Asking Silly Questions
Chapter 6 in Armistead, N. (Ed.) Reconstructing Social Psychology, Harmondsworth: Penguin.
This article takes as its starting point the frustration that many people feel when they try to answer attitude questionnaires. it attempts to do justice to those who have laboriously tried to reassert their own attitudes, feeling them unrepresented, or even caricatured by the items offered. For the most part these criticisms, added as comments at the end, or expressed in other ways, have been ignored by psychologists. For one thing they are not easy to score, but more importantly the subject is not considered qualified to criticise the questionnaire designed and standardised by the expert psychologist.
A personal experience is relevant here. Some time ago I was asked, as a subject, to fill in a questionnaire drawn up by a well-known social psychologist. It was entitled 'The Student and Society'. I recall a feeling of intense annoyance on finding the following item: ' I am in favour of destroying the present political system even without knowing what will replace it'. My pencil hovered nihilistically over the response 'strongly agree', then wavered uneasily between 'mildly agree' and 'cannot answer'. In vain did I search the questionnaire for a reasonable expression of my views. Caricatures of extremity and moderation predominated, intermingled with potted homilies and laughable simplifications. I added a side of my own views on the back, but I doubt whether they were quantifiable, as that term is generally understood by psychologists.
In this article the theory underlying conventional attitude measures is examined. The efforts of psychologists to avoid ambiguity and bias in the design of attitude statements are criticised. Three examples of objections from subjects to attitude questionnaires are examined. Finally it is asked whether an alternative rationale for the measurement of attitude is needed, and its implications discussed.
Attitude scales typically comprise a number of brief statements gathered and standardised by the psychologist. For instance, the subject may be asked to register degrees of agreement or disagreement with the phrase:
'Everyone in this country should be treated equally regardless of colour', and so on. From his pattern of responses a numerically precise score on a scale can be calculated to mark the subject's racial attitude. Often subjects, unappreciative of the rationale of psychology, have complained that such statements are too brief and the response categories too limited. Despite the brevity of the statements subjects have claimed them to be frequently ambiguous or biased.
Psychologists have generally been agreed that an attitude item should be precise, since otherwise its measuring ability is impaired. Ideally it is seen as having a fixed, externally ascertainable value, just like a physical stimulus, a weight or a light. Thus the response made can be checked against the real value. In fact one of the founding fathers of attitude measurement, Louis Thurstone, began as a psychophysicist, transferring physical judgemental concepts to the study of attitude (Thurstone and Chave, 1929). This helps to explain why the objections that subjects have made to items have been undervalued. The subject cannot, after all, 'disagree with' a stimulus. He either perceives it correctly or incorrectly. The subject is asked to express agreement or disagreement with the attitude ‘behind' the stimulus item. That the item adequately represents the attitude in question is left to the psychologist and his standardising procedures. To bring together the attitude behind the statement and the statement itself could involve viewing the attitude as a communication, since the statement of attitude clearly has these properties. This would complicate the simple view of the statement as stimulus. The idea that social anti behavioural phenomena, like attitudes and attitude statements, should be dealt with as discrete, quantifiable, stimulus-objects, is of course a basic tenet of behaviourist theory.
In addition to evoking objections from subjects these stimulus-like attitude items have run into other difficulties. Unlike physical stimuli they have no external referent; their referent is the 'internal' attitude. The determination of the value of such stimuli must thus be done through the judgement of subjects. But experiments have shown that subjects of different attitudes make different evaluations of attitude statements (Hovland and Sherif, 1952). Moreover the differences are complex and do not conform to a simple formula (Selltiz, Edrich and Cook, 1965; Zavalloni and Cook, 1965). More recently it has been shown that subjects themselves can be aware that there is an intrinsic lack of precision in the perceived values of attitude statements; that is, they not only make different judgements from other subjects, but know that they are doing it (Dawes et at, 1972).
Thus the criticisms raised by subjects do find a parallel in studies on the judgement of attitude statements. Maybe the subjects have realised, much sooner than the psychologists, that there is an inevitable ambiguity attached to many, if not most, statements of attitude, and their criticisms should be taken seriously. To examine this criticism is not easy. Data have often been discarded and never, to the writer's knowledge, sought systematically. But there are enough examples to at least start the discussion.
