COLORIMETRIC ANALYSIS O F H A I R COLOR B. B. GARDNER Hurvard Oniversit y AND D. L. M A c A D S M Massachusetts Institute of Technology SIX FIGURES The following report is the result of a study of a series of fifty-three hair samples for the purpose of determining the colorimetric characteristics of hair, in the hope that such an analysis might point the way to a metrical classification capable of statistical manipulations. The work was done under the auspices of Dr. E. A. Hooton, of the Division of Anthropology, H a r v a r d University, and was carried out in the Colorimetric Laboratory of the Massachusetts Institute of Technology, whose equipment was placed at our disposal by Prof. A. C. Hardy. Spectrophotometric curves of the samples were secured by use of the recording spectrophotometer designed by Professor Hardy, and built under his direction at the Massachusetts Institute of Technology. This apparatus was described before the Optical Society of America, meeting in New York, February 24, 1934. A full description of the optical and mechanical details will be published in the near future. This recording photoelectric apparatus f o r measuring the spectral reflectance of all kinds of samples is a n improvement over the spectrophotometer described by Professor Hardy ( '29). The present system consists of a double monochromator, which is followed by a photometer unit consisting of two Rochon prisms and a Wollaston prism. The Wollaston prism sepa187 A X E R I C A N J O U R N A L OF PHYSICAL AXTHROPOLOQY, VOL. XIX, N O . JULY-SEPTEMBER, 1934 2 188 B. B. GARDNER A N D D. L. MAC ADAM rates the beam leaving the monochromator into two beams, one of which falls on a standard white magnesium oxide surface and the otner on the sample. The beams are made to flicker alternately by passing them through a rapidly rotating Rochon prism which, by its property of transmitting only light polarized in a plane fixed to it, alternately extinguishes the beams from the Wollaston, which are polarized in planes perpendicular to each other. The light reflected from the standard of reflectance and the sample is gathered by an integrating sphere and acts on a photoelectric cell. The current from the cell is amplified and acts on the field of a n alternating current motor, which, through a cam, rotates the Rochon placed ahead of the Wollaston prism in such a manner as to alter the relative intensities of the beams falling on the sample and standard until the amounts of light reflected from these a r e equal. I t is evident that when this condition is satisfied the reflectance of the sample relative to the standard is the ratio of the intensities of the beams. This ratio is a function of the angle of rotation of the cam controlled Rochon ; therefore, a shaft, driven by a balancing motor, can move a recording pen along a per cent reflectance axis, while another motor, which continuously changes the wavelength of the light used f or the measurement, causes a table containing the record paper to move along a wavelength axis perpendicular in direction to the reflectance axis. Since both beams traverse the same optical parts, the effect of selective absorption or selective reflection is completely eliminated; and since only one photoelectric cell is used, and that one only for detecting inequality in reflected energy coming alternately from sample and standard at a high rate of alternation, the readings a r e not affected by variations in the light source, the photoelectric cell, or in the associated amplifier. A series of fifty-three hair samples were collected which ranged in color from black to very blond and included a wide selection of reds. It was impossible a t the time to obtain a series including all possible shades, but special efforts were made to include the extreme types as well as the intermediate 189 COLORIMETRIC ANALYSIS OF HAIR COLOR ones. The result was a series containing a few very blonds, several ash blonds, a large number of light and dark browns, and a great many reds and reddish browns. A few samples of bleached hair were taken to determine the effect of bleaching on the color. A colorimetric analysis was then made for each sample and the curves carefully studied. The most striking characteristic of all curves is their uaiformity and simplicity. Figure 1 shows six of the curves for samples which a r e classed a s black, dark red, red, reddish brown, light brown, and very blond, respectively. It is seen that all the curves a r e quite simple; the black and blonds are - VfOLET- - -8 L UE GREEN -- Y - -0- - R€D L 30. $? 25. w u 5 Z 20. ‘5. w -I 10. LL w CL 5 . 