Neuropsychologia 47 (2009) 995–1003
Contents lists available at ScienceDirect
Neuropsychologia
journal homepage: www.elsevier.com/loca
Interac ma
An fMR
Jonathan
Department of es
articl
Article history:
Received 19 M
Received in re
22 September
Accepted 28 O
Available onlin
Keywords:
Human
Semantic me
Concept repr
DLPFC
s both
he cu
sema
l feat
), and
a par
ipants viewed a list of features and decided whether the list accurately described a target concept. We
focused on the effect of semantic context on processing of features critical to a concept’s representa-
1. Introdu
Our ab
both on th
the flexibl
demands o
tic memor
primarily o
1972). Rec
involvesad
mation (co
service of
Chao, 2001
process an
howdiffer
different p
∗
Correspo
15 Chaucer R
E-mail ad
0028-3932/$
doi:10.1016/jmory
esentation
tion. Task demands were manipulated by giving participants instructions that encouraged rule-based or
similarity-based judgments. Activation patterns for feature processing were found to depend on the type
of noun being evaluated and whether or not critical features were consistent with surrounding informa-
tion: When processing critical features that contradicted other information, complex nouns resulted in
additional recruitment in frontal and temporal cortex compared to nominal kinds. We observed modest
effects of instruction condition, with rule-based instructions resulting in increased frontal processing and
similarity-based instructions recruiting more temporal and parietal regions. Together, these results sup-
port the hypothesis that various classes of nouns are represented differently in semantic memory, and
emphasize the dynamic interaction of process and content in semantic memory.
© 2008 Elsevier Ltd. All rights reserved.
ction
ility to understand and interact with the world rests
e information that we have acquired about objects and
e application of this knowledge within the contextual
f our immediate environment. Early models of seman-
y deemphasized the importance of context and focused
n the storage of features that compose objects (Tulving,
ent work has demonstrated that semantic knowledge
ynamic interactionbetweenstorageof conceptual infor-
ntent) and the active manipulation of this knowledge in
a task (process) (Koenig & Grossman, 2007; Martin &
). In the current study we focus on the interaction of
d content in semantic memorywith specific attention to
ent types of semantic content can engender qualitatively
rocessing strategies.
nding author. Present address:MRC Cognition and Brain Sciences Unit,
oad, Cambridge CB2 7EF, United Kingdom. Tel.: +44 1223 355294.
dress: jonathan.peelle@mrc-cbu.cam.ac.uk (J.E. Peelle).
Theprinciple that disparate types of semantic content are stored
differently in the brain is well established. For example, it has
been repeatedly demonstrated that the noun categories “animals”
and “tools” rely on at least partially dissociable regions of cor-
tex in the ventral visual pathway (Caramazza & Shelton, 1998;
Martin, 2007). However, differences in content may also arise
from the semantic structure of a concept—that is, how individ-
ual features contribute to a concept’s representation (Crutch &
Warrington, 2005; Keil, 1989). We suggest that such differences
in content necessitate differences in process because divergent
types of information must be evaluated. In the current study we
examined the effects of context on semantic processing on mul-
tiple levels. First, we investigated semantic context effects within
two types of nouns in order to assess the degree to which implicit
processing of feature knowledge depends on prior semantic con-
text. Second, we examined processing differences resulting from
dissimilarities in the semantic structure of two classes of nouns.
Finally, we used two sets of experimental instructions designed
to encourage distinct processing strategies for this task to see
if implicit processing differences can be further altered using
explicit task demands. This approach enabled us to examine both
processing requirements that differ based on intrinsic concept
– see front matter © 2008 Elsevier Ltd. All rights reserved.