The first concerns a judgemental study rather than an attitude survey. Subjects were asked not whether they agreed with items, but what attitudes they thought the items signified. The psychologist, E.D. Hinckley, was in fact constructing a scale for measuring 'attitudes toward the social position of Negroes'. Subjects were asked to rate items on a 1-11 scale according to their favourability-unfavourability. The study was strictly carried out: 'every student was required to stay in the classroom until the end of the hour, even though he had finished rating' (Hinckley, 1932, p. 287). Thus there was perhaps little room for objection. But, in the final sample, there was a group, mainly black, whose ratings were considered deviant. These subjects had rated many items as extremely unfavourable, many as extremely favourable, and few in the middle categories. They had ' polarised' their ratings. The others had spread their items evenly. These subjects were considered to show poor discrimination and carelessness and were excluded from the sample. In addition, as might he expected, Hinckley also excluded items that could be said to be faulty:
“It is necessary that the statements which are used in the final scale be clear, concise and unambiguous, and that they contain no irrelevant elements . . . The care which was taken in this refining process cannot be overemphasised. The search for statements and the reduction of the original list extended over a year (1932, p. 286).
Subsequent work suggests that he was not only wrong in excluding the 'objecting' subjects, but that he was also wrong in including some of the items that he did, particularly in the light of more recent studies. Readers may like to judge for themselves whether the following of Hinckley’s items meet the criteria 'clear, concise and unambiguous'
A wide-awake Negro is physically superior and in other respects equal to the white man.
The Negro is fully capable of social equality with the white man, but he should not be so recognised until he is better trained.
The rich spiritual life of the Negro compensates adequately for the defects in his nature.
Subjects in the early sixties recorded substantial shifts in the unfavourable direction in the rating of these statements. They were, with many others, removed from the item pool on the grounds that ' many of the supposedly favourable items had a strong derogatory or condescending undertone' (Selltiz, Edrich and Cook, 1965, p.409). But what remains to be explained is how the items got included in the first place. The solidly favourable ratings that they gained among white subjects, and blacks if those excluded are forgotten, and their survival of the elimination procedures, suggest that their derogatory overtones, their ambiguity and bias may really have been hard to detect in the racialist culture of the times. (Many of the colleges used in the study came from Georgia, Virginia, Florida and Carolina.) The disparity in ratings of different attitude groups, the obvious presence of bias in the items, seems to have appeared much more markedly later on. If this is the case then the 'loading' of these items has come about through the social investment of certain phrases with derogatory affect; wide-awake', 'better trained', ' rich spiritual life'. If bias in an item can result from the change in meaning of a couple of words then the attitude item, brief as it may be, becomes a complex phenomenon. Hinckley's subjects only objected in an indirect sense, as has been said. But, partly through their actions, at the time written off as 'careless and showing poor discrimination', it has been shown that ratings of statements can differ markedly between people of different attitudes. This rating difference which effectively denotes ambiguity in the statement can, it seems likely, appear at different stages in time. Thus the task of producing unbiased, unambiguous statements becomes more difficult. So in one instance subjects making awkward evaluations of attitude statements, who were removed from the sample, were subsequently found to be justified.
A more obvious example of objection to it questionnaire, and of the psychologist's response, is to be found in Eysenck (1954). He says:
“[some] respondents claimed that the questionnaire forced them to record complex attitudes as simple dichotomies; many said that they could 'write a book' on each of the statements. . . . In a number of cases the writer himself discussed questionnaire answers with respondents When asked to amplify their answers, either in writing or verbally, the response was very scant; the elaborations put forward did not justify the claims of the respondents. Nor would they, in the writer's opinion, have altered the position on the attitude scale marked by the respondents. Often remarks were frankly irrelevant: sometimes they merely restated in a slightly bombastic fashion what the questionnaire statement had said more simply; frequently they consisted of unimportant quibbles over wording,” (pp. 125--6).
Examination of Eysenck's list does, in the writer's view, evoke some sympathy for the critical respondents. Without knowing which items they found objectionable, or what the content of their objections was, here are some of the items which might strike people as unsatisfactory. The response categories allowed were; 'strongly agree’, 'agreee on the whole', 'cannot decide', 'disagree on the whole', 'strongly disagree':
Ultimately private property should be abolished and complete socialism introduced.
All human beings are born with the same potentialities.
The dropping of the first atom bomb on a Japanese city, killing thousands of innocent women and children, was morally wrong and incompatible with our kind of civilization (1954, p. 278).