400 500 600 WAVELENGTH - mp 700 Figure 1 almost straight lines while the reds show a decided upward curvature in the region of the green. I t is also evident that differences in the color a r e due rather to changes in the reflectance in the red end of the spectrum than in the violet end, since the total range of variation is 5.5 per cent a t 400 mp and 31 per cent a t 7OOmp. Figure 2 shows very strikingly the differences between the curves for red hair and those for brown hair. Both samples show a reflectance of 9 per cent at 700mp and only a slight difference a t 620 mp which is the limit of the red band. F rom there on they diverge until, at 540mp they reach a maximum divergence of 1.3 per cent, which is maintained until they reach 190 B. B. G A R D N E R A N D D. L. M A C ADAiM the violet band where they draw together again. I n spite of the slight difference in the curves, the colors a r e definitely unlike in appearance; no. 23 is a dark red, while no. 28 is a medium brown. It is evident that the distribution between the color of red a n d brown hair is due not so much to the disparity in reflectance in the red bands of the spectrum as to a difference in the blue and green bands. y w Lu U 600 500 400 700 WAVELENGTH - m y FSgure 2 35 30 F u W 15 -I 2 - 10 5 I 400 I , 500 600 WAVELENGTH - my 700 Figure 3 Figure 3 presents the curves f o r a series of samples ranging from dark brown (no. 5 ) to very blond (no. 43). When graded by eye, the samples fall in the same order as that shown by the curves, with no. 14 a light brown and no. 13 blond. It is apparent that each of the browns and blonds gives a practically straight line which differs from the other lines either in slope or in total reflectance. For instance, the distinction COLORIMETRIC AWALYSIS OF H A I R COLOR 191 between nos. 14 and 13 is purely a difference in slope of the curves, while the difference between nos. 7 and 14 is principally one of total reflectance, since the curves a r e parallel. The blonds show a greater reflectance for all the wavelengths than do either the browns or reds ; the lightest blond studied had 15 per cent greater reflectance at 700mp than did the brightest red. Comparison of all the curves studied shows that they range from almost straight lines to those with a definite curvature. Actually, none of them are perfectly straight lines I 400 600 500 WAVELENGTH 70 0 -my Figure 4 as is shown by figure 4 where they have been redrawn to a n enlarged scale. No. 2 is a dark red, no. 38 a dark brown, and no. 37 is a medium brown ; and although the two browns show a, slight curvature, the a r e still sharply distinguishable from the red. A study of different types of curves show that each curve could be expressed approximately by taking two values, one for the point of maximum reflectance, 700 mp, and another for the point of maximum curvature which lies between 540mp and 550mp for all curves. Because 546mp may be used as a measure of brightness of the color and may be conveniently secured by use of a mercury arc and filters, it was arbitrarily selected as the point of maximum curvature, and 192 B. B. G l R D N E R A N D D. L. -MACADAM the height of each curve was taken a t this point and at 700 mp. These figures were then plotted on a scattergram, as shown in figure 5, where they grouped themselves with the darker colors in the lower left-hand corner. The samples were then carefully sorted by eye into groups according to color, and these groupings, marked off on the scattergram, resulted in the eleven groupings labeled A to K. While the samples in any’group are, in most cases, distinguishable from each other, they do form distinct series with a gradation of shades ; and each series is distinguishable from the groups on eit,her side. F o r example, in group R no. 39 is darker than no. 37, but together with-nos. 15 and 22 they form a series which, a s a whole, is distinguishable from either group A or group C. Also, no. 39, although only slightly lighter in shade than no. 41, is still closer to no. 15 than to 41 and fits better in group B than in group A. A comparison of all the groups shows that they too fall into three graded series. Group A to F form one series ranging from black t o d a r k brown in group A, through medium browns in groups B a n d C, to light browns in D and E, and blond in group F. This, as a series, was easily separated from a second series formed by groups G and H, which a re reddish brown, that is, brown with a definite reddish tinge. The third series is formed by groups I, J, and K which are all true reds with a range of dark, medium, and light. After determining the various groups and series i t is possible to draw lines OX and OY which seem to delimit the three series. Line OX, with a slope of 0.47, is the upper limit of reddish shades. That means that when hair reflects less than 47 per cent as much light a t 546 mp wavelength as it does a t 700 mp wavelength, the eye distinguishes some of the red, but when the ratio of the reflectance a t the two wavelengths is greater than 47 per cent the resulting impression is brown. Similarly the line OY, with a slope of 0.40, separates the reds from the reddish browns. The separation of these two series is quite sharp as is well shown by nos. 16 and 54. Though these two samples vary only a t 0.6 per cent a t 546 mil and are together a t 700 mp, C lu REFLECTANCE- AT- 546 P m m 0 lu & I I I I I I I mp - R, - m 0) I 1 0 x 1 N P ru P a P 4 s N m c 0 a\" I 1 I I I I I I I c='* Y * .!