How to plan it?
While planning to do observational research, it is important to:
- Identify the study main objectives. Defining the objectives will help determine what type of observational research should be used; what type of situations, individuals and behaviours should be observed; and how the data will be recorded. For example, one study objective could be to evaluate if new food safety practices that were previously communicated to food safety workers are actually implemented in their work environment.
- Determine the type of observation to be conducted. Observational studies can vary in terms of: 1) participants’ awareness of being observed (overt vs. covert observation); 2) the role of the researcher during the observation (participant vs. non-participant); and 3) the control that can be achieved over the situation and context. The combination of these factors will determine the type of observation study to be conducted and also how the data will be recorded. The appropriate combination of factors should take into account the study’s main objectives and the advantages and disadvantages of each option.
Overt vs. Covert Observation: Participants may (overt) or may not (covert) know that they are being observed. When participants are aware that they are being observed there is a risk that they become conscious of their actions and do not behave and/or decide has they normally would but rather in accordance with the observer’s expectations. Thus, whenever possible it is better to conduct a covert observation. However, in some cases, for ethical and practical reasons it may be necessary to inform participants that they will be observed and get their informed consent, such as when observing private or professional situations.
Non-Participant vs. Participant Observation: In non-participant observation, the observer does not interfere with or manipulate the event being observed. This type of observation is particularly suited when it is important not to influence the setting or the participants. In participant observation, the observer takes in some extent, an active part in the situation being studied. This type of observation is particularly useful when studying complex or private situations and behaviours that require an inside perspective or interaction with participants (e.g., teaching new food handling practices). However, the fact that the observer takes an active role can alter the dynamics of the situation and behaviours in some way and may make it hard for the observer to remain objective. On the other hand, if it is a covert observation the observers have to rely on memory, as they cannot take notes while observing, which may promote observer bias.
Control and manipulation of the situation: Observational studies can also vary in terms of the degree of control over and manipulation that the researcher has over the situation, events, etc. By exerting control over the environment, the observer can control for factors that could alter the participant's behaviour that are not of interest to the study. This type of studies is useful to look at specific links between the variables of interest or when the goal is to compare participants or groups. The main drawbacks of exerting control are the lower ecological validity and the risk of participants behaving in an unnatural way if the setting is unfamiliar to them (e.g., laboratory).
- Determine the broad categories of behaviour to be studied. At this stage, and based on the study’s main objectives, it is important to identify the broad categories of behaviour to be studied. An example of a category could be “food safety behaviours”. If there is little knowledge about the categories that are relevant, informal observation should be conducted to help identify and develop them.
- Define the behaviours and determine how they will be recorded and measured. One important consideration at this stage relates to the breath of information sought within the broad categories, usually referred to as units of behaviour. This could be small segments of behaviour (e.g., specific behaviours such as hand washing and food handling practices) that are relatively easy to define and measure reliably; or larger segments (e.g., cooperativeness, openness to change in food safety practices) that are usually more psychologically meaningful. However, this type of segments usually requires a fair degree of inference on the part of the observer and, therefore, may reduce the reliability and validity of this sort of data. Once the behaviours have been defined, a decision must be made about how to record and measure them. Two main classes of measures can be selected: casual and systematic recordings.
Casual recording: Here observations are recorded in a narrative form or on audiotape or videotape and subsequently transformed in to data by categorizing and coding the various elements of behaviour (see the Analyzing section). Typically, when using narrative accounts the aim is to reproduce behavioural events in a written form in much the same way, and in the same sequence, as they originally occur, often aiming for little or no interpretative content. Video recordings offer a relatively cheap and semi-permanent record that can be played repeatedly, allowing for analysis at a level of detail and reliability not possible from direct observations. However, they can be very difficult to use in certain natural contexts. Casual recordings are particularly useful when studying large segments and complex behaviour, and when it is important to collect all the real-life elements of the situation.
Systematic recordings: Here a scoring system and prearranged specific categories are applied consistently and in a standardized way. Thus, this type of recording is particularly useful when studying predetermined and specific behaviours. This type of recording usually requires an observation checklist, on which information is recorded under the proper headings. Checklists typically consist of a number of behavioural units or categories with clear descriptions of each. Depending on the characteristics of behaviour units, a classification and measurement format is used to recode the behaviour. Popular types of classification and measurement formats are:
Frequency: counting the number of times a behaviour occurs. It is an appropriate method when the behaviour is relatively brief and clearly defined. However, it does not provide any information on duration, intensity and quality of behaviour.
Duration: measures how long the behaviour was performed and determines when it started and ended. It can also be transformed into a frequency measure. However, it does not provide any information on intensity and quality of behaviour. This is a most appropriate method to recode larger segments of behaviour.
Ratings: subjective measures of the qualitative aspects of behaviour, such as degree of cooperativeness. This type of measure requires the observer to make a subjective evaluation of the observed behaviour, and thus is more prone to observer bias.
- Determine how to sample the events and behaviours. This should be based on the relevance and representativeness of the cases and situations in relation to the research objectives, and on the resources available. Sampling issues should take into account the main goal of the research. When the goal is to describe some phenomenon, one will likely need multiple observations at the same site in order to get a thick and rich description that allows enough understanding. When the goal is to explore, one will likely gain more by investigating more settings with numerous observations in order to evaluate variation and consistency of relevant categories among observations. When the goal is to explain a phenomenon, a few well-chosen comparative cases are preferable in order to examine causal connections and explanations. Ideally, these cases should only differ in the key categorical variables one wants to study. There are a number techniques of sampling behaviour, the most popular and relevant being event sampling and time sampling.
Event sampling is used when observing a particular event and activity from the beginning to the end. One of its main advantages is that there is an inherent validity in observing a complete event rather than the more piecemeal observation performed in time sampling. It also enables the observation of rare events that would probably not be found by time sampling. However, when using event sampling the researcher needs to know when the event is likely to occur or be prepared for when it occurs.
Time sampling refers to the selection of periods of observation at different points in time. This could be done in a more systematic way, such as 5-minute observation periods at specified times (e.g., every hour), or in a random way, such as 5-minute observations selected at random times. By using time sampling, the researcher can observe representative samples of behaviours, but only those that occur relatively frequently. The main disadvantage of using this type of sampling is that it lacks the continuity and completeness found in event sampling.
- Hire and train observers (if necessary). When the study requires observing a large sample of people or situations it might be necessary to have a team of observers. If that is the case, it is important to hire observers that have past experience with observation and train them. During training it is important to make sure that all the observers understand correctly the procedure, what behaviours should be measured and how to record them. It is crucial to ensure that they are reporting observations consistently. To improve consistency between observers, one should train observers in a group and have them practice using the same form. Compare their results to assure that the data is being recorded in a consistent and reliable way.
- Conduct a pilot-test and assess reliability among multiple observers. At least two independent observers should recode the data in order to check the instruments reliability, i.e., that there is consistency in the recoding of the data between observers. Also, conducting a pilot-test ensures that the delineated procedure for collecting data works properly, and that the instruments used to measure behaviours (e.g., checklists) cover all the relevant behaviours observed.
- Revise the procedure and instruments used in accord with the pilot-test results. If substantial changes are made, run another pilot test.