Stigmatization of people with mental illnesses: a follow-up study within the Changing Minds campaign of the Royal College of Psychiatrists

Abstract

A population survey before the start of the Changing Minds campaign showed that negative opinions about people with mental illnesses were widely held, and that opinions about different disorders differed in important ways. We repeated the survey 5 years later, when the campaign had ended. Interviews were again conducted with a representative population sample (1725 interviews; response rate 65%), enquiring about demographic variables, about eight opinions concerning seven common mental disorders, and whether the respondents knew anyone with one of these mental disorders. The pattern of response in this second survey resembled that in the first. However, there were significant changes. Though often small, apart from reported opinions concerning treatment and outcome, they were all reductions in the percentages of stigmatizing opinions. Seventy seven percent of respondents reported knowing someone with one of the seven disorders. Those who did so in respect of severe depression or panic and phobias were less likely to have stigmatizing opinions about people with the corresponding disorder, but the same did not apply to the other disorders. The greatest proportion of negative opinions was in the 16-19 year age group, and respondents with higher education were less likely than the rest to express such views. We conclude that stigmatizing opinions are frequent in the community but the various disorders are not stigmatized in the same way. Campaigns to reduce stigma should take account of these differences, and of the need to address young people.

In 1998, the Royal College of Psychiatrists commissioned the Office for National Statistics to carry out a national survey of public opinions about people with mental illnesses. Negative opinions were widespread, especially in relation to people with schizophrenia, alcoholism and drug dependence, who were widely regarded as unpredictable and dangerous (1,2). The survey was repeated five years later. We describe the principal results of this new survey, the changes since the first report, and certain characteristics of individuals that are associated with stigmatizing opinions. Between the two surveys there have been several anti-stigma campaigns, including the "Changing Minds: Every Family in the Land" Royal College of Psychiatrists' campaign (3).

METHODS

The survey was carried out for the College by the Office for National Statistics using their established National Statistics Omnibus Survey methodology (4). With one exception, described below, the questions were the same as those of the first survey (1). The main features of the method can be summarized as follows.

The sample

A nationally representative sample of 3,000 addresses (100 addresses in each of 30 postal sectors) in Great Britain was drawn from the Post Code Address File of the Office for National Statistics. At addresses shared by more than one household, one household was selected at random. In each household, one person aged 16 years or over (an "adult") was chosen randomly. This means that people in households with few adults had a higher probability of selection than people in households with many adults. A weighting factor was applied to correct for this inequality. Proxy interviews were not taken.

The interview

The interviews were carried out in July 2003, five years after the first survey. Questions were asked about seven disorders: severe depression, panic attacks or phobias, schizophrenia, dementia (e.g., Alzheimer's disease), eating disorder (e.g., anorexia nervosa, bulimia nervosa), alcoholism, and drug addiction.

Interviewers asked whether respondents knew anyone with any of these seven psychiatric disorders. This question was the only one that differed between the two surveys. In 1998, interviewers had asked the more general question whether respondents knew anyone with mental illness, without mentioning specific disorders. In neither survey were respondents asked whether they themselves had experienced a mental illness.

Respondents were then asked how far each of eight statements applied to people with each of the seven disorders. The statements were: dangerous to others, unpredictable, hard to talk with, feel different from the way we feel at times, have only themselves to blame for their condition, could pull themselves together if they wanted, would not improve if given treatment, will never recover fully.For each item, respondents were asked to choose a point on a five point scale, the extremes of which bore anchoring statements: for example dangerous to others - not dangerous to others.

Region of residence, age, gender and ethnicity were recorded. Regarding occupation, respondents in the second survey were assigned to three groups: managerial/professional; intermediate; and routine/manual (since the classification in general use at the time of the first survey was slightly different, this analysis was of the second set of data only). Respondents were also assigned to one of four groups according to annual income: less than £5,000; £5,000 but less than £10,000; £10,000 but less than £20,000; £20,000 and more.

Also recorded was the age at which respondents left full-time education, as a proxy for highest educational level achieved, and whether there were children under 16 years of age in the household.

Statistical methods

The responses recorded on the 5 point scales were converted into scores as follows: scores 1 and 2 were coded as negative, scores 4 and 5 were coded as positive, and those who were unsure or could not answer the question were assigned to the central code 3. Sampling errors between surveys were calculated using the standard method assuming random samples. Except where stated, attention is drawn, in the results section, only to differences that are significant at the 95% confidence level. With so many comparisons, results must be interpreted cautiously.

