Thomas J. Schoeneman, Suzanne Segerstrom, Paul Griffin, and David Gresham

Lewis and Clark College

Journal of Social and Clinical Psychology, 1993, 12, 429-453

Abstract

Three studies investigated implicit categories of mental illness. In the first, 124 informants in public places and college classrooms generated 162 category labels for mental illness. Next, we narrowed the list to 48 items mentioned more than once in Study 1 and rated as familiar and prototypical by 135 undergraduates; schizophrenia, manic-depression, and depression were the most prototypical disorders. In the third study, 75 undergraduates sorted the 48 categories by similarity. Cluster analyses indicated that 12 low distance clusters combined into three large groups: Violence and Disorders of Mood and Control, Intellect Disorders, and Schizophrenoid Disorders. A multidimensional scaling analysis of Study 3 yielded three dimensions which we interpreted as Onset Controllability/ Responsibility, Cognitive Deficit versus Excess, and Potency/Severity. In discussing our findings, we note a correspondence to the 20th century triad of psychosis, neurosis, and organic disorders as well as to the three historical stereotypes of maniac, melancholic, and fool. We speculate about the assumptions and values revealed by subjects' responses and their social implications.

Our opening quotation suggests that conceptions of mental disorder--both lay and professional--are the products of social discourse. This view is consistent with an emerging orientation in psychology known as social constructionism (Gergen, 1985). A basic assumption of constructionism is that knowledge originates in social interchanges. That is, people's everyday knowledge about "the way things are" is not given by the real world but is the result of an ongoing process of communication: People speak, write, and use signs and symbols actively and cooperatively and end up creating "reality" out of negotiated understandings. The constructionist position is that the "real world" of the five senses and the thinking brain emerges from social discourse. In addition, constructionists argue that the negotiated understandings that produce conventional wisdom are grounded not only in the participants' immediate situations but also in significant historical and cultural contexts. As one pair of investigators assert, "Representations of reality are shared meanings that derive from language, history, and culture" (Hare-Mustin & Maracek, 1988, p. 456).

A typical social constructionist research agenda would begin with a careful scrutiny of some aspect of the taken-for-granted world of a particular group--for instance, conceptions of mental disorder in modern Western culture. The investigator would attempt to interpret what people say about mental illness and try to extract meaning from the responses in light of the personal, historical, and cultural contexts in which they are embedded. After uncovering some of the ideological forms and social processes that underlie this corner of the "real world," the investigator might examine their consequences. When a particular understanding of reality emerges from social interaction, some actions are prescribed while others are excluded, and the investigator could try to describe the overall complex of communication, cognition, context, and behavior (Gergen, 1985).

The research that we report here is a preliminary step in a constructionist account of current Western conceptions of mental disorder. Refer again to the initial quotation from MacDonald (1981): It suggests that a society's conceptions of insanity are a joint function of psychiatric nosology and lay stereotypes. We began our investigations from the nonprofessional side by asking about the vocabulary and categories that people use today in their everyday thinking about mental disorder.

More specifically, we posed the following questions: When people think about mental illness, what categories do they use? How is their knowledge organized? What dimensions of difference underlie this knowledge? There have been many investigations of attitudes toward the mentally ill (see Rabkin, 1972, 1980; Wahl, 1992) and of the structure of professional taxonomies of mental disorder (Buss & Craik, 1986; Cantor & Genero, 1986), but implicit categories of mental illness have received little attention. This is surprising because there are important consequences that could follow from popular stereotypes of psychiatric problems. In the first place, categorizations should differentially affect not only actions toward target persons but also the self-concepts and behavior of recipients of category labels (Goffman, 1963; Jones et al., 1984; Scheff, 1966). In addition, popular understandings of mental disorder could conflict with professionals' attempts at education and treatment (Cumming & Cumming, 1957). It is also possible that professional views of mental disorder might reflect or express popular conceptions in nonobvious ways (Gilman, 1988; MacDonald, 1981). Before any of these implications can be investigated, though, the categories and dimensions of mental disorder that are implicit in people's day-today knowledge must be revealed. We conducted three studies to begin this venture.

Studies of attitudes toward mental illness often treat the "mental patient" as a single stereotype (Gerbner, Gross, Morgan, & Signorelli, 1988; Nunnally, 1961; Rabkin, 1972) while formal taxonomies such as DSM-III-R (American Psychiatric Association, 1987) offer a profusion of categories that are related in ways that may not be obvious to casual observers (Klerman, 1988). While it is possible that everyday conceptions of mental illness tend toward the undifferentiated concept of "mental patient," both psychological investigations of natural categorization (e.g., Rosch, 1978) and common sense suggest that "mental patient" could well be an abstract concept that includes a number of more concrete categories. If there are multiple stereotypes of mental illness, then two questions arise: "How many?" and "What kind?"

It is likely that the views of 20th century laypersons would be influenced by the concepts of twentieth century psychiatry. Three broad categories have characterized psychiatric classification for much of this century: psychoses, neuroses, and organic disorders (Klerman, 1988). Interestingly, this triad parallels cultural conceptions that are centuries old. In historical investigations of popular views of insanity (MacDonald, 1981), visual stereotypes of madness in art and medical illustration (Gilman, 1982, 1988), and formal diagnostic systems (Babb, 1951; Jackson, 1986), three figures emerge repeatedly.

The first is the maniac, usually characterized as active, uncontrolled, and unpredictable; pictures of maniacs show contorted and disheveled individuals who have flailing limbs, wide open eyes and mouth, and hair and clothes in wild disarray (or even missing). Historically, the diagnostic terms that have applied to the maniac include madness, lunacy, mania, distraction, and light-headedness; we suspect that current synonyms may include psychosis and schizophrenia.

