QUESTIONNAIRE

QUESTIONNAIRE

SURVEYS

Psychological concerns in questionnaire research Michael Kirk-Smith PhD, CPsychol Reader in Behavioural Sciences

Hugh McKenna DPhil, BSc, DipN, RMN, RGN, AdvDipEd, RNT

Professor of Nursing Both at the School of

Health

Sciences, University of Ulster,

Newtownabbey, Northem Ireland, BT370QB

The questionnaire survey is a common and important tool in nursing research.

However, there are many psychologically related issues and limitations which may not always be considered. This paper brings these issues together under four headings: theory-building; validity of self-report; measurement and analysis that is, spanning the formulation of the research, the collection and analysis of data through to the interpretation of the results. In planning questionnaire research the explanatory limitations must be recognised, the aims must be linked directly to the measures, other methods should be considered as checks, and, most importantly, professional advice should be sought during the planning. —

INTRODUCTION

As nursing research develops it is likely that nursing will become more evidence based. None the less, in terms of the history of empirical investigation, research within nursing is a relatively new phenomenon. Consequently, research-based evidence for guiding nursing practice is conspicuous by its scarcity (lBlcKenna, 1995). In a healthcare system that emphasises evidencebased practice for clinical effectiveness (Parker, 1997; Hicks et al., 1996; Stevens, 1997) this is potentially disadvantageous for the profession.

Mulhall et al. (1997) have recently drawn attention to the predominance of questionnaire surveys in nursing research and identified a general lack of research skills in their use. Following on from their discussion, this paper addresses the psychological issues regarding the use of questionnaires and ’the form of evidence that they can provide.

Questionnaire surveys are useful. However, eve if they are well-specified and structured, there are limitations in using self-report to support evidencebased nursing. If these are not understood, questionnaires may be used inappropriately and/or results misinterpreted. The modem use of questionnaires originally comes from psychology and the social sciences. Nursing researchers are typically not psychologists specialised in self-report measurement.

The intention of this paper is to draw attention to psychologically related limitations of questionnaire-based research, using examples of the psychologycal issues involved drawn from both relevant seminal and more recent psychological research. For convenience, these limitations are discussed under four sections relevant to applied nursing research, that is spanning the formulation of the research, the collection and analysis of data, through to the interpretation of the results: 1. Theory--building.

2. The validity of self-report.

3. Measurement.

4. Analysis.

1. Theory-building

KEY WORDS

Questionnaire surveys,

Psychological issues, Self-report

There are limits to the sort of theory that can be constructed from questionnaire data. People hare always asked each other questions and assessed what they havc said. Questionnaire surveys are a more effective way of doing this, in that more people can be asked and in that the answers can be subjected to statistical analysis. However, surveys are not a universal method to answer Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 Š 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

203

in scientific research. For example, scientific research may be charby the questions: 1. Is there an effect? This is an empirical question.

2. V’hat causes it? This requires an understanding of the causal processes or mechanisms involved, that is, having a theoretical understanding.

3. Can we vary it? This is the application aspect and usually involves improving the effect if positive, or reducing or removing it if negative.

These questions have to be answered in order, as there is no way in which an effect can be optimised without understanding what is causing it. How these questions are answered depends on the situation. In many cascs, experimental methods may be needed to identify definitely what causes something, and this involves the use of planned interventions and controls (this is often closely identified with the ’scientific method’).

By contrast, questionnaire surveys provide a characterisation, or diagnosis, of a situation or problem and may also be very useful in suggesting why something is happening. For example, if a survey of people’s eating habits revealed that 10% of a large sample developed cancer, while none who regularly ate apricots did, this would be strongly indicative. The survey would have suggested a causal mechanism or ‘hy-pothesis’ and a strategy of social that is, to get other people to eat apricots. intervention However, the apricot eaters might be doing something else that reduces the incidence of cancer and which has not been measured. To check that it really is the apricot that is reducing cancer, the next step might be to set up an intervention-based trial whcre two groups are randomly taken from the general population that is, they are effectively the ’same’ people.

One group is asked to eat apricots and the other eats a ’control’ fruit, which is known not to reduce the incidence of cancer. If the ’apricot’ group is shown to have a reduced incidence of cancer, then we can be pretty sure it is the apricot that had the effect. Following on from this, questionnaire surveys cannot, by themselves, suggest how a problem can be solved or cured. They can indicate ’what’ needs to be done but not ’how’ to do it.

