Impressions of Self-Directed Action

Lory Britt
David R. Heise

Department of Sociology
Indiana University
Bloomington IN 47405

Social Psychology Quarterly, 55 (1992): 335-350


Ratings of evaluation, potency, and activity (EPA) for 256 combinations of identity and self-directed behavior were predicted from EPA measurements of the component identities and behaviors. The prediction equations show that individuals who engage in self directed activity generally seem less good, powerful, and active than before the event, but outcomes also are influenced by the character of the actor, the behavior, and interactions between the two. Data from both female and male respondents generated similar equations, but impressions form differently for female as opposed to male actors.

The numerically simplest structures which can still be designated as social interactions occur between two elements. Nevertheless, there is an externally even simpler phenomenon that belongs among sociological categories, however paradoxical and in fact contradictory this may seem--namely, the isolated individual. (Simmel, 1950: 118)

Social psychological literature gives the impression that people spend most of their time interacting with others. Wondering if other kinds of activity are being overlooked, we coded a seven year old boy's behaviors for three hours and twenty-four minutes after waking, using the archive in Barker and Wright's (1951) classic observational study, One Boy's Day: A Specimen Record of Behavior. Only about one-third of the boy's activities were social interactions. The rest were monadic events in the sense that the boy directed his action at no one, other than himself. Approximately two-thirds of the monadic events were devoted to spatially relocating the self or adjusting body orientation (e.g., the boy went into the kitchen, went to the table; leaned back in his chair, looked out the window). About half of the rest of the monadic events could be construed as actions on non-human objects (e.g., petted his dog, hung up his coat). The remainder were self directed actions like dressing self, grooming self, feeding self, talking to self, entertaining self.

Of course, such monadic activity is not really asocial, being generated within a socially inherited culture. As Simmel (1950: 119) said regarding extreme forms of monadism: "Isolation thus is a relation which is lodged within an individual but which exists between him and a certain group or group life in general." Monadic activity has several other social facets as well.

Self-directed activity amounts to "interactional" events in which the self is both actor and object of action. This kind of action constantly is available as an alternative during social interaction, and social interaction implicitly involves forgoing such activity. An example of what could occur is provided by displacement behavior among chickens (Eibl-Eibesfeldt, 1970: 178)--a cock faced off by another cock suddenly may begin pecking at the ground, determinedly feeding himself as if no confrontation occurred. In fact, our interest in self-directed action arose from questions about the limits of commitment to social interaction. Under what conditions might an individual interrupt social interaction in favor of self-directed action? More pointedly, why do people so rarely abandon social interaction in favor of self directed action?

Some self-directed activities actually are socially demanded because others want to be free of certain responsibilities. The boy observed by Barker and Wright had to dress, groom, and feed himself because no one wanted to do these things for him (parents dress, groom and feed babies interactionally but not seven year olds). Similarly, health-care practitioners may transform interactive medical activities into self care so that the medical practitioners can attend to other duties, and a literature has developed regarding issues in self-care, especially self medication among the elderly (e.g., Coons and McGhan, 1988; Fincham, 1986; Segall, 1990; Segall and Goldstein, 1989). Framed this way, self-directed actions positively connotate taking responsibility for oneself, and care givers would like to know how to facilitate such behavior.

On the other hand, the relative ease with which individuals can engage in deviant self actions is a prominent social control concern. Clinical psychologists try to classify and help self mutilators who attack their own eyes, genitals, and body surfaces (e.g., see Conn & Lion, 1983; Favazza, 1989; Feldman, 1988; Silva et al, 1989; Zerbe, 1988); clinical psychologists recently have developed treatments for purgers (e.g., Favazza, DeRosear, and Conterio, 1989; Gotestam and Agras, 1989); and a voluminous social science literature deals with suicide (see Stack, 1982; Cockerham, 1989). Masturbation and self-administration of recreational drugs waver at the edge of social acceptability, sometimes emerging as social problems that demand attention from authorities. In these instances the key question that is raised about monadic activity is how to prevent it or extinguish it.

