Semantic Memory and Language Processing:

A Primer
Sharon M. Antonucci, Ph.D.,1 and Jamie Reilly, Ph.D.2

ABSTRACT

Semantic memory refers to our long-term knowledge of word

and object meaning. There is increasing evidence that rather than being
a passive warehouse of knowledge, semantic memory is a dynamic
system whose effectiveness relies on the coordination of multiple
components distributed across a large network of cortical regions.
Damage to one or more of these components produces distinct profiles
of impairment in aphasia and dementia. Furthermore, such differences
are associated with different responses to behavioral treatment. That is,
effective treatment for semantically based language disorders in aphasia
may have very limited success in dementia. We argue that treatment
specificity demands a comprehensive understanding of the structure of
semantic memory and the nature of its compromise. Here, we review
several neuroanatomically informed theories of semantic organization
with respect to the effects of semantic impairment on language
processing in aphasia and neurodegenerative disease.

KEYWORDS: Semantic memory, dementia, language, aphasia,

memory

Learning Outcomes: As a result of this activity, the reader will be able to (1) describe the components of
semantic memory, (2) distinguish among cognitive models of semantic memory, and (3) identify the common
neural substrates of semantic processing.

M ost of us have at one time heard sions immediately come to mind when we hear
descriptions of another person couched in that another person has memory impairment.
terms of memory difficulties (e.g., Grandpa One might expect to meet a disorganized
has trouble with his memory). Several impres- individual who constantly forgets his keys

1
Department of Speech-Language Pathology and 10012 (e-mail: sma208@nyu.edu).
Audiology, New York University, New York, New Semantic Memory and Language Processing in
York; 2Departments of Communicative Disorders and Aphasia and Dementia; Guest Editor, Jamie Reilly,
Neurology, University of Florida, Gainesville, Florida; Ph.D.
Brain Rehabilitation Research Center, Malcom Randall Semin Speech Lang 2008;29:5–17. Copyright #
Veterans Affairs Medical Center, Gainesville, Florida. 2008 by Thieme Medical Publishers, Inc., 333 Seventh
Address for correspondence and reprint requests: Avenue, New York, NY 10001, USA. Tel: +1(212)
Sharon M. Antonucci, Ph.D., Department of Speech- 584-4662.
Language Pathology and Audiology, New York DOI 10.1055/s-2008-1061621. ISSN 0734-0478.
University, 665 Broadway, 9th floor, New York, NY
5
6 SEMINARS IN SPEECH AND LANGUAGE/VOLUME 29, NUMBER 1 2008

and is unreliable with appointments. Another that a patient with semantic impairment sec-
impression is that of a confused and confab- ondary to extensive temporal lobe damage
ulatory patient unable to relay events sur- could lace his shoe without substantive knowl-
rounding his current hospital admission. A edge of what a shoe is.
third impression, perhaps fueled more by Squire6,16 proposed a taxonomy of human
television drama than by neuropsychological memory based on the distinction between de-
reality, is that of an individual with severe clarative and nondeclarative systems. Declara-
retrograde amnesia who is unable to recall tive memory is information that one has
extensive periods of his past. Very few of us conscious or explicit access to, whereas non-
have the impression of memory loss present- declarative memory includes implicit processes
ing as degradation of object meaning (e.g., such as motor schemata (e.g., lacing a shoe) or
Grandpa has forgotten what an apple is). classic conditioning effects (e.g., salivating
Such fundamental knowledge is represented when hearing the dinner bell). Within this
within semantic memory, a system that sup- classification, semantic memory is a type of
ports long-term knowledge of word and declarative knowledge that exists alongside
object meaning.1,2 The deterioration of se- other forms of declarative knowledge such as
mantic memory is associated with several working memory and episodic memory. In a
neurologic conditions addressed in this issue. further fractionation of declarative memory,
It is our aim here to provide an introduction Tulving2 demonstrated that semantic memory
to gross structure and neuroanatomic sub- (general knowledge) and episodic memory (i.e.,
strates of semantic memory from a perspec- memory for the details of a specific event) are
tive clinically relevant to speech-language also dissociable. That is, one’s knowledge of
pathology. DOGS is dissociable from one’s memory of a
particular dog (e.g., Bob) or a particular event in
which one encountered a memorable dog.
SEMANTIC MEMORY IN THE More specifically, semantic memory encom-
TAXONOMY OF HUMAN MEMORY passes our stored knowledge of meaning about
The different impressions of impairment de- the world, extending from encyclopedic facts
scribed above highlight variability on everyday (e.g., George Washington was the first U.S.
tasks that rely on human memory. Rather than president), to concrete object knowledge (e.g.,
being a unitary system, there is strong evi- this red sphere is an apple) through abstract
dence to suggest that human memory frac- lexical-semantic meaning (e.g., melancholy is a
tionates into several distinct subsystems.3–7 A type of sadness).17–20
multitude of neuropsychological and func- Although partially dissociable, there is an
tional imaging studies suggest that these in- intimate relationship between episodic and
dividual memory systems have both semantic memory. Many researchers have ar-
overlapping and distinctive neuroanatomic gued that context-specific event memories
substrates.8–14 The nature of discrete repre- eventually consolidate into a semantic mem-
sentation versus interactivity within different ory that is context-free.21,22 That is, memory
memory systems remains an issue of central of single episodes or exemplars evolves into
debate in cognitive neuroscience. Nonethe- abstract conceptual knowledge. As an illustra-
less, due to this ‘‘patchy’’ anatomic overlap, tive example, consider one’s episodic memory
neurologic damage may result in impairment of a childhood dog, Max. Max is inextricably
of one or more domains of human memory linked to a particular context and time (i.e.,
with relative preservation of other domains. Max was a chocolate Labrador retriever who
For example, the ability to tie one’s shoe relies lounged around the house and snatched food
heavily on procedural motor memory that is from the counter when we were not looking).
strongly linked to the cerebellum,15 whereas However, one’s semantic memory of DOGS
conceptual knowledge of SHOES (hereafter we is more abstract, generalizing beyond Max
represent concepts in SMALL CAPITAL letters) to other category members. Ultimately,
relies on semantic memory. Thus, it is feasible conceptual representation is shaped by one’s
 SEMANTIC MEMORY AND LANGUAGE PROCESSING/ANTONUCCI, REILLY 7

