The Significance of Intentionality

Full text

In B. F. Malle, L. J. Moses, & D. A. Baldwin (Eds.), Intentions and Intentionality:
Foundations of Social Cognition (pp. 1-24). Cambridge, MA: MIT Press, 2001.
Introduction: The Significance of
Intentionality
Bertram F. Malle, Louis. J. Moses, and Dare A. Baldwin
Considerations of intentions and intentionality permeate human social life.
Picture a first date, in which the partners try to find out their own and the
other’s desires, or a business negotiation, in which proclaimed intentions
must be separated from hidden ones. Scan the human affairs columns for
stories about conflicting desires and surmised intentions, or for legal cases
about intent and insanity. Or simply read literature to see that human social
interaction fundamentally requires that people infer and avow intentions
as well as probe and affirm the intentionality of actions. If one took a
Kantian approach to social cognition, searching for the fundamental con-
cepts without which such cognition is impossible, intentionality would be
one of those concepts, on par with space, time, and causality in the realm
of non-social cognition.
Intentionality is a foundation for social cognition in several ways. For
one, the concept of intentionality unlocks a central part of the folk ontol-
ogy of mind, because intentionality’s constituent components represent
basic mental categories, such as belief, desire, and awareness. Moreover,
the concept of intentionality brings order to the perception of behavior in
that it allows the perceiver to detect structure—intentions and actions—in
humans’ complex stream of movement. Further, the intentionality concept
supports coordinated social interaction by helping people explain their own
and others’ behavior in terms of its underlying mental causes. And inten-
tionality plays a normative role in the social evaluation of behavior through
its impact on assessments of responsibility and blame.
Intentionality is thus a tool with manifold functions, ranging from the
conceptual to the interpersonal and even to the societal, and it is a tool with
various domains of application, ranging from perception to explanation to

2 Introduction
interaction. Contemporary research on the role of intention and intention-
ality in human social cognition has touched on all these functions and
domains, but findings from this research are often discussed in isolation
from one another. For example, much philosophical work has been devoted
to analyzing the conceptual components of intentional action, but these
analyses have rarely guided psychological research on the social perception
of intentional behavior. Within psychology, the role of intentionality in
explanations and in assignments of responsibility has been studied in devel-
opmental psychology within the paradigm of “theory of mind” and in
social psychology within the paradigm of “attribution theory,” but little
communication has occurred between these paradigms. The central aim of
this volume is to bring together the various disciplines, approaches, and tra-
ditions that have examined intentions and intentionality and to integrate
current knowledge of this central facet of human cognition.
With this integrative aim in mind, we have organized this introductory
chapter and the book by the major research questions being asked about
intentionality across disciplines. A first set of questions concern how the
concept of intentionality is defined within the folk theory of mind, what
components make up this concept, how the components are related, and
how they are acquired between infancy and adulthood. Earlier attempts to
answer these questions can be found both in philosophy (e.g., Brand 1984;
Mele and Moser 1994; Schueler 1995; Searle 1983) and in psychology (e.g.,
Astington and Gopnik 1991; Malle and Knobe 1997a; Maselli and
Altrocchi 1969; Moses 1993). A second set of questions concern how peo-
ple perceive human action and detect its underlying intentions and motives.
These issues, which are a major focus of current research in developmental
psychology (e.g., Barresi and Moore 1996; Baldwin and Baird 1999;
Zelazo, Astington, and Olson 1999), have attracted considerable attention
in philosophy (Bogdan 2000; Carruthers and Smith 1996; Davies and Stone
1995), as they did in early social psychology (Heider 1958). A third set of
questions concern the role of intentionality in people’s explanations of
behavior, and especially its role in distinguishing “reason explanations”
from “cause explanations.” Here, too, pertinent work spans the disciplines
of development (Bartsch and Wellman 1989; Kalish 1998), social psychol-
ogy (Buss 1978; Malle 1999; White 1991), and philosophy (Audi 1993;
Davidson 1963; Lennon 1990). A final set of questions concern the role of

Introduction 3
judgments of intentionality in the evaluation of human action in terms of
responsibility and blame. These issues have been explored in legal, philo-
sophical, and psychological writings (e.g., Duff 1990; Hart 1968; Shaver
1985; Wallace 1994; Williams 1993).
Conceptual Elements of Intentionality
If the study of intentionality is to be a successful cross-disciplinary enter-
prise, it will require conceptual clarification to establish a common con-
ceptual language and a shared map with which to scout the territory. This
is not merely a methodological desideratum. Conceptual clarity represents
a necessary step toward a model of what intentionality consists of, what it
means to people, and how it functions in the social world.
Intentionality has two quite different meanings. Brentano (1874) intro-
duced it as a technical term that could be used to refer to the property of all
mental states as being directed toward something. Desires, for example, may
be directed toward attractive objects, and beliefs toward states of affairs
(Searle 1983; Lyons 1995). Second, intentionality is the property of actions
that makes ordinary people and scholars alike call them purposeful, meant,
or done intentionally. The focus of this volume is on this second sense—more
specific, on people’s conceptions of this sense of intentionality.
Another important distinction is that between intentionality and inten-
tion. The two terms are sometimes equated in psychological writing, even
though folk use and philosophical analysis mark them as distinct. Intention-
ality, as we have mentioned, is a quality of actions (those that are intentional
or done on purpose), whereas intention is an agent’s mental state that rep-
resents such actions. This type of mental state often precedes its correspon-
ding action or even occurs without it. One can therefore ascribe intention
to an agent without making a judgment of intentionality. The reverse is not
true, however: The judgment of an action’s intentionality typically implies
the ascription of an intention to the agent. Whether this implication holds
under all circumstances is still debated among philosophers (Adams 1986;
Bratman 1987; Harman 1976; Mele, this volume), but studies of folk use
have thus far supported it (Malle and Knobe 1997a).
Both intention and intentionality are complex concepts in that people
apply them only when a number of conditions are met. The ascription of

