The current paper describes the development and validation of the first instrument for measuring the level of the public's personal and perceived stigma for Generalised Anxiety Disorder. The resulting GASS-Personal and Perceived subscales were shown to have adequate internal consistency, 4-month test-retest reliability and construct validity.
Convergent validity was demonstrated by moderate or high correlations between: (1) the GASS-Personal scale and other measures designed to assess personal stigma or proxy discrimination including the DSS-Personal and the Social Distance Scales; (2) the GASS-Perceived stigma scale and other measures designed to assess perceived stigma including the DSS-Perceived subscale and the Devaluation-Discrimination Scale; and (3) the GASS-Personal subscale and level of contact and past history of GAD. Divergent validity was demonstrated by zero or very small correlations between: (1) the GASS-Personal measure and measures of perceived stigma including the GASS-Perceived and the Devaluation-Discrimination Scale; (2) the GASS-Perceived measure and measures of personal stigma including Social Distance and the DSS-Personal scales; and (3) the GASS-Perceived and level of contact and past history of GAD. The above findings suggest that the GASS measure may be a suitable tool for community studies of the stigma associated with Generalised Anxiety Disorder including studies of its prevalence, predictors and the interventions for reducing it.
The stability of each subscale of the GASS was demonstrated by moderately high levels of test-retest reliability and stable scores over 4 months. Evidence of such reliability is lacking for many measures of stigma or in cases where it has been measured it has been assessed over shorter periods. For example, Corrigan and his colleagues measured test-retest reliability of the Psychiatric Disability Attributions Questionnaire (PDAQ) over one day  and King and his collaborators measured reliability over a period of 2 weeks .
The percentage of participants reporting that they personally agreed with negative statements about people with GAD was substantially lower than the percentage who believed that most other people in the community would endorse stigmatising attitudes to GAD. In this respect the findings strongly resemble those previously reported by Griffiths and her collaborators for depression [12, 33, 34].
The relatively low level of personal stigma reported by respondents for most items is encouraging although the extent to which these findings were influenced by social desirability biases and the low response rate is unclear (see Limitations below). It is of interest that on average a greater percentage of people exhibited discriminatory responses to GAD on the Social Distance scale than endorsed stigmatising statements on the GASS. Thus 14.4% of respondents were definitely or probably unwilling to socialise with a person with GAD, and 14.4% were unwilling to make friends, 23.2% to move next door, 23.7% to work closely and 36.1% to have someone with GAD marry into the family.
It is unclear why there is a disparity in the prevalence of respondents endorsing negative views on the GASS-Personal subscale items and the GAD Social distance items. It is typically hypothesised that stigmatising attitudes underpin discriminatory behaviour [eg., ]. Why then are the greatest levels of proxy discriminatory responses (unwillingness to have a person with GAD marry into the family 36%) over double that of the most highly endorsed anxiety stigma item (unstable - 16.7%)? There are several possible explanations for the observed pattern of findings. One is that the items employed in the Personal subscale of the GASS do not tap the most important elements of stigma associated with GAD. The items were derived from a qualitative analysis of the text on websites identified using a public search engine. Most of this text was written by mental health stakeholders rather than by members of the public who held negative views about mental disorder. Thus, the identified sites may have more strongly represented the domain of perceived stigma than personal stigma. A second possibility is that social distance and personal stigma are underpinned by different factors. Third, perhaps some people hold non-stigmatising attitudes about individual facets of GAD (e.g., that people with GAD are not to blame, not lazy and cannot simply snap out of it) but hold the view that active interaction with a person with GAD will generate a type or level of burden (e.g., a need for emotional support, proxy stigma) that they would prefer to avoid. A final possibility is that social distance measures are more resistant to social desirability bias.
The primary limitation of the current study is the low response rate to the questionnaire. Although the surveys were sent to a randomly selected sample of members of the public the respondents were not a representative sample of the community. Notably the sample comprised more women than men and the respondents showed somewhat greater levels of distress than has previously been documented in the Australian population . In addition, the participants in the retest subsample had significantly lower levels of personal anxiety stigma and a higher level of perceived anxiety stigma and education than those who were not included in this substudy (p < 0.001). The lack of representativeness of the overall sample precludes the use of the current statistics as an indicator of the prevalence of stigma associated with Generalised Anxiety Disorder. However, it does not detract from the demonstration that the GASS has adequate psychometric properties. If anything, the greater homogeneity of the sample employed for documenting the test-retest reliability of the GASS may have led to an underestimate of the true reliability. Additional research is required to further investigate the psychometric properties of the GASS including a confirmatory factor analysis of data collected from a new sample, further investigation of the scale in different populations and a demonstration of the sensitivity of the scales to change.
Another potential limitation is that self-report measures of personal stigma may underestimate the level of an individual's personal stigma and that the disparity in GASS personal and perceived stigma is a reflection of social desirability bias rather than two stigma factors. Link and his collaborators  have postulated that such bias may operate in self-reports of attitudes to mental illness. This possibility cannot be excluded for the GASS or any other stigma scale based on self-report. However, we can be confident that when used in a representative sample the GASS will provide a reliable indication of the minimum prevalence of anxiety stigma in the community. Secondly, previous research has demonstrated a significant association between attitudinal and physical proximity measures of stigma . Further research using behavioural, physiological, or implicit measures of anxiety stigma may shed further light on the validity of the GASS measure.