Bertram F. MalleSelf-Other Asymmetries1
The Social Self and the Social Other:
Actor-Observer Asymmetries in Making Sense of Behavior
Bertram F. Malle
University of Oregon
In navigating the social world, people assume two fundamental perspectives: As observers they try to make sense of other people’s behavior; as actors they try to make sense of their own behavior. This fundamental duality is codified in the grammar of languages as the first-person and the third-person form and has generated numerous philosophical puzzles such as the nature of introspection and the possibility of knowing other minds. More recently, psychologists have begun to investigate the empirical parameters of the actor-observer duality (e.g., Nisbett, Caputo, Legant, & Marecek, 1973; Robins, Spranca, & Mendelsohn, 1996; Sande et al. 1988; Storms, 1973). Typically, the discussion focuses on a single asymmetry—that between behavior explanations by actors, who are said to be using more “situation causes,” and behavior explanations by observers, who are said to be using more “person causes.” But this is far too simple a picture. The studies reported in this paper show that social interactants face a variety of actor-observer asymmetries—in the events people attend to, the events they try to explain, and in the ways they explain those events.
Investigations of such actor-observer asymmetries are not just explorations into the curiosities of social cognition. The negotiation of self-other asymmetries is a basic challenge for social agents. Organisms who have complex social interactions must form representations about their interaction partners as well as about themselves and must integrate them in their mental model of the social situation they experience. Many researchers have argued that humans form such integrative representations with the help of a theory of mind, which provides a common conceptual framework for the processing of information about self and other (Barresi & Moore, 1996; Heider, 1958). There is debate over the extent to which information access, and perhaps information processing, differs between self and other (e.g., Gopnik, 1993; Goldman, 1993). What is not debated is that the conceptual categories in which self- and other-information is interpreted are essentially the same for actors and observers. In light of this assumption the present paper examines empirically how actors and observer differ in their use of these shared concepts when making sense of behavior.
At a first level of analysis, I examine the very events people attend to in social interactions, because making sense of a behavioral event presupposes first attention to its occurrence. I offer a conceptual scheme that classifies these events using two folk distinctions—between intentional and unintentional events and between observable and unobservable events. At a second level of analysis I ask which behavioral events actors and observers wonder about and try to explain. Finally, a third level of analysis concerns the way people explain behavior. Here I rely on a model of behavior explanations that significantly expands the classic attribution dichotomy of person vs. situation causes by considering the folk-conceptual distinctions that people themselves use when explaining behavior.
Besides documenting multiple actor-observer asymmetries in attention and explanation, this paper also suggests that the asymmetries can be best understood as reflections of two fundamental determinants of self- and other-cognition in social settings: epistemic access and motivational relevance. In brief, social interactants have access to different information about the self than about the other person, and they find different information relevant for regulating the self than for regulating the other person.
Actor-Observer Asymmetries in Attention to Behavioral Events
Social interactions can be taxing on people’s attention. Interactants must process what others are saying and doing, infer what they are feeling and thinking, and predict impending actions. All the while, they must plan their own utterances, monitor their actions, and confront the vast inner landscape of their thoughts, feelings, and bodily states. How people regulate attention to this complex pattern of behavioral events in self and other has been largely unexplored. In a recent series of studies we examined one aspect of this regulation during social interaction: how people, as actors and observers, distribute attention to various behavioral events and build up memory representations about these events (Malle & Pearce, in press). (With the term behavioral event I refer, broadly, to behaviors as well as mental states, such as greeting, crying, thinking, feeling, but not to traits or stable attitudes.
Some indications in the literature suggested that there would be actor-observer asymmetries in the kinds of behavioral events to which they attend. Jones and Nisbett (1972) argued that observers have access to the other person’s behavior but little access to the other’s internal states; actors, by contrast, have difficulties monitoring their own behavior but no difficulties accessing their own internal states. Indirect evidence for this proposition comes from studies by Sheldon and Johnson (1993), who asked people to estimate which of several objects they usually think about when speaking with another person. The two most frequently chosen objects of attention in conversation were people’s own thoughts and feelings and the other person’s appearance. Similarly, people’s long-term memory representations of self contain more private aspects (e.g., thoughts and feelings) than public aspects (e.g., actions and appearance), whereas representations of others contain more public aspects than private aspects (Andersen, Glassman, & Gold, 1998; McGuire & McGuire, 1986; Prentice, 1990). People find it especially difficult to track accurately their own observable behaviors (Gosling, John, Craik, Robins, 1998). For example, while actors are acutely aware of their own emotional states, they cannot easily observe their own facial expressions, leading them to overestimate their face’s expressiveness and their interaction partner’s ability to infer emotional states from those expressions (Barr & Kleck, 1995; Gilovich, Savitzky, & Medvec, 1998). Conversely, observers find it difficult to reliably infer others’ internal states, as seen for example in their limited empathic accuracy (Ickes, 1993).