The processes of diagnosis and cure may require very different theoretical approaches and understanding. A survey might reveal or ’diagnose’ that people with a particular lifestyle smoke more than others. Further surveys might find out more details about them, and be based on the insights of sociological and cultural/anthropological theory and research. However, no amount of further survey work will allow the formulation of a useful, that is, causal, theory, that will give guidance as to how to design a planned intervention to reduce or ’cure’ smoking. Instead, interventions are likely to involve findings and theoretical approaches which have come from experimentally based research related to changing addictive behaviours, feelings and thinking, perhaps using influence and persuasion techniques from experimental social psychology (Cialdini, 1993) and behaviour modification (Kazdin, 1989).

More generally, although they may be suggestive, surveys being ’crosssectional’ look at the current state of a population, rather than at the processes that are controlling how any individual got to that state. The situation is analogous to studying a snapshot of a city street. No amount of surveying and describing a cross-section of situations/people at a certain time can give definitive rules about how a particular person will behave in the future, or why, or how to change that behaviour. Similarly, taking several surveys at different times cannot, by themselves, identify definitely what is causing change.

Although one may suggest that several factors could affect another, it is always possible that something unknown is affecting all of them.

In terms of the four levels of theory relevant to nursing (Dickoff and

questions acterised

-

-

’

204

Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

James, 1968), questionnaires can be useful in describing names and concepts (factor-isolating theory); identifying possible explanations relating concepts to one another (factor-relating theory); and in showing correlational interrelationship among concepts (situation-relating theories). However, questionnaire methods cannot deliver a theory that can prescribe actions to reach certain outcomes (situation-producing theory), that is, tell you what to do, which Dickoff and James maintain is the most important and useful theory for an applied discipline like nursing.

Although these levels of theory give the impression of separate types of research, in practice, self-report approaches are very useful in getting ’deep’ descriptions and for tailoring evaluations to the specific situation. For example, suppose a nurse wanted to change a patient’s health-related behaviour through the use of compliance procedures. These are known, through much experimentation, to be generally applicable and effective ways of changing behaviours (Cialdini, 1993). However, the nurse would have to be sensitive to the local situation in which they will be applied; for example, knowing how people think about the behaviour you are trying to change, and so on. Here self-report methods can help tailor the application of an experimentally- tcstcd and generalisable intervention.

In summary, surveys are very useful in describing and diagnosing situations and can suggcst causes, but, depending on the purpose of the research, they may need to be backed up by inten-ention-based research for confirmation. ’

2. The

validity of self-report

Questionnaire

research is

commonly

used

to

identify

what

people do and questionnaire

what influences their behaviour. It is often assumed that

accurately reflect both behaviour and influences on behaviour. In practice, this depends on factors such as what is being asked, the situation, why people think the questions are being asked that is, their view about

responses

-

of the research and the use which will be made of it and how anonymous their comments will remain. for People are likely to be quite reliable when predicting some things when whether will be on in going holiday August. However, example they such as opinions, values and more personal aspects are being investigated attitudes more caution is needed in interpretation. that are the social and psychological factors that may affect the accuracy and effectiveness of questionnaire surveys? It may be surprising to non-psychologists that it is a long-standing and well-documented finding that expressed attitude or changes in attitude bear little or no relation to behaviour (Dickoff and James, 1968; Wicker, 1969; Lloyd et al., 1990); the reasons for this discrepancy and its resolution are a continuing topic of debate and research (Street, 1994; Kraus, 1995; Debono and Snyder, 1995; Lloyd, 1994).

Indecd, one could say that attitude theorists measure attitude to find out why it does not predict behaviour; whereas others use attitude measures presuming that it does.

One sociall; based explanation for the lack of relation between expressed attitude and actual behaviour is that the expression of an attitude is a ’speech act’ produced more in reaction to the immediate social situation; for example, reacting to the perceived motives of the researcher or interviewer, rather than being related to past behaviour or future behaviour (Lalljee et al., 1984; Abraham and Hampson, 1996). In essence, this explanation suggests there is no possibility of a ’neutral’ context where a person’s ’true’ attitude can be elicited. These ’impression management’ effects will be more pervasive the more socially or politically important the questions are to the respondent. So even the

nature

-

-

-

-

’

Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

205

behaviour, they may not be expressed in the questionnaire responses.

This may be especially so in an organisational environment (Bozeman and Kacmar, 1997). For example, in surveying attitudes of NHS managers to their organisation, two responses other than a simply ’truthful’ answer might be expected. First, most managers will simply not answer the questionnaire because they get no benefit from answering it any information leaving the confer to others.