In order to compare the responses of people with different characteristics, two composite scores were derived from the original scores. An overall opinion score was obtained by summing the scores on the 5-point scales relating to 5 of the 8 opinions, omitting the two opinions about outcome and also the rating concerned with feeling different. The resulting scores, which vary from 5 to 25, with 5 as the most negative response, were grouped into negative (5-12), neutral (13-17) and positive (18-25). We omitted the two statements that refer to prognosis, because negative responses may reflect knowledge rather than prejudice, for example the opinion that dementia has a poor prognosis. We omitted responses to the statement about feeling different for the reasons presented later. This method of calculating an overall opinion score gives equal weight to each of the five opinions, and it can be argued that some opinions should be given greater weight, for example opinions about dangerousness. However, there is no generally agreed way of weighting the opinions, so we did not attempt it.

A summary score was obtained by adding the five overall opinion scores for each of the seven disorders. The resulting score varies from 7 to 21, with 7-10 the most negative and 18-21 the most positive.

The effects of individual characteristics were examined by comparing the percentages of respondents with and without the relevant variable, whose responses led to negative overall opinion scores and summary scores in the most negative group.

RESULTS

The sample

Interviews were obtained with 1725 people aged 16 years and over. This response rate of 65% is close to that obtained in similar surveys carried out by Office for National Statistics, and to the 67% response rate in the 1998 survey.

Personal knowledge of someone who has had mental illness

In 1998, 52% of respondents said yes to the single question "Do you know someone with mental illness?". In 2003, the question was rephrased and asked about each of the seven illnesses. In response to these seven more specific questions, 77% said that they knew at least one person with one of the specified illnesses. Almost half the sample (47%) reported knowing someone who had experienced severe depression; about a third knew someone with alcoholism (37%), dementia (34%), or panic and phobias (33%); about a quarter knew someone with drug addiction (23%), or an eating disorder (22%); and 15% knew someone with schizophrenia.

In general, men were rather less likely than women to answer that they knew someone with a mental disorder (74% vs. 80%) and to know someone with any of the individual conditions, with the exceptions of drug addiction (M/F: 25/22) and alcoholism (M/F: 40/34). Respondents aged 16-24 were more likely than those 65 and over to answer that they knew someone with drug addiction (43% vs. 8%), alcoholism (42% vs. 21%) and eating disorder (36% vs. 8%). The percentages of respondents in other age groups who knew someone with these disorders lay between these extremes. On the other hand, respondents aged 65 and over were more likely than those aged 16-24 to answer that they knew someone with dementia (40% vs. 20%).

In terms of socio-economic status, the main finding was that respondents in managerial and professional classes were rather more likely than other respondents to say that they knew someone with at least one of the mental illnesses (82% vs. 75%).

Opinions about people with mental illnesses

Opinions expressed in 2003

Table 1 reveals that the percentage of respondents endorsing the eight negative opinions differed between the seven disorders. The greatest variation in this percentage was in relation to danger to others, ranging from 75% for drug addiction and 66% for schizophrenia, to only 7% for eating disorder. The corresponding percentages in relation to unpredictability form a similar pattern, while the percentages for blameworthiness were greatest for drug addiction (60%) and alcoholism (54%) and smallest for schizophrenia (6%) and dementia (4%). The percentage endorsing negative statements about treatability also differ between the disorders, with 50% endorsing the opinion that dementia will not recover with treatment, but only 11% choosing this response for drug addiction. The statement "will never recover completely" was endorsed by 79% in relation to dementia but only 15% for eating disorder, while 42% chose this response for schizophrenia.

Table 1.

