The strengths and weaknesses for the assessment of personality characteristics of self-reports, reports by knowledgeable informants, and measures based on behavioral observation data are discussed and compared It is suggested that measurement methods should be evaluated with respect to Loevinger's concept of structural validity The concept of structural validity is expanded to include evidence that the scores have properties paralleling the theoretical definition of the characteristic with respect to reliability over occasions, range of referent attributes sampled, and generality across situations or the specification of situation parameters Each of the methods is evaluated for its potential to construct measures with structural validity for different models of constructs The methods are also discussed in relation to several other common threats to validity, specifically, threats stemming from the respondent, from the investigator, and from sampling errors It is likely that demonstrations of convergence between these methods will be possible for some but not all constructs It is also suggested that the examination of structural validity will contribute to the evolution and refinement of constructs appropriate for interactionist theories