In the previous chapter, we looked at the use of questionnaires as a research tool. A questionnaire is a very general method of obtaining information from people, and, as we have seen, it can be quite useful in providing us with largely factual information about people's behaviours or habits. But when it comes to obtaining more subtle information about people, questionnaires are rather more limited.
Part of that limitation comes about because of response bias, in that people are conscious of the way that they are responding to a questionnaire, and generally try to make sure that their answers project the kind of image that they want to give. Another part of the reason is that some information about ourselves is just not open to our own conscious awareness. It can be brought out by the right kind of questioning, or deduced from patterns of responses, but it isn't open to direct self‑reporting.
There are ways of obtaining much more detailed, or personal information using questionnaire‑type instruments. But these instruments are very different from conventional questionnaires, and constructed in an entirely different way. They can be sorted, loosely, into two groups: attitude scales and psychometric tests. We will look at each of them in this chapter, beginning with attitude scales.
Attitude measurement is a little more challenging than simply gathering information about someone's likes and dislikes, or their consumer choices. The main reason for this is that many people are not fully aware of their own attitudes, so they find it difficult to report them
fully. A simple questionnaire would be a completely inadequate method of measuring attitudes. Attitude scales are measuring instruments designed, as their name suggests, to evaluate attitudes. Some of them may appear like questionnaires on the surface, but the process by which they are constructed is more tightly specified and much more rigorous.
Even when people are aware of their attitudes, they still may not like to admit to them. One of the major problems of the development of attitude scales is response bias ‑ the tendency that people have to present themselves in the most favourable light. At its worst, this can mean that people simply lie, denying holding attitudes or ideas which they believe will be judged as unacceptable, or claiming to agree with socially acceptable attitudes when really they don't agree at all.
In its milder and more common form, though, response bias has much more to do with conformity than with deliberate deception. There is evidence, for example, that people of all ethnic groups respond differently to attitude surveys about colour prejudice when approached by a black researcher than they do when approached by a white presenter. While response bias may sometimes involve lying, it seems more often to be an unconscious wish to appear socially acceptable, which means that people focus on the more positive aspects of their attitudes, and less on the negative ones. As we saw in the last chapter, very few of us are consistent all of the time, and not many people hold rigidly consistent attitudes which never vary. Indeed, if someone does, there are sometimes grounds for wondering whether they are suffering from some kind of psychological problem ‑ normal human awareness is much more complex, and sensitive to its environment, than that.
One way of getting round the response bias is to introduce a response bias scale ‑ sometimes known as a lie scale. This is a set of questions, scattered throughout the attitude scale, designed to show up whether someone is making themselves out to seem better than they really are. They are quite common in psychometric tests, and are sometimes used in attitude scales too; but that depends a great deal on the form of the attitude measure. Sometimes, response bias scales simply aren't practical or appropriate.
Another source of response bias is to do with common patterns of responding. For example, we have a tendency to give consistent answers regardless of the question, and people answer 'yes' much more readily than they answer 'no' ‑ it's another aspect of our tendency to conform to others, and to avoid social confrontation. Because of this tendency, the way that the question is phrased can make quite a lot of difference to the answers. If an attitude scale is organised such that positive attitudes are always indicated by 'yes' answers, its results are likely to become unbalanced. The scale has to be carefully designed, so that 'yes' and 'no' answers are evenly balanced.
Some attitude measures are reasonably straightforward in their design and construction, while others are more indirect. In most cases, it is the straightforward ones where we need to take most precautions against response bias and other types of problem. These are less problematic for the indirect measures since, as we shall see, it is much less obvious how they work or what they are measuring. Nowadays, however, researchers tend to favour direct measures over indirect ones, and perhaps the most popular type of attitude measure of all is the Likert scale.
The Likert scale
A Likert scale is a five‑point scale, used to express agreement or disagreement with a particular statement (Figure 6. 1). Likert scales have the advantage that they can cope with different strengths of opinion, or even if someone has no opinion at all about a topic. By looking at the combined responses to different items on a Likert scale, it is possible for a researcher to obtain a measure of attitude which is often quite thorough.
I believe that ecological questions are the most important issues facing human beings today.
Strongly agree ‑Agree ‑ Don't know ‑ Disagree ‑ Strongly disagree
The construction of a Likert scale involves much more than just writing down a sentence and drawing up a five‑point scale. Technically, a five‑point scale item does not really qualify as a Likert scale unless it gives results which are normally distributed. And there is a series of definite steps which need to be carried out to construct a Likert scale properly. They help to ensure that the items which have been selected for the scale are appropriate to the topic, and a good measure of attitudes. The steps are summarised in Table 6. 1, but we will look at them in more detail here.
The first step in devising a Likert scale is to decide on the topic of investigation. That might sound obvious, but, as with any other research method, it is very easy to be too vague about what is being investigated. Strictly speaking, the topic of each item in a Likert‑style questionnaire should contribute to our theoretical understanding of the phenomenon which is being researched, so it should be possible to explain and produce a full justification of why each item has been included.