In the first sentence the words 'abolish' and ' introduce' have a kind of legalistic ring which seems out of key with the general sentiment. The second statement really appears to be a trick one. Its fact it is very reminiscent of the jibe about '100 per cent environmentalists' that is heard in the debate about heredity and environment. The third sentence appears to have an internal contradiction since it was 'our civilization' that dropped the bomb. The trouble is that the sentences do form a recognisable caricature. if the subject is tempted away from the 'cannot decide' category and puts 'agree on the whole' then according to Eysenck’s method of scoring the response is treated precisely the same as if he had marked 'strongly agree' (1954, p. 276). As previous results suggest, the perception of ambiguity and bias is by no means universal. Others might find the statements above quite lucid expressions of opinion, with which they can agree or disagree unambiguously.
As with Hinckley, Eysenck cannot be accused of completely ignoring the problem of bias. He quotes examples of different answers obtained from variations of the same basic question. But he does treat the problem as less than severe, as can be seen from the statement: 'fortunately it is usually easy to detect biased wording by simple inspection' (1954, p. 57).
For the third example of objections to attitude items it was possible to acquire a sample of completed questionnaires and examine the comments written in by the respondents. Though this is a somewhat unsystematic way of collecting data it was thought possible to test the three following rival explanations of objections to items. If objections are based on some characteristic unrelated to attitude, like 'awkwardness', then there should be an even distribution of object ions. If it is based on strength of attitude then those at both ends of the scale will object most. if it is based on subjects of a particular attitude objecting to particular items then a more complex picture may emerge. The forty-two objectors were subdivided into three groups according to the attitude under examination, pro-religious, neutral and anti-religious. This was done on the basis of measured attitude and indices of church membership and attendance. Within the total sample the anti-religious objectors numbered 17 out of 166, or 10.2 per cent, the neutral objectors numbered 14 out of 160 (8.8 per cent) and the pro-religious objectors numbered 11 out of 5l (21.6 per cent).
Table 1 Characteristics of subjects objecting to religious attitude survey
subjects own attitudes
anti-religious neutral pro-religious
raw score 40-63 64-103 104-127
total subjects in each category 166 160 51
number of objectors in each
category 17 14 11
percentage of objectors 10.2 8.8 21.6
memberships of objectors 0 4 17
church attendances in a month
of objectors 0 8 88
percentage of items criticised 20.5 17.7 25.0
Memberships included denominations and other groups. Many of the pro-religious sample had two memberships
The picture that emerges suggests that although there are subjects of all attitudes who write comments, and that neutral subjects are, marginally, the least critical, the pro-religious subjects are far more prone to write comments than anyone else. When quantity of commentary is allowed for this effect becomes more pronounced. The writing of long explanatory scripts was entirely restricted to this category.
The further questions that should be asked is how valid are these criticisms. Criteria of validity are difficult, but it can be asked whether the objections appear to rise above the level of 'irrelevant', 'bombastic' and 'unimportant quibbles over wording'. Beyond that it can be asked whether there are any agreed criticisms of particular items within or across groups.
An item-by-item analysis shows that the pro-religious subjects have most comments to make per questionnaire, followed by the anti-religious and the neutral subjects. Furthermore the pro-religious subjects concentrate their criticisms on certain items while the other subjects spread their comments evenly. Six items (out of twenty-one) received comment from at least half of the pro-religious subjects, and five items received from them no comment at all. In both the other groups no item was criticised by more than half the subjects and all the items received some comment. Thus the pro-religious group was much more 'agreed' on its criticisms than the other groups. Further, of interest to the present argument is the fact that these criticism of an item by one group was no guarantee that it would be criticised by another. Two items were criticised by all groups:
'If you lead a good and decent life it is not necessary to go to church' and 'Parents have a duty to teach elementary Christian truths to their children'; and one neutral item went virtually uncommented. But many items received uneven criticism. At least two were heavily criticised by the pro-religious but not by the others: 'The existence of disease, famine and strife in the world makes one doubt some religious doctrines' and ' International peace depends on the world-wide adoption of religion'. Two items were criticised by the non-religious, but not by the religious: 'Jesus Christ was an important and interesting historical figure, but in no way divine' and 'Religious faith is merely another name for belief which is contrary to reason'. This last finding is important since it suggests that there would be difficulties in redesigning the questionnaire to suit all groups. Also relevant to this are differences of criticism made about the use of the term 'religious', illustrated by these quotes from pro and anti subjects respectively, one finding the term too broad, the other too narrow:
“...does the term religious mean Christian belief or the wide spectrum of religious belief?”
