To try to make visible what is latent: Coming back to the foundations of psychometrics 

 

Most of the psychological variables used in basic and applied research are latent and a strongly consensual theoretical definition is needed in order to obtain reliable and valid evidence of the measures of such constructs. Nothing new under the sun, but the problem is that the patterns of development and use of psychometric instruments by psychological research show important shortcomings, especially related to the definition of constructs, to the quality of their operationalization, and to the application of a magical thought that believes that the theoretical weakness of a measure will be mitigated by the statistical sophistication of the validation procedures used. The main goals of this address are to present and discuss: a) the difficulties of making visible and tangible a psychological latent variable; b) the need to rethink psychometric validation designs based on the scientific method as expressed by Joseph Lee Cronbach; c) to banish magical thinking about the power of statistical-psychometric sophistication; d) to flee from automated test replications that are focused solely on reliability or latent structure; and in consequence, e) not to succumb to the uncontrolled growth of supposed constructs without sufficient theoretical foundation, both in Psychology and in other disciplines such as Medicine or Nursing, among others. In sum, this talk tries to come back to the foundations of Psychometrics, searching for the optimal ways to obtain high quality evidence, strongly related to well-founded psychological theories. In fact, to validate a test helps to validate the underlying theory, but only when the accuracy and scientific quality of the procedure used is really achieved.