W5-DQ

W5-DQ-TheMediator.pdf

Journal of Personality and Social Psychology 1986, \<*. 51, No. 6, 1173-1182

The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations

Reuben M. Baron and David A. Kenny University of Connecticut

In this article, we attempt to distinguish between the properties of moderator and mediator variables at a number of levels. First, we seek to make theorists and researchers aware of the importance of not using the terms moderator and mediator interchangeably by carefully elaborating, both concep- tually and strategically, the many ways in which moderators and mediators differ. We then go beyond this largely pedagogical function and delineate the conceptual and strategic implications of making use of such distinctions with regard to a wide range of phenomena, including control and stress, attitudes, and personality traits. We also provide a specific compendium of analytic procedures ap- propriate for making the most effective use of the moderator and mediator distinction, both sepa- rately and in terms of a broader causal system that includes both moderators and mediators.

The purpose of this analysis is to distinguish between the properties of moderator and mediator variables in such a way as to clarify the different ways in which conceptual variables may account for differences in peoples' behavior. Specifically, we differentiate between two often-confused functions of third variables: (a) the moderator function of third variables, which partitions a focal independent variable into subgroups that es- tablish its domains of maximal effectiveness in regard to a given dependent variable, and (b) the mediator function of a third variable, which represents the generative mechanism through which the focal independent variable is able to influence the dependent variable of interest.

Although these two functions of third variables have a rela- tively long tradition in the social sciences, it is not at all uncom- mon for social psychological researchers to ust the terms mod- erator and mediator interchangeably. For example, Harkins, Latane, and Williams (1980) first summarized the impact of identifiability on social loafing by observing that it "moderates social loafing" (p. 303) and then within the same paragraph proposed "that identifiability is an important mediator of social loafing." Similarly, Findley and Cooper (1983), intending a moderator interpretation, labeled gender, age, race, and socio- economic level as mediators of the relation between locus of control and academic achievement. Thus, one largely pedagogi-

This research was supported in part by National Science Foundation Grant BNS-8210137 and National Institute of Mental Health Grant RO1MH-40295-01 to the second author. Support was also given to him during his sabbatical year (1982-83) by the MacArthur Foundation at the Center for Advanced Studies in the Behavioral Sciences, Stanford, California.

Thanks are due to Judith Harackiewicz, Charles Judd, Stephen West, and Harris Cooper, who provided comments on an earlier version of this article. Stephen P. Needel was instrumental in the beginning stages of this work.

Correspondence concerning this article should be addressed to Reu- ben M. Baron, Department of Psychology U-20, University of Connect- icut, Storrs, Connecticut 06268.

cal function of this article is to clarify for experimental re- searchers the importance of respecting these distinctions.

This is not, however, the central thrust of our analysis. Rather, our major emphasis is on contrasting the moderator-mediator functions in ways that delineate the implications of this distinc- tion for theory and research. We focus particularly on the differential implications for choice of experimental design, re- search operations, and plan of statistical analysis.

We also claim that there are conceptual implications of the failure to appreciate the moderator-mediator distinction. Among the issues we will discuss in this regard are missed op- portunities to probe more deeply into the nature of causal mechanisms and integrate seemingly irreconcilable theoretical positions. For example, it is possible that in some problem areas disagreements about mediators can be resolved by treating cer- tain variables as moderators.

The moderator and mediator functions will be discussed at three levels: conceptual, strategic, and statistical. To avoid any misunderstanding of the moderator-mediator distinction by er- roneously equating it with the difference between experimental manipulations and measured variables, between situational and person variables, or between manipulations and verbal self-re- ports, we will describe both actual and hypothetical examples involving a wide range of variables and operations. That is, moderators may involve either manipulations or assessments and either situational or person variables. Moreover, mediators are in no way restricted to verbal reports or, for that matter, to individual-level variables.

Finally, for expository reasons, our analysis will initially stress the need to make clear whether one is testing a moderator or a mediator type of model. In the second half of the article, we provide a design that allows one to test within the structure of the same study whether a mediator or moderator interpreta- tion is more appropriate.

Although these issues are obviously important for a large number of areas within psychology, we have targeted this article for a social psychological audience because the relevance of this distinction is highest in social psychology, which uses experi-

1173

1174 REUBEN M. BARON AND DAVID A. KENNY

mental operations and at the same time retains an interest in organismic variables ranging from individual difference mea- sures to cognitive constructs such as perceived control.

The Nature of Moderators

In general terms, a moderator is a qualitative (e.g., sex, race, class) or quantitative (e.g., level of reward) variable that affects the direction and/or strength of the relation between an inde- pendent or predictor variable and a dependent or criterion vari- able.

Specifically within a correlational analysis framework, a moderator is a third variable that affects the zero-order correla- tion between two other variables. For example, Stern, McCants, and Pettine (1982) found that the positivity of the relation be- tween changing life events and severity of illness was considera- bly stronger for uncontrollable events (e.g., death of a spouse) than for controllable events (e.g., divorce). A moderator effect within a correlational framework may also be said to occur where the direction of the correlation changes. Such an effect would have occurred in the Stern et al. study if controllable life changes had reduced the likelihood of illness, thereby changing the direction of the relation between life-event change and ill- ness from positive to negative.

In the more familiar analysis of variance (ANOVA) terms, a basic moderator effect can be represented as an interaction be- tween a focal independent variable and a factor that specifies the appropriate conditions for its operation. In the dissonance- forced compliance area, for example, it became apparent that the ability of investigators to establish the effects of insufficient justification required the specification of such moderators as commitment, personal responsibility, and free choice (cf. Brehtn& Cohen, 1962).

An example of a moderator-type effect in this context is the demonstration of a crossover interaction of the form that the insufficient justification effect holds under public commitment (e.g., attitude change is inversely related to incentive), whereas attitude change is directly related to level of incentive when the counterattitudinal action occurs in private (cf. Collins & Hoyt, 1972). A moderator-interaction effect also would be said to oc- cur if a relation is substantially reduced instead of being re- versed, for example, if we find no difference under the private condition.1

Toward Establishing an Analytic Framework for Testing Moderator Effects

A common framework for capturing both the correlational and the experimental views of a moderator variable is possible by using a path diagram as both a descriptive and an analytic procedure. Glass and Singer's (1972) finding of an interaction of the factors stressor intensity (noise level) and controllability (periodic-aperiodic noise), of the form that an adverse impact on task performance occurred only when the onset of the noise was aperiodic or unsignaled, will serve as our substantive exam- ple. Using such an approach, the essential properties of a mod- erator variable are summarized in Figure 1.