.neuropsychologia.2008.10.027tion between process and content in se
I study of noun feature knowledge
E. Peelle
∗
, Vanessa Troiani, Murray Grossman
Neurology, University of Pennsylvania School of Medicine, Philadelphia, PA, United Stat
e info
ay 2008
vised form
2008
ctober 2008
e 11 November 2008
abstract
Effective semantic processing require
information to our environment. In t
cept’s features was modulated by the
of nouns: complex nouns, in which al
(apples are usually red, but not always
role (an uncle must be the brother ofte/neuropsychologia
ntic memory:
stored conceptual knowledge and the ability to relate this
rrent study we examined how neural processing of a con-
ntic context in which they were presented using two types
ures contribute in a variable manner to an object’s meaning
nominal kinds, for which a single feature plays a diagnostic
ent). We used fMRI to monitor neural activity while partic-

996 J.E. Peelle et al. / Neuropsychologia 47 (2009) 995–1003
properties and those that differ based on externally imposed crite-
ria.
One dominant framework used to approach the study of seman-
tic memor
characteriz
tributions o
and growso
McRae, de
(Weuse cap
nent featur
manner to
itself can d
most peopl
larly salient
apple becau
contributin
in this fash
size the po
to a concep
determine f
strongly to
apple) as w
object (aux
not). Within
of an objec
that contrib
account the
In gener
cal feature
auxiliary fe
degree to t
may be mo
mismatch b
the other fe
this is not
cal feature
1989). A co
father. The
this person
features tha
kindly deme
its often. Ho
gatherings,
had a large
GRANDFAT
nominal kin
feature. Un
able to ove
ture. Accord
less upon a
Thedegr
kindconcep
amatter of
rule-based
of the diagn
ing” feature
this perspec
ent way fro
all concepts
features, bu
to individua
example, fo
weighted m
fact the cate
(Hampton,
that the two types of nouns elicit differences in cortical processing,
regardless of the nature of these differences.
Empirical evidence supporting the difference between neural
ntat
d in
& M
heth
ed a
artic
t’s m
s stro
x no
ely to
tera
nall
than
that
y fro
that
ce a
he cu
in o
s. W
ature
ral,
n th
siste
ted. T
ntly
pecte
e for
reat
d to
ve w
tures
t (co
f the
ition
e als
ation
ema
. Eac
sed
, Pat
ed to
ays a
imila
qual
t and
sub
ronta
ity-b
l ass
an,
here
nditi
tion
ested
od
curren
icipan
icipan
om 18y, which we adopt for the current study, involves
ing the meanings of concrete nouns in terms of dis-
f features: an APPLE hasastem, is red, contains seeds,
natree(Hampton, 1995; McClelland & Rogers, 2003;
Sa, & Seidenberg, 1997; Smith, Shoben, & Rips, 1974).
ital letters to refer to a concept, and italics to its compo-
es.) Each of these features contributes in a probabilistic
the representation of APPLE, but no single feature by
etermine whether an object is an APPLE. For example,
e consider the red color of an apple to be a particu-
feature, but do not have difficulty recognizing a green
se it possesses a sufficient number of other positively
g features. Although most nouns can be characterized
ion, we refer to these as “complex nouns” to empha-
tentially large number of features that can contribute
t’s meaning. Through empirical testing, it is possible to
eatures of a complex noun that contribute particularly
its meaning (critical features, such as the color of an
ell as features that are less strongly associated with an
iliary features, such as whether an apple has a stem or
this feature-based approach, determining the identity
t is accomplished by assessing the number of features
ute to its meaning, while simultaneously taking into
relative importance of these features.
al it may be assumed that the importance of a criti-
to a concept is tied to the importance of surrounding
atures, because all of these features contribute in some
he representation of a concept. Thus, a critical feature
re important in evaluating a concept when there is a
etween the information provided by this feature and
atures present. However, there are concepts for which
the case. These include nouns in which a single criti-
plays a diagnostic role, known as nominal kinds (Keil,
mmon example of a nominal kind is the word grand-
meaning of GRANDFATHER is constrained such that
must be a parent’s father. There exist other auxiliary
t people tend associate with GRANDFATHER, such as
anor, attends family gatherings, brings presents, and vis-
wever, a personwhowas unkind, never present at family
didn’t bring presents, and never visited—in other words,
number of anti-characteristic features—could still be a
HER if he were the father of a parent. Object identity for
ds therefore relies predominantly on a single critical
like complex nouns, auxiliary features should never be
rwhelm the contribution of the diagnostic critical fea-
ingly, we expect processing of nominal kinds to rely
uxiliary attributes and more upon critical features.
