance is low, the objects of a same cluster are similar; when intercluster variance is high, objects of different clusters are distant. This process yields a binary segmentation tree reflecting the hierarchy of similarities between objects.

The best partitions are obtained when the ratio between the intercluster variance and the total variance is high (this ratio is the rate of data variability expressed by the partition) and when adding a new cluster does not add new information (the intercluster variance gain is insignificant).

In our study the best partitions are found in two, four, five, or eight clusters. However, we found that those in eight clusters or groups were likely to provide the best compromise between statistical results and visual appearance of hair. Compared to a five-cluster partition, the eight-cluster partition allowed us to better differentiate frizzy hair from curly hair and also to distinguish groups where obvious differences were visually perceptible.

To illustrate visually these eight groups or clusters, we identified them on Figure 4 with ellipses. This plane shows the projection of samples onto the first two principal components derived from the principal components analysis carried out on the four variables CD, w, i, and t. The first axis (PC^sub 1^) includes about 87% of the total variance. Hairs with a higher CD are located on the left-hand side of the plot. Hairs with a higher i, w, or t are located on the right-hand side of the plot.

Figure 5 shows that CD is a key parameter for describing groups I to TV. Variable i allows two large groups (V + VI) and (VII + VIII) to be distinguished, whereas w and t, which are highly correlated (linear correlation coefficient = 0.91), are useful to split each of those two groups into two groups.

Segmentation Tree Analysis. Segmentation tree analysis is a statistical method (Breimanetal. 1984) used to determine which dependent variable, among multiple dependent variables, best discriminates between preestablished groups of objects (in our case the groups are the eight hair types and the objects are the subjects). This approach also gives rules for making predictions about the group membership of new objects.

Figure 6 summarizes the resulting classification rules, showing the variables and cut-values whereby the eight hair types are optimally discriminated. Because t and w are highly correlated, only three variables (CD, i, and w) are needed to describe and/or define hair groups. Thus classifying a given hair sample becomes simple: It clearly appears that four groups (types I, II, III, and IV) are distinguished by CD values, whereas the combination of i and w further differentiates the four other groups (types V, VI, VII, and VIII). For example, a value of CD greater than 10.5 leads to assignment to type I. A value of CD lower than 1.2 and an i value less than 5.9, together with a value of w less than 3.3, rank hair as type V.

Validation of the Segmentation Tree. The segmentation tree was validated using a cross-validation approach (Lebart et al. 1995) that consists in estimating how well the classification rules of the segmentation tree perform on future asyet-unseen data. The tree was unaffected by extreme values and is suitable for widespread use.

The relevance of the segmentation tree was evaluated by comparing reality and prediction, that is, by comparing the hair type of each subject as found by hierarchical ascendant classification (actual category) and the hair type proposed by the segmentation tree analysis (predicted category).

The percentage of subjects correctly classified by the segmentation tree analysis was excellent and varied from 100% for type I to 78% for type VIII. For incorrect classifications by segmentation tree analysis, there was never more than one class difference above or below (e.g., type II instead of type III or type IV instead of type III). More interestingly, Table 3 shows the distribution of these typological groups within the different populations.

“African” hairs collected from South Africa, Ghana, and Chicago were mainly classified as types V to VIII Although types VI to VIII prevailed in samples collected from South Africa and Ghana, type V was most frequent in the samples collected in Chicago and likely reflects a past mixed-origin population.

Hairs collected from China were essentially type I and type III. On the other hand, the amount of type III hairs reached 20% in samples from Japan and Korea; type in hair is still more frequent in samples from Thailand, where it is found in one-third of the samples.

Hairs collected from Brazilian volunteers showed a great diversity because they include types I to VI. This is likely the result of past mixed origins, leading to the well-known multiethnic populations.

In three countries (Korea, South Africa, and Ghana), where the cohorts of men and women were well balanced, no sex differences were found.

Photographs of representative volunteers are included to clearly illustrate the global appearance of hair from each type of hair grouping (Figure T).

Discussion

The approach reported here leads to the identification of eight types of human hair with good statistical confidence and fairly clear limits. Combining CD, i, t, and w values showed that hair types can be readily defined using only CD, i, and w. Parameters t and w were found to be correlated. Even though recent studies have given unequivocal evidence that curvature in curly hair is independent of ethnic origin and is genetically programmed from the basal area of the hair bulb (Thibaut et al. 2005), the genetic causes of hair curling, twisting, and kinking are still unknown. The correlation between t and w suggests only that the number of twists and waves measures the same biological characteristic. At the cellular level curly hair shows intrinsic asymmetry of the proliferative compartment from the earlier stage, involving delayed differentiation and different protein structure and distribution in the concave versus the convex side of the curvature (Thibaut et al. 2005). In addition to a different genetic program, different mechanical stresses exerted on the concave side from the early stage of hair growth can reflect on both t and w.

