Page 63 - PhD GT

upon observation. With reference to opiate users, Maisto et al. (1990) raise the question of what a collateral source can know about an individual’s drug use.
On the question of methods of validating the self-report of alcoholics, Watson et al. (1984) suggest that when self-report has been compared to collateral report the assumption that, of the two, collateral report is the more accurate may be a false one to make. Collaterals, whether they are family members, friends or colleagues may well overestimate the extent of drinking for many reasons, for example the fear that their report will not be taken sufficiently seriously, or they may simply be making guesses. They found that 74% of alcoholics’ self-report did coincide with collaterals’ reports, but that in three quarters of the remaining cases, self-descriptions underated collateral reports. The systematic way in which the reporting varied in those cases where there was no concurrence of reports led the authors to the conclusion that alcoholics framed their drinking in a more favourable light than collaterals. They do, however, admit that they used the collateral ratings of consumption as their criterion without having any validation data for so doing. Moreover, they found that the differences in reporting varied according to whether high consumption or low consumption and abstinence were being reported. Self-report and collateral report were more consistent in cases of the latter. Their conclusion that the lack of concurrence between self-report and collateral report is sufficient ground to doubt the utility of self-report in follow up studies rather flies in the face of the critique they presented of their own work. They have though highlighted some important methodological questions in the validation of self-report.
Midanik (1988), in her review and assessment of the literature on the validity of self-reported alcohol use, criticises the way that correlations between two measures or two reports have been used as measures of validity, assuming that one of the measures is the criterion measure when it may well not have been validated as such. The example she gives in alcohol research is the commonly held assumption that the source reporting higher alcohol consumption is necessarily the more accurate one where validation of the criterion measure may not have been reported. She then identifies those factors which either enhance or diminish accuracy and makes the case for ensuring they are addressed with the aim of maximising the validity of self-report rather than settling a dispute about whether or not it can be accurate.
In an earlier review of the literature on agreement between self and collateral report, Midanik found no consistent direction of error in reporting (Midanik 1982). She cited the work of Polich (1982) who found disagreements between self and collateral report to be in both directions with “little or no net bias in self-reports” (p. 124). He concluded that, in view of the many instances of discrepant reports, random errors may attenuate correlations. Midanik reported the finding that
51