The complement to the maniac is the passive, pensive melancholic usually portrayed as seated, head in hands, with darkened face and downcast eyes. Melancholy, along with its less fashionable cognates of mopishness, insensibility and lethargy, became known as depression early in this century; however, earlier descriptions of melancholy (e.g., Burton 1621/1977) are so broad as to suggest that the correct modern synonym would be "neurosis."

The third figure that occurs frequently in both lay and psychiatric conceptions of disorder is the mentally deficient fool or idiot. In art and literature, the fool is sometimes portrayed for satirical or polemical purposes as a jester with cap, bells, and fool's bladder and staff. More "realistic" portrayals of mental deficiency and senility often include standard, general icons of mental disorder such as hidden hands, unfocused gaze, the posture of mania or melancholy, darkness, and a staff of madness.

The main outlines of the stereotypic figures of the maniac, melancholic, and fool are continuous over the centuries of Western history; as Gilman (1982) has said of the visual elements of stereotypes of madness, "they seem never to die, only to recede from the center of perception" (p. xii).^l However, historical analyses also make it clear that the details of conceptions of insanity reflect the social and intellectual contexts in which they are embedded. An example of such contextual effects comes from MacDonald's (1981) study of early modern England. MacDonald compiled a list of over 100 symptoms that were current in 17th century England in the medical and popular literatures on mental disorder and in the case notes of 2,483 consultations kept by an eminent practitioner, Richard Napier. These materials indicated that there were two "clusters" of mental disorder in common currency: the first was a pair of chronic disorders, madness and mania, characterized by high activity and incoherent speech; the second pair, melancholy and mopishness, were acute and less severe problems involving apathy and disordered emotions. MacDonald noted that delusions and hallucinations, "which today are regarded as the token of the worst kinds of insanity, were considered to be symptoms of melancholy rather than madness" (p. 170). This seems odd to modern minds but makes sense when we take into account the cosmology of the times (cf. Babb, 1951; Tillyard, n.d.): The universe was arranged hierarchically, with humans at the intersection of the supernatural and natural realms and thus subject to influences from both. Therefore, when someone reported seeing or conversing with angels, demons, saints, and the like, the "delusion" or "hallucination" always remained within the realm of reasonable possibility; even if an account of a particular vision were doubted, people believed that there was an unseen world that was active, immediate, and potent in human affairs. MacDonald (1981) points out that "the most compelling evidence for this argument is that when the educated elite abandoned their beliefs in divine inspiration and demonology, they also elevated delusion to a prominent place among signs of madness" (p. 170).

These considerations of historical continuities versus context-dependent features of conceptions of mental disorder resulted in two predictions. First, we expected to find the basic triad of maniac/ psychotic, melancholic/neurotic, and fool/organic in the more abstract and general aspects of our subjects’ responses.² However, the more concrete details, such as the vocabulary of terms used to describe disorders and judgments about close similarities among categories, should display features of modern American views and concerns.

Our research consisted of a sequence of category nomination, prototypicality rating, and similarity judgment that is common in psychological research on natural categorization (cf. Buss & Craik; 1986, Shaver, Schwartz, Kirson, & O’Connor, 1987; Sternberg, Conway, Ketron, & Bernstein, 1981)^.3 In Study 1, we asked people in two public locations and in college classrooms to list as many kinds of mental disorders as they could. Undergraduates in Study 2 reduced this considerable set of items by rating the typicality of each and indicating which terms were unfamiliar. In Study 3 undergraduates sorted mental disorders into similarity sets. Hierarchical cluster analyses revealed the content and structure of these sorts and, consistent with the idea that stereotypes exaggerate similarities within groups and distinctions between groups (Cantor & Genero, 1986), we chose an analysis strategy that emphasized the formation of distinct clusters. Finally, we submitted our subjects’ similarity judgments to multidimensional scaling in order to characterize their conceptions of abnormal psychology in a different way. Categorical and dimensional approaches are sometimes presented as adversaries in efforts to describe phenomena (see Klerman, 1988, on this rivalry in the area of psychiatric diagnosis). Following Shaver et al. (1987), we view the two approaches as different but complementary ways of representing people’s knowledge.

Study 1 supplied our initial pool of mental illness category labels by using a strategy adapted from research into everyday theories of intelligence (Sternberg et al., 1981). We approached 124 people in a downtown post office (n = 23), weekend outdoor market (32), or college classroom (69). Our informants in the two public locations were 22 men and 27 women (6 did not identify their gender on our questionnaires) who ranged in age from 15 to 69 years, with medians of 41 for the post office and 36 for the market. The college sample included 21 men and 47 women (1 was unidentified) from 17 to 21 years old (mdn = 18). In order to elicit a variety of category labels, we asked volunteers to list as many kinds of either "mental illness" (n = 46), "insanity," (39) or "psychological problems" (39) as they could. Informants also used 5-point scales (1 = "very little" to 5 = "very much") to tell us how much of their knowledge came from school work, the media, and personal observation.

We retained responses that were clearly categories and discarded any that were illegible, uninterpretable (e.g., "impressions of other people," "pornography," "medication"), colloquial ("crazy"), or polemical ("voting for Bush," "making war"). The majority of discarded responses were idiosyncratic; none were mentioned by more than two subjects. Informants gave 162 valid category labels. Schizophrenia was by far the most frequent response (f = 85), followed by manic depression (44), depression (36), paranoia (32), psychosis (18), phobias (16), neurosis (14), Alzheimer’s disease (13), obsessive-compulsive disorder (13) and psychopathy (11). The majority of our chi-square analyses of the effects of survey location or stimulus label (mental illness, insanity, psychological problems) were unreliable due to small expected cell frequencies; analyses of the ten most frequently mentioned categories yielded only a marginally significant tendency for "psychopathy" to be cited as a type of "insanity." Given this, we decided to use college samples and the descriptor "mental illness" in subsequent investigations.