Attempts to detect sex-related differences in hair type in well-balanced population groups did not reveal any variation of type distribution, suggesting that the intrinsic three-dimensional shape of human hair is not sex related.

More interestingly, Table 3 shows the overall distribution of hair types among people of various ethnic origins, assuming that the human cohort involved in the study is fairly representative of worldwide human diversity. Our results show that types I to III are the most widely distributed. Types IT and III are found in at least 15 countries of the 18 included in our study.

Some points have to be emphasized. As mentioned, so-called straight hair, as embodied in types I and III, is clearly not exclusive to Asian hair. Although straight hair is shared by 93% of Chinese, it accounts for 57%, 68%, and 57% of tested Caucasian samples from Denmark, Poland, and Germany, respectively. Such a weight again justifies the initial choice not to include any hair color parameter in our tier approach, because blond hair is a frequent feature in these three countries.

Table 3 depicts some close similarity between Japan, Korea, and possibly Thailand regarding hair type distribution (types I, II, and III). More important, it provides (some) evidence that Chinese hair is different from other Asian hairs. The same observation applies to African people, because clear shifts in hair type distribution are seen between hairs from South Africans and Ghanaians. Such a preliminary observation should obviously be confirmed by larger studies on African populations. A great diversity is also observed in Europeans. It thus emerges from hair type classification that the three conventional human subgroups all appear to be heterogeneous. It vindicates the idea that the earlier ethnic distinction between the three racial subgroups (Asian, African, European) is oversimplified and less and less appropriate, especially as far as hair is concerned.

Increasing mixed-race descent and mixed-origin populations have a clear impact on hair type distribution, as reflected by the results on hairs from Brazil and Chicago, the only locations where six hair types were found. Conversely, the more homogeneous frequencies found in Asian countries, mostly limited to three types, suggest a lower statistical weight of past admixture.

Despite the fact that we did not include populations from intermediate parts of the world (such as Northeast Africa, the Middle East, and West Asia) in this study, Figure 5 clearly shows that there is a continuum in the distribution of hair shape variation throughout the world according to a main curling intensity factor along axis 1. The eight hair types we identified look like stepping stones along this axis, except maybe for types VI and VII, which are differentiated through another axis. This view should be consolidated by including data on people from areas such as Northeast Africa, an area that has been both a historical path of first human colonization, some 150,000-200,000 years ago, and a mixing place for a long time, as proven by the highest genetic variability being found in Africa (Cavalli-Sforza et al 1994; Excoffier 2002).

Hair shape diversity probably epitomizes worldwide migration and settlements over time, involving both continuous intermixture and isolation of fringes of population. Itis likely to fit an overall isolation by geographic distance (Malécot 1948) model of variation, in which divergence of populations is a function of the distance between their spots of origin, following the general pattern demonstrated for gene frequency variation throughout the world (Morton et al. 1971; CavalliSforza et al. 1994; Sanchez-Mazas and Langaney 1988; Poloni et al. 1995; Malécot 1948; Langaney 1988). These conclusions are in close agreement with those of the last important meeting on the implication of origin in health and disease (Hill et al. 1999).

This study is a first step in establishing a reliable method to describe various hair types using morphological measurements. There are many other traits concerning hair form, such as diameter, medulla characteristics, crimp, and cuticular scale count (Hrdy 1973), that could have been studied. It is possible that further work incorporating these other hair form traits would lead to an even more complex classification, for example, one including subclasses.

Conclusion

This study has shown that it is possible to classify the various hairs found worldwide into eight main coherent hair types. The approach involves objective descriptors of hair shape and is more reliable than traditional methods, which rely on categories such as curly, wavy, and kinky (Coon 1962).

The method is simple to use and only needs the measurement of three easily accessible parameters: curve diameter (CD), curl index (i’), and number of waves (w). Even though the method has some limitations (the required hair length must be more than 6 cm (sex bias) and the handling of single hairs may be tricky in some cases (e.g., wave may be mistakenly counted), this method of assigning hair type has a number of advantages. Applied to worldwide human diversity, our method avoids referring to putative, unclear ethnic origin of subjects while providing reliable and relevant data. Briefly, a straight hair type I is a straight hair type I and whether it originates from a European or an Asian subject is not at issue. The hair type classification described here can be considered a meaningful and universal standard for defining hair in a similar way as phototype has been for skin (Fitzpatrick 1988). Skin phototype 3 behaves the same way (when exposed to sunlight) irrespective of ethnic origin.