The model diagrammed in Figure 1 has three causal paths that feed into the outcome variable of task performance: the

Predictor.

Moderator

Predictor X

Moderator

Outcome 1 Variable

Figure 1. Moderator model.

impact of the noise intensity as a predictor (Path a), the impact of controllability as a moderator (Path b), and the interaction or product of these two (Path c). The moderator hypothesis is supported if the interaction (Path c) is significant. There may also be significant main effects for the predictor and the moder- ator (Paths a and b), but these are not directly relevant concep- tually to testing the moderator hypothesis.

In addition to these basic considerations, it is desirable that the moderator variable be uncorrelated with both the predictor and the criterion (the dependent variable) to provide a clearly interpretable interaction term. Another property of the moder- ator variable apparent from Figure 1 is that, unlike the media- tor-predictor relation (where the predictor is causally anteced- ent to the mediator), moderators and predictors are at the same level in regard to their role as causal variables antecedent or exogenous to certain criterion effects. That is, moderator vari- ables always function as independent variables, whereas medi- ating events shift roles from effects to causes, depending on the focus of the analysis.

Choosing an Appropriate Analytic Procedure: Testing Moderation

In this section we consider in detail the specific analysis pro- cedures for appropriately measuring and testing moderational hypotheses. Within this framework, moderation implies that the causal relation between two variables changes as a function of the moderator variable. The statistical analysis must measure and test the differential effect of the independent variable on the dependent variable as a function of the moderator. The way to measure and test the differential effects depends in part on the level of measurement of the independent variable and the mod- erator variable. We will consider four cases: In Case 1, both moderator and independent variables are categorical variables; in Case 2, the moderator is a categorical variable and the inde- pendent variable a continuous variable; in Case 3, the modera-

1 At a conceptual level, a moderator may be more impressive if we go from a strong to a weak relation or to no relation at all as opposed to finding a crossover interaction. That is, although crossover interactions are stronger statistically, as they are not accompanied by residual main effects, conceptually no effect shifts may be more impressive.

THE MODERATOR-MEDIATOR DISTINCTION 1175

tor is a continuous variable and the independent variable is a categorical variable; and in Case 4, both variables are continu- ous variables. To ease our discussion, we will assume that all the categorical variables are dichotomies.

Casel

This is the simplest case. For this case, a dichotomous inde- pendent variable's effect on the dependent variable varies as a function of another dichotomy. The analysis is a 2 X 2 ANOVA, and moderation is indicated by an interaction. We may wish to measure the simple effects of the independent variable across the levels of the moderator (Winer, 1971, pp. 435-436), but these should be measured only if the moderator and the inde- pendent variable interact to cause the dependent variable.

Case 2

Here the moderator is a dichotomy and the independent vari- able is a continuous variable. For instance, gender might moder- ate the effect of intentions on behavior. The typical way to mea- sure this type of moderator effect is to correlate intentions with behavior separately for each gender and then test the difference. For instance, virtually all studies of moderators of the attitude- behavior relation use a correlational test.

The correlational method has two serious deficiencies. First, it presumes that the independent variable has equal variance at each level of the moderator. For instance, the variance of inten- tion must be the same for the genders. If variances differ across levels of the moderator, then for levels of the moderator with less variance, the correlation of the independent variable with the dependent variable tends to be less than for levels of the moderator with more variance. The source of this difference is referred to as a restriction in range (McNemar, 1969). Second, if the amount of measurement error in the dependent variable varies as a function of the moderator, then the correlations be- tween the independent and dependent variables will differ spuri- ously.

These problems illustrate that correlations are influenced by changes in variances. However, regression coefficients are not affected by differences in the variances of the independent vari- able or differences in measurement error in the dependent vari- able. It is almost always preferable to measure the effect of the independent variable on the dependent variable not by correla- tion coefficients but by unstandardized (not betas) regression coefficients (Duncan, 1975). Tests of the difference between re- gression coefficients are given in Cohen and Cohen (1983, p. 56). This test should be performed first, before the two slopes are individually tested.

If there is differential measurement error in the independent variable across levels of the moderator, bias results. Reliabilities would then need to be estimated for the different levels of the moderator, and slopes would have to be disattenuated. This can be accomplished within the computer program LISREL-VI (Joreskog & Sorbom, 1984) by use of the multiple-group op- tion. The levels of the moderator are treated as different groups.

Case 3

In this case, the moderator is a continuous variable and the independent variable is a dichotomy. For instance, the indepen-

Effeet of the independent variable on the dependent variable

Effect of the Independent variable on the dependent variable

Level of the moderator variable

Figure 2. Three different ways in which the moderator changes the effect of the independent variable on the dependent variable: linear (top), qua- dratic (middle), and step (bottom).

dent variable might be a rational versus fear-arousing attitude- change message and the moderator might be intelligence as measured by an IQ test. The fear-arousing message may be more effective for low-IQ subjects, whereas the rational message may be more effective for high-IQ subjects. To measure modera- tor effects in this case, we must know a priori how the effect of the independent variable varies as a function of the moderator. It is impossible to evaluate the general hypothesis that the effect of the independent variable changes as a function of the moder- ator because the moderator has many levels.

Figure 2 presents three idealized ways in which the modera- tor alters the effect of the independent variable on the dependent variable. First, the effect of the independent variable on the de- pendent variable changes linearly with respect to the moderator. The linear hypothesis represents a gradual, steady change in the effect of the independent variable on the dependent variable as the moderator changes. It is this form of moderation that is gen- erally assumed. The second function in the figure is a quadratic function. For instance, the fear-arousing message may be more generally effective than the rational message for all low-IQ sub- jects, but as IQ increases, the fear-arousing message loses its ad- vantage and the rational message is more effective.

