Individuals aged > =18 years, drawn from an established market research panel available to Ipsos, were contacted via email and invited to complete an online survey between May 8–12, 2009. In the absence of response to the initial contact, individuals were contacted on one further occasion three days later. Recruitment continued until 1000 responses were obtained. To try to ensure the recruited sample was representative of the general adult population of France, sampling was stratified initially for place of residence (taking into account population density) then according to gender, age (in 5 years groupings) and socio-economic status (according to occupation of the head of household).

In order to ensure this was a brief, user friendly, internet questionnaire that was acceptable to a target population the survey instrument was limited to 21 items written in French. The final set of questions selected targeted key domains such as mental health knowledge (n = 8), attitudes (n = 5) and behaviours (n = 4) towards those with mental disorders and familiarity with mental disorders (n = 4). These items were designed to capture data on key themes examined in previously published questionnaires exploring knowledge, attitudes and beliefs 1 2 3 9 11 18 21 22 23 . However, the researchers did not use established questionnaires as (a) using a set of assessment tools for each issue we wished to target would extend the duration of participation to 30–50 minutes, which would lead to significant loss of participants on the internet and was also counter to the idea of a very brief questionnaire, (b) we especially wanted to explore views of disorders spanning childhood and adulthood and different types of presentations and none of the questionnaires or assessment tools available covered this range, (c) we wished to explore whether individuals understand the ‘terminology’ of mental disorders and what actually constitutes the disorder named eg schizophrenia (ie do people know not just the words or names used to describe mental disorders such as schizophrenia, but can they actually spontaneously describe the symptoms and problems that are integral to the disorder with that name), as such we designed some new questions to look at this, (d) we decided that to make this a simple user friendly survey the information on names and symptoms of disorders would be used as an alternative for case vignettes.

Knowledge items (n = 8) included the following- views of the likely prevalence of mental disorders in the general population, causal attributions (eg views of risk factors such as genetic vulnerability, external stressors, etc.), beliefs about controllability (by the individual themselves or via different treatments), beliefs about stability and predictability. Attitudes (n = 5) and behavior (n = 4) were explored by questions that assessed terminology used to describe mental disorders, and reactions such as avoidance or social distancing. Familiarity questions (n = 4) explored issues such as personal or family experiences of mental disorders and predictions about future vulnerability (factors that may act as modifiers of reactions); these were supplemented by questions that explored views on likely help-seeking and probability of self-disclosure.

The item format included ‘yes/no/don’t know’ questions, rank ordering of statements or Likert scale ratings. Respondents were also asked to endorse adjectives, verbs or expressions (from a list provided) that describe responses to mental disorders in general and then individuals with mental disorders specifically. The responses were then further classified qualitatively (eg ‘anger’, ‘pity’, ‘fear’, etc.) to allow the identification of key themes such as sympathetic, empathic or prejudicial 24 . Likewise some questions about predicted behaviours (eg would the respondent be prepared to work alongside someone with a bipolar disorder, schizophrenia or autism) assessed differences in reaction to, or degree of discrimination towards each disorder.

Data was analyzed using the SAS 9.1 statistical software package (SAS Corporation, Cary NC). Statistical significance was set a priori at the p < 0.05. For ease of interpretation, all data are presented as percentages unless stated otherwise. Basic statistics, such as chi-squared and rank order testing, were used to explore any specific differences regarding level of knowledge about risk factors for and attitudes towards each of the three disorders studied and when differences were found we then also examined whether there was an age group or gender effect. Missing values were dealt with by excluding a case with missing values from a specific analysis of that variable.

The target sample of 1000 respondents was recruited within the five day time frame. The sample was 52% female (see Table 1) and the mean age was 45 years (SD 14), about 30% were aged <35 and a similar proportion were aged 55–64 years. About 20% had received tertiary education and the mean monthly household income was 2,300€. About 18% lived in the Paris region, 11% in the south west, with the rest of the sample distributed evenly across the North West, north east and south east. The sample demographics are therefore similar to the adult population of France. Nine hundred and 16 of the 1000 questionnaires were fully completed or provided sufficient responses to allow some or all the questions completed by the individual to be included in the analyses.