The second step is to get several people together, and to generate a broad set of statements about the topic. It's important that more than one person is involved, because it is necessary to have several
different people's points of view. They should generate far more sentences than will be needed in the final attitude scale. Once the statements have been produced, the researcher then discards repetitive or badly worded statements, and collates the rest of them together into a lengthy list. That list is presented to a different group of people, who are asked to judge whether each separate statement is positive or negative, with respect to the topic being investigated. Any statements which are ambiguous are discarded, and only those statements which all raters agree on are included.
Table 6.1 Steps in constructing a Likert scale
1. Decide on the topic of your investigation.
2. Collect a panel of volunteers.
3. Ask them to generate statements about the topic.
4. Collect all the statements and discard those which are repetitive or badly worded.
5. Put the statements into a list.
6. Ask a second panel to evaluate each statement as positive or negative.
7. Discard ambiguous statements, keeping only clear ones.
8. Select an equal number of positive and negative statements and arrange them in random order.
9. Allocate a five‑point scale to each statement.
10. Combine your statements into a full Likert‑based questionnaire.
The next task is for the researcher to choose an equal number of positive and negative statements which will contribute to the final attitude scale. These are then arranged in random order, so that there is no regular pattern of positive and negative statements which might affect how people respond. The statements are each allocated a five‑point scale, which usually goes from 'strongly agree' to 'strongly disagree', but which might be something like 'strongly believe' or 'strongly disbelieve', or some other phrases depending on the actual statement involved. The points on the scale are then assigned score values ‑ usually 5 for an extreme positive response, 4 for a less extreme positive one, and so on ‑ so that the person's responses can be collected to give a general answer.
The semantic differential
The Likert is a fairly direct type of attitude scale: The respondent is asked straightforwardly, about the topics of interest. Some attitude
scales, however, approach the question of attitudes much more indirectly. One of the more interesting examples of this is the semantic differential, which was developed by Osgood in 1952, as a way of allowing a researcher to develop a much more rounded awareness of an attitude. Instead of seeing attitudes as a purely straightforward matter of cognitive belief, Osgood took advantage of people's ability to think in metaphors, and to express their understanding of concepts by drawing parallels with other aspects of experience.
The semantic differential asks the respondent to express how the target of the attitude measure would rate on a number of dimensions, as indicated in Figure 6.2. It represents an attempt to assess the emotional and associative nuances of an attitude ‑ its connotative meaning ‑ rather than its literal (denotative) meaning. As such, it tries to capture more of the depth in someone's attitude than can be
measured by conventional scales.
angular ............. rounded
weak ................. strong
rough ................ smooth
active ................ passive
small ................. large
cold ................... hot
good .................. bad
tense ..... ……….. relaxed
According to Dawes and Smith (1985), factor analysis of a large number of semantic differential studies, conducted in twenty‑six different cultures, indicates that there seem to be three major underlying factors: evaluation (good‑bad), potency (strong‑weak) and activity (active‑passive). In fact, these appear to be so powerful that if it is known how someone rates a particular concept on these three dimensions, it is possible to make a reasonable prediction of how they will respond to the more indirect dimensions of the semantic differential. But there are many advantages of using the full semantic differential rather than just asking someone to rate items on those dimensions, the chief one being that doing so encourages the respondent to think around the topic more deeply, so they are likely to gain a fuller awareness of their personal perceptions of it.
These are just two examples of attitude measurement, but there are many others. Each method of measuring attitudes has its limitations, as well as its advantages. One of the main problems about attitude
measurement in general, for example, is that what people say does not always correspond with what they actually do.
Assumptions of attitude scales
There are three basic assumptions made by researchers using attitude scales (see Table 6.2). The first is that attitudes can be expressed in verbal statements ‑ that there is a way of putting the attitude into words. This is fairly tricky, since emotions or feelings are often a major part of an attitude, but it is possible, as we saw with the semantic differential, for words themselves to be used as an indirect expression of feelings.
Table 6.2 Assumptions of attitude scales
1. It is possible to express attitudes using verbal statements.
2. A verbal statement will have the same meaning for all participants.
3. Attitudes expressed as verbal statements can be measured or quantified.
The second assumption is that the same statement has the same meaning for all participants. People often interpret the same words very differently; but for the most part, construction procedures for an attitude measure are designed to minimise this, by using multiple raters for attitude statements and checking on the ways that statements are perceived. But as we will see when we look at repertory grid analysis, everyone makes their own sense of situations, and sometimes people see things very differently from one another.
The third assumption of attitude scales is that attitudes, when expressed in the form of verbal statements, can be measured and quantified. This assumption lies at the heart of attitude measurement, and it applies to psychometric testing too. It also lies at the heart of some objections to attitude measurement and to psychometric testing, since there are some researchers who believe that it is only really possible to look at this aspect of human experience using qualitative approaches, and that attempts to quantify the subtleties of human thinking are misguided. Others disagree, and it is a matter for individual judgement where one stands on the issue. We will be looking at qualitative approaches in Chapters 10‑13.
Attitude scales, then, are able to provide us with much more subtle information than can be obtained from a conventional questionnaire, as long as they have been constructed with attention to detail and following appropriate procedures ‑ including the piloting process which
is so important in any kind of test or questionnaire development. In a sense, attitude scales are a kind of half‑way house between the ordinary questionnaire and that even more rigorous and specialised kind of research tool, the psychometric test.
……from Nicky Hayes (2000) “Doing Psychological Research” Open University Press