Wagner, Bunge, & Badre
1
Cognitive control, semantic memory, and priming:
Contributions from prefrontal cortex
Anthony D. Wagner1,2, Silvia A., Bunge3, & David Badre4
1
Department of Psychology and Neurosciences Program, Stanford University;
2
3
Martinos Center for Biomedical Imaging, MGH/MIT/HMS;
Department of Psychology and Center for Mind & Brain, UC Davis;
4
Department of Brain and Cognitive Sciences, MIT
Correspondence:
Anthony D. Wagner
Department of Brain and Cognitive Sciences
NE20-463
Cambridge, MA 02139
[email protected]
617-452-2492 (ph)
617-258-8654 (fax)
Wagner, Bunge, & Badre
2
Cognitive control refers to mechanisms that permit an individual to access and work with
internal representations in a goal-directed manner. In so doing, cognitive control
supports context-relevant stimulus processing, the retrieval and on-line maintenance of
knowledge, and the transformation of representations to satisfy our goals. Prefrontal
cortex (PFC) is a central component of the neural circuitry underlying cognitive control,
including the control of memory; flexible behavior often depends on PFC mechanisms
that support access to long-term knowledge. Accordingly, the specification of cognitive
control processes, their dependence on PFC, and their interactions with long-term
memory is of fundamental importance (Fuster, 1997; Goldman-Rakic, 1987; Miller and
Cohen, 2001; Petrides, 1994; Schacter, 1987b; Shallice, 1988; Shimamura, 1995; Stuss
and Benson, 1984; Wagner, 2002).
In this chapter, we consider neuroimaging evidence regarding PFC contributions
to cognitive control, and the interactions between control mechanisms and memory. We
first delineate the elemental processes that constitute working memory and cognitive
control, and their organization within PFC. We then provide an in-depth discussion of
one control mechanism to illustrate how cognitive control can interact with long-term
memory. This discussion considers how cognitive control supports retrieval from
declarative (or explicit) memory, as revealed through PFC contributions to the controlled
retrieval of task-relevant semantic knowledge. We conclude by exploring the relation
between this PFC control mechanism and priming, a nondeclarative (or implicit) form of
Wagner, Bunge, & Badre
3
memory, illustrating how reductions in cognitive control demands sometimes follow
experience-based changes in long-term memory.
Working memory and cognitive control
Language comprehension, problem solving, goal satisfaction, and other high-level
cognitive functions depend on working memory (WM), which refers to our ability to
maintain and manipulate active representations (Fuster and Alexander, 1971; GoldmanRakic, 1987; Miller et al., 1960). WM is a multifaceted, rather than unitary, faculty. A
prominent model of human WM proposes that this ability depends on at least three
components: a buffer for the short-term maintenance of verbal information, a buffer for
the maintenance of visuo-spatial information, and a ‘central executive’ that gates and
manipulates representations held in these buffers (Baddeley, 1986; Baddeley and Hitch,
1974). This influential perspective has proven an effective framework for examining the
architecture of WM, although, as we will see, WM can be alternatively conceptualized as
an integrated set of cognitive control processes.
VERBAL WORKING MEMORY
Verbal WM constitutes the maintenance of speech-
based (phonological) representations, as when we subvocally rehearse a phone number
while preparing to dial it. Verbal WM depends on a system termed the phonological
loop, which consists of two components: a phonological store that represents active
phonological information, and an articulatory control process that rehearses the contents
of the store and reactivates long-term phonological (Baddeley and Hitch, 1974). Support
for this architecture comes from behavioral and neuropsychological observations that the
Wagner, Bunge, & Badre
4
two components of the phonological loop are dissociable (Baddeley, 1986; Smith and
Jonides, 1995).
Neuroimaging studies of verbal WM consistently implicate frontal, parietal, and
cerebellar regions that work together to support the phonological loop (Figure 1) (Smith
and Jonides, 1999). Demands on the articulatory control process elicit activation in
structures important for speech production, including left ventrolateral PFC
(~Brodmann's area [BA] 44; Broca's area) and left premotor and supplementary motor
cortices (BA 6) (Awh et al., 1996; Fiez et al., 1996; Paulesu et al., 1993). These frontal
regions are thought to maintain active phonological representations stored in left or
bilateral inferior parietal cortices (Awh et al., 1996; Bunge et al., 2001; Jonides et al.,
1998a; but see Chein and Fiez, 2001). Cerebellar subregions may integrate inputs from
ventrolateral PFC and parietal regions, providing a corrective signal that refines the
rehearsal process (Desmond et al., 1997).
The contributions of specific PFC subregions to verbal WM partially depend on
the amount of information that must be maintained. Whereas control processes mediated
by left ventrolateral PFC appear sufficient to maintain low WM loads, when the amount
of to-be-maintained information begins to approach an individual's WM capacity limit,
activation also emerges in bilateral dorsolateral PFC (BA 9/46) (Rypma et al., 1999).
Dorsolateral PFC activation may reflect a shift to reliance on additional cognitive control
processes, such as those supporting the ‘chunking’ of information (but see Veltman et al.,
2003). This interpretation is consistent with the idea that central executive mechanisms
are engaged when WM maintenance demands approach or exceed capacity limits
(Baddeley and Hitch, 1974), and further suggests that, in humans, dorsolateral PFC
Wagner, Bunge, & Badre
5
supports non-maintenance cognitive control processes – a hypothesis to which we return
below.
SPATIAL AND OBJECT WORKING MEMORY Visuo-spatial WM allows us to
maintain visual representations of objects and knowledge of the position of objects in
space. As with verbal WM, this ability depends on PFC–posterior neocortical
interactions (Chafee and Goldman-Rakic, 2000; Miller and Cohen, 2001). However,
whereas verbal WM is associated with left PFC, visuo-spatial WM in humans is
differentially associated with right-lateralized or bilateral PFC (D'Esposito et al., 1998;
Prabhakaran et al., 2000; Smith and Jonides, 1999). Visuo-spatial WM also engages
occipito-temporal and parietal cortices that support object and spatial representations;
PFC is thought to control the maintenance of spatial and object codes through reciprocal
interactions with these posterior representations. One hypothesis regarding rehearsal in
spatial WM is that spatial information is maintained through shifts of spatial selective
attention to location-specific representations (Awh and Jonides, 1998).
In contrast to the traditional two-store model of WM, which proposes that object
and spatial representations are maintained by a single buffer, termed the visuo-spatial
sketchpad (Baddeley, 1986), recent evidence suggests that object and spatial WM may be
separable (e.g., Baddeley, 1994; Carlesimo et al., 2001; Farah, 1988). Behaviorally,
spatial distraction impairs spatial WM more than visual WM, whereas visuo-object
distraction has the reverse effect (e.g., Tresch et al., 1993). At the neural level, singlecell recordings in infrahuman primates indicate that some PFC neurons maintain spatial
representations, others maintain object representations, and yet others maintain an
Wagner, Bunge, & Badre
6
integration of spatial and object information (Rainer et al., 1998; Rao et al., 1997). Thus,
segregation and interaction characterize WM for space and objects.
Object and spatial WM are not strictly topographically segregated in PFC. In
infrahuman primates, object, spatial, and integrative neurons are present in ventrolateral
and dorsolateral PFC (Figure 2) (Rainer et al., 1998), with there being a moderate bias for
a greater proportion of object WM neurons in ventrolateral PFC and spatial WM neurons
in dorsolateral PFC (Wilson et al., 1993). In humans, many neuroimaging studies have
failed to observe a marked segregation between spatial and object WM within PFC, with
both forms of maintenance often eliciting activation in ventrolateral and dorsolateral PFC
(e.g., D'Esposito et al., 1998; McCarthy et al., 1996; Nystrom et al., 2000). By contrast,
other studies suggest that spatial and object WM are at least partially separable, with a
subregion of the superior frontal sulcus being differentially associated with spatial WM
(Courtney et al., 1998; Haxby et al., 2000; Sala et al., 2003; see also Alain et al., 2001).
Although the degree of topographic segregation between spatial and object WM
in PFC remains uncertain, the collective data indicate that a comprehensive model of
human WM requires separable systems for object, spatial, and phonological information,
and an understanding of how these different types of information are integrated.
Neuropsychological and neuroimaging data also suggest that a fourth system may
subserve WM for semantic knowledge (e.g., Demb et al., 1995; Martin et al., 1994).
COGNITIVE CONTROL
The central executive, as originally proposed, is a limited-
capacity attentional system that coordinates WM maintenance buffers and manipulates
their contents (Baddeley, 1986). The executive also is thought to mediate selective
Wagner, Bunge, & Badre
7
attention and the inhibition of task-irrelevant representations. Importantly, although once
conceptualized as a monolithic component of WM, there is little behavioral evidence for
a single executive processor (e.g., Towse and Houston-Price, 2001). Rather, 'executive'
functions, which may comprise both WM manipulation and maintenance processes,
likely emerge from a set of elemental cognitive control mechanisms.
