Comparing Networks: Structural Correspondence between Behavioral and Recall Networks
Relationship between Behavior and Recall
The question of whether we, as individuals, are reliable sources of information concerning our behavior has been debated for many decades in social network research. Because most research in this field uses individuals’ recollection of their behavior as a source of data (Marsden, 2005), assessing the reliability of our informants is not a trivial issue. This study focuses on comparing individuals’ actual behavior to their recall of that behavior. As such, I am not concerned with perception-based networks (e.g., friendship or trust) because they do not have to be related to the occurrence or frequency of a specific behavior. The reliability of individual recollection is, however, critical for research that relies on an accurate account of interactions between individuals. For example, the study of the diffusion of information relies on the account by informants of their actual communication behavior - was the message transmitted? - and not just on the interactions that they can recall.
Existing research comparing behavior and recall has mainly sought to understand how and why recall and behavior differ from each other at the individual or dyadic level. The BKS studies - a series of articles published by Bernard, Killworth, and Sailer (1979,1981,1982) - are the main body of work in this strand of research. They made an early attempt to assess the reliability of informants by observing the communication behavior of several groups of actors in different contexts and comparing it to the interactions that the actors reported during the same period of time. Their findings showed that respondents had a poor recollection of whom they communicated with and the frequency of these communications (Bernard et al., 1979), as well as questioned the reliability of social network data collected based on respondents’ recall (Bernard & Killworth, 1977; Kill- worth & Bernard, 1976).
Further research suggests that specific factors may affect individuals’ recall of their interactions. Freeman, Romney, and Freeman (1987) showed that the difference between individuals’ behaviors and their recall in the BKS studies could be explained by biases in individuals’ recollection patterns. They found that respondents had a tendency to remember their interactions based on stable and repetitive patterns. Hence, individuals will be more likely to remember others with whom they interact regularly, even though they might not have interacted during a specific observation period. A corollary conclusion is that individuals tend to forget infrequent communication episodes (Tversky & Kahneman, 1974). Although respondents are more likely to report interactions that are frequent and regular, other factors contribute to making certain interactions stand out in an individual’s memory. In an organizational context, hierarchical status has been shown to affect individuals’ recollection of their interactions. Webster (1995) reported that respondents with similar status tended to remember reciprocal interactions more than if they had different hierarchical status. This indicates that certain attributes of the relationship (e.g., frequency, regularity of behaviors) and of the individuals (e.g., seniority) can alter the patterns of recall of respondents.
My approach in this chapter is to explore whether networks based on the observation of a behavior and networks based on the recall of this behavior are structurally different. More specifically, I examine whether actors’ recall relates to the behavioral networks, and if it does, whether any differences between recall and behavior are systematic.