4 Introduction
an intention to A requires minimally that one grants the agent a desire for
some outcome O and a belief that A will likely lead to O (Malle and Knobe
1997a). But an intention cannot be reduced to this belief-desire pair nor are
intention ascriptions just shorthands for elaborate desire ascriptions. Three
chapters in part I explore what unique conditions underlie people’s ascrip-
tions of intention and how these conditions distinguish the folk concepts
of intention and desire.
In chapter 2, Bertram Malle and Joshua Knobe contend that the two
concepts are distinguished by three features: First, intentions are directed
at the intender’s own action whereas desires can be directed at anything.
Second, intentions are based on some amount of reasoning whereas desires
are typically the input to such reasoning. Third, intentions come with a
characteristic commitment to perform the intended action whereas desires
do not. Malle and Knobe provide conceptual arguments and empirical
data to support the validity of this tripartite model. They also speculate
about the psychological functions of the folk distinction between desire
and intention.
In chapter 3, Louis Moses argues that, although preschool children have
some appreciation of motivational aspects of intention, they are not espe-
cially cognizant of the fact that agents’ beliefs constrain their intentions.
Such belief constraints further distinguish intentions from desires. For
example, although agents cannot intend what they believe to be impossible
(Davis 1984; Grice 1971; Velleman 1989), there is nothing that prevents
them from desiring it. Moses concludes that children aged 3 and younger
may collapse desire and intention within a generic pro-attitude concept.
Janet Astington, in chapter 4, also traces developmental stages in mas-
tering the folk concept of intention and disentangling it from desire.
Paradoxically, intention can be thought of as either an early-arriving or a
late-arriving concept within children’s developing theory of mind: Infants
seem to be able to detect intentions by age 1 (see part II), but not until age
5 do children reliably master the distinction between intentions and desires.
The complex adult concept of intention must therefore be acquired gradu-
ally, with some aspects (e.g., object-directedness) acquired well before oth-
ers (e.g., self-referentiality). Genuine understanding of intention, Astington
hypothesizes, depends on the emergence of “metarepresentational” under-
standing—the child’s understanding that people’s beliefs and desires are

Introduction 5
mental representations of the world that mediate their actions in the world
(Bartsch and Wellman 1995a; Perner 1991). Astington’s hypothesis con-
verges here with Moses’s claim that an intention concept requires a belief
concept, as the latter also rests on metarepresentational understanding.
The importance of metarepresentational understanding after age 3 poses
a puzzle, however. Even 2-year-olds seem to have an understanding of men-
tal states like goals or desire (Bartsch and Wellman 1995a; Wellman and
Woolley 1990); thus, either children have already acquired a representa-
tional theory by age 3 or else their early understanding of desire, although
mentalistic, is not representational. The latter would be an unusual combi-
nation of features in light of the philosophical tradition of defining mental
states as representational states (Brentano 1874; Chisholm 1981; Searle
1983). In chapter 10, as one part of his larger argument, Alvin Goldman
challenges the evidence and logic in support of such a non-representational
desire concept.
The conditions for ascribing intentionality are even more complex than
those for ascribing intention, as Alfred Mele demonstrates in chapter 1.
Following Aristotle and Hume, philosophers have focused on desire and
belief as primary features of intentionality; however, the mere presence of
appropriate belief and desire states is not sufficient for an action to be inten-
tional. For one thing, intentions are considered an additional condition of
intentionality. A behavior may be performed in accordance with a belief-
desire pair, but it would not count as intentional unless it was brought about
by an intention grounded in that belief-desire pair (Brand 1984; Bratman
1987; Searle 1983; Thalberg 1984). Suppose Brenda fouls an opponent dur-
ing a basketball game. Suppose further that we are certain Brenda wants to
win the game and believes that fouling her opponent would help her win.
Still, we can’t be certain that she committed the foul intentionally unless we
know of Brenda’s specific intention, her decision to act on her desire and
belief. Moreover, even a behavior that was based on appropriate desires
and beliefs plus an intention may not count as intentional: The intention
must also cause the action via skill rather than luck. For instance, a golf
novice may (at most1) intend to hit a hole in one, but few would call the
accomplished feat intentional. Both conceptual analyses and empirical stud-
ies have converged on identifying skill as a further necessary condition for
intentionality (Malle and Knobe 1997a; Mele and Moser 1994). Finally,

6 Introduction
the folk concept of intentionality appears to require a particular kind of
awareness on the part of the agent, namely, the awareness of acting as
intended (informants call it “knowing what she is doing” (Malle and Knobe
1997a)). This condition shares some similarity with Searle’s (1983) notion
of an “intention-in-action” even though, somewhat surprisingly, Searle did
not characterize intentions-in-action as conscious. Future research should
address the stringency of the awareness condition as it applies to habitual
and automatic daily actions (e.g., eating or driving). In his analysis of these
conditions of intentionality, Mele poses a number of additional questions
about the intentionality concept that invite further research, such as the
generality of the intentionality-intention implication, the boundaries of
skill, and the relevance of moral considerations for judgments of inten-
tionality (as opposed to the relevance of intentionality considerations for
moral judgments—see chapter 16).
Despite some disagreement over details, there is consensus across disci-
plines that intention and intentionality are complex states that are ascribed
only if a set of simpler component states are present (in contrast to models
of “direct perception” of intentionality, which are discussed in the follow-
ing section). The ascription conditions for intention minimally include the
presence of desire, belief, and some form of commitment; those for inten-
tionality minimally include the presence of desire, belief, intention, skill,
and awareness. This does not mean, of course, that perceivers always com-
pute each and every component before they ascribe the resulting complex
state. Many routine actions and familiar social contexts permit the spon-
taneous judgment of intentions or intentionality without explicitly check-
ing as to whether each component is present. However, the constituent
components are very likely to be considered when such judgments are dif-
ficult to make, or when they are debated (as in interpersonal conflict or in
a court of law).
Analysis of the components of intention and intentionality offers many
advantages. For one, it helps us to separate concepts and phenomena that
adult social perceivers distinguish, such as desires, intentions, and inten-
tionality. As a result, we can ask precise questions, such as “How does chil-
dren’s theory of mind develop from broad category distinctions to
differentiated component concepts?” Moreover, unlike the full-blown con-
cepts themselves, their constituent components can be relatively easily