Less than one in five respondents (17%) correctly identified the estimated prevalence of mental disorders to be 21-30%. However, there was almost universal recognition of the names of most mental disorders including the three specific disorders investigated, with 100% respondents recognizing the term autism and 97% and 96% respectively recognizing the terms schizophrenia and bipolar disorders. However, when respondents were asked if they could describe some of the characteristics of these disorders, the proportions decreased to 67% for autism, 53% for schizophrenia and 43% for bipolar disorders. Both awareness and knowledge of the disorders tended to increase with age (up to 55 years) and socioeconomic status but findings were inconsistent across disorders. When age and social status were controlled for, women were more frequently aware than men of mental disorders in general (p < 0.01) and the specific characteristics (p < 0.05).

As shown in Table 2, views about disorders showed significant differences (X² = 55.4; df = 8; p < 0.002): views about the need for treatment (less treatment) and prognosis (better outcome) favoured bipolar disorders; schizophrenia was regarded as similar to autism in having an early onset but viewed as dissimilar in ‘not being diagnosed early’. When offered a list of possible risk factors for each disorders, about two thirds of the sample rated drug or alcohol misuse as the most important risk factor for schizophrenia, 44% rated genetic factors as most important for autism, whilst 67% endorsed emotional stress as the most important risk factor for bipolar disorders; the rank ordering of risk factors for each disorders was statistically significant (p < 0.001).

*see text for details: views about course of disorder show % giving a ‘yes’ response; views about risk factors show % ranking each item first or equal first.

Only 33% respondents considered information available about mental disorders and their treatments to be adequate, but they rated the media as a more frequent source of sufficient information (30%) ahead of doctors (27%) or health professionals (21%) or other public services (13%). When asked to rate the effectiveness of a list of different treatments approaches, a minority (<1 in 5) of interviewees expected medication or psychotherapies to be ‘very effective’, whilst >30% considered all treatments to be ineffective or did not know if they were likely to be beneficial. When controlling for age group, women were significantly more likely than men to believe psychological treatments would be very effective (70% v 58%; p < 0.01) and also more likely to endorse medications (74% v 66%).

When asked to provide descriptors of mental disorders, 61% respondents reported pejorative labels (e.g. 47% used the words mad or lunatic). Compassionate descriptors, (e.g. sad or sorrowful), were used by 29% respondents. In comparison, when respondents were asked to provide descriptors of individuals with a mental disorder rather than of mental disorders per se, respondents were more likely to use compassionate (34%) and less likely to use negative labels (18%). There was a four-fold likelihood of respondents being compassionate rather than pejorative if considering individuals rather than disorders (Odds Ratio 3.97, 95% Confidence Intervals 1.92 to 8.18; X² = 14.6, df = 1, p < 0.001)

As shown in Figure 1, respondents viewed individuals with autism or bipolar disorders to be able to live in society, and viewed individuals with bipolar disorders as most likely to be able to work- views that were less likely to be endorsed for schizophrenia. Familiarity with mental disorders was associated with a non-significant trend for greater endorsement of the ability of individuals with autism or bipolar disorders to live in society, but no increase in rate of endorsement for individuals with schizophrenia.

Of the three disorders studied, schizophrenia generated the greatest fear, distrust and desire for social distancing: 65% respondents considered individuals with schizophrenia to be dangerous to others (as shown in Figure 1, this compares with 29% for bipolar disorders and 7% for autism; X² = 12.1, df = 2, p < 0.001). Furthermore, 30% would refuse to work with an individual with schizophrenia (bipolar disorders = 12%, autism = 6%; X² = 11.8, df = 2, p < 0.002), 31% would not want their children in the same class at school (bipolar disorders = 15%, autism = 3%; X² = 21.5, df = 2, p < 0.0002) and 24% would not agree to live with a relative with schizophrenia (bipolar disorders = 11%, autism = 6%; X² = 11.3, df2, p < 0.003). Figure 1 shows that respondents rated 22% of individuals with autism as a danger to themselves, compared with 74% of individuals with schizophrenia and 71% with bipolar disorders (X² = 10.3; df = 2; p < 0.004).

Although 40% of respondents reported familiarity with mental disorders only 5% (n = 55) reported a previous or current personal history of one or more mental disorder; an additional 34% did report they knew a friend or relative with a mental disorder. Of those with a personal history, 80% (n = 44/55) reported depression or mixed depression with anxiety, 9% (n = 5/55) bipolar disorders and 1.8% (n = 1) schizophrenia. A majority of respondents (65%) considered that they were potentially at risk of developing a mental disorder and two thirds also endorsed the statement that this was also true for members of their family. Interestingly, whilst it was unsurprising that 61% thought the most likely disorder to be experienced in the future would be depression, bipolar disorders (19%) was ranked second, ahead of anxiety (17%) or any other disorder. However, if problems were to occur, 30% of respondents stated that they would not mention it to their relatives, 60% would not mention it to friends, and 95% would not mention it at their workplace.