“...the term ''religion'' is very narrow The survey does not allow for mystical or spiritual experiences outside the framework of religion.”
In terms of content of criticism the most frequent complaint was that items were vague and ambiguous. Following these were comments concerning items being 'weighted', 'loaded', 'awkward' or 'biased'. In many instances subjects complained about items with which they might have been expected to agree. One highly religious subject commented that an item favouring religious education implied 'a sort of Sunday School network dominating society'. He responded 'uncertain'. The isolated situation of the subject and the way in which the subject wished to treat the questionnaire as communication was well illustrated by the comment: ' In general debate I would not contemplate many of the statements as phrased, but there is little room for discussion with a survey '.
In general the criticisms put forward were serious and coherent, though a few were flippant and merely hostile. in particular the comments of the pro-religious subjects were strenuously argued and backed up with Bible references. The broad point which, in the writer's opinion, the criticisms succeeded in making, was that the framework of the questionnaire did not satisfactorily allow the subjects concerned to express their own viewpoint, Neither this discussion, nor of course the questionnaire, is concerned with the justification or consistency of the attitudes being measured. But what is claimed by attitude scales, and what its this case was not accomplished, was an adequate description of all the attitudes concerned. The objectors certainly formed a minority of the respondents, but if the behavioural indices of attitude shown its the table are considered, for the pro-religious at least, they consist of a minority whose attitudinal involvement and activity merit recognition beyond their numbers. These criticisms were not generally trivial or bombastic, points of considerable interest were made, and, particularly among the pro-religious group, there was agreement on a number of points of criticism.
These three examples of studies of different attitudes have registered, in one way or another, dissatisfaction with items on attitude scales. But they cannot be said to be conclusive. Data was not directly sought and the nature of the objection was not always clear. The crucial hypotheses have not yet been put to systematic test. However, the indications are there and further studies are clearly needed in 'which respondents have adequate opportunity to be involved in the designing of the questionnaire.
In what direction could these studies lead, beyond confirming the results already suggested? There are two broad points which are suggested and need further elaboration. The first is the distinction between ambiguity and bias. The second is the role of the social group in the generation of meaning. For both these points there are important leads from the data surveyed already.
It was quite noticeable, though no one remarked on this distinction, that there were two kinds of objection, made both by psychologists and by subjects. Hinckley warned of the problem of ambiguity and Eysenck of bias (though neither avoided them, as has been seen). Among the subjects many complained of vagueness and imprecision (i.e. ambiguity) and there were also complaints of 'loading ', ‘weighting' and 'bias' (though probably both ambiguity and bias occur in the subjects' everyday expression of attitude). It is possible to suggest here a distinction between ambiguity and bias, conceptually, experimentally and in terms of social usage.
Ambiguity is taken as denoting imprecision in a broad sense. In its simplest form an ambiguous statement is two-sided : it might mean one thing or it might mean another. Bias, an attitudinally more pointed form of ambiguity, has a superficial two-sidedness which contains within it an indication of the dominant intended meaning. The two-sidedness becomes the two-faced. A crude analogy is that of a dice whose numbers have been partly worn away in comparison to one which is loaded.
Thurstone gave an effective way of denoting the ambiguous statement. This type of statement receives widely differing evaluations from subjects, some seeing it as unfavourable, some as favourable. This property can be given a simple numerical index, the 'Q-value' or ambiguity rating (1929, p. 37). An example from his study is: 'I do not believe in any kind of church but have never given the matter serious thought'. What attitude lies behind this statement is unclear: it is rather neutral, vague and ambiguous. Neither Thurstone, nor his junior co-worker Hinckley, investigated the problem of bias, which involves the different evaluation of statements by people of different attitude. They thought it did not occur. The biased statement has two sides which are differently perceived by those of different attitude. Experimentally it is likely to receive a higher Q-rating than the simply ambiguous statement. This is certainly the case for the items from the Hinckley study quoted earlier. Evaluations of these statements were shifted three or four points on the eleven-point scale of favourability. This confirmed their apparent quality of 'derogatory undertones'. The shift in evaluation was especially marked for the groups of black subjects, understandably. Here the distinction between ambiguity and bias becomes clear. Ambiguity can be denoted by differences in evaluation irrespective of the attitudes of the judges. Bias, on the other hand, is denoted by differences in evaluation which have particular reference to the attitudinal groups from which the judges are drawn. The black subjects in the study quoted were thus particularly aware of the unfavourable connotations of the statements concerned. The prejudiced white subjects, on the other hand, were not that aware of them and so gave much more favourable ratings. They were not, apparently, aware of any internal inconsistency in the statement which was obvious to the other subjects, and to the experimenters. It can be said that this very inconsistency within the statement prompts the unfavourable rating. If a person attempts to dress up an unfavourable attitude in favourable terms then he can be assumed to be doubly unfavourable. It is interesting to note in passing that most examples of biased statements seem to be phrased in this direction. They involve the partial covering up of an unfavourable attitude in pleasant words, though there is no reason why it should not be done the other way round.