Introduction 7
grounded in lower-order precursors, such as belief in perception, desire in
bodily needs, and intention in acts of reaching—precursors that children
may perceive both in themselves as agents (Russell 1996) and in others
through social interactions (Bruner 1981; Dunn 1991). Analysis of com-
ponents also sharpens our understanding of extraneous variables that influ-
ence judgments of intention and intentionality, among them emotions and
stereotypes. Stereotypes may provide default assumptions about certain
components, such as expectations about an agent’s stable desires or beliefs,
which may bias the perceiver toward or against ascribing intentions and
intentionality. Finally, component models highlight a fundamental feature
of folk theories of mind: that they consist of conceptual networks system-
atically relating beliefs, desires, intentions, and other mental states to one
another and to observable behavior (Gopnik and Wellman 1994). These
networks provide tools for parsing and organizing what might otherwise be
a chaotic stream of mental experiences (in the case of self) and behaviors (in
the case of both self and others).
Traditionally, intentions have been regarded as private mental states that
one ascribes to individual persons. However, intentions can also be ascribed
to pairs or groups of people, who may have a joint intention to see a movie
together, win a game, or publish an edited volume. Recent work in philos-
ophy and psychology (e.g., Abelson, Dasgupta, Park, and Banaji 1998;
Bratman 1993; Gilbert 1989; O’Laughlin and Malle 2000; Searle 1990;
Velleman 1997) has begun to explore the nature of such joint intentions
and the intentionality ascribed to whole groups and even nations. In addi-
tion, psycholinguists have examined the emergence of shared meaning out
of individual intentions, a necessary process for successful conversation
and social coordination (Clark and Brennan 1991; Gibbs 1998; Krauss
and Fussell 1996). Interesting questions about the “location” of joint
intentions and the “location” of shared meaning arise. One wonders, for
example, whether there actually exist group minds that “have” mental
states or whether social perceivers merely metaphorically extend their folk
ascriptions of mental states to group agents. These puzzles notwithstand-
ing, the ascension from individual to shared mental phenomena is essen-
tial to human relations. Along these lines, Raymond Gibbs (chapter 5)
examines the function of shared meaning in communication and argues
that at least some intentions are not in the head but rather are emergent

8 Introduction
social phenomena. In addition, Daniel Ames, Eric Knowles, Michael
Morris, Charles Kalish, Andrea Rosati, and Alison Gopnik (chapter 15)
review recent psychological evidence that people comfortably apply their
theory of mind to individual agents as well as to group agents and, more-
over, that cultures seem to differ in their tendency to designate either
groups or individuals as the primary agent category.
Reading Intentions and Intentionality
People typically read the intentions underlying the behavior of others read-
ily and with little conscious effort. Of course failures occur, and these are
sometimes serious enough to give rise to argument, legal action, or inter-
national conflict. However, such interpretive failures are rare when mea-
sured against the countless actions to which perceivers smoothly assign
relevant intentional meanings—actions such as tooth brushing, newspaper
reading, and kitchen cleaning. Even actions motivated by complex and
potentially obscure intentions, such as the casting-about behavior occa-
sioned by a search for a television remote control, often pose little inter-
pretive difficulty for perceivers. The same can be said of novel actions. On
first viewing a skiboarder in action, for instance, it is easy to recognize the
intention of thrill seeking.
How do people so effortlessly detect intentions within the dynamic
behavior stream and so readily apprehend their content, and how is such
skill acquired in children’s development? Surprisingly, these questions
received little systematic examination before the relatively recent attempts
by social psychologists. For example, following ideas that Asch (1952) bor-
rowed from gestalt psychology, Newtson and his colleagues (see, e.g.,
Newtson 1973; Newtson and Engquist 1976) argued that people directly
perceive others’ intentions on witnessing their actions. Intentionality, and
the specific intentions at play, are thought to be there within the behavior
stream, waiting to be detected. Working within a similar direct-perception
framework, Premack and his colleagues (see, e.g., Premack 1990; Premack
and Premack 1995) provided nativist speculations about the origins of the
human ability to read others’ intentions. They suggested that infants arrive
in the world biologically prepared to perceive certain kinds of animate

Introduction 9
motion (in particular, self-propelled motion) as intentional, and that infants
and can recognize at least a small set of specific intentions (e.g., helping vs.
hurting) on the basis of the different behavioral patterns associated with
them.
Here it is important to separate possible claims about detecting inten-
tionality (how perceivers recognize that an intention is being enacted) from
those concerning the content of an agent’s intention (how perceivers recog-
nize which specific intention, or set of intentions, is being enacted). The
direct-perception framework seems at least potentially fruitful in accounting
for the former, but seriously flawed as an approach to the latter. Let us con-
sider these points in turn. Concerning the detection of intentionality, organ-
isms wired to read “intentional” anytime they encounter self-propelled
motion would be off to something of a start, because of the correlation
between such motion and intentional action (see also Mandler 1992 and
Wellman and Phillips in this volume). Of course, they would then need to
learn to suppress an intentionality reading for countless important excep-
tions involving self-propelled motion or the appearance of it: involuntary
behaviors such as sneezing, accidental and incidental motions such as the
inadvertent knocking of objects off counters, and the motion of many inan-
imates for which the cause of motion has either been missed or is not yet
understood (e.g., feathers and leaves blown by the wind, falling rocks, cars
and trains, computer cursors). Clearly, then, the direct-perception frame-
work requires substantial embellishment to successfully account for the full
spectrum of intentionality judgments that adult perceivers actually make.
Nonetheless, it might well capture the essence of how infants get their start
in the business of detecting intentionality.
In contrast, the direct-perception framework seems fundamentally
unworkable as an account of how perceivers detect the specific intentions
motivating others’ behavior. Behavior patterns and intentions stand in a
many-to-many relation (Baird 1999; Baldwin and Baird 1999; Searle 1984):
One and the same action (e.g., pressing a hypodermic needle into another’s
arm) admits of multiple intentional interpretations (e.g., the intent to heal vs.
harm), and one and the same intention (e.g., to heal) can give rise to many
possible actions (e.g., referral, advice, medication, surgery). Moreover, an
infinite number of possible intentions are consistent with any given action,