Reports about the sources of category knowledge tended toward the middle of our 5-point scales, with mass media and personal observation emphasized more than schoolwork: means were 3.0, 2.9, and 2.4, respectively. Multivariate ANOVAs of the effects of survey location and superordinate category label on these source ratings revealed no significant effects at the .006 level (a Bonferroni correction of .05 divided by 9 tests).

Participants were 135 volunteers (81 females,44 males,10 unidentified) from introductory psychology, statistics, personality, and physiological psychology courses. A majority of the sample (62%) had no previous psychology coursework; 50 (37%) were psychology majors and only 17 (13%) had taken an abnormal psychology course. Subjects’ ages ranged from 16 to 35 years, with a median of 19.

Informants received an alphabetical list of the 162 category labels generated in Study 1 and either rated the item’s prototypicality (1 = "I definitely would not call this a mental illness" to 4 = "I definitely would call this a mental illness") or indicated that they didn’t understand its meaning (cf. Shaver et al., 1987).

We narrowed the list of 162 category labels to 48 items that (a) had been mentioned more than once by subjects in Study 1, (b) were understood by at least two-thirds of Study 2 subjects, and (c) had a mean prototypicality rating of 2.25 or greater. Informants gave the highest average prototypicality ratings to schizophrenia, insanity, mental illness, mental insanity, and paranoid schizophrenia (all at 3.8); multiple, split, and dual personality (3.6); psychopathy (3.5); and manic-depression and psychosis (3.4). Table 1 gives the values that correspond to (a), (b), and (c) above.⁴

Study 3: Similarity Clusters of 48 Types of Mental Illness

Participants were 75 undergraduates (59 females, 16 males) who were recruited from courses in art history, English, music, biology, astronomy and introductory psychology. We asked subjects about their previous psychology courses: one-third (25) had had none, 57% (43) had taken one and the remainder (7) two or three. Subjects’ ages ranged from 18 to 32 years, with a median of 19.

Experimenters met with participants individually and read the following instructions:

Participants received a stack of 3 x 5 in. index cards which had been thoroughly shuffled. Each card contained the name of a mental disorder (e.g., "schizophrenia") followed by its adjectival form in the statement "Person who is" (e.g., "schizophrenic").

We used hierarchical agglomerative cluster analysis (Aldenderfer & Blashfield, 1984) to reveal the contents and structure of similarity sorts. This procedure offers a variety of merger or linkage rules that determine how items are clustered in multivariate space: Space contracting methods such as average linkage tend to reduce the space between clusters, while space-dilating merger rules such as complete linkage or Ward’s method tend to form more distinct clusters. Our analyses proceeded in two stages. First, we compared the results generated by different merger rules in order to identify subjects’ low distance, high similarity clusters of mental disorders--those that would occupy the lowest level in a hierarchy of similarity sorts. Then, consistent with our belief that naive categories of mental illness are stereotypic, we used the space-dilating Ward’s method to determine the higher level structure of our subjects’ similarity judgments.

In the descriptions and figures that follow, items that subjects sorted appear in lowercase. Capitalized terms are cluster names that we created to summarize our subjects’ implicit categories.

Sorters made from 4 to 21 stacks of cards, with a mean of 10.7 (SD = 3.4). Cluster analyses using average linkage, complete linkage, or Ward’s merger rules all resulted in 12 nearly identical clusters at the lowest combination distances: Eating Disorders (eating disorder, anorexia, bulimia), Compulsive Disorders (compulsive lying, kleptomania, obsessive-compulsive), Mood Disorders (depression, manic-depression, mania, post-traumatic stress syndrome), Addictive Disorders (alcoholism, drug addiction, addiction, drug abuse), Violence (rape, violence, homicide, abuse, perversion, passive-aggression, suicide), Intellect Disorders (Alzheimer’s disease, senility, mental retardation, retardation, Down’s syndrome, autism), Fear Disorders (claustrophobia, phobia, fear of leaving the house, paranoia), Multiple Personality (dual personality, multiple personality, split personality, personality disorder), Schizophrenia (schizophrenia, paranoid schizophrenia), Psychotic Symptoms (delusions of grandeur, out of touch with reality, hallucination), Insanity (insanity, mental insanity, emotional insanity, temporary insanity), and Generic Terms (mental illness, neurosis, psychosis, psychopathy).

Cluster analysis using Ward’s method is shown in Figure 1. The scale at the top of the figure gives the rescaled cluster combination distance, a benchmark generated by the SPSSX CLUSTER program; distances range from 0 to 25, with low values indicating closer associations between items in subjects’ sorts. Items in the left margin are the ones that subjects sorted; those marked with an asterisk are the most frequently cited disorders in the 10 clusters which we named at a distance of 5. The dendrogram indicates that clusters combined to form five larger groups at a middle distance: A = Disorders of Control and Mood (Eating Disorders, Mood Disorders, Compulsions, and Addictions), B = Violence, C = Intellect Disorders, D = Fear Disorders, and E = Schizophrenoid Disorders (Schizophrenia, Multiple Personality, Psychotic Symptoms, Insanities, and Generic Terms). Finally, at the highest distance, the five clusters combined to form two, with a sub-branch: I = Violence and Disorders of Mood and Control, IIa = Intellect Disorders, and IIb = Schizophrenoid and Fear Disorders. Note that the most frequently cited members of these clusters are manic depression (followed closely by depression), Alzheimer’s disease, and schizophrenia, respectively (Table 1).⁵

Multidimensional Scaling of Similarity Sorts from Study 3

Our final analysis attempted to identify the dimensions which might characterize our subjects’ similarity sorts in Study 3. We subjected co-occurrence matrices to classical nonmetric multidimensional scaling (Kruskal & Wish, 1991) using the ALSCAL program of SPSSX. An elbow test plotting Kruskal stress coefficients against number of dimensions showed that a three-dimensional solution best fit the data from Study 3: The stress coefficient was .399 for the one-dimensional solution, .218 for two dimensions, and .115 for three.