Over the past decade, the notion of a central executive has evolved into the
concept of cognitive control, which entails processes that constrain our thoughts and
responses in accordance with our goals. Putative control functions include: dual-task
coordination and task-switching (task management), retrieval of goal-relevant
information from long-term memory (controlled retrieval), selection of appropriate
responses (response selection) and inhibition of inappropriate responses (response
inhibition), resolution of distraction from competing representations (interference
resolution), transformation, reordering, or updating of information maintained in WM
(manipulation), and evaluation of whether information in WM meets the criteria
associated with one's goal (monitoring).
Biased competition theory
According to the biased competition framework (Desimone
and Duncan, 1995), a similar neural mechanism may underlie each of the abovehypothesized control functions (Miller and Cohen, 2001). Biased competition theory
asserts that, in the absence of cognitive control, the representation most strongly activated
by response cues in an automatic, bottom-up fashion will serve as the basis for behavior
irrespective of whether the representation is task-relevant. However, cognitive control, in
the form of top-down excitatory signals deriving from PFC, can bias information
Wagner, Bunge, & Badre
8
processing in favor of weaker, but task-relevant representations, enhancing these
representations and indirectly suppressing task-irrelevant competitors through lateral
inhibitory interconnections between representations (Cohen et al., 1990). This top-down
bias signal is argued to constitute the fundamental basis of cognitive control, enabling the
selection of appropriate goal representations, the retrieval of target knowledge from WM
and long-term memory (LTM), and the execution of context-appropriate responses (see
also Shimamura, 1995).
Intimately related to the biased competition framework is the hypothesis that the
anterior cingulate cortex (ACC) and PFC play distinct roles in cognitive control (Figure
3) (Cohen et al., 1996; MacDonald et al., 2000). From this perspective, the rostral zone
of ACC serves to detect the presence of conflict (i.e., the simultaneous activation of
multiple, competing representations), and signals PFC to resolve the conflict by upregulating control (i.e., increasing the top-down biasing of task-appropriate
representations). Computational models and some neuroimaging data suggest that ACC
may selectively detect conflict between competing responses (Botvinick et al., 2001;
Milham et al., 2001; van Veen et al., 2001). However, other evidence points to a more
general function, with ACC detecting conflict at multiple processing stages (see van
Veen and Carter, 2002).
Dissociable mechanisms in dorsolateral and ventrolateral PFC
A complementary
perspective on cognitive control is the hypothesis that “maintenance” and “manipulation”
constitute interacting, but dissociable, control mechanisms that differentially depend on
distinct PFC subregions. From this perspective, the rehearsal processes that support
Wagner, Bunge, & Badre
9
maintenance in verbal, object, and spatial WM are, themselves, forms of cognitive
control. Accordingly, ventrolateral PFC regions are argued to control the retrieval of
task-relevant representations and their active maintenance. By contrast, dorsolateral PFC
functions monitor the contents of WM and/or manipulate active representations in the
service of completing the current goal (D'Esposito et al., 1998; Owen et al., 1996;
Petrides, 1994; but see Raye et al., 2002; Veltman et al., 2003).
Neuroimaging data provide empirical support for differential roles of ventrolateral
and dorsolateral PFC subregions in cognitive control. As already discussed,
neuroimaging studies indicate that dorsolateral PFC is recruited at higher verbal WM
loads, suggesting that it is not involved in maintenance per se, but instead may play a role
in reorganizing or manipulating representations to facilitate maintenance (Rypma et al.,
1999). Dorsolateral regions are also differentially engaged when we reorder (D'Esposito
et al., 1999a; Postle et al., 1999; Wagner et al., 2001a) or update (Garavan et al., 2000;
Salmon et al., 1996) representations in WM (Figure 4). Thus, dorsolateral PFC
mechanisms may operate on and transform the contents of WM.
In addition to its role in active maintenance, ventrolateral PFC supports the
control of thought and action in ways that are just beginning to be understood.
Subregions of ventrolateral PFC are implicated in the controlled retrieval and selection of
task-relevant long-term knowledge (Badre and Wagner, 2002; Petrides, 1994; ThompsonSchill et al., 1997), a point to which we return in the next section. Other subregions of
left ventrolateral PFC appear to resolve interference in verbal WM (Bunge et al., 2001;
D'Esposito et al., 1999b; Jonides et al., 1998b), with homologous right ventrolateral
regions supporting interference resolution between nonverbal representations (Hazeltine
Wagner, Bunge, & Badre
10
et al., in press). Right ventrolateral PFC regions are also active when we shift between
alternate response criteria, as in the Wisconsin Card-Sorting Task (Konishi et al., 1998a),
and when we halt the execution of an inappropriate response (Bunge et al., 2002;
Garavan et al., 1999; Konishi et al., 1998b; see also Aron et al., 2003).
Frontopolar control processes
Extant data also implicate frontopolar cortex (FPC)
as central to cognitive control. However, perhaps owing to a paucity of comparative
data, FPC remains poorly understood at the functional level, even though neuroimaging
studies consistently observe FPC activation during episodic retrieval, complex working
memory, and higher-level cognitive tasks (Fletcher and Henson, 2001). Findings from
the task-management literatures suggest that FPC is active when subjects must maintain a
primary task goal or task-relevant information while simultaneously attending to a
secondary goal or task (Braver and Bongiolatti, 2002; Koechlin et al., 1999). An
emerging literature on analogical reasoning further suggests that FPC is important for
integrating between disparate representations or relations (Bunge et al., 2003; Christoff et
al., 2001; Kroger et al., 2002). Given these findings, FPC involvement in episodic
retrieval may reflect processes that integrate retrieved information with response criteria
or context information held in WM, thus permitting decisions about whether the retrieved
information satisfies the mnemonic goal (e.g., Rugg and Wilding, 2000).
As suggested by this formulation of frontopolar function, one intriguing
development over the past decade is the emerging conceptualization of cognitive control
as an ensemble of processes that intimately interact with mnemonic and cognitive
systems. Accordingly, efforts to delineate cognitive control may benefit from approaches
Wagner, Bunge, & Badre
11
that examine the role of control processes in a variety of memory domains. With this in
mind, we now turn to consider evidence regarding the interaction between cognitive
control and LTM, providing an in-depth discussion of a particular form of control:
controlled retrieval. In the next section, we delineate how controlled retrieval serves to
activate declarative knowledge structures through a top-down biasing of semantic
memory, and then conclude the chapter by considering how nondeclarative memory – in
particular, priming – influences controlled retrieval demands.
Cognitive control and semantic memory
Recovering context-relevant meaning about the world allows us to flexibly use long-term
knowledge to generate goal-consistent responses. Consider, for example, that you need
to pound in a nail and the only objects available are standard office supplies, such as a
stapler, computer, and the like. Although information strongly associated with each
concept (e.g., that staplers are used to bind documents) is of little assistance in this
context, you are capable of accessing other weakly associated, but context-relevant,
knowledge about the objects. Thus, you may recall that staplers are heavy and fairly
wieldy objects, whereas computers tend to be fragile. Such knowledge may ultimately
support the inference that your stapler is a suitable substitute for a hammer. Importantly,
your ability to retrieve this information requires processes that guide access to, or control
the retrieval of, relevant knowledge when strongly associated semantic information is
insufficient to meet task demands.
This example illustrates how some situations require control processes that guide
retrieval from semantic memory, which is a declarative (or explicit) form of LTM that
Wagner, Bunge, & Badre
12
represents our general knowledge about the world, including facts, concepts, and
vocabulary (Tulving, 1972). In this section, we consider when and how cognitive control
mediates semantic retrieval. We begin by briefly introducing basic concepts regarding
the organization of semantic knowledge, and then consider how semantic information can
be retrieved through automatic or controlled processes. The consequences of left PFC
lesions for semantic retrieval are subsequently discussed as a foundation for
understanding neuroimaging data regarding the role of PFC in controlled semantic
retrieval.
STRUCTURE OF SEMANTIC KNOWLEDGE
Our aim is to consider the
contributions of PFC to cognitive control and to specify how control processes interact
with LTM; as such, a full discussion of the structure of semantic knowledge is beyond the
scope of this chapter (see Martin and Chao, 2001; Smith and Medin, 1981). However, a
few characteristics of semantic memory are relevant for understanding the processes that
retrieve semantic knowledge. First, semantic information is stored in a distributed,
associative network that links conceptual representations and components of
representations. Second, the associations between representations vary in strength
depending on the frequency of their prior co-occurrence, their overlap in features, and/or
their categorical relations. Finally, multiple representations can compete for processing
and retrieval through mutually inhibitory interactions.