10 Introduction
yet only one of these candidates (or at most a very small set) is actually rel-
evant and usually considered by perceivers. Searle (1984, p. 58) captures this
beautifully:
If I am going for a walk to Hyde Park, there are any number of things that are
happening in the course of my walk, but their descriptions do not describe my
intentional actions, because in acting, what I am doing depends in large part on
what I think I am doing. So for example, I am also moving in the general direc-
tion of Patagonia, shaking the hair on my head up and down, wearing out my
shoes, and moving a lot of air molecules. However, none of these other descrip-
tions seems to get at what is essential about this action, as the action it is.
Searle’s example makes obvious that the content of agents’ intentions can-
not be recovered directly from the behavior stream itself; too many possi-
ble intentions are recoverable in any given case. In other words: From the
standpoint of the perceiver, the content of agents’ intentions is radically
underdetermined by their behavior.
An alternative approach views the detection of intentions and intention-
ality as the outcome of an inferential system. (See, e.g., Baldwin 1993b;
Baldwin and Baird 1999; Dittrich and Lea 1994; Meltzoff 1995; Tomasello,
Kruger, and Ratner 1993.) Unlike the direct-perception account, this infer-
ential framework readily accommodates, at least in principle, our ability as
perceivers to deal with the complex link between actions and intentions.
The specific intention motivating a given action is thought to be inferred
not just from the flow of behavior itself but also from external information,
including other cues in the immediate context (e.g., a medical setting such
as a clinic, the presence of doctors, nurses, and medical supplies), prior
knowledge about the agent (e.g., a physician vs. a violent offender), and
the script within which the agent’s motions are embedded (e.g., a physical
exam vs. a session of interrogation and torture). Sensitivity to such “extra-
behavioral” characteristics could enable perceivers to constrain their
inferences about intentions in the face of the limitless possibilities.
In addition to providing a possible account for the ability to interpret the
content of agents’ intentions, the inferential approach seems amenable to
explaining how people distinguish intentional from unintentional or inci-
dental behavior. Because behaviors from these different classes seem struc-
turally different in many cases (e.g., self-propelled motion is often intentional
whereas motion caused by direct physical contact with another moving body
rarely is), processing of the behavior stream can play an important role in

Introduction 11
making these distinctions. This is why the direct-perception framework also
can offer something in accounting for such ability. However, the inferential
approach again has the advantage in that it also has the potential to explain
the finer judgments about intentionality that social perceivers make—
for example, the ability to recognize the behavior of sleepwalkers and
“zombies” as unintentional despite the surface similarity of such motion to
that of conscious agents. In such cases, an inferential account would point
to information external to the behavior stream (e.g., night-time setting, his-
tory of night-time talking and walking, lack of response to questions,
known ingestion of mind-altering substances) as crucial in shaping the per-
ceiver’s inferences about the agent’s intentionality.
The chapters in part II all speak (at least implicitly) to the inferential
framework. Each offers new ideas and new evidence regarding the processes
involved in detecting intentions. In chapter 9, Jodie Baird and Dare Baldwin
highlight some of the qualities requisite to an inferential system for inten-
tional understanding. In particular, they propose that such an inferential
system crucially depends on the operation of a low-level structure-
detection mechanism capable of analyzing the dynamic behavior stream
into relevant units—units coinciding with the initiation and the completion
of intentions—for further analysis. Moreover, they present new evidence
that adults as well as 10–11-month-old infants spontaneously parse con-
tinuous intentional action in terms of just such “intention-relevant” units.
To date, the preponderance of work within the inferential approach has
focused on development, exploring infants’ and young children’s emerging
abilities to detect and interpret the intentions motivating others’ behavior.
Many of the chapters in part II reflect this trend, as they deal primarily with
developmental evidence. In chapter 7, Amanda Woodward, Jessica
Somerville, and José Guajardo present research indicating that infants as
young as 9 months understand the goal-oriented quality of some intentional
actions and can distinguish between intentional and unintentional action in
at least some cases. Further, they find that at this early age infants already
use information external to the behavior stream to determine the relevance
of a goal object. For example, previously provided information about an
agent’s interest in the contents of a box led infants to construe a subse-
quent box-grasping action as goal-oriented; in the absence of such prior
information, infants failed to register the action’s goal-oriented quality.

12 Introduction
This “action-in-context” effect meshes nicely with the predictions of the
inferential framework described above, which is grounded in the idea that
intentions are inferred by interpreting action within its larger context.
Inferential theorists concerned with how children come to detect inten-
tions and intentionality typically think in constructivist—as opposed to
purely nativist—terms. To the extent that judgments about intentions are
derived through complex inferential processes depending on experience and
world knowledge, conceptual change within this arena is to be expected as
development proceeds. Many of the chapters in part II manifest this par-
tiality for constructivist speculation regarding the origins of intentional
understanding. For example, in discussing the origins of early intentional
understanding, Henry Wellman and Ann Phillips (chapter 6) offer an
analysis of the kind of input regarding intentions and intentionality that
might be available to infants through observation of others’ behavior. They
go on to present new evidence that infants as young as 12 months are sen-
sitive to two perceptible features of behavior—object-directedness and
action-connectedness—that are typical of intentional action. As Wellman
and Phillips suggest, early recognition of these features may not actually
reflect a genuine understanding of the agent’s intentions; rather, it may rep-
resent crucial steps toward such understanding. In a similar vein, Wood-
ward and colleagues (chapter 7) have found that infants at 9 months
process some actions (such as grasping) in ways that are relevant to inten-
tions but fail to process other actions (such as pointing) in these terms.
This is among the first evidence suggesting that abilities enabling the detec-
tion of intentional content are constructed through infants’ world experi-
ence. In chapter 9, Baird and Baldwin suggest that a low-level mechanism
for analyzing action plays a crucial role in making possible developmen-
tal change of the kind that Woodward et al. demonstrate.
In chapter 8, Andrew Meltzoff and Rechele Brooks provide important
ballast to the constructivist stance by embracing a hybrid account that cred-
its newborns with crucial skills for interpreting others’ actions. Such an
account helps to explain how typically developing children so easily and
naturally come to understand others’ intentions. Meltzoff and Brooks pro-
pose a “starting-state nativism” that consists of innate foundations that are
modified by extensive development and practice in social interactions. The
innate foundations are at work, for example, when a newborn imitates

