Multiple observation

Multiple observation

Multiple observation is the more appropriate technique for studying events that occur in well-defined settings, are clearly bounded in time and space, and replicate essentially the same relationship. The exact number of observers does not matter as much as do other conditions that have to be met.

First, observations have to be standardized much in the same way as the interviews in a survey. The dimensions or attributes focal to the investigation need to be specified in advance, so that observation schedules can be drawn up and observers instructed as to what to look for.

Second, the events and occurrences to be enumerated must be accessible to observers from preselected vantage points to which they can be assigned. Again, such assignment to sampling points is broadly

Studying events in their natural settings 197

analogous to the selection of respondents from lists of individuals or by households but with one significant difference: events, not individuals, are the units of observation. Hence, sampling strategy must aim at an unbiased selection of stations or situations where the relevant encounters, transactions, or behavioral responses are likely to be found with acceptable frequency.

Insofar as there is a predictable pattern of physical movement, one can sample by locale, stationing observers at sites where people normally pass in their rounds of daily activity (a marketplace or a railroad station), in centers of certain institutional activities (a cathedral or a lecture hall), or at sites where particular kinds of transactions typically take place (a check-out counter or a hiring hall). Any of these qualify as strategic vantage points from which to observe particular kinds of behavior. This is also how many organizations generate their own records, usually on standardized forms, by focusing on points of routine contact with their public or clientele. But their procedures do not of themselves ensure that the observations are standardized. They may ask for information about which the record keepers are sensitive, or the record itself may be used to measure administrative performance. Sometimes definitions and/or baselines are changed for reasons that have nothing to do with research objectives. As everyone knows, unless an enumeration stands in a constant relation to the phenomenon it is meant to track, the cross- group and time-series comparisons based on it lose much of their value. Indeed, they may actually obscure the true state of affairs.

When it comes to events that have no distinct locale but are clearly linked to role behavior, observers have to follow their subjects in their normal round of activity. This is what Reiss (1971) did in studying police-citizen encounters. Latane and Darley (1970) went even further in studying bystander behavior in simulated attacks against persons in public places. Such contrivance increased their control over the stimulus event. The recording procedures were, of course, the same as in autonomously generated situations.

Although one suspects that the presence of “outsiders” in circumstances as touchy as law enforcement and crime would have some influence on police behavior, this may be less than expected. As time passed, so Reiss (1971) maintains, the scientific observers became part of the team, and officers ceased to accord them any undue attention. There are, nevertheless, limits to the utility of observations conducted in a “natural” setting. For one thing, it creates a dependence

198 A handbook of qualitative methodologies

on what one can see and is able to overhear. It precludes statistically random selection. Replication with comparable samples is difficult. Researchers are also deterred by its high cost. The use of administrative records that cover long time spans and the time-series data produced by surveys are so much more economical. When both are lacking, knowledgeable information can fill the gap. Yet, systematic observation, even when confined to a single point in time, can be a useful supplement, even an indispensable corrective, to more conventional methods of social research.