At the neural level, semantic knowledge is distributed throughout posterior
neocortex, including lateral and ventral regions of the temporal lobes (for review, see
Martin and Chao, 2001). Patients suffering from the temporal-variant of semantic
Wagner, Bunge, & Badre
13
dementia demonstrate a marked loss of semantic knowledge due to degeneration of
temporal cortices, and more focal temporal lesions can result in category-specific
dementia – the loss of knowledge associated with certain taxonomic categories. These
latter deficits suggest that different semantic primitives correlate more strongly with
certain categories, with their representation being differentially dependent on particular
posterior cortical regions (Farah and McClelland, 1991; Martin and Chao, 2001;
Thompson-Schill, 2003; but see Caramazza and Shelton, 1998). Convergent with this
interpretation, neuroimaging studies demonstrate a systematic relation between activation
in temporal cortical subregions and the processing of specific semantic categories or
features (Martin and Chao, 2001). For example, subregions within inferotemporal cortex
are differentially active during the processing of living versus non-living concepts
(Figure 5), perhaps revealing markers of stored primitives related to visual versus
functional semantics (e.g., Chao et al., 1999; Martin et al., 1996; Thompson-Schill et al.,
1999a; but see Pilgrim et al., 2002). Other putative primitives, such as knowledge about
color versus action, result in functional dissociations in lateral temporal cortex (e.g.,
Martin et al., 1995; see also Kellenbach et al., 2003; Kourtzi and Kanwisher, 2000).
Collectively, these data indicate that semantic knowledge is stored in a distributed, but
systematic, manner in posterior neocortex.
Complementary data indicate that long-term semantic knowledge is not stored in
PFC, although PFC mechanisms appear to impact our ability to work with semantic
information. For example, multidimensional scaling reveals that the structure of
semantic memory is similar in patients with PFC damage and in healthy control subjects
(Figure 5), although such patients sometimes demonstrate difficulties in retrieving stored
Wagner, Bunge, & Badre
14
knowledge (Sylvester and Shimamura, 2002). These latter deficits beg the question:
When and how does cognitive control interact with semantic memory to support
knowledge retrieval?
MULTIPLE ROUTES TO MEMORY
Semantic retrieval constitutes the recovery
of conceptual representations, and accordingly depends on processes that activate stored
knowledge, essentially bringing long-term information into WM. Behavioral evidence
indicates that retrieval of task-relevant semantic knowledge can occur in a relatively
automatic (or bottom-up) fashion or in a more controlled (or top-down) manner, with
these two retrieval routes representing the ends of a continuum (for review, see Neely,
1991).
Automatic retrieval occurs when the association between a retrieval cue and
relevant knowledge is sufficiently strong that presentation of the cue serves to
automatically activate, or make accessible, the target knowledge (Figure 6). Automatic
retrieval is thought to derive from an associative mechanism, wherein bottom-up
activation of the cue’s representation serves to activate other, strongly associated
representations. Automatic retrieval (a) occurs rapidly, (b) is obligatory, and (c) is
context-independent, resulting in recovery of strongly associated knowledge regardless of
whether the knowledge is task-relevant.
By contrast, controlled retrieval mechanisms are recruited when automatically
retrieved knowledge is insufficient to meet task demands or when an individual comes to
expect that the strategic retrieval of certain conceptual representations will aid
performance. Controlled retrieval is thought to depend on a top-down bias mechanism,
Wagner, Bunge, & Badre
15
wherein task or context representations serve to facilitate activation of weakly associated,
task-relevant knowledge (Figure 6). Relative to automatic retrieval, controlled retrieval
(a) is slower and more effortful, (b) can bias retrieval of task-relevant information even in
the presence of more strongly associated but task-irrelevant information, and (c) can
either directly or indirectly inhibit the retrieval of prepotent, task-irrelevant information.
SEMANTIC RETRIEVAL AND PREFRONTAL CORTEX
Left ventrolateral PFC
lesions result in impairments on semantic tasks that require some form of cognitive
control during semantic or lexical retrieval (e.g., Metzler, 2001; Swick and Knight, 1996;
Thompson-Schill et al., 1998). For instance, although patients with Broca’s aphasia – a
language disorder that follows left PFC damage – show intact semantic priming effects
that are due to automatic retrieval processes (Blumstein et al., 1982; Hagoort, 1997), such
patients fail to show normal levels of semantic priming when cue–target associative
strength is weak or ambiguous (Metzler, 2001; Milberg et al., 1987) or when the context
requires strategic assessment of the utility of primes in cueing upcoming targets (Milberg
et al., 1995). Left PFC lesions also result in difficulty retrieving a weak associate of a
cue when faced with competition from other associated knowledge (Thompson-Schill et
al., 1998). Thus, deficits occur in situations in which healthy individuals likely use
cognitive control to effectively access relevant semantic knowledge. Consistent with this
interpretation, even temporary disruption of anterior left ventrolateral PFC with
transcranial magnetic stimulation results in a performance deficit when controlled access
to semantic knowledge is required, but not when stimuli are processed in a nonsemantic
manner (Devlin et al., 2003).
Wagner, Bunge, & Badre
16
Neuroimaging studies provide complementary high-spatial resolution evidence
that more precisely maps semantic processing functions to ventrolateral PFC. Extensive
imaging data indicate that anterior left ventrolateral PFC (BA 47/45) – which falls rostral
and inferior to the left PFC region associated with phonological WM – is engaged when
access to semantic knowledge is required (Buckner et al., 1995; Fiez, 1997; Poldrack et
al., 1999; Wagner, 1999). For example, greater left PFC activation is observed during the
retrieval of semantic knowledge associated with a word than during word reading (e.g.,
Petersen et al., 1988), and when making semantic relative to nonsemantic decisions about
stimuli (e.g., Gabrieli et al., 1996; Kapur et al., 1994). This is the case even when task
difficulty is greater during nonsemantic processing, indicating that left PFC activation is
modulated by semantic retrieval demands rather than by demands on global attentional
resources (Demb et al., 1995). Thus, extant data implicate anterior left ventrolateral PFC
in some form of cognitive control that facilitates access to task-relevant semantic
knowledge.1
COGNITIVE CONTROL AND RECOVERING MEANING
The precise nature of
this left ventrolateral PFC control process and its role in semantic retrieval remain
controversial (e.g., Badre and Wagner, 2002; Gabrieli et al., 1998; Gold and Buckner,
2002; Thompson-Schill et al., 1997). From one perspective, left ventrolateral PFC
supports controlled semantic retrieval, wherein PFC represents the conceptual context
and biases retrieval of context-relevant information when that information is not
recovered through automatic retrieval processes. By contrast, the selection hypothesis
posits that left ventrolateral PFC does not support semantic retrieval (Thompson-Schill et
Wagner, Bunge, & Badre
17
al., 1997). Rather, retrieval is argued to emerge entirely within posterior neocortex,
where knowledge associated with a given cue is retrieved upon cue presentation through
bottom-up processes. Once representations are retrieved, left PFC serves to select those
that are task-relevant from amidst competing, irrelevant representations.
A critical prediction of the controlled retrieval hypothesis, but not the selection
account, is that increased left PFC activity should occur when the associative strength
between a retrieval cue and target knowledge is weak relative to when it is strong, and
that this should be the case even when using a semantic processing task that requires
minimal selection. A recent fMRI study (Wagner et al., 2001b) tested this prediction,
varying controlled retrieval demands by (a) manipulating the pre-experimental
associative strength between the cue and the correct target and (b) further manipulating
retrieval demands by varying the number of targets to be considered (Figure 7) (cf.,
Thompson-Schill et al., 1997). FMRI results revealed that, even when selection demands
are held constant and to a minimum, activation in anterior left ventrolateral PFC increases
with the number of to-be-considered targets, and, critically, is greater when the
cue–target associative strength is weak rather than strong. Conceptually similar effects of
associative strength on left PFC activation are also seen when subjects make categorical
decisions about less prototypical exemplars (e.g., BIRD-OSTRICH) than about more
prototypical exemplars (e.g., BIRD-ROBIN) (Roskies et al., 2001), and when subjects
make analogical reasoning decisions about weakly associated relative to strongly
associated word pairs (Bunge et al., 2003). Such findings indicate that left ventrolateral
PFC is engaged when target semantic knowledge must be retrieved in a controlled
manner.
Wagner, Bunge, & Badre
18
It should be emphasized that controlled retrieval and selection demands are likely
to be highly correlated, suggesting a possible reconciliation of the two perspectives
(Badre and Wagner, 2002). Specifically, controlled retrieval is necessary when taskrelevant information does not come to mind through automatic, bottom-up processes.