Introduction 13
someone’s actions, thus translating a perceived act into an act of its own.
This imitation relies on a mapping between others and self—at birth, on
the level of actions. By 18 months, this mapping occurs on the level of goals.
Meltzoff and Brooks describe experiments in which infants observed
another person engaging in action that failed with respect to goal attain-
ment (e.g., an attempt to open a device was unsuccessful) but nevertheless
inferred the goal and spontaneously performed the action that successfully
led to the goal (opening the device). Meltzoff and Brooks argue that the piv-
otal element in early imitation, in later goal inference, and in many other
achievements of infant social cognition is the “like me” analogy—the ten-
dency to see others’ acts as being like acts the infants can produce them-
selves. As they experience their own attempts to control behavior, infants
build maps that link effort experiences, goals, and their own actions, and
they use these maps to infer others’ goals from observed actions.
The idea of an innate analogical process shares important similarities
with a simulation theory account such as the one suggested by Alvin
Goldman in chapter 10. Like Meltzoff and Brooks, Goldman offers his
account as an alternative to a purely inferential framework. A key element
of his model is the distinction between first-person and third-person attri-
butions of mental states. Goldman argues that first-person attributions are
based not on inference but rather on a form of direct perception, or intro-
spection. He defends this proposal against recent skepticism regarding the
possibility of introspection (e.g., Gopnik 1993). Goldman goes on to argue
that social perceivers make third-person attributions of mental states with
the help of first-person access, using simulation processes to re-create and
thereby represent others’ mental states. He then musters various lines of
evidence for the plausibility of simulation as a core mechanism of intention
reading. This evidence includes the intriguing possibility of “mirror neu-
rons” in primates that seem to fire both when the organism perceives cer-
tain actions performed by others and when it performs that action itself—a
mechanism not unlike the supramodal representation system discussed by
Meltzoff and Brooks in chapter 8.
Any attempt to account for the ability to read others’ intentions must
grapple with the extent to which such skill is special to humans. On the one
hand, there is an obvious gulf dividing humans from other species in terms
of the complexity of reasoning about others’ intentions. Imputing to others

14 Introduction
complex intentions with multiple subparts, such as intentions to embezzle
funds for charitable purposes or to run for president of the United States in
good faith but with little hope of success is everyday fare for humans, but as
far as we can tell nothing of comparable complexity has ever been seen in
other species. On the other hand, behavior that powerfully suggests the
essence of intention-reading ability is ubiquitous in interactions between
humans and other species and in interactions among members of other
species. A dog’s suspicion at signs indicating an intention to bathe him and
his ecstasy when he notes preparations for a walk are examples familiar to
many. Other examples are easy to find throughout modern literature, even
at the academic level; witness two influential volumes concerning the social
understanding and the “Machiavellian intelligence” of higher primates
(Byrne and Whiten 1986; Whiten and Byrne 1997).
In chapter 11, Daniel Povinelli describes a substantial body of work
investigating these issues. Based on the evidence, he suggests that chim-
panzees lack a genuine appreciation for the intentions motivating others’
actions, despite their evident skill at processing such actions in ways that
enable them to predict and influence others’ behavior. To account for this
apparent paradox, Povinelli proposes the operation of two independent
mechanisms. One is the skill of analyzing, correlating, and predicting
complex patterns of behavior on the merely behavioral level; the other
comprises genuinely mentalistic reasoning about intentions and inten-
tionality underlying behavior. On this proposal, both humans and chim-
panzees are skilled at behavior analysis but only humans are capable of
mentalistic reasoning. And Povinelli speculates that the developmental
progress from infants’ behavior analysis to preschoolers’ mentalistic rea-
soning may be a qualitative step—the emergence of a separate mecha-
nism—rather than a gradual elaboration of a single mechanism from its
incipient to its mature stage.
Intentionality and Explanations
Explanations of behavior are regarded by many as a key function of folk
psychology, and the concept of intentionality plays a pivotal role in the con-
struction of such explanations. Behavior explanations take cognitions of
behavior (including judgments about the behavior’s intentionality) as input

Introduction 15
Figure 1
Four approaches to the study of explanation.
and render as output a model of what generated the behavior, often includ-
ing reference to mental states such as beliefs, desires, and intentions. Such
a model, in turn, influences judgments of responsibility, predictions of
future behavior, and attempts to change the behavior.
Various traditions of explanation research across disciplines can be orga-
nized along two dimensions: whether or not the role of intentionality in
explanations is considered, and whether or not explanations are studied in
their natural context of conversation and social interaction. Figure 1 shows
the resulting four combinations.
The first cell contains approaches to explanation that do not consider the
role of intentionality and analyze explanations independently of the inter-
active context in which they occur. According to this causal judgment
approach, explanations are cognitive processes (often unconscious) that
apply equally to all objects of explanation, be they physical or behavioral,
intentional or unintentional events. Prime representatives of this approach
include Kelley’s (1967) ANOVA model of causal attribution and its vari-
ants (e.g., Hewstone and Jaspars 1987), the cognitive study of causal rea-
soning (Spellman 1997; Cheng and Novick 1990), and normative models
of scientific explanation (Hempel and Oppenheim 1948). The strength of
this approach is that it focuses on the common cognitive principle of all
explanations: that they identify causal antecedents of the explanandum
according to rules of logic and evidence. At the same time, this focus is also
its major weakness, because distinctions among explananda (e.g., actions
vs. physical events) and corresponding distinctions between causal models