Focusing on self-directed actions from the perspective of affect control theory (Heise, 1979; Smith-Lovin and Heise, 1988), this paper addresses the question: What kinds of impressions are made when people engage in behaviors toward the self? It seemed to us that impressions formed from self-directed actions are different from impressions created by actions directed toward others. For example, a lawyer praising himself does not seem as admirable as a lawyer praising another lawyer, even though a lawyer is praising a lawyer in both cases. The results of our research do indicate that impression-formation processes based on self-directed actions are different than those based on other directed actions. Moreover, we believe that the kinds of impressions that are produced by self-directed actions are relevant to the facilitation and control of monadic activity.

Modeling Impression Formation Processes

Three universal dimensions of human response to stimuli involve judgments of goodness versus badness--Evaluation; powerfulness versus powerlessness--Potency; and liveliness versus quietness--Activity (Osgood, May, and Miron, 1975). Obtaining Evaluation-Potency-Activity (EPA) measurements of identities and behaviors in isolation and when combined in descriptions of events provides an approach to understanding how impressions are formed as a result of events. Specifically, regressing the EPA ratings of an identity or behavior obtained within an event depiction on out of-context ratings yields a model for predicting the impressions created by events (Smith-Lovin, 1987).

Previous research on transitive social action where actor and object are different has demonstrated that EPA impressions created by events can be both simple and interactive (Heise, 1979; Smith Lovin 1987). Thus a predictive model for self-directed action also should allow for both linear and non-linear effects, as indicated in the following schematic formulation. Image



Our initial sample of 409 respondents drawn from a psychology subject pool at a large midwestern state university consisted of 208 females and 201 males. "Good" data, however, were obtained from only 184 females and 181 males. The other respondents (eleven percent) were deleted from analyses because more than one-third of their responses were "don't know" or else more than one-third of their ratings of stimuli were neutral or essentially neutral. These indices indicate that the deleted respondents did not cooperate fully in the task, probably because of the length and the complex nature of the stimulus set (described below).

Respondents came to a computer laboratory and sat in front of a computer of their choice, restricted only by the messages "female version" or "male version" visible on the computer screen. Such self-seating assured random assignment of subjects across various stimuli sets used in the study.


EPA ratings can be positive or negative. For example, favorable evaluations are coded positive (+) and unfavorable evaluations are coded negative (-); potency is positive and impotency is negative; activity is positive and inactivity is negative. Configurations of plus and minus signs therefore may be used to factorialize the design of stimuli.

Our design combined every EPA configuration of identity with every EPA configuration of behavior to produce 64 types of self directed action. For example, one of the events referred to "the hero thrilling himself", which represented a (+++) actor engaging in a (+++) behavior; another of the stimulus events referred to "the tot laughing at herself", representing a (+-+) actor and a ( -+) behavior. The design was replicated four times to improve statistical power, and gender of actor was incorporated into the design by wording sentences in two of the replications with the female pronoun "herself" and wording sentences in the other two replications with the male pronoun "himself".

We used previously assembled dictionaries of EPA profiles (Heise and Lewis, 1988) to select social identities and social behaviors with desired profiles. Since not every social behavior can be directed toward the self, the social behaviors first were screened to isolate the subset of actions which could be self directed.

The identities and self-directed behaviors used to compose event descriptions in the study are listed in Table 1. Five of the eight EPA configurations yielded few self-directed behaviors. For example, for the configuration +-+, only the behaviors idolize and idealize could be identified. Where this was the case, the available verbs were used repeatedly to generate event depictions fitting the study design. Further, some of the behaviors barely fit the configurations, or fit only in a relative sense. For example, the EPA configuration of idealize is 0.45, 0.08, 0.57, yet the verb had to be used to represent approximately the +-+ configuration since there are no good, lively self-directed behaviors that connote actually impotent action.

