Just like survey research, structured observation necessitates decisions about sampling. However, with structured observation, issues surrounding sampling do not revolve solely around how to sample people. Several other sampling issues are involved.
Sampling people
When people are being sampled, considerations very similar to those encountered in Chapter 8 in respect of probability sampling come to the fore. This means that the observer will ideally want to sample on a random basis. In Croll and moses’s (1985) research on children with special educational needs, thirty-four classrooms from a number of different schools were selected for observation. All children were within the same age range. Initially, each teacher was interviewed to determine which children in her or her class were regarded as having special needs. In addition, test of both reading ability and non-verbal reasoning were administered to children to identify those who appeared to have special needs, but who had not been identified by the teacher. Up to six children with special needs in each class were then randomly sampled as being the focus of structured observation; so too were four children who had not been identified as having special needs---these children acted as a kind of control group. In this way, 280 children were sampled, of whom 151 were identified as having special needs; the other 129 served as control subjects. The teacher did not know exactly which children were being observed. Each child was observed for a few minutes, after which the observer proceeded to the next child to be observed in a predetermined random order. In the end, each child was observed for a total of two hours. This was made up of a large number of short observation periods. In the research reported in Research in focus 12.1, the six students in each class were selected randomly but with the stipulation that three boys and three girls would be sampled. In the study of job characteristics (Research in focus 12.2), the individuals who were observed at work were randomly selected (Jenkins et al. 1975).
Sampling in terms of time
As implied by the idea of time sampling (see above), it is often necessary to ensure that, if certain individuals are sampled on more than one occasion, they are not always observed at the same time of the day. This means that, if particular individuals are selected randomly for observation on several different occasions for short periods, it is desirable for the observation periods to be randomly selected. For example, it would not be desirable for a certain pupil in a school classroom always to be observed at the end of the day. He or she may be tried, and this will give a false impression of that pupil’s behaviour. In the research reported in Research in focus 12.1, each child was observed at separate times on three different days. As a result, the researchers’ ratings of any child are unlikely to be distorted by unusual behaviour that he or she might exhibit on just one or two occasions.
Further sampling considerations
The sampling procedures mentioned so far conform to probability sampling principles, because it is feasible to construct a sampling frame for individuals. However, this is not always possible for different kinds of reason. Studies in public areas, like the research on superstition mentioned above, do not permit random sampling, because we cannot very easily construct a sampling fame of people walking along a street. Similarly, it is not feasible to construct a sampling fame of interactions. Reise (1976), for example, has written about the difficulty of developing a random sample encounters between police officers and the public. The problem with doing structured observation research on such a topic is that it does not lend itself to the specification of a sampling frame, and therefore the researcher’s ability to generate a probability sample is curtailed.
As suggested in Chapter 8, considerations relating to probability sampling derive largely from concerns surrounding the external validity (or generalizability) of findings. Such concerns are not necessarily totally addressed by resorting to probability sampling, however. For example, if a structured observation study is conducted over a relatively short and span of time, issues of the representativeness of finding are likely to arise. If the research was inducted in schools, observation conducted towards the end of the school year, when examinations are likely to loom large in the thinking of both teachers and students, may affect the results obtained compared to observations at a different point in the academic year. Consequently, consideration has to be given to the question of the timing of observation. This potential problem was dealt with in the ORACLE research by ensuring that teachers and each target pupil were observed on six occasions during each of the three school terms. Furthermore, how are the sites in which structured observation is to take place selected? Can we presume that they are themselves representative? Clearly, a random sampling procedure for the selection of schools may assuage any worries in this connection. However, in view of the difficulty of securing access to settings such as schools and business organizations, it is likely that the organizations to which access is secured may not be representative of the population of appropriate ones.
A further set of distinctions between types of sampling in structured observation have been drawn by Martin and Bateson (1986) between:
• ad libitum sampling, whereby the observer records whatever is happening at the time;
• focal sampling, in which a specific individual is observed for a set period of time; the observer records all examples of whatever forms of the behaviour are of interest in terms of a schedule;
• scan sampling, whereby an entire group of individuals is scanned at regular intervals and the behaviour of all of them is recorded at that time. This sampling strategy allows only one or two types of behaviour to be observed and recorded; and
• behaviour sampling, whereby an entire group is watched and the observer records who was involved in a particular kind of behaviour.
Most structured observation research seems to employ the first two types: Flanders’s FIAC scheme is an example of ad libitum sampling; the research by Galton et al. (1980), Croll and Moses (1985), Blatchford et al. (2003; see Research in focus 12.1), Jenkins et al. (1975; see Research in focus 12.2), and the research by Buckle and Farrington (1994) cited in Research in focus 12.4 are illustrations of focal sampling.
Research in focus 12.4
A study of shoplifting
Buckle and Farrington (1994) report the results of a replication of an earlier study of shoplifting in a department store in Peterborough (Buckle and Farrington 1984). The replication was conducted in a similar store in Bedford. Customers were selected at random as they entered the store and followed by two observers until they left. The observers recorded such details as: cost of items bought; race, and estimated age; and behaviour. In Peterborough 486 people formed the basis of the sample and Bedford it was 502. Nine people shoplifted in Peterborough and six Bedford. Somewhat surprisingly, shoplifters were more likely to be male and either under 25 (in Peterborough) or over 55 (in Bedford). Most shoplifters also purchased goods. Most shoplifting was of small items of relatively little monetary value. The sampling and observation strategies entailed random sampling of people followed by continuous recording for a short or long period depending on how long the person remained in the store.