experience with many dogs and the integra-

tion of information from other modalities into
a representation that is not constrained to a
specific context. Thus, the formation of a
stable concept often reflects an amalgamation
of sensory-motor features (e.g., DOGS have
tails, fur, can bark), affective experience (e.g.,
DOGS are friendly), and encyclopedic facts
(e.g., A DOG is a mammal.) among other traits.
This shift from memory of specific exemplars
to more general knowledge is crucial because it
allows us to rapidly generalize to new category
members by establishing a semantic field that
spans the family of DOGS.23–26

SEMANTIC MEMORY AND

LANGUAGE
Much of the literature on conceptual develop-
ment does not draw sharp distinctions be-
tween words and their associated concepts,
whose relationship is generally considered
to be arbitrarily symbolic. Saussure (1857–
1913), one of the critical figures in modern
linguistics, compared the relation between
words and concepts to the opposite sides of
the same page.27 This associative link be-
tween language and meaning enables us to
make strong inferences about the structure of
semantic memory from language. That is, in
the absence of impaired access to word mean-
ing (e.g., aphasia), lexical knowledge is also
semantic knowledge.
Confrontation naming is perhaps the most
common tool used to assess the integrity of
language functioning (and by default the in-
tegrity of semantic memory). Although naming
is seemingly straightforward, naming difficul-
ties can reflect impairment to one or more
processes. These processes are often depicted
using the ubiquitous ‘‘box and arrow’’ format,
wherein boxes represent distinct levels of proc-
essing and arrows represent connectivity among
these levels. Figure 1 displays a classic example
of a cognitive model of single-word processing Figure 1 Cognitive model of intact single-word
from the visual modality. Consider, for exam- processing for the concept CAT, from visual percep-
tion to spoken naming.
ple, how the concept of CAT would be processed
in this model. When viewing a picture of a cat, form to a stored visual-perceptual representa-
sensory processes linked to visual perception of tion. Disruption to this pathway results in
color, line orientation, and texture (e.g., yellow, inability to name from a visual stimulus as
round, and fluffy) are followed by matching of demonstrated in patients with visual agnosia,
8 SEMINARS IN SPEECH AND LANGUAGE/VOLUME 29, NUMBER 1 2008

who are able to name the same concept from patients, as well as in those who experience
auditory or tactile stimuli. Assuming the in- degraded access to conceptual knowledge, is
tegrity of this pathway, stored visual object explored. The remainder of this review will
representations activate meaning associated focus on the development of theories of the
with CATS. This encompasses hierarchical taxo- organization of conceptual knowledge and the
nomic representations (e.g., is an ANIMAL), neural substrates thereof.
features shared among other category members
(e.g., has legs, walks), and distinctive features
that discriminate cats from other category THEORETICAL MODELS OF
members (e.g., meows, elliptical pupils). In SEMANTIC ORGANIZATION
addition to these feature-based distinctions, Theories of semantic organization revolve pri-
meaning is represented also in encyclopedic marily around two fundamental questions.
memory (e.g., cats land on their feet), episodic First, is there one semantic system that proc-
memory (e.g., my childhood cat, Tom), and esses conceptual information for all modalities
through abstract semantic associations (e.g., regardless of the input/output route, or are
cats are evil). Successful word retrieval requires there multiple semantic systems dedicated to
effective storage of discrete pieces of conceptual processing modality-specific information? Sec-
knowledge and the dynamic integration of ond, how is semantic knowledge itself (i.e.,
these parts into a cohesive representation. concepts and the features that define concepts)
One common assumption underlying tradi- organized?
tional aphasia therapy is that patients have
preserved conceptual knowledge and that im-
pairment results from disrupted access to that Interaction of the Semantic System
knowledge. with Input/Output Modalities
Figure 2 represents possible stages of The semantic system described in Fig. 1 cor-
breakdown associated with several classic cort- responds with an amodal system in which
ical aphasia syndromes. Some patients appear separate modality-specific perceptual processes
to demonstrate disconnection between intact all access a unitary semantic store. As described
semantic knowledge and preserved phonologic above, this system contains information about
representations (e.g., pure anomia). In these objects’ physical properties, functions or ac-
individuals, naming attempts are characterized tions, and associations with other concepts, all
by predominance of semantic circumlocutions of which is represented in an abstract, modal-
(Fig. 2A). In contrast, other patients who ity-independent format. Concepts created in
experience degraded phonologic representa- this format may be mapped onto any type of
tions (e.g., Broca’s aphasia) frequently produce output, accessed for example by phonologic
phonemic paraphasias (Fig. 2B). Common lexical output for oral naming or orthographic
among aphasia syndromes is that although lexical output for written naming. One influ-
word retrieval is impaired, patients are typically ential example of such a model is the organized
able to demonstrate preserved conceptual unitary content hypothesis (OUCH),29 which
knowledge through alternative modalities proposes that certain types of input/output
such as gesture, appropriate object use, or modalities have ‘‘privileged access’’ to certain
drawing, if not through verbal description.28 types of semantic features (e.g., visual presen-
There are clinical populations, however, in tation rapidly accesses semantic knowledge).
which the assumption of intact conceptual Reports of patients with optic aphasia,30,31
knowledge is not upheld. This issue of Seminars an impairment characterized by difficulty nam-
addresses language function within several pa- ing visually presented objects in the context of
tient populations who experience progressive preserved ability to pantomime the action as-
loss of conceptual knowledge that is not sociated with those objects and name from
reducible to selective impairments of input tactile and auditory stimuli, seemed to argue
(i.e., agnosia) or output (i.e., lexical retrieval). against the existence of an amodal semantic
Treatment for language impairment in these system. This observation led to theories
 SEMANTIC MEMORY AND LANGUAGE PROCESSING/ANTONUCCI, REILLY 9