Table 2 presents the three-dimensional solution for Study 3; Figure 2 is a plot of the 10 disorders that were the most frequently mentioned members of low distance clusters (items marked with asterisks in Figure 1) plus three other disorders that were frequently cited (depression, psychosis, phobia; see Table 1). Since subjects in Study 3 did not have an opportunity to characterize the dimensions that were implicit in their card sorts, that task fell to us. We based our decisions about how to label dimensions on both common sense and the literature on attitudes toward the mentally ill. Dimension labels are therefore speculative and will remain so pending further investigation.

Dimension 1 places items such as schizophrenia, multiple personality, and mental insanity at one extreme against alcoholism, anorexia, and abuse at the other. In an attributional analysis of stigma, Weiner, Perry, and Magnusson (1988) found that subjects viewed problems such as blindness as distinct from drug abuse, obesity and child abuse on the basis of whether the target had potential control over the initial cause of the problem. They also found that onset controllability was related to the amounts of responsibility and blame that subjects assigned to target persons. In addition, Rabkin (1980) identified lack of accountability as a major component of public stereotypes of the mentally ill. Based on the parallels between our findings and these studies, we labeled dimension 1 as "Onset Controllability/Responsibility."

The second dimension isolates members of the Intellect Disorders cluster from all other items. It is interesting to note that the items that are most remote from Intellect Disorders include multiple personality and schizophrenia. Subjects may be suggesting that individuals with Intellect Disorders suffer from cognitive deficits (i.e., negative symptoms) while multiple personality and schizophrenia are problems of individuals with cognitive excesses (positive symptoms). We therefore called this dimension "Cognitive Deficit versus Excess".

Dimension 3 contrasts violence, rape, and homicide with phobias. It is possible that this dimension represents "Potency" (strong vs. weak), a polarity that typically emerges in semantic differential studies (e.g., Shaver et al., 1987). Alternatively, the dimension could represent a continuum of disorders from high to low "Severity." A final possibility invokes the concept of harmfulness. Rabkin (1980), Gerbner et al. (1988), and Nunnally (1961) describe dangerousness as central to stereotypes of mental illness. Along these lines, the strong/severe pole may also indicate harmful or dangerous behaviors.

In this discussion of our subjects’ constructions of mental disorder, we begin by characterizing the broader outlines of the similarity sorts in Study 3 and assessing their correspondence to the historical stereotypes of maniac, melancholic, and fool and the corresponding modern categories of psychosis, neurosis, and organic disorders. We look next at the finer details of subjects’ responses for clues about the values that inform current conceptions of mental disorder. Finally, we speculate about the social implications of our findings. Suggestions for further research appear throughout this section.

Our use of the terms "basic categories and stereotypes" is misleading if it fosters the impression that our research revealed groups that are clearly bounded classical sets. On the contrary, our methods and analyses, with their basis in earlier studies of the prototype approach to categorization, produced clusters that shade into one another at their edges (Rosch, 1978; Fehr & Russell, 1991). Our task is to try to characterize the core features of these fuzzy sets.

The structure of similarity sorts at distances of 15 to 25 in Figure 1 shows that there are between two and five basic clusters of mental disorder. Any decision about which number best characterizes subjects’ basic categories is admittedly arbitrary. We will proceed as if there were three: This intermediate number allows for fairly distinct groupings with fuzzy boundaries and makes it easier to assess a possible match to the three historical stereotypes and their modern equivalent categories. We will characterize these three clusters in two ways (cf. Shaver et al., 1987): We will look for commonalities among the members of each cluster in terms of the dimensions that emerged from multidimensional scaling, and we will assume that the best representatives of a basic stereotype are the disorders that were most typical in terms of frequent responses, familiarity, and high prototypicality ratings (see Table 1).

The first basic stereotype, corresponding to Roman numeral I in Figure 1, is composed of Eating Disorders, Compulsive Disorders, Mood Disorders, Addictive Disorders, and Violence. In multidimensional scaling analyses, the disorders in this cluster tended toward controllable onset and high responsibility, with modest cognitive excesses; potency and severity were low to moderate, but increased markedly for the Violence Disorders at the fringe of the cluster. The most highly visible disorders in this broad cluster, in terms of frequency of citation and prototypicality ratings, are manic-depression and depression (Table 1). This suggests that the historical stereotype of the melancholic is still extant.

The overall impression of this first basic stereotype is of a mix of what abnormal psychology textbooks used to call "neuroses" and "social problems." Subjects seem to be implying that while members of this class are currently preoccupied with modest cognitive excesses, they originally were responsible for the onset of their plight; attributional analyses suggest that these people would thus receive little pity and assistance from others (Weiner et al., 1988). An obvious exception to our characterization of cluster I as "neurotic" involves phobias, which formed a category of Fear Disorders with the addition of paranoia. Fear Disorders was a transitional member of a different cluster (IIb in Figure 1), most likely because of linguistic matching: a number of subjects sorted "paranoia" and "paranoid schizophrenia" together.⁶ Respondents’ identification of phobias with paranoia is interesting and merits further investigation, including replication.

The second basic cluster consisted of Intellect Disorders (IIa in Figure 1). In multidimensional scaling analyses, this cluster is isolated on the pole of the second dimension that we labeled as "cognitive deficit." The disorders in this group fell at the midpoints of the other two dimensions (Figure 3). The most frequently cited disorders in this basic cluster were Alzheimer’s disease, autism, and mental retardation. The correspondence of this cluster to older conceptions of folly and idiocy and to the 20^th century construct of organic disorders is clear.

The core categories of the third basic cluster were Schizophrenia, Multiple Personality, Insanity, Generic Terms, and Psychotic Symptoms, with Fear Disorders as a transitional group. Multidimensional scaling analyses defined these disorders as of uncontrollable onset and low responsibility, modest to high cognitive excess, and moderate to high potency and severity. The likely consequences of such a conception from an attributional theory of emotion would be the somewhat incompatible reactions of pity and fear (Weiner, 1979; Weiner et al., 1988). In terms of prototypical disorders, the cluster is defined especially by schizophrenia and psychosis.