Second, the motives organisation may competitive advantage of those (few) who do answer must be questioned. Their loyalty is to their organisation and not the researcher. Therefore, their answers to the questions may well reflect the public relations ’story’ that the organisation wishes to for example: ’INe have excellent staff communication’. This may present be the case even if anonymity means that they could, in principle, give realistic answers. iniore generally, those who choose to respond to questionnaires if ’true’ attitudes did reflect

-

-

are a

self-selected group, and may also differ in ways relevant

to

the research

question from those who choose not to answer. Given such factors, although it is possible that answers may reflect ’reality’, it is arguable whether a questionnaire survey of organisational members will give easily interpretable results.

Attitude also may not predict behaviour if it does not directly relate to the behaviour. Measures of behavioural intention or attitude, for example, whether one will buy something, predict behaviour most accurately if they are taken immediately before the behaviour to which they refer. This is not usually the case in most questionnaire studies. Ajzen and Fishbein (1977), on whose attitude theory much questionnaire research is based, point out that many situational factors must be considered in the determination of behav-

iour, and that these are usually unavailable in practice.

Experimental psychological research has also show n that people often do not know why they do things, although they can efl~ortlessl5· give apparently ’true’ reasons. For example, people can be (a) unaware of the stimulus that influenced their responses, (b) unaw are of their responses to a stimulus, and (c) unaware that the stimulus affected their response. These studies have led cognitive psychologists to question the value of self-reports about individuals’ perceptions. Nisbett and Wilson (1977) published a classic and influential review of this research that showed although people may not know the reasons for their behaviour, they are well able to provide a plausible, but unrelated, post-rationalisation. They commented that the evidence was consistent with ’the most pessimistic view* concerning subjects’ ability to report accurately their cognitive processes. Consequently, it may be quite misleading for social scientists to ask about the influences on people’s evaluations, choices or behaviour. They concluded that such reports, as well as predictions, ’may have little value except for whatever utility they may have in the study of verbal explanations per se’.

The gulf betw een what people do and what people say they do is a major problem in modem cognitive psychology. Accordingly, one of the preoccupations of cognitive psychology has been to develop alternative theories and techniques to overcome the unreliability of self-reports about mental events.

For example, Abelson (1976) pointed out that the verb concept of ’attitude’ is a somewhat abstract notion and may not relate to the ’concrete’ way in which people actually understand and act. He suggested ’script theory~’ (Abelson, 1981) as an alternative concept to ’attitude’ in understanding people’s mental representations. Script theory has led to indirect methods of understanding representations such as lexical decision-making (Faust and Babkoff, 1977) and recall times (Custers et al., 1996). A ’script’ can be Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

206

thought of as a ’package’ of commonly understood information about the routine social situations and actions typically encountered by people at work or play such as the ’restaurant script’. It can be likened to a series of pictures with associated captions. These situations and actions arc commonly the subjects of questionnaires.

The more frequently or routinely an activity takes place, the more people rely on such ’scripts’. There is much evidence that most ordinary human belayiour is routine and habitual (Langer, 1978; Slugoski, 1995) even that which appears quite cvgniti~-ely· complex (Langer et al., 1978). Habits include not only patterns of behaviour, but also patterns of thought, fantasy and emotion. Much of working behaviour often the subject of questionnaire research in nursing - is also routine and habitual, and perhaps not very involving.

By contrast, questionnaires are novel to respondents and thus require thought, and they also assume high involvement and cognition. It is arguable to what extent the type of thinking required to complete questionnaires is rel-

-

the routine of social or work situations being inquired about. People think about their attitude to a (social) object while they are behaving towards it.

In summary, the above issues indicate that the validity of the responses to a questionnaire will depend very much on what is being asked about and the situation that is being asked about. So the appropriateness of the questionnaire method to a particular research problem must be carefully considered. evant to

do

not

ordinarily

3. Measurement

Measurement is a complex and contentious issue in nursing and it is perhaps at its most contentious when addressing the measurement of self-report.

The following sections give examples of the issues involved.

-

Obsen’er bias A common justification for questionnaire surveys is to collect data and, through textual analysis, to derive a theory post-hoc (Glaser and Strauss, 1967). There is need for caution here as the act of collecting data requires a decision as to which data are appropriate, and this requires at least some idea about why they are appropriate so there must already be an ’implicit theory’ in the researcher’s mind before data collection commences.