Figure 2 Cognitive models exemplifying breakdown of spoken single-word naming as the result of
(A) disconnection between intact semantic knowledge and preserved phonologic representations and
(B) degradation of phonologic representations.

espousing separate semantic systems dedicated tent. However, there was less consistency in
to processing specific types of information (e.g., responses across modalities. Response consis-
visual versus auditory/verbal versus tactile). A tence within modality was considered evidence
seminal paper by Warrington and Shallice,32 of degradation of, rather than disrupted access
comparing naming of the same concepts from to, semantic knowledge, and response incon-
visual and verbal stimuli across multiple ses- sistency between modalities was considered
sions, supported this hypothesis. Patients’ re- evidence of separate visual and verbal semantic
sponses to the same items presented multiple systems that could be independently damaged.
times in the same modality were fairly consis- Whereas some more contemporary work
10 SEMINARS IN SPEECH AND LANGUAGE/VOLUME 29, NUMBER 1 2008

continues to support the notion of multiple objects.12,41 In this model, object concepts are
modality-specific semantic systems,33–35 uni- formed, in part, by core features or semantic
tary accounts of the semantic system have primitives stored in the same neural mecha-
become more commonly accepted likely in nisms active when the concepts’ features were
large part because they more parsimoniously first acquired. Specifically, semantic features
explain semantic deficits observed across mo- related to perception of visual form and motion
dalities (i.e., impairment to comprehension and are stored within visual processing systems, and
production regardless of how the information is features related to manipulation and function
accessed).28 are stored in motor systems42–45 (also see
A third possibility has also been proposed, Pulvermüller46,47). In this view, semantic rep-
in which the semantic system is not thought of resentations are grounded in the implicit acti-
either as modality-specific or amodal but rather vation of these semantic primitives. However,
multimodal. Early forms of this model were such features do not encompass our entire
proposed by Karl Wernicke, who described the conceptualization of concrete objects, which
semantic system as a network in which con- also includes explicit encyclopedic and associa-
cepts emerge from the interaction of modality- tive knowledge, or our understanding of more
specific sensory perceptions36 (see also Rogers abstract concepts.
et al37). An analogous comparison might be
that between a library and a librarian. Though
the format and content may be represented Organization of Semantic Knowledge
differently, both modality-specific and amodal We now shift our attention to the internal
models of the semantic system suggest that it structure of the semantic system, focusing on
functions like a library, as a repository of facts. how concepts and the features that define
Alternatively, multimodal theories of the se- concepts are themselves organized. Attempts
mantic system could liken it to the librarian, to discern these patterns dominate the litera-
who does not contain the information, but ture, much of which has been borne from
knows where to go to get it and helps to attempts to interpret category-specific deficits in
cross-reference and synthesize it. A set of patients with brain injury. The most frequently
related theories have emerged, which are reported are those in which patients have
known as simulationist or embodied models of difficulty processing information about con-
semantic processing. Within this framework, cepts within the domain of living things with
conceptual representations are grounded in the less frequent reports of the opposite dissocia-
sensory-motor processes that were active when tion in which nonliving concepts are impaired
a concept was first acquired. The strongest (for review, see Capitani et al48). The variable
interpretations propose that each time we en- patterns of impairment in these individuals
counter a concept such as CAT, whether by provide important clues about how semantic
looking at a picture, hearing a description, or knowledge is organized. Any comprehensive
recalling how frequently Fluffy made us sneeze, model of the semantic system must account
the sensory-motor areas of the brain engaged for or predict all permutations of impairment, a
during perception of an actual cat are reacti- mandate that has yet to be realized.
vated in a simulation of that experience. Some Models of the internal organization of
authors propose that higher-order association semantic knowledge fall into two broad camps,
areas in the brain act to integrate this informa- feature-based and category/domain-based. An or-
tion,38,39 and even more radical proponents ganizing principle of feature-based models is
contend that sensory-motor areas themselves that they consider category-specific impair-
are heteromodal and interaction among them is ments to emerge from damage to the under-
sufficient not only to fully represent concrete lying features that define concepts. A prime
concepts but also ultimately abstract, meta- example of these was the differential weighting
phorical concepts (e.g., love) as well.40 A hypothesis offered by Warrington and col-
less extreme formulation is the sensory/ leagues,49–51 in which they observed a close
motor model of semantic representations of relationship between the types of categories
 SEMANTIC MEMORY AND LANGUAGE PROCESSING/ANTONUCCI, REILLY 11