This group is reminiscent of the historical stereotype of the maniac. It may also be the most central to people’s everyday conceptions of mental disorder because it contains the disorder that was by a wide margin the most frequently cited, schizophrenia, in close association with the generic "mental illness" and the various "insanities." We will discuss the centrality of schizophrenia and its conflation with multiple personality in the next section.

We conclude that there is overlap, albeit imperfect, between our subjects’ broader clusters and the modern triad of psychoses, neuroses, and organic disorders. There may be a fourth, transitional concept that encompasses Violence and Addictive Disorders, although the disorders in these categories were only modestly frequent and prototypical (Table 1). We also note that the most frequently cited and prototypical disorders in each of the three large clusters--schizophrenia, manic depression, and Alzheimer’s disease--bear a passing resemblance to the historical stereotypes of maniac, melancholic, and fool. Future investigations should assess the characteristics of these stereotypes more directly. For instance, subjects could supply modern labels in response to summary accounts and pictures of historical stereotypes; or, presented with modern labels such as "schizophrenia," "depression," and "Alzheimer’s disease," subjects could list symptoms and visible signs for comparison with older stereotypes.⁷

At first glance, our subjects’ vocabulary of mental disorders and low distance similarity judgments seem rather commonplace and uninformative. The vocabulary is a mix of colloquial and professional descriptors and similarity sorts included a lot of linguistic matches. Social constructionists and their forerunners claim, however, that cultural ideology resides in such details (Gergen, 1985). Our task in this section is to identify some of the values and assumptions lurking in our subjects’ responses. We frankly acknowledge that much of what follows is speculative and incomplete. Our findings are clues that need to be followed beyond these initial attempts to detect pattern and meaning.

Two aspects of our findings are particularly interesting. The first is subjects’ willingness to nominate violence, homicide, rape, and abuse as instances of mental disorder. At first glance, this seems inconsistent with the public outcry that regularly occurs when the insanity defense is used in trials for violent crimes, but our multidimensional scaling analyses may help to resolve the contradiction: Violent offenses are seen as potent, severe, and initially controllable acts for which people can be held accountable; such behaviors typically evoke fear, anger, and calls for punishment rather than rehabilitation (Jenkins, 1988; Weiner, 1979; Weiner et al., 1988). In addition, violent acts were only modestly prototypical of mental disorder in terms of the values presented in Table 1, suggesting that they occupy an area on the border between "crime" and "madness." Our main question, however, is about the border itself: Why are violent actions connected to mental disorder at all, both in our subjects’ responses and in other sectors of society such as the mass media (Fleming & Manvell, 1985; Gerbner et al., 1988; Hyler, Gabbard, & Schneider, 1991; Shain & Phillips, 1991; Wahl, 1992)? The answer cannot fully reside in empirical reality. Although mental disorders--especially alcohol and drug abuse--seem to be risk factors for violent behavior in recent epidemiological studies, their impact is so modest that one expert concluded that "mental health status makes at best a trivial contribution to the overall level of violence in society" (Monahan, 1992, p. 518).

The persistence of perceptions of an association between violence and mental illness may be due to the usefulness of the concept of mental illness as a "social tranquilizer." Szasz (1960) has argued that people in this culture want to believe that life is inherently harmonious and conflict-free, but it obviously is not. Real, day-to-day interactions are full of contention and, sometimes, violence. The metaphor of mental illness conveniently suggests that the problem is identified and that experts are working on it. There is no need to face the enormous task of restructuring a violent society when we can isolate a group as the Other and entrust them to professionals who, we hope, will neutralize the threat (cf. Foucault, 1979; Gilman, 1982). A "homicidal maniac" stereotype also deflects attention from real but problematic risk groups for violence such as "young males" and "family, friends, and acquaintances." These groups are problematic because they encompass the people that we know intimately in our daily life. As a result of this familiarity, it is very difficult to stereotype these groups as dangerous Others--most people would have little difficulty visualizing a generic homicidal maniac but would frown in puzzlement at the concept of a "homicidal friend."

Another interesting area of our findings concerns the centrality of schizophrenia to conceptions of mental disorder. Schizophrenia appears to be the single most prototypic mental disorder. Not only was its frequency of citation double that of any other item (Table 1), but it was rated as highly prototypical and associated in our subjects’ sortings with "insanity" and "mental illness" (Figure 1). Why was schizophrenia so predominant in our subjects’ responses?

One possibility is that the high profile of schizophrenia is a reflection of professionals’ preoccupations and interests, which is in turn a reflection of the actual severity of the disorder. Professional interest is not hard to demonstrate: For example, in textbooks of abnormal psychology, it is standard to devote two chapters to schizophrenia, while other disorders merit a chapter or less. One pair of authors justify this imbalance by characterizing schizophrenia as "such a serious disorder . . . [that it] has long-term impacts not only on those who develop it, but also on their families, and on society" (Sarason & Sarason, 1993, p. 323). Yet any explanation of the prototypicality of schizophrenia that cites the disorder’s real world costs will immediately encounter difficulties. We could make a case, for instance, that in terms of prevalence, cost to the American economy, and number of fatal outcomes, depression and drug abuse are much more serious problems than schizophrenia. Why aren’t these disorders the most prototypical disorders?

The perceived seriousness of schizophrenia may be a reflection of wider cultural values. Recall, for instance, MacDonald’s (1981) suggestion that an increase in the perceived severity of delusions and hallucinations accompanied a decline in beliefs about the potency and immediacy of the supernatural world. The loss of a superior realm which could accommodate unseen voices and strange experiences gave way to a relocation of these phenomena in an interior space. Furthermore, one interpretation of our multidimensional scaling results is that subjects saw schizophrenics as cognitively impaired victims of an uncontrollable process who are not responsible for their actions. If this is an accurate reading of subjects’ similarity sorts, then the prototypicality of schizophrenia makes even more sense: In a society that values personal agency and rationality (Sampson, 1981, 1988), the loss of control and thought disorder of schizophrenia should seem particularly severe.