It is impossible for a researcher to examine a situation with an unbiased mind. One’s cultural, social and educational background must Inevitably colour the nature of this implicit theory and thus how a situation will be approached for example, w hat is looked at and how the results will be analysed and interpreted.

In short, an in~~estigator’s biases and expectations can never be completely ruled out as no one comes to a situation with a ’blank slate’. This applies to all research to some extent, including the ’hard’ sciences. However, the social situations being investigated in questionnaire research are complex and unique, so there will always be a debate as to the most appropriate approach.

There is no easy- answer and the best that can be done is to be aware of the problem and make one’s implicit assumptions explicit by careful analysis and discussion, where all criticism must be examined. This may lead to uncomfortable conclusions. &dquo;

-

-

-

Selectio1l of measures The results of a questionnaire survey will depend on what has been asked in the first place. It is a matter of opinion about the number of questions to ask and why these particular questions. For - why 15, not 30 questions?

-

Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

207

needed to describe human personality depend how it is measured. Cattell’s personality theory (1981) has 16 factors, whereas Eysenck’s theory has three (1975). The humans are not different; the difference is due to the type and numbers of questions asked (and also the statistical analysis employed; see Section 4).

In summary, the results of questionnaire surveys and their interpretation are always open to debate - in other words, inconclusive in some sense.

example, the number of factors on

Errors ;11 measurement There are many ’technical’ problems associated with the measurement of selfreport. Here are some examples covering issues of question design and- contextual influences.

It is notoriously difficult to make any clear quantitative statement about For example, unless a psyit is not possiabsolute zero that is a ’ratio scale’ ble to make statements such as ’A is x°ro more than B’. It is unlikely that a questionnaire scale about opinions and so on can be assumed to have ratio properties, so such useful statements cannot usually be made. There are also problems in scale linearity; for example, is the psychological distance between 1 and 2 the same as between 9 and 10? End affects must also be considered if the maximum score is 10 and it is given, a respondent cannot give a score of 15 if a more extreme case is then encountered. In addition, the questions may not be sensitive to differences, such as ‘improved or not improved’ and may not pick up subtle differences, say, in a behaviour that has graduations.

Questions often involve comparisons. The direction of a comparison, for example of A to B versus B to A - ’Is A more important than B’ can give different results (Wanke ct al., 1995). Similarly, people may make reversal errors if left-right scales are used to rate A and B, perhaps especially if the preferred term is placed on the right (Kirk-Smith, 1995). ’Context’ effects also influence responses to questionnaires, both in the method of administration (Grossarthmaticek et al., 1993) and in the ordcr of questions (Sheeran and Orbell, 1996). In one study (Schwartz et al., 1991), a questionnaire asked a) how happy people were in their marriage and b) how happy they were with life as a whole. If the marriage question was asked first, then the correlation between the two questions w as a significant 0.67; if the general question was asked first, the correlation was an insignificant 0.32.

Thus researchers asking the same questions, but in a different order, would come to completely different conclusions.

This effect is thought to be caused by Increased cognitive accessibility or ’priming. When people have to think about their marriage, relevant thoughts come easily to mind and will influence the answer to the general question. iilhen the general question is asked first people cannot retrieve all potentially relevant information and so there is less influence on the subsequent marriage question. Such context effects are hard to predict, although the underlying ’social cognition’ processes are becoming understood (Schwarz and Strack, 1991).

Beliefs, attitudes or intentions may also be created by the questionnaires themselves, if such ’constructs’ do not already exist in long-term memory.

For example, people answering a questionnaire about redress for medical negligence may not have considcred the issue before, but the reading and answering of the questions may now create ncw constructs about the issue in their minds. In addition, constructs already existing in memory may be altered in content or importance by doing the questionnaire.

changes on a subjective scaling (Torgerson, 1958).

chological scale

has

an

-

-

-

-

-

-

Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

208

Questionnaires may often be given before and after an intervention (Cook Campbell, 1979). However, repeated experience of questionnaires influences the way that they are answered, even with a control group. For example, the various questions asked may change in relevance from pre- to postintervention. Such ’experience treatment’ interactions are uninterpretable. If a change is found before and after, one cannot be sure that this change was due to the intervention rather than a change in relevance of the questions.

These problems primarily arise from the fact that humans are not machines from which answers can be ’read out’, and because everyone’s experience is different. As a result, individuals’ ways of thinking are unpredictably changed by the methods used to measure their way of thinking, and this puts a limit on certainty in interpreting questionnaire results. and

4.