& I COLORIMETRIC ANALYSIS OF HAJX COLOR 197 nevertheless no. 16 is a reddish brown and no. 54 is a bright red. Of course, these lines are purely tentative and can be definitely established only through study of a larger number of samples. Actually, the lines may not be straight or may vary slightly from the slopes given here. Study of sufficient samples may, also, reveal the fact that there are no separate groups in each series, but that there is, rather, a gradual transition from one extreme to the other. A calculation of the complete colorimetric specifications of the color of twenty of the hair samples was carried out following a recent international standardization of the method of objective color specification (Judd, '33). This is a method by which a specification of the appearance of the color t o a normal observer may be calculated from the reflectance values measured at every visible wavelength. The end result of this computation from spectrophotometric data is a set of three quantities. The first of these is the brightness of the sample relative to the brightness of a perfect reflector. The second is the dominant wavelength which f o r convenience of visualization may be thought of as the wavelength of the spectrum color most closely resembling the sample color. The third is the purity, which indicates the degree of similarity between the color of the sample and the color of the pure spectrum light having its wavelength equal to the dominant wavelength of the sample. Thus colorless grays and whites have zero purity, browns and other colors closely resembling grays and whites have low purities up to perhaps 30 per cent. Colors closely resembling pure spectrum colors have high purity values u p to 100 per cent. The twenty samples studied included one or more from each of the groups ( A to K ) indicated on the scattergram, figure 5. The computation was performed on the assumption that the I.C.I. standard illuminant C, equivalent to average daylight, was the source of illumination at the time of visual examination. Table 1 exhibits several interesting correlations between the visual classifications and the colorimetric specifications. It is immediately noticed that the brightness, comAMERICAN JOURNAL OF P H Y S I C 1 1 . A N T H R O P O L O Q Y . VOI,. IIX, NO. 2 198 B. B. GBRDNER A N D D. L. LMACADAM pared to a magnesium oxide surface in the same illumination, is higher i n almost every case than the reflectance a t 546 mp. In most cases this reflectance is about 1 per cent lower than the computed relative brightness. Class A seems to include all samples having less than 3 per cent relative brightness, TABLE 1 Colorimetric Ypecifications f o r twenty hair s a m p l e r - ~ _ _ _ _ _ _ I . ~ ~~ 1 GROWP -. -~ -. percent A 25 B B 32 40 C 35 D 14 E E E 35 45 13 48 E 580 I j 580 580 580 583 F F 36 44 43 580 580 581 G 46 583 H 47 583 F I I I i J J K j I I I I i I I i 3.8 4.4 5.5 9.1 10.9 11.4 7.3 41 42 43 17.6 15.4 20.8 588 593 587 i 33 11 585 585 j I 57 58 51 585 I 60 I I 52 50 55 - - 2.2 42 32 42 45 2 31 32 I I per cent 7.8 55 , I .~ 31 45 I 1 4.7 7.3 3.2 3.6 4.4 I ! i 5.4 5.4 8.6 with purity of less than 30 per cent. Classes B to F, inclusive, are marked by possessing a common dominant wavelength ranging from 580 mp to 583 mp, while they show uniformly increasing relative brightness and colorimetric purity, Groups G and H a r e characterized by a dominant wavelength of 583 COLORIMETRIC ANALYSIS OF HAIR COLOR 199 mp, almost inappreciably different from the wavelength range of groups B and F. The groups G and H differ most markedly from the former groups in having high values of purity associated with moderately low values of relative brightness. Group H differs from group G in having higher brightness and purity. Groups I, J, and K are characterized by having high purity, 50 per cent to 60 per cent, together with low relative brightness, 3.2 per cent to 8.6 per cent, and a longer dominant wavelength, 585 mp to 593 mp. It is especially interesting to note that the dominant wavelengths cover a very narrow range in spite of the wide differences in visual colors, and also that the range coincides almost exactly with the yellow band of the spectrum. 