None of these findings assessed attention during social interactions, so we set out to do so systematically and using a solid theoretical foundation. We introduced two pieces of theory: a framework that outlines the different types of behavioral events to which people direct their attention and two main factors that guide their attention allocation. The framework is a four-fold classification of occurrent behavioral events based on two central folk distinctions—intentionality and observability—that have been referred to in prior literature (Andersen & Ross, 1984; Buss, 1978; Funder & Dobroth, 1987; Gilovich & Regan, 1986; Heider, 1958; John & Robins, 1993). Behavioral events can thus be either intentional or unintentional, and they can be publicly observable or unobservable. By crossing these distinctions, four event types result (see Figure 1), which can be labeled as follows: (1) actions (observable and intentional; e.g., asking for a favor, greeting), (2) mere behaviors (observable and unintentional; e.g., shivering, crying), (3) intentional thoughts (unobservable and intentional; e.g., searching for things to say, imagining Bali), and (4) experiences (unobservable and unintentional; e.g., being nervous, feeling angry).
IntentionalUnintentional
Observable / actions / mere behaviorsUnobservable / intentional thoughts / experiences
Figure 1. Postulated Folk-Classification of Behavioral Events
As our second piece of theory, we identified two factors that are known to govern attention allocation in general (e.g., Fiske & Taylor, 1991; Posner, 1980) and are important to social interaction: epistemic access and motivational relevance. First, to turn one’s attention to a particular behavioral event one needs to have access to it—i.e., become in some way aware of it taking place (through introspection, perception, or at least inference). Second, attention to an event increases if the perceiver considers it relevant (i.e., informative, helpful) for processing or coordinating the current interaction (e.g., Cantril, 1947; Jones & Thibaut, 1958; Wyer, Srull, Gordon, & Hartwick, 1982). Using these two factors governing attention allocation and the above classification of behavioral events, we formulated two main hypotheses regarding the behavioral events actors and observers attend to during social interaction.
First, epistemic access for actors is greater to their own unobservable events than to their own observable events, because actors are constantly presented with their stream of consciousness but cannot easily monitor their own facial expressions, gestures, or posture; Bull, 1987; DePaulo, 1992; Gilovich et al., 1998). For observers, obviously, access is greater to other people’s observable events than to their unobseverable (mental) events. We therefore hypothesized that, in interaction, people attend as observers to more observable events than as actors, whereas they attend as actors to more unobservable events than as observers (“observability gap”).
Second, for observers the perceived relevance of intentional events is greater than that of unintentional events, because intentional events define the main business of an encounter (Goffman, 1974), because they are directed at the other and thereby demand a response, and because they have powerful effects on the other’s emotions and moral evaluations (Shaver, 1985). By contrast, for actors the perceived relevance of unintentional events is greater than that of unintentional events, because unintentional events were not controlled and therefore must be monitored and understood, whereas the execution of intentional events frequently relies on automatic programs (Norman & Shallice, 1986). We therefore hypothesized that, in interaction, people attend as observers to more intentional events than as actors, and they attend as actors to more unintentional events than as observers (“intentionality gap”).
To test these hypotheses, we developed an experimental paradigm in which pairs of participants had a conversation and, immediately afterward, were asked to report in writing everything “that was going on” with their partner (on one page) and with themselves (on another page), in counterbalanced order. The reports were then coded for references to behavioral events (verb phrases that referred to actions, mere behaviors, intentional thoughts, or experiences) and classified according to their intentionality and observability, using a coding scheme in the public domain ( In Study 1, we found strong support for both actor-observer gaps. Of approximately 8 events reported per page, actors reported 2.2 more unobservable events than did observers, and observers reported 2.2 more observable events than did actors, F(1,57) = 79.4, p < .001, 2 = 58%. (These means represent the actual interaction effect, computed after removing main effects; see Rosnow & Rosenthal, 1989.) Second, actors reported 0.8 more unintentional events than did observers, and observers reported 0.8 more intentional events than did actors, F(1,57) = 13.6, p < .001, 2 = 19%.