Percentages agreeing with negative statements, 1998 and 2003

	Severe		Panic		Schizophrenia		Dementia		Eating		Alcoholism		Drug
	depression		Attacks						disorder				addiction
	1998	2003	1998	2003	1998	2003	1998	2003	1998	2003	1998	2003	1998	2003
Danger to others	23	19**	26	23*	71	66**	19	20	7	7	65	64	74	75
Unpredictable	56	53	50	50	77	73*	53	54	29	27	71	70	78	78
Hard to talk to	62	56**	33	26**	58	52**	60	56*	38	33**	59	55*	65	64
Feel different from us	43	30**	39	25**	57	37**	61	42**	49	33**	35	25**	48	33**
Selves to blame	13	11	11	10	8	6*	4	4	34	33	60	54	68	60**
Could pull self together	19	17	22	20	8	8	4	5	38	35	52	50	47	45
Not improved if treated	16	15	14	15	15	12*	56	50**	9	10	11	12	12	11
Never fully recover	23	25	22	21	51	42**	82	79*	11	15**	24	29**	23	26**

Differences from the 1998 data

* p<0.05

** p<0.005

The greatest percentage of negative opinions was in relation to drug addiction, especially opinions about dangerousness, unpredictability and being hard to talk with. Alcoholism and schizophrenia were close behind on these three variables, but whereas more than half the respondents thought that people with drug addiction and alcoholism were to blame for their condition, only 6% expressed this opinion about people with schizophrenia.

Comparison of responses in 1998 and 2003

The pattern of responses in 2003 is similar to that found in 1998, suggesting that the seven disorders are recognized consistently and that there are real and enduring differences of opinion about them. The percentages of people expressing negative opinions changed in a number of instances over the 5 years. This change was generally a decrease, often small. Of the 27 (out of a total of 56 items) which changed significantly, the 24 concerning opinions about people with the given mental illnesses were all decreases. All three increases were in relation to opinions about prognosis, with more respondents choosing the statements that people with eating disorder, alcoholism and drug addiction will never fully recover. Of the 24 decreases, seven were responses to the statement feel different from the way we feel at times and these decreases were often greater than those for any of the other statements. Some interviewers reported that some respondents found this statement difficult to understand. Nevertheless, the overall pattern of the data on this statement was consistent with clinical expectations. Thus, three times as many people reported that they felt unable, at times, to empathize with people with schizophrenia and dementia as reported being able to do so. For eating disorder and drug addiction the ratio was two to one. For depression it was three to two, whilst for panic attacks and alcoholism the numbers were nearly equal. Moreover, the changes in the empathy item are closest to those in the hard to talk to item, which can be considered to reflect related aspects of communication with people with mental illnesses. Even when the empathy result is discarded, the pattern of change between the two surveys remains the same, with decreased percentages for 14 of a possible 49 items and increased percentages for just the items concerning prognosis. Nevertheless, in view of the doubt about some respondents' understanding of the statement concerned with feeling different and also because the relatively large changes in it might unduly distort the two sets of summary scores, we excluded it from that aspect of the analysis.

Overall opinions about each disorder

Table 2 shows that, in 2003, drug addiction was viewed unfavourably by the greatest percentage of respondents, with 74% of respondents recording overall negative opinions and only 5% recording positive ones. Alcoholism, with 66% overall negative opinions and only 6% positive ones, is next. Dementia, with only 3% overall negative and 35% positive opinions, was viewed most favourably, followed by panic attacks (14% negative and 36% positive) and eating disorder (13% negative and 32% positive). Severe depression (16% negative and 28% positive) is in an intermediate position. Schizophrenia has 21% negative and only 9% positive overall opinions and 70% of neutral ones. This is the highest percentage of people with overall neutral responses for any of the seven disorders; the lowest are for alcoholism and drug addiction.

Table 2.

Overall attitudes (%) toward each disorder by year of interview

	Year of	Negative	Neutral	Positive
	interview	(5 to 12)	(13 to 17)	(18 to 25)
Severe depression	2003	16	56	28
	1998	20	59	21
Panic attacks	2003	14	50	36
	1998	14	54	31
Schizophrenia	2003	21	70	9
	1998	22	71	7
Dementia	2003	3	62	35
	1998	4	59	37
Eating disorder	2003	13	55	32
	1998	14	58	28
Alcoholism	2003	66	28	6
	1998	69	25	6
Drug addiction	2003	74	21	5
	1998	77	19	4

Changes in overall opinions

Table 2 shows that the pattern of overall opinions was similar in the two surveys. For none of the disorders was there an increase in the percentage of respondents expressing overall negative opinions over this period and there were small decreases for depression (from 20% to 16%), alcoholism (69% to 66%) and drug addiction (77% to 74%). The percentage of overall positive opinions increased slightly for depression (21% to 28%), panic attacks (31% to 36%) and eating disorder (28% to 32%). Only for dementia was there a small decrease. In both surveys, the proportion of respondents expressing overall neutral opinions was high for all disorders except alcoholism and drug addiction. This percentage decreased slightly over the five years for depression (59% to 56%), panic attacks (54% to 50%) and eating disorder (58% to 55%). On both occasions, schizophrenia attracted the greatest percentage of overall neutral opinions (71% in 1998 and 70% in 2003).