ee towhich the cognitive processes underlyingnominal
tsdiffer fromthose supportingcomplexnounshasbeen
some debate. One viewholds that nominal kinds rely on
processes to determine category membership because
ostic role played by a single, or small number, of “defin-
s (Keil, 1989; Keil & Batterman, 1984; Rips, 1989). From
tive, such nouns are processed in a qualitatively differ-
m other nouns. An alternative view suggests that in fact
are processed using the same summed weighting of
t that differences arise due to the weightings assigned
l features (Hampton, 1997; Rosch & Mervis, 1975). For
r nominal kinds, a feature that is particularly strongly
ay appear to play a special diagnostic role, when in
gorization process is no different than any other noun
1995). For the current study our focus is only on the fact
represe
reporte
Work,
ated w
describ
while p
concep
ity wa
comple
positiv
ity in la
Additio
vation
theory
ferentl
notion
influen
In t
(2007)
feature
iary fe
In gene
ity whe
are con
presen
differe
We ex
becaus
tively g
affecte
Abo
tic fea
conten
tance o
In add
task ar
inform
these d
groups
rule-ba
(Smith
intend
that pl
trast, s
more e
patien
tions to
useof f
similar
parieta
Grossm
data, t
tion co
instruc
being t
2. Meth
The
2.1. Part
Part
in age frions of nominal kinds and complex nouns has been
one fMRI study to date (Grossman, Troiani, Koenig,
oore, 2007). In this prior report, participants evalu-
er a list of sequentially presented features accurately
target concept. The authors examined neural activation
ipants readdescriptions that contributedpositively to a
eaning. When examining characteristic features, activ-
nger in parietal regions for nominal kinds relative to
uns. By contrast, examining features that contributed
themeaning of complex nouns resulted inmore activ-
l temporal and frontal regions relative to nominal kinds.
y, critical features were found to result in greater acti-
auxiliary features. These results are consistent with the
complex nouns and nominal kinds are represented dif-
m complex nouns in semantic memory, as well as the
individual features differ in the degree to which they
concept’s representation.
rrent study we re-analyze data from Grossman et al.
rder to more closely examine the processing of critical
e hypothesized that the context provided by auxil-
s would modulate the processing of critical features.
we predict critical features should require more activ-
ey contradict auxiliary features compared towhen they
nt, related to theevaluationof the semantic information
hat is, the critical feature is the same, but is processed
due to its importance in evaluating the target concept.
d this effect to be less prominent in nominal kinds,
these nouns the critical feature always plays a rela-
er role in concept representation, and thus should be
a lesser degree by the auxiliary features.
e have suggested that the neural processing of seman-
can be modulated by the nature of the conceptual
mplex noun or nominal kind) and the relative impor-
se features to a concept, given their semantic context.
to these considerations, participants’ goals during a
o likely to affect the relative evaluation of semantic
. In the current experiment we explicitly manipulated
nds by randomly assigning participants to one of two
h group received instructions designed to promote
or similarity-based approaches to concept evaluation
alano, & Jonides, 1998). Rule-based instructions were
bias participants towards identifying a specific feature
prominent role in the meaning of a concept. By con-
rity-based instructions were intended to encourage a
lydistributedevaluationof all factors. Basedonprevious
neuroimaging studies we expected rule-based instruc-
tly bias participants in this condition towards increased
l brain regionsassociatedwithexecutive resources, and
ased instructions to shift activity towards temporal and
ociation cortices (Koenig et al., 2005; Koenig, Smith, &
2006). However, in the previous report on the current
were no differences found between the two instruc-
ons (Grossmanet al., 2007). Thus, althoughweexpected
effects to be apparent regardless of the type of concept
, we presumed any effects would be subtle.
t study is a reanalysis of data collected for Grossman et al. (2007).
ts
ts were 25 healthy adult volunteers, 14 females and 11males, ranging
to 33 years (M=23.9, SD=3.6). All were native speakers of English,