and specific impairment to characteristic se- contrast, nonliving things tend to share more
mantic features (e.g., color discriminates many functional features (e.g., things used to cut,
different types of fruits and vegetables). The things used to write), and form-function rela-
premise is that domain-specific impairments tionships tend to be specific to distinctive
result from damage to the underlying features features of the objects (e.g., containing ink is
that are most salient in defining the concepts associated with writing). Overlap among fea-
belonging to those domains, and this damage tures leads to similar concepts being stored in
was interpreted in the context of modality- close proximity to each other, so that damage to
specific semantic systems that process different a particular area of semantic space will result in
feature types. The authors proposed that proc- impairment to similar concepts (i.e., catego-
essing of living things was more reliant on ries). Differences in the strength of intercorre-
sensory (i.e., visual-perceptual) features used lation among features are also considered
to differentiate among them, whereas process- important in these models. The predicted di-
ing of nonliving things was more reliant on rection of category-specific deficits (e.g., living
differentiation based on function. As such, vs. nonliving categories) depends more on the
deficient sensory feature processing would re- severity of damage to the semantic system than
sult in semantic impairment for living things on damage to a particular feature type, though
while deficient processing of functional features different interpretations have made opposite
would result in impairment for nonliving predictions regarding the relationship between
things. This theory proposed a broad dichot- severity and direction of impairment.58,59 In
omy between domain (living vs. nonliving) and addition to these more cognitively based ap-
feature (sensory vs. functional) and came to be proaches, early formulations of the sensory/
known as the sensory/functional hypothesis. Sub- motor model of semantic processing attempted
sequent work revealed inconsistencies with the to discern the neuroanatomic structure of fea-
predictions of the sensory/functional hypoth- ture-based organization, proposing that activa-
esis, such that some patients demonstrate im- tion of different brain regions during
pairment for living things without an processing of objects from living versus non-
accompanying disproportionate visual-percep- living categories reflected automatic processing
tual feature deficit,52,53 and others demonstrate of the features that defined those concepts.12,41
visual-perceptual feature impairment that does In response to the fact that feature-based
not result in a domain-specific deficit for living models did not provide a unified account for all
things.54,55 permutations of category-specific deficits, the
Furthermore, it was observed that deficits domain-specific knowledge hypothesis was pro-
could fractionate within the domain of living posed.53 In this view, the semantic system
things, specific to either animal or plant cate- evolved categorically as the result of evolu-
gories.53 To account for these dissociations, tionary imperatives, for quick perception and
feature-based models emerged proposing that analysis of those categories that were most
the structure of object representation was based salient for survival: animals (predator or prey),
on intercorrelations among semantic fea- plants (food or medicine), and nonliving ob-
tures.29,56,57 In this view, categories emerge jects (that could be manipulated to achieve
because similar concepts share more features some goal). The finer-grained impairment
and these features tend to co-occur in particular within the living domain between animals
patterns. For example, dogs share more features and plants was thus explained. The dispropor-
with cats (e.g., have legs, walk, have fur) than tionate occurrence of impairment for living
they do with knives. In addition, in animals, things was accounted for by the proposal that
visual features shared by all members of the processing of living things requires more highly
category tend to co-occur (e.g., if you have eyes, specialized neural mechanisms, which are more
you also probably have a nose and a mouth), likely to be localized and hence more suscep-
and there tends to be a relationship between tible to selective damage.53 Whereas the orig-
shared visual and functional features (e.g., hav- inal proposal was concerned with the tripartite
ing a nose is associated with breathing air). In distinction among animals, plants, and
12 SEMINARS IN SPEECH AND LANGUAGE/VOLUME 29, NUMBER 1 2008

nonliving things, subsequent iterations have types of semantic features and contributing to
addressed the additional category of conspecifics executive control of semantic information.
(i.e., other humans).8 Caramazza and col- Disproportionate difficulty processing
leagues have also pointed out that even though conceptual representations of actions has been
categorical distinctions are considered the first- observed after lesion to left frontal cortex,
order constraint on semantic organization in particularly within the frontal operculum and
this model, feature-based theories may be rel- inferior frontal gyrus, in patients with chronic
evant in describing the organization of seman- lesions60,61 and degenerative cortical atro-
tic knowledge within categories.53 The original phy.62,63 To examine conceptual knowledge
proposal also did not address which might be of actions distinct from lexical retrieval, partic-
the particular neural substrates of categorical ipants in these studies typically respond to
knowledge. Recent developments of the sen- instructions by choosing from an array of pic-
sory/motor model of semantic organization tured or photographed actions. Patients with
suggest that within brain regions that process both chronic and degenerative lesions to frontal
particular featural information, there may be cortex have also demonstrated disproportionate
categorical structure as well.12,42–45 difficulty retrieving action names (i.e., verbs).
At this stage, no one theory appears Several studies suggest that this difficulty re-
adequately developed or specified to compre- sults from damage to semantic representations,
hensively describe semantic organization. Con- though debate continues regarding the cogni-
tinued progress toward this goal will likely tive mechanism of impairment in patients with
result from convergence of behavioral study of word class effects (e.g., verbs worse than
brain-injured and neurologically healthy indi- nouns), with some authors suggesting that
viduals, complemented by a better understand- this pattern reflects syntactic/grammatical
ing of how semantic knowledge is represented rather than semantic impairment (see Gainotti64
in the brain. We turn now to a review of these for review).65–67 Lesions in patients with
neural substrates. impaired semantic processing of actions are
not always restricted to frontal cortex but may
also extend to left parietal and posterior
NEURAL SUBSTRATES OF middle temporal cortices,61 a distributed net-
LEXICAL-SEMANTIC KNOWLEDGE work that has also been observed in functional
Evidence that the neural substrates of semantic neuroimaging of healthy adults (see Martin
processing are distributed throughout the brain and Chao12 for review). Though not fre-
comes from neuropsychological lesion studies quently observed in the neuropsychological
of patients with brain damage and functional literature, individuals who demonstrate dis-
brain imaging studies of neurologically healthy proportionate semantic impairments for
individuals. In this review, we focus primarily nonliving things, particularly manipulable
on studies of patients with brain damage that objects such as tools, often demonstrate
provide a window into the relationship between lesion to the aforementioned dorsolateral
the brain and behavior. network (see Gainotti68 for review). The
fact that this same cortical network is
engaged when we observe, imagine, and per-
Frontal Networks and Semantic form actions provides some support for
Processing the notion that sensory-motor processes con-
As reviewed above, one way to think about tribute to concept formation (see Gainotti64
semantic processing might be as the integration and Martin42,69 for reviews of lesion and
of stored factual knowledge with the control functional imaging work, respectively).
processes that allow us to access and make Left prefrontal and inferior frontal regions
connections among pieces of knowledge in have also been implicated in supporting exec-
task-appropriate ways. Evidence suggests that utive control of semantic knowledge, and dam-
left frontal cortex may contribute at each of age to these regions may manifest in several
these levels, supporting storage of particular ways. In a recent study comparing semantic
 SEMANTIC MEMORY AND LANGUAGE PROCESSING/ANTONUCCI, REILLY 13

