Page 60 - PhD GT

recent events or the perceived stigma attached to the condition or its consequences, the actual or potential illicit nature of the behaviour or its consequences are sources of social desirability bias in self-report (Oppenheim 1992). The setting and the type of information sought may compound these sources of potential bias. In the clinical setting, a similar source of bias may stem from the perceived utility of giving a particular report. Thus for heroin, over-report may be accounted for by fear of the medical practitioner underestimating the severity of withdrawal distress and prescribing insufficient doses of a drug to relieve this, while for alcohol, under-reporting may be due to reluctance to pursue an abstinence goal of treatment. On the other hand, Maisto et al. (1990) point out that, depending on the situation in which self-report is requested, patients in treatment may not wish to disclose heroin use at all for fear of being removed from the treatment programme. The same principle would hold for patients in abstinence oriented treatment of alcohol problems. Such bias may be relevant at the follow-up and would not operate at intake in a clinical sample.
3.2.3 Other sources of bias
Embree and Whitehead (1993) point out two further sources of bias in the reliability of self-report data: random error in data collection and recording and misinterpretation of interviewer questions or respondents’ replies. Babor et al. (1987b) classify the factors affecting the validity of self-report as either task variables or respondent variables. Task variables refer to the situation or the procedures used to elicit the information. Situational factors include the manner of the interviewer, the extent to which the nature and purpose of the task are explained, confidentiality and the belief that anonymity will be protected, the question of whether the purpose of information gathering is to decide a suitable treatment or to assess outcome. Procedural factors are the duration and complexity of the task including the criterion interval and the conceptual level of the questions, the clarity of the instructions and the sensitivity of the questions. Respondent variables include the ability to understand, modified by language and conceptual ability, drug or alcohol related cognitive impairment, state of intoxication, anxiety and general physical condition, self esteem and perceived social desirability of responses.
Hesselbrock et al. (1983) refer to the demand characteristics of the situation as being important in enhancing or reducing the accuracy of self-report. In a study of the validity of self-report measures, they were able to demonstrate that in the hospital situation, using volunteer patients in a research setting where a “scientific aura” (p. 607) was established, where patients had been admitted because of their alcohol problems, the giving of an inaccurate report would be neither functional nor credible, unlike in the home, employment or similar situation.
48