Table 1 Classification of validity evidence

Convergent/concurrent validity – Convergent validity is usually used to refer to the extent to which theoretically related tests (e.g. tests of numerical and verbal intelligence) correlate. Concurrent validity refers to the extent to which two (or more) tests of the same construct (e.g. two verbal intelligence tests) correlate. Given the conceptual difficulties in defining empathy precisely, and the subtly different definitions used by different groups, distinguishing between convergent and concurrent validity was difficult and unlikely to be reliable. Therefore, convergent and concurrent validity were classed together.

Divergent validity – This is a measure of validity based on the principle that theoretically unrelated constructs should correlate poorly. For example, some groups proposed that empathy scores should not correlate with scores on tests of biomedical knowledge as they treated the two constructs as independent.

Formally assessed face validity – This is a measure of the extent to which the test appears to assess the construct of interest. Any formal process of assessing the extent to which a relevant group (e.g. patients, clinicians) recognised the test as measuring 'empathy' was considered in this class.

Factor analysis – Factor analysis may be treated as a means of assessing convergent and divergent validity within the test under consideration. It involves the statistical analysis of inter-item correlations in order to identify underlying structures within the test. This is usually in the form of subscales containing test items that correlate highly with one another but less well with other items in the test. However, factor analysis is open to accusations of 'results fixing' because of the number of statistical decisions to be made before a result is produced and may, without an agreed theoretical framework, be difficult to interpret. For this reason, factor analysis was treated as a separate validation tool.

Other validity – This class of validation methods was used to capture validity assessments that, while of potential relevance, were unusual or not intrinsically relevant. For example, some groups, on the basis of previous findings of an empathy differential between men and women, used a difference in empathy scores between men and women as a measure of criterion-related validity.