86 JOURNAL OF COSMETIC SCIENCE DEFINING HAIR GLOSS Surfaces and objects often show gloss, which is attributed to specular reflectance at the respective surface. Due to the diversity of surface structures, different kinds of gloss can be observed. Hunter, as early as 1937 (2), intensively investigated the phenomenon of gloss and arrived at a classification with six different types of gloss that were defined in terms of measurable parameters. Stamm et al. (5) reviewed the semantics of gloss, luster, shine, etc. for human hair. Measuring by goniophotometry light reflected from single hairs or hair strings, Bustard and Smith (15) and Stamm et al. (5), respectively, chose one of these classes, namely "contrast gloss," to represent the "luster" of hair. Their approach relates to the work by Nickerson on the gloss of cotton yarn and fabric (3 ), where she found this type of gloss well applicable to test textile luster in correspondence with the visual assessment. Hunter (2) defines contrast gloss as "contrast between specularly reflecting areas and other areas." It is measured by comparing the intensity of the light that is specularly reflected with the intensity of that which is diffusely scattered, so that: gt = s!d (1) gt is Hunter contrast gloss. s and dare the intensities of specular and diffuse reflection, respectively, measured at two positions, namely (a) at the expected receptor angle for specularly reflected light and (6) normal to the material surface. Following the notation introduced by Stamm et al. (5 ), lower case letters are employed to designate these so-called spot values. Since Equation 1 goes to infinity as d approaches zero, Nickerson suggested as an alternative: (2) where gt is the "Nickerson contrast gloss," considered to represent luster. For cotton, Nickerson (3) found a good correlation between contrast gloss values and visual esti­ mates of luster. Nickerson contrast gloss becomes zero when d equals s} unity when d equals zero, and negative when d s. In view of the complex structures of the GP curves for different types of hair (1), all three cases are likely to occur in practice. The approach applying spot values is valid, notably for fl.at surfaces, where the angle of incidence and the receptor angle for specularly reflected light are equal and where the intensity of the scattered light shows no angular dependence. However, both of these conditions are obviously not fulfilled for human hair, implying that spot measurements, as implemented in commercial testing devices (10,13,14) are expected to be at best of very limited use for determining hair gloss. In this view, it is not unexpected, that Stamm et al. (5), Bustard and Smith (15), and Guiolet et al. (17) found that the gloss parameters defined in equations 1 and 2, based on spot values, have high precision but poor sensitivity. Various approaches have been devised to determine integral values for specularly and diffusely reflected light (5, 11, 15), applying the principles inherent to equation 1 and especially equation 2 in order to derive parameters to describe hair luster. The definition of these parameters was either based on theoretical considerations (5, 15) or on empirical observations (11). Inherent to the approaches is the principle that diffuse reflection from hair occurs uniformly.

MEASURE OF HAIR LUSTER 87 Dismissing this assumption in view of the positions and shapes of the light intensity distributions underlying the GP curves (see Figure 2) and of our interpretation of Figure 1, we propose that the effect of hair luster and of the shine of a given hairstyle is related to the ratio of the total amount of light reflected specularly from single hairs to the overall amount of reflected light: GL = S/(D + S) (3) with (4) Following the notation of Stamm et al. (5 ), capital letters are employed to mark the integral values of light intensity. Gv the gloss index, is considered to describe "specular gloss," as a measure of luster, in correspondence to earlier definitions and subjective descriptions as "brilliance of specularly reflected light" (2) and "brilliance of highlights" (3). S and the components of D are determined from the complete goniophotometric curves for single human hairs through separating and integrating the intensities of the specu­ larly and diffusely reflected light components, respectively, by fitting three Gaussian distributions (see Figure 2). With the exception of light blonde or white hair, D i can usually be neglected, greatly simplifying the determination of the components of equa­ tion 3. The properties of equation 3 are plausibly connected with practical observations. If there is no specularly reflected light, S = 0 and hence GL = 0. In the case of D = 0, there is no diffusely reflected light, and perfect luster is achieved with GL = 1. If the situation arises that the integral intensities of specular and diffuse reflections are equal so that S = D} equation 3 yields 50% gloss. By darkening or lightening hair, the diffuse com­ ponent, due to changes in light absorption in the fiber, will decrease or increase, respectively . If the specular component remains unchanged, gloss will increase or de­ crease, respectively, by the coloring of the hair alone. This corresponds to the practical observation that dark hair always tends to appear more lustrous than blonde hair. RESULTS AND DISCUSSION By fitting Gaussian distributions to the GP curves, locations and widths for the three types of reflected light were determined for the different types of hair. The parameter values and their changes along the hair length, including cuticle angle, are presented and discussed in detail in Part I of this paper (1 ). By far, most GP curves could well be analyzed by considering just Sand D s} that is, by a two-component approach. Only for light blonde hair did D i as a further component of diffuse reflection have to be considered. Introducing the areas of the fitted peaks for a given GP curve into equation 3 yields the respective value for the gloss index G v All data were checked for outliers prior to further analysis by assessing them in so-called normal probability plots as implemented in the applied statistics software (21 ). In this type of plot, cumulative data frequencies follow a straight line when the data are normally distributed. On this basis, three obvious outliers were readily identified (GL 10%) in these plots for brown hair and removed prior to further analysis of the data.