Hair types defined here also more suitably reflect the hair shape diversity around the world and possibly help to trace past mixed origins among human subgroups. The general picture of hair shape variation throughout the world that we obtained in this study is in good accordance with the general picture of human biological variation as measured by genetic markers or by biometric methods (Langaney 1988). Hair shape is a quantitative, continuous trait, just as skin color intensity or stature is. It is not a discontinuous, discrete feature described by split categories. Its variation within and between populations is gradual along the continents, according to the prehistoric and historic spread of Homo sapiens since 100,000 years ago and to the subsequent history of migrations and admixture.

Even though our study was conducted on a limited number of volunteers, the findings perfectly parallel those obtained from hair growth data (Loussouarn 2001; Loussouarn, unpublished data, 2006). South Africans, East Asians, and Western Europeans represent the most extreme cases of continuous variation in which all intermediate cases can be observed, either in intermediate or mixed-origin populations. As a result, the same hair properties can be seen in people of different ancestry, whereas people of similar origin may be different depending on local polymorphism or genetic recombination.

Further work is required to develop an even finer classification of various hair types, using other hair form descriptors instead of, or in addition to, those used in the present study.

Acknowledgments This research involved many contributors from L’Oréal, but the list is too long to name each individual; we are very grateful to all of them for their commitment and dedication to the project. Among them we would like to mention V. Holloway Barbosa and S. Diridollou (L’Oréal Institute for Ethnic Hair and Skin Research, Chicago, Illinois), F Leroy and P. Barbarat (L’Oréal Recherche, Aulnay, France), F. Aghassian and K. Giering (L’Oréal Recherche, Asnières, France), and C. Collaudin and I. Lozano (L’Oréal Recherche, Clichy, France). Also we would like to acknowledge the support of L’Oréal Operational Facilities in countries where these studies were conducted, which allowed us to implement this worldwide approach. Finally, we would like to especially thank our proofreaders, C. Bouillon, C. Dubief, and J. Gawtrey (L’Oréal Recherche, Clichy), who helped us in putting together this paper in an appropriate format for publication. We also thank the editor and two anonymous reviewers for their valuable corrections and suggestions.

Received 3 May 2006; revision received 23 January 2007.

Literature Cited

Bailey, J., and S. Schliebe. 1985. The precision of average curvature measurement. In Human Hair: Proceedings of the Symposium on Forensic Hair Comparisons, Federal Bureau of Investigation, ed. Washington, DC: US Government Printing Office, 147-148.

Barbarat, P., C. Hadjur, R Fiat et al. 2005. From the molecular structure to the macroscopic shape of hair curl patterns. In Proceedings of the 11th International Wool Research Conference, University of Leeds, Leeds, UK (in press).

Benzécri, J. P. 1984. L’analyse des données. 1. La taxinomie. Paris: Dunod.

Bernard, B. A. 2003. Hair shape of curly hair. J. Am. Acad. Dermatol. 48:120-126.

Bouillon, C., and J. Wilkinson. 2005. The Science of Hair Care, 2nd ed. Boca Raton: CRC Press and Taylor and Francis.

Breiman, L., J. H. Friedman, R. A. Olshen et al. 1984. Classification and Regression Trees. Belmont, CA: Wadsworth.

Carvalho-Silva, D. R., F. R. Dantos, J. Rocha et al. 2001. The phyIogeography of Brazilian Y-chromosome lineages. Am. J. Hum. Genet. 68:281-286.

Cavalli-Sforza, L. L., P. Menozzi, and A. Piazza. 1994. The History and Geography of Human Genes. Princeton, NJ: Princeton University Press.

Coon, C. 1962. The Origin of Races. New York: Alfred Knopf. Quoted and criticized in A. Alland Jr., Human Diversity (New York: Columbia University Press, 1971).

Dekio, S., and J. Jidoi. 1990. Amounts of fibrous proteins and matrix substances in hairs of different races. J. Dermatol. 17:62-64.

Diday, E., G. Govaert, Y Lechevallier et al. 1981. Clustering in pattern recognition. In Digital Image Processing, J. C. Simon and R. Haralic, eds. Dordrecht, Netherlands: D. Reidel and Kluwer, 19-58.

