Event Perception: A Mind-Brain Perspective

Psychological Bulletin Copyright 2007 by the American Psychological Association
2007, Vol. 133, No. 2, 273–293 0033-2909/07/$12.00 DOI: 10.1037/0033-2909.133.2.273
Event Perception: A Mind–Brain Perspective
Jeffrey M. Zacks, Nicole K. Speer, Khena M. Swallow, Todd S. Braver, and Jeremy R. Reynolds
Washington University in St. Louis
People perceive and conceive of activity in terms of discrete events. Here the authors propose a theory
according to which the perception of boundaries between events arises from ongoing perceptual
processing and regulates attention and memory. Perceptual systems continuously make predictions about
what will happen next. When transient errors in predictions arise, an event boundary is perceived.
According to the theory, the perception of events depends on both sensory cues and knowledge structures
that represent previously learned information about event parts and inferences about actors’ goals and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
plans. Neurological and neurophysiological data suggest that representations of events may be imple-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
mented by structures in the lateral prefrontal cortex and that perceptual prediction error is calculated and
evaluated by a processing pathway, including the anterior cingulate cortex and subcortical neuromodu-
latory systems.
Keywords: events, attention, memory, prefrontal cortex, orienting response
What Is an Event? beginning and an end” (Zacks & Tversky, 2001, p. 17). This
definition is useful but surely does not exhaust the common
The world as presented to human sense organs is continuous, conception of an event. The everyday notion of an event probably
dynamic, and fleeting. Yet people seem to perceive events as has a family-resemblance structure, with some highly typical
stable entities, to identify parts of events and their relations to other members such as weddings and breakfasts and some atypical
parts. For example, one might describe baking cookies by listing members such as the decay of a radioactive atom or the melting of
the following parts: “Preheating the oven, mixing the ingredients a pond. Typical events seem to share some common features. They
in a bowl, putting the dough on a cookie sheet . . .” This could range from a few seconds (eating a strawberry) to a few hours
reflect mere happenstance or accidents of linguistic structure, but (going for a hike). They are directed toward a goal; the goal of a
a growing body of research suggests that talk of discrete events wedding is to formalize a union, and the goal of breakfast is to sate
reflects a deeper psychological reality, that people perceive activ- one’s hunger. Events involve animate agents, often human. These
ity in terms of discrete events, that ongoing processing resources features, however, are neither necessary nor sufficient. Some
are devoted to this perceptual process, and that the online percep- events are quite short (cutting a ribbon) or quite long (World War
tion of events determines how episodes are encoded in memory II). Events that are natural occurrences, such as landslides, may
(Zacks & Tversky, 2001). Thus, events are key components of lack both goals and animate agents. So, the taxonomic boundaries
perception, attention, and memory. of the category event are fuzzy. The spatial and temporal bound-
In this article, we present a theory of the perception of everyday aries of events also can be fuzzy—it is sometimes difficult to say
events, review psychological data that have informed the theory, where or when one event ends and another begins. Neither taxo-
and discuss possible neural substrates of the theory’s components. nomic fuzziness nor boundary fuzziness is a particular problem for
We begin with the formal definition of an event: “a segment of the psychology of events, and both are exactly analogous to the
time at a given location that is conceived by an observer to have a psychology of objects. Here, we are concerned with the core of the
category event: events that involve goal-directed human activity
and are of modest duration (seconds to tens of minutes). (For
Jeffrey M. Zacks and Todd S. Braver, Department of Psychology and
comparative reviews of conceptions of events in psychology, see
Department of Radiology, Washington University in St. Louis; Nicole K. Stränger & Hommel, 1996; Shipley, in press.) In the next section,
Speer, Khena M. Swallow, and Jeremy R. Reynolds, Department of Psy- we present a theory of how human observers segment continuous
chology, Washington University in St. Louis. activity into discrete events. In subsequent sections, we discuss
Nicole K. Speer and Jeremy R. Reynolds are now at the Department of evidence in support of this theory and its implications.
Psychology, University of Colorado.
The research summarized here was conducted with a number of collab-
orators, including Margaret Sheridan, Barbara Tversky, and Jean Vettel. It A Theory of Event Segmentation
was supported in part by the James S. McDonnell foundation, National
Perception can be described as a roughly hierarchical process in
Institutes of Health (Grant MH62318-01), and National Science Founda-
tion (Grant 0236651). The article benefited from the comments of Corey
which sensory information is successively transformed into repre-
Maley and Barbara Tversky. sentations that form the basis for action. Particularly important are
Correspondence concerning this article should be addressed to Jeffrey representations of states of the world in the near future, which may
M. Zacks, Department of Psychology, Washington University, St. Louis, be called perceptual predictions. Perceptual predictions are valu-
MO 63130-4899. E-mail: [email protected] able because they allow an organism to anticipate the future and to
273
274 ZACKS, SPEER, SWALLOW, BRAVER, AND REYNOLDS
plan appropriate actions rather than merely react to incoming ceptual processing to produce multimodal representations with
stimuli. Such representations are critical for avoiding interception rich semantic content, encoding information such as object identity
by predators, intercepting prey, and coordinating behavior with and location, motion trajectories, and the identities and attitudes of
others. To the extent that information processing is hierarchical, other people. According to the theory, processing is oriented in
perceptual predictions arise late in the processing hierarchy be- time such that it results in predictions about the future state of
cause incoming sensory information is transformed to generate perceptual representations. For example, extracting a motion con-
predictions. In addition to being hierarchical, perception can be tour leads to predictions about the future locations of objects, and
described as recurrent: Later processing stages affect the flow of inferring the goals of a person leads to predictions about his or her
processing in earlier stages. Finally, perception can be described as future movements.
cyclical: Perceptual predictions are constantly compared with what We propose that perceptual processing is guided by a set of
actually happens, and these comparisons are used to guide ongoing representations called event models that bias processing in the
processing. These three notions— hierarchy, recurrence, and cy- perceptual stream. An event model is a representation of “what is
clicality—are working assumptions in many different theories of happening now,” which is robust to transient variability in the
perception (Neisser, 1967), neurophysiology (Carpenter, & Gross-
sensory input. The stability of event models over time is a source

berg, 2003; Fuster, 1991), and language processing (van Dijk & of perceptual constancy; an ongoing event is a single entity despite
Kintsch, 1983). They have been developed perhaps most fully in potential disruptions in sensory input such as occlusion or distrac-
recurrent neural network models, which have been applied to word tion. In this regard, event models are similar to the object files
learning (Elman, 1990), to action learning (Jordan & Rumelhart, proposed to mediate object constancy in visual perception (Kah-
1992), and to event perception (C. Hanson & Hanson, 1996). neman, Treisman, & Gibbs, 1992) or the short-term action repre-
The theory presented here, which we call event segmentation
sentations proposed to mediate perceptual constancy in biological
theory (EST), shares these three properties. In this section, we
motion (Stränger & Hommel, 1996, see section 5). However, event
describe the theory in information-processing terms. Later in the
models are hypothesized to be active over much longer time
article, we recast the theory in terms of the neural systems that may
frames than object files or biological motion representations. In
implement these information-processing components.
terms of the current theory, object files and biological motion
representations are hypothesized to be part of high-level perceptual
Architecture and Principles processing components.
EST proposes that event segmentation arises from the percep- Event models are working memory representations, which are
tual processing stream depicted in Figure 1. Its core is a pathway implemented by transient changes in neural activation rather than
whose input is a set of sensory representations and whose output is long-term changes in synaptic weights. They are not necessarily
a set of perceptual predictions. The sensory inputs correspond to accessible to consciousness, though people may have partial
the information conveyed by the peripheral nervous system to the awareness of their contents under some circumstances. Event
cortex. In the visual modality, for example, this corresponds to models are multimodal, integrating information from visual, audi-
basic information about brightness, color, and possibly some pre- tory, and the other sensory modalities. In these regards they are
liminary edge extraction. Sensory inputs are transformed by per- akin to the representations recently proposed by Baddeley (2000)
Error Detection
Predicted Future Inputs
Perceptual Processing Event Models Event Schemata
Sensory Inputs
Figure 1. Schematic depiction of the theory. Thin gray arrows indicate the flow of information between
processing areas. Dashed lines indicate projections that lead to the resetting of event models. The connection
from sensory inputs to event models is gated, such that event models receive sensory input only during the reset
phase.
EVENT PERCEPTION 275
as forming an episodic buffer. Most of the time, the contents of needed to perform skilled activities is rapidly encoded into long-
event models are insensitive to immediate sensory and perceptual term memory using knowledge structures that anticipate future
input, providing a stable representation of the current event to retrieval demands. This information remains readily accessible as
guide perceptual processing. This is indicated by the gated arrow long as some part of it is available in short-term memory and can
terminating on the event models component in Figure 1. Event be rapidly reinstated once some part of the structure is retrieved.
models also receive input from event schemata—semantic mem- Developing effective long-term working memory for information
ory representations that capture shared features of previously en- in a particular domain is a skill that requires repeated practice with
countered events. Event schemata contain previously learned in- that domain. We presume that most human adults have had exten-
formation about the sequential structure of activity. Unlike event sive experience with a wide range of events. Thus, event schemata
models, event schemata are implemented by permanent synaptic help expand the effective capacity of event models by storing
changes. The information they store includes distinctive physical predictive information about the future relevance of certain aspects
features such as object and actor movement, statistical information of events.
about which patterns of activity are likely to follow a given pattern, Here is an example of how the mechanism in EST might work
and information about actors’ goals. when one observes an everyday activity: Imagine watching a man
The quality of perceptual prediction depends critically on wash dishes. He takes one plate from a pile next to the sink,
whether one’s current event models are a good fit to what is scrapes food from it, and then places it in the sink. He does the
actually happening. Prediction quality is evaluated by an error same with a second plate. At this point, a number of cues make it
detection mechanism that compares the perceptual processing likely that he will continue to scrape the plates. First, continuing to
stream’s predictions with what actually happens in the world. Most scrape would maintain a coherent movement pattern. Second, it
of the time, the event models represent the current state of events would be consistent with previous observations in which it was
well, and perceptual prediction is easy and accurate. From time to statistically likely that scraping a plate was followed by more
time, however, activity becomes less predictable, and the current scraping of plates. Third, the observer might infer that the man had
contents of the event models become less useful for perceptual the goal of scraping all of the plates. Thus, for the duration of the
prediction. At these points, prediction error increases. EST pro- plate-scraping activity, affairs would be predictable. However,
poses that at these points a gating mechanism detects these tran- when the man scraped the last plate, things would become less
sient increases in prediction error and reacts to them by updating predictable. The coherent movement pattern would cease, the
the event models. Updating consists of (a) resetting the current statistical dependency would be broken, and the inference of the
representations (indicated by the dashed black line in Figure 1) and actor’s goal to scrape all of the plates would no longer have
(b) transiently increasing the influence of the pathway from sen- predictive value. At this point, perceptual prediction would de-
sory inputs to the event models. Together, these operations drive cline, leading to the activation of the gating mechanism and
the event models into a new stable state. As the event models are updating of the event model.
updated, prediction error typically improves, and the influence of
sensory inputs on the event models diminishes. The signal path- A Partial Computational Implementation
ways into the event models can be thought of as controlled by a
gate that swings open in response to increases in error prediction Reynolds, Zacks, and Braver, (in press) have developed a neural
and then quickly swings shut again. Thus, the system alternates network simulation that implements the core features of EST:
between long periods of stability and brief periods of change. perceptual prediction, activation-based event models, and error-
Periods of stability are perceived by observers as events, and based gating of those models. The network is presented with
periods of change are perceived as the boundaries between events. animations of a human actor performing simple actions (e.g.,
The contents of event models are determined by a combination jumping jacks), represented as the three-dimensional location of 18
of bottom-up and top-down processing. When the event models’ points on the actor’s body (see Figure 2). At each timepoint, the
sensory inputs are transiently opened, they receive information network attempts to predict the actor’s body position at the next
from the current state of sensory and perceptual representations in timepoint. The network is trained on a corpus of stored events,
a bottom-up fashion. Event schemata affect event models in a such that each event is presented from start to finish, but each
top-down fashion. (We propose that the influence of event sche- event can be followed by any randomly chosen event from the
mata on event models is continuous and unaffected by the gating corpus. The network has pools of units corresponding to sensory
mechanism. However, this claim is based largely on parsimony inputs, perceptual processing, predicted future inputs, and event
and may need to be revised in the future.) At the same time that models, as described in EST (see Figure 1). (Event schemata are
schemata influence the current contents of the event models, the not implemented in the simulation.) The pathway from sensory
event models’ contents update the event schemata through a slow inputs to predicted future outputs is fully connected in a feed-
incremental learning process. forward fashion. The event model units have connections from
Event models are active and accessible representations of the sensory inputs that are gated; they open only at transient increases
events that are currently underway, yet the amount of information in prediction error. Simulations using this network provide support
contained in these models almost certainly exceeds what can be for the basic architecture of EST: Throughout training, boundaries
actively maintained in a limited capacity working memory system. between individual events are associated with larger prediction
The effective capacity of working memory can be augmented by errors than are within-event time points. Further, the network uses
the efficient use of previously stored knowledge representations. these transient spikes in prediction error to update stable represen-
Ericsson and Kintsch (1995) proposed the construct of long-term tations differentiating among possible events. These event repre-
working memory to describe these interactions. Information that is sentations improve performance on the prediction task. With ap-
A B C D
Figure 2. Graphical representation of the inputs and outputs to the Reynolds, Zacks, and Braver (in press)
neural network model. A–D: Four consecutive frames of an actor chopping down a tree. The target output
(dashed lines) on each frame is the model input (solid lines) on the subsequent frame. Frames C and D have
similar inputs but dissimilar target outputs, illustrating the need to represent necessitating the representation of
long-term sequential dependencies to achieve accurate prediction. From “A Computational Model of Event
Perception From Perceptual Prediction,” by J. R. Reynolds, J. M. Zacks, and T. S. Braver, in press, Cognitive
Science. Copyright 2006 by Cognitive Science Society. Reprinted with permission.
propriate gating, the event model representations can self-organize inputs to event models is inactivated, and the event models are
without any explicit labeling or categorization of the events. Thus, stable, which conserves resources. More intensive processing oc-
the proper gating and maintenance of event models aids percep- curs transiently when prediction error increases, and the event
tion. models are reset. This regulation of resources over time can be
viewed as a form of attention, focusing processing resources
Multiple Timescales adaptively at those moments when incoming sensory information
is most behaviorally relevant (Coull & Nobre, 1998; Nobre, 2001).
Note that the dish-scraping example and the Reynolds et al. (in In other words, event segmentation does not itself require atten-
press) network focus on one timescale in which activity becomes tion; rather, it implements a mechanism of attention. Second, event
less predictable at the end of the dish-scraping or at the end of the segmentation controls the updating of information in working
individual events in the network’s corpus. However, variations in memory by resetting the event models. The term cognitive control
predictability are to be expected on finer and coarser timescales as has been used to describe the control of attention and working
well. For example, it is likely that as each dish-scraping comes to memory in a variety of task domains (Cohen, Braver, & O’Reilly,
an end, there is a small, brief increase in prediction error and that 1996; Posner & Snyder, 1975). In our view, EST’s proposal that
as the dishwashing activity comes to an end, there is a larger, event segmentation controls resource allocation and updates mem-
longer increase in prediction error. We hypothesize that the archi- ory is a claim that event segmentation is a core, domain-general
tecture in EST is implemented simultaneously on a range of mechanism of cognitive control.
timescales, spanning from a few seconds to tens of minutes. For
each timescale, the error signal is integrated to provide a reset Related Approaches
signal tuned to the appropriate grain. Fine-grained representations
are tuned such that they can be updated in response to small, brief Our proposal that event models maintain stable representations
increases in prediction error. For coarse-grained representations, that influence perceptual processing and are updated in response to
the error signal is integrated with a longer time constant such that errors of prediction is similar to proposals of several neural net-
resets happen only in response to larger, more sustained increases work models. In these networks, transient updating based on
in error. failures of prediction is a means of balancing stability and flexi-
In addition to simultaneous parsing on multiple timescales, it is bility: Representations need to be stable across moment-to-
possible that event segmentation sometimes tracks simultaneous moment fluctuations in perceptual input but need to be updated
activities in parallel. For example, when attending a child’s birth- when they are no longer appropriate. These considerations have
day party, one might simultaneously segment the actions of chil- played a large role in the development of adaptive resonance
dren playing a birthday game and those of parents having a theory (ART; Grossberg, 1999; Carpenter & Grossberg, 2003). In
conversation at the same time. It is an empirical question whether ART networks, perceptual input is “cleaned up” by recurrent
such parallel processing occurs or whether observers are limited to interactions with a stable high-level representation. This high-level
processing one activity stream at a time. representation forms an abstraction of the perceptual input over
some period of time. When the cleanup process leads to large
distortions of the perceptual input, the current high-level represen-
Cognitive Control
tation is deemed no longer appropriate, and a search is initiated for
According to EST, the segmentation of activity into events a new high-level representation. Our proposal differs from the one
happens on an ongoing basis and plays two general and central instantiated in ART networks most significantly in that error
roles in regulating perception and cognition. First, event segmen- comparison is made between a predicted perceptual state and the
tation controls the allocation of cognitive resources over time. actual perceptual input. In ART networks, there is no intrinsic
During periods of low prediction error, the pathway from sensory orientation of the processing stream with respect to time.
Similar concerns motivate the architecture of several recent The gating mechanism resets event models and thus is the
models of prefrontal cortex (PFC). In these models, PFC repre- means by which the cognitive system exerts control over
sents current goals for action and the means to achieve them. Goal the disposition of processing resources and the updating
representations must be stable until the goal is achieved (or of working memory.
blocked) in order to be effective; however, once a goal is no longer
relevant, a new goal representation is desirable. In a broad theo- 3. Event segmentation happens simultaneously on multiple
retical review, Miller and Cohen (2001) suggested that updating of timescales, though an observer may attend to a particular
PFC representations could be gated by phasic dopamine signals timescale.
from midbrain dopamine neurons (Braver & Cohen, 2000), trig-
gered by encountering unexpected rewards; other models address 4. Event segmentation incorporates information from mul-
the possibility that unexpected lack of reward may also trigger the tiple senses. This results from the fact that the mecha-
updating of memory representations (O’Reilly, Noelle, Braver, & nisms in EST are general across sensory modalities and
Cohen, 2002; Rougier, Noelle, Braver, Cohen, & O’Reilly, 2005; incorporate information from multiple modalities.
Rougier & O’Reilly, 2002). EST differs from this family of models
5. Event segmentation depends on change. When the world

