Orientation Observation In-depth interviews Document analysis and semiology Conversation and discourse analysis Secondary Data Surveys Experiments Ethics Research outcomes



Social Research Glossary

About Researching the Real World



© Lee Harvey 2012–2017

Page updated 9 October, 2017

Citation reference: Harvey, L., 2012–2017, Researching the Real World, available at
All rights belong to author.


A Guide to Methodology

3. Observation

3.1 Introduction
3.2 Aspects
3.3 Methodological approaches

3.3.1 Positivism and observation
3.3.2 Phenomenology and observation Observation for theory generation: emergent theory Observation for theory development: a naturalistic perspective Naturalistic research The problem of generalisability in naturalistic research Objectivity and naturalistic research Observation as a basis for identifying subjects' meanings Observation as a basis for empathetic interpretation Observation as the basis for deconstructing everyday life Observation as a process of establishing identity Observation as a means of discovering how communities operate

3.3.3 Critical social research and observation

3.4 Access
3.5 Recording data
3.6 Analysing observational or ethnographic data
3.7 Summary

3.3 Methodological approaches

3.3.2 Phenomenology and observation Observation for theory development: a naturalistic perspective
Observation is seen as a vital ingredient of naturalistic approaches to theory development. A naturalistic approach is about getting insights into people's real lives as they live them, rather than the contrived interactions that sociologists have with research subjects when they use questionnaires and interview surveys. Naturalistic research
Naturalistic research refers to methods of social enquiry that attempt to grasp the 'natural' processes of social action and interaction. The aim is to collect information from social settings without creating artificial situations (such as an interview or an experiment) and with the minimum of disruption to the natural setting by the researcher.

It is argued that the naturalistic approach (using observation or unstructured, in-depth interviews (see Section 4)) places the researcher in a better position to interpret or understand social actions in their own context. Participant observation enables the researcher to experience the natural setting at first hand through being directly involved and thus improves ecological validity (see Section

In addition, there is now an opportunity to develop naturalistic research through the analysis of some forms of 'big data', such as social media postings (see Section 7.4.3).

Top The problem of generalisability in naturalistic research
Critics of a naturalistic approach argue that the participant observer's situation only permits a partial picture and this is not necessarily valid, reliable or generalisable (see Sections 1.8, 1.9 and 1.10).

Ken Pryce (1979), for example, admitted that that as a male researcher he had only limited access to the women in the West Indian community for research purposes because in any working-class community there is a tendency for males and females in all age groups to associate in single-sex peer groups. He also noted that the generalizations he made were based on very small numbers 'a mere fraction compared with the total number making up the community'. For Pryce, the evidence in his book is 'illustrative rather than systematic'.

Naturalistic researchers tend to respond by saying that some insight is better than none. A view that itself is countered by a suggestion that limited insight might actually be heavily biased and misleading, especially if any attempt at generalisation or theorising ensues.

Goldthorpe (2000), for example, argued that naturalistic ethnography fails to accept that the social world exists independently of our ideas about it, thus is unable to make inferences about the world beyond the data at hand. For Goldthorpe, this failure of, what he called, the 'logic of inference' renders ethnography impotent. In essence, those who are sceptical of naturalistic research adopt the standard positivist view that requires the use of probabilistic samples to permit generalisability. Given that probabilistic samples are virtually unobtainable in practice in any form of research, (see Section 8), this is an unreasonable expectation to place on observation/ethnographic research.

More recently, Williams (2000) has argued that generalisation is inevitable, desirable and possible in interpretive research. He proposed that 'interpretivism' must employ a special kind of generalisation, which he calls moderatum (see Section 1.10.1). He acknowledges that there are limits on generalisations from interpretive research, which reflect 'the limits of interpretivism itself'. As a result he calls for the 'adoption of methodological pluralism' (Williams, 2000, p. 209).

Williams notes that some qualitative sociologist, notably Denzin (1983), or Guba and Lincoln (1982, 1994) claim that generalisation is impossible. Denzin argues, for example that naturalistic enquiry results in too too much variability to allow the possibility of generalisation from a specific situation to others. However, others, 'notably Hammersley (1990, 1992) and Mason (1996), and at least one approach (analytic induction) is founded on very strong generalising claims'.

