MEASUREMENT OF HAIR LUSTER 299 ethcom s+nl n s+n2 s+n3 ' I ' I ' I ' I ' I ' I ' I ' 0,0 D• ,• ! ,1• ! ,15 • .0 2 .{• :• .D• Lllstr (DE/CIELAIS) Figure 5. Luster scale using color-difference values obtained for black hair tress. Sample B3. Treatments: ethcom (commercial ethanol/8 h), n (Neutrox © conditioner), s + nl (l•lseve © shampoo + Neutrox©/first wash), s + n2 (15,1seve © shampoo + Neutrox©/second wash), s + n3 (15-1seve © shampoo + Neutrox©/third wash). Results of ten measurements. D65 illuminant. 10 ø viewing angle. 5 x 10-mm viewing aperture. Internal reference. CIELAB system. Table VIII Values of Color Difference (DE) Obtained Using Internal Reference for CIELAB and FMCII Systems After the Described Treatments Treatments DE/FMCII DE*/CIELAB Commercial alcohol 2.49 + 0.8 0.50 + 0.3 zero-luster value Ethanol PA/8 h 1.48 + 0.9 0.81 _+ 0.4 zero-luster value Neutrox © conditioner 1.29 + 0.3 1.18 + 0.3 Shampoo + Neutrox©? 3.71 + 0.9 0.61 + 0.4 Shampoo + Neutrox©?? 4.16 + 0.8 2.77 + 0.4 Shampoo + Neutrox©??? 4.40 + 0.9 2.89 + 0.5 ?fw (first wash), ?? sw (second wash), and ??? tw (third wash). Operating conditions: horizontal sample position, D65 illuminant, 10 ø viewing angle, 5 x 10-mm viewing aperture, and adapted holder. Black hair sample B3. Averages and estimated standard deviation among a set of ten measurements. Internal reference: CIELAB (commercial alcohol: L* = 21.55, a* = 1.73, and b* = 1.67) and FMCII (X = 3.13, Y = 3.19, and Z = 3.08). CIELAB and from ethanol PA/8-h treatment (fourth reading) for FMCII. As observed, the choice of the reference produces significant differences in the results. However, in both cases, the data show the same profile. After the treatments with shampoo plus Neutrox ©, there was a consistent increase in DE values, which demonstrates a cumu- lative luster effect on hair. Analysis of variance (ANOVA) was performed in order to select the best reference (8,9). Table VII shows the results of statistical analysis, where DE,,• is the average color difference among a set of ten measurements for each sample, DE M is the average color difference among the samples, S2R is the variance within the sample, and S2T is the variance among the samples, always using the external reference. As observed, S2T • S2R in every case. This means that each sample is statistically different from the others. Therefore, it is not reasonable to calculate the parameters of color difference using an external reference. The color-difference calculations should be done, using as reference the sample that will be afterwards treated, in order to build the luster scale. Note that all the "samples" come from a single "tress." These results agree with those of Bustard and Smith, who used an internal reference to make sure that any changes between the

300 JOURNAL OF COSMETIC SCIENCE light-scattering properties of the reference and the treated hair were attributed only to the effects of the treatment (5). The cumulative effect of luster can also be observed in Table VII. Color-difference (DE) values increased after each treatment, especially after sequential shampoo-plus- conditioner treatments. Therefore, in order to build a scale, the luster should be satu- rated. Table VII shows the DE and DE* values obtained after sequential treatments in sample B3. As seen before, in both cases the data show the same profile. The treatments with shampoo plus Neutrox © produce an increase in DE and DE* values, which demonstrates a cumulative luster effect on hair. Figure 5 shows the luster scale built with Table VIII DE*/CIELAB data. The luster scale was built from 0 (zero luster value as internal reference) to 3 (maximum luster value) for the CIELAB system. The DE* estimated standard deviation was about 0.5, as calculated for each sample from a set of DE* values using the internal reference. In the FMCII system the luster scale ranged from 0 (zero luster value as internal reference) to 6 (maximum luster value). In this case, the DE estimated standard deviation was about 1.5. Visual luster observations of the treated samples agree with the DE and DE* data. CONCLUSIONS We have show that diffuse reflectance is a suitable technique for the measurement of hair luster. Being a spectrophotometric technique, it avoids the subjective factor that is associated with human perception. This can be an advantage in cases in which a quick check is needed. As expected, the cylindrical geometry of human hair hindered the use of the specular component as the main luster variable. The overall color difference is the best parameter both in CIELAB and in FMCII color equation systems. From this, the luster scale is not an absolute scale, since it depends on hair color characteristics. The luster scale should be built using as zero-luster the color values of the sample before applying the luster treatment. Diffuse reflectance equipment is relatively cheap and easy to handle. The luster scale has enough sensitivity to discriminate among treatments, although the estimated standard deviation of luster between measurements is somewhat high, in the range of 20%. This error range forces one to do a relatively high number of measurements on each sample, but since the measurements are taken just by moving the sample in the sample holder, this takes no more than one minute per replicate. ACKNOWLEDGMENTS The authors acknowledge the financial support of Inddstria e Com•rcio de Cosm•ticos Natura Ltda, Insddstrias Gessy Lever Ltda-Divis•o Elida Gibbs, and Conselho Nacional de Pesquisa e Desenvolvimento (CNPq).