This article argues for the necessity of phonetic analysis in signed language linguistics and presents a case study of one analytical system being used in a preliminary attempt to identify natural classes and investigate variation in ASL handshapes. Robbin Battison (1978) first described what is now a widely accepted list of basic handshapes, including those involved in the ASL signs for the letters A, B, C, S, and O and the numbers 1 and 5. These handshapes are said to be theoretically similar to the more common (and thus more basic) phonemes of spoken languages, and an equally wide cross-linguistic distribution is expected. However, without a subhandshape level of description and analysis, precise differences and similarities among and between handshapes are impossible to examine. Robert E. Johnson and Scott K. Liddell (forthcoming) provide a componential, feature-based transcription system in which groups of symbols describe phonetic elements of the handshape (such as finger selection and joint flexion/extension), allowing the identification of natural classes and potential phonemes. This study compares standard adult forms of signs with the forms produced by a two-year-old child and focuses on the phonetic features of hand configuration. Child-production errors have been noted to often involve the substitution of less marked hand configurations for more highly marked ones (see Siedlecki and Bonvillian 1993; Takkinen 2003); I suggest that the specific handshapes involved in the substitutions may be motivated by features shared between the target and the produced forms and that this phenomenon is apparent only with a phonetic feature-level (as opposed to the more usual phonemic category-level) transcription of the data. (Contains 18 figures and 1 footnote.)