Opinions about the outcome of mental illnesses

We summed the scores on the scales relating to prognosis and response to treatment to give a composite outcome score varying from 2 to 10. We grouped these composite scores into three categories: poor (2-4), neutral or uncertain (5-7) and good (8-10). These outcome scores differed between the seven disorders: 70% of respondents had poor outcome scores for dementia and only 6% had good outcome scores. In contrast, only 9% had poor outcome scores for eating disorder and 65% had good outcome scores. For schizophrenia, 25% had poor outcome scores and 29% had good ones. For the other disorders, between 16 and 18% had poor outcome scores and 52-53% had good ones. For all except one of the disorders, between about a quarter and a third of ratings were neutral or uncertain. The exception was schizophrenia, with almost a half (46%) of overall ratings in this category.

Effects of personal characteristics

Age and gender

Respondents aged 16-19 years were more likely than the rest to have negative overall opinion summary scores for every disorder except panic attacks (Table 3). For example, for depression, 36% of 16-19 year olds had negative overall opinion scores, compared with fewer than 20% of those in every other age group. Likewise, 31% of 16-19 year olds had negative overall opinion scores for schizophrenia, compared with 20-26% of those aged 25 years and over. For drug addiction and alcoholism, there was a trend across the age groups with the greatest percentage of negative summary scores among the younger respondents (87% for drug addiction and 85% for alcoholism). On the summary scores, 49% of 16-19 year olds were in the two negative categories compared with 36% for the other age groups combined. Also, compared with the rest, far more of the 16-19 year olds were in the most extreme negative category (22% vs. 8%).

Table 3.

Percentage with negative views by age (combined data 1998 + 2003)

	16-19	20-24	25-44	45-64	65 and
					over
Severe depression	36	18	17	15	19
Panic attacks	15	9	15	13	16
Schizophrenia	31	26	21	21	20
Dementia	7	1	2	3	3
Eating disorder	21	12	11	13	15
Alcoholism	85	77	74	64	55
Drug addiction	87	83	81	74	64

Compared with women, men were rather more likely to have overall negative opinions for depression (20% vs. 16%), panic attacks (17% vs. 12%), schizophrenia (24% vs. 20%) and eating disorder (15% vs. 12%). For dementia, alcoholism and drug addiction, however, the corresponding differences were not significant. On the summary scores, 40% of men were in the two negative groups compared with 34% of women (the percentages in the extreme negative category were: men 10%; women 8%).

Region of residence, ethnicity, occupation and income

There was no consistent pattern of variation by region of residence, either in the overall negative opinion scores or in the summary scores and the few and scattered positive findings may not reflect real differences between regions. Thus, respondents in Scotland were more likely than those in other regions to have negative attitudes to people with a drug addiction, whilst respondents in Wales were less likely to report negative attitudes towards those with depression or dementia.

It was not possible to examine relationships between reported opinions and ethnic backgrounds, because the number of non-white respondents was too low, even when the data from the two surveys were combined.

Regarding occupation, those in routine or manual occupations were more likely to have overall negative opinions for severe depression, compared with people whose current or last job was managerial or professional (18% vs. 11%). This was also true for schizophrenia (25% vs. 17%), and eating disorder (17% vs. 9%), but not for alcoholism (65% and 66%) or drug addiction (73% and 74%). Few of either group had overall negative opinions for dementia (4% and 1%). On the summary scores, 38% of those in routine and manual occupations had scores in the two negative categories, compared with 30% in each of the other two socio-economic classes. Most of the difference was in the percentages in the most extreme negative category (11% vs. 6% respectively).

Compared with those in the lower income groups, a smaller proportion of respondents with a gross personal income of £20,000 a year or more had overall negative opinions for depression (13% vs. 19%), schizophrenia (17% vs. 23%), and eating disorder (10% vs. 14 %). The responses of the high-income group concerning the other disorders fell between those of the three lower income groups. On the summary scores, 32% of the group with incomes of £20,000 per annum or more were in the most negative category, compared with 35-41% of people in the other income groups.