Table 1
athlete, boyfriend, brother, champion, fireman, hero, heroine, husband, magician, prodigy, spouse, superstar, sweetheart, teammate, whizkid, wife, winneralert, amaze, amuse, dazzle, defend, delight, entertain, excite, love, rescue, thrill
advisor, church deacon, father, grandfather, grandparent, grown-up, judge, minister, parent, physician, professor, psychiatrist, psychoanalystcalm, caution, comfort, console, explain, forgive, listen to, pray for, soothe
baby, child, freshman, girl scout, grandchild, infant, lass, pupil, schoolgirl, tot, undergraduate, youngsteridealize, idolize
granny, housemaid, janitor, library assistant, old maid, oldtimer, patient, servant, tailor, widowgrieve for, observe, revere, study
assailant, assassin, bouncer, brute, bully, cutthroat, desperado, gangster, gunman, mobster, mugger, pusher, roughneck, ruffian, slavedriver, toughanger, corrupt, destroy, kill, radicalize, shoot, terrorize
auditor, disciplinarian, scrooge, tightwad, warden, witchhypnotize, silence
acidhead, braggart, chatterbox, crybaby, delinquent, drug addict, fink, jerk, junkie, liar, shoplifter, slut, stripper, whorelaugh at, lie to, ridicule
beggar, coward, deadbeat, deadhead, do-nothing, drudge, dullard, hobo, loafer, old fogy, weakling, winofear, pamper

Five sets of 145 stimuli each were prepared. Each set contained 51 event descriptions (one set contained 52), and half of the stimulus sentences presented female actors, half male actors, as cued by female and male pronouns. Each sentence was presented twice for EPA rating: once to assess the impression of the actor and once to assess the impression of the behavior. These two presentations were in different stimuli sets so that a given respondent made only one the ratings.

Also contained in each stimulus set were eight out-of-context behaviors, and 32 out-of-context identities, and the organization of the stimulus sets was such that an out-of-context identity or behavior was rated by different subjects than were the events employing that word. Through the use of the appropriate pronoun, half of the identities were represented as female actors and half as male actors (e.g. She is a physician, He is a professor). A few additional stimuli (behaviors presented in the passive voice) were included in each set to explore issues unrelated to self-directed action.

All five stimuli sets together presented a total of 40 out of-context behaviors, 160 out-of-context identities, and 256 events involving self-directed action. Each stimulus was rated by approximately 30 females and 30 males.


The computer program ATTITUDE (Heise and Lewis, 1988) was used to collect EPA ratings of the stimuli. Respondents rate stimuli by using cursor keys to move a pointer back and forth along bipolar graphic rating scales defined by the following clusters of adjectives at either end:

Good, Nice versus Bad, Awful
Big, Powerful versus Little, Powerless
Lively, Fast, Young versus Quiet, Slow, Old
The adjectives on the left represent positive values, while those on the right define negative values. The midpoint on the graphic rating scale is coded zero. Intervals between points labeled as "neutral," "slightly," "quite," and "extremely" are coded as differences of 1.0. The differences between "extremely" and "infinitely" (the scale endpoints) are coded 1.33; all of these intervals correspond to spatial distances on the scale. Ratings between labeled points are converted into appropriate fractional values, so ratings range in fine increments from -4.33 to 0 to +4.33. The program presents scales in randomized order.

Stimulus sentences were presented in a box at the top of the screen in randomized order. Event descriptions were preceded with the phrase "you see" to enhance respondents' visual imagery--e.g. "You see the champion delighting herself. She seems" or "You see the gangster forgiving himself. The act seems". Out-of-context identities were presented in the frame "She/he is a ___. She/he is"; e.g. "She is a wife. She is" Out-of-context behaviors were presented in the frame "You see someone ___ing someone else. The act seems"; e.g. "You see someone silencing someone else. The act seems". Transitive phrasing for out-of-context behaviors was used under the assumption that the meanings of the behaviors are interpersonal.

After brief oral instructions, respondents worked through a tutorial offered by ATTITUDE, and then went on to rate the actual stimuli.


Our goal was to predict mean EPA ratings of actors and behaviors in a self-directed activity from mean EPA ratings of the actor's identity and behavior when rated out of the context of the event. Thus we reduced the individual data to averages before preceding with other analyses. By collating in-context mean ratings with out-of-context mean ratings while matching the sex of the actor in both the event and the out-of-context stimulus, we obtained our sample of 512 records for initial analyses, each record summarizing all of the obtained data about one event involving a self-directed action by either a female or a male actor, as rated by females or by males.