The third function in Figure 2 is a step function. At some critical IQ level, the rational message becomes more effective than the fear-arousing message. This pattern is tested by dichot- omizing the moderator at the point where the step is supposed to occur and proceeding as in Case 1. Unfortunately, theories in social psychology are usually not precise enough to specify the exact point at which the step in the function occurs.

The linear hypothesis is tested by adding the product of the moderator and the dichotomous independent variable to the re-

1176 REUBEN M. BARON AND DAVID A. KENNY

gression equasion, as described by Cohen and Cohen (1983) and Cleary and Kessler (1982). So if the independent variable is de- noted as X, the moderator as Z, and the dependent variable as Y, Yis regressed on X, Z, and XZ. Moderator effects are indi- cated by the significant effect of XZ while X and Z are con- trolled. The simple effects of the independent variable for different levels of the moderator can be measured and tested by procedures described by Aiken and West (1986). (Measurement error in the moderator requires the same remedies as measure- ment error in the independent variable under Case 2.)

The quadratic moderation effect can be tested by dichotomiz- ing the moderator at the point at which the function is pre- sumed to accelerate. If the function is quadratic, as in Figure 2, the effect of the independent variable should be greatest for those who are high on the moderator. Alternatively, quadratic moderation can be tested by hierarchical regression procedures described by Cohen and Cohen (1983). Using the same notation as in the previous paragraph, Y is regressed on X, Z, XZ, Z2, and XZ2. The test of quadratic moderation is given by the test of XZ1. The interpretation of this complicated regression equa- tion can be aided by graphing or tabling the predicted values for various values of X and Z.

Case 4

In this case both the moderator variable and the independent variable are continuous. If one believes that the moderator al- ters the independent-dependent variable relation in a step func- tion (the bottom diagram in Figure 2), one can dichotomize the moderator at the point where the step takes place. After dichot- omizing the moderator, the pattern becomes Case 2. The mea- sure of the effect of the independent variable is a regression co- efficient.

If one presumes that the effect of the independent variable (X) on the dependent variable (Y) varies linearly or quadrati- cally with respect to the moderator (Z), the product variable approach described in Case 3 should be used. For quadratic moderation, the moderator squared must be introduced. One should consult Cohen and Cohen (198 3) and Cleary and Kessler (1982) for assistance in setting up and interpreting these regres- sions.

The presence of measurement error in either the moderator or the independent variable under Case 4 greatly complicates the analysis. Busemeyer and Jones (1983) assumed that the moderation is linear and so can be captured by an XZ product term. They showed that measuring multiplicative interactions when one of the variables has measurement error results in low power in the test of interactive effects. Methods presented by Kenny and Judd (1984) can be used to make adjustments for measurement error in the variables, resulting in proper esti- mates of interactive effects. However, these methods require that the variables from which the product variable is formed have normal distributions.

The Nature of Mediator Variables

Although the systematic search for moderator variables is rel- atively recent, psychologists have long recognized the impor- tance of mediating variables. Woodworm's (1928) S-O-R

Independent Variable

Mediator

Outcome Variable

model, which recognizes that an active organism intervenes be- tween stimulus and response, is perhaps the most generic for- mulation of a mediation hypothesis. The central idea in this model is that the effects of stimuli on behavior are mediated by various transformation processes internal to the organism. Theorists as diverse as Hull, Tolman, and Lewin shared a belief in the importance of postulating entities or processes that inter- vene between input and output. (Skinner's blackbox approach represents the notable exception.)

General Analytic Considerations

In general, a given variable may be said to function as a medi- ator to the extent that it accounts for the relation between the predictor and the criterion. Mediators explain how external physical events take on internal psychological significance. Whereas moderator variables specify when certain effects will hold, mediators speak to how or why such effects occur. For example, choice may moderate the impact of incentive on atti- tude change induced by discrepant action, and this effect is in turn mediated by a dissonance arousal-reduction sequence (cf. Brehm& Cohen, 1962).

To clarify the meaning of mediation, we now introduce a path diagram as a model for depicting a causal chain. The basic causal chain involved in mediation is diagrammed in Figure 3. This model assumes a three-variable system such that there are two causal paths feeding into the outcome variable: the direct impact of the independent variable (Path c) and the impact of the mediator (Path b). There is also a path from the independent variable to the mediator (Path a).

A variable functions as a mediator when it meets the follow- ing conditions: (a) variations in levels of the independent vari- able significantly account for variations in the presumed media- tor (i.e., Path c), (b) variations in the mediator significantly ac- count for variations in the dependent variable (i.e., Path b), and (c) when Paths a and b are controlled, a previously significant relation between the independent and dependent variables is no longer significant, with the strongest demonstration of media- tion occurring when Path c is zero. In regard to the last condi- tion we may envisage a continuum. When Path c is reduced to zero, we have strong evidence for a single, dominant mediator. If the residual Path c is not zero, this indicates the operation of multiple mediating factors. Because most areas of psychology, including social, treat phenomena that have multiple causes, a more realistic goal may be to seek mediators that significantly decrease Path c rather than eliminating the relation between the independent and dependent variables altogether. From a theo- retical perspective, a significant reduction demonstrates that a given mediator is indeed potent, albeit not both a necessary and a sufficient condition for an effect to occur.

THE MODERATOR-MEDIATOR DISTINCTION 1177

Testing Mediation

An ANOVA provides a limited test of a mediational hypothesis

as extensively discussed in Fiske, Kenny, and Taylor (1982).

Rather, as recommended by Judd and Kenny (1981 b), a series

of regression models should be estimated. To test for mediation,

one should estimate the three following regression equations:

first, regressing the mediator on the independent variable; sec-

ond, regressing the dependent variable on the independent vari-

able; and third, regressing the dependent variable on both the

independent variable and on the mediator. Separate coefficients

for each equation should be estimated and tested. There is no

need for hierarchical or stepwise regression or the computation

of any partial or semipartial correlations.

These three regression equations provide the tests of the link-

ages of the mediational model. To establish mediation, the fol-

lowing conditions must hold: First, the independent variable

must affect the mediator in the first equation; second, the inde-

pendent variable must be shown to affect the dependent variable

in the second equation; and third, the mediator must affect the

dependent variable in the third equation. If these conditions all

hold in the predicted direction, then the effect of the indepen-

dent variable on the dependent variable must be less in the third

equation than in the second. Perfect mediation holds if the inde-

pendent variable has no effect when the mediator is controlled.