Education

Respondents who stayed in education after the age of 18 years (Table 4) were less likely than those who left earlier to have negative overall opinions for severe depression, schizophrenia and eating disorder, but not for alcoholism or drug addiction. Few in any group had negative overall opinions for dementia. Fewer of those whose education continued after the age of 18 years had summary scores in the most negative category, compared with those whose education did not extend beyond 18 years (4-5% vs. 8-9%). Respondents who were still in education at the time of the survey contained the highest percentage of those with the summary scores in the most negative category (15% vs. 4-9% for those with education beyond 18 years and 8-9% for the rest).

Table 4.

Percentage with negative views by age left full-time education (combined data 1998 + 2003)

	Up to 14	15 to 18	19 to 25	26 or over	Still in
					education
Severe depression	20	19	12	12	27
Panic attacks	16	14	14	14	13
Schizophrenia	19	24	17	17	19
Dementia	5	3	2	2	6
Eating disorder	16	14	8	11	16
Alcoholism	56	69	67	66	85
Drug addiction	62	78	75	78	89

Knowing someone with a mental illness

Compared with the rest of the respondents, a smaller proportion of those who knew someone with depression had overall negative opinions for this disorder (12% vs. 18%). Similarly, a smaller proportion of those who knew someone with panic attacks had overall negative opinions for panic disorder (10% vs. 16%). Those who knew someone with any of the other disorders were no less likely than the other respondents to have overall negative opinions for the corresponding disorder. Compared with the other respondents, a slightly smaller proportion of those who knew someone with schizophrenia held overall neutral opinions for this disorder (65% vs. 71%) and a slightly higher percentage had positive ones (15% vs. 8%). Similar differences were found for the responses of people who knew someone with an eating disorder (neutral scores 47% vs. 58%; positive scores 41% vs. 29%).

People who knew someone with the corresponding mental illness were more likely than the rest to rate as poor the outcome of depression (20% vs. 14%), schizophrenia (30% vs. 20%), dementia (75% vs. 63%) and alcoholism (21% vs. 16%) (Table 5).

Table 5.

Overall attitudes toward each disorder (%) by whether respondent knows someone with that disorder

	Respondent knows
	someone with	Negative	Neutral	Positive
	the disorder	(5 to 12)	(13 to 17)	(18 to 25)
Severe depression	Yes	12	53	35
	No	18	58	23
Panic attacks	Yes	10	41	49
	No	16	55	29
Schizophrenia	Yes	20	65	15
	No	21	71	8
Dementia	Yes	2	62	36
	No	3	62	35
Eating disorder	Yes	11	47	41
	No	13	58	29
Alcoholism	Yes	66	27	6
	No	66	28	6
Drug addiction	Yes	75	18	6
	No	74	22	4

Respondents with a child under 16 in the household

Respondents with at least one child under the age of 16 years in the same household were rather more likely than the rest to have overall negative opinions for alcoholism (73% vs. 66%) and drug addiction (80% vs. 74%) but not for the other disorders. However, the group with a child under 16 years in the same household did not have a greater percentage of respondents with summary scores in the most negative categories.

DISCUSSION

The survey was carried out on our behalf by the Office for National Statistics, using well-tried procedures for data collection and analysis. Its shortcomings are those of other large-scale surveys of representative samples of the population. The response rate was 65%, which is close to the usual rate achieved in comparable surveys by the Office for National Statistics but nevertheless incomplete. Also, as in all opinion surveys, we cannot be certain that expressed opinions accurately reflect true opinions or that opinions reflect actual behaviour. Therefore, as in any population survey, conclusions should be drawn cautiously. Nevertheless, the broad similarities in the pattern of responses obtained in the present survey and in that conducted in 1998 encourage confidence in the reliability of the procedures.

We asked about opinions on seven psychiatric disorders whereas most other investigations of stigma have enquired about a general concept of mental illness. We did this because many people with major psychological distress presenting to doctors now receive a psychiatric diagnosis and it is important to know what stigma attaches to it. It is sometimes suggested that the stigma attached to a psychiatric diagnosis does harm that outweighs any benefits from the diagnosis. We do not share this view. It is true that diagnostic labels focus attention on general features, often of an adverse kind and, in this way, are potentially stigmatizing (5). However, labels with damaging and dismissive connotations have long been attached to people with mental disorders (6), and modern diagnoses are more specific and can be a valuable guide to prognosis and treatment. In any case, since stigma is attached to diagnoses, it is most important to find out more about it so that it can be reduced and eventually overcome.