Regression analyses estimated a dependent variable from a constant, first order terms, and all second order interactions in which an attribute of actor combines with an attribute of behavior. Also included was a sex-of-actor variable (corresponding to the pronoun used in event descriptions) and its products with the other variables. The following stepwise procedure was used to determine which terms to retain in prediction equations.
1.The initial model was put through Systat's STEP procedure (Wilkinson, 1988: 480-1), forcing first-order and consistency terms (i.e., the products of actor evaluation with behavior evaluation, actor potency with behavior potency, and actor activity with behavior activity) and allowing other terms to enter or be removed with the default alpha value of 0.15.
2.Using the model defined in step 1, Systat's STEP procedure was run again, forcing no variables and using an enter-remove alpha value of 0.01.
3.Using the model defined in step 2, Systat was used to estimate the coefficients reported here.


Sex Effects

Analyses of covariance that preceded the regression analyses revealed no appreciable differences in the rating of stimuli by male or female respondents. That is, the psychological processing of self-directed action appears to be similar for females and males. Accordingly, data were pooled, disregarding raters' gender, for subsequent analyses, so mean ratings were based on 60 respondents rather than 30.

The sex of the actor in an event involving self-directed action does make a difference, however, inasmuch as several interaction terms involving sex were retained throughout the stepwise selection procedure. The interactions with sex have the following meanings.

We will present separate results for female and male actors since there are a number of differences in how impressions are formed when females versus males direct actions at themselves. The regression coefficients obtained for each gender are given in Tables 2 and 3. The terms in each regression equation are indicated at the left margin: A stands for actor, B for behavior; the letters e, p, and a in lower case identify which measure is involved. For example, Ae stands for mean evaluation of the actor's identity when presented out of context; Bp stands for the mean potency rating of the behavior when considered out of context; and AeBp stands for the product of the Ae and Bp mean ratings

Impressions of Actor

The regression equations for predicting evaluation, potency, and activity of actors directing behaviors toward themselves all have rather large negative constants, as shown in Table 2. This suggests that people engaging in self-directed action generally are perceived as less good, less powerful, and less active than they were before the event, and the decrements are considerably larger than when a person engages in interpersonal actions (Smith Lovin, 1987). The biggest effect is a reduction in potency- directing actions toward the self makes the actor appear weak. Since the actor in a self-directed event is simultaneously the object in the event, the lowering of potency could be viewed as corresponding to the finding in previous impression formation research that a person loses potency merely by being the object of action (Smith-Lovin, 1987).

Table 2
TermImpression of Female ActorImpression of Male Actor

Actor Evaluation. Evaluations of the actor who has engaged in a self-directed behavior depend highly upon pre-event evaluations of the actor and the behavior. Aside from other effects, nice people doing nice things to themselves tend to stay favorably evaluated.

The AeBe interaction replicates a consistent finding in impression formation research that people tend to be evaluated more positively when they behave in ways that match their status (Gollob, 1968; Smith-Lovin, 1987). In the case of self-directed action, this interaction simultaneously replicates the finding that actors are evaluated more positively when they match the goodness of their behavior to the goodness of the object person. The interaction between actor evaluation and behavior potency (AeBp) suggests that actors are perceived relatively unfavorably when they have high status and direct powerful behaviors toward themselves. Actor potency interacts with behavior activity (ApBa) in predicting evaluations of the actor: actors are viewed more negatively when they are powerful and engage in lively self directed behavior or if they are powerless and direct quiet, slow activity at themselves. Finally, there is an interaction between actor activity and behavior evaluation (AaBe): actors gain in evaluation if they are active and do bad things to themselves, or if they are quiet and do good things to themselves.

Actor Potency. Impressions of an actor's potency after self directed behavior depend on the actor's prior potency. Moreover, directing nice, active behaviors at the self increases impressions of powerfulness. Potency of the behavior only influences impressions of actor potency for female actors: when females direct strong, powerful activity toward themselves, they are perceived as less potent. Meanwhile, for female actors who are evaluated positively (AeBe), doing good things to the self tends to increase impressions of powerfulness.

Actor Activity. Weak but lively people engaging in nice, weak, lively self-directed actions impress us as lively. Beyond these main effects, powerful actors who engage in nice behaviors toward themselves appear still more lively (ApBe). Nice females who do good things to themselves are perceived as especially lively (AeBe).

Impressions of Self-Directed Behavior

Similar to equations predicting impressions of actors, the regression equations that predict impressions of self-directed behaviors all have negative constants--see Table 3. Behaviors tend to lose favorability, potency, and activity when they are directed toward the self.

