2 1.2 Author contribution
4.3. Methods of recording observations
Decisions regarding the method of recording behaviour will depend on the purpose of the observational study, on the nature of the data that the observer aims to collect, either qualitative or quantitative, and on how the results of the study will eventually be summarised and analysed. A general classification can be done according to whether the observer seeks to obtain a comprehensive description of the behaviour or only selected aspects that are related to the specific goals of a study. In the former case the behaviour exhibited in a given setting is described in considerable detail whereas in the latter only particular units of the behaviour are recorded (Shaughnessy & Zechmeister, 1997).
Narrative records
Narrative records form a continuous description of the phenomenon being observed. The observer aims to provide a more or less accurate description of what takes place in a particular setting. For that reason, narrative records need to include descriptions and accounts of people, tasks, behaviours and conversations (Dane, 1990). Descriptions may be
either written or in the form of spoken and visual records obtained with video tape and movie cameras.
A major difference between narrative records and other forms of behaviour measurement is that particular hypotheses and expectations about the behaviours under observation can be tested after obtaining the data (Shaughnessy & Zechmeister, 1997).
Instead of classifying or coding the behaviour at the time of observation, everything that takes place is targeted for recording, minimising the amount of inference required from observers (Bentzen, 2000). Once narrative records are obtained the researcher can review the data and organise, classify and analyse the particular information that is critical for evaluating the study’s hypotheses (Bowling, 2002). By excluding any inferences or impressions on the part of the observer, it is hoped that the content of the narrative record can be classified and coded in a more objective manner (Shaughnessy &
Zechmeister, 1997).
An important drawback of narrative records is that the observer is expected to record a large amount of information in a very short time. As this information is unstructured, its organisation and analysis may also be a very demanding task (Bowling, 2002). Moreover, narrative records have to be made during or as soon as possible after behaviour is observed (Shaughnessy & Zechmeister, 1997). This is particularly difficult when observational activity is concealed and participants are not aware of being observed. If the recording occurs too long after the observation, it may be harder to reproduce the original sequence of actions and important features may become underestimated or forgotten.
According to Robson (2002) a way to deal with this problem may be the inclusion of several participant observers in the setting or the use video or audio tapes. This allows independent analyses to be conducted and then compared in order to assess agreement (Robson, 2002).
Coding Schemes
Researchers may want to focus on certain behaviours and specific aspects of individuals and settings. In that case the observation variables will take the form of pre-specified categories of behaviour or interaction which will form the basic units of a coding scheme (Humphreys, 1992). Decisions about coding definitions and the included categories must be made in relation to the theoretical purposes of the study, in advance of data gathering (Cozby,2009). When defining the categories, a researcher should be aware of the “level of
analysis” problem. Observed phenomena may consist of different levels, hierarchically arranged, with larger and more inclusive concepts occupying higher levels of abstraction and smaller and more detailed occupying lower levels (Bakeman & Gottman, 1997).
Categories should not continually multiply and, therefore, researchers need to decide what conceptual level seems appropriate for the purposes of the study. As proposed by Bakeman & Gottman (1997) choosing categories on slightly more detailed level than required can increase the chances of capturing an accurate and comprehensive description of selected activities and interactions and facilitate reliability checks. In order to permit a quantitative analysis of the observations, categories must be exhaustive, meaning that the coding scheme covers all the possible events, and mutually exclusive, which means that only one code can be associated with a particular activity or behaviour (e.g. Irwin &
Bushnell, 1980).
The use of coding schemes has methodological and practical advantages. A simple and straightforward coding scheme allows the observer to record very quickly the target behaviours and consequently the collection of a large amount of data, on many individuals.
By setting out a predetermined set of categories and an explicit set of criteria for assigning occurrences to them an amount of structure is imposed on what is observed (Smith &
Davis, 2007). It is only explicit behaviour and not the observer’s interpretation of the meaning of the behaviour that is recorded, minimising one aspect of the subjectivity which normally occurs in an individual’s descriptions of events (Croll, 1986). This method of recording produces data which can be presented in quantitative form and which can be summarised and related to other data using statistical techniques (Croll, 1986).
However, the use of a limited number of pre-defined categories has given rise to a certain number of criticisms. According to Irwin and Busnell (1980) “observers may end up looking for things to fit the categories rather than describe what is occurring” (p.159). As a result, behaviours that may be important in understanding the phenomenon under study could be totally overlooked. Coding categories that are too vague provide little specification of what to record allowing space for observer bias (Kerlinger & Lee, 1999).
Conversely, when the categories are too specific, although they may reduce ambiguity and uncertainty, they may also be too rigid and inflexible. In that case researchers may end up collecting irrelevant or trivial information. It has also been suggested that coding schemes can only give a partial view of the situation as they fail to capture contextual factors or relations between behaviours (Bowling, 2002; Croll, 1986). Consequently, the phenomena being observed may appear as de-contextualised isolated bits of behaviour stripped of the
richness and spontaneity of the specific situation. Irwin & Bushnell (1980) suggested piloting the coding scheme in the observational setting in order to allow the researchers to spot any limitations and make any necessary additions or modifications prior to the main study.