in two regards. First, stable representations contain information

is static, prediction is easy.
about the state of the world rather than goals and the means to
achieve them. Second, gating in this family model is based on 6. Event segmentation depends on prior knowledge. Event
failure to predict the reward value of a situation, whereas gating in models are constructed through the interaction of sensory
EST is based on failure of perceptual prediction. input with stored knowledge (including the knowledge
Finally, a model of frontal cortex proposed by Frank and col- stored in event schemata).
leagues also includes stable representations that are occasionally
updated through a gating mechanism (Frank, Loughry, & O’Reilly, In the following two sections, we review research on the cog-
2001; Frank, Seeberger, & O’Reilly, 2004). However, in this nitive and neural correlates of event segmentation with these
model, the gating mechanism is implemented by loops through implications in mind. The first section (Causes and Consequences
frontal cortex, the basal ganglia, and thalamus, with subcortical of Event Segmentation) describes behavioral data that provide
structures modulating the excitability of frontal cortex and thereby support for the model. The second section (Neural Correlates of
determining whether it is stable or plastic. This approach, like Event Segmentation) describes proposals for how the nervous
Miller and Cohen’s (2001) model and related proposals, focuses on system may implement the information-processing model and a
representations required for action rather than for perception. Un- review of the relevant neuropsychological and neurophysiological
like dopamine accounts, this model provides a natural mechanism data that motivate these proposals.
for selectively updating some representations without disrupting
others. However, in the context of event perception, a more global
signal might be more appropriate. In EST, we propose that the
Causes and Consequences of Event Segmentation
complete representation of an event at a given timescale is updated Perceptual Measures of Event Segmentation
based on a relatively global signal, such as could be provided by
the catecholamine neurotransmitters dopamine and norepineph- The first question that comes up in asking how people perceive
rine. By varying the time constant over which this signal is temporal structure in events is, How can one measure it? Newtson
integrated, selective updating of fine-grained or coarse-grained (1973) introduced a simple and surprisingly powerful solution to
event models can be accomplished. EST thus differs from the this problem, which he dubbed unitization (for reviews, see Newt-
model proposed by Frank and colleagues in three substantial ways: son, 1976; Stränger & Hommel, 1996; Zacks & Tversky, 2001).
first, by basing the gating mechanism on the failure of perceptual Participants are asked to watch movies of everyday activities (such
prediction rather than on reward; second, by characterizing per- as a person filling out a questionnaire) and to press a button
ception rather than action; and third, by adopting a less selective whenever they judge that a boundary between successive events
updating mechanism more compatible with implementation by has occurred. We use the term unitization to refer to this task and
neuromodulatory neurotransmitters than by corticothalamic loops. to distinguish it from event segmentation, which we hypothesize is
an ongoing perceptual process that is independent of any inten-
Implications tional task. When performing a unitization task, participants may
The principal novel features of EST are that event models be asked to identify the largest events they find meaningful
maintain stable representations of “what is happening now” and (coarse-grained unitization) or the smallest (fine-grained unitiza-
are updated based on transient increases in perceptual prediction tion). With a little bit of practice, people generally have no prob-
error. The theory has several implications for perception and lem following the instruction, and this simple task has produced
cognition: rich and replicable phenomena. By two measures, reliability of
unitization is good. First, participants show good agreement re-
1. Most important, the theory implies that the segmentation garding the location of event boundaries (Newtson, 1976). Second,
of ongoing activity into discrete events is a spontaneous the differences that do exist among observers can be attributed, in
concomitant of ongoing perception and does not require part, to stable individual differences rather than to noise: test–retest
conscious attention. studies have found good reliability both in the length of the units
people identify (Newtson, 1976) and in the particular locations
2. Event segmentation is a mechanism of cognitive control. they mark as event boundaries (Speer, Swallow, & Zacks, 2003).
The reliability of the unitization procedure provides support for grain, on different viewings. If viewers segmenting at a fine grain
EST’s claim that event segmentation is a spontaneous concomitant were spontaneously grouping fine-grained events into larger units,
of ongoing perception; the procedure corresponds intuitively to one would expect coarse-grained event boundaries to be a subset
replicable aspects of observers’ ongoing experience. of fine-grained event boundaries. As a result, one would expect
The original interest in event unitization concerned social cog- coarse and fine event boundaries to be aligned such that each
nition: How does the grain at which people segment an activity coarse boundary would be placed close to a fine boundary. This
affect observers’ attributions about why actors perform particular pattern has been observed in several studies (Hard et al., in press;
actions? Newtson (1973) asked participants to segment movies of Speer et al., 2003; Zacks, Tversky, & Iyer, 2001). One also would
a man filling out a questionnaire or building a model molecule into expect that coarse-grained events would tend to enclose a set of
either fine-grained or coarse-grained events and then to make fine-grained events, such that coarse boundaries would follow
judgments about the activity that had been performed. When rather than precede the fine boundary to which they were closest;
observers were asked to segment activity into fine-grained events, this has also been observed (Hard et al., 2006; Lozano et al., in
they were more likely to draw conclusions about actors’ permanent press). The presence of relations in segmentation that span time-
traits (e.g., personality characteristics). They also formed impres- scales indicates that the coarse-grained segmentation occurred
sions of the actors’ traits that were more differentiated and were even when participants were attending to fine-grained segmenta-
more confident in their judgments. This association between fine- tion.
grained units and dispositional attributions was supported correla-
tionally in one subsequent study (Wilder, 1978b) but not in another Consequences of Event Segmentation for Long-Term
(Wilder, 1978a). The grain at which observers segment events also Memory
affects the dispositional attributions observers make. In one study,
participants judged an actor as more likable if they had unitized her According to EST, working memory representations (the event
actions into fine-grained units rather than coarse-grained units models) are updated selectively at those points in time that corre-
(Lassiter, 1988). spond to perceptual event boundaries. This means that perceptual
If people form systematically different impressions when they information at those times receives more extensive processing than
attend to fine-grained than coarse-grained events, one reasonable perceptual information from other points in time. This extra pro-
possibility is that observers adaptively modulate the grain at which cessing should result in better long-term memory for this informa-
they segment activity in response to the needs of the situation. tion. Several converging measures suggest that perceptual infor-
Newtson (1976) proposed that fine-grained unitization is more mation from event boundaries is preferentially accessible in long-
resource demanding than coarse-grained unitization and that ob- term memory. In one experiment, participants watched a movie
servers unitize at the coarsest grain they can sustain while main- and then saw still pictures from event boundaries, points in be-
taining a coherent representation of the ongoing activity. Activity tween boundaries, and from a similar movie they had not seen.
that is relatively coherent and predictable can be parsed at a coarse They then reported which pictures came from the movie they had
grain, whereas activity that is confusing or surprising must be watched (Newtson & Engquist, 1976). Accuracy was higher for
parsed at a finer grain. This assumes that observers can perceive pictures taken from event boundaries. Another study found that
boundaries only at one grain at any time, in which case it is descriptions of event boundaries from memory were richer and
intuitive that identifying few units would be less resource demand- more detailed than descriptions of nonboundaries (Schwan, Gar-
ing than identifying many units. Evidence for this proposal has soffky, & Hesse, 2000). Also, the overall amount of information
accrued from studies in which a single surprising action was that can be recalled is affected by the grain of segmentation:
inserted into an otherwise predictable activity (Newtson, 1973) and Multiple studies have reported that fine-grained unitization of
from studies in which the overall predictability of the activity was movies leads to more detailed recall of the events depicted than
manipulated more systematically (Wilder, 1978a, 1978b). does coarse-grained unitization (C. Hanson & Hirst, 1989; Las-
EST provides an alternative account of these findings. The siter, 1988; Lassiter, Stone, & Rogers, 1988). However, it is
theory asserts that people do not perceive event boundaries on only currently a matter of debate whether fine-grained unitization also
one timescale. Rather, they perceive event boundaries on multiple improves recognition memory for depictions of events (C. Hanson
timescales simultaneously but selectively attend to one timescale & Hirst, 1991; Lassiter & Slaw, 1991). These effects are consistent
in response to instructions or other experimental manipulations. with the view that observers can selectively attend to one timescale
According to this view, when activity is coherent, participants while segmenting simultaneously at several timescales.
segment it at multiple timescales and can choose to attend to finer According to the theory, when the surface structure of an event
or coarser grains. Attending to coarser grains may be preferred depiction aligns with its underlying event structure, the gating
because it reduces the frequency of decision making and button mechanism should operate efficiently, resulting in enhanced long-
pressing. However, when activity is less coherent, coarse-grained term recall for the events. Conversely, surface cues that conflict
event segmentation may break down because prediction error will with the event structure of an activity may lead to poorer memory.
be uniformly high on coarse timescales, leaving only fine-grained Three studies have used film editing techniques to test these
segmentation intact. Data from experiments in which people seg- hypotheses. In the first study (Schwan & Garsoffky, 2004), par-
ment the same activities multiple times at different timescales ticipants viewed short movies of everyday events and then recalled
support this hypothesis (Hard, Lozano, & Tversky, 2006; Hard, them. The movies were presented either intact, with deletions that
Tversky, & Lang, in press; Lozano, Hard, & Tversky, in press; corresponded to intervals surrounding event boundaries, or with
Speer et al., 2003; Zacks, Tversky, & Iyer, 2001). In these exper- deletions of intervals in between event boundaries. Memory for the
iments, participants segmented activities at both a coarse and a fine edited movies with preserved event boundaries was as good as
memory for intact movies, but memory for edited movies with pants’ descriptions suggested that the benefit of hierarchical en-
deleted intervals around event boundaries was poorer. The second coding was mediated by a tendency for those participants who
study examined longer events and manipulated cues to event segmented hierarchically to simulate the events from the actor’s
structure by marking the event boundaries rather than deleting point of view. A subsequent series of experiments experimentally
actions (Boltz, 1992). In this series of experiments, participants manipulated perspective taking and supported the hypothesis that
viewed a feature film with no commercial breaks or a film with hierarchical encoding facilitated simulating the actor’s perspective
commercial breaks. The breaks corresponded to or conflicted with (Lozano et al., in press).