Analysis and interpretation

If a

questions it is likely that its aims are describe a situation or problem better, rather than to test specific hypotheses in which case fewer measures would likely be involved.

The aims of a multi-question survey w ill necessarily be diffuse, and the purpose of the data analysis is to look for patterns in the data, and then to speculate as to their cause.

Questionnaires may be summarised by simple cross-tabulation. However, the complex patterns and relationships between many questions are more appropriately analysed by multivariate statistical methods. The increased availability of computing power and good statistical software means that data can be readily subjected to such methods.

This ease of collecting and analysing huge amounts of questionnaire data, without a clear theoretical guidance to provide specific aims, has been picturesquely described as ’dustbowl empiricism’ or ’the notorious barrenness of multivariate data-grubbing’. Because questionnaire data can now be collected and analysed so easily there is a risk that the aims may not be clearly identifield leading to ineffective and even pointless research, as well as the problems of measurement and analysis not being addressed fully. For example, interval data might be analysed by statistical methods appropriate for ratio data, non-parametric data might be subjected to parametric analyses, or unreliable generalisations made for example, when many questions from too few subjects are factor analysed. ~Vhen data are multivariate there can be many interpretations. These depend on which method is used for example, cluster analysis versus factor analysis and to the options which are taken in using a particular for example, the decision as to the degree of correlation between method factors in factor analysis. These can completely alter the ’picture’ that comes a different selection of options might from analysis of the same data of the number change patient ’groups’ with similar attitudes from four to 14, even though the same data are input.

In summary, questionnaires allow the easy collection of data. There are also many ways of analysing these data, and the consequent possibility of many ways of patterning and grouping them. This multiplies the possible ways of making errors in interpretation. to

questionnaire

characterise

survey involves many

or

-

-

-

-

-

-

-

-

-

CONCLUSION

Although questionnaire methods are useful, the provisos and limitations described above mean that they must be used with caution. With this in mind, several questions may usefully be posed to avoid the major errors: 1. Are we expecting the results of the survey to tell us what to do, rather Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

209

than to identify where and what the problem is? An accurate diagnosis can be made, and may suggest what should be done, but will it give guidance on how to do it?

2. Do we have clear aims? It is easy to have many questions in a questionnaire and each question will be reasonable. However, the real issue is the

for each question. If a question cannot be directly linked to answering an aim or describing the sample, it should be omitted. A good way of ensuring this is to draw up mock results and checking that these arc likely to answer the aims. This will help prevent the wrong people being asked the wrong questions in the wrong way.

3. Are there other methods that might be used in conjunction with questionnaires as a check? In a working context, clinicians’ questionnaire responses about what they do might be checked with data on clinics taken and patients seen, time taken over activities, or the analysis of written records, such as care plans.

4. Finally, we hope it is clear from the examples given that knowledge of the psychological issues is important for the correct use of questionnaires. For this reason we urge nurse researchers to seek advice from, or collaborate with, psychologists, social scientists and statisticians with specialist experreason

tise in these

areas.

REFERENCES

Abelson, R.P. (1976) Script processing in attitude formation and decision making. In: Carroll, J., Payne J. (eds.) Cognition and Social Behaviour.

Hillsdale, NJ: Erlbaum.

Abelson, R.P. (1981) Psychological status ofthe script concept. American Psychologist 36: 7, 715-729.

Abraha m, C., Hampson, S.E. (1996) A social

cognition approach to health psychology— Philosophical and methodological issues.

2, 223-241.

Psychology and Health 11: Ajzen, I., Fishbein, M. (1977) Attitude-behaviour relations: A theoretical analysis and review of empirical research. Psychological Bulletin 84: 888-918.

Bozeman, D.P., Kacmar, K.M. (1997)A cybernetic model of impression management processes in

organisations. Organiza tional Behavior and Human Decision Processes 69:1, 9-30.

Cattell, R.B. (1981) Personality and Learning

Theory. Berlin: Springer.

Cialdini, R. (1993) Influence: Science andpractice.

New York: Harper Collins.

Cook, T.D., Campbell, D.T (1979) QuasiExperimenta tion: Design and analysis issues for

fieldsettings. Chicago: Rand McNally.

Custers, E.J.F.M., Boshuizen, H.P.A., Schmidt, H.G. (1996) The influence of medical expertise, case typicality, and illness script component on case

-

processing and disease probability estimates.

Memory and Cognition 24:3,384-399.