Because the independent variable is assumed to cause the me-

diator, these two variables should be correlated. The presence

of such a correlation results in multicollinearity when the

effects of independent variable and mediator on the dependent

variable are estimated. This results in reduced power in the test

of the coefficients in the third equation. It is then critical that

the investigator examine not only the significance of the co-

efficients but also their absolute size. For instance, it is possible

for the independent variable to have a smaller coefficient when

it alone predicts the dependent variable than when it and the

mediator are in the equation but the larger coefficient is not

significant and the smaller one is.

Sobel (1982) provided an approximate significance test for

the indirect effect of the independent variable on the dependent

variable via the mediator. As in Figure 3, the path from the

independent variable to the mediator is denoted as a and its

standard error is sa; the path from the mediator to the depen-

dent variable is denoted as b and its standard error is s/,. The

exact formula, given multivariate normality for the standard er-

ror of the indirect effect or ab, is this:

Sobers method omits the term SaSt, but that term ordinarily

is small. His approximate method can be used for more compli-

cated models.

The use of multiple regression to estimate a mediational

model requires the two following assumptions: that there be no

measurement error in the mediator and that the dependent vari-

able not cause the mediator.

The mediator, because it is often an internal, psychological

variable, is likely to be measured with error. The presence of

measurement error in the mediator tends to produce an under-

estimate of the effect of the mediator and an overestimate of

the effect of the independent variable on the dependent variable

when all coefficients are positive (Judd & Kenny, 1981 a). Obvi-

ously this is not a desirable outcome, because successful media-

tors may be overlooked.

Generally the effect of measurement error is to attenuate the

size of measures of association, the resulting estimate being

closer to zero than it would be if there were no measurement

error (Judd & Kenny, 1981 a). Additionally, measurement error

in the mediator is likely to result in an overestimate in the effect

of the independent variable on the dependent variable. Because

of measurement error in the mediator, effects of the mediator

on the dependent variable cannot totally be controlled for when

measuring the effects of the independent variable on the depen-

dent variable.

The overestimation of the effects of the independent variable

on the dependent variable is enhanced to the extent that the

independent variable causes the mediator and the mediator

causes the dependent variable. Because a successful mediator is

caused by the independent variable and causes the dependent

variable, successful mediators measured with error are most

subject to this overestimation bias.

The common approach to unreliability is to have multiple

operations or indicators of the construct. Such an approach re-

quires two or more operationalizations or indicators of each

construct. One can use the multiple indicator approach and es-

timate mediation paths by latent-variable structural modeling

methods. The major advantages of structural modeling tech-

niques are the following: First, although these techniques were

developed for the analysis of nonexperimental data (e.g., field-

correlational studies), the experimental context actually

strengthens the use of the techniques. Second, all the relevant

paths are directly tested and none are omitted as in ANOVA.

Third, complications of measurement error, correlated mea-

surement error, and even feedback are incorporated directly

into the model. The most common computer program used to

estimate structural equation models is LISREL-VI (Joreskog

& Sorbom, 1984). Also available is the program EQS (Bentler, 1982).

We now turn our attention to the second source of bias in

the mediational chain: feedback. The use of multiple regression

analysis presumes that the mediator is not caused by the depen-

dent variable. It may be possible that we are mistaken about

which variable is the mediator and which is the dependent vari-

able.

Smith (1982) has proposed an ingenious solution to the prob-

lem of feedback in mediational chains. His method involves the

manipulation of two variables, one presumed to cause only the

mediator and not the dependent variable and the other pre-

sumed to cause the dependent variable and not the mediator.

Models of this type are estimated by two-stage least squares or

a related technique. Introductions to two-stage least squares are

in James and Singh (1978), Duncan (1975), and Judd and

Kenny (198la). The earlier-mentioned structural modeling

procedures can also be used to estimate feedback models.

Overview of Conceptual Distinctions Between Moderators and Mediators

As shown in the previous section, to demonstrate mediation

one must establish strong relations between (a) the predictor

1178 REUBEN M. BARON AND DAVID A. KENNY

and the mediating variable and (b) the mediating variable and

some distal endogenous or criterion variable. For research ori-

ented toward psychological levels of explanation (i.e., where the

individual is the relevant unit of analysis), mediators represent

properties of the person that transform the predictor or input

variables in some way. In this regard the typical mediator in

cognitive social psychology elaborates or constructs the various

meanings that go "beyond the information given" (Bruner,

1957). However, this formulation in no way presupposes that

mediators in social psychology are limited to individualistic or

"in the head" mechanisms. Group-level mediator constructs

such as role conflict, norms, groupthink, and cohesiveness have

long played a role in social psychology. Moreover, with the in-

creasing interest in applied areas, there is likely to be an increas-

ing use of mediators formulated at a broader level of analysis.

For example, in the area of environmental psychology, territo-

rial constructs such as defensible space (Newman, 1972) or the

role of sociopetal versus sociofugal sitting patterns (Sommer,

1969) clearly take the mediator concept beyond the intraorga-

nismic level. Despite this range of application of the mediator

concept, it is in principle capable of rigorous tests at the group

level. For example, Zaccaro (1981) has attempted to support a

mediator interpretation of cohesiveness using a strategy com-

bining experimental manipulations with causal modeling.

In addition, whereas mediator-oriented research is more in-

terested in the mechanism than in the exogenous variable itself

(e.g., dissonance and personal-control mediators have been im-

plicated as explaining an almost unending variety of predic-

tors), moderator research typically has a greater interest in the

predictor variable per se. However, whether a given moderator-

oriented investigation is strongly committed to a particular pre-

dictor is likely to vary widely. Although a pragmatic-predictor

orientation is typical in industrial psychology, where the predic-

tor is often a test, in social psychology moderators are often as

theoretically derived as mediators.