This can occur either because task-relevant information is weakly associated with a
presented retrieval cue, resulting in insufficient bottom-up activation, or because more
strongly associated information competes with recovery of the target knowledge,
resulting in high selection demands. In both cases, a top-down bias signal may be
required to effectively retrieve the relevant representation (Figure 6). Accordingly,
controlled retrieval demands typically increase as selection demands increase, but also
increase due to other factors that make automatic access to relevant information
insufficient or ineffective, such as weak cue–target associative connections.
Cognitive control and priming: On the road to automaticity
The evidence discussed in the preceding section indicates that a controlled retrieval
process interacts with declarative memory to support knowledge recovery, thus guiding
flexible, goal-directed behavior. Because controlled retrieval is one end of a processing
continuum, it is important to understand how knowledge recovery shifts from depending
on more controlled to more automatic retrieval mechanisms. As we discuss in this final
section, neuroimaging studies of repetition priming indicate that reductions in controlled
retrieval demands follow experience-based modifications of semantic memory.
Repetition priming is a nondeclarative (or implicit) form of LTM that is expressed
as an experience-based change in behavior, that is not necessarily accompanied by
Wagner, Bunge, & Badre
19
conscious recollection, and that is preserved following medial-temporal lobe damage
(Gabrieli, 1998; Schacter, 1987a; Squire, 1992). At the behavioral level, priming refers
to instances in which an earlier encounter with a stimulus alters later responding to that
same stimulus or to a related stimulus. Behavioral changes include an increased speed of
responding, increased response accuracy, or biased responding.
Although there are multiple forms of priming, most instances fall into one of two
broad categories: perceptual and conceptual (Roediger and McDermott, 1993).
Perceptual priming refers to an enhanced ability to identify a stimulus due to previous
exposure to the stimulus, whereas conceptual priming refers to facilitated processing of
the meaning of a stimulus or enhanced access to a concept due to prior processing of the
stimulus’s meaning or retrieval of the concept. In this section, we discuss perceptual and
conceptual priming, and describe neuroimaging correlates of these forms of
nondeclarative memory. We then focus on the relation between conceptual priming and
controlled retrieval, emphasizing how priming-related changes in PFC activation mark a
shift from controlled to automatic retrieval.
PERCEPTUAL PRIMING
Perceptual priming is often revealed as (a) an increased
likelihood of completing a perceptual fragment, such as a word stem (e.g., STA___), with
a previously encountered stimulus (e.g., STAMP), or (b) faster and more accurate
identification of degraded stimuli due to prior exposure to those stimuli. Perceptual
priming is sensitive to the degree of perceptual overlap between the initial and repeated
encounter with a stimulus. For example, this form of priming is modality-specific, being
greater when the initial and repeated encounters occur in the same modality, as opposed
Wagner, Bunge, & Badre
20
to different modalities (Jacoby and Dallas, 1981). By contrast, perceptual priming is
comparable irrespective of whether attention at study is focused on the meaning of the
stimulus or on its structural form (Jacoby and Dallas, 1981).
Perceptual priming is thought to reflect learning in perceptual representation
systems (Tulving and Schacter, 1990). Perceptual representation systems encode and
retain pre-semantic perceptual information (in the form of perceptual records) about
encountered words and objects, and are thought to depend on modality-specific sensory
cortices. Perceptual representation systems include a word form system that can be
further differentiated into visual and auditory systems that represent seen and spoken
words, respectively, and a structural representation system that represents the visual form
of objects.
Neuroimaging studies of perceptual priming typically demonstrate decreased
activation in modality-sensitive cortical regions that were engaged during initial stimulus
processing (Schacter and Buckner, 1998; Wagner and Koutstaal, 2002) (Figure 8),
although increases are sometimes observed upon repetition of unfamiliar stimuli (Henson
et al., 2000). Priming-related activation decreases persist across delays of multiple days,
indicating that such experience-based reductions in activation depend on a form of LTM
(van Turennout et al., 2000).
Neural priming effects in human sensory cortex resemble the phenomenon of
repetition suppression in infrahuman primates, wherein repeated exposure to a stimulus
leads to a reduction in the firing rate of neurons in higher-order visual regions
(Desimone, 1996). Given this parallel, one hypothesis is that priming-related activation
reductions in neuroimaging studies reflect a "sharpening" or "tuning" of perceptual
Wagner, Bunge, & Badre
21
representation systems (Wiggs and Martin, 1998). From this perceptive, multiple
neurons are engaged during initial perception of a stimulus, with some coding less
relevant features than others. In the course of this initial processing, the perceptual
representation system is sharpened such that neurons that code unnecessary features of
the stimulus are pruned, resulting in reduced responding by those neurons upon reencounter with the stimulus (Figure 8). This sharpening of the representation may result
in efficient stimulus processing, thus giving rise to facilitated behavior. Although an
intriguing possibility, priming-related activation reductions could also reflect a reduction
in the duration of neural processing due to experience-based strengthening of synaptic
connections (Figure 6), thus resulting in more rapid network settling (Henson and Rugg,
2003).
CONCEPTUAL PRIMING Conceptual priming refers to the increased likelihood of
generating a concept in response to a test cue due to previous processing of the concept
during an unrelated study phase. For example, when participants are presented with a
category cue (e.g., Fruit), and are asked to generate the first few exemplars that come to
mind, the probability of spontaneously generating a given exemplar (e.g., Cherry) is
higher if the exemplar had appeared on a prior study list. Conceptual priming is also
revealed on semantic classification tasks, wherein prior processing of a stimulus's
meaning results in faster classification of the stimulus along a particular semantic
dimension.
In contrast to perceptual priming, conceptual priming is modality-independent.
For example, cross-modality priming (auditory study, visual test) is comparable to
Wagner, Bunge, & Badre
22
within-modality priming (visual study, visual test) on semantic classification tasks
(Vaidya et al., 1997). Moreover, whereas perceptual priming is insensitive to the level of
semantic elaboration during initial processing, conceptual priming is greater when
semantic features of a stimulus are attended during study compared to when perceptual
features are attended. Finally, conceptual priming is intact following focal lesions to
modality-specific sensory cortices, whereas such lesions impair perceptual priming (e.g.,
Gabrieli et al., 1995). By contrast, conceptual priming is impaired in patients with
Alzheimer’s disease (AD), who suffer pathological changes in amodal association areas
of frontal, temporal, and parietal cortex, as well as in the medial-temporal lobe (Heindel
et al., 1989). This pattern suggests that temporal association cortices that represent
semantic knowledge, and perhaps frontal cortices that permit controlled access to this
knowledge, may be important for conceptual priming (Swick and Knight, 1996).
In a pattern that parallels that seen during comparisons of automatic versus
controlled retrieval conditions, neuroimaging investigations of conceptual priming reveal
that the anterior and posterior extents of left ventrolateral PFC show reduced activation
during repeated (primed) relative to initial (unprimed) semantic processing of stimuli
(e.g., Gabrieli et al., 1996; Raichle et al., 1994; Schacter and Buckner, 1998; Wagner et
al., 1997). For example, activation is weaker when making semantic classification
decisions about primed relative to unprimed objects or words (Figure 8), and this is the
case even following study–test retention intervals of over 24 hours (Wagner et al., 2000).
Reductions in left ventrolateral PFC activation are observed in patients with global
amnesia, suggesting that these changes reflect computational benefits deriving from the
nondeclarative memory processes that support conceptual priming at the behavioral level
Wagner, Bunge, & Badre
23
(Buckner and Koutstaal, 1998; Gabrieli et al., 1998). In addition to changes in PFC
activation, conceptual priming is correlated with activation reductions in posterior
cortical regions that may represent long-term conceptual knowledge, including ventral
and lateral temporal cortices (e.g., Blaxton et al., 1996; Raichle et al., 1994). These
outcomes indicate that conceptual priming stems from experience-based "tuning" of
semantic memory, mirroring at the conceptual level the sort of processes hypothesized to
underlie perceptual priming.
The changes in semantic memory that support conceptual priming could emerge
in a number of ways. For example, the tuning of semantic memory may constitute a
sharpening of conceptual representations, perhaps through retrieval-induced suppression,
whereby the initial retrieval of relevant stimulus features serves to suppress or inhibit less
relevant features (Anderson et al., 1994). Suppression of irrelevant features would
reduce the degree to which these features compete for retrieval, thus facilitating the
subsequent recovery of recently retrieved knowledge. In addition, modifications in
semantic memory are likely to emerge at least partially from a strengthening of the
connections or associations between a response cue and the initially retrieved relevant
features associated with the cue/concept. This strengthening of cue–target associations
facilitates subsequent retrieval of these target representations by increasing the effects of
bottom-up inputs on knowledge recovery.
