Abstract
Due to modest predictability by Rotter's I-E generalized locus of control scale in some situations, a plethora of domain-specific scales have been developed. Although providing high predictability of unique behaviors in unique domains, these specific locus of control scales, using idiosyncratic independent variables and developing inconsistent dimensions, may not serve the field of psychology well.
Locus of control is one of the most studied variables in psychology, and has been successful in predicting behavior across several different populations (Rotter, 1990). Based on social learning theory, locus of control was originally conceptualized by Rotter (1966) as:
“When a reinforcement is perceived by the subject as…not being entirely contingent upon his action, then, in our culture, it is typically perceived as the result of luck, chance, fate, as under the control of powerful others, or as unpredictable because of the great complexity of the forces surrounding him. When the event is interpreted in this way by an individual, we have labeled this a belief in external control. If the person perceives that the event is contingent upon his own behavior or his own relatively permanent characteristics, we have termed this a belief in internal control (p. 1).”
Rotter introduced locus of control as a generalized expectancy, more trait-like than statelike, and therefore similar to a personality trait (Wallston, 1992), a determining tendency, or predisposition to respond in a certain manner. If locus of control is a fairly stable, consistent pattern of perception, or an immutable personality trait, then an individual with an internal (or an external) locus of control would be expected to react somewhat similarly across situations. Rotter (1990) pointed out that in social learning theory, the unit of investigation for personality study is the interaction of the individual and his/her relevant environment: although behavior will be different in different situations, there may be a gradient of generalization across situations.
Rotter (1966) believed that although locus of control could vary based on circumstance, people would generally trend toward internality or externality as a personality trait. His I-E scale was not designed for specific contexts, but rather to measure generalized locus of control and predict behavior across a broad range of situations, which is a general theory of locus of control. General theories are limited since they represent only one of many variables that are relevant in prediction of behavior (Rotter, 1975). He pointed out that the items in his scale “dealt exclusively with the subjects' belief about the nature of the world…they are concerned with the subjects' expectations about how reinforcement is controlled” (Rotter, 1966, p. 10).
The locus of control construct, measured by various domain-specific scales, is increasingly used as an antecedent variable to predict specific outcomes. Context-specific locus of control scales have been developed for sales (Chung & Ding, 2002), pilot safety (Hunter, 2002), weight control (Saltzer, 1982), smoking (Georgiou & Bradley, 1992), drinking (Donovan & O'Leary, 1978), reckless driving (Özkan & Lajunen, 2005), child development (Furnham, 2010), academic success (Trice, 1985), health student academic success (Eachus & Cassidy, 1997), prison (Pugh, 1992), control of health (Wallston, 2005), headaches (Martin, Holroyd, & Penzien, 1990), cancer (Watson, Greer, Pruyn, & Van den Borne, 1990), maternal labor and delivery outcomes (Schroeder, 1985), fetal health (Labs & Wurtale, 1986), parent health (DeVillis, DeVillis, Blanchard, Klotz, Luchok, & Voyce, 1993), spiritual health (Holt, Clark, Kreuter, & Rubio, 2003), and many other contexts. This trend towards specificity has resulted in development of an increasing number of scales, not designed to measure generalized expectancies of control (Furnham & Steele, 1993). Rather, they assess a specific group or domain, using unique variables construed as locus of control, and attempt to predict or explain narrow, domain-specific outcomes.
The independent variables used in context-specific scales generally provide satisfactory psychometrics, and acceptable correlations with, or prediction of, outcome variables of interest. However, there is a question of overlap among the context-specific scales and generalized locus of control. Consider an overweight student who smokes, has headaches, drives recklessly, and drinks too much. If this individual took the locus of control scale developed for each condition or situation, plus Rotter's (1966) I-E global scale, although there may be some overlap, there would almost certainly be conflicting results among the domain-specific scales, as well as with Rotter's global scale. Rotter's unidimensional scale treats reinforcement valence as a separate construct, while most locus of control scales are concerned only with expectancies. Also, the structure, scoring (Yes-No forced choice, true-false, and 5-, 6-, and 7-item Likert- type), and dimensionality (ranging from one to six dimensions) of the many domain-specific scales are inconsistent (Furnham & Steele, 1993).
As examples of the inconsistencies, the 18-item Fetal Health Locus of Control Scale (Labs & Wurtle, 1986) was developed with a sample of 63 women, using a 9-point graphic rating scale, which produced three subscales: internal, chance, and powerful others. There was no comparison of this scale to Rotter's (1966) scale. The 30-item Parent Health Locus of Control Scale (DeVellis, et al., 1993) was developed with a sample of 822, using a 6-point Likert-type scale, which produced six subscales: divine, professional, child, parental, media, and fate influence. These subscales were intended to measure influences on the health and well-being of children. There was no comparison of this scale to Rotter's (1966) measure. A new, unidimensional, 25-item Smoking-Specific Locus of Control scale (Bunch & Schneider, 1991) was developed with a sample of 70 respondents, and a 4-point Likert-type rating scale. A comparison of the new Smoking-Specific scale with Rotter's (1966) scale yielded no significant correlations, which, according to the authors, “suggests independence” (p. 1080). The 22-item Cancer Locus of Control Scale (Watson, et al., 1990) was originally developed using 118 cancer patients. This scale comprised three subscales: course of the illness, cause of the illness, and religious control. There was no comparison of this scale to Rotter's (1966) scale.
So the issue must be faced: what is the value and contribution of context-specific locus of control scales beyond the unique situation? Do the specific scales, comprised of inconsistent, situation-specific variables, provide measures sufficiently overlapping generalized locus of control to be considered the same construct? Or do they simply produce situation-specific sets of correlates or sets of independent variables predicting various behaviors, and represented as locus of control? Ross and Nisbett (1991) argue that behavior is determined to a large extent by external situational factors, and dispositional traits are not stable across situations. They conclude that it is the stability of situations that produce similar behaviors. Consequently, we would expect narrow, situation-specific scales purporting to measure locus of control to be assessing a specific situation, and reflecting “locus of control” in name only.
Rotter (1975) recognized the limitations of general theory, which allows prediction in a large number of situations but with low precision, and he realized that a more specific generalized expectancy should allow greater predictability. Perhaps a few broad-context locus of control scales (e.g., work, health), consistent with the general locus of control theory, could be properly developed and used rather than the plethora of narrow, context-specific measures. Otherwise, we will be left with an increasing number of test batteries purporting to measure locus of control using domain-specific independent variables—many of which appear to be only loosely related to expectancy—to predict behavior and other outcomes in specific domains and situations.
The development of an adequate measure is a very expensive undertaking, requiring years of research (Rotter, 1975). Although specific tests predict specific problems or behaviors better if they have been carefully constructed, a process which usually requires several versions with different validity criteria and several test developments, this has not been the case for the majority of the context-specific tests (Rotter, in Furnham & Steele, 1993). Many of the context-specific scales have been used only once or a few times. Any value of these context-specific measures and the relationships among their variables accrue only to the specific domains. Unless the scientific psychological community ceases to accept untested, domain-specific measures, we can expect continued proliferation of these narrow theories and measures.
Guthrie's (1946) warning in his presidential address to the American Psychological Association is valid today: “Theories may flourish if their basis lies not in scientific fact but in opinions and interpretations acceptable only to the members of a limited faction; but they will be bad theories…. Unless psychologists maintain an interest in general theory, the fields of psychology will increasingly become independent collections of undigested information (p. 20).”