Strategic Considerations

Moderator variables are typically introduced when there is

an unexpectedly weak or inconsistent relation between a pre-

dictor and a criterion variable (e.g., a relation holds in one set-

ting but not in another, or for one subpopulation but not for

another). The recent use by Snyder (1983) and others (cf. Sher-

man & Fazio, 1983) of the self-monitoring variable as a means

to improve the ability of personality traits to predict behavioral

criteria is illustrative. Mediation, on the other hand, is best done

in the case of a strong relation between the predictor and the

criterion variable.

Moderator to mediator. In addition, there may be a wide vari-

ation in the strategic functions served by moderators and medi-

ators. In this regard one may begin with a moderator orienta-

tion and end up elucidating a mediator process, or begin with a

mediator approach and derive moderator-type interventions.

For example, let us assume that race functioned as a moderator

for the efficacy of certain instructional techniques, such that a

given technique (e.g., programmed instruction) works better for

one racial group than for another. One could view such a finding

as just the first step toward specifying the underlying dimen-

sion^) that account for the instructional effect. For example, it

could be argued that the real issue is a difference in anxiety

level; that is, when black and white children are placed in mid-

dle-class learning environments, black children may experience

a higher level of evaluative anxiety. Therefore, evaluative-anxi-

ety level may be postulated to mediate the differential effec-

tiveness of a given instructional technique. Thus, here we have

a situation where a moderator variable has been useful in sug-

gesting a possible mediator variable. What is at stake in this

regard is selecting moderators that do more than improve pre-

dictive power. For example, race would be preferred over social

class as a moderator if race was more able to tell us something

about the processes underlying test performance.

A similar point can be made in regard to the current use of

moderator variables in personality research. That is, if two vari-

ables have equal power as potential moderators of a trait-behav-

ior relation, one should choose the variable that more readily

lends itself to a specification of a mediational mechanism. For

example, the self-monitoring variable both improves predictive

efficacy and suggests mediational processes involving attention

deployment. Indeed, such a strategy of selection points to one

way to circumvent the oft-made criticism of moderator vari-

ables that we have no principled procedure for reducing their

proliferation (cf. Epstein, 1983).

Mediator to moderator. The relation may also work in the

opposite direction. Differences in perceived control may be

found to mediate the relation between social density and decre-

ments in task performance. In this situation a mediator may

suggest an environmental intervention to prevent density from

having adverse effects. For example, what appears to be needed

is an intervention that would serve to increase the controllabil-

ity of social encounters. This might take the form of architec-

tural variation, for example, suite versus corridor dorm ar-

rangements, or involve various types of restrictions on change

or unpredictable social encounters, for example, institution of

quiet hours. What is at stake here is the choice of mediators that

point to the possibility of environmental intervention.

Thus, at times moderator effects may suggest a mediator to

be tested at a more advanced stage of research in a given area.

Conversely, mediators may be used to derive interventions to

serve applied goals.

Operational Implications

There are a number of implications of the moderator-media-

tor distinction at the level of the choice of research operations.

First, the moderator interpretation of the relation between the

stressor and control typically entails an experimental manipula-

tion of control as a means of establishing independence between

the stressor and control as a feature of the environment separate

from the stressor. When control is experimentally manipulated

in service of a moderator function, one need not measure per-

ceived control, which is the cognitive intraorganismic concept

If it is measured, perceived control serves as a manipulation

check.

A theory that assigns a mediator role to the control construct,

however, is only secondarily concerned with the independent

manipulation of control. The most essential feature of the hy-

pothesis is that perceived control is the mechanism through

which the stressor affects the outcome variable. For such a the-

THE MODERATOR-MEDIATOR DISTINCTION 1179

Manipulation of Control

Control (P)

Outcome (0)

x -̂— Manipulation of Control

X Stressor

(CS)

Figure 4. Path diagram combining mediation and moderation.

ory, an independent assessment of perceived control is essential for conceptual reasons, as opposed to methodological reasons as in the moderator case. Because of the conceptual status of this assessment in the mediator case, one's main concern is the demonstration of construct validity, a situation that ideally re- quires multiple independent and converging measurements (Campbell & Fiske, 1959). Thus, when mediation is at issue we need to increase both the quality and the quantity of the data.

A Framework for Combining Mediation and Moderation

Figure 4 presents a combined model with both mediation and moderation. The variable control has both mediator and moderator status in the model. The stressor in the figure is the independent variable, and the dependent variable is labeled the outcome. We denote manipulated control as C, the stressor as S, the C X S interaction as CS, measured perceived control as P, the P X S interaction as PS, the C X P interaction as CP, the C X P X S interaction as CPS, and the outcome as O. We assume that both the manipulation of control and the stressor are di- chotomies and that all moderator effects are linear.

The analysis proceeds in three steps. In Step 1, the effects of the manipulated variables on O are assessed. In Step 2, the effects to and from P are assessed. In Step 3, the effect from PS is assessed.

Step 1. The Step 1 regression is illustrated in Figure 1. This step is a simple 2 x 2 ANOVA on the outcome variable. If C has a significant effect on O, then control may be a mediating variable of the stressor effect on the outcome. If S affects O, then it is sensible to evaluate the mediating effects of perceived control. These two effects are supportive of the mediation hy- pothesis, but direct evidence for mediation is provided in the next step. Finally, the CS effect indicates moderation.

Step 2. The Step 2 regressions are illustrated in Figure 4. In this step, two equations are estimated. First, P is regressed on C, S, and CS. This can be more easily accomplished by a 2 X 2 ANOVA. Second, O is regressed on C, S, P, and CS. For P to mediate the S to O relation, S must affect P and P must affect O. If there is complete mediation, then S does not affect O when

P is controlled. To strengthen the claim that it is perceived con- trol that mediates the relation, C should strongly affect P but should not affect O. If C affects O, then it is indicated that some aspect the control manipulation is different from perceived con- trol.

There are two remaining paths in Step 2. They are the paths from CS to P and to O. If CS affects P, then the control manipu- lation is not equally effective in determining perceived control across the levels of the stressor. The stressor moderates the effectiveness of the manipulation. The final Step 2 path is the one from CS to O. Let us assume that CS affects O in the Step 1 regression, and in Step 2 CS has a weaker effect on O. Then the interpretation is that P has mediated the CS effect on O. We have what might be termed mediated moderation. Mediated moderation would be indicated by CS affecting O in Step 1, and in Step 2 CS affecting P and P affecting C. So it is possible for P to mediate both the effect of S on O and the effect of CS on O.