event boundaries. Commercials at event boundaries improved later In a final procedural learning study (Zacks & Tversky, 2003),
recall of the activity, whereas commercials at nonboundaries im- participants learned about an everyday procedure, such as putting
paired memory. Commercial breaks at event boundaries also im- together a musical instrument using a computer program. The
proved memory for the temporal order of events. One other study program was designed to visually depict a set of presupposed
(Schwan et al., 2000) manipulated the placement of cuts between coarse-grained event boundaries in the procedure or to conflict
different camera positions in short movies. Cuts at event bound- with those boundaries. In one experiment, event boundaries were
aries improved memory for those points in time but had little effect identified by asking a separate group of participants to segment a
on overall memory performance. It is notable that the placement of movie of the activity to be taught. In this experiment, visual
cuts in the Schwan et al. (2000) study produced relatively weak structure that coincided with these boundaries improved memory
effects compared with the placement of commercials in the Boltz for the order of events. In another experiment, event boundaries
study. One possibility is that simple cuts are not sufficiently salient were taken from manufacturer’s instructions, which may have
to affect the operation of the gating mechanism—at least for conflicted with perceptually natural event boundaries. In this ex-
participants used to viewing movies and television. Another pos- periment, reinforcing the event boundaries reduced memory for
sibility is that the effects of event segmentation on memory are the order of events. These results suggest that memory is facilitated
more pronounced at longer timescales. Together, these results when surface cues support the intrinsic event structure of an
converge in supporting a role for event segmentation in long-term activity, and memory is impaired when surface structure conflicts
memory encoding. with the intrinsic event structure.
Patterns of individual differences suggest a strong relation be- Together, studies of long-term memory for events and studies of
tween one’s ability to segment activity during learning and one’s learning procedures from events suggest that event boundaries are
ability to recall it later. In one study (Zacks, Speer, Vettel, & used to structure memory encoding. The fact that event boundaries
Jacoby, 2006), older adults with and without dementia of the are remembered better than nonboundaries supports EST’s claim
Alzheimer’s type (DAT) segmented movies of everyday events that event segmentation is a cognitive control mechanism. Further
and then performed a recognition memory test for visual details support for this claim is that individuals who are good at segment-
from the movies. An individual’s ability to properly segment the ing events remember them better than do individuals who are poor
movies was evaluated by comparing her or his segmentation with at segmenting. Moreover, cuing the appropriate event structure
the segmentation of the group as a whole. Group-typical segmen- facilitates memory, whereas miscuing event structure impairs
tation was found to be a unique predictor of later memory, even memory. Finally, people consistently segment activity hierarchi-
after the presence of dementia and overall cognitive level were cally, making connections that span timescales. This supports the
controlled. This is consistent with EST’s proposal that event claim that event segmentation proceeds simultaneously on multi-
boundaries receive richer processing and thus that identifying the ple timescales.
proper boundaries results in more effective encoding for long-term
memory. Features That Correlate With Perceptual Event
The theory’s proposal that event boundaries receive differential Segmentation
processing during perception has implications for the ability to use
perceptual experiences when one learns new skills. The ability to According to EST, event segmentation depends on changes in
segment an activity into the right units should be valuable in the environment and on prior knowledge. Physical changes in the
learning how to perform the actions that occur during the activity. environment can drive event segmentation in a bottom-up fashion
Several studies of procedural learning from events support this by increasing the potential for prediction error. When sensory
hypothesis. In one series of experiments, participants learned from inputs are constant, a well-tuned perceptual prediction system will
a movie how to assemble a TV cart or a construction-block model adhere to the old adage about weather forecasting—“tomorrow
(Hard et al., 2006). During the learning phase, they segmented the will be the same as today”—and will be correct in this prediction.
movie into events; afterwards, they built the TV cart or the model. When changes occur there is more opportunity for error. At the
A number of features of the situation were varied: whether the same time, prior knowledge can influence event segmentation in a
participants segmented at a coarse grain, a fine grain, or both (and, top-down fashion by biasing the system’s interpretation of what a
if so, in what order); whether the participants described the activity change portends. One type of prior knowledge can be tied to
while segmenting; and whether they were explicitly instructed to domain-specific reasoning about goals and plans. When watching
group fine-grained events into larger structures. Across experimen- goal-directed human activity, observers may infer an actor’s goal
tal conditions, a consistent pattern was observed: Those individu- on the basis of past experience or explicit instructions and use this
als whose segmentation was more hierarchically structured were information to bias perceptual prediction. For example, baseball
better at assembling the TV cart than those whose segmentation fans can anticipate events that might come as a surprise to baseball
was less hierarchically structured. This was true across experimen- novices (e.g., all the players running off the field at the end of an
tal conditions and across individuals within conditions. Partici- inning), and this is likely to influence event segmentation. Another
type of prior knowledge involves domain-general statistical learn- quences taken from these movies with brief tones placed either at
ing. Humans have powerful mechanisms for learning sequential points when goals were completed or at midpoints between goal
dependencies in streams of information, which can influence mo- completions. After each sequence, they were asked to watch it
tor performance (Seger, 1994) and language learning (e.g., Brent, again and press a button to mark where the tone had occurred.
1999; Saffran, Aslin, & Newport, 1996). Similar mechanisms may Tones were remembered more accurately when they were placed
guide perceptual prediction. According to the model, prior knowl- at goal completions, whereas midpoint tones were remembered as
edge about goals and plans and prior statistical learning are en- having occurred closer to completions than they actually did.
coded in event schemata. A small body of evidence indicates that These effects on memory suggest that participants perceived the
both physical changes and prior knowledge are correlated with the activity in terms of intention-based units, and this affected how the
perception of event boundaries. locations of tones were stored during the initial encoding phase.
One distinctive physical change that correlates with the percep- However, an alternative explanation is that knowledge structures
tion of event boundaries is movement. In one study, participants representing goals biased performance only during the retrieval
unitized brief movies, and the experimenters identified those phase of the task. More direct evidence that goal completions are
points that most observers agreed were event boundaries (Newt- perceived as event boundaries comes from a second study, con-
son, Engquist, & Bois, 1977). The movies then were coded using ducted with 10 –11-month-old infants. In this experiment, the
a dance notation that provided a discrete representation of which infants were repeatedly exposed to a sequence taken from a movie
actor joints had changed position by more than 45° during each 1-s of an everyday activity. They then were tested with one of two
interval. The number of joints that changed between successive altered versions of the sequence. In one version, a pause was
intervals is a rough index of the amount of motion at that point in inserted at a moment that had been identified as the completion of
the movie. The number of changing joints was averaged at each a goal; in the other, a pause was inserted in the middle of two
grain for transitions into and out of event boundaries and for completions. The experimenters hypothesized that if the infants
successive intervals within an event. Transitions into and out of encoded the activity in terms of the actor’s intentions, a pause in
event boundaries had statistically greater numbers of changes (i.e., the middle of accomplishing a goal would be more surprising than
more movement) than did successive within-event intervals. Con- a pause at the end of completing a goal. Thus, they should look
verging evidence for a relation between movement features and longer at the version with a pause in the middle. This was exactly
segmentation comes from a recent study in which participants what the experimenters found: Infants looked longer at the altered
segmented animations of geometric objects playing simple games sequences when those sequences contained a pause that interrupted
(Hard et al., in press). The animations were coded by eye for an ongoing goal compared with when the pause occurred at a goal
movement change such as stops, direction changes, and changes in completion.
speed. These movement changes were associated with increases in A recent study provided evidence that both physical changes
both fine-grained and coarse-grained segmentation. and goals are systematically related to event segmentation (Zacks,
A pair of studies (summarized in Baird & Baldwin, 2001) 2004). In three experiments, participants viewed animations that
suggests that event segmentation is correlated with actors’ goals showed the movements of a pair of objects (see Figure 3A) and
and plans. In one study, one group of participants was asked to unitized them to mark coarse-grained or fine-grained units. Two
segment movies of everyday activities based on when the actor types of animation were shown: Goal-directed activity stimuli
completed a goal. A second group of participants watched se- were generated by recording the actions of two people controlling
Figure 3. A: Trace of two randomly moving objects. From the time-varying x and y locations of the object, an
exhaustive set of movement features was calculated, which included the objects’ position, speed, and acceler-
ation, the distance between the objects, and their relative speed and relative acceleration. B: Movement features
were quite predictive of where observers segmented the activity when the activity was randomly generated and
was unitized at a fine grain. The dependence of unitization on movement features was reduced for goal-directed
human activity and for coarse-grained unitization. Adapted from “Using Movement to Understand Simple
Events,” by J. M. Zacks, 2004, Cognitive Science, 28, pp. 979 –1008. Copyright 2006 by Cognitive Science
Society. Adapted with permission.
the objects in a simple video game, and random activity stimuli event boundaries are correlated with goals and with sequential
were generated by a random process designed to match the veloc- statistical structure indicates that event segmentation depends on
ity and acceleration of the video game stimuli. Rather than hand prior experience.
coding movement features of interest to the experimenters (Hard et
al., in press; Newtson et al., 1977), a quantitative movement
analysis was performed. For each animation, an exhaustive set of Causes and Consequences of Segmentation in Story
movement features was calculated, including the position, veloc- Understanding
ity, and acceleration of each object, the distance between the
objects, their relative velocity, and their relative acceleration. For Narrative stories depict events through a medium that is surely
each experimental condition, the relation between the movement not equivalent to the actual experience of those events; however,
features and the probability that a participant would identify an there is good reason to think that narrative comprehension may
event boundary was characterized using linear models. Movement share representations and processes with the comprehension of live
features were significantly related to event unitization for all events. Recent theories of narrative comprehension claim that
conditions, providing evidence that distinctive physical features readers and listeners mentally simulate an experience based on the
description provided in the text (e.g., Glenberg & Kaschak, 2002;