Stigmatizing opinions in 2003

Stigmatizing opinions were endorsed commonly by men and women from all social classes, living in all parts of the country. There were, however, differences in the nature and extent of the stigma attached to the seven disorders. The most stigmatized of the disorders were drug addiction, alcoholism and schizophrenia. People with schizophrenia are likely to find that two thirds of the people they meet think that they are dangerous, that three quarters think they are unpredictable, and that half think that they will be hard to talk with. People with alcoholism and drug addiction are even more stigmatized, for not only do many people think that they are dangerous, unpredictable and hard to talk with, but three in five people think that they are to blame for their condition - an opinion endorsed by only 6% in relation to schizophrenia.

Compared with people with schizophrenia, those with severe depression will encounter fewer people who think them dangerous - about one in five - but as with schizophrenia, about half will think that they are unpredictable and hard to talk with. These last two opinions are likely to discourage those who hold them from making the closer contact with depressed people that could lead them to change their other opinions. Looked at from the opposite standpoint, however, a substantially greater proportion of people expressed positive opinions than negative ones about people with depression, panic and phobic disorders, dementia and eating disorder.

Knowing someone with mental disorder

In the first survey we asked the general question whether respondents knew someone with mental illness and 52% said yes. This figure is close to the 49% of respondents who said that they knew someone with mental illness in another recent survey (7). In the present survey, interviewers asked separately about knowledge of someone with each of the seven disorders, and about three-quarters of respondents replied affirmatively. It is possible that, on the first occasion, some respondents did not consider, for example, eating disorder as a mental disorder. If so, this could account for the differences between the two survey responses. Opinions about the people with psychiatric disorders are subject to many influences, including accounts in the media and, in some cases, personal knowledge of a person with the illness. The importance of personal knowledge varies between the seven disorders: almost half the respondents knew someone with severe depression, but only one in six knew someone with schizophrenia. For the rest, about one in three knew someone with alcoholism, dementia or panic attacks; about one in four knew someone with an eating disorder or with drug addiction. As would be expected, the elderly were more likely to know someone with dementia, and the young were more likely to know someone with drug addiction, alcoholism or eating disorder.

Respondents who knew someone with depression or panic and phobias were rather less likely to express stigmatizing opinions about the corresponding disorder, but this was not the case with personal knowledge of someone with one of the other disorders. However, people who knew someone with schizophrenia or eating disorder were a little less likely to record neutral opinions and a little more likely to record positive ones. People who knew someone with depression, schizophrenia, dementia and alcoholism were also rather more likely to think that the outcome of the corresponding condition was poor. It seems therefore that personal acquaintance does not always modify negative opinions that have been arrived at in other ways. Meanwhile, there is also some evidence that expertly guided contact with people with chronic mental illness severe enough for them to need collective supervised community care, can lead to increased knowledge and empathy in members of the public (8).

Neutral opinions

The neutral category is for respondents whose opinion is undecided or uncertain. When the overall measure is used, 70% of respondents recorded overall neutral opinions about schizophrenia, compared with only 21% for drug addiction and 28% for alcoholism. For the other dis- orders, about a half of the respondents recorded such opinions. These substantial percentages of people with neutral or uncertain opinions are a potential target for anti-stigma campaigns, for it seems possible that their opinions might be more amenable to positive change than those of people whose opinions are negative. If this is correct, it is noteworthy that the largest percentage of neutral opinions was about schizophrenia.

Changes between 1998 and 2003

The most striking change over the 5-year interval was that fewer people endorsed the statement that people with any of the seven disorders feel different from us. This change ranged from 10 to 20%, which is substantially greater than the changes in any other opinion. The opinion feel different from us was included because it might reflect an aspect of empathy, and lack of empathy is an important aspect of stigma. Because the changes are larger than those in any other variable, and because some respondents had difficulty in understanding the statement, it is necessary to consider alternative reasons for the decrease. It is unlikely that it is due to a change in procedure, since the statement was worded and presented to respondents in exactly the same way on the two occasions. The fact that the percentages changed for all seven disorders suggests that the findings are not random responses. Nevertheless, because the greater size of the changes is unexplained, we thought it safer to exclude this opinion from analyses in which overall opinions were considered. However, we do not wish to ignore it. If it does reflect real change, then we regard it as a potential forerunner to other less negative public opinions concerning people with mental illnesses.