Step 3. In this step, one equation is estimated. The variable O is regressed on C, S, P, CS, and PS. This equation is identical to the second Step 2 equation, but the PS term has been added. The key question is the extent to which the CS effect on O is reduced in moving from Step 2 to Step 3. If it has been, then we can say that P and not C moderates the S to O relation. In a sense, P mediates the moderating effects of C on S. For this to happen, CS must have less of an effect on O at Step 3 than at Step 2, and PS must affect O. Finally in Step 2, C should affect P, which will result in CS and PS being correlated.

There are then two ways in which the CS effect on O can be explained by P. It can be explained by P because the control manipulation is differentially affecting perceived control for the levels of the stressor. Or, the CS interaction can be funnelled through the PS interaction. The former explanation would change what was a moderator effect into a mediator effect, and the latter would keep the moderator explanation but enhance the meaning of the moderator construct.

We present the three step hypotheses because they represent a series of reasonable hypotheses. If one wished, further models could be estimated. For instance, one could regress O on C, S, P, CS, and CP. The presence of the CP effect, as well as media- tional effects by P of the S to O relation, would be indicative of moderated mediation (James & Brett, 1984). That is, the medi- ational effects of P vary across the levels of C. The second-order interaction effect, CPS, could also be estimated and tested.

Implications and Applications of the Moderator-Mediator Distinction

In this section, we take the themes developed in the three pre- vious sections and apply them to three areas of social psycholog- ical research. These areas are personal control, the behavior- intention relation, and linking traits and attitudes to behavior.

Clarifying the Meaning of Control

Many investigations of the impact of personal control in so- cial and environmental psychology have been methodologically (but not theoretically) ambivalent with respect to the control variable's causal status. Investigators have tended to use experi- mental manipulations of personal control along with ANOVA-

1180 REUBEN M. BARON AND DAVID A. KENNY

type analyses. This practice leads to serious difficulties of inter-

pretation when a researcher intends to investigate one function

of control but studies only the other function. For example,

Langer and Saegert (1977) and Rodin, Solomon, and Metcalf

(1978) sought to examine the mediational role of lessened con-

trol for crowding. Given this mediator interpretation, it is not

enough to demonstrate by use of an experimental manipulation

that high density creates more perceived crowding than does

low density only when there is a low availability of control, for

example, the ability to escape from the high-density situation.

To provide stronger evidence of mediation, an independent as-

sessment of the impact of the stressor on some index of organis-

mic control is required. Only when this is done can we establish

the crucial link between perceived control and the criterion. Be-

cause the Langer and Saegert and Rodin, Solomon, and Metcalf

studies failed to provide an independent assessment of control,

they lack the requisite information to establish a strong case for

control as a mediator. Moreover, because Langer and Saegert

failed to find differential effects for density under varying levels

of their control manipulation, that is, a Control X Density inter-

action, they are not even in a position to make moderator-vari-

able claims.

Finally, there is another important role that the present mod-

erator-mediator distinction can play in the domain of crowding

theory and research. Although a control-mediation model of

crowding is generally accepted (e.g., Baron & Rodin, 1978; Sto-

kols, 1976), there are significant dissenters such as Freedman

(1975). Given the present status of the evidence, it appears

much easier to support the claim that control moderates, as op-

posed to mediates, the density-crowding relation. Such an inter-

pretation would leave open the possibility that other factors,

such as an arousal-labeling or an arousal-amplification mecha-

nism, mediate the effects of density (i.e., Freedman, 1975; Wor-

chel & Teddlie, 1976).

Behavior Intention-Behavior Relation

Because Fishbein and Ajzen's (1975; Ajzen & Fishbein,

1980) attitude theory of reasoned action is in general highly so-

phisticated at both the conceptual and quantitative levels, it

provides a good example of the extent of confusion regarding

mediators and moderators. Moreover, this model, as Bentler

and Speckart (1979) have demonstrated, readily lends itself to

a causal modeling approach. Specifically, behavioral intention

(BI) is a clear-cut example of a mediator concept in social psy-

chology. Fishbein and Ajzen assumed that the impact of both

attitudes and normative factors on behavior (B) is mediated

through behavioral intentions. Although one can disagree with

Fishbein and Ajzen's assertion that attitudes and norms can in-

fluence behavior only indirectly through behavioral intention

(see Bentler & Speckart, 1979; Songer-Nocks, 1976), their for-

mulation represents a correct statement of a strong mediator

position.

Surprisingly, however, given the elegance of their general

model, similar care was not taken regarding the nature of the

BI-B link. For example, Fishbein and Ajzen's treatment of this

relation failed to distinguish between variables that are likely to

moderate and those likely to mediate this relation. Such diverse

variables as gender, time delay, perceived likelihood of co-work-

ers complying, skill, and resources were all treated as mediating

factors (Fishbein & Ajzen, 1975, pp. 377-381).

From the present perspective, such an approach ignores the

possibility that some of these factors are best conceptualized

and treated statistically as moderators whereas others are best

viewed as mediators. For example, gender of subjects is best

viewed as a moderator of the BI-B relation. Given this distinc-

tion, different analysis strategies are entailed at the statistical

level. Specifically, Fishbein and Ajzen tested the importance of

given factors by looking at the impact on the multiple correla-

tion of dropping or adding a variable. This type of strategy,

which is analogous to treating a covariate as a potential media-

tor, is best used to infer mediation as opposed to moderation.

For testing a moderator interpretation, what is required is a

term involving the product of BI and the hypothesized modera-

tor, for example, one would construct a Gender X BI interaction

term to test for gender as a moderator variable.

Finally, although Fishbein's (1966) finding that intentions are

better predictors for women than for men is in itself best viewed

as a moderator effect, a sensitivity to the present set of issues

prompts further analyses. For example, if we ask why gender

has such effects on sexual intentions, it is possible that we will

be led to postulate a mediator that transcends gender. For exam-

ple, it might be argued that intentions predict better for women

because women are less impulsive than men in regard to the

timing of sexual behavior.