CONTROLLED RETRIEVAL AND CONCEPTUAL PRIMING The consistent
observation that reductions in left PFC activation accompany conceptual priming
suggests that cognitive control demands can decline following experience-based changes
Wagner, Bunge, & Badre
24
in semantic memory (Figure 6). That is, PFC priming effects may mark the beginnings
of a transition from reliance on more controlled to more automatic retrieval mechanisms
(Demb et al., 1995; Fletcher et al., 2000; Raichle et al., 1994). From this perspective,
controlled semantic retrieval tunes semantic memory, strengthening the representations of
initially recovered knowledge, and perhaps weakening those of competing knowledge.
Critically, this increased accessibility of recently relevant knowledge reduces the
computational demands on controlled retrieval when this knowledge must be retrieved in
the future. Such reductions in controlled retrieval demands reflect a) an increased
automaticity of target knowledge retrieval due to strengthened cue–target associations
(Raichle et al., 1994), and b) reduction of competition from task-irrelevant
representations (Thompson-Schill et al., 1999b). Thus, these neural priming effects
illustrate the point that interactions between cognitive control and LTM are bidirectional: cognitive control impacts retrieval from LTM, with the resulting changes in
LTM impacting subsequent demands on cognitive control.
Conclusions
Memory systems can produce outputs based on bottom-up mechanisms that unfold
automatically in response to inputs, although in many situations these outputs are
insufficient to support flexible cognition. In such instances, cognitive control processes
must be recruited to retrieve, maintain, and manipulate relevant knowledge and to select
against or resolve interference from competing representations. As reviewed in this
chapter, the past decade of cognitive neuroscience research highlights that an
understanding of cognitive control processes requires consideration of their intimate
Wagner, Bunge, & Badre
25
relation with memory. Future investigations undoubtedly will continue to delineate the
elemental forms of control and their dependence on prefrontal cortex. In so doing, these
efforts promise to provide important insights into the workings of memory, revealing
how memory can be brought to bear, in a controlled manner, to support flexible behavior.
Wagner, Bunge, & Badre
26
Wagner, Bunge, & Badre
27
References
Alain C, Arnott SR, Hevenor S, Graham S, Grady CL. "What" and "where" in the human
auditory system. Proceedings of the National Academy of Sciences, USA 2001;
98: 12301-12306.
Anderson MC, Bjork RA, Bjork EL. Remembering can cause forgetting: Retrieval
dynamics in long-term memory. Journal of Experimental Psychology: Learning,
Memory, and Cognition 1994; 20: 1063-1087.
Aron AR, Fletcher PC, Bullmore ET, Sahakian BJ, Robbins TW. Stop-signal inhibition
disrupted by damage to right inferior frontal gyrus in humans. Nature
Neuroscience 2003; 6: 115-116.
Awh E, Jonides J. Spatial working memory and spatial selective attention. In:
Parasuraman R, editor. The Attentive Brain. Cambridge, MA: MIT Press, 1998:
353-380.
Awh E, Jonides J, Smith EE, Schumacher EH, Koeppe RA, Katz S. Dissociation of
storage and rehearsal in verbal working memory: Evidence from PET.
Psychological Science 1996; 7: 25-31.
Baddeley AD. Working Memory. Oxford: Oxford University Press, 1986.
Baddeley AD. Working memory: The interface between memory and cognition. In:
Schacter DL and Tulving E, editors. Memory Systems 1994. Cambridge, MA:
The MIT Press, 1994: 351-367.
Baddeley AD, Hitch GJ. Working memory. In: Bower GH, editor. The Psychology of
Learning and Motivation: Advances in Research and Theory. Vol 8. New York:
Academic Press, 1974: 47-89.
Badre D, Wagner AD. Semantic retrieval, mnemonic control, and prefrontal cortex.
Behavioral and Cognitive Neuroscience Reviews 2002; 1: 206-218.
Badre D, Wagner AD. Selection and conflict monitoring in prefrontal cortex: Assessing
the specificity of cognitive control. in prep 2003.
Blaxton TA, Bookheimer SY, Zeffiro TA, Figlozzi CM, Gaillard WD, Theodore WH.
Functional mapping of human memory using PET: Comparisons of conceptual
and perceptual tasks. Canadian Journal of Experimental Psychology 1996; 50: 4256.
Blumstein SE, Milberg W, Shrier R. Semantic processing in aphasia: Evidence from an
auditory lexical decision task. Brain and Language 1982; 17: 301-315.
Botvinick MM, Braver TS, Barch DM, Carter CS, Cohen JD. Conflict monitoring and
cognitive control. Psychological Review 2001; 108: 624-652.
Braver TS, Bongiolatti SR. The role of frontopolar cortex in subgoal processing during
working memory. NeuroImage 2002; 15: 523-536.
Buckner RL, Koutstaal W. Functional neuroimaging studies of encoding, priming, and
explicit memory retrieval. Proceedings of the National Academy of Sciences,
USA 1998; 95: 891-898.
Buckner RL, Raichle ME, Petersen SE. Dissociation of human prefrontal cortical areas
across different speech production tasks and gender groups. Journal of
Neurophysiology 1995; 74: 2163-2173.
Bunge SA, Badre D, Wagner AD. Analogical reasoning and prefrontal cortex: Evidence
for separable retrieval and integration mechanisms. submitted 2003.
Wagner, Bunge, & Badre
28
Bunge SA, Hazeltine E, Scanlon MD, Rosen AC, Gabrieli JD. Dissociable contributions
of prefrontal and parietal cortices to response selection. NeuroImage 2002; 17:
1562-1571.
Bunge SA, Ochsner KN, Desmond JE, Glover GH, Gabrieli JD. Prefrontal regions
involved in keeping information in and out of mind. Brain 2001; 124: 2074-2086.
Caramazza A, Shelton JR. Domain-specific knowledge systems in the brain: The
animate-inanimate distinction. Journal of Cognitive Neuroscience 1998; 10: 1-34.
Carlesimo GA, Perri R, Turriziani P, Tomaiuolo F, Caltagirone C. Remembering what
but not where: Independence of spatial and visual working memory in the human
brain. Cortex 2001; 37: 519-534.
Chafee MV, Goldman-Rakic PS. Inactivation of parietal and prefrontal cortex reveals
interdependence of neural activity during memory-guided saccades. Journal of
Neurophysiology 2000; 83: 1550-1566.
Chao LL, Haxby JV, Martin A. Attribute-based neural substrates in temporal cortex for
perceiving and knowing about objects. Nature Neuroscience 1999; 2: 913-919.
Chein JM, Fiez JA. Dissociation of verbal working memory system components using a
delayed serial recall task. Cerebral Cortex 2001; 11: 1003-1014.
Christoff K, Prabhakaran V, Dorfman J, Zhao Z, Kroger JK, Holyoak KJ, et al.
Rostrolateral prefrontal cortex involvement in relational integration during
reasoning. NeuroImage 2001; 14: 1136-1149.
Cohen JD, Braver TS, O'Reilly RC. A computational approach to prefrontal cortex,
cognitive control and schizophrenia: Recent developments and current challenges.
Philosophical Transactions of the Royal Society of London (Biological Science)
1996; 351: 1515-1527.
Cohen JD, Dunbar K, McClelland JL. On the control of automatic processes: A parallel
distributed processing account of the Stroop effect. Psychological Review 1990;
97: 332-361.
Courtney SM, Petit L, Maisog JM, Ungerleider LG, Haxby JV. An area specialized for
spatial working memory in human frontal cortex. Science 1998; 279: 1347-1351.
D'Esposito M, Aguirre GK, Zarahn E, Ballard D, Shin RK, Lease J. Functional MRI
studies of spatial and nonspatial working memory. Cognitive Brain Research
1998; 7: 1-13.
D'Esposito M, Postle BR, Ballard D, Lease J. Maintenance versus manipulation of
information held in working memory: An event-related fMRI study. Brain and
Cognition 1999a; 41: 66-86.
D'Esposito M, Postle BR, Jonides J, Smith EE. The neural substrate and temporal
dynamics of interference effects in working memory as revealed by event-related
functional MRI. Proceedings of the National Academy of Sciences, USA 1999b;
96: 7514-7519.
Demb JB, Desmond JE, Wagner AD, Vaidya CJ, Glover GH, Gabrieli JD. Semantic
encoding and retrieval in the left inferior prefrontal cortex: A functional MRI
study of task difficulty and process specificity. Journal of Neuroscience 1995; 15:
5870-5878.
Desimone R. Neural mechanisms for visual memory and their role in attention.
Proceedings of the National Academy of Sciences, USA 1996; 93: 13494-13499.
Wagner, Bunge, & Badre
29
Desimone R, Duncan J. Neural mechanisms of selective visual attention. Annual Review
of Neuroscience 1995; 18: 193-222.