In a sample of 1,000 French persons, name recognition for mental disorders was high and marginally greater than reported in Australia (61%; 12 ) and Scotland (72%; 22 ). However, awareness of mental disorders (by name) was not necessarily accompanied by knowledge about the characteristics of the specific disorders and there were also limited expectations, especially in men, for treatment efficacy. Negative views of schizophrenia in this sample seem to be similar to those reported in other studies many decades earlier (eg 16 ). Interestingly, prejudice, which was evident from the frequent use of negative labels about mental disorders in general, was significantly less overt when descriptors were related to an individual with a mental disorder. This is a relevant finding when considering how we might attempt to reduce stigma, but this simple question represents a novel method of examining this issue.

There were interesting and previously unreported differences in public views of risk factors for each disorder studied, with genetic factors rated as most important in autism but psychosocial factors predominating in bipolar disorders and substance misuse more often rated as a risk factor for schizophrenia. This finding is especially interesting given the fact that there is increasing concern that messages regarding the biogenic nature of mental disorders may have increased rather than decreased negative views of adult mental disorders- this effect was not consistently observed in our study as it appeared to differ across disorders, possibly suggesting that age of individuals with a disorder may influence levels of stigma 3 5 6 .

The acknowledged personal experience of mental disorders was lower than predicted and this, plus the fact respondents also stated they would not readily share information with friends or work colleagues (and only 1 in 3 stated they would admit such problems to their family), suggests potentially high levels of self-stigma and fears about negative judgments or rejection by other people. Intriguingly, individuals ranked bipolar disorders as the second most likely future mental health problem (after depression), ahead of other common mental disorders, perhaps indicating that in the public’s mind bipolar disorders are more like depression rather than being severe mental disorders with many similarities with psychoses.

The responses to questions focusing on bipolar disorders, autism and schizophrenia showed a marked contrast between views of schizophrenia compared with the other disorders. Individuals with schizophrenia were considered dangerous by two out of three respondents and responses indicted that participants continue to have negative view of this disorder with low expectation of patient functioning and a desire for social distance for themselves and their children. Such attitudes were less likely to be expressed regarding bipolar disorders and remarkably rare in response to autism. The fact that negative stereotypes of schizophrenia prevail (despite research that challenges these notions e.g. 25 ), may indicate that media views of schizophrenia are still a more powerful influence. Direct support for this hypothesis comes from our finding that this sample acknowledged the media as their main source of information regarding mental disorders. Portrayal by the media of persons with psychosis as dangerous or helpless seem to be the norm while images of such people at school, work or enjoying themselves in the community are rare or absent 26 27 . Historically, there has been less media coverage of autism and bipolar disorders; not only have these disorders only recently begun to attract more public attention, but so far, the images presented in the media have tended to be positive rather than negative and are more often focused on an individual experiences (eg films such as ‘Rain Man’ or reports of celebrities with bipolar disorders) or agreement that there is a need to provide resources to support individuals (eg special school support for autism).

The company undertaking the majority of opinion surveys in France (Ipsos Public Affairs) provided a sample of sufficient size to ensure only a 3% margin of error. However, the survey was conducted online using individuals already recruited to a panel for market surveys, so by definition respondents were computer literate and accessed the internet regularly. Rates of declared personal history of mental health problems were also lower than expected. This sample may therefore be more homogeneous than in other community studies which may explain why we failed to find more consistent socio-demographic trends as reported in some previous studies 2 23 . The main study limitation is however that in order to make this a brief, internet based questionnaire we have used the concepts and ideas prevalent in the literature but did not use a set of previously validated reliable questionnaires, also some concepts needed to be translated into French (which in some cases may change the subtlety of the original meaning eg whether a person is described as being ‘able to work’ as opposed to ‘unable to work). Hence there may be issues in the nature of the questionnaire that will reduce our ability to compare some of the findings with other studies. Also we used diagnostic labels in some instances instead of case vignettes which will affect the views expressed or items endorsed 13 .