We interpreted these results as supporting the hypothesized asymmetries in attention during interaction, even though strictly speaking the measure assesses people’s mental models of the interaction immediately afterward (which is the only such measure currently used in the literature; e.g., Frable, Blackstone, & Scherbaum, 1990; Smart & Wegner, 1999). But we secured our interpretation against several alternative explanations. First, varying instructions (e.g., asking participants to report about their own and their partner’s “behavior” or “experiences”) had little biasing effect on people’s overall reporting rates and did not interact with the actor-observer gaps, ensuring that participants were not merely constructing events on the spot. Second, we also examined “intrusive events”—behavioral events that were reported about the other person on the actor page and about the self on the observer page. Because intrusive events were obviously not offered in compliance with instructions, they should be free of demand characteristics and of strategic reporting. Among these intrusive events, too, both asymmetries replicated.
In two subsequent studies we explored factors that might close the two actor-observer gaps in attention. First, we examined intimacy and found that among intimates the observability gap was cut in half and the intentionality gap disappeared. Second, making the conversation highly personal among strangers had no impact on either gap. Third, empathy instructions slightly reduced the observability gap but left the intentionality gap intact. The documented actor-observer differences in attention to events (especially the observability gap) thus vary somewhat but appear to be fairly resistant to change, although more research is necessary to determine just how resistant it is.
Actor-Observer Asymmetries in Which Behavioral Events People Explain
So far we have seen that actors and observers differ in the behavioral events they attend to during social interactions. A second question of interest is which events actors and observers wonder about and try to explain. Attention is broader and more inclusive than wondering and explanations. Social interactants likely attend to many events whose meaning is so clear that no need for explanation arises. But for a subset of events interactants will wonder why the event occurred, what it means, and so they will try to explain it.
Given that there are actor-observer asymmetries for both intentionality and observability in the events people attend to, one would expect parallel asymmetries in the events people wonder about and try to explain. Moreover, these asymmetries should be derivable from principles similar to the ones that applied to the domain of attention. That is what we set out to do in another series of studies (Malle & Knobe, 1997b). Specifically, we posited that for an event to elicit a wondering why (and, under most circumstances, an explanation), three conditions must be met: there must be access (people must be aware of the event to wonder about it), nonunderstanding (people must not already have an explanation for the event), and relevance (people must find it useful and important to generate an explanation for the event). From these three conditions we derived asymmetries parallel to the ones for attention as follows (see Malle & Knobe, 1997b, pp. 289-290):
Actors have less access to their own observable events than to their own unobservable events (e.g., sensations, thoughts, feelings). Observers, on the other hand, have less access to other people’s unobservable events than to their observable events. People cannot wonder about events they don’t have access to, so we predicted that actors would tend to wonder more often about unobservable than observable events, while observers tend to wonder more about observable than unobservable events. In addition, actors are rarely in a state of non-understanding with respect to their intentional behaviors because they typically know, or at least believe they know, why they performed those behaviors; thus, they are unlikely to wonder about intentional behaviors. Observers, on the other hand, may find both intentional and unintentional equally difficult to understand, but intentional behaviors will often be more relevant than unintentional behaviors, because intentional behaviors are socially more consequential and because they are highly diagnostic of a person’s desires, beliefs, abilities, and character (Jones & Davis, 1965; Malle & Knobe, 1997a). Therefore, we predicted that actors would tend to wonder more often about unintentional than intentional events, while observers tend to wonder more about intentional than unintentional events.
We confirmed both of these predictions in two studies (using a coding scheme for events explained: In one study participants kept thought protocols of spontaneous wonderings during the day and described them in more detail in the evening; in the other study wonderings were extracted from three 20th-century novels. In these two studies, actors wondered about more unobservable events (67%) than observable events (33%), whereas observers wondered about more observable events (74%) than unobservable events (26%). And actors also wondered about more unintentional events (63%) than intentional events (27%), whereas observers wondered about more intentional events (67%) than observable events (74%).
When examining actual explanations, we drew a distinction between explanations that are directed to oneself (in private thought) and explanations that are directed to a partner (in communication). Explanations to oneself, collected in two studies from memory protocols and diaries, showed the same actor-observer asymmetries as wonderings because these explanations answer one’s own wonderings. Explanations to others in communication, however, eliminated the asymmetries in that both actors and observers explained in recalled and actual conversations more observable and intentional behavioral events. That is because in communicative explanations actors answer their partners’ wonderings, which come from the observer perspective, and so actors explain behavioral events about which observers wonder, namely, intentional and observable ones. We might say that the social self adapts to the social other by using a tool of social cognition (i.e., behavior explanations that normally help find meaning for oneself) for pragmatic, interactive purposes.
But this adaptation appears limited. By shifting away from explaining unobservable and unintentional events, actors answer their partners’ wonderings, but they fail to explain the very information that observers normally have little access to, namely, unobservable (mental) events. In a sense, actors bolster the other-mind problem. Or do they?