Variables, Latent

Latent variables are variables that are unobserved—at least not directly. Also sometimes referred to as hidden variables, unmeasured variables, factors, hypothetical variables, or hypothetical constructs, latent variables are measured in practice not by a single variable, but by multiple observed (or manifest) variables. Their use in communication, and in the social sciences more generally, is ubiquitous.

For instance, communication researchers are frequently interested in the influence of personality or other psychological traits on the reception of different kinds of messages—a perennial concern for students of attitude change and persuasion. With few exceptions, however, complex psychological or attitudinal traits of the kind that often serve as a filter for messages received from the mass media or through an individual’s social environment are not adequately captured by a single, readily observed indicator. To get at these underlying, latent concepts, researchers must engage in efforts to infer their existence from analyses of variables that can be observed—much in the same way as physicists infer the existence of far-away planets by observing the “wobble” induced by the gravitational pull of these unobserved celestial bodies on neighboring stars. In the social sciences, making inferences about the existence of latent constructs often necessitates that researchers employ large batteries of variables culled from surveys or multiple, imperfectly measured indicators from other sources.

Apart from providing a partial solution to the problems inherent in capturing unobserved and perhaps even unobservable concepts, latent variables are also used in a purely predictive capacity and for data reduction—stripping down large data sets for the sake of parsimony and tractability. This entry provides an overview of the different ways in which latent variables have been conceptualized and employed in communication research. It also touches upon some of the philosophical underpinnings of latent variables and briefly introduces several methods for measuring latent variables or statistical techniques that incorporate measures of latent variables into a broader framework.