A measure may discriminate between group but it should discriminate only in terms of a relevant aspect of health and not to some other characteristic. Accordingly, many validation reports reviewed in this book consider whether an item or a measure as a whole records the health characteristic consistently in different group. For example, questions in a depression scale are sometimes worded in an colloquial manner("I feel downhearted and blue”) that may not have equivalent meaning in different cultures or age groups and thus may be difficult to translate into other languages. Ideally, responses to an item should not vary for different types of people with the same level of depression. The theme of differential item functioning, or DIF, refers to n which sources of variation other Situations than the trait being measured influence the response to the item, this therefore reduces the validity of the item as a measure of health. Even though people of different ages(for example) may be expected to respond differently to questions on their health, this should only occur because of actual differences in health that occur with age, and not because people of different ages use language differently. Similarly, men might tend to underreport mental distress if an tem were phrased in a manner that implies that agreeing to it implied weakness. DIF can be illustrated graphically(150, Figure 1). The various ways to identify DIF fall under the heading of tem analysis techniques; they may use variants a regression approach, a factor analytic approach(150), or item response theory models(151; 152). People in contrasting groups (e.g., those interviewed with versions of a questionnaire in different languages) are matched in terms of their overall score, and their responses to each item compared; this forms a variant of item total correlations. Groups for which DIF is typically tested during item analysis include gender, age, education, ethnicity, and language.