Debono, K.G., Snyder, M. (1995) Acting on one’s attitudes —the role of a history of choosing situations. Personality and Social Psychology Bulletin 21:6,629-636.

Dickoff, J., James, P. (1968) A theory of theories: A position paper. Nursing Research 17:3,197-203.

Eysenck, HJ., Eysenck, S.B.G. (1975) Manual for the Eysenck Personality Questionnaire. London: Hodder and Stoughton.

Faust, M., Babkoff, H. (1977) Script as a priming stimulus for lexical decisions with visual hemifield stimulation. Brain and Language 57:

3,423-437.

Glaser, B.C., Strauss, A.L (1967) The Discoveryof Grounded Theory: Strategies for qualitative research. Chicago: Aldine Press.

Grossarthmaticek, R., Eysenck, HJ., Barrett, P. (1993) Prediction of cancer and coronary heart disease as a function of method of questionnaire administration. Psychological Reports 73:3(1), 943-959.

Hayre, J., Parker, T. (1997) Putting research into practice: How to bridge the gap between policy and practice. Nursing Times Research 2:1, 5-6.

Hicks, C., Hennessy, D., Cooper, et al. (1996) Investigating attitudes to research in primary healthcare teams. Journal of Advanced Nursing 24:1033-1041.

Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 © 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

210

Kazdin, A.E. (1989) Behavior Modification in AppliedSettings. Pacific Grove, CA: Brooks/Cole Publishers.

Kirk-Smith, M. (1995) Handedness bias in

preference rating scales. Journal

of Market Research 37: 2, 195-202.

SJ. and the prediction of Kraus, (1995) Attitudes behavior—a meta-analysis ofthe empirical literature. Personality and Social Psychology Bulletin 21: 1, 58-75.

Lalljee M., Brown LB., Ginsburg G.P. (1984) Attitudes: Disposition, behaviour or evaluation.

British Journal 23: 233-244. of Social Psychology Langer, EJ. (1978) Rethinking the role of thought in social interaction. In: Harvey, J.H., Ickes, WJ., Kidd, R.F (eds.) New Directions in Attribution Research (Vol. 2). New York : Halsted.

Langer, E.J., Blank, A., Chanowitz, B. (1978) The mindlessness of ostensibly thoughtful action.

Journal of Personalityand Social Psychology 36: 635-642.

K.E. Do as not as I do Behavior (1994) Lloyd, I say,

17:1,131-139.

Analyst Lloyd, P., Mayes, A., Manstead, A.S.R. et al. (1990) Introduction to Psychofogy: An integrated approach. London: Fontana.

McKenna, H.P (1995) Application and dissemination of research findings. British Journal of Nursing 4: 21,1257-1263.

Mulhall, A., Alexander, C. Le May, A. (1997) Evidence for practice: Are we acknowledging the importance of methodology and context?

4, 242-244.

Nursing Times Research 2: Nisbett, R.E., Wilson, T.D. (1977) Telling more than

know: Verbal reports on mental processes. 84: Psychological Review 231-259.

Schwarz, N., Strack, F. (1991) Context effects in attitude surveys: Applying cognitive theory to social research. European Review of Social Psychology 2: 31-50.

Schwarz, N., Strack, F., Mai, H.R (1991) Assimilation and contrast effects in part-whole question sequences: A conversational logic we can

analysis. Public Opinion Quarterly 55: 1, 3-23.

Sheeran, P., Orbell, S. (1996) How confidently can we infer health beliefs from questionnaire responses. Psychologyand Health 11: 2, 273-290.

Slugoski, B.R. (1995) Mindless processing of requests — don’t ask twice. British Joumal of Social

3, 335-350.

34: Psychology Stevens, J. (1997) Improving integration between research and practice as a means of developing evidence-based healthcare. Nursing Times Research 2:1, 7-15.

Street, W.R. (1994) Attitude-behavior congruity. mindfulness, and self-focused attention—a behavior-analytic reconstruction. Behavior 1, 145-153.

Analyst 17: The Theoryand Method of Torgerson, W.S. (1958)

Scaling. New York: Wiley.

Wanke, M., Schwarz, N., Noelleneumann, E. (1995) Asking comparative questions—the impact of the direction of comparison. Public Opinion 59: 3, 347-372.

Quarterly Wicker, A.W. (1969) Attitudes verses actions: The relationship of verbal and overt responses to

attitude Journal 41-78.

25: of Sociallssues objects.

Downloaded from http://jrn.sagepub.com by Juan Pardo on November 14, 2007 Š 1998 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution.

211