Desmond JE, Gabrieli JD, Wagner AD, Ginier BL, Glover GH. Lobular patterns of
cerebellar activation in verbal working-memory and finger-tapping tasks as
revealed by functional MRI. Journal of Neuroscience 1997; 17: 9675-9685.
Devlin JT, Matthews PM, Rushworth MF. Semantic processing in the left inferior
prefrontal cortex: A combined functional magnetic resonance imaging and
transcranial magnetic stimulation study. Journal of Cognitive Neuroscience 2003;
15: 71-84.
Farah MJ. Is visual memory really visual? Overlooked evidence from neuropsychology.
Psychological Review 1988; 95: 307-317.
Farah MJ, McClelland JL. A computational model of semantic memory impairment:
Modality specificity and emergent category specificity. Journal of Experimental
Psychology: General 1991; 120: 339-357.
Fiez JA. Phonology, semantics, and the role of the left inferior prefrontal cortex. Human
Brain Mapping 1997; 5: 79-83.
Fiez JA, Raife EA, Balota DA, Schwarz JP, Raichle ME, Petersen SE. A positron
emission tomography study of the short-term maintenance of verbal information.
Journal of Neuroscience 1996; 16: 808-822.
Fletcher PC, Henson RN. Frontal lobes and human memory: Insights from functional
neuroimaging. Brain 2001; 124: 849-881.
Fletcher PC, Shallice T, Dolan RJ. "Sculpting the response space": An account of left
prefrontal activation at encoding. NeuroImage 2000; 12: 404-417.
Fuster JM. The prefrontal cortex: Anatomy, physiology, and neuropsychology of the
frontal lobe. Philadelphia: Lippincott-Raven, 1997.
Fuster JM, Alexander GE. Neuron activity related to short-term memory. Science 1971;
173: 652-654.
Gabrieli JD. Cognitive neuroscience of human memory. Annual Review of Psychology
1998; 49: 87-115.
Gabrieli JD, Poldrack RA, Desmond JE. The role of left prefrontal cortex in language and
memory. Proceedings of the National Academy of Sciences, USA 1998; 95: 906913.
Gabrieli JDE, Desmond JE, Demb JB, Wagner AD, Stone MV, Vaidya CJ, et al.
Functional magnetic resonance imaging of semantic memory processes in the
frontal lobes. Psychological Science 1996; 7: 278-283.
Gabrieli JDE, Fleischman DA, Keane MA, Reminger SL, Morrell F. Double dissociation
between memory systems underlying explicit and implicit memory in the human
brain. Psychological Science 1995; 6: 76-82.
Garavan H, Ross TJ, Li SJ, Stein EA. A parametric manipulation of central executive
functioning. Cerebral Cortex 2000; 10: 585-592.
Garavan H, Ross TJ, Stein EA. Right hemispheric dominance of inhibitory control: An
event-related functional MRI study. Proceedings of the National Academy of
Sciences, USA 1999; 96: 8301-8306.
Gold BT, Buckner RL. Common prefrontal regions coactivate with dissociable posterior
regions during controlled semantic and phonological tasks. Neuron 2002; 35: 803812.
Wagner, Bunge, & Badre
30
Goldman-Rakic PS. Circuitry of primate prefrontal cortex and regulation of behavior by
representational memory. In: Plum F and Mountcastle V, editors. Handbook of
Physiology, Section 1: The Nervous System, Vol. V. Higher functions of the
Brain, Part 1. Bethesda, MD: American Physiological Society, 1987: 373-417.
Hagoort P. Semantic priming in Broca's aphasics at short SOA: No support for an
automatic semantic access deficit. Brain and Language 1997; 56: 287-300.
Haxby JV, Petit L, Ungerleider LG, Courtney SM. Distinguishing the functional roles of
multiple regions in distributed neural systems for visual working memory.
NeuroImage 2000; 11: 380-391.
Hazeltine E, Bunge SA, Scanlon MD, Rosen AC, Gabrieli JD. Material-dependent and
material-independent selection processes in the frontal lobes: An event-related
fMRI investigation of response competition. Neuropsychologia in press.
Heindel WC, Salmon DP, Shults CW, Wallicke PA, Butters N. Neuropsychological
evidence for multiple implicit memory systems: A comparison of Alzheimer’s,
Huntington's and Parkinson’s disease patients. Journal of Neuroscience 1989; 9:
582-587.
Henson R, Shallice T, Dolan R. Neuroimaging evidence for dissociable forms of
repetition priming. Science 2000; 287: 1269-1272.
Henson RNA, Rugg MD. Neural response suppression, haemodynamic repetition effects,
and behavioral priming. Neuropsychologia 2003; 41: 263-270.
Jacoby LL, Dallas M. On the relationship between autobiographical memory and
perceptual learning. Journal of Experimental Psychology: General 1981; 110:
306-340.
Jonides J, Schumacher EH, Smith EE, Koeppe RA, Awh E, Reuter-Lorenz PA, et al. The
role of parietal cortex in verbal working memory. Journal of Neuroscience 1998a;
18: 5026-5034.
Jonides J, Smith EE, Marshuetz C, Koeppe RA. Inhibition in verbal working memory
revealed by brain activation. Proceedings of the National Academy of Sciences,
USA 1998b; 95: 8410-8413.
Kapur S, Rose R, Liddle PF, Zipursky RB, Brown GM, Stuss D, et al. The role of the left
prefrontal cortex in verbal processing: Semantic processing or willed action?
NeuroReport 1994; 5: 2193-2196.
Kellenbach ML, Brett M, Patterson K. Actions speak louder than functions: The
importance of manipulability and action in tool representation. Journal of
Cognitive Neuroscience 2003; 15: 30-46.
Koechlin E, Basso G, Pietrini P, Panzer S, Grafman J. The role of the anterior prefrontal
cortex in human cognition. Nature 1999; 399: 148-151.
Konishi S, Nakajima K, Uchida I, Kameyama M, Nakahara K, Sekihara K, et al.
Transient activation of inferior prefrontal cortex during cognitive set shifting.
Nature Neuroscience 1998a; 1: 80-84.
Konishi S, Nakajima K, Uchida I, Sekihara K, Miyashita Y. No-go dominant brain
activity in human inferior prefrontal cortex revealed by functional magnetic
resonance imaging. European Journal of Neuroscience 1998b; 10: 1209-1213.
Kourtzi Z, Kanwisher N. Activation in human MT/MST by static images with implied
motion. Journal of Cognitive Neuroscience 2000; 12: 48-55.
Wagner, Bunge, & Badre
31
Koutstaal W, Wagner AD, Rotte M, Maril A, Buckner RL, Schacter DL. Perceptual
specificity in visual object priming: Functional magnetic resonance imaging
evidence for a laterality difference in fusiform cortex. Neuropsychologia 2001;
39: 184-199.
Kroger JK, Sabb FW, Fales CL, Bookheimer SY, Cohen MS, Holyoak KJ. Recruitment
of anterior dorsolateral prefrontal cortex in human reasoning: A parametric study
of relational complexity. Cerebral Cortex 2002; 12: 477-485.
MacDonald AW, 3rd, Cohen JD, Stenger VA, Carter CS. Dissociating the role of the
dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Science
2000; 288: 1835-8.
Martin A, Chao LL. Semantic memory and the brain: Structure and processes. Current
Opinion in Neurobiology 2001; 11: 194-201.
Martin A, Haxby JV, Lalonde FM, Wiggs CL, Ungerleider LG. Discrete cortical regions
associated with knowledge of color and knowledge of action. Science 1995; 270:
102-105.
Martin A, Wiggs CL, Ungerleider LG, Haxby JV. Neural correlates of category-specific
knowledge. Nature 1996; 379: 649-652.
Martin RC, Shelton JR, Yaffee LS. Language processing and working memory:
Neuropsychological evidence for separate phonological and semantic capacities.
Journal of Memory and Language 1994; 33: 83-111.
McCarthy G, Puce A, Constable RT, Krystal JH, Gore JC, Goldman-Rakic P. Activation
of human prefrontal cortex during spatial and nonspatial working memory tasks
measured by functional mri. Cerebral Cortex 1996; 6: 600-611.
Metzler C. Effects of left frontal lesions on the selection of context-appropriate meanings.
Neuropsychology 2001; 15: 315-328.
Milberg W, Blumstein SE, Dworetzky B. Processing of lexical ambiguities in aphasia.
Brain and Language 1987; 31: 138-150.
Milberg WP, Blumsteing SE, Katz D, Gershberg F, Brown T. Semantic facilitation in
aphasia: Effects of time and expectancy. Journal of Cognitive Neuroscience 1995;
7: 33-50.
Milham MP, Banich MT, Webb A, Barad V, Cohen NJ, Wszalek T, et al. The relative
involvement of anterior cingulate and prefrontal cortex in attentional control
depends on nature of conflict. Cognitive Brain Research 2001; 12: 467-473.