Table 3
TermImpression of Female's BehaviorImpression of Male's Behavior
Bp 0.17-0.150.16-0.15

Behavior Evaluation. Self-directed behaviors end up being viewed positively when they are good to begin with and are done by a nice person. Self-directed behaviors also seem nicer when they are consistent with the evaluation of the actor (AeBe), and this effect is stronger when the actor is female. The interaction between actor evaluation and behavior potency (AeBp) suggests that when nice people do powerful things to themselves, the behavior seems less good.

Behavior Potency. Behaviors directed toward the self seem more powerful if they start off being good, powerful, and lively, and if they are enacted by a nice, powerful person. When the actor is a nice female doing something nice to herself, or a deviant female doing something bad to herself, the behavior increases in potency (AeBe).

Behavior Activity. Self-directed behaviors impress us as more lively when they are committed by a weak, lively actor and when the behavior was good, weak, and lively to begin with. When the behavior evaluation parallels the status of the actor (AeBe), the behavior seems more active.

Explained Variance

The regression equations represented in and (including the interaction terms involving sex of actor) explain substantial proportions of variance in the dependent variables: 0.66 for actor evaluation, 0.67 for actor potency, 0.84 for actor activity; and 0.55 for behavior evaluation, 0.55 for behavior potency, 0.82 for behavior activity. Thus the equations do offer a model of how people typically respond to events involving self directed actions.

Nevertheless, our equations explain only about half of the variance in behavior evaluation and potency outcomes, which is rather low when compared to earlier impression formation research which has traditionally explained over seventy percent of the variance in outcomes. Variances in this study are not much less than in prior studies, so the reduction in R-squares is not a result of there being less meaningful variance to explain. Perhaps cognitive processing is more intricate for self-directed actions, creating outcomes that are beyond straightforward prediction from pre-event EPA measurements. For example, behavior potency is rated much higher than equations predict for a junkie saving self and for a magician hypnotizing self: in both cases the behavior seems exceptionally powerful in achieving so great an effect with such an unpromising subject.


Sex of Actor

While there were no significant differences found for sex of rater, there were a number of differences found for sex of actor, cued by male and female pronouns. This combination of findings has the rather striking implication that somewhat different schemes are used to process self-actions of males and females--schemes that are applied similarly by both males and females. Specifically, people view evaluative consistency between identity and behavior as more important in assessing females, and, additionally, females actually lose potency by engaging in potent actions on the self.

Among other things, this suggests that feminist emphasis on pronouns may be well placed. The 1970's was a decade characterized by debate over sexism inherent in the English language (Martyna, 1980), and feminist scholars advocated the implementation of nonsexist language. The research findings reported here support the feminist contention that the masculine pronoun should not be considered generically representative of both sexes: male and female pronouns invoke different psychological processing.

This is the first impression-formation study to treat sex of actor as a variable in estimating structural equations. Our results encourage revised studies of impression formation in transitive events with controls for sex of actor and sex of object person. We suspect that the occasional small sex effects found among respondents assessing transitive events (Smith-Lovin, 1987) actually may be sex-of-actor or sex-of-object effects arising because raters implicitly identify characters with their own gender.


Our prediction equations show that when behaviors are self directed, they decrease in favorability, in potency, and in liveliness. These changes in meaning of behavior have interesting implications with respect to moral judgments, allowing that observers sometimes classify an actor and object person as so similar as to warrant interpreting the actor's behavior as self-directed. Homosexuals might be perceived as virtually having sex or making love with themselves. The same might be said for perpetrators of incest. In a similar vein, nepotists are virtually hiring themselves. Condemnation of such activities might be due in part to changes in the meanings of behaviors that are self-directed, thereby explaining how condemnation can arise from actions that are approved when directed away from the self at nonsimilar others.

We employed the equations in with EPA profiles from Heise and Lewis (1988) to compute some illustrations. Were a woman to make love to self (i.e., another woman) the evaluation of the behavior declines from 2.58 to 1.24--a difference of -1.34 units, and the potency of the behavior declines by -1.22 units; were a man to make love to self (i.e., another man) the evaluation of the behavior drops by -1.92 units and the potency drops by -1.06 units. For a woman hiring self (i.e., a relative) evaluation of the behavior drops -0.68 units and the potency of the behavior drops -1.03 units; for a man doing the same thing, evaluation of the behavior drops -1.03 units and potency drops -1.40 units. Thus viewing the behaviors as self-directed makes the behaviors seem corrupted.

