Textile fibers are a key form of trace evidence, and the ability to reliably associate or discriminate them is crucial for forensic scientists worldwide. While microscopic and instrumental analysis can be used to determine the composition of the fiber itself, additional specificity is gained by examining fiber color. This is particularly important when the bulk composition of the fiber is relatively uninformative, as it is with cotton, wool, or other natural fibers. Such analyses pose several problems, including extremely small sample sizes, the desire for nondestructive techniques, and the vast complexity of modern dye compositions. This review will focus on more recent methods for comparing fiber color by using chromatography, spectroscopy, and mass spectrometry. The increasing use of multivariate statistics and other data analysis techniques for the differentiation of spectra from dyed fibers will also be discussed.

Clear coats have been a staple in automobile paints for almost thirty years and are of forensic interest when comparing transferred and native paints. However, the ultraviolet (UV) absorbers in these paint layers are not typically characterized using UV microspectrophotometry, nor are the results studied using multivariate statistical methods. In this study, measurements were carried out by UV microspectrophotometry on 71 samples from American and Australian automobiles, with subsequent chemometric analysis of the absorbance spectra. Sample preparation proved to be vital in obtaining accurate absorbance spectra and a method involving peeling the clear coat layer and not using a mounting medium was preferred. Agglomerative hierarchical clustering indicated three main groups of spectra, corresponding to spectra with one, two, and three maxima. Principal components analysis confirmed this clustering and the factor loadings indicated that a substantial proportion of the variance in the data set originated from specific spectral regions (230-265 nm, 275-285 nm, and 300-370 nm). The three classes were well differentiated using discriminant analysis, where the cross-validation accuracy was 91.6% and the external validation accuracy was 81.1%. However, results showed no correlation between the make, model, and year of the automobiles.