Miller EK, Cohen JD. An integrative theory of prefrontal cortex function. Annual Review
of Neuroscience 2001; 24: 167-202.
Miller GA, Galanter E, Pribram KH. Plans and the Structure of Behavior. New York:
Holt, Rinehart & Winston, 1960.
Neely JH. Semantic priming effects in visual word recognition: A selective review of
current findings and theories. In: Besner D and Humphreys GW, editors. Basic
Processing in Reading: Visual Word Recognition. Hillsdale, NJ: Lawrence
Erlbaum Associates, 1991: 264-336.
Nystrom LE, Braver TS, Sabb FW, Delgado MR, Noll DC, Cohen JD. Working memory
for letters, shapes, and locations: fMRI evidence against stimulus-based regional
organization in human prefrontal cortex. NeuroImage 2000; 11: 424-446.
Wagner, Bunge, & Badre
32
Owen AM, Evans AC, Petrides M. Evidence for a two-stage model of spatial working
memory processing within the lateral frontal cortex: A positron emission
tomography study. Cerebral Cortex 1996; 6: 31-38.
Paulesu E, Frith CD, Frackowiak RS. The neural correlates of the verbal component of
working memory. Nature 1993; 362: 342-345.
Petersen SE, Fox PT, Posner MI, Mintun M, Raichle ME. Positron emission tomographic
studies of the cortical anatomy of single-word processing. Nature 1988; 331: 585589.
Petrides M. Frontal lobes and behaviour. Current Opinion in Neurobiology 1994; 4: 207211.
Pilgrim LK, Fadili J, Fletcher P, Tyler LK. Overcoming confounds of stimulus blocking:
An event-related fMRI design of semantic processing. NeuroImage 2002; 16:
713-723.
Poldrack RA, Wagner AD, Prull MW, Desmond JE, Glover GH, Gabrieli JDE.
Functional specialization for semantic and phonological processing in the left
inferior frontal cortex. NeuroImage 1999; 10: 15-35.
Postle BR, Berger JS, D'Esposito M. Functional neuroanatomical double dissociation of
mnemonic and executive control processes contributing to working memory
performance. Proceedings of the National Academy of Sciences, USA 1999; 96:
12959-12964.
Prabhakaran V, Narayanan K, Zhao Z, Gabrieli JD. Integration of diverse information in
working memory within the frontal lobe. Nat Neurosci 2000; 3: 85-90.
Raichle ME, Fiez JA, Videen TO, Macleod AMK, Pardo JV, Fox PT, et al. Practicerelated changes in human brain functional anatomy during nonmotor learning.
Cerebral Cortex 1994; 4: 8-26.
Rainer G, Asaad WF, Miller EK. Memory fields of neurons in the primate prefrontal
cortex. Proceedings of the National Academy of Sciences, USA 1998; 95: 1500815013.
Rao SC, Rainer G, Miller EK. Integration of what and where in the primate prefrontal
cortex. Science 1997; 276: 821-824.
Raye CL, Johnson MK, Mitchell KJ, Reeder JA, Greene EJ. Neuroimaging a single
thought: Dorsolateral PFC activity associated with refreshing just-activated
information. NeuroImage 2002; 15: 447-453.
Roediger HL, McDermott K. Implicit memory in normal human subjects. In: Boller F
and Grafman J, editors. Handbook of Neuropsychology. Vol 8. New York:
Elsevier, 1993: 63-131.
Roskies AL, Fiez JA, Balota DA, Raichle ME, Petersen SE. Task-dependent modulation
of regions in the left inferior frontal cortex during semantic processing. Journal of
Cognitive Neuroscience 2001; 13: 829-843.
Rugg MD, Wilding EL. Retrieval processing and episodic memory. Trends in Cognitive
Science 2000; 4: 108-115.
Rypma B, Prabhakaran V, Desmond JE, Glover GH, Gabrieli JD. Load-dependent roles
of frontal brain regions in the maintenance of working memory. NeuroImage
1999; 9: 16-26.
Wagner, Bunge, & Badre
33
Sala JB, Rama P, Courtney SM. Functional topography of a distributed neural system for
spatial and nonspatial information maintenance in working memory.
Neuropsychologia 2003; 41: 341-356.
Salmon E, Van Der Linden M, Collette F, Delfiore G, Maquet P, Degueldre C, et al.
Regional brain activity during working memory tasks. Brain 1996; 119: 16171625.
Schacter DL. Implicit memory: History and current status. Journal of Experimental
Psychology: Learning, Memory, and Cognition 1987a; 13: 501-518.
Schacter DL. Memory, amnesia, and frontal lobe dysfunction. Psychobiology 1987b; 15:
21-36.
Schacter DL, Buckner RL. Priming and the brain. Neuron 1998; 20: 185-195.
Shallice T. From Neuropsychology to Mental Structure. New York: Cambridge
University Press, 1988.
Shimamura AP. Memory and frontal lobe function. In: Gazzaniga MS, editor. The
Cognitive Neurosciences. Cambridge, MA: MIT Press, 1995: 803-813.
Smith EE, Jonides J. Working memory in humans: Neuropsychological evidence. In:
Gazzaniga MS, editor. The Cognitive Neurosciences. Cambridge, MA: MIT
Press, 1995: 1009-1020.
Smith EE, Jonides J. Neuroimaging analyses of human working memory. Proceedings of
the National Academy of Sciences, USA 1998; 95: 12061-12068.
Smith EE, Jonides J. Storage and executive processes in the frontal lobes. Science 1999;
283: 1657-1661.
Smith EE, Medin DL. Concepts and Categories. Cambridge, MA: Harvard University
Press, 1981.
Squire LR. Memory and the hippocampus: A synthesis from findings with rats, monkeys,
and humans. Psychological Review 1992; 99: 195-231.
Stuss DT, Benson DF. Neuropsychological studies of the frontal lobes. Psychological
Bulletin 1984; 95: 3-28.
Swick D, Knight RT. Is prefrontal cortex involved in cued recall? A neuropsychological
test of PET findings. Neuropsychologia 1996; 34: 1019-1028.
Sylvester CY, Shimamura AP. Evidence for intact semantic representations in patients
with frontal lobe lesions. Neuropsychology 2002; 16: 197-207.
Thompson-Schill SL. Neuroimaging studies of semantic memory: Inferring "how" from
"where". Neuropsychologia 2003; 41: 280-292.
Thompson-Schill SL, Aguirre GK, D'Esposito M, Farah MJ. A neural basis for category
and modality specificity of semantic knowledge. Neuropsychologia 1999a; 37:
671-676.
Thompson-Schill SL, D'Esposito M, Aguirre GK, Farah MJ. Role of left inferior
prefrontal cortex in retrieval of semantic knowledge: A reevaluation. Proceedings
of the National Academy of Sciences, USA 1997; 94: 14792-14797.
Thompson-Schill SL, D'Esposito M, Kan IP. Effects of repetition and competition on
activity in left prefrontal cortex during word generation. Neuron 1999b; 23: 513522.
Thompson-Schill SL, Swick D, Farah MJ, D'Esposito M, Kan IP, Knight RT. Verb
generation in patients with focal frontal lesions: A neuropsychological test of
Wagner, Bunge, & Badre
34
neuroimaging findings. Proceedings of the National Academy of Sciences, USA
1998; 95: 15855-15860.
Towse JN, Houston-Price CMT. Reflections on the concept of working memory. In:
Andrade J, editor. Working memory in perspective. Hove, England: Psychology
Press, 2001: 240-260.
Tresch MC, Sinnamon HM, Seamon JG. Double dissociation of spatial and object visual
memory: Evidence from selective interference in intact human subjects.
Neuropsychologia 1993; 31: 211-219.
Tulving E. Episodic and semantic memory. In: Tulving E and Donaldson W, editors.
Organization of memory. New York: Academic Press, 1972: 382-403.
Tulving E, Schacter DL. Priming and human memory systems. Science 1990; 247: 301306.
Vaidya CJ, Gabrieli JD, Keane MM, Monti LA, Gutierrez-Rivas H, Zarella MM.
Evidence for multiple mechanisms of conceptual priming on implicit memory
tests. Journal of Experimental Psychology: Learning, Memory, and Cognition
1997; 23: 1324-1343.
van Turennout M, Ellmore T, Martin A. Long-lasting cortical plasticity in the object
naming system. Nature Neuroscience 2000; 3: 1329-1334.
van Veen V, Carter CS. The anterior cingulate as a conflict monitor: fMRI and ERP
studies. Physiology and Behavior 2002; 77: 477-482.
van Veen V, Cohen JD, Botvinick MM, Stenger VA, Carter CS. Anterior cingulate
cortex, conflict monitoring, and levels of processing. NeuroImage 2001; 14:
1302-1308.
