Application and Extension of Existing Theory 465the predictive power of dispositional factors should be mani-
fest only when situational pressure is weak, not when it is
strong. Carlo, Eisenberg, Troyer, Switzer, and Speer (1991)
claimed support for this distinction between weak and strong
pressure when predicting prosocial behavior. Within these
more recent studies, then, dispositional predictors have
fared better than in earlier work. Still, correlations between
personality measures and prosocial behavior—however
measured—rarely rise above .30 to .40, leaving 85% to 90%
of the variance unaccounted for.
At the same time that dispositional predictors were being
revived, the health of situational predictors took a turn for
the worse: Their ecological validity was questioned (Bar-Tal,
1984). Could one expect a situational predictor of single-act
helping by college students in a controlled laboratory experi-
ment to be equally powerful in predicting naturally occurring
prosocial behavior outside the lab, such as volunteerism
(Clary & Snyder, 1991)?
Proliferating Predictors and Predictions
Since 1970, proposed predictors of prosocial behavior have
proliferated well beyond the initial dichotomy between
dispositional and situational factors. Krebs and Miller (1985)
presented an interlocking three-tier classification. Most distal
from the specific prosocial behavior are biological and cul-
tural predictors (see also Fiske, 1992). These predictors com-
bine to produce enduring dispositional characteristics, which
are more proximal. Dispositional factors then combine with
situational factors to produce cognitive and affective reac-
tions, which are considered the most proximal predictors of
prosocial behavior. Within each of these broad classes, nu-
merous specific variables can be identified.
In additions to proliferating predictors, there are also many
different forms of prosocial behavior to be predicted, and the
variables that predict one form may not predict another. For
example, within the domain of helping are rescuing, donating,
assisting, volunteering, and giving social support (Pearce &
Amato, 1980). Moreover, each of these categories includes a
wide range of specific behaviors. One can assist by holding a
door, answering a request for directions, splinting a broken
leg at the scene of an automobile accident, securing false pa-
pers for a Jew in Nazi Europe, or enabling a suicide. One can
volunteer to serve on the board of directors for the local sym-
phony, to call potential blood donors, to be a buddy for some-
one who has AIDS, or to join the rescue squad. Critics
claim—and research supports the claim (Levine, Martinez,
Brase, & Sorenson, 1994; Omoto & Snyder, 1995)—that
variables accounting for variance in one form of prosocial be-
havior in one setting are not likely to account for the same
amount of variance (if any) in other forms of behavior or in
other settings. Talk of prediction based on interactions among
person, situation, and behavior has become common (e.g.,
Bandura, 1991; Carlo et al., 1991).
One need not pursue this logic very far—adding predic-
tors, behaviors to be predicted, situations in which prediction
can be made, and populations for which predictions can be
made—to realize that a general variance-accounted-for an-
swer to the question of why people act prosocially is impos-
sible. All one can hope for is the identification of predictors
that account for a specific prosocial behavior in a specific
situation for a specific population at a specific time (Snyder,
1993). Although useful to address some applied questions,
such research is apt to become ideographic rather than nomo-
thetic (Allport, 1961), with very little generalizability.APPLICATION AND EXTENSION
OF EXISTING THEORYWell aware of the limited, ad hoc nature of a variance-
accounted-for approach, Lewin (1951) reminded us, “There
is nothing so practical as a good theory” (p. 169). In opposi-
tion to the Aristotelian approach to science that guides the
variance-accounted-for strategy, in which the scientist’s goal
is to identify essential features to predict outcomes, Lewin
advocated a Galilean approach. Galileo’s goal was to identify
underlying genotypic (conditional-genetic) constructs and
the highly general—even universal—relations among them
that account for observable phenotypic events. Lewin was
convinced that explanatory theories developed and tested
following Galileo are of far more practical value than are
explanations developed following Aristotle, even though the
Galilean model relies on contrived laboratory experiments
rather than on direct, real-world observation.
Psychologists approaching the study of prosocial behavior
from Lewin’s Galilean perspective are not likely to look to
empirical research to identify predictors accounting for the
most variance. They are likely instead to look to existing the-
ory about genotypic psychological processes, using research
to illustrate and document the relevance of these processes to
understanding prosocial behavior. At least seven broad theo-
retical perspectives have been applied in this way: social
learning, tension reduction, norms and roles, exchange or
equity, attribution, esteem enhancement/maintenance, and
moral reasoning. Let us briefly consider each of these.Social LearningSocial learning theory suggests that if you want to know
why people act prosocially, you should consider their learn-
ing history. You should consider not only the rewards and