dementia and stroke-aphasia patients, Jefferies described above, these patients with pure anomia
and Lambon Ralph70 suggested that the group are fluent with well-spared auditory compre-
of chronic aphasia patients, the majority of hension, and their naming errors tend to be
whom had large left frontal lobe lesions, semantic circumlocutions (e.g., bed: It’s some-
demonstrated deficient semantic control charac- thing to sleep on). In contrast, central semantic
terized by difficulty apprehending relevant impairment resulting from loss of semantic
semantic associations and rejecting semantic knowledge has been associated with damage
distracters. Converging evidence is derived to more anterior temporal regions resulting
from patients with progressive nonfluent apha- from disease processes including temporal lobe
sia (PNFA), a neurodegenerative disorder epilepsy, herpes simplex encephalitis, and se-
characterized by atrophy of left inferior frontal mantic dementia.82–87 In contrast with pure
and premotor cortices.71–73 In PNFA patients, anomia, individuals with semantic anomia tend
categorization deficits were apparent when in- to produce semantic paraphasias (e.g., apple:
dividuals were required to make ‘‘rule-based’’ orange, fruit) or ambiguous, empty circumlocu-
judgments by attending to relationships among tions, and concomitant comprehension deficits
salient semantic features.74 In addition, naming are also observed. Although severity of semantic
errors made by individuals with PNFA tend impairment has been correlated with left ante-
to be semantic coordinate errors (e.g., zebra: rior temporal atrophy,88,89 recent evidence
horse), suggesting that damage to left frontal obtained from patients with unilateral left tem-
regions may also disrupt the ability to differ- poral lesions90 supports the notion that severe
entiate among category members, a process semantic impairment requires bilateral temporal
requiring selective attention to distinguishing damage as observed in patients with herpes
features, inhibition of competing responses, simplex encephalitis and semantic dementia.
and working memory processes. Observation The broad role of inferior temporal cortex
of these deficits in patients with damage to left in lexical-semantic processing of objects can also
premotor and inferior frontal cortices is con- be further specified by examining particular
sistent with evidence from functional neuro- patterns of impairment. Damage to inferior
imaging studies suggesting that these regions temporal cortex has been associated with
subserve ‘‘controlled’’ semantic retrieval75 and/ category-specific deficits for living things
or selection among competing alternatives.76,77 (see Capitani et al for review48) and dispro-
Furthermore, a recent case report of an indi- portionately impaired processing of visual-
vidual with right inferior prefrontal damage perceptual semantic features (e.g., Lambon
suggests that participation of right frontal Ralph et al55,91). As noted above, these deficits
homologues is critical when tasks require sometimes co-occur,32,90,92 but both profiles
particularly demanding executive semantic have also been observed separately.93 Impair-
processing, such as when semantic distracters ment of visual/perceptual features is often
are more strongly associated with a stimulus attributed to the fact that the inferior tempo-
than is the correct response.78 ral lobes are part of the ventral visual process-
ing stream critical for object recognition (for
comprehensive anatomic review, see Gloor94).
Temporal Cortex and Semantic As noted above relative to the dorsolateral
Processing processing network, such interpretations con-
In contrast with the relative importance of verge with sensory-motor models of semantic
frontal regions for conceptualization and nam- organization positing that semantic features’
ing of actions, the left inferior temporal lobe has representations are linked to the sensory-mo-
been implicated in lexical-semantic processing tor brain regions active during initial acquis-
of objects.56,79 Damage restricted to posterior ition of the features.41 Category-specific
regions, typically observed in chronic stroke deficits for living things that result from
patients, has been reported to result in a dis- damage to temporal cortex are sometimes
connection between preserved semantic knowl- interpreted in the context of theories that
edge and phonologic word forms.80,81 As posit disproportionate importance of visual/
14 SEMINARS IN SPEECH AND LANGUAGE/VOLUME 29, NUMBER 1 2008

perceptual features in conceptualization of follow in this issue of Seminars in Speech and