16 Introduction
are overlooked, as is the social context that gives explanations their func-
tion, regardless of how they are cognitively (or physiologically) realized.
The second cell contains paradigms that still analyze explanations inde-
pendently of their interactive context but that consider the role of inten-
tionality by distinguishing between two types of explananda (intentional
human action vs. all other events) and their corresponding types of expla-
nation (often labeled reasons and causes). Representatives of this inten-
tional approach include the hermeneutic movement in the social sciences
(e.g., Gadamer 1989; Harre and Secord 1972; von Wright 1971) and the
substantial contingent of philosophers who consider reason explanations of
intentional action to be unique and irreducible (e.g., Audi 1993; Davidson
1963; Mele 1992a; Taylor 1964). The strength of this approach is that it
recognizes important conceptual differences between two types of expla-
nation and tries to work out the implications for theories of science, moti-
vation, and action. However, the psychological reality of the two types of
explanation has not been explored within this tradition, because it regards
explanations primarily as logical entities rather than as verbal behaviors.
Philosophers of action sometimes assume that ordinary people, too, dis-
tinguish between reasons and causes, but no systematic empirical data are
ever mustered in support of this assumption.
The third cell contains the communicative approach, which emphasizes
the social context and function of explanations, especially their dialogical
nature in both everyday and scientific settings (Antaki 1994; Bromberger
1965; Hilton 1990; Kidd and Amabile 1981; Turnbull and Slugoski 1988).
Explanations are seen as answers to why-questions, filling a knowledge gap
exhibited by the questioner. (The explainer and the questioner are typically
separate individuals, but in the case of private explanations they are identi-
cal.) Scholars within this approach—e.g., Slugoski, Lalljee, Lamb, and
Ginsburg (1993)—have recognized that causal judgments are responsive to
the social demands created by why-questions, such that explainers search
for and present different explanations depending on their inferences about
the questioner’s background and the particular knowledge gap that is to be
filled. For example, when Q asks E “How come Mary bought a Mercedes?”
E might answer “Because it’s a good car.” However, if E considers that Q
asks the question because Q knows Mary is poor, a more appropriate answer
would be “Because she inherited a load of money” (McClure and Hilton

Introduction 17
1997). Variations in explanations are therefore understood not just in terms
of different causal perceptions but also in terms of different social demands
that the explainer tries to meet (Malle, Knobe, O’Laughlin, Pearce, and
Nelson 2000). Perceived gaps in knowledge present the most obvious
demands; authority and accountability have also been examined (Edwards
and Potter 1993; Scott and Lyman 1968; Tedeschi and Reiss 1981).
The communicative approach shares with the causal-judgment approach
its major strength and its major weakness: It identifies general principles of
explanation but does not distinguish between different types of explananda;
hence, it still assumes that humans have a uniform conceptual model of
causality. The consideration of social context, however, is a distinct strength
of the communicative approach. Important questions follow from this con-
sideration, such as whether the “truth” of an explanation depends solely on
a speaker’s and an audience’s assumptions and to what extent social demands
may modify not only verbal behavior but actual causal perceptions.
The fourth cell contains approaches emphasizing both the role of inten-
tionality in explanations and the social-interactive context in which expla-
nations are embedded. We call this the folk-theoretical approach because it
considers explanations as an integral part of folk theories for core domains
of cognition, such as psychology, physics, and biology (Carey 1995; Malle
1997; Wellman, Hickling, and Schult 1997). Within folk psychology, expla-
nations heavily implicate the concept of intentionality in that all human
behavior is classified as either intentional or unintentional and (depending
on the classification) explained in conceptually distinct ways (Buss 1978;
Locke and Pennington 1982; Kalish 1998; Malle 1999; Read 1987; White
1991). In particular, unintentional behavior is explained by mere causes,
which are seen as simply bringing about the effect in a mechanical way (sad-
ness causes crying; sunshine causes happiness). In contrast, intentional
behavior is explained by reasons—the beliefs and desires in light of which
the agent formed an intention to act.
Unlike scholars who take the intentional approach, those who endorse
the folk-theoretical approach do not try to clarify the nature of explanation
in a philosophical sense, nor do they necessarily postulate the objective
existence of intentionality. Instead, they analyze explanations and intention-
ality as cognitive tools that guide people’s perception, prediction, and con-
trol of their environment. Explanations are thus assigned a psychological

18 Introduction
reality that is grounded in a shared folk-conceptual framework. In addi-
tion, and in agreement with the communicative approach, some researchers
consider explanations also to be a social tool, expressed and strategically
used in social interaction (Bartsch and Wellman 1995a; Malle and Knobe
1997b; Malle et al. 2000).
A possible weakness of the folk-theoretical approach is that, in appreci-
ating the complexity and the variability of explanations in particular
domains, it may lose the generality that other approaches seek. However,
domain specificity is not incompatible with generality. Some functions of
explanations (e.g., the need to anticipate and control one’s environment
(Heider 1958) and the desire to propel one’s theoretical understanding of it
(Gopnik 1998)) hold across domains; other functions (e.g., the face-saving
and evaluative character of behavior explanations) are specific to particular
domains. The same might be said of the cognitive processes underlying
explanations, of which some may be domain-general (e.g., considerations
of temporal order in causal processes) and others domain-specific (e.g., con-
siderations of rationality in explanations of intentional behavior). Future
research will have to clarify the exact similarities and differences among dif-
ferent types of explanation, but the contribution of the folk concept of inten-
tionality to some of these differences is already apparent.
Several recent advances in folk-explanation research are driven by the
assumption that people learn to master not one but a variety of modes of
explanation (e.g., Kalish 1998; Malle 1999; McClure and Hilton 1997;
Wellman et al. 1997). One question to be settled is whether these modes of
explanation can be distinguished solely by their domain of application
(Wellman et al. 1997; Schult and Wellman 1997) or whether they exhibit
distinct conceptual structures (Kalish 1998; Malle 1999). Broad domain
distinctions (e.g., physics, biology, psychology) provide useful approxima-
tions of explanatory types, but they may be less helpful for distinguishing
modes of explanation within the richest of domains: that of human behav-
ior. Because humans can be characterized as physical, biological, and psy-
chological systems, explanations of human behavior encompass all the
forms of explanation employed in other domains but also make use of
unique modes that apply only to intentional action.
All the chapters in part III deal in one way or another with the interplay
and the distinctions among the various modes of explaining human behav-