Linking Global Dispositions to Behavior:

Attitudes and Traits

Of all the current areas in social psychology, the one where

the use of what we have referred to as the combined model (see

Figure 4) is perhaps the strongest is the prediction of social be-

havior from global dispositional variables. In this regard, the

trait-behavior and the attitude-behavior relations have recently

been explicitly approached from the moderator-variable per-

spective. For example, the predictive efficacy of both traits and

attitudes have improved when self-monitoring (Snyder, 1983)

and self-consciousness (Scheier, 1980), respectively, have been

used as moderator variables. Moreover, investigators such as

Snyder and Ickes (1985) and Sherman and Fazio (1983, p. 327)

have asked the following questions: By what process or pro-

cesses do attitudes toward an object affect behavior toward the

object? Likewise, what conceivable processes link traits to be-

havior?

What such suggestions lack is precisely the kind of unified

conceptual and analytic framework presented in our combined

moderator-mediator example (see Figure 4). By using such a

path analytic framework, one could take a variable such as

differences in self-monitoring orientation and simultaneously

establish both its role as a moderator and the nature of the me-

diation process through which it has an impact on a given class

of behavior. At an operational level, such a strategy compels one

to go beyond merely measuring differences in self-monitoring

(the moderator paths) to operationalizing a mediator mecha-

nism, for example, providing some measure of differential at-

tention or variables in impression management.

Further, placing both moderator and mediator variables

within the same causal system helps to make salient the more

THE MODERATOR-MEDIATOR DISTINCTION 1181

dynamic role played by mediators as opposed to moderators

(Finney, Mitchell, Cronkite, & Moos, 1984). Specifically, intro-

ducing a moderator variable merely involves a relatively static

classification procedure. For example, self-monitoring as a

moderator sets up a partition of people holding a given personal-

ity trait into subgroups of those more or less likely to translate

their psychological dispositions into overt actions; that is, the

emphasis is on who does what. On the other hand, linking the

Self-Monitoring X Trait relation to a specific mediating mecha-

nism implies that variations in self-monitoring elicit or insti-

gate different patterns of coping or information processing that

cause people to become more or less consistent with their atti-

tudes in their behavior. Here the prior condition allows us to

discover different states that cause individuals to act differ-

ently—a more dynamic conception of how third variables op-

erate.

Summary

In this article we have attempted to achieve three goals. First,

by carefully elaborating the many ways in which moderators

and mediators differ, we have tried to make theorists and re-

searchers aware of the importance of not using the terms mod-

erator and mediator interchangeably. We then went beyond this

largely pedagogical function and delineated the conceptual and

strategic implications of making use of this distinction with re-

gard to a wide range of phenomena, including control and

stress, attitudes, and personality traits. We have also provided

the first specific compendium of analytic procedures appropri-

ate for making the most effective use of the moderator-media-

tor distinction both separately and in terms of a broader causal

system that includes both moderators and mediators.

References

Aiken, L. S., & West, S. G. (1986). Use and interpretation of regression

analysis models containing interactions and power polynomials. Un- published manuscript, Arizona State University, Tempe.

Ajzen, I., & Fishbein, M. (1980). Understanding attitudes andpredict-

ing social behavior. Englewood Cliffs, NJ: Prentice-Hall. Baron, R. M., & Rodin, J. (1978). Personal control as a mediator of

crowding. In A. Baum, J. Singer, & S. Valins (Eds.), Advances in envi-

ronmental psychology (pp. 145-190). Hillsdale, NJ: Erlbaum. Bentler, P. M. (1982). Theory and implementation ofEQS:A structural

equations program. Unpublished manuscript, University of Califor-

nia, Los Angeles.

Bentler, P. M., & Speckart, G. (1979). Models of attitude-behavior re-

lations. Psychological Review, 86, 452—464.

Brehm, J. W., & Cohen, A. R. (1962). Explorations in cognitive disso-

nance. New York: Wiley.

Bruner, J. S. (1957). Going beyond the information given. InH. Gruber,

G. Terrell, & M. Wertheimer (Eds.), Contemporary approaches to

cognition (pp. 258-290). Cambridge, MA: Harvard University Press.

Busemeyer, J. R., & Jones, L. E. (1983). Analysis of multiplicative com-

bination rules when the causal variables are measured with error. Psy-

chological Bulletin, 93, 549-563.

Campbell, D. X, & Fiske, D. W. (1959). Convergent and discriminant

validation by the multitrait-multimethod matrix. Psychological Bul-

letin, 56. 81-105. Cleary, P. D., & Kessler, R. C. (1982). The estimation and interpretation

of modified effects. Journal of Health and Social Behavior, 23, 159-

169.

Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation

analysis for the behavioral sciences (2nded.). Hillsdale, NJ: Erlbaum.

Collins, B. E., & Hoyt, M. G. (1972). Personal responsibility for conse-

quences: An integration and extension of the forced compliance liter-

ature. Journal of Experimental Social Psychology, 8, 558-593.

Duncan, O. D. (1975). Introduction to structural equation models. New

Tfork: Academic Press.

Epstein, S. (1983). Aggregation and beyond: Some basic issues in the

prediction of behavior. Journal of Personality, 51, 360-392.

Findley, M. J., & Cooper, H. M. (1983). Locus of control and academic

achievement: A literature review. Journal of Personality and Social

Psychology, 44, 419-427.

Finney, J. W., Mitchell, R. E., Cronkite, R., & Moos, R. H. (1984).

Methodological issues in estimating main and interactive effects: Ex-

amples from the coping/social support and stress field. Journal of

Health and Social Behavior, 10, 85-98.

Fishbein, M. (1966, November). Sexual behavior and prepositional

control. Paper presented at the meeting of the Psychonomic Society,

Chicago.

Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behav-

ior. Reading, MA: Addison-Wesley.

Fiske, S. T., Kenny, D. A., & Taylor, S. E. (1982). Structural models for

the mediation of salience effects on attribution. Journal of Experi-

mental Social Psychology, 18, 105-127.

Freedman, J. L. (1975). Crowding and behavior. San Francisco: Free-

man.

Glass, D., & Singer, J. (1972). Urban stress: Experiments on noise and

social stressors. New \brk: Academic Press.

Harkins, S. G., Latane, B., & Williams, K. (1980). Social loafing: Allo-

cating effort or taking it easy? Journal of Experimental Social Psy-

chology, 16, 457-465.