play a role in event segmentation. Two further results suggested

that the processing of movement information interacted with goal Zwaan, Stanfield, & Yaxley, 2002). From a researcher’s point of
processing to determine event segmentation. First, the relation view, narratives offer opportunities for quantification and control
between movement features and unitization was consistently stron- over some features that may be important for event segmentation,
ger for fine-grained unitization than coarse-grained unitization (see so the available data provide important insights into event under-
Figure 3B). This suggests that movement features may play a standing.
particularly strong role in identifying the smallest units of activity In narrative comprehension, the effects of prior knowledge on
but that other features may be important in identifying which of segmentation have been studied under the rubrics of schemata
those low-level event boundaries are also boundaries between (Rumelhart, 1975), scripts (Schank & Abelson, 1977), and situa-
larger units of activity. Second, the relation between movement tion models (van Dijk & Kintsch, 1983). At least since Bartlett’s
features and unitization was weaker for goal-directed activity than (1932) classic studies in the early 20th century, experimental
for random activity. This suggests that prior knowledge may play psychologists have argued that readers understand narratives by
a role in modulating how physical characteristics are processed in constructing representations of the events described in the text and
order to identify event boundaries. later remember these constructions rather than simply remember-
Knowledge about actors’ goals is one source of prior knowledge ing the text itself. Schemata, scripts, and situation models differ in
that can affect event segmentation. Simple statistical information is many particulars, but they share a common assumption: Readers
another complementary source of prior knowledge, which is also use prior knowledge to segment a narrative into discrete events.
related to the perceptual segmentation of events. One series of Prior knowledge may include information about actors’ goals and
studies indicated that similar mechanisms may play a role in the purely conventional (statistical) sequential dependencies. Our use
identification of events and their boundaries (Avrahami & Kareev, of the term event schema is closely related to the usage described
1994). In three experiments, participants viewed movies made by here, and our use of the term event model is analogous to the term
concatenating short clips from cartoon films. In the first experi- situation model used in the narrative comprehension literature.
ment, the experimenters constructed a sequence of clips that in- The term schema was originally introduced in neurology to
cluded a perceptually salient change. These changes were identi- describe the coordination of movement in goal-directed action and
fied by untrained participants as a part boundary. A second group adopted by Bartlett (1932) to describe structured representations of
of participants viewed a stream of clips that contained repeated the actions that typically happen in stories. The notion of schema
presentations of the sequence. They were then shown the sequence in cognitive psychology was developed more fully by Rumelhart
in a short stream of clips and were asked where they thought it and colleagues (e.g., Rumelhart, 1980). Rumelhart (1977) argued
should be divided. These participants placed the boundary not at that stories consist of episodes; in each episode, a protagonist is
the salient change but at the end of the sequence. A second confronted with a situation that causes him or her to desire some
experiment showed that participants were more likely to recognize goal, and the protagonist tries to achieve that goal. Thus, each
a sequence that repeated throughout a movie if the clips appearing episode is interpreted in terms of a schema for trying to achieve a
before and after the sequence were varied. A final experiment goal. This schema has a recursive structure, such that components
showed a similar effect using a recall paradigm. These data suggest of trying to achieve a goal (e.g., buying ice cream) include sub-
that the repetition of an arbitrary sequence across different con- goals that the protagonist tries to achieve in order to fulfill the
texts is sufficient for it to be conceived of as a coherent unit. larger goal (e.g., getting money). When comprehending a story,
However, it is important to note that in all three experiments the readers or listeners encode it in terms of a series of events that are
tasks depended heavily on memory, so the extent to which these segmented and hierarchically arranged in accord with this goal-
data speak to perception, as such, is not clear. based structure. According to schema theory, summarizing a story
To summarize, studies of the unitization of events provide consists of pruning low levels in the hierarchical representation.
evidence consistent with EST’s claims that event segmentation The theory also predicts that schematic representation can lead to
depends on change and that event segmentation depends on prior two kinds of distortion in delayed recall. First, pruning can occur
knowledge. The fact that event boundaries are correlated with because high levels in the schema are better represented than low
changes in movement features in two quite different paradigms levels, leading recall to increasingly resemble summarization. Sec-
indicates that event segmentation depends on change. The fact that ond, distortions can occur that normalize recall to the participant’s
schema for that type of activity. Experimental tests of these pro- relevant, the causes of events, and the intentions of protagonists. A
posals supported the theory (Rumelhart, 1977). change in any dimension requires that the reader update the current
The term script was introduced in computer science and psy- model, deactivating one representation and activating a new one or
chology by Schank and Abelson (1977), who used it to refer to a reactivating a previous representation. Considering the conceptual
structured representation of an activity that has predictable rela- similarity between situation models and event models, it is a small
tions among settings, actors, props, and actions. As with event step to propose that readers segment a narrative into events based
schemata, scripts are knowledge structures with a nested organi- on changes in these features. According to the theory proposed
zation such that the contents of a script include other knowledge here, features such as time, space, objects, and characters affect
structures. A script specifies a list of the events that are typically segmentation in narrative through bottom-up processing. How-
part of the activity represented by the script. For example, a script ever, according to the theory, features such as causes and inten-
for visiting the doctor might include signing in with the reception- tions depend also on inferences based on prior knowledge.
ist, filling out forms, waiting in the waiting room, and being Until recently, there was little evidence for the proposal that
examined by the doctor. Scripts may include information about the changes in the dimensions of events, as proposed by the event
order in which these events occur (Abelson, 1981). Script theory indexing model, are perceived by readers as boundaries between
motivated a number of studies of reading comprehension and events. Two sets of recent experiments provide strong support for
memory. One set focused on readers’ knowledge about the every- this proposal. One series of studies (Speer, Zacks, & Reynolds,
day activities that might be represented by scripts (Bower, Black, 2006) took a correlational approach, using narratives from preex-
& Turner, 1979). These experiments established that people agree isting descriptions of a child’s activities over the course of a day
about the typical actions performed in common everyday activities (Barker & Wright, 1966). Each story was divided into clauses
and about the boundaries between events in a script-based narra- (defined as a verb plus its argument structure). Raters coded each
tive. Moreover, after reading a story, people remember the events clause for changes in the five dimensions of situation models
and their order as having been more similar to the standard script described previously. Readers then segmented the activity in the
for the activity than was actually described in the story. Another stories. Readers tended to identify event boundaries at those
set of studies provided evidence for the hierarchical organization clauses in which one of the situation model dimensions was
of stories in memory (Abbott, Black, & Smith, 1985). Participants changing. The probability of identifying an event boundary in-
read stories based on scripts, and their memories were tested. In creased parametrically with the number of dimensions that
one experiment, reading about low-level (fine-grained) events changed. These findings support the model’s claim that event
primed retrieval of high-level (coarse-grained) events but not vice segmentation depends on change. In line with the model’s claim
versa. In another experiment, participants received sentences that that prediction becomes more difficult at event boundaries, clauses
were not in the story but could be inferred from it. They were more following changes in a dimension were rated as being less pre-
likely to report that these sentences, which were never presented, dictable than other clauses. A second series of studies used narra-
had occurred in the story if they corresponded to high-level events tives in which shifts in time were manipulated using temporal
than if they corresponded to low-level events. This suggests that references (Speer & Zacks, 2005). An example is provided in
the high-level information had been inferred on the basis of the Figure 4. Participants were asked to segment the activity in these
underlying hierarchical representation. Similar effects have been stories using the unitization procedure described previously. Sen-
obtained with filmed narratives, including direct comparisons be- tences containing a temporal reference were likely to be identified
tween narratives presented as texts and as films (Brewer & Dupree, as the onsets of new events, and this was especially true when the
1983; Lichtenstein & Brewer, 1980). In sum, research on event temporal reference indicated a long interval (an hour as opposed to
schemata and scripts supports the view that event segmentation in a moment).
narrative is correlated with inferences about the goals of actors and According to the theory, the perception of an event boundary
with statistical dependencies between events. leads to a cascade of processing that resets the event models’
Several current theories of narrative comprehension claim that contents. Thus, information presented prior to an event boundary
understanding a story involves the construction of a situation
model. Unlike event schemata and scripts, which are representa-
tions of classes of events, situation models are representations of …
particular events that are described in a narrative. These theories She had just bought a new camera, and she hoped
Introduction
build on Johnson-Laird’s (1989) characterization of mental models the pictures would turn out well.
as simulations of actual situations and on the reading comprehen- She could hear water running, and figured there
Object
sion models of Kintsch and colleagues (e.g., Kintsch, 1994), who must be a creek nearby.
introduced the term situation model into the literature. One theory Time shift A moment later/An hour later , she was collecting
of comprehension that incorporates a situation model representa- wood for a fire.
tion, called the event indexing model, is particularly helpful for Anaphor She heard a noise near the stream.
thinking about the relation between event perception and narrative …
understanding (Zwaan, 1999; Zwaan & Radvansky, 1998; Zwaan,
Figure 4. Excerpt from a narrative describing a woman on a backpacking
Magliano, & Graesser, 1995). The event indexing model proposes trip. Sequences such as this were embedded throughout the narratives. A
that readers and listeners construct a model that represents the critical object (in bold) is introduced, followed by a sentence that contains
situation described in a text. To do so, they track indices repre- either a time shift (an hour later) or a control phrase (a moment later).
senting five dimensions of events: the current time, the current Finally an anaphor (underlined) refers back to the critical object. (See
location in space, the objects and characters currently present and Speer & Zacks, 2005.)
should be difficult to retrieve following that boundary. In the Speer patients with disorders of event understanding and action planning
and Zacks (2005) study, this hypothesis was tested by measuring and the few available neuroimaging studies of event understanding
reading time and recognition memory (see Figure 5). Reading provide surprisingly strong support for the theory proposed here.
times for sentences containing a shift in narrative time were longer As will be seen in this section, neuroimaging data support the
than nearly identical control sentences, consistent with the hypoth- theory’s claims that event segmentation is a spontaneous concom-
esis that increased processing occurred. References to information itant of ongoing perception, happens simultaneously on multiple
presented just prior to a temporal reference were read more slowly timescales, and depends on change. Patient data support the the-
when the temporal reference indicated a time shift than when no ory’s claims that event segmentation incorporates information
time shift occurred. This provides indirect support for the hypoth- from multiple senses, depends on prior knowledge, and is a mech-
esis that memory is diminished following an event boundary. anism of cognitive control. Patient studies and neuroimaging data
Direct support for this hypothesis came from an experiment in also provide critical information about how the brain may imple-
which participants’ recognition memory was tested immediately ment the information-processing theory proposed here. In this
after they read a temporal reference sentence. Information pre- section, we propose a tentative mapping between the processing
sented just before the temporal reference was recalled less well components described earlier (see the A Theory of Event Segmen-
when the temporal reference indicated a time shift than when no tation section) and review the relevant neurophysiological data
time shift occurred (see also Bower & Rinck, 2001; Rinck & with this mapping in mind. In particular, we focus on the possible
Bower, 2000; Zwaan, 1996; and see Zwaan & Radvansky, 1998, neural substrates of the unique components of the theory, event
for a review of related effects). Together, these results support the models and event schemata.
theory’s claim that event segmentation is a form of cognitive control,
modulating cognitive processing and memory maintenance. Sensory Inputs and Perceptual Processing
In sum, research on narrative comprehension provides support
for four implications of the theory. First, the fact that reading or Figure 6 reproduces Figure 1, annotated with the brain regions
listening to stories leads to robust event segmentation supports the that are hypothesized to correspond to each of the computational
implication that event segmentation is multisensory—it does not units. Recall that the core of the segmentation mechanism in EST
depend on specific visual or auditory features present in live is the perceptual processing stream illustrated in the left of the
activity. Second, event segmentation depends on changes, partic- figure. Early sensory representations are transformed in a percep-
ularly on changes in dimensions identified by the event indexing tual processing stream leading to representations that predict the
model (Zwaan, 1999). Third, event segmentation depends on prior state of the world a short time hence. The early-stage representa-
knowledge, particularly knowledge about goals and statistical de- tions correspond to the outputs of primary sensory areas, including
pendencies. Finally, event segmentation is a mechanism of cogni- primary auditory cortex (A1), primary visual cortex (V1), and
tive control, regulating the contents of memory. primary somatosensory cortex (S1). The nature of these represen-
tations is relatively well understood, particularly for V1. Repre-
sentations in these areas are topographically organized maps of the
Neural Correlates of Event Segmentation
distal environment, representing its spatial structure as well as
Research examining the neural mechanisms by which events are modality-specific features (Kolb & Whishaw, 2003). In the case of
perceived and conceived is still in its infancy. However, studies of vision, the midstage representations are also fairly well under-
Figure 5. The left panel shows that readers were slower to process sentences if they contained a time shift (an
hour later) than a control phrase (a moment later). As is seen in the middle panel, they were also slower to read
a following sentence that contained a reference to a previously mentioned object. Finally, the right panel shows
that after reading a time shift sentence, critical objects were correctly recognized less often (see Speer & Zacks,
2005).
Error Detection
SN, VTA, LC
Predicted Future Inputs

ACC, ...
Perceptual Processing Event Models Event Schemata

IT, MT+, pSTS, ... Lateral PFC
Sensory Inputs
A1, V1, S1, ...
Figure 6. Schematic depiction of the model, with hypotheses about the neurophysiological structures corre-
sponding to the different components of the model. Thin gray arrows indicate the flow of information between
processing areas, which are proposed to be due to long-range excitatory projections. Dashed lines indicate
projections that lead to the resetting of event models. PFC ⫽ prefrontal cortex; IT ⫽ inferotemporal cortex;
MT⫹ ⫽ human MT complex; pSTS ⫽ posterior superior temporal sulcus, ACC ⫽ anterior cingulate cortex,
SN ⫽ substantia nigra, VTA ⫽ ventral tegmental area, LC ⫽ locus ceruleus; A1 ⫽ primary auditory cortex; S1
⫽ primary somatosensory cortex; V1 ⫽ primary visual cortex.
stood. After V1, visual processing is segregated into a dorsal and established that these areas, which previously had been explored
ventral stream, though there is some communication between the using highly controlled picture judgment tasks, responded selec-
two (Felleman & Van Essen, 1991; Ungerleider & Mishkin, 1978). tively in the context of realistic dynamic movies (see also Bartels
Regions within these streams represent specialized aspects of the & Zeki, 2004).
visual world (and also integrate some information from other Studies of neural processing during the passive viewing of event
modalities via recurrent projections). Regions in inferotemporal boundaries specifically address the theory’s claim that event seg-
cortex in the ventral stream respond selectively to properties of mentation controls the deployment of processing resources over
objects, including properties that are invariant over changes in time. In one experiment (Zacks, Braver, et al., 2001), participants
orientation, lighting, and so forth (Tanaka, 1996). Regions in the passively viewed movies of everyday events while brain activity
dorsal stream corresponding to areas MT and MST in the monkey, was recorded with fMRI. Following passive viewing, they watched
dubbed “the MT complex” (MT⫹), selectively represent features the movies again, this time segmenting the activity by pressing a
of object and observer movement (Tootell et al., 1995). Adjacent button to mark boundaries between events. A network of brain
areas in the posterior superior temporal sulcus are selectively regions showed transient increases in activity at perceptual event
activated by nonrigid biological motion (Beauchamp, Lee, Haxby, boundaries during passive viewing (see Figure 7). In other words,
& Martin, 2003; Grossman et al., 2000). These perceptual process- the activity of regions in this network correlated with participants’
ing streams are oriented in time such that they not only represent later identification of segment boundaries. Evoked responses were
the current state of the world but also represent predictions about greater for coarse-event boundaries than for fine-event boundaries,
what is likely to happen a short time later. consistent with the behavioral finding that participants encode
Two lines of research directly address the neurophysiology of activity hierarchically (Zacks, Tversky, & Iyer, 2001). Activated
perceptual processing during ongoing activity. In one study, par- areas included right posterior frontal cortex (Brodmann’s area
ticipants passively viewed 30 min of a Hollywood movie while [BA] 6) and a collection of regions in extrastriate visual cortex,
whole-brain activity was recorded with functional MRI (fMRI; including temporal, occipital, and parietal areas (BAs 19, 31, 37,
Hasson, Nir, Levy, Fuhrmann, & Malach, 2004). The fMRI data 39). In particular, the strongest activity was identified near the
were warped into a common atlas space and analyzed to identify MT⫹, an area in posterior visual cortex that is selectively respon-
regions in which the brain responses were similar across observers. sive to visual motion. A second study directly localized MT⫹ in
The areas identified included a region in the inferior temporal lobe individual participants who also completed the event viewing and
(fusiform gyrus) that had previously been found to be sensitive to segmentation tasks (Speer et al., 2003). Evoked responses in the
the presentation of faces; it responded when close-ups of faces individually identified MT⫹ regions were large and statistically
were on the screen. Another temporal region (parahippocampal reliable. Another study (Zacks, Swallow, Vettel, & McAvoy,
gyrus) that had previously been found to be sensitive to pictures of 2006) localized MT⫹ in a set of observers who then went on to
buildings and spaces responded selectively when establishing watch simple two-object animations of the sort described previ-
shots or other wide-angle views were presented. This paradigm ously (see the Features That Correlate With Perceptual Event
A 0.7
Passive viewing
0.7
Segmentation
0.5 0.5
0.3 0.3
Coarse Coarse
0.1 0.1 Fine
Fine
-0.1 -0.1
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
B
0.7
Percent signal change
0.7
0.5 0.5
Coarse
0.3 Coarse 0.3
Fine
0.1 0.1
Fine
-0.1 -0.1
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
C
0.7 0.7
Coarse
0.5 0.5
0.3 0.3
0.1 Coarse 0.1