It may have been noticed that this discussion of ambiguity and bias has leant strongly on the very tradition of attitude measurement that it seeks to bring into question. Obviously the existing tradition of research has to be studied, but a further discussion is needed. This concerns the social groups within which attitudes are developed. That attitudes are generated within social groups is hardly a controversial point to make. However, when it is examined carefully it does come into conflict with the techniques of attitude measurement used to date. In the data examined previously it was very noticeable that the pro-religious group is distinguished as clearly, if not more so, from other groups by the behavioural criteria of religious membership as by the scalar index of attitude. Furthermore in the studies by Zavalloni and Cook (1965) and by Selltiz et al. (1965), it made very little difference to the results whether the subjects were divided up according to measured attitude or according to the groups to which they belonged. Indeed in one study the results were more marked when subjects were divided according to membership (Zavalloni and Cook, 1965, p. 46). If it is the case that subjects' own group-membership patterns are as informative about their behaviour as their measured attitudes then the whole effort of constructing an attitude scale has not accomplished a great deal. Since membership of given groups appears, in these instances, to provide an adequate alternative metric of attitude, a more 'natural' scale moreover, then perhaps psychologists should pay more attention to the attitudinal consequences of membership than they have done traditionally.
Further it can be pointed out that membership of groups often provides quite specific information as to attitudes held. Zavalloni and Cook quote as one of their membership groups college fraternities that had taken 'pledges' in campaigns against the admission of Negro students. If the content of such a pledge were incorporated into an attitude inventory it would presumably evoke maximum agreement from the signatories. Likewise the other organisations in their study would have been likely to require some degree of formal adherence to attitudinal conditions of membership. Examples are numerous. In the religious sphere it could be presumed that clergymen 'would 'strongly agree' with the 'Thirty-nine Articles' were they included in a questionnaire. In Eysenck's political-attitudes study there are items more or less resembling the policy statements of political groups to which his subjects could be expected to belong.
In a sense then an attitude questionnaire can be seen as selections from the policy statements of the groups whose attitudes are being surveyed. But in practice it does not work like this. This is because one of the main reasons for forming an attitude group is that the attitudes involved are not those of the world at large, either in content or style, and thus are precisely those attitudes which are not likely to be included in a questionnaire. The more distant from consensus thinking is the group then the less likely is it that its own statements of attitude will find their way onto a 'general' questionnaire.
Any group will, of course, formulate its programme, pledge or conditions of membership in a form which is internally favourable and consistent. Correspondingly it will portray opposing groups in terms which are unfavourable, and internally inconsistent, self-discrediting and biased. To try to construct an attitude questionnaire from statements drawn from the programmes, say, of the John Birch Society and the Black Panther Party, covering equivalent ranges of content in each case, would be difficult. But if an attitude scale claims generality then this is what it must try to do. If, on the other hand, the questionnaire seeks to avoid group-specific terminology (or 'jargon') and aims solely at consensus viewpoints, then it will effectively omit, or at least misrepresent, attitudinally significant groups.
If this outline is correct then an explanation can be offered for a puzzling feature of the criticisms of the religious-attitude survey discussed above. The anti-religious subjects offered fewer criticisms of the questionnaire than the pro-religious subjects, though in scalar terms they were as far from the neutral attitude. An explanation of this may he that there were among this sample no relevant membership organisations from which particular formulations of anti-religious attitude could be developed. Indeed the absence of such organisations is reflected in the very structure of the anti-religious attitudes which have been found to be more vague, more fragmented and 'less probing' than those at the pro-religions end (Poppleton and Pilkington, 1963; Fraser and Stacey, 1973) Such dimensional asymmetry should not, ideally, occur in a scale of attitude But in this instance it finds a simple explanation in terms of the lack of anti religious membership groups.