James, L. R., & Brett, J. M. (1984). Mediators, moderators, and tests

for mediation. Journal of Applied Psychology, 69, 307-321.

James, L. R., & Singh, B. K. (1978). An introduction to logic, assump-

tions, and basic analytic procedures of two-stage least squares. Psy-

chological Bulletin, 85, 1104-1123.

Joreskog, K. J., & Sorbom, D. (1984). L1SREL-VI—Estimation of lin-

ear structural equations by maximum likelihood methods (3rd ed.).

Mooresville, IN: Scientific Software.

Judd, C. M., & Kenny, D. A. (1981a). Estimating the effects of social

interventions. New "Vbrk: Cambridge University Press.

Judd, C. M., & Kenny, D. A. (198 Ib). Process analysis: Estimating me-

diation in evaluation research. Evaluation Research, 5, 602-619.

Kenny, D. A., & Judd, C. M. (1984). Estimating the nonlinear and inter-

active effects of latent variables. Psychological Bulletin, 96, 201-210.

Langer, E. J., & Saegert, S. (1977). Crowding and cognitive control.

Journal of Personality and Social Psychology, 35, 175-182.

McNemar, Q. (1969). Psychological statistics (4th ed.). New York: Wi- ley.

Newman, O. (1972). Defensible space. New York: Macmillan.

Rodin, J., Solomon, S., & Metcalf, J. (1978). Role of control in mediat-

ing perceptions of density. Journal of Personality and Social Psychol-

ogy, 36, 988-999.

Scheier, M. F. (1980). The effect of public and private self-consciousness

on the public expressions of personal beliefs. Journal of Personality

and Social Personality, 39, 514-521.

Sherman, S. J., & Fazio, R. H. (1983). Parallels between attitudes and

traits as predictors of behavior. Journal of Personality, 51, 308-345.

Smith, E. R. (1982). Beliefs, attributions, and evaluations: Nonhierar-

chical models of mediation in social cognition. Journal of Personality

and Social Psychology, 43, 248-259.

Snyder, M. (1983). The influence of individuals on situations: Implica-

tions for understanding the links between personality and social be-

havior. Journal of Personality, 51,497-516.

1182 REUBEN M. BARON AND DAVID A. KENNY

Snyder, M., & Ickes, W. (1985). Personality and social behavior. In G. Stokols, D. (1976). The experience of crowding in primary and second- Lindzey & E. Aronson (Eds.), Handbook of social psychology (3rd ary environments. Environment and Behavior, 8,49-86. ed., pp. 883-948). Reading, MA: Addison-Wesley. Winer, B. J. (1971). Statistical principles in experimental design (2nd

Sobel, M. E. (1982). Asymptotic confidence intervalsfor indirect effects ed.). New York: McGraw-Hill. in structural equations models. In S. Leinhart (Ed.), Sociological Woodworth, R. S. (1928). Dynamic psychology. In C. Murchison (Ed.), methodology 1982 (pp. 290-312). San Francisco: Jossey-Bass. Psychologies of 1925. Worcester, MA: Clark University Press.

Sommer, R. (1969). Personal space. Englewood Clifls, NJ: Prentice- Worchel, S., & Teddlie, C. (1976). The experience of crowding. Journal Hall. of Personality and Social Psychology, 34, 30-40.

Songer-Nocks, E. (1976). Situational factors affecting the weighting of Zaccaro, S. J. (1981). The effects of cohesion source on process loss in predictor components in the Fishbein model. Journal ofExperimen- group performance. Unpublished doctoral dissertation, University of tal Social Psychology, 12,56-69. Connecticut, Storrs.

Stern, G. S., McCants, T. R., & Pettine, P. W. (1982). The relative contri- bution of controllable and uncontrollable life events to stress and ill- Received August 7, 1985 ness. Personality and Social Psychology Bulletin, S, 140-145. Revision received July 14, 1986 •

Instructions to Authors

Authors should prepare manuscripts according to the Publication Manual of the A merican Psy-

chological Association (3rd ed.). Articles not prepared according to the guidelines of the Man-

ual will not be reviewed. All manuscripts must include an abstract of 100-150 words typed on

a separate sheet of paper. Typing instructions (all copy must be double-spaced) and instructions

on preparing tables, figures, references, metrics, and abstracts appear in the Manual. Also, all

manuscripts are subject to editing for sexist language.

APA policy prohibits an author from submitting the same manuscript for concurrent consid-

eration by two or more journals. Also, authors of manuscripts submitted to APA journals are

expected to have available their raw data throughout the editorial review process and for at least

5 years after the date of publication. For further information on content, authors should refer

to the editorial in the March 1979 issue of this journal (Vol. 37, No. 3, pp. 468-469).

The reference citation for any article in any JPSP section follows APA's standard reference

style for journal articles; that is, authors, year of publication, article title, journal title, volume

number, and page numbers. The citation does not include the section title.

Authors will be required to state in their initial submission letter or sign a statement that they

have complied with APA ethical standards in the treatment of their sample, human or animal.

A copy of the APA Ethical Principles may be obtained from the APA Ethics Office, 1200 17th Street, N.W., Washington, DC 20036.

Anonymous reviews are optional, and authors who wish anonymous reviews must specifically

request them when submitting their manuscripts. Each copy of a manuscript to be anonymously

reviewed should include a separate title page with authors' names and affiliations, and these

should not appear anywhere else on the manuscript. Footnotes that identify the authors should

be typed on a separate page. Authors should make every effort to see that the manuscript itself

contains no clues to their identities.

Manuscripts should be submitted in quadruplicate, and all copies should be clear, readable,

and on paper of good quality. A dot matrix or unusual typeface is acceptable only if it is clear

and legible. Dittoed and mimeographed copies will not be considered. Authors should keep a

copy of the manuscript to guard against loss. Mail manuscripts to the appropriate section editor.

Editors' addresses appear on the inside front cover of the journal.

Section editors reserve the right to redirect papers among themselves as appropriate unless

an author specifically requests otherwise. Rejection by one section editor is considered rejection

by all, therefore a manuscript rejected by one section editor should not be submitted to another.