Fine
Fine
-0.1 -0.1
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
Time relative to boundary (s) Time relative to boundary (s)
Figure 7. Focal brain activity in three regions, showing transient changes at event segment boundaries
(identified by vertical lines). A–C: The bilateral extrastriate and right frontal clusters of activated voxels. All
regions showed reliable responses to event boundaries during passive viewing and larger responses during active
segmentation. The left image shows the extent of the cluster, superimposed on an averaged anatomical image for
the 16 participants. The two graphs to the right show the evoked response during coarse and fine event units for
passive viewing and active segmentation (see Zacks, Braver, et al., 2001).
Segmentation section). MT⫹ responded selectively when the ob- see the Causes and Consequences of Segmentation in Story Un-
jects were moving faster, and also selectively responded to event derstanding section, above) while brain activity was measured
boundaries. These data indicate that extrastriate visual areas and with fMRI. After scanning, participants re-read the narratives and
right frontal cortex either perform computations that contribute to segmented them into events. The results indicated that neural
the detection of event boundaries or are up-regulated as part of systems that are specialized for processing particular features of
boundary detection. One possibility is that the activity in extrastri- live action were transiently activated when reading about changes
ate visual cortex reflects the transient opening of event models to in those features. For example, reading that a character began
sensory and perceptual information. interacting with a new object activated readers’ somatomotor cor-
One recent study used a narrative reading paradigm to look at tex, whereas reading that the primary character in a story moved
the relation between online processing of event features and pro- from one place to another activated the medial temporal cortex,
cessing of event boundaries (Speer, Reynolds, Swallow, & Zacks, near regions activated during spatial changes in navigation in
2006). Participants read extended narratives that had been coded virtual reality (Burgess, Maguire, & O’Keefe, 2002; Shelton &
for changes in event dimensions (Speer, Zacks, & Reynolds, 2006; Gabrieli, 2002). This indicates that changes in the narrated situa-
tion were being processed in real time. Further, the data suggested Strick, 1996). We believe that participants’ identification of event
that the processing of these changes led to the perception of event boundaries corresponds to the resetting of event models when they
boundaries: An overlapping network, including posterior parietal, perform event segmentation tasks. This account is consistent with
right anterior temporal, and frontal cortex transiently increased in a recent theory of locus ceruleus norepinephrine function, which
activity at event boundaries, and these increases were fully medi- proposes that norepinephrine serves as a general reset signal,
ated by activity associated with event changes. This pattern of allowing neural networks to move out of one stable state and settle
results supports EST’s claim that changes lead to transient in- into a new stable state on the basis of the current network input
creases in prediction error, which in turn lead to the detection of (Bouret & Sara, 2005). We hypothesize that this reset is approx-
event boundaries and the updating of event models. imately hierarchically structured, such that resets of representa-
tions with longer timescales are generally a subset of resets of
Perceptual Prediction and Error Detection representations with shorter timescales. This could be imple-
mented by evaluating prediction error at a range of timescales in
EST claims that perceptual predictions are constantly compared the ACC, with subcomponents of the ACC projecting to distinct
with actual sensory input, providing an evaluation of how well targets. Alternatively, ACC could signal only prediction errors on
perception is functioning. As indicated in Figure 6, we hypothesize short timescales, and computations within lateral PFC could com-
that predicted future inputs are represented in the anterior cingulate pute which of these fine-grained breaks are also coarse-grained
cortex (ACC) and that the error detection function is implemented breaks.
by neuromodulatory nuclei in the midbrain—the substantia nigra, There are a small number of neurophysiological data that bear
ventral tegmental area, and locus ceruleus. We begin by reviewing on the relation between error detection and event segmentation.
data that bear on the error detection mechanism and then work These include studies using electroencephalography (EEG)
back to the perceptual prediction representations. evoked-response potential (ERP) methods in combination with
Several mechanisms have been identified that could compute sentence-processing paradigms. In sentence processing, an ERP
prediction error in event-structure perception (Schultz & Dickin- component called the N400 has been associated with the process-
son, 2000). Increases in prediction error lead to a cascade of ing of local breakdowns in predictability (Kutas & Hillyard, 1980).
processing that has been characterized as an orienting response One study of narrative comprehension contrasted sentence-level
(Sokolov, Spinks, Naeaetaenen, & Lyytinen, 2002). According to breakdowns in predictability with discourse-level breakdowns by
the theory, the resetting of event models and the transient increase presenting participants with sentences that were locally coherent
in sensitivity to sensory input are two components of that response. but whose meaning conflicted with the larger narrative (van Ber-
We hypothesize that this resetting is implemented by midbrain kum, Hagoort, & Brown, 1999). These breakdowns in narrative
neuromodulatory systems. These systems appear to broadcast error predictability produced N400 responses that were very similar to
signals—including prediction errors—through widespread projec- those produced by sentence-level breakdowns. A follow-up study
tions to the cortex. In particular, circuits based on dopamine and found similar results with auditory presentation rather than with
norepinephrine have received substantial attention. Dopamine neu- reading (van Berkum, Zwitserlood, Hagoort, & Brown, 2003).
rons in the substantia nigra and ventral tegmental area are sensitive Breakdowns in predictability also have been studied with pictures
to differences between actual and predicted rewards (Schultz, and movies. In one study, participants viewed a series of grayscale
1998). Norepinephrine neurons in the locus ceruleus appear to images that told a brief story, with the final image being either
track performance in attention-demanding tasks and have been congruent or incongruent with the preceding ones (West & Hol-
proposed to regulate the sensitivity of an organism to external comb, 2002). Incongruous final images produced N400s and an
stimuli (Usher, Cohen, Servan-Schreiber, Rajkowski, & Aston- earlier negative response at approximately 325 ms. In another
Jones, 1999). The ACC is sensitive to both sorts of error and may study, participants viewed short movies in which an object ap-
play a role in adaptively modulating behavior in response to peared that was either expected or unexpected, given the context of
prediction error (Botvinick, Braver, Barch, Carter, & Cohen, the movie (Sitnikova, Kuperberg, & Holcomb, 2003). Unexpected
2001). More specifically, different subpopulations within the ACC objects led to a frontal-medial N400 and a large late positive
respond when a monkey is learning a new sequential structure than response at lateral electrodes. The N400 may have reflected a
when it has discovered the sequence and is using the learned mismatch between the semantic features of the final object and its
information to guide performance (Procyk, Tanaka, & Joseph, context, whereas the late positive component may have reflected a
2000). ACC and nearby regions have been proposed to underlie lack of fit of the final object into the causal/logical structure of the
learning of sequential structure in simple motor tasks and in activity (Sitnikova, Holcomb, & Kuperberg, in press). Together,
cognitive domains (Koechlin, Danek, Burnod, & Grafman, 2002). these data suggest that drops in predictability in text, pictures, or
One possibility is that the ACC computes the discrepancy between movies produce a reliable brain response that is maximal approx-
perceptual predictions and actual inputs (Cohen, Botvinick, & imately 400 ms after presentation of the unexpected stimulus. A
Carter, 2000) and that when prediction error spikes, this triggers recent source-localization ERP study (Frishkoff, Tucker, Davey, &
the nuclei of the catecholamine neurotransmitter systems. These Scherg, 2004) localized the earliest correlates of semantic incon-
subcortical nuclei have diffuse projections throughout cortex gruity in a sentence-processing paradigm to the ACC, with activity
(Schultz & Dickinson, 2000) and receive inputs from the ACC then spreading to prefrontal sites (and, to a lesser degree, posterior
(Holroyd & Coles, 2002). The resetting of event models may be sites). These results suggest that discrepancies between the current
mediated by projections from the substantia nigra or locus ceruleus event model and incoming information can lead to error signals
to the striatum, which modulates activity in frontostriatal circuits, that originate in the ACC and then propagate widely throughout
or by specific reciprocal connections with lateral PFC (Picard & the brain.
Event Models and Event Schemata 1997). The specific association of PFC with script processing has
been proposed by Grafman and colleagues (e.g., Grafman, 1995;
Data from several paradigms suggest that both event models Wood & Grafman, 2003). They argued that PFC is the storage site
(representations of what is happening now) and event schemata for event representations call managerial knowledge units, a type
(representations of semantic knowledge about events in general) of script. They pointed out that the cytoarchitecture of PFC is
may be implemented by anterior lateral PFC (BA 45/46). Many of comparable to the rest of cortex, suggesting that it processes
the experimental tasks that have been used do not permit one to information in ways similar to other cortical areas. However,
distinguish between schemata and models; therefore, we focus subdivisions of PFC are uniquely positioned to integrate multimo-
here on neuropsychological and neurophysiological data that es- dal information about the typical unfolding of everyday events. In
tablish the importance of PFC for tasks that require representations support of this view, Grafman and colleagues have built on the
of events. previous clinical literature on action disorders, collecting new
Before proceeding to neuropsychological data that do appear to lesion and neuroimaging data to test their proposal directly.
constrain the neural substrate of event representations, we discuss One study compared patients with prefrontal lesions with pa-
two disorders in which the pattern of deficits does not appear to tients with posterior lesions and non-brain-damaged controls (Si-
provide constraints with respect to which brain areas represent rigu et al., 1995). When participants were asked to list the events
events. The first of these is dementia of the Alzheimer’s type that are typical of everyday activities, the prefrontal group was
(DAT). On its face, DAT is a reasonable candidate for a disorder more likely than the other groups to end their lists early (before the
affecting event representations because patients with the disorder activity was complete) or late (including extra actions outside the
sometimes fail to remain oriented with regard to the time, location, activity). When the participants were asked to place the events in
and people present in a given situation. Two studies examined the a script into the correct temporal order, all but 1 of the prefrontal
ability of patients with DAT to generate scripts for a particular patients made errors, whereas the other groups performed without
activity or to verify the order in which two actions typically error. Other studies have found that patients with prefrontal lesions
occurred in an activity (Grafman et al., 1991; Weingartner, Graf- are less able than controls to identify violations of normal sequen-
man, Boutelle, Kaye, & Martin, 1983). Such tasks should depend tial structure in scripts and to identify events that are not part of a
on the presence of intact event schemata and may involve the script (Allain, Le Gall, Etcharry-Bouyx, Aubin, & Emile, 1999;
construction of event models. Both studies found evidence that Sirigu et al., 1996).
patients with DAT had impaired script processing; however, this Converging with the data from patients with focal PFC lesions
impairment was associated with other disorders of semantic are data from patients with neurological diseases that include PFC
knowledge and did not appear to be a specific impairment of event pathology, which suggest that PFC dysfunction impairs access to
representations. A third study assessed the ability of patients with either event models or event schemata. Patients with schizophrenia
very mild DAT to segment and to remember everyday activities have relatively intact abilities to identify fine-grained event bound-
(Zacks, Speer, et al., 2006). Those with dementia were poorer at aries but are selectively impaired at identifying the correct location
segmenting than neurologically healthy older adults and had of coarse-grained boundaries (Zalla, Verlut, Franck, Puzenat, &
poorer memory for the events. However, these deficits were again Sirigu, 2004). Patients with Parkinson’s disease, which is charac-
part of a general pattern of cognitive decline. Within both groups terized by degeneration of frontostriatal dopamine circuits, also are
of participants, event segmentation was predictive of later memory impaired on tasks that require ordering events within a script and
even after controlling for overall level of cognitive function. This detecting events from outside the script (Zalla et al., 1998). The
suggests that event understanding plays a unique role in memory; authors of this study proposed that the Parkinson’s disease pa-
however, these data do not indicate that event understanding is tients’ deficits reflected damage not to the underlying representa-
selectively impaired in DAT. tions in PFC but to the switching mechanisms implemented by
Another patient population in which script processing has been frontostriatal circuits; a similar logic could be applied to the
investigated is autism. In one study (Trillingsgaard, 1999), high- schizophrenia data, as frontal dopamine projections are compro-
functioning autistic children (with IQs in the normal range) and mised in schizophrenia (Knable & Weinberger, 1997).
control participants matched for mental age were asked to describe Several neuroimaging studies of event understanding have used
what typically happens in a set of everyday activities. The controls script judgment tasks. In one study, participants were asked to
almost always gave lists of events that met minimal criteria for imagine the sequence of actions involved in dressing and preparing
conforming to scripts, but the autistic individuals did so only half for an emotionally neutral event (dinner) or a sad event (their
of the time. However, the autistic children also were impaired on mother’s funeral) while brain activity was recorded using positron
a test of second-order theory of mind (knowing what someone else emission tomography (Partiot, Grafman, Sadato, Flitman, & Wild,
knows about what you know), which does not, on its face, appear 1996). Thinking about the sequence of activities in both the neutral
to depend specifically on event representations. Thus, DAT and and sad contexts led to increased activity in PFC (including the
possibly autism may produce disordered event representations, but superior, medial, and middle frontal gyri) compared with a set of
this appears to be the result of general cognitive disturbances loose control tasks. An fMRI study of script processing used an
rather than damage to event representations in particular. order verification task in which participants read a list of words
Although studies of DAT and autism do not appear to constrain describing events and were asked to indicate whether the events
