As a behaviorist , Edward Tolman wanted to achieve with this purely operationally defined construct that in his theory unobservable variables such as personality traits , consciousness contents, desires, attitudes , etc. can be dispensed with.
The individual observations that someone always behaves in a similar way in comparable situations can be reduced to a rule by introducing an intervening variable. Example: You create a class of situations in which a person is exposed to the judgment of a group (giving a speech, music or theater performance, teaching, penalty shoot-out, etc.) and you find that a person always has a high pulse and blood pressure, Sweating etc. reacts. If you label the situations class with “performance in front of an audience” and the reactions class with “stress reaction”, you get the dependency function “performance in front of an audience” → intervening variable X → “stress reaction”. From a personistic point of view, however, this variable X would be interpreted as a personality trait (such as “shyness” or “lack of self-confidence”); However, these terms have a larger extension than is covered by the observations ( surplus of meaning , English surplus meaning ).
The distinction between hypothetical construct and intervening variable goes back to Paul E. Meehl and Kenneth MacCorquodale (see literature). Unlike the construct, no causal effect is ascribed to the intervening variable. Example: The superego , as a way of talking about behavioral inhibition, has no influence on behavior itself. Intervening variables are only "notation aids" that must not be objectified .
An Intervening Variable (or Mediator ) mediates the relationship between two or more variables . A distinction is made between full mediation and partial mediation. Complete mediation describes the case in which the apparent influence of one variable on another disappears when the mediator variable is controlled ( third variable control ). Partial or partial mediation describes the case in which the relationship between two variables is reduced when the mediator variable is introduced. However, if common causes of the mediator variable and the outcome are not included in the mediation analysis, this can lead to a non-causal, purely statistical relationship between the two variables of interest.
- Friedrich Dorsch: Psychological dictionary. Publisher Hans Huber, 1994.
- A. Kaplan (1964): The Conduct of Inquiry: Methodology for Behavioral Science. Chandler, p. 61.
- F. Elwert and C. Winship (2014): Endogenous Selection Bias: The Problem of Conditioning on a Collider Variable In: Annual Review of Sociology, 40, pp. 31–53., Doi : 10.1146 / annurev-soc-071913- 043455
- Kenneth MacCorquodale, Paul E. Meehl: On a distinction between hypothetical constructs and intervening variables . In: Psychological Review . 55, No. 2, 1948, pp. 95-107. doi : 10.1037 / h0056029 .
- Reuben M. Baron, David A. Kenny: Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations . (PDF) In: Journal of Personality and Social Psychology . 51, No. 6, 1986, pp. 1173-1182. doi : 10.1037 / 0022-35220.127.116.113 .