The Royal College of Psychiatrists and other organizations have campaigned against stigmatization during the 5 years between the two surveys. However, the positive changes between the two occasions cannot be ascribed to these campaigns, because opinions are subject to so many other influences. Nevertheless, the results can suggest whether stigmatization is changing. A recent study (7) reported that, while the majority of respondents expressed caring and sympathetic views of people with mental illness, attitudes had become less positive between 2000 and 2003. Our findings were about individual disorders, rather than a global concept of mental illness, and we found some small improvements. For example, although overall opinions about schizophrenia changed little over the 5- year period, about 5% fewer respondents endorsed the opinions that people with schizophrenia are dangerous and unpredictable. Further surveys will be needed to determine whether the changes that we observed are part of a lasting trend to improvement.

Our findings suggest that anti-stigma campaigns should be directed not to mental illness in general, but to specific disorders. This is because the stigma attached to the various disorders is not all the same. Thus, the majority of our respondents endorsed the opinion that people with schizophrenia are dangerous, unpredictable and hard to talk with, but few thought that they are capable of pulling themselves together. On the other hand, very few respondents endorsed the opinion that people with eating disorder are dangerous but a third thought that they could pull themselves together. However, there were some common themes, notably that respondents thought that people with all seven disorders would be hard to talk with. Campaigns to reduce stigmatization need to address both the specific and the general opinions, whilst recognizing the social handicaps that can be features of mental illness, in messages tailored to each disorder. They will require that both knowledge and contact skills (9) are appropriately enhanced.

Opinions about prognosis and the effects of treatment

Opinions about prognosis and treatment were generally realistic. Thus, although half of the respondents endorsed the opinion that dementia responds poorly to treatment, only 10-15% said this of the other disorders. Opinions about outcome were similar. These findings suggest that there is a basis of understanding of at least some aspects of mental disorders on which future anti-stigma campaigns might build.

Characteristics associated with stigmatizing opinions

Among the respondents to these two surveys, stigmatizing opinions were endorsed by men and women of all ages, living in all parts of Great Britain. There were nevertheless some features that help to characterize people who endorsed negative opinions. Regarding age, the most negative opinions were endorsed by 16-19 year olds: one in three endorsed overall negative opinions about people with schizophrenia and depression, compared with about one in five in the other age groups, and a striking 85% endorsed overall negative opinions about people with alcoholism and drug addiction. Also, respondents in the 16-19 age group were more likely than the rest to choose the most extreme overall opinions. The findings concerning alcoholism and drug addiction are in contrast with the reported wide use of alcohol and drugs by young people. The finding may suggest that the young people who use these substances do not think of themselves as potential abusers and do not identify with those who are. If so, the findings would be relevant to campaigns that seek to prevent drug abuse among young people by warning of the consequences of addiction.

Stigmatizing opinions were expressed by a smaller proportion of those who had received higher education than of those whose education had stopped at age 18 years. While this finding may point to the value of education in reducing stigmatization, it could also reflect the processes of selection for higher education, or the greater choice of socially acceptable responses to questions by some respondents in the higher education group. Nevertheless this finding, together with those concerned with age, suggests that anti-stigma campaigns should pay particular attention to young people. This was a feature of the recent campaign by the Royal College of Psychiatrists (see 10).

Stigmatizations of people with mental illnesses are fuelled by many things. Media attention still often focuses public attention on the most negative attributes of mental illnesses (11). Moreover, people observe or read about evident problems of the "visible" minority of people with chronic and severe mental illness, and also those with personality disorder (which has attracted much media attention these last few years and which we did not include in our survey). They may then generalize their conclusions, assuming that everyone with a mental illness will have the same severe problems. If Murray and Lopez (12,13) are correct, then 'one in four' of us will experience a mental illness personally at some stage of our lives. Correspondingly, we are likely to encounter such illness at some stage in 'one in four' of those around us, or, as the Royal College campaign put it, within 'every family in the land'. Stigmatizations by others, such as we have identified here, may be coupled with self-stigmatizations (14,15). Together they make it even harder for people with a mental illness to acknowledge their problems and seek help. Real progress in the care of people with mental illnesses requires that stigmatizations are reduced alongside improvements in prevention, treatments and self-help strategies.