Williams points out that 'If one takes generalisation in a broad non-scientific sense to mean 'a general notion or proposition obtained by inference from particular cases' (Concise Oxford Dictionary), then interpretive research is replete with generalisations' (Williams, 2000, p. 212) and he demonstrates this using published studies.

Top Objectivity and naturalistic research
Observation studies are often seen, by positivists, as 'subjective' compared to the 'objectivity' of quantitative techniques. Participant observation, in particular, is singled out as a 'subjective method'. This notion of the subjectivity of observation studies arises for several reasons.

First, data derived from naturalistic observation is seen as being more personal and more idiosyncratic than answers to questions on a survey schedule. (Section 8). (This, of course, ignores that someone subjectively determines what are the appropriate questions and how they are worded.) There is an assumption that naturalistic observation gives too much leeway to the researcher to be subjective in selecting what to report.

Observation that takes place in a conventional experimental setting is rarely questioned, on the other hand, because the experiment is set up so that particular effects are being looked for from the outset (see Chapter 9). This predetermination of what to look for is reproduced in the use of interview schedules or questionnaires. There is a clear bias here in suggesting that predetermining what to look for is 'more objective' than discovering the important issues as the research progresses.

Second, naturalistic observation, it is argued, relies heavily on the ability of the researcher to get to see what is going on and thus the role being played by the researcher effects what is observed.

Third, if observation studies are concerned with subjects' meanings (Section then the data of the study is itself subjective. There is no way of objectively measuring meanings and indeed there is no way that researchers can even be sure that they have grasped someone else's meanings. This means that the results of the observation are dependent entirely on the observer's interpretation and there is no possibility that the information can be reinterpreted by another researcher. This raises issues about reliability and accuracy (see Section 1.9).

Non-positivists are not convinced by these arguments. Sallie Westwood (1984) in her study of female factory workers, for example, set out to grasp a specific 'cultural space' and this required immersion in the life of the shop floor. Participant observation was, for her, the only technique that allows the researcher to inhabit and record such a cultural space as a factory shop floor. Westwood aimed to illuminate the 'lived experiences' of women workers who come together to generate and sustain a specific feminine working-class culture. This required that she be involved as a participant so as to unravel the world of symbols and meanings that made up the culture.

Such a view makes the positivist objections irrelevant. First, the participant observer is no more selective than the experimenter or social surveyor. The experimenter and social surveyor are selective in what they are going to look for before the data collection takes place. The participant observer is less selective, has fewer rigid preconceptions, and tends to look at a much wider array of 'facts' in order to filter out what is significant in the development of a theory during the data collection. The observer thus allows the data to inform the development of theory rather simply test a predetermined theory.

Second, the idea that social facts can be measured objectively is disputed. The non-positivist view suggests that no method can provide the researcher with 'objective facts'. Facts do not exist in isolation. The words a respondent utters, or the actions a subject is observed to make have no existence on their own but only 'make sense' in relation to other utterances and actions (Section 1.4.1).

Someone waving at a group of people may be waving goodbye but it might also be an act of defiance or recklessness (for example, a member of a gang from one neighbourhood who is drawing attention to herself in the territory of another gang) or an acknowledgement of adulation (for example, a film star waving to the crowd at a film première).

Furthermore, the meanings that actions and utterances have for the researcher is also not self-evident. Data only have meaning when they are interpreted. The meaning of an event or piece of evidence depends on theory. This may be the everyday 'theory' of unreflective common sense or it may be a reflective attempt to develop a social theory.

For example, two groups of football supporters may confront each other in a side street outside a football ground after a match. For the supporters, this is a sign of their commitment to their team; for the news media, this might be seen as another mindless brawl between 'football hooligans'; for the police, a potential problem of public order; for the home owners in the area, it is a nuisance; for a sociologist, it might be an incidence of working-class group machismo; for a psychologist, an example of territorial conflict; and for a magistrate, a wanton disregard of the law. In other words, what a particular event or piece of data means to the interpreter of the data is not self-evident in the data itself. It depends upon a theoretical framework within which it is located. It makes sense in different ways to different people who approach the 'fact' from different perspectives.


Next Observation as a basis for identifying subjects' meanings