Introduction 19
ior. In chapter 12, G. F. Schueler takes a close look at the tension between
two ways of explaining intentional action: by means of the agent’s own rea-
sons, and by means of objective causal processes that seem to leave no room
for genuine reasons. He argues that reason explanations are distinct from
and not reducible to standard causal explanations because they funda-
mentally incorporate normative features. Even reason explanations offered
by an observer must mimic the normative reasoning process the agent went
through when deciding to act.
In chapter 13, Bertram Malle argues that folk explanations of intentional
behavior encompass three different modes, with reason explanations the pri-
mary one and causal history of reason explanations and enabling factor
explanations the secondary ones. Malle proposes a model that distinguishes
these modes by their conceptual, linguistic, and functional features and con-
trasts this model with alternative theories of attribution and explanation.
He also explores what implications this plurality of explanation modes has
for the well-known debate between “theory theorists” (who assume that
people ascribe mental states by relying on an organized set of generaliza-
tions) and “simulation theorists” (who assume that people ascribe mental
states by relying on their capacity to simulate these states in their own mind).
In chapter 14, Andrea Rosati, Eric Knowles, Charles Kalish, Alison
Gopnik, Daniel Ames, and Michael Morris explore the possible reconcil-
iation between behavior explanations referring to the agent’s mental states
and behavior explanations referring to the agent’s enduring traits. Rosati
et al. identify mental-state components in trait concepts and demonstrate
that trait inferences rely crucially on mental-state inferences. In the past,
mental-state explanations were featured in the theory of mind tradition
and trait explanations in the social-psychological attribution tradition.
Rosati et al. attempt to integrate these rather disparate paradigms of
research into social cognition.
Intentionality, Responsibility, and Social Context
Interpersonal perception encompasses not only cold assessments of others’
mental states but also affective and moral responses to those states and to the
social actions the states generate. These responses, which include praise,
blame, pride, shame, resentment, and gratitude, may well be unique to

20 Introduction
humans. They stake out the evaluative and corrective functions of the folk
theory of mind, and they place issues of intentionality in a larger societal
context. Moral responses are inextricably linked to the constituents of inten-
tionality, to mental-state inferences, and to explanations of action, the top-
ics discussed in parts I–III of this volume. But exactly how these connections
should be understood—that is, exactly how intentionality relates to moral
responsibility—is a matter of some debate in psychology, philosophy, and
the law. Below we provide a conceptual framework for studying the inten-
tionality–responsibility connection, and we locate some recent literature as
well as this volume’s contributions within that framework.
Responsibility has many meanings (cf. Frey and Morris 1991; Hamilton
and Sanders 1992; Hart 1968), and so the role of intentionality in respon-
sibility will differ depending on the meaning in view. Responsibility or being
responsible always refers to a socially ascribed relation that holds either (1)
between an agent and a specific action or outcome, (2) between an agent
and that agent’s general capacity for acting, or (3) between a cause and an
effect. The third relation casts responsibility merely as causality (e.g., “The
hurricane was responsible for 10 deaths”2), a derivative meaning that
exports the first (agent/outcome) meaning of responsibility from the psy-
chological realm into the realm of natural events but removes its implica-
tion of intentional agency. We therefore focus here on the two principal
meanings, which might be labeled normative responsibility and responsi-
bility as agency. (For similar distinctions, see Bratman 1997, note 7.)
Normative responsibility establishes a normative relation between an
agent and a specific action or outcome. This relation is cast either as duty
(“Mulder and Scully are responsible for investigating every paranormal
event in the country”) or as liability (liableness to blame3) (“That consti-
tutes child neglect and the parents should be held responsible!”). Duty and
liability are normative in that they refer to social rules or expectations that
dictate what the agent should do or should have done (Hamilton 1978).
Responsibility as duty is typically used in a forward-looking manner to
direct the agent’s future actions (“What are your responsibilities at your
new job?”). One has such responsibilities because they are given or
assumed. Responsibility as liability is typically used in a backward-looking
manner to respond to negative outcomes (“Police and public officials are
not responsible for the attack”). One is held responsible or accepts respon-

Introduction 21
sibility for those outcomes (which usually means that one is blamed or
accepts blame). In practice, the two normative relations are often blended,
as pre-existing duties are considered when assigning liability (Haidt and
Baron 1996; Hamilton 1978).
Responsibility as duty presupposes the capacity for intentional action.
Assigning specific duties to a person is pointless unless one assumes that
the person can intentionally fulfill them. Responsibility as liability, too, pre-
supposes the capacity for intentional action: Liability is assigned when the
agent could and should have acted so as to prevent the outcome but didn’t
(Hamilton and Sanders 1992; Weiner 1995). “Could have” corresponds to
an assumption of preventability or intentional controllability; “should
have” corresponds to an assumption of duty. Liability does not, however,
presuppose factual intentionality—agents can be liable to blame even for
outcomes they did not bring about intentionally but rather caused through
negligence or recklessness (Duff 1990; Hart 1968). Intentionality amplifies
liability, to be sure, and lack of intention can ameliorate it (Heider 1958;
Schlenker, Britt, Pennington, Murphy, and Doherty 1994). But liability
holds even for unintentionally caused outcomes, as long as the agent had the
intentional capacity and duty to prevent that outcome.
Beginning with the classic work of Piaget (1932), research on children’s
developing moral reasoning has emphasized the transition from judgments
based solely on outcome severity to judgments incorporating or even focus-
ing on the agent’s intentions and motives. This transition occurs in the
preschool period (Nelson-LeGall 1985; Yuill and Perner 1988), apparently
after a child begins to distinguish intentional from unintentional behavior
(Shultz 1980) and to ascribe motives to agents (Bartsch and Wellman
1995a). Not surprisingly, it seems to take a child some time to learn to apply
the intentionality concept to the new function of distinguishing a blame-
worthy from a blameless state of mind.
An even more complex step occurs when a child learns that sometimes
even unintentional behaviors leave the actor subject to blame. In chapter 17,
Michael Chandler, Bryan Sokol, and Darcey Hallett explore this intriguing
developmental step. As was mentioned above, adults hold agents responsi-
ble when they are perceived to have been capable of preventing the behav-
ior and to have a duty to do so. Chandler et al. demonstrate that by age 5
children begin to take these counterfactuals into account when making

