Structured observation

Structured observation does impose a framework onto the social setting being observed. The problem is that the framework may be inappropriate or even irrelevant. Because of its focus on behaviour, it cannot easily study intentions of human actions, in other words, the meanings behind behaviour. Another problem is that lots of fragmentary data is gathered, which can be hard to integrate into a coherent whole. Interpretivist sociologists are often reluctant to use this method because the focus on observable behaviour often means neglecting to consider the subjective meanings that people give to their actions. This may suggest that observation should usually be accompanied with another data-gathering method, whether the research strategy is quantitative or qualitative.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 279

Quantitative research automatically brings probability sampling to mind. If we have defined the population closely and located a sampling frame, random sampling becomes feasible. The problem with observation, is that it is often focussed on an incident and as a result, we cannot know what other kinds of interaction might have gone on, and so we cannot know whether each episode was ‘representative’. It is not possible to develop a “snowball” sample either, because we are limited to observation alone. However, just as other data-gathering tools use non-probability samples, so too can direct observation. Answer (d) may be right – perhaps researchers try to steer clear of random sampling. This question is not concerned with research preference, however, but with research possibility.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 274

When respondents read a survey questionnaire, they may not always interpret particular questions correctly and they may, inadvertently, skip a key word in the question and so answer inappropriately. Usually this can be traced to a defect in the manner of phrasing the question but the point, here, is the damage has been done. “Tips and skills” on page 267 lists the most significant problems of survey research as a tool in studying behaviour, including the problem of omission. Objectivity and non-involvement (and hence little risk of “going native”) are claimed as advantages of survey methodologies.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 267

One of the problems encountered in structured observation concerns the degree of inter-observer consistency. The point is that we need to feel confident that separate observers do not see things so very differently from each other as to use different codes for essentially the same behaviour. Cohen’s kappa is a statistical measure of the degree of agreement between two people’s coding of the same situation, over and above what could have happened by chance. Naturally, the same observer might code the same things a little differently over time, so intra-observer consistency can also be a problem. These are problems of reliability, not validity.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 276, Key concept 12.3

An observation schedule is the back-bone of structured observation. It specifies the categories of behaviour to be observed and how behaviour should be allocated to those categories through a coding frame. Bryman advises (p271) engaging in a bit of unstructured observation before constructing the observation schedule proper, to get a general feel for the likely range of behaviours observable. Many of the features and rules of structured interviewing can be seen to apply in structured observation as well.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 271, 272

Bryman asks “Do people change their behaviour because they know they are being observed?” (p276) If, or when, they do, we call this a “reactive effect”. The problem then becomes one of research participants behaving other than they would in normal circumstances, rendering the data invalid. Webb et al (1966, cited on page 277) argued for greater use of unobtrusive measures of observation to minimise the reactive effect. Others (like McCall, 1984, cited on page 276) suggest the effect diminishes over time as participants grow used to the presence of the observer.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 276,277

Martin & Bateson (2007, cited on page 274) identify four main types of sampling that apply to structured observation. It is important to bear in mind that the behaviour itself is the focus of study, rather than the person exhibiting the behaviour, so samples are drawn up according to occurrences in time. “Ad libitum sampling”, records all behaviours observed in a particular time period whereas, “focal sampling” concentrates on one individual only, in a set time period. “Scan sampling” observes behaviours within a group at set intervals. Finally, “behaviour sampling”, somewhat confusingly, occurs by the researcher observing which individuals engage in which sort of behaviours. We can observe behaviour but there seems no way we can observe impulses to behaviour, like emotions.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 274

It has been observed (see the reference to LaPiere, 1934) that what people say they do and what they actually do may differ. Quite why this should be the case is outside the scope of the current question. Accepting the statement as at least having hypothetical value could suggest that surveys will elicit the truth of what people feel they are likely, or prone, to do but direct observation of their behaviour would be required to see how close their survey statements are reflected by actual behaviour. To gather quantitative data, observation needs to be structured into a standardized format in order to have measurement validity.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 268

This is the classical study (LaPiere, 1934, cited on page 273 and discussed in Thinking deeply 12.2 on page 268) of the discrepancy between what we do and what we say we do. In his study, LaPiere concentrated on the “incident” of acceptance by hotels and restaurants of Chinese guests. In other words, he did not attempt to find out why entry might be granted or refused. His work, therefore, is quantitative. It demonstrates a method whereby we can record the way people respond to particular events or incidents in a social setting. This may only refer to one point in time but will still allow for comparisons to be made because of high degrees of reliability. It should be pointed out that LaPiere conducted a survey as the counterpoint for his argument.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 273

In “field stimulation”, participants do not know they are being studied, so there is no reactive effect as discussed in the previous question. In this type of observation study, the researcher directly intervenes in and/or manipulates an element in the environment in order to observe changes in participant behaviour. Salancik (1979, cited on page 277) classified “field stimulations” as a qualitative method but Bryman (p277) believes they work better as part of a quantitative strategy because of the concentration on numbers of instances of particular behaviours.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 277,278