the localization of event representations, studies of patients with were listed in the order in which they typically occurred (Crozier
focal damage to PFC do. PFC has been associated with the et al., 1999). A tight control task was administered: verifying
processing of temporally structured information and the mainte- whether a list of words formed a syntactically valid sentence. The
nance of information over long delays (for a review, see Fuster, script task led to increases in four areas not activated in the
sentence task: the left and right middle frontal gyri in PFC (BA 8), Mayer, 1991). Direct evidence that similar representations support
the left supplementary motor area, the posterior frontal cortex (BA the sequence of events in comprehension and in action comes from
6), and the left angular gyrus in the parietal lobe (BA 39). Acti- a study of frontal lesion patients with and without action disorga-
vation in these areas has subsequently been replicated in another nization syndrome (Humphreys & Forde, 1998). In this study,
fMRI study (Knutson, Wood, & Grafman, 2004). A recent positron patients with frontal lesions who had difficulty performing sequen-
emission tomography study focused on the temporal grain of script tially structured activities also performed poorly at judgments
processing (Ruby, Sirigu, & Decety, 2002). Participants judged about event structure, including the script-listing task described
whether three-event sequences (depicted as pictures or two-word previously. However, it is important to note that the question of
descriptions) were shown in their typical order. The sequences whether action disorganization is the result of specific damage to
showed events at either a long timescale (e.g., growing a crop) or event representations in PFC remains a matter of debate (Schwartz
a short timescale (e.g., brushing one’s teeth). Long-term order et al., 1998).
verification was associated with stronger activity bilaterally in the In sum, data from neuropsychological and neurophysiological
angular gyrus (BA 39), the precuneus, and the medial superior studies support the view that representations of events are sub-
frontal gyrus. Short-term order verification was associated with served by the lateral PFC. Our characterization of PFC represen-
stronger activity in the left middle frontal gyrus, inferior temporal tations has some similarity to that of Miller and Cohen (2001);
gyrus, middle occipital gyrus (BA 19), and supramarginal gyrus however, as noted previously, we hypothesize that PFC represents
(BA 40). Particularly intriguing was a pattern suggesting a topo- events rather than goals and the means to achieve them. Our
graphic organization of short-term and long-term events in parietal characterization of event representations in PFC also draws
cortex, with short-term event structure apparently represented heavily on the arguments of Grafman and colleagues (Grafman,
more anteriorly. However, this design does not permit conclusions 1995; Wood & Grafman, 2003; Grafman, Partiot, & Hollnagel,
about which areas the order verification task activated overall. 1995) as indicated previously. The present theory augments those
Together, these results suggest that PFC and perhaps the lateral theoretical proposals in two ways: First, it proposes a mechanism
parietal cortex are selectively activated by thinking about the by which event models are reset, and second, it proposes that
sequential structure of events. These data also provide some indi- working memory representations can be dissociated from long-
cation that this activation is stronger in the left hemisphere; how- term memory representations.
ever, further research is needed to clarify the role and lateral
organization of these areas. Toward Refining the Theory
PFC does not generally increase in activity at event boundaries
during passive viewing (Speer, Reynolds, & Zacks, in press; The theory in its current form has some clear limitations. First,
Zacks, Braver, et al., 2001; Zacks, Swallow, et al., 2006). How- the mechanism by which event models are reset is an important
ever, increases in fMRI activity were observed in lateral and aspect of the account, but it needs more development. More formal
medial PFC during an event segmentation task, and an EEG modeling, particularly aimed at exploring the relationship between
indicated that a negative wave of activity moved from prefrontal the reset mechanism proposed here and that used in ART models
regions to posterior regions prior to the decision to identify a (Carpenter & Grossberg, 2003; Grossberg, 1999), would be very
segment boundary (S. J. Hanson, Negishi, & Hanson, 2001). One productive.
possibility is that this PFC activity reflects the activation of new A second limitation is that the theory in its current form is a
event models and/or event schemata at event boundaries. How- purely passive perceiver. In our view, perception and prediction
ever, another possibility is that it reflects task-specific decision are tightly interleaved with motor simulation. We hypothesize that
making or response planning rather than the activation of event the event schemata described in EST are used not just to guide
representations. EST does not make strong predictions about perception of new activities based on past experience but also to
whether PFC should be transiently activated at event boundaries; plan actions. We recognize that perceivers are usually also actors,
the theory predicts that the content of event models will change at and so their event models include information about their own
this point, but such changes may not produce global increases or goals and allow them to anticipate the consequences of their
decreases in overall neural activity. actions as well as the actions of others. The general proposal that
In addition to data from event comprehension, studies of pa- common representations support perception and action has broad
tients with difficulty sequencing actions support the hypothesis empirical support (for reviews, see Gallese, 2001; Hommel, Mues-
that PFC is specialized to maintain event representations. Schwartz seler, Aschersleben, & Prinz, 2001; Prinz, 1997). Moreover, there
and colleagues have characterized action disorganization syn- are hints that the specific representations posed by event models
drome as a selective impairment of the ability to sequence goal- may play important roles in guiding action planning (see also
directed actions in the face of an intact ability to perform individ- Zacks & Tversky, 2001). In infant and child development, parsing
ual actions and an intact understanding of the objects and activity into goal-based events may be critical for learning by
movements involved (Schwartz, 1995; Schwartz et al., 1995; imitation (Baldwin & Baird, 1999; Meltzoff, 1995). In the proce-
Schwartz, Reed, Montgomery, Palmer, & Mayer, 1991; Schwartz, dural learning studies described previously (see the Consequences
Segal, Veramonti, Ferraro, & Buxbaum, 2002). For example, when of Event Segmentation for Long-Term Memory section), higher
one patient with a bilateral lesion to frontal cortex tried to brush his quality segmentation was associated with better learning of an
teeth, he showed disturbances of sequential ordering that included action sequence (Hard et al., 2006; Lozano et al., in press; Zacks
a dramatic tendency to perseverate, repeating individual actions & Tversky, 2003). Finally, in the neural network modeling we
such as rinsing the brush under the faucet or repeating whole have reviewed (see the A Theory of Event Segmentation section),
sequences of actions (Schwartz, Reed, Montgomery, Palmer, & the gating of memory representations on the basis of prediction
error has been found to be a powerful technique for action control. Robertson et al., 2000; Speer, 2005; St. George, Kutas, Martinez,
Other models of everyday action sequencing have stable event & Sereno, 1999), and there are no direct comparisons of working
representations that are similar in spirit to those of EST, though memory and semantic memory representations of events in the
they do not share its commitment to a gating mechanism (Botvin- neuroimaging or neuropsychological literatures. An alternative
ick & Plaut, 2002, 2004; Cooper & Shallice, 2000). We believe an possibility is that both event models and event schemata are
important goal for future research should be to extend EST to implemented by bilateral PFC but by different regions within PFC
account for (a) how event models explicitly represent one’s own or different populations of neurons within a region.
goals and (b) how perceptual prediction can include predicting the A fourth line of research suggested by the theory concerns the
consequences of one’s actions. acquisition of event schemata. Schema acquisition is a difficult and
The theory makes one prediction that does not square well with understudied problem, but the current framework offers some
the available data: If sudden increases in prediction error detected points of traction. In particular, it may be valuable to adapt
in the ACC lead to the perception of an event boundary, one would procedures from the sequence learning literature for studying the
expect that increased ACC activity would be observed at event acquisition of schemata for everyday events, because EST claims
boundaries during passive viewing. This has not been the case that prefrontal event schemata are the same representations that are
(Speer et al., in press; Speer et al., 2003; Zacks, Braver, et al., involved in some kinds of sequential learning. EST also makes
2001). However, as was the case for prefrontal cortex, ACC specific neuroanatomic predictions (based primarily on proposals
activity has been reported during the active detection of event by Hazeltine & Ivry, 2003, mentioned previously). For example,
boundaries (S. J. Hanson et al., 2001). It is possible that the failure the cerebellum has been implicated particularly strongly in the
to detect ACC activity at event boundaries during passive viewing early stages of motor learning when temporal coordination is
reflects a lack of power. The passive viewing studies, to date, have effortful. Thus, activity in the cerebellum should be substantial
used relatively predictable and stereotyped events; thus, these when one is exposed to new sequentially structured events and
signals may have been especially weak. Alternatively, it may be should diminish with learning. Also, the supplementary motor area
that the error calculation is performed in the ACC, but the fMRI and lateral premotor cortex, both posterior to PFC in the frontal
signal is detected at the downstream afferents from the ACC. lobe, have been associated with implicit learning of sequential
However, this is at odds with findings that the fMRI signal in the relations in motor sequences. One possibility is that these regions
ACC is modulated by conflict in other domains (Botvinick et al., implement part of the sequential learning mechanisms described
2001). Further research is needed to verify whether the ACC previously and therefore should be active during the learning of
responds selectively at event boundaries; if it does not, research new sequential relations in everyday events, even if no overt motor
should focus on evaluating other candidate mechanisms for the output is required.
error detection mechanism.
Limitations notwithstanding, the theory in its current form pro- Conclusion
vides a heuristic framework to guide future research. First, the
theory predicts that the perception of an event boundary is asso- Events are natural kinds, just like objects: Everyday life consists
ciated with an orienting response. This could be tested with mea- of picnics and meetings just as it consists of chairs and birds. The
sures of eye behaviors (eye movements, blinks, and pupil diame- perception of event structure has only recently emerged as an
ter), postural responses, and peripheral physiological responses independent scientific problem. It draws on a broad body of
(galvanic skin response, heart rate). research in linguistics, psychology, and neuroscience as well as a
Second, EST predicts that the functional connectivity between small but growing literature on the perception of event structure
sensory processing pathways and right PFC increases at event per se. We are entering a stage in which these diverse findings can
boundaries. This could be tested with electrophysiology or func- be integrated into theories that provide comprehensive accounts of
tional MRI by measuring whether the states of these regions event perception and point the way for future research.
become transiently more correlated at event boundaries.
Third, EST claims that event models and event schemata are References
dissociable. If so, are they localized to different parts of the brain?
Abbott, V., Black, J. H., & Smith, E. E. (1985). The representation of
Current data on this question are mixed. One proposal comes from scripts in memory. Journal of Memory and Language, 24, 179 –199.
the text processing literature. In a recent review, Gernsbacher and Abelson, R. P. (1981). Psychological status of the script concept. American
Kaschak (2003) argued that the processes by which isolated ele- Psychologist, 36, 715–729.
ments are connected to build a coherent representation of the Allain, P., Le Gall, D., Etcharry-Bouyx, F., Aubin, G., & Emile, J. (1999).
situation described by a text depend particularly on brain areas in Mental representation of knowledge following frontal-lobe lesion: Dis-
the right hemisphere. Right PFC has been associated with the sociations on tasks using scripts. Journal of Clinical and Experimental
active maintenance of multimodal information over time (Gruber Neuropsychology, 21, 643– 665.
& von Cramon, 2003; Prabhakaran, Narayanan, Zhao, & Gabrieli, Avrahami, J., & Kareev, Y. (1994). The emergence of events. Cognition,
2000). Conversely, there is evidence that the left PFC is special- 53, 239 –261.
Baddeley, A. (2000). The episodic buffer: A new component of working
ized for the storage of semantic knowledge (Cabeza & Nyberg,
memory? Trends in Cognitive Sciences, 4, 417– 423.
2000). Together, such results suggest that event models may be Baird, J. A., & Baldwin, D. A. (2001). Making sense of human behavior:
right lateralized, and event schemata may be left lateralized within Action parsing and intentional inference. In B. F. Malle, L. J. Moses, &
PFC. However, the data regarding the specialization of the right D. A. Baldwin (Eds.), Intentions and intentionality: Foundations of
hemisphere for establishing narrative coherence are mixed (Ferstl social cognition (pp. 193–206). Cambridge, MA: MIT Press.
& von Cramon, 2001, 2002; Maguire, Frith, & Morris, 1999; Baldwin, D. A., & Baird, J. A. (1999). Action analysis: A gateway to
intentional inference. In P. Rochat (Ed.), Early social cognition (pp. Elman, J. L. (1990). Finding structure in time. Cognitive Science, 14,
215–240). Hillsdale, NJ: Erlbaum. 179 –211.
Barker, R. G., & Wright, H. F. (1966). One boy’s day: A specimen record Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory.
of behavior. Hamden, CT: Archon Books. Psychological Review, 102, 211–245.
Bartels, A., & Zeki, S. (2004). Functional brain mapping during free Felleman, D. J., & Van Essen, D. C. (1991). Distributed hierarchical
viewing of natural scenes. Human Brain Mapping, 21, 75– 85. processing in the primate cerebral cortex. Cerebral Cortex, 1, 1– 47.
Bartlett, F. C. (1932). Remembering: A study in experimental and social Ferstl, E. C., & von Cramon, D. Y. (2001). The role of coherence and
psychology. New York: Macmillan. cohesion in text comprehension: An event-related fMRI study. Cognitive
Beauchamp, M. S., Lee, K. E., Haxby, J. V., & Martin, A. (2003). fMRI Brain Research, 11, 325–340.
responses to video and point-light displays of moving humans and Ferstl, E. C., & von Cramon, D. Y. (2002). What does the frontomedian
manipulable objects. Journal of Cognitive Neuroscience, 15, 991–1001. cortex contribute to language processing: Coherence or theory of mind?
Boltz, M. (1992). Temporal accent structure and the remembering of Neuroimage, 17, 1599 –1612.
filmed narratives. Journal of Experimental Psychology: Human Percep- Frank, M. J., Loughry, B., & O’Reilly, R. C. (2001). Interactions between
tion and Performance, 18, 90 –105. frontal cortex and basal ganglia in working memory: A computational
Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D.
model. Cognitive, Affective and Behavioral Neuroscience, 1, 137–160.