living things, suggesting that the deficit seen Language address qualitative differences and
for living things may be an emergent property similarities in the effects of semantic impair-
of disrupted processing of visual-perceptual ment on language processing in dementia and
features subserved by these regions. stroke aphasia.
A burgeoning literature suggests that the
temporal pole and anteromedial regions of
temporal cortex support binding of semantic REFERENCES
feature information that is stored in sensory-
motor cortices.94,95 The anterior temporal 1. McClelland JL, Rogers TT. The parallel distrib-
uted processing approach to semantic cognition.
lobes serve as the end point of the ventral visual
Nat Rev Neurosci 2003;4:310–322
stream described above, and medial perirhinal 2. Tulving E. Episodic and semantic memory. In:
cortex serves as a multimodal association area Tulving E, Donaldson W, eds. Organization of
receiving converging inputs from all sensory Memory. New York, NY: Academic Press; 1972:
modalities.94,95 These interconnections render 381–403
anterior temporal cortex well-suited for making 3. Baddeley AD. Working Memory. Oxford, UK:
fine-grained distinctions among visually and Oxford University Press; 1986
4. Schacter DL. Memory: memory systems. In:
conceptually similar exemplars.96 Damage to
Kazdin AE, ed. Encyclopedia of Psychology.
these regions, most frequently observed in in- Washington, DC: American Psychological Associ-
dividuals with semantic dementia and herpes ation; 2000:169–172
simplex encephalitis, can manifest in difficulty 5. Squire LR. Mechanisms of memory. Science
processing distinguishing semantic features of 1986;232:1612–1619
objects, as well as in category-specific deficits 6. Squire LR. The organization of declarative and
for living things, which may be more difficult nondeclarative memory. In: Ono T, Squire LR,
Raichle ME, Perrett DI, Fukuda M, eds. Brain
to distinguish because they share more seman-
Mechanisms of Perception and Memory: From
tic features than do nonliving things.55,97 In Neuron to Behavior. New York, NY: Oxford
accordance with proposals that temporal cortex University Press; 1993:219–227
is organized hierarchically along a posterior- 7. Tulving E. How many memory systems are there?
anterior gradient, with progressively anterior Am Psychol 1985;40:385–398
regions subserving progressively more fine- 8. Caramazza A, Mahon BZ. The organisation of
grained discriminations, the temporal pole has conceptual knowledge in the brain: The future’s
past and some future directions. Cogn Neuro-
been implicated in discriminating unique enti-
psychol 2006;23:13–38
ties, and damage to the left temporal pole has 9. Carpenter PA, Just MA, Reichle ED. Working
been associated with deficient naming of famil- memory and executive function: evidence from
iar faces79 and unique geographic landmarks.98 neuroimaging. Curr Opin Neurobiol 2000;10:
195–199
10. Dronkers NF, Wilkins D, van Valin RD, Redfern
CONCLUSION B, Jaeger JJ. Lesion analysis of the brain areas
There is an intimate relation between con- involved in language comprehension. Cognition
ceptual representation and language process- 2004;92:145–177
11. Graham KS, Patterson K, Hodges JR. Episodic
ing. Furthermore, semantic impairment and
memory: new insights from the study of semantic
its associated language deficits can emerge dementia. Curr Opin Neurobiol 1999;9:245–250
from a variety of causes, including stroke and 12. Martin A, Chao LL. Semantic memory and the
neurodegenerative disease. We emphasize that brain: structure and processes. Curr Opin Neuro-
semantic memory is a dynamic system that biol 2001;11:194–201
involves both the storage and the active ma- 13. Moscovitch M, Nadel L, Winocur G, Gilboa A,
nipulation of this knowledge. A comprehen- Rosenbaum RS. The cognitive neuroscience of
remote episodic, semantic and spatial memory.
sive understanding of the breakdown of these
Curr Opin Neurobiol 2006;16:179–190
processes is necessary to target effective, eti- 14. Warrington EK. The selective impairment of
ology-specific treatments for semantically semantic memory. Q J Exp Psychol 1975;27:
based language disorders. The articles that 635–657
 SEMANTIC MEMORY AND LANGUAGE PROCESSING/ANTONUCCI, REILLY 15

15. Krakauer JW, Shadmehr R. Consolidation 30. Beauvois MF. Optic aphasia: a process of
of motor memory. Trends Neurosci 2006;29: interaction between vision and language. Proc R
58–64 Soc (London) 1982;B298:35–47
16. Squire LR. The neuropsychology of human 31. Lhermitte F, Beauvois MF. A visual-speech
memory. Annu Rev Neurosci 1982;5:241–273 disconnexion syndrome. Report of a case with
17. Cree GS, McRae K. Analyzing the factors optic aphasia, agnosic alexia and colour agnosia.
underlying the structure and computation of the Brain 1973;96:695–714
meaning of chipmunk, cherry, chisel, cheese, and 32. Warrington EK, Shallice T. Category specific
cello (and many other such concrete nouns). J Exp semantic impairments. Brain 1984;107:829–
Psychol Gen 2003;132:163–201 854
18. Crutch SJ, Ridha BH, Warrington EK. The 33. McCarthy RA, Warrington EK. Evidence for
different frameworks underlying abstract and modality-specific meaning systems in the brain.
concrete knowledge: evidence from a bilingual Nature 1988;334:428–430
patient with a semantic refractory access dysphasia. 34. Shallice T. Specialisation within the semantic
Neurocase 2006;12:151–163 system. Cogn Neuropsychol 1988;5:133–142
19. Crutch SJ, Warrington EK. Abstract and 35. Shallice T. Multiple semantics: whose confusions?
concrete concepts have structurally different Cogn Neuropsychol 1993;10:251–261
representational frameworks. Brain 2005;128: 36. Eggert GH. Wernicke’s Works on Aphasia:
615–627 A Sourcebook and Review. The Hague, The
20. Humphreys GW, Riddoch MJ. Features, objects, Netherlands: Mouton; 1977
action: the cognitive neuropsychology of visual 37. Rogers TT, Lambon Ralph MA, Garrard P, et al.
object processing, 1984–2004. Cogn Neuropsy- Structure and deterioration of semantic memory: a
chol 2006;23:156–183 neuropsychological and computational investiga-
21. Burgess N, Maguire EA, O’Keefe J. The human tion. Psychol Rev 2004;111:205–235
hippocampus and spatial and episodic memory. 38. Damasio AR. Time-locked multiregional retro-
Neuron 2002;35:625–641 activation: a systems-level proposal for the neural
22. Garrard P, Lambon Ralph MA, Hodges JR, substrates of recall and recognition. Cognition
Patterson K. Prototypicality, distinctiveness, and 1989;33:25–62
intercorrelation: analyses of the semantic attributes 39. Simmons WK, Barsalou LW. The similarity-in-
of living and nonliving concepts. Cogn Neuro- topography principle: reconciling theories of con-
psychol 2001;18:125–174 ceptual deficits. Cogn Neuropsychol 2003;20:451–
23. Gaskell MG, Marslen-Wilson WD. Integrating 486
form and meaning: a distributed model of speech 40. Gallese V, Lakoff G. The brain’s concepts: the
perception. Lang Cogn Process 1997;12:613– role of the sensory-motor system in conceptual
656 knowledge. Cogn Neuropsychol 2005;22:455–
24. Keil FC. Concepts, Kinds, and Cognitive Develop- 479
ment. 1st ed. Cambridge, MA: MIT Press; 1989 41. Martin A, Ungerleider LG, Haxby JV. Category
25. Koenig P, Smith EE, Moore P, Glosser G, specificity and the brain: the sensory/motor
Grossman M. Categorization of novel animals model of semantic representations of objects. In:
by patients with Alzheimer’s disease and cortico- Gazanniga MS, ed. The New Cognitive Neuro-
basal degeneration. Neuropsychology 2007;21: sciences. 2nd ed. Cambridge, MA: MIT Press;
193–206 2000:1023–1036
26. Steyvers M, Tenenbaum JB. The large-scale 42. Martin A. The representation of object concepts
structure of semantic networks: statistical analyses in the brain. Annu Rev Psychol 2007;58:25–45
and a model of semantic growth. Cogn Sci 2005; 43. Martin A, Haxby JV, Lalonde FM, et al. Discrete
29:41–78 cortical regions associated with knowledge of color
27. Saussure Fd. Cours de linguistique generale (1907). and knowledge of action. Science 1995;270:102–
New York, NY: Philosophical Library; 1916 105
28. Raymer AM, Gonzalez-Rothi LJ. Cognitive 44. Martin A, Weisberg J. Neural foundations for
approaches to impairments of word comprehen- understanding social and mechanical concepts.
sion and production. In: Chapey R, ed. Language Cogn Neuropsychol 2003;20:575–587
Intervention Strategies in Aphasia and Related 45. Martin A, Wiggs CL, Ungerleider LG, Haxby JV.
Neurogenic Communication Disorders. Philadel- Neural correlates of category-specific knowledge.
phia, PA: Lippincott Williams & Wilkins; 2001: Nature 1996;379:649–652
524–550 46. Pulvermüller F. Words in the brain’s language.
29. Caramazza A, Hillis AE, Rapp BC, Romani C. Behav Brain Sci 1999;22:253–279
The multiple semantics hypothesis: Multiple 47. Pulvermüller F. Brain reflections of words and
confusions? Cogn Neuropsychol 1990;7:161–189 their meaning. Trends Cogn Sci 2001;5:517–524
16 SEMINARS IN SPEECH AND LANGUAGE/VOLUME 29, NUMBER 1 2008