5 20. 2" I5 Lu a t- u lo' i 5. LLI W - 0: 400 500 600 WAVELENGTH - my 700 3igure 6 To determine the effects of bleaching, two samples, taken from the original sample no. 11,were bleached with peroxide. The results were lighter reds than the original and showed typical red curves as seen in figure 6, in which no. 11 is the original and nos. 51 and 52 are the bleached samples. On the scattergram the effect of bleaching raised the color from the original group J to the higher group K, but did not change it to another series. As shown in table 1, bleaching did not change the dominant wavelength, but increased both purity and brightness. Another sample, no. 47, was bleached before it was received and its original color is not known. Although the number of samples studied is too few to justify any definite conclusions, the study does suggest certain interesting probabilities. For an exact metric classification 200 B. B. G A R D N E R A N D D. L. M A C A D A M of hair color, it would, apparently, be necessary to measure only the reflectance of the samples at the wavelengths 546 mu and 700 mp. These two values should be sufficient to identify the color and to distinguish it from all others, and would a t the same time be susceptible to statistical analysis. F o r a descriptive classification there would be three main groups (red, reddish brown, and brown) and a number of subgroups corresponding to those in figure 5 or even based on arbitrary divisions of the main groups. Study of a larger number of samples would make i t possible to determine whether the subgroups shown a r e normal or whether there is merely a gradual transition from one end of the series to the other. If the groups a r e normal, a color chart could be made up with a color corresponding to that of the midpoint of each subgroup in each of the three series. Each of these colors could be given the numerical values a n d colorimetric specifications of an average of the group, and thus, a sample could be given numerical value merely by comparison to the chart. If further study reveals, on the other hand, that there a r e no subdivisions, but that each of the main groups shows, rather, a gradual transition, then a color chart could be devised for arbitrary intervals within the groups, and the colors thus selected could be given numerical values accordingly. SU Y MA RY 1. Prof. A. C. Hardy’s recording spectrophotometer, used for the analysis of hair color in this study was described. 2. A series of fifty-three hair samples, ranging in color from black to very blond, and including a wide selection of reds, was subject to colorimetric analysis. 3. The most striking characteristic of all of the resulting curves is their uniformity and simplicity. The blacks and blonds a r e almost straight lines, while the reds show a decided upward curvature in the region of the green. 4. The curves show that the distinction between the color of red and brown hair is due not so much to the disparity in reflectance in the red bands of the spectrum as to a difference in the blue and green bands. COLOBIMETFXC ANALYSIS O F HAIR COLOR 201 5. Each of the browns and blonds gives a practically straight line which differs from the other lines either in slope or in total reflectance. 6. Each curve can be expressed approximately by taking two values, one f o r the point of maximum reflectance, 700 mp, and another for the point of maximum curvature arbitrarily selected a t 546 mp. 7. The hair samples were sorted by eye into color groups, and the colorimetric values of the individual samples plotted on a scattergram. I t was then possible to draw lines delimiting three principal series: blond, red, and brown. Each of these series showed in its component samples a n approximately continuous gradation. 8. Calculation of complete colorimetric specifications was carried out in twenty hair samples. The specifications of the several visually determined groups were discussed. 9. Eleaching of hair did not change the dominant wavelength, but increased both purity and brightness. 10. The samples studied are insufficient to justify definite conclusions, but the method points the way to a larger investigation which would result in a scientifically graded hair color series, from which the colorimetric specifications of matched samples could be read approximately. LITERATURE CITED HARDY, A . C. 1929 A recording photoelectric color analyzer. J.O.S.A., XVIII, 96. JVDD,D. €3. 1933 The 1931 I.C.I. standard observer and coijrdinate system for colorimetry. J.O.S.A., XXIII, 359.