(2001). Conflict monitoring and cognitive control. Psychological Re-
Frank, M. J., Seeberger, L. C., & O’Reilly, R. C. (2004, December 10). By

view, 108, 624 – 652. carrot or by stick: Cognitive reinforcement learning in parkinsonism.
Botvinick, M. M., & Plaut, D. C. (2002). Representing task context: Science, 306, 1940 –1943.
Insights from a connectionist model of action. Psychological Research, Frishkoff, G. A., Tucker, D. M., Davey, C., & Scherg, M. (2004). Frontal
66, 298 –311. and posterior sources of event-related potentials in semantic comprehen-
Botvinick, M. M., & Plaut, D. C. (2004). Doing without schema hierar- sion. Cognitive Brain Research, 20, 329 –354.
chies: A recurrent connectionist approach to routine sequential action Fuster, J. M. (1991). The prefrontal cortex and its relation to behavior.
and its pathologies. Psychological Review, 111, 394 – 429. Progress in Brain Research, 87, 201–211.
Bouret, S., & Sara, S. J. (2005). Network reset: A simplified overarching Fuster, J. M. (1997). The prefrontal cortex: Anatomy, physiology, and
theory of locus coeruleus noradrenaline function. Trends in Neuro- neuropsychology of the frontal lobe (3rd ed.). Philadelphia: Lippincott-
sciences, 28, 574 –582.
Raven.
Bower, G. H., Black, J. B., & Turner, T. J. (1979). Scripts in memory for
Gallese, V. (2001). The “shared manifold” hypothesis. Journal of Con-
text. Cognitive Psychology, 11, 177–220.
sciousness Studies, 8, 33–50.
Bower, G. H., & Rinck, M. (2001). Selecting one among many referents in
Gernsbacher, M. A., & Kaschak, M. P. (2003). Neuroimaging studies of
spatial situation models. Journal of Experimental Psychology: Learning,
language production and comprehension. Annual Review of Psychology,
Memory, and Cognition, 27, 81–98.
54, 91–114.
Braver, T. S., & Cohen, J. D. (2000). On the control of control: The role
Glenberg, A. M., & Kaschak, M. P. (2002). Grounding language in action.
of dopamine in regulating prefrontal function and working memory. In
Psychonomic Bulletin & Review, 9, 558 –565.
S. Monsell & J. Driver (Eds.), Attention and performance XVIII (pp.
Grafman, J. (1995). Similarities and distinctions among current models of
713–738). Cambridge, MA: MIT Press.
prefrontal cortical functions. Annals of the New York Academy of Sci-
Brent, M. R. (1999). An efficient, probabilistically sound algorithm for
ences, 769, 337–368.
segmentation and word discovery. Machine Learning, 34, 71–105.
Grafman, J., Partiot, A., & Hollnagel, C. (1995). Fables of the prefrontal
Brewer, W. F., & Dupree, D. A. (1983). Use of plan schemata in the recall
cortex. Behavioral and Brain Sciences, 18, 349 –358.
and recognition of goal-directed actions. Journal of Experimental Psy-
chology: Learning, Memory, and Cognition, 9, 117–129. Grafman, J., Thompson, K., Weingartner, H., Martinez, R., Lawlor, B. A.,
Burgess, N., Maguire, E. A., & O’Keefe, J. (2002). The human hippocam- & Sunderland, T. (1991). Script generation as an indicator of knowledge
pus and spatial and episodic memory. Neuron, 35, 625– 641. representation in patients with Alzheimer’s disease. Brain and Lan-
Cabeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical guage, 40, 344 –358.
review of 275 PET and fMRI studies. Journal of Cognitive Neuro- Grossberg, S. (1999). The link between brain learning, attention, and
science,12, 1– 47. consciousness. Consciousness and Cognition, 8, 1– 44.
Carpenter, G. A., & Grossberg, S. (2003). Adaptive resonance theory. In Grossman, E. D., Donnelly, M., Price, R., Pickens, D., Morgan, V.,
M. A. Arbib (Ed.), The handbook of brain theory and neural networks Neighbor, G., et al. (2000). Brain areas involved in perception of
(2nd ed., pp. 87–90). Cambridge, MA: MIT Press. biological motion. Journal of Cognitive Neuroscience, 12, 711–720.
Cohen, J. D., Botvinick, M., & Carter, C. S. (2000). Anterior cingulate and Gruber, O., & von Cramon, D. Y. (2003). The functional neuroanatomy of
prefrontal cortex: Who’s in control? Nature Neuroscience, 3, 421– 423. human working memory revisited: Evidence from 3-T fMRl studies
Cohen, J. D., Braver, T. S., & O’Reilly, R. C. (1996). A computational using classical domain-specific interference tasks. NeuroImage, 19,
approach to prefrontal cortex, cognitive control and schizophrenia: Re- 797– 809.
cent developments and current challenges. Philosophical Transactions Hanson, C., & Hanson, S. J. (1996). Development of schemata during
of the Royal Society, Series B, 351, 1515–1527. event parsing: Neisser’s perceptual cycle as a recurrent connectionist
Cooper, R., & Shallice, T. (2000). Contention scheduling and the control of network. Journal of Cognitive Neuroscience, 8, 119 –134.
routine activities. Cognitive Neuropsychology, 17, 297–338. Hanson, C., & Hirst, W. (1989). On the representation of events: A study
Coull, J. T., & Nobre, A. C. (1998). Where and when to pay attention: The of orientation, recall, and recognition. Journal of Experimental Psychol-
neural systems for directing attention to spatial locations and to time ogy: General, 118, 136 –147.
intervals as revealed by both PET and fMRI. Journal of Neuroscience, Hanson, C., & Hirst, W. (1991). Recognizing differences in recognition
18, 7426 –7435. tasks: A reply to Lassiter and Slaw. Journal of Experimental Psychol-
Crozier, S., Sirigu, A., Lehéricy, S., van de Moortele, P.-F., Pillon, B., ogy: General, 120, 211–212.
Grafman, J., et al. (1999). Distinct prefrontal activations in processing Hanson, S. J., Negishi, M., & Hanson, C. (2001). Connectionist neuroim-
sequence at the sentence and script level: An fMRI study. Neuropsy- aging. In S. Wermter, J. Austin, & D. Willshaw (Eds.), Emergent neural
chologia, 37, 1469 –1476. computational architectures based on neuroscience: Towards
neuroscience-inspired computing (pp. 560 –577). Berlin, Germany: Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal
Springer. cortex function. Annual Review of Neuroscience, 24, 167–202.
Hard, B. M., Lozano, S. C., & Tversky, B. (2006). Hierarchical encoding Neisser, U. (1967). Cognitive psychology. New York: Appleton-Century-
of behavior: Translating perception into action. Journal of Experimental Crofts.
Psychology: General, 135, 588 – 608. Newtson, D. (1973). Attribution and the unit of perception of ongoing
Hard, B. M., Tversky, B., & Lang, D. (in press). Segmenting abstract behavior. Journal of Personality and Social Psychology, 28, 28 –38.
events: Building event schemas. Memory & Cognition. Newtson, D. (1976). Foundations of attribution: The perception of ongoing
Hasson, U., Nir, Y., Levy, I., Fuhrmann, G., & Malach, R. (2004, March behavior. In J. H. Harvey, W. J. Ickes, & R. F. Kidd (Eds.), New
12). Intersubject synchronization of cortical activity during natural vi- directions in attribution research (pp. 223–248). Hillsdale, NJ: Erlbaum.
sion. Science, 303, 1634 –1640. Newtson, D., & Engquist, G. (1976). The perceptual organization of
Hazeltine, E., & Ivry, R. B. (2003). Neural structures that support implicit ongoing behavior. Journal of Experimental Social Psychology, 12, 436 –
sequence learning. In L. Jimenez (Ed.), Attention and implicit learning 450.
(pp. 73–107). Philadelphia: John Benjamins. Newtson, D., Engquist, G., & Bois, J. (1977). The objective basis of
Holroyd, C. B., & Coles, M. G. (2002). The neural basis of human error behavior units. Journal of Personality and Social Psychology, 35, 847–
862.
processing: Reinforcement learning, dopamine, and the error-related

Nobre, A. C. (2001). Orienting attention to instants in time. Neuropsycho-
negativity. Psychological Review, 109, 679 –709.

Hommel, B., Muesseler, J., Aschersleben, G., & Prinz, W. (2001). The logia, 39, 1317–1328.
theory of event coding (TEC): A framework for perception and action O’Reilly, R. C., Noelle, D. C., Braver, T. S., & Cohen, J. D. (2002).
planning. Behavioral and Brain Sciences, 24, 849 –937. Prefrontal cortex and dynamic categorization tasks: Representational
Humphreys, G. W., & Forde, E. M. E. (1998). Disordered action schema organization and neuromodulatory control. Cerebral Cortex, 12, 246 –
and action disorganisation syndrome. Cognitive Neuropsychology, 15, 257.
771– 811. Partiot, A., Grafman, J., Sadato, N., Flitman, S., & Wild, K. (1996). Brain
Johnson-Laird, P. N. (1989). Mental models. In M. I. Posner (Ed.), Foun- activation during script event processing. Neuroreport, 7, 761–766.
dations of cognitive science (pp. 469 –500). Cambridge, MA: MIT Press. Picard, N., & Strick, P. L. (1996). Motor areas of the medial wall: A review
of their location and functional activation. Cerebral Cortex, 6, 342–353.
Jordan, M. I., & Rumelhart, D. E. (1992). Forward models: Supervised
Posner, M. I., & Snyder, C. R. R. (1975). Attention and cognitive control.
learning with a distal teacher. Cognitive Science, 16, 307–354.
In G. H. Bower (Ed.), Information processing and cognition: The Loyola
Kahneman, D., Treisman, A., & Gibbs, B. J. (1992). The reviewing of
Symposium (pp. 55– 85). Hillsdale, NJ: Erlbaum.
object files: Object-specific integration of information. Cognitive Psy-
Prabhakaran, V., Narayanan, K., Zhao, Z., & Gabrieli, J. D. E. (2000).
chology, 24, 175–219.
Integration of diverse information in working memory within the frontal
Kintsch, W. (1994). Text comprehension, memory, and learning. American
lobe. Nature Neuroscience, 3, 85–90.
Psychologist, 49, 294 –303.
Prinz, W. (1997). Perception and action planning. European Journal of
Knable, M. B., & Weinberger, D. R. (1997). Dopamine, the prefrontal
Cognitive Psychology, 9, 129 –154.
cortex and schizophrenia. Journal of Psychopharmacology, 11, 123–
Procyk, E., Tanaka, Y. L., & Joseph, J. P. (2000). Anterior cingulate
131.
activity during routine and non-routine sequential behaviors in ma-
Knutson, K. M., Wood, J. N., & Grafman, J. (2004). Brain activation in
caques. Nature Neuroscience, 3, 502–508.
processing temporal sequence: An fMRI study. NeuroImage, 23, 1299 –
Reynolds, J. R., Zacks, J. M., & Braver, T. S. (in press). A computational
1307.
model of event segmentation from perceptual prediction. Cognitive
Koechlin, E., Danek, A., Burnod, Y., & Grafman, J. (2002). Medial
Science.
prefrontal and subcortical mechanisms underlying the acquisition of Rinck, M., & Bower, G. (2000). Temporal and spatial distance in situation
motor and cognitive action sequences in humans. Neuron, 35, 371–381. models. Memory and Cognition, 28, 1310 –1320.
Kolb, B., & Whishaw, I. Q. (2003). Fundamentals of human neuropsy- Robertson, D. A., Gernsbacher, M. A., Guidotti, S. J., Robertson, R. R.,
chology (5th ed.). New York: Worth Publishers. Irwin, W., Mock, B. J., et al. (2000). Functional neuroanatomy of the
Kutas, M., & Hillyard, S. A. (1980, January 11). Reading senseless cognitive process of mapping during discourse comprehension. Psycho-
sentences: Brain potentials reflect semantic incongruity. Science, 207, logical Science, 11, 255–260.
203–205. Rougier, N. P., Noelle, D. C., Braver, T. S., Cohen, J. D., & O’Reilly, R. C.
Lassiter, G. D. (1988). Behavior perception, affect, and memory. Social (2005). Prefrontal cortex and flexible cognitive control: Rules without
Cognition, 6, 150 –176. symbols. Proceedings of the National Academy of Sciences, USA, 102,
Lassiter, G. D., & Slaw, R. D. (1991). The unitization and memory of 7338 –7343.
events. Journal of Experimental Psychology: General, 120, 80 – 82. Rougier, N. P., & O’Reilly, R. C. (2002). Learning representations in a
Lassiter, G. D., Stone, J. I., & Rogers, S. L. (1988). Memorial conse- gated prefrontal cortex model of dynamic task switching. Cognitive
quences of variation in behavior perception. Journal of Experimental Science, 26, 503–520.
Social Psychology, 24, 222–239. Ruby, P., Sirigu, A., & Decety, J. (2002). Distinct areas in parietal cortex
Lichtenstein, E. D., & Brewer, W. F. (1980). Memory for goal-directed involved in long-term and short-term action planning: A PET investi-
events. Cognitive Psychology, 12, 412– 445. gation. Cortex, 38, 321–339.
Lozano, S. C., Hard, B. M., & Tversky, B. (in press). Perspective-taking Rumelhart, D. E. (1975). Notes on a schema for stories language, thought,
promotes action understanding and learning. Journal of Experimental and culture. In D. G. Bobrow & A. Collins (Eds.), Representation and
Psychology: Human Perception and Performance, 32, 1405–1421. understanding; studies in cognitive science (pp. 211–236). New York:
Maguire, E. A., Frith, C. D., & Morris, R. G. (1999). The functional Academic Press.
neuroanatomy of comprehension and memory: The importance of prior Rumelhart, D. E. (1977). Understanding and summarizing brief stories. In
knowledge. Brain, 122, 1839 –1850. D. Laberge & S. J. Samuels (Eds.), Basic processes in reading: Percep-
Meltzoff, A. N. (1995). Understanding the intentions of others: Re- tion and comprehension (pp. 265–303). Hillsdale, NJ: Erlbaum.
enactment of intended acts by 18-month-old children. Developmental Rumelhart, D. E. (1980). Schemata: The building blocks of cognition. In
Psychology, 31, 838 – 850. R. J. Spiro, B. C. Bruce, & W. F. Brewer (Eds.), Theoretical issues in
reading comprehension: Perspectives from cognitive psychology, lin- activity time-locked to narrative event boundaries. Psychological Sci-
guistics, artificial intelligence, and education (pp. 33–58). Hillsdale, NJ: ence.
Erlbaum. Speer, N. K., Swallow, K. M., & Zacks, J. M. (2003). Activation of human
Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996, December 13). motion processing areas during event perception. Cognitive, Affective, &
Statistical learning by 8-month-old infants. Science, 274, 1926 –1928. Behavioral Neuroscience, 3, 335–345.
Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals, and under- Speer, N. K., & Zacks, J. M. (2005). Temporal changes as event bound-
standing: An inquiry into human knowledge structures. Hillsdale, NJ: aries: Processing and memory consequences of narrative time shifts.
Erlbaum. Journal of Memory and Language, 53, 125–140.
Schultz, W. (1998). Predictive reward signal of dopamine neurons. Journal Speer, N. K., Zacks, J. M., & Reynolds, J. R. (2006). Situation changes
of Neurophysiology, 80, 1–27. predict the perception of event boundaries, reading time, and perceived
Schultz, W., & Dickinson, A. (2000). Neuronal coding of prediction errors. predictability in narrative comprehension. Manuscript submitted for
Annual Review of Neuroscience, 23, 473–500. publication.
Schwan, S., & Garsoffky, B. (2004). The cognitive representation of filmic St. George, M., Kutas, M., Martinez, A., & Sereno, M. I. (1999). Semantic
event summaries. Applied Cognitive Psychology, 18, 37–55. integration in reading: Engagement of the right hemisphere during
Schwan, S., Garsoffky, B., & Hesse, F. W. (2000). Do film cuts facilitate
discourse processing. Brain, 122, 1317–1325.