22 Introduction
judgments of blame. Furthermore, they argue that this advance in moral
reasoning is tied to the onset of a “constructivist” theory of mind—one that
fully appreciates that human agents interpret and construct what they know
about the world but that their interpretations and their ensuing moral con-
duct can be more or less justified.
Responsibility as agency, the second principal meaning of responsibility,
refers to an agent’s general capacity to perform autonomous, rational action
(e.g., “The successful performance of chores is another way that patients
can demonstrate they are ready for greater financial responsibility”). To act
with responsibility in this sense requires the ability to consider the con-
sequences of one’s actions and to choose an action with desirable conse-
quences (“Taking responsibility means accepting the consequences of your
own choices”). Responsible agency thus presupposes intentionality in the
form of planning, deliberation, and reasoning (Bratman 1997; Hart 1968).
Indeed, these planning features of intentionality define a particular version
of responsible agency: acting with good judgment (“If you choose to drink
alcohol, drink responsibly”). However, exactly what goes into the concept
of responsible agency beyond the capacity for intentional action is still
debated among philosophers (see, e.g., Bratman 1997; Fischer 1994;
Wallace 1994). Proposed criteria include rationality, communicative capac-
ity, and responsiveness to reasons. The choice of these criteria has great
practical importance; for example, it affects sentiments and decisions about
the responsibility of children and that of mental patients. However, there
have been no empirical studies exploring what the criteria for responsible
agency might be. Psychological research (see, e.g., Fincham and Jaspars
1980; Shaver 1985; Weiner 1995) has focused almost exclusively on the
conditions under which people assign liability, leaving discussions of factors
that constitute responsible agency to legal and philosophical scholars.
Normative responsibility, responsibility as agency, and intentionality are
closely intertwined concepts. Responsible agency presupposes intentional-
ity (the capacity to deliberate about and choose one’s course of action).
Normative responsibility in turn presupposes responsible agency (hence
intentional capacity), for unless one considers an agent equipped with
responsible agency in general one cannot assign to this agent any specific
duties or any liability for specific outcomes. Thus, the folk-psychological
assumption that humans are capable of intentional action underlies both

Introduction 23
of the principal meanings of responsibility; as a result, judgments of inten-
tion and intentionality permeate the social practices of praise, blame,
reward, and punishment (Marshall 1968; Williams 1993).
The fundamental role that the assumption of intentionality plays in
responsibility is further highlighted when we consider the social role and
function of responsibility attributions. The practice of assigning responsi-
bility—in all its meanings—serves the coordination and organization of
social activities, the maintenance of social order, and the enforcement of
social rules (Heider 1944; Schlick 1966; Semin and Manstead 1983).
Normative responsibility, in particular, lays the foundation for a social feed-
back system in which desirable outcomes yield a premium and undesirable
outcomes are sanctioned. But this normative system applies only to those
outcomes that are in principle controllable by intentional agency (exclud-
ing, for example, natural disasters) and involves only those agents who are
equipped with responsible agency (excluding, for example, young children
and some mental patients). This feedback system has often been discussed
in theoretical terms, but few scholars have explored it in detail. In chapter
16, Bernard Weiner does exactly that, detailing some of the system’s cog-
nitive, emotional, and behavioral elements and using a model centered on
the concept of controllability to account for moral sentiments (such as anger
or pity), philosophies of punishment, and individual differences in political
ideology.
Two chapters put intentionality and responsibility in their larger social
and cultural contexts. In chapter 15, Ames, Knowles, Morris, Kalish, Rosati,
and Gopnik try to integrate considerations of norms and context into the
often purely cognitive models of social perception. They examine, in partic-
ular, how people’s social and cultural contexts shape their judgments of
intentionality and responsibility as well as their mental-state ascriptions and
their behavior explanations. In chapter 18, Leonard Kaplan analyzes how
conceptions of intentional agency and responsibility interrelate and, more
important, how they are in tension with expectations of justice in the mod-
ern state. Kaplan examines several models of moral action and identifies
their varying assumptions about intentional agency, responsibility, and jus-
tice. Because all these models center on responsibility as the ethical duty to
be responsive to another’s needs, classic issues of social cognition arise when
the models specify to what extent an individual is capable of recognizing the

24 Introduction
suffering of others and distinguishing it from deception or exploitation. Even
though the particular ethical problem of enabling justice in the modern state
must remain unsolved, Kaplan’s analysis illustrates how ethical discourse
presupposes the agent’s capacity to act intentionally and to perceive and
interpret the social world and other beings within it.
Conclusions
Theories and research programs on the role of intention and intentionality
in social cognition are distributed over many scholars, traditions, and dis-
ciplines. These individual efforts, though united by the goal of elucidating
interpersonal understanding, have often remained isolated from one
another. A unifying theory of how humans understand other humans will
have to emerge from communication and collaboration across the tradi-
tional boundaries of paradigms and disciplines. The research brought
together in this volume, we hope, both attests to the fundamental role of
intentionality in human social cognition and offers noteworthy progress
toward a broad and interdisciplinary account of human social relations.
Acknowledgments
Preparation of this chapter was supported by a National Science Foundation
CAREER award (No. 9703315) to Bertram Malle and by a National Science
Foundation New Young Investigator Award (No. 9458339) and a John
Merck Scholars Award to Dare Baldwin. The chapter was prepared while
Dare Baldwin was a fellow at the Center for Advanced Study in the Behav-
ioral Sciences; she is grateful for the financial support provided by the
William T. Grant Foundation under award 95167795.
Notes
1. On the appropriateness of ascribing to such an agent only an intention to try to
A, not an intention to A, see Mele 1989.
2. Quoted examples were found by searching the World Wide Web and various
newspapers for sentences containing the words responsibility or responsible.
3. This meaning of responsibility is also labeled accountability, answerability, or
blameworthiness in the literature. Its legal version is liableness for punishment.