ABSTRACT
In the preceding chapters of this book, we have discussed a number of topics that are of relevance to the process of developing measuring instruments in the behavioral and social sciences. In particular, we have spent a considerable amount of time on a major problem of measurement in these and cognate disciplines-the presence of measurement error when evaluating latent constructs that represent entities of major interest. In earlier chapters we were mainly concerned, however, with random error of measurement as well as its effect on observed scores, and we discussed relatively briefly the issue of systematic error. What we have not addressed yet in the needed detail is what may well be viewed as the bottom line of measurement, viz., whether an instrument is indeed measuring what it purports to evaluate, that is, the construct of actual interest. In this sense, we have not posed so far the question of how good the recorded scores with an instrument under consideration are in this particular respect. This issue is of paramount importance for measurement in the social and behavioral sciences, as well as outside them, and is the subject of the present chapter.