If this interpretation is correct, then those who engage in such actions (e.g., homosexuals) probably differentiate people in such a way that actors in the events are distinguished from object persons, whereas condemners (e.g. homophobes) are equating people such that actors and objects in the events seem the same.

Facilitation and Control

The impression formation equations obtained in this study have implications for facilitating and controlling self-directed action, assuming that people try to create events that maintain sentiments about identities and behaviors (Heise, 1979).

Most of the behaviors that care-takers want to facilitate as self-directed actions can be approximated by the generic EPA profile of 1.2, 0.9, -0.3 (this is the mean EPA profile, averaged across females and males, of dressing, feeding, and medicating). Employing mathematical procedures defined by Heise (1987) we can compute the EPA profile for the kind of person who could direct such an action at the self with maximum confirmation of sentiments: 1.7, -0.1, -0.3 for females; 1.5, -0.1, -0.4 for males. A woman could acquire approximately such an identity by seeing herself as a quiet woman or a dependent woman during the action; a man by seeing himself as relaxed man or a shy man; a child by seeing herself or himself as an old child or experienced child (see Heise and Thomas, 1989, for the equations used to compute these attributions). Presumably, care takers could encourage such self-perceptions in order to foster responsibility for self care (at least when working with an adult who views herself as a woman or himself as a man or with a youth who views her or himself as a child).

Similarly, controlling deviant self actions would be a matter of identity therapy, this time by inhibiting the identities that get confirmed by disapproved self actions. For example, someone who purposively hurts self presumably is confirming a negative identity (EPA profile -1.1, -0.7, 0.8 for females; EPA profile -0.9, -0.5, 1.7 for males) such as might emerge from thinking of self as a sinner, a divorcee, a dropout, etc. While it may not be possible to shed such identities once acquired, the salience of negative identities can be reduced by committing more time to action in positive identities (Stryker, 1980), and therapists can encourage patients to adopt an agenda of actualizing their positive selves in order to reduce salience of negative selves.

Observer Versus Actor

Since our stimuli emphasized observation ("You see A doing B to self"), the equations might predict self-directed actions only when the actor is conscious of being observed and concerned with the kind of impression that she or he is making. A follow-up study investigated whether a different dynamic determines private feelings about one's own self-directed action. We changed the event stimuli to "Imagine you're an IDENTITY ACTing yourself. You feel (ACTing feels)"; e.g., "Imagine you're a minister lying to yourself. Lying feels". We employed the same basic design as outlined in Procedures but without replications so that there are only 64 cases (events). Equations were re-estimated the same way as before except that no gender factor was included because of the small number of cases. With these procedures, the impression-formation equations for self-directed action were different than those represented in Tables 2 and 3. Image

These equations make impressions largely a function of the behavior being enacted, behavior activity is a major factor determining all impressions, and evaluative consistency no longer is a factor at all. Thus impression formation might operate differently when considering self as actor rather than someone else. Even if this is the case (and this small auxiliary study certainly requires replication), the commentary on facilitation and control still roughly applies. According to additional analyses, self care would be facilitated by encouraging people to see themselves as having quiet traits, and controlling deviant self-action would require suppressing bad, weak identities.

Acting on Self Versus Others

In principle, one might act on the self to maintain a sense of self esteem and self efficacy, but this study shows that self directed action tends to reduce the impression of an actor's goodness, potency, and activity, and therefore it rarely is an optimal way to confirm the good, potent, lively identities that people normally want to maintain. People normally can confirm their sense of self much better through social interaction than through self-directed actions, using social relationships as a resource for building meaningful experiences that actualize positive selves (Heise and Wiggins, 1987). Self-directed actions might be preferable only in confirming relatively depressed qualities like being fatigued, sleepy, or melancholy.

The preference for acting on another rather than on self should be especially pronounced when already engaged in a social interaction. Then self-directed action not only fails to confirm personal concepts of self, it also may produce a pejorative impression in the eyes of the other who is present, and thereby lead to a loss of face in the relationship (Goffman, 1967). Proving that one deserves to be seen as good and potent provides a compelling goal that drives one into social interaction and keeps one at it.