Levels of Analysis
Conflict communication research uses data of a psychological, interactional, or sociological sort. Psychological data are descriptive of individuals and include emotional states and traits, personality states and traits, level and type of motivation, and types and degree of knowledge.
Examples of such data are measures of hostility (A. H. Buss & Durkee, 1957), propensity for abusiveness (Dutton, Landolt, Starzomski, & Bodnarchuk, 2001), and ethnocentrism (Neuliep & McCroskey, 1997). Interactional data include attributes of verbal and nonverbal communicative behavior of people in simulated or actual interaction. Examples of such data are integrative and distributive behaviors in negotiation (Cai et al., 2000; Donohue & Roberto, 1996), emotion change between interactants (Rogan & Hammer, 1995a), and coded linguistic measures (Cook-Gumperz & Szymanski, 2001; Scarry, 1985). Finally, sociological data involve attributes of groups, organizations, states, and cultures. Examples are Cashman’s (1993) review of national attributes and international conflict (see also Diehl, 2004; Speer, 1986) and Doreian’s (1981) analysis of network data to predict the mobilization of individuals taking sides in a conflict.The typical methods used to gather these three types of data differ. However, all three types of data require evidence of validity. For theoretical concepts, validity is typically assessed by construct validation methods (or by related techniques, such as the multitrait- multimethod matrix approach, as found in Campbell & Fiske, 1959, and more recent methods using similar ideas, as found in Maruyama, 1997). It is quite common for measurement validity to be ignored prior to data collection, because a measure of a theoretical variable may be validated by finding support for a set of hypotheses in which the theoretical variable is measured by an operationalized version of the variable. This strategy is risky: Without independent evidence of measurement validity, we cannot determine if failure to find support for a set of hypotheses is due to poor measurement, inadequate theory, or both.