The categorical consequences and psychometric properties of different sociometric classification methods were evaluated. Children aged 9 to 12 years (N=254) completed three sociometric questionnaires and a peer assessment measure on two occasions 5 weeks apart. The sociometric data were analysed using 13 different methods. Analysis of kappa values indicated relatively poor agreement across methods on subject classification. Temporal stability of the classifications was also poor. Assessment of construct validity involved analysis of the peer assessment items, using MANOVA to test hypotheses based on ideas from social exchange theory. Cross-sex rating biases and difficulties with the neglected and controversial classifications are discussed as indicating a need for the application of theoretically based approaches which consider features of the peer group social system and a need for caution in selecting methods for clinical use.