Our subjects also associated schizophrenia with multiple personality (Figure 1) despite years of insistence by professors and clinicians that "A schizophrenic is not a split personality." The idea of one skin containing more than one person violates basic Western assumptions about the nature of the individual as a self-contained entity (Sampson, 1988), so that the modern-day absorption of the multiple personality into the maniac stereotype may reflect an association-by-extremity. That is, the delusions, hallucinations, and thought disorder of schizophrenia and the identity violations of multiple personality are all regarded as fundamentally bizarre and grotesque in our society (Kayser, 1963), and so they "go together." Popular culture also plays a part in maintaining the conflation. For instance, journalists and other wordsmiths know that "schizo-" means something like "split," which of course calls to mind the "split personality" (see Figure 1). Thus we see press statements such as "The administration’s foreign policy is schizophrenic," which is a metaphor intended to invoke the inconsistency of multiple personalities rather than the delusions, hallucinations, and incoherence of schizophrenia. In addition, we should not underestimate the considerable influence of such movies as Psycho (Rebello, 1990), which portrays the multiple personality problem as homicidal maniac. The identification of Norman Bates’s split personality as both "schizophrenia" and "homicidal maniac" is not a big leap: imitations of Hitchcock’s movie have such titles as Schizoid, Maniac, and Deranged.

There are other aspects of our subjects’ responses that we will only mention for future consideration. For example, some investigators may want to replicate and probe further our subjects’ pairings of paranoia with phobias and of post-traumatic stress with mood disorders. In addition, we are aware that if we had conducted this study in 1975 or 1980, certain disorders-such as eating disorders, Alzheimer’s disease, post-traumatic stress disorder may not have appeared. Repetition of Studies 1 and 2 every decade or so could give interesting historical data that could be compared with content analyses of mass media (e.g., newspaper articles on mental health, listings of disease-of- the-week movies on TV) and of professional journals in an attempt to trace the transmission of mental health information in this society. Such investigations could also separate out enduring public conceptions of mental disorder from passing trends.

In the past, studies of attitudes toward the mentally ill have assessed reactions to a generic "mental patient" and found public concerns about accountability, predictability, and dangerousness (Nunnally, 1961; Rabkin, 1972, 1980). Our research suggests that the subjects in those studies, given no other recourse, probably identified generic mental disorder with schizophrenia or the broader stereotype of the maniac. What the attitudinal research has not shown, however, is that people who are not limited to a considering a single category can conceive of different kinds of mental illness and that accountability and harmfulness may be dimensions rather than fixed aspects of a unitary stereotype. Thus, the knowledge that someone is schizophrenic should have implications for attitudes and actions that are in some ways different from and in other ways similar to those that are activated by the labels "Alzheimer’s patient" or "manic-depressive" or "alcoholic" (Weiner et al., 1988). Our research also hints that laypersons may be unimpressed by distinctions that seem obvious to professionals who are well versed in the psychiatric nosology of DSM-III-R. People surely know that there are differences between bulimia, depression, combat stress, and compulsive stealing--after all, they have different names. But on a global, impressionistic level, people also know that they "go together" in ways that are hard to articulate. Professionals’ attempts to educate the public and to communicate with clients in the first stages of treatment are quite likely to be complicated by such global impressions (Cumming & Cumming, 1957; Monahan, 1992).

Another implication is suggested by research into natural categorization. Categories are not just ways for people to organize knowledge-they help people to process information and store it in memory and they give meaning to the world (Medin, 1989; Rosch, 1978). These cognitive processes in turn influence expectancies and actions. This is evident in studies of the interpersonal effects of stigmatization of the mentally ill. For instance, Farina and his colleagues have found that individuals identified as ex-mental patients not only receive negative treatment in conversations and job interviews, but also actually elicit negative responses when they know they have been stigmatized (Jones et al., 1984). Farina’s work highlights the reciprocal nature of such interactions and reminds us that category knowledge can be possessed by and affect the behavior of those who are categorized as well as those who apply categories to others.

The interpersonal effects of mental illness stereotypes may also apply to professionals. Gilman (1988) has written that "however much clinicians (not to mention the lay public) believe themselves to be free of such gross internal representations of difference, they are present, and they alter the relationship with the patient or client" (p. 48). He cites as an example the difficulty that mental health professionals have had in identifying tardive dyskinesia and other Parkinsonian side effects of the antipsychotic medications which are used to treat schizophrenia. These disorders involve significant disruptions of motor control such as lip-smacking and tongue protrusion, and some researchers have used the word "epidemic" in describing their frequency. Why, then, has research shown that clinicians have done a poor job of identifying patients who have tardive dyskinesia (Hansen, Casey, & Weigl, 1986)? Gilman (1988) believes that the "clinicians were unable to see the gross movement disorders in their patients because their patients were supposed to ‘look crazy’ " (p. 49). In general, the possibility that professional education and experience may not eradicate prior stereotypes of mental disorder is an intriguing area for future study. The research we have reported here focused on the perceptions of mental disorder held by nonprofessionals, but it would not be too difficult to adapt it for use with mental health professionals. A comparison of the ways in which laypersons and professionals sort different kinds of mental disorders would be instructive.

Our consideration of the possibility of interactions between popular knowledge and professional conceptions of mental disorder returns us to our initial assumption that both are products of social discourse. The emphasis of this research on nonprofessional views may have fostered a false division: All conceptions of mental disorder are, ultimately, expressions of culture (Foucault, 1979). This is not a claim that professional taxonomies are specious. Klerman (1988) notes that social constructs such as mental illness, the university, and the electron

We hope that our focus on naive diagnostic categories will move readers to reflect further on the negotiated understandings that construct professional taxonomies.