It may be objected that only a minority of people in any sample are likely to belong to appropriate membership groups. In a sense this is true, though many studies have found widespread membership among subjects. However, membership is only the most formal and extreme form of a more general phenomenon. A person need not be a 'card carrying member' to share the attitudes of a given group. It has been pointed out that membership groups and reference groups act as psychological perspectives (Shibutani, 1967). They create an internal consensus, a ‘universe of discourse', within which their own view of the world pertains. Those sharing that view are 'members'. The membership group is the most socially obvious form of this phenomenon. It represents the point at which attitudes begin to form frameworks for action, to implement the 'aims and objects' of the group.
The relationships between groups are communicated by attitudes. In this sense, as was suggested earlier, the attitude can be seen as a communicative act, aimed at defining the relationship between people or groups. Each group will attempt to 'define the situation' in such a way as to make its own attitudes the most consistent. It will attempt to capture the consensus to its own way of thinking. Seen in this way the attitude statement cannot have the property required of a stimulus, a publicly observable, unambiguously quantifiable value. It is intrinsically part of a debate. Likewise the attitude itself cannot be seen as some inaccessible entity in the 'black box' of the individual. Statements 'emanating' from it cannot thus be seen as simple stimuli from which the larger attitude is inferred. The attitude and its statements are all part of the debate. This approach might appear to put attitude wholly in the realm of ideas, divorcing it from the solid biological reality in which many psychologists have attempted to ground the concept. A debate after all does concern ideas and is, theoretically, quite swiftly resolved. But the debates between attitudinal groups are grounded in the real problems of the social world. The resolution is thus an extremely slow process, depending on the ability of particular groups to succeed in overcoming the problems they set themselves. This view of social groups as collective attempts to overcome real problems is well outlined by Hans Toch (1971) in his book on social movements.
What the psychologist has done traditionally in measuring attitude is to take the line of least resistance among the groups battling to define the situation. He has, without justification, suggested this as an objective measure of attitude. In reality all he has achieved is a consensus measure, unable to cope with the alternative perspectives, with the sub-consensual.
The thesis that attitudes can be objectively measured on a precise interval scale (Thurstone and Chave, 1929) can be derived from behaviourist theory, which treats attitude statements as quantifiable stimuli. This has already been discussed. But a very similar thesis can also be derived from a socio-political theory, one which maintains that there is only a single valid consensus of meaning within a given society, that from the point of view of the community of meanings we are 'all one nation'. though there are differences of attitude these exist within an agreed framework; they can be measured on the same scale. Deviant perspectives are denied, so the plurality of attitudes does not offend the accepted structure of meaning. The socio-political theory concerned is conservative and constitutional. It admits only of gradual change from recognised differences of viewpoint.
This convenient coincidence of academic and political ideology has distorted the study of attitudes in quite a marked way, I would suggest. But more serious is the way in which measurements of attitude have been used under the slogan of 'public opinion counts'. Opinion polls, containing the most limited and biased selection of items, administered doorstep by doorstep, are presented as 'the voice of the people'. No doubt the designers of ‘proper' attitude scales would deny connection with the opinion pollsters. I hope that enough evidence has been presented here to show that such connection cannot be denied, and that the failure of psychologists to critically evaluate their subject has opened the way for manipulative exercises of this sort.
Finally, I return to the starting-point, the respondent in a survey, confronted with a questionnaire. Any person asked to participate in a survey has the right to expect that it will contain an unambiguous and unbiased statement of his own attitude. Maybe this expectation will make it very difficult to continue to design questionnaires of general applicability in their present form. In the meantime the infuriated fillers-in of questionnaires should be recommended not only to write in what they find wrong with which statements, but they should also insist that their objections be treated seriously by the psychologist concerned. This may serve to impress on him that attitudes are really part of a debate, from which the psychologist himself is not free to withdraw, and convince him that it is a caricature and a prejudice to foist on people consensual formulations.
 I am grateful to Professor Selltiz for making available a list of the items used in the original study by Hinckley, together with the ratings gained in that and later studies.
 I am grateful to Judith Gould and Margaret McCourt for giving me access to this data.
 This distinction between the consensual (of generally agreed meaning) and the sub-consensual (of specific meaning to the individual) is well illustrated by Kuhn and McPartland's study of self-attitudes (1967). There is no reason why the term 'sub-consensual' should not apply to the group also.