48. Capitani E, Laiacona M, Mahon B, Caramazza A. 63. Hillis AE, Sangjin O, Ken L. Deterioration of
What are the facts of semantic category-specific naming nouns versus verbs in primary progressive
deficits? A critical review of the clinical evidence. aphasia. Ann Neurol 2004;55:268–275
Cogn Neuropsychol 2003;20:213–261 64. Gainotti G. Anatomical functional and cognitive
49. Crutch SJ, Warrington EK. The selective impair- determinants of semantic memory disorders.
ment of fruit and vegetable knowledge: a multiple Neurosci Biobehav Rev 2006;30:577–594
processing channels account of fine-grain category 65. Bird H, Howard D, Franklin S. Why is a verb like
specificity. Cogn Neuropsychol 2003;20:355– an inanimate object? Grammatical category and
372 semantic category deficits. Brain Lang 2000;72:
50. Warrington EK, McCarthy RA. Category specific 246–309
access dysphasia. Brain 1983;106:859–878 66. Damasio AR, Tranel D. Nouns and verbs are
51. Warrington EK, McCarthy RA. Categories of retrieved with differently distributed neural sys-
knowledge. Further fractionations and an attempted tems. Proc Natl Acad Sci U S A 1993;90:4957–
integration. Brain 1987;110(Pt 5):1273–1296 4960
52. Barbarotto R, Capitani E, Spinnler H, Trivelli C. 67. Daniele A, Giustolisi L, Silveri MC, Colosimo C.
Slowly progressive semantic impairment with Evidence for a possible neuroanatomical basis for
category specificity. Neurocase 1995;1:107–119 lexical processing of nouns and verbs. Neuro-
53. Caramazza A, Shelton JR. Domain-specific psychologia 1994;32:1325–1341
knowledge systems in the brain: the animate- 68. Gainotti G. What the locus of brain lesion tells us
inanimate distinction. J Cogn Neurosci 1998; about the nature of the cognitive defect underlying
10:1–34 category-specific disorders: a review. Cortex 2000;
54. Lambon Ralph MA, Graham KS, Patterson K, 36:539–559
Hodges JR. Is a picture worth a thousand words? 69. Martin A. Neural foundations for conceptual
Evidence from concept definitions by patients representations: evidence from functional brain
with semantic dementia. Brain Lang 1999;70: imaging. In: Hart JJ, Kraut MA, eds. Neural Basis
309–335 of Semantic Memory. Cambridge, UK: Cam-
55. Lambon Ralph MA, Patterson K, Garrard P, bridge University Press; 2007:302–330
Hodges JR. Semantic dementia with category 70. Jefferies E, Lambon Ralph MA. Semantic impair-
specificity: a comparative case-series study. Cogn ment in stroke aphasia versus semantic dementia:
Neuropsychol 2003;20:307–326 a case-series comparison. Brain 2006;129:2132–
56. Tyler LK, Moss HE. Towards a distributed 2147
account of conceptual knowledge. Trends Cogn 71. Gorno-Tempini ML, Dronkers NF, Rankin KP,
Sci 2001;5:244–252 et al. Cognition and anatomy in three variants of
57. Tyler LK, Moss HE, Durrant-Peatfield MR, Levy primary progressive aphasia. Ann Neurol 2004;
JP. Conceptual structure and the structure of 55:335–346
concepts: a distributed account of category-specific 72. Gorno-Tempini ML, Murray RC, Rankin KP,
deficits. Brain Lang 2000;75:195–231 Weiner MW, Miller BL. Clinical, cognitive and
58. Devlin JT, Gonnerman LM, Andersen ES, anatomical evolution from nonfluent progressive
Seidenberg MS. Category-specific semantic defi- aphasia to corticobasal syndrome: a case report.
cits in focal and widespread brain damage: a Neurocase 2004;10:426–436
computational account. J Cogn Neurosci 1998; 73. Grossman M, Mickanin J, Onishi K, Hughes E.
10:77–94 Progressive nonfluent aphasia: language, cognitive,
59. Gonnerman LM, Andersen ES, Devlin JT, and PET measures contrasted with probable
Kempler D, Seidenberg MS. Double dissociation Alzheimer’s disease. J Cogn Neurosci 1996;8:
of semantic categories in Alzheimer’s disease. 135–154
Brain Lang 1997;57:254–279 74. Koenig P, Smith EE, Grossman M. Semantic
60. Saygin AP, Wilson SM, Dronkers NF, Bates E. categorisation of novel objects in frontotemporal
Action comprehension in aphasia: Linguistic and dementia. Cogn Neuropsychol 2006;23:541–
non-linguistic deficits and their lesion correlates. 562
Neuropsychologia 2004;42:1788–1804 75. Wagner AD, Pare-Blagoev EJ, Clark J, Poldrack
61. Tranel D, Kemmerer D, Adolphs R, Damasio H, RA. Recovering meaning: left prefrontal cortex
Damasio AR. Neural correlates of conceptual guides controlled semantic retrieval. Neuron 2001;
knowledge for actions. Cogn Neuropsychol 2003; 31:329–338
20:409–432 76. Thompson-Schill SL, D’Esposito M, Aguirre GK,
62. Bak TH, Hodges JR. The effects of motor Farah MJ. Role of left inferior frontal cortex in
neurone disease on language: further evidence. retrieval of semantic knowledge: a re-evaluation.
Brain Lang 2004;89:354–361 Proc Natl Acad Sci U S A 1997;94:14792–14797
 SEMANTIC MEMORY AND LANGUAGE PROCESSING/ANTONUCCI, REILLY 17