the perceptual and cognitive organization of activity sequences? Memory
Stränger, J., & Hommel, B. (1996). The perception of action and move-
& Cognition, 28, 214 –223. ment. In W. Prinz & B. Bridgeman (Eds.), Handbook of perception and
Schwartz, M. F. (1995). Re-examining the role of executive functions in action: Volume 1 (pp. 397– 451). San Diego, CA: Academic Press.
routine action production. Annals of the New York Academy of Sciences, Tanaka, K. (1996). Inferotemporal cortex and object vision. Annual Review
769, 321–325. of Neuroscience, 19, 109 –139.
Schwartz, M. F., Montgomery, M. W., Buxbaum, L. J., Lee, S. S., Carew, Tootell, R. B., Reppas, J. B., Kwong, K. K., Malach, R., Born, R. T.,
T. G., Coslett, H. B., et al. (1998). Naturalistic action impairment in Brady, T. J., et al. (1995). Functional analysis of human MT and related
closed head injury. Neuropsychology, 12, 13–28. visual cortical areas using magnetic resonance imaging. Journal of
Schwartz, M. F., Montgomery, M. W., Fitzpatrick-DeSalme, E. J., Ochipa, Neuroscience, 15, 3215–3230.
C., Coslett, H. B., & Mayer, N. H. (1995). Analysis of a disorder of Trillingsgaard, A. (1999). The script model in relation to autism. European
everyday action. Cognitive Neuropsychology, 12, 863– 892.
Child & Adolescent Psychiatry, 8, 45– 49.
Schwartz, M. F., Reed, E. S., Montgomery, M., Palmer, C., & Mayer, N. H.
Ungerleider, L. G., & Mishkin, M. (1978). Two cortical visual systems. In
(1991). The quantitative description of action disorganisation after brain
D. Ingle, M. A. Goodale, & R. J. W. Mansfield (Eds.), Analysis of visual
damage: A case study. Cognitive Neuropsychology, 8, 381– 414.
behavior (pp. 549 –586). Cambridge, MA: MIT Press.
Schwartz, M. F., Segal, M., Veramonti, T., Ferraro, M., & Buxbaum, L. J.
Usher, M., Cohen, J. D., Servan-Schreiber, D., Rajkowski, J., & Aston-
(2002). The Naturalistic Action Test: A standardised assessment for
Jones, G. (1999, January 22). The role of the locus coeruleus in the
everyday action impairment. Neuropsychological Rehabilitation, 12,
regulation of cognitive performance. Science, 283, 549 –554.
311–339.
van Berkum, J. J. A., Hagoort, P., & Brown, C. M. (1999). Semantic
Seger, C. A. (1994). Implicit learning. Psychological Bulletin, 115, 163–
integration in sentences and discourse: Evidence from the N400. Journal
196.
of Cognitive Neuroscience, 11, 657– 671.
Shelton, A. L., & Gabrieli, J. D. E. (2002). Neural correlates of encoding
van Berkum, J. J. A., Zwitserlood, P., Hagoort, P., & Brown, C. M. (2003).
space from route and survey perspectives. Journal of Neuroscience, 22,
When and how do listeners relate a sentence to the wider discourse?
2711–2717.
Evidence from the N400 effect. Cognitive Brain Research, 17, 701–718.
Shipley, T. F. (in press). An invitation to an event. In T. F. Shipley & J. M.
Zacks (Eds.), Understanding events: How humans see, represent, and van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse compre-
act on events. New York: Oxford University Press. hension. New York: Academic Press.
Sirigu, A., Zalla, T., Pillon, B., Grafman, J., Agid, Y., & Dubois, B. (1995). Weingartner, H., Grafman, J., Boutelle, W., Kaye, W., & Martin, P. R.
Selective impairments in managerial knowledge following pre-frontal (1983, July 22). Forms of memory failure. Science, 221, 380 –382.
cortex damage. Cortex, 31, 301–316. West, W. C., & Holcomb, P. J. (2002). Event-related potentials during
Sirigu, A., Zalla, T., Pillon, B., Grafman, J., Agid, Y., & Dubois, B. (1996). discourse-level semantic integration of complex pictures. Cognitive
Encoding of sequence and boundaries of scripts following prefrontal Brain Research, 13, 363–375.
lesions. Cortex, 32, 297–310. Wilder, D. A. (1978a). Effect or predictability on units of perception and
Sitnikova, T., Holcomb, P., & Kuperberg, G. (in press). Neurocognitive attribution. Personality and Social Psychology Bulletin, 4, 281–284.
mechanisms of human comprehension. In T. F. Shipley & J. M. Zacks Wilder, D. A. (1978b). Predictability of behaviors, goals, and unit of
(Eds.), Understanding events: How humans see, represent, and act on perception. Personality and Social Psychology Bulletin, 4, 604 – 607.
events. New York: Oxford University Press. Wood, J. N., & Grafman, J. (2003). Human prefrontal cortex: Processing
Sitnikova, T., Kuperberg, G., & Holcomb, P. J. (2003). Semantic integra- and representational perspectives. Nature Reviews Neuroscience, 4,
tion in videos of real-world events: An electrophysiological investiga- 139 –147.
tion. Psychophysiology, 40, 160 –164. Zacks, J. M. (2004). Using movement and intentions to understand simple
Sokolov, E. N., Spinks, J. A., Naeaetaenen, R., & Lyytinen, H. (Eds.). events. Cognitive Science, 28, 979 –1008.
(2002). The orienting response in information processing. Mahwah, NJ: Zacks, J. M., Braver, T. S., Sheridan, M. A., Donaldson, D. I., Snyder,
Erlbaum. A. Z., Ollinger, J. M., et al. (2001). Human brain activity time-locked to
Speer, N. K. (2005). Text comprehension processes in the brain. Unpub- perceptual event boundaries. Nature Neuroscience, 4, 651– 655.
lished doctoral dissertation, Washington University in St. Louis, St. Zacks, J. M., Speer, N. K., Vettel, J. M., & Jacoby, L. L. (2006). Event
Louis, Missouri. understanding and memory in healthy aging and dementia of the Alz-
Speer, N. K., Reynolds, J. R., Swallow, K. M., & Zacks, J. M. (2006). heimer type. Psychology and Aging, 21, 466 – 482.
Reading stories activates neural representations of perceptual and mo- Zacks, J. M., Swallow, K. M., Vettel, J. M., & McAvoy, M. P. (2006).
tor experiences. Manuscript submitted for publication. Visual movement and the neural correlates of event perception. Brain
Speer, N. K., Reynolds, J. R., & Zacks, J. M. (in press). Human brain Research, 1076, 150 –162.
Zacks, J. M., & Tversky, B. (2001). Event structure in perception and event-indexing model. In S. R. Goldman, A. C. Graesser, & P. van den
conception. Psychological Bulletin, 127, 3–21. Broek (Eds.), Narrative comprehension, causality, and coherence: Es-
Zacks, J. M., & Tversky, B. (2003). Structuring information interfaces for says in honor of Tom Trabasso (pp. 93–110). Mahwah, NJ: Erlbaum.
procedural learning. Journal of Experimental Psychology: Applied, 9, Zwaan, R. A., Magliano, J. P., & Graesser, A. C. (1995). Dimensions of
88 –100. situation model construction in narrative comprehension. Journal of
Zacks, J. M., Tversky, B., & Iyer, G. (2001). Perceiving, remembering, and Experimental Psychology: Learning, Memory, and Cognition, 21, 386 –
communicating structure in events. Journal of Experimental Psychol- 397.
ogy: General, 130, 29 –58. Zwaan, R. A., & Radvansky, G. A. (1998). Situation models in language
Zalla, T., Sirigu, A., Pillon, B., Dubois, B., Grafman, J., & Agid, Y. (1998). comprehension and memory. Psychological Bulletin, 123, 162–185.
Deficient in evaluating pre-determinated sequences of script events in Zwaan, R. A., Stanfield, R. A., & Yaxley, R. H. (2002). Language com-
patients with Parkinson’s disease. Cortex, 34, 621– 628. prehenders mentally represent the shape of objects. Psychological Sci-
Zalla, T., Verlut, I., Franck, N., Puzenat, D., & Sirigu, A. (2004). Percep- ence, 13, 168 –171.
tion of dynamic action in patients with schizophrenia. Psychiatry Re-
search, 128, 39 –51.
Zwaan, R. A. (1996). Processing narrative time shifts. Journal of Experi- Received November 28, 2005
mental Psychology: Learning, Memory, and Cognition, 22, 1196 –1207. Revision received June 20, 2006
Zwaan, R. A. (1999). Five dimensions of narrative comprehension: The Accepted July 11, 2006 䡲

Event Perception: A Mind-Brain Perspective

Uploaded by

Copyright:

Available Formats

Event Perception: A Mind-Brain Perspective

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Event Perception: A Mind-Brain Perspective

Uploaded by

Copyright:

Available Formats

Psychological Bulletin Copyright 2007 by the American Psychological Association

2007, Vol. 133, No. 2, 273–293 0033-2909/07/$12.00 DOI: 10.1037/0033-2909.133.2.273

Event Perception: A Mind–Brain Perspective

Keywords: events, attention, memory, prefrontal cortex, orienting response

sensory input. The stability of event models over time is a source

Predicted Future Inputs

Perceptual Processing Event Models Event Schemata

5. Event segmentation depends on change. When the world

in two regards. First, stable representations contain information

description provided in the text (e.g., Glenberg & Kaschak, 2002;

play a role in event segmentation. Two further results suggested

Predicted Future Inputs

Perceptual Processing Event Models Event Schemata

0.1 Coarse 0.1

model. Cognitive, Affective and Behavioral Neuroscience, 1, 137–160.

Frank, M. J., Seeberger, L. C., & O’Reilly, R. C. (2004, December 10). By

processing: Reinforcement learning, dopamine, and the error-related

negativity. Psychological Review, 109, 679 –709.

discourse processing. Brain, 122, 1317–1325.

You might also like