Footnotes

1. Other categories that appear frequently in medical taxonomies of the past-for instance, epilepsy and hysteria-often seem to be incorporated as variations on the maniac or melancholic in art and public perceptions (Gilman, 1982).

2. Such continuity may seem implausible at first--but notice that humoural theory and its associated anatomy, physiology, psychology and cosmology still appear in English vocabulary (e.g., such adjectives as mercurial, venereal, martial, jovial, saturnine, ethereal, splenetic, galling, bilious, sanguine, phlegmatic, choleric, melancholic, etc.) and metaphor (e.g., expressions involving head, heart, and guts). In addition, Martindale and Martindale (1988) found that undergraduates unfamiliar with humoural concepts can reliably sort words associated with the four elements and four temperaments into their proper correspondences, that is, fire/choleric, earth/melancholic, air/sanguinic, and water/ phlegmatic.

3. Although our investigation adopted the methods of research into the prototype approach to natural categorization, we make no claims about the processes involved in our subjects’ categorical judgments or representations. Our primary mission is to reveal the contents of our subjects’ stereotypic judgments rather than the processes of stereotyping or categorization. This focus on content places our work closer to the interests of European psychologists (e.g., see Jankowicz, 1986; but see also Fehr & Russell, 1991, for an American defense of descriptive analysis). Note that if we had conducted this research in the mid-1980s, we could have reasonably claimed that category knowledge about mental illness is organized by resemblances to a prototype-that is, to an abstract "average" representation that consists of a set of correlated features (Medin, 1989; Rosch, 1978). However, there has been much theoretical activity and research in this area in the past few years (Cantor & Genero, 1986; Medin, 1989), so that there is no longer any clear consensus about the nature and processes of implicit category knowledge.

4. Tests for gender effects in Studies 1 and 2 turned up no significant effects.

5. We conducted a fourth study in an attempt to clarify the higher distance structure found in Study 3. Rather than ask people to sort 48 disorders, we had them sort nine cards which contained lists of the disorders that comprised low-distance clusters from Figure 1. Three clusters from Study 3 were omitted: Psychotic Symptoms, Generic Terms, and Insanity. Cluster analysis (Ward’s method) revealed that subjects sorted the nine cards into four middle-level clusters similar to those in Study 3: A = Disorders of Control, Mood and Fear, B = Violence and Addictions, C = Intellect Disorders, and E = Schizophrenoid Disorders. Two principal differences appeared at this middle level. Fear Disorders, the fifth mid-level cluster in Study 3, had an early merger with Mood Disorders. In addition, Addictions moved away from Disorders of Mood and Control (Study 3) to an alliance with Violence. The four mid-level clusters of Study 4 merged into two high-distance clusters that were similar to those in Study 3. Category I united Disorders of Control, Mood and Fear with Violence and Addictions. Category II was comprised of Intellect Disorders and Schizophrenoid Disorders. Details of this study are available from the first author.

6. Note, however, that in the follow-up study described in footnote 5, Fear Disorders appeared in the "neurotic" cluster.

7. That the maniac melancholic, and fool are alive and well and in the public eye should be evident upon reflection to anyone who watches movies or TV shows, reads newspaper comics, or scans advertisements in any of the mass media. A more formal demonstration of the prevalence of these categories comes from a content analysis of pictures in nine abnormal psychology textbooks published from 1986 to 1988 (Schoeneman, Gibson, Brooks, Jacobs, & Routbort 1992). There were 534 depictions of abnormal people in our sample; of these, the most commonly presented diagnostic categories were schizophrenia and paranoia (104 or 19%), organic brain syndromes and mental retardation (80 or 15%), and mood disorders (67 or 13%).

Aldenderfer, M. S., & Blashfield, R. K. (1984). Cluster analysis. Sage University Paper series on Quantitative Applications in the Social sciences, series no. 07-044. Newbury Park, CA: Sage Publications.

American Psychiatric Association (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.) Washington, DC Author.

Babb, L (1951). The Elizabethan malady: A study of melancholia in English literature from 1510 to 1642. East Lansing, Ml: Michigan State College Press.

Burton, R. (1621/1977). The anatomy of melancholy. New York: vintage.

Buss, D. M., & Craik, K. H. (1986). Acts, dispositions, and clinical assessment: The psychopathology of everyday conduct. Clinical Psychology Review, 6, 387-406.

Cantor, N., & Genero, N. (1986). Psychiatric diagnosis and natural categorization: A close analogy. In T. Millon & G. L. Klerman (Eds.), Contemporary directions in psychopathology: Toward the DSM-IV. New York: Guilford.

Cumming, E., & Cumming, J. (1957). Closed ranks. Cambridge, MA: Harvard University Press.

Fehr, B., & Russell, J. A. (1991). The concept of love viewed from a prototype perspective. Journal of Personality and Social Psychology, 60, 425-438.

Fleming, M., & Manvell, R. (1985). Images of madness: The portrayal of insanity in the feature film. Rutherford, NJ: Fairleigh Dickinson University Press.

Foucault M. (1979). Discipline and punish: The birth of the prison (A. Sheridan, Trans.) . New York: Vintage.

Gerbner, G., Gross, L., Morgan, M., & Signorelli, N. (1988). Health and medicine on television. New England Journal of Medicine, 305, 901-904.

Gergen, K. J. (1985). The social constructionist movement in modern psychology. American Psychologist, 40, 266-275.

Gilman, S. L. (1982). Seeing the insane. New York: Wiley.

Gilman, S. L. (1988). Disease and representation: Images of illness from madness to AIDS. Ithaca, NY: Cornell University Press.

Goffman, E. (1963). Stigma: Notes on the management of spoiled identity. Englewood Cliffs, NJ: Prentice-Hall.

