Observational studies can be classified usefully on five bases: (1) whether the situation in which the observation is made is natural or contrived, (2) whether the observation is obtrusive or unobtrusive, (3) whether the observation is structured or unstructured, (4) whether the factor of interest is observed directly or indirectly and (5) whether observations are made by observers or by mechanical means. Each of these classifications has some impact on the quality of data collected under the circumstances specified. In the following materials, various combinations of these classifications are considered, along with their advantages and disadvantages.
Natural, Direct, Unobtrusive Observation: When an observer is stationed in a grocery store to note how many different brands of canned soup each shopper picks up before selecting one, there is unobtrusive, direct observation in a natural situation. If the observer looks and acts like another shopper, the regular shoppers do not realize they are being observed. If a camera is positioned to record shopping actions, the observation is by mechanical means. If the observer counts the specific cans picked up, the observation is structured; but if the observer has a less clear assignment, such as to observe how shoppers go about selecting a brand of soup, the situation is unstructured.
Structured direct observation is used when the problem at hand has been formulated precisely enough to enable researchers to define specifically the observations to be made. Observers in a supermarket, for example, might note the number of different cans of soup picked up by each customer buying soup. Such observations are not as apt to have a subjective bias by observers as are unstructured observations. A form can be easily printed for simple recording of such observation. Even if the observers must wait for an unnoticed moment to record each observation, they are not apt to have as much bias in their memory as with unstructured observations. Not all structured observations are as simple as this example, but experiments have shown that even observes with different points of view on a given question will tend to make similar observations under structured conditions.
Unstructured, direct observation is similar to unstructured questioning; observers are placed in situations and observe whatever they deem pertinent. For example, in an effort to find ways of improving the service in a retail store, observer may mingle with customers in the store and look for activities that suggest service problems. Such observation is subject to subjective errors in both the actual observation and recording. No one can observe everything that is going on; hence the observer must select certain things to note. The actual observation also has subjective elements because of the difficulty of separating observation from inference; customers standing at a counter with annoyed looks on their faces may be observed as irritated because of lack of service. The latter inference follows so easily from the observations that it may not be separated from it.
The following is the report of one out of a large number of observations of shopping behavior in supermarkets:
A school-age boy and his parents enter the aisle.
The parents hurry down the aisle, looking straight ahead and not even glancing at the cereals
“Can’t I have some cereal?” asks the boy very winningly.
No, answers the father very sternly and quickly continues up the aisle.
‘You dirty crumb’ is the boy’s reply as he walks up the aisle with his head lowered.
This is somewhat dramatic observation. Did the drama cause the observer to overlook more mundane aspects of the family’s behavior? Would similar adjectives have been used by others observers to describe the behavior? The example suggests the possibilities for observer bias. In this particular study, the researchers found two major problems: (1) to get observers to record their observations in detail and (2) to prepare permanent records of observations at the earliest opportunity. Extensive training was necessary to overcome these problems. The second problem above is common to observations where an effort is made to keep the observer unnoticed. In such cases, the observer cannot easily record the data except from memory another invitation to error.
Sampling is another problem common to natural observation studies. Because it is necessary to let events happen as they do naturally, it is often difficult to get a cross sectional sample. Efforts are usually made to sample at different geographical locations (e.g. different stores) and to make a representative sampling of time periods.