When is evaluation a scientific activity,when is it not?

Because of the nature of social work, evaluating its impact on users presents a range of definitional and design problems calling into question the scientific basis of its evaluation. Competing paradigms, claiming to be the only seekers of the truth, may not always suit the problem under investigation or may provide limited answers to questions of outcome. Concepts such as truth, objectivity and certainty are elusive even in the physical sciences and are more so when it comes to human interactions with which social work is concerned. With the limitations of each of the established paradigms in mind, the paper puts forward the idea of a composite/pluralist approach. Such an approach can produce relatively reliable and valid results by combining statistical and qualitative methods, including the use of control and comparative samples and of pre- and pro-tests.