Hansen, T. E., Casey, D. E., & Weigl, R. M. (1986). TD prevalence: Research and clinical differences [Summary]. Proceedings of the American Psychiatric Association Annual Meeting, 113.

Hare-Mustin, R. T., & Maracek, J. (1988). The meaning of difference: Gender theory, postmodernism, and psychology. American Psychologist, 43, 455-464.

Hyler, S. E., Gabbard, G. O., & Schneider, I. (1991). Homicidal maniacs and narcissistic parasites: Stigmatization of mentally ill persons in the movies. Hospital and Community Psychiatry, 42, 1044-1048.

Jackson, S. W. (1986). Melancholia and depression: From Hippocratic times to modern times. New Haven, CT: Yale University Press.

Jankowicz, A. D. (1987). Whatever became of George Kelly? American Psychologist, 42, 481-487.

Jenkins, P. (1988). Myth and Murder: The serial killer panic of 1983-1985. Criminal Justice, 3 (11), 1-7.

Jones, E. E., Farina, A., Hastorf, A. H. Markus, H., Miller, D. T. & Scott, R. A. (1984). Social stigma: The psychology of marked relationships. New York: W.H. Freeman.

Kayser, W. (1963). The grotesque in art and literature. (U. Weisstein, Trans.) New York: Columbia University Press.

Klerman, G. L. (1988). Classification and DSM-IIIR. In A.M. Nicholi, Jr. (ed.), The new Harvard guide to psychiatry. Cambridge, MA: The Belknap Press.

Kruskal, J. B. & Wish, M. (1991). Multidimensional scaling. Sage University Paper series on Quantitative Applications in the Social Sciences, series no. 07-077. Newbury Park, CA: Sage.

MacDonald, M (1981). Mystical bedlam: Madness, anxiety and healing in seventeenth-century England. New York: Cambridge University Press.

Martindale, A. E. & Martindale, C. (1988) Metaphorical equivalence of elements and temperaments: Empirical studies of Bachelard's theory of imagination. Journal of Personality and Social Psychology, 55, 836-848.

Medin, D.L. (1989). Concepts and conceptual structure. American Psychologist, 44, 1469-1481.

Monahan, J. Mental disorder and violent behavior: Perceptions and evidence. American Psychologist, 47, 511-521.

Nunnally, J. C. (1961). Popular conceptions of mental health: Their development and change. New York: Holt, Rinehart & Winston.

Rabkin, J. G. (1972). Opinions about mental illness: A review of the literature. Psychological Bulletin, 77, 153-171.

Rabkin, J. G. (1980). Determinants of public attitudes about mental illnesses: Summary of the research literature. In J.G. Rabkin, L. Gelb & J.B. Lazar (Eds.). Attitudes toward the mentally ill: Research perspective. Report of an NIMH workshop, January 24-25, 1980 (DDHS Publication No. ADM 80-1031). Washington D.C: U.S. Government Printing Office.

Rebello, S. (1990). Alfred Hitchcock and the making of Psycho. New York: Dembner.

Rosch, E. (1978) Principles of categorizations. In E. Rosch & B.B. Lloyd (Eds.). Cognition and categorization (pp. 27-48). Hillsdale, NJ: Erlbaum.

Sampson, E. E. (1988). The debate on individualism: Indigenous psychologies of the individual and their role in personal and societal functioning. American Psychologist, 43, 15-22.

Sarason, I. G. & Sarason, B. R. (1993). Abnormal Psychology: The problem of maladaptive behavior (7th ed.). Englewood Cliffs, N. J.:Prentice Hall.

Scheff, T. J. (1966). Being mentally ill. Chicago: Aldine.

Schoeneman, T. J, Gibson, C., Brooks, S., Jacobs, D. & Routbort, J. (1992, April). Seeing the insane in textbooks of abnormal psychology: 1. Diagnosis and gender in visual stereotypes of mental illness. Paper presented at the meeting of the Western Psychological Association, Portland, OR.

Shain, R. & Phillips, J. (1991). The stigma of mental illness: Labeling and stereotyping in the news. In L. Wilkins & P. Patterson (Eds.), Risky business: Communicating issues of science, risk, and public policy (pp. 61-74).

Shaver, P., Schwartz, J., Kirson, D., & O'Connor, C. (1987). Emotion knowledge: Further exploration of a prototype approach. Journal of Personality and Social Psychology, 52, 1061-1086.

Sternberg, R., Conway, B. E., Ketron, J. L. & Bernstein, M. (1981). People's conceptions of intelligence. Journal of Personality and Social Psychology, 41, 37-55.

Szasz, T. S. (1960). The myth of mental illness. American Psychologist, 15, 113-118.

Tillyard, E. M. W. (n.d.) The Elizabethan world picture. New York: Vintage.

Wahl, O. F. (1992). Mass media images of mental illness: A review of the literature. Journal of Community Psychology, 20, 343-352.

Weiner, B. (1979). A theory of motivation for some classroom experiences. Journal of Educational Psychology, 71, 3-25.

Weiner, B., Perry, R. P., & Magnusson, J. (1988). An attributional analysis of stigma. Journal of Personality and Social Psychology, 55, 738-748.

This research was presented at the annual meeting of the American Psychological Society, June 13-16, 1991, Washington, DC. The authors would like to thank Shelley Taylor for her helpful comments; Sara Neill, Antoinette Farah, Carolyn Hull, and Stephanie Lewis for their work on a pilot project; and the other members of the 1987 History of Insanity seminar at Lewis and Clark College--Scott Anderson, Shannon Brooks, Laura Christie, Carla Gibson, Greg Peoples, Liesl Prather, and Jason Saunders--for their contributions in reviewing literature and discussion. Correspondence concerning this research should be addressed to Thomas J. Schoeneman, Department of Psychology, Lewis and Clark College, Portland, OR97219.