Veltman DJ, Rombouts SA, Dolan RJ. Maintenance versus manipulation in verbal
working memory revisited: An fMRI study. NeuroImage 2003; 18: 247-256.
Wagner AD. Working memory contributions to human learning and remembering.
Neuron 1999; 22: 19-22.
Wagner AD. Cognitive control and episodic memory: Contributions from prefrontal
cortex. In: Squire LR and Schacter DL, editors. Neuropsychology of Memory.
New York: Guilford Press, 2002: 174-192.
Wagner AD, Desmond JE, Demb JB, Glover GH, Gabrieli JDE. Semantic repetition
priming for verbal and pictorial knowledge: A functional MRI study of left
inferior prefrontal cortex. Journal of Cognitive Neuroscience 1997; 9: 714-726.
Wagner AD, Koutstaal W. Priming. In: Ramachandran VS, editor. Encyclopedia of the
Human Brain. Vol 4. San Diego: Academic Press, 2002: 27-46.
Wagner AD, Maril A, Bjork RA, Schacter DL. Prefrontal contributions to executive
control: fMRI evidence for functional distinctions within lateral prefrontal cortex.
NeuroImage 2001a; 14: 1337-1347.
Wagner AD, Maril A, Schacter DL. Interactions between forms of memory: When
priming hinders new learning. Journal of Cognitive Neuroscience 2000; 12:S2:
52-60.
Wagner AD, Paré-Blagoev EJ, Clark J, Poldrack RA. Recovering meaning: Left
prefrontal cortex guides controlled semantic retrieval. Neuron 2001b; 31: 329338.
Wiggs CL, Martin A. Properties and mechanisms of perceptual priming. Current Opinion
in Neurobiology 1998; 8: 227-233.
Wagner, Bunge, & Badre
35
Wilson FAW, Scalaidhe SPO, Goldman-Rakic PS. Dissociation of object and spatial
processing domains in primate prefrontal cortex. Science 1993; 260: 1955-1958.
Wagner, Bunge, & Badre
36
Figure captions
Figure 1
Neural correlates and component processes of phonological WM. (A)
Left VLPFC and posterior parietal cortex demonstrate greater activation during the
encoding and maintenance of visually presented words. (Data are from Badre and
Wagner, 2003). (B) Component processes supporting phonological WM and their neural
substrates. Visually presented information is translated to phonological representations,
which are actively represented (stored) in posterior parietal cortex. These representations
passively decay over the course of seconds in the absence of subvocal rehearsal, which is
mediated by left VLPFC (Broca’s area). Recurrent loops between left VLPFC and
posterior parietal cortex allow information to be actively maintained for a prolonged
period. (Adapted from Smith and Jonides, 1998).
Figure 2
PFC neurons maintain object and spatial information, and the integration
of object and space. (A) Delayed Match-to-Sample task. An initial sample is presented
and must be maintained over a delay until a "matching" test stimulus appears. In this
variant, the monkey must remember the sample stimulus ("what") and the location of the
stimulus in space ("where"), and respond by releasing a lever only to stimuli that "match"
on both dimensions. (B) Recording sites where neurons with what, where, or what-andwhere delay activity were found. The arcuate sulcus (A.S.) and principal sulcus (P.S.) are
delineated. (From Rainer et al., 1998).
Figure 3
Model of PFC and ACC involvement during performance of the Stroop
task. Circles represent processing units, which correspond to a population of neurons
assumed to code a given piece of information. Lines represent connections between
units, with heavier lines indicating stronger connections. Looped connections with black
circles indicate mutual inhibition among units within that layer. Subjects must name the
color that a word is printed in, rather than read the word. During presentation of a
conflict stimulus (the word ‘GREEN’ displayed in red ink), the "colors" context unit is
activated in PFC (indicated by gray fill), representing the current intent to name the color.
This passes activation to the intermediate units in the color-naming pathway (indicated by
arrows), which increases the activation of those units (indicated by larger size), and
Wagner, Bunge, & Badre
37
biases processing in favor of activity flowing along this pathway. This biasing effect
favors activation of the response unit corresponding to the color input, even though the
connection weights in this pathway are weaker than in the word pathway. Importantly,
the need for this top-down bias from PFC is initially detected by ACC, which computes
the level of conflict (or the presence of simultaneously active representations in the
response layer). (Adapted from Botvinick et al., 2001; Miller and Cohen, 2001).
Figure 4
Hemodynamic responses in left VLPFC and right DLPFC during WM
maintenance alone or maintenance with manipulation. During each 8-sec trial (indicated
by black box), subjects either rehearsed a triplet of nouns (‘maintenance’ task) or
maintained the triplet and reordered the words along a semantic dimension. The
manipulation task recruited both VLPFC and DLPFC, whereas the maintenance task
primarily recruited VLPFC. (Adapted from Wagner et al., 2001a).
Figure 5
Multidimensional representation of semantic space and functional
neuroanatomic dissociations in occipito-temporal cortices. (A) Two-dimensional
semantic space for controls and patients with frontal lesions as revealed by
multidimensional scaling. The close correspondence between patients and controls
provides strong evidence that frontal cortex does not support storage of semantic
knowledge. (From Sylvester and Shimamura, 2002). (B) Group-averaged statistical
maps show differential activation across semantic domains in distinct occipito-temporal
cortices. Lateral fusiform (1) and right superior temporal (4) cortices show greater
activation during the processing of animals versus tools (white), whereas medial fusiform
(2) and left middle to inferior temporal (3) cortices show preference for tools over
animals (black). (From Martin and Chao, 2001).
Figure 6
Two routes to the retrieval of semantic knowledge. (A) Schematic
representation of automatic retrieval processes that support classification of a stimulus
along a semantic dimension (here, abstract/concrete). Features are connected by
associative links of varying strengths. The presentation of a retrieval cue (e.g.,
“GIRAFFE”) results in bottom-up activation of strongly associated representations
Wagner, Bunge, & Badre
38
(denoted by think lines). When the intrinsic connections are sufficient to retrieve relevant
representations –– due to a strong cue–target association, low competition from irrelevant
representations, or priming due to prior retrieval –– limited demands are placed on topdown (controlled) retrieval processes. (B) Schematic representation of controlled
retrieval. Due to weak pre-existing associative links between the cue and relevant
knowledge, competition from irrelevant information, or the absence of priming,
knowledge necessary to meet task demands may not be successfully retrieved through
automatic processes. In this case, a top-down bias mechanism, depending on left PFC
regions that represent the conceptual context, facilitates retrieval of task-relevant
information.
Figure 7
Left ventrolateral PFC is differentially engaged when controlled semantic
retrieval demands are high. (A) Semantic decision task designed to manipulate
controlled retrieval demands. On each trial, a cue (top word) was presented above either
two or four target words. Participants determined which of the targets was most globally
related to the cue word, with the correct response (circled) being either a strong or a weak
associate of the cue. The factors of cue–target associative strength and number of targets
were crossed. (B) FMRI revealed greater activation in left ventrolateral PFC (circled) as
controlled retrieval demands increased either due to having to retrieve information about
4 versus 2 possible targets, or due to a weak versus strong pre-existing associative
relation between the cue and the correct target. (Data are from Wagner et al., 2001b).
Figure 8
Repetition priming paradigm, neural priming effects, and a hypothesized
neural “sharpening” mechanism. (A) On priming tasks, subjects initially study a set of
stimuli; in this instance, making a semantic decision (size judgment) about visually
presented objects. Subsequently, during the critical test, semantic decisions are made
about previously studied (primed) and novel (unprimed) stimuli. (B) FMRI scanning
during the critical test revealed priming-related activation reductions in fusiform cortex
(circled) and left ventrolateral PFC (arrow); the former reflect neural correlates of
perceptual priming and the latter reflect correlates of conceptual priming. (Data are from
Koutstaal et al., 2001). (C) Hypothesized experience-based changes in a neural network
Wagner, Bunge, & Badre
39
representing visual object features. Initially, neurons coding for relevant and irrelevant
features respond during object processing, but each encounter serves to “prune” or
“sharpen” the representation such that only neurons coding relevant features subsequently
fire. Such a change may result in more efficient perception (i.e., behavioral priming) and
an overall reduction in the mean firing rate of a population of neurons (i.e., repetition
suppression or neural priming). (From Wiggs and Martin, 1998).
Note
1
Other evidence indicates that functional distinctions exist within left PFC. One putative distinction is that
between anterior and posterior left ventrolateral PFC based on semantic and phonological processing
demands (for discussion, see Bookheimer, 2002; Buckner et al., 1995;Fiez, 1997;Gabrieli et al., 1998; Gold
and Buckner, 2002 Poldrack et al., 1999;Wagner et al., 2000).