Excoffier, L. 2002. Human demographic history: Refining the recent African origin model. Curr. Opin. Genet. Dev. 12:675-682.

Fitzpatrick, T. B. 1988. The validity and practicality of sun-re active skin types I through VI [editorial]. Arch. Dermatol. 124:869-871.

Food and Drug Administration. 2003. Guidance for Industry: Collection of Race and Ethnicity Data in Clinical Trials. Draft Guidance, January. Cited in Federal Register 68 (January 30, 2003), 4788.

Franbourg, A., P. Hallegot, F. Baltenneck et al. 2003. Current research on ethnic hair. J. Am. Acad. Dermatol. 48:115-119.

Goldstein, D. B., and L. Chikhi. 2002. Human migrations and population structure: What we know and why it matters. Annu. Rev. Genomics Hum. Genet. 3:129-152.

Gray, J. 2000. Human Hair Diversity. Oxford, UK: Blackwell Science.

Hill, A. V. S., A. Sanchez-Mazas, G. Barbujani et al. 1999. Human genetic variation and its impact on public health and medicine. In Evolution in Health and Disease, S. C. Stearns, ed. Oxford: Oxford University Press, 62-74.

Hrdy, D. 1973. Hair form variation in seven populations. Am. J. Phys. Anthropol. 39:7-18.

Kreplak, L., F. Briki, Y. Duvault et al. 2001. Profiling lipids across Caucasian and Afro-American hair transverse cuts, using synchrotron infrared micro spec trometry. Int. J. Cosmet. Sci. 23:369-374.

Langaney, A. 1988. Les Hommes. Paris: Belin.

Lebart, L., A. Morineau, J.-P. Fenelon et al. 1982. Traitement des données statistiques. Paris: Dunod Méthodes et Programmes.

Lebart, L., A. Morineau, and M. Piron. 1995. Statistique exploratoire multidimensionnelle. Paris: Dunod.

Lindelöf, B., B. Forslind, M. A. Hedblad et al. 1988. Human hair form: Morphology revealed by light and scanning electron microscopy and computer aided three-dimensional reconstruction. Arch. Dermatol. 124:1359-1363.

Loussouarn, G. 2001. African hair growth parameters. Br. J. Dermatol. 145:294-297.

Malécot, G. 1948. Les mathématiques de l’hérédité. Paris: Masson.

McMichael, A. J. 2003. Ethnie hair update: Past and present. J. Am. Acad. Dermatol. 48:S127-S133.

Morton, N. E., D. E. Harris, and R. Lew. 1971. Bioassay of skinship. Theor. Pop. BioL 2:507-524.

Ogle, R. R., J. R. Michelle, and J. Fox. 1999. Atlas of Human Hair Microscopic Characteristics. Boca Raton, FL: CRC Press.

Poloni, E. S., L. Excoffier, J. L. Mountain et al. 1995. Nuclear DNA polymorphism in a Mandenka population from Senegal: Comparison with eight other human populations. Ann. Hum. Genet. 59:43-61.

Robbins, C. R. 2002. Chemical and Physical Behavior of Human Hair. New York: Springer.

Rosenberg, N. A., J. K. Pritchard, J. L. Weber et al. 2002. Genetic structure of human populations. Science 298:2381-2385.

Sanchez-Mazas, A., and A. Langaney. 1988. Common genetic pools between human populations. Hwm. Genet. 78:161-166.

Thibaut, S., O. Gaillard, P. Bouhanna et al. 2005. Human hair shape is programmed from the bulb. Br. J. Dermatol. 152:632-638.

Trotter, M. 1938. A review of the classification of hair. Am. J. Phys. Anthropol. 24(1):105-126.

ROLAND DE LA METTRIE,1 DIDIER SAINT-LÉGER,1 GENEVIÈVE LOUSSOUARN,1 ANNELISE GARCEL,2 CRYSTAL PORTER,3 AND ANDRÉ LANGANEY4

1 L’Oreal Recherche, Clichy, France.

2 L’Oreal Recherche, Asnieres, France.

3 L’Oreal Institute for Ethnic Hair and Skin Research, Chicago, IL.

4 Museum National d’Histoire Naturelle, Paris, France, and Département d’Anthropologie, Université de Genève, Switzerland

Human Biology, June 2007, v. 79, no. 3, pp. 265-281.

Copyright © 2007 Wayne State University Press, Detroit, Michigan 48201-1309

Copyright Wayne State University Press Jun 2007

Provided by ProQuest Information and Learning Company. All rights Reserved