77. Thompson-Schill SL, D’Esposito M, Kan IP. morphometry study of Semantic Dementia:
Effects of repetition and competition on activity in Relationship between temporal lobe atrophy
left prefrontal cortex during word generation. and semantic memory. Ann Neurol 2000;47:36–
Neuron 1999;23:513–522 45
78. Samson D, Connolly C, Humphreys GW. When 89. Williams GB, Nestor PJ, Hodges JR. Neural
‘‘happy’’ means ‘‘sad’’: neuropsychological evidence correlates of semantic and behavioural deficits in
for the right prefrontal cortex contribution to frontotemporal dementia. Neuroimage 2005;24:
executive semantic processing. Neuropsychologia 1042–1051
2007;45:896–904 90. Antonucci SM, Beeson PM, Labiner DM,
79. Damasio H, Grabowski TJ, Tranel D, Hichwa Rapcsak SZ. Lexical retrieval and semantic knowl-
RD, Damasio AR. A neural basis for lexical edge in patients with left inferior temporal lobe
retrieval. Nature 1996;380:499–505 lesions. Aphasiology 2008;22:281–304
80. Foundas AL, Daniels SK, Vasterling JJ. Anomia: 91. Lambon Ralph MA, McClelland JL, Patterson K,
case studies with lesion localization. Neurocase Galton CJ, Hodges JR. No right to speak? The
1998;4:35–43 relationship between object naming and semantic
81. Raymer AM, Foundas AL, Maher LM, Green- impairment: neuropsychological evidence and a
wald ML. Cognitive neuropsychological analysis computational model. J Cogn Neurosci 2001;
and neuroanatomic correlates in a case of acute 13:341–356
anomia. Brain Lang 1997;58:137–156 92. De Renzi E, Lucchelli F. Are semantic systems
82. Bell BD, Hermann BP, Woodard AR, et al. Object separately represented in the brain? The case of
naming and semantic knowledge in temporal lobe living category impairment. Cortex 1994;30:3–
epilepsy. Neuropsychology 2001;15:434–443 25
83. Davies RR, Graham KS, Xuereb JH, Williams 93. Moss HE, Tyler LK, Durrant-Peatfield M, Bunn
GB, Hodges JR. The human perirhinal cortex and EM. ‘Two eyes of a see-through’: impaired and
semantic memory. Eur J Neurosci 2004;20:2441– intact semantic knowledge in a case of selective
2446 deficit for living things. Neurocase 1998;4:291–
84. Glosser G, Donofrio N. Differences between 310
nouns and verbs after anterior temporal lobec- 94. Gloor P. The Temporal Lobe and Limbic System.
tomy. Neuropsychology 2001;15:39–47 New York, NY: Oxford University Press; 1997
85. Hodges JR, Patterson K, Oxbury S, Funnell E. 95. Bussey TJ, Saksida LM. The organization of
Semantic dementia: progressive fluent aphasia visual object representations: a connectionist
with temporal lobe atrophy. Brain 1992;115: model of effects of lesions in perirhinal cortex.
1783–1806 Eur J Neurosci 2002;15:355–364
86. Pietrini V, Nertempi P, Vaglia A, Revello MG, 96. Tyler LK, Stamatakis EA, Bright P, et al.
Pinna V, Ferro-Milone F. Recovery from herpes Processing objects at different levels of specificity.
simplex encephalitis: selective impairment of J Cogn Neurosci 2004;16:351–362
specific semantic categories with neuroradiological 97. Noppeney U, Patterson K, Tyler LK, et al.
correlation. J Neurol Neurosurg Psychiatry 1988; Temporal lobe lesions and semantic impairment:
51:1284–1293 a comparison of herpes simplex virus encephalitis
87. Snowden JS, Goulding PJ, Neary D. Semantic and semantic dementia. Brain 2007;130:1138–
dementia: a form of circumscribed cerebral 1147
atrophy. Behav Neurol 1989;2:167–182 98. Tranel D. Impaired naming of unique landmarks
88. Mummery CJ, Patterson K, Price CJ, Ashburner is associated with left temporal polar damage.
J, Frackowiak RSJ, Hodges JR. A voxel-based Neuropsychology 2006;20:1–10