When Action Turns into Words. Activation of Motor-Based Knowledge during Categorization of Manipulable Objects

Abstract

Functional imaging studies have demonstrated that processing of man-made objects activate the left ventral premotor cortex, which is known to be concerned with motor function. This has led to the suggestion that the comprehension of man-made objects may rely on motor-based knowledge of object utilization (action knowledge). Here we show that the left ventral premotor cortex is activated during categorization of “both” fruit/vegetables and articles of clothing, relative to animals and nonmanipulable man-made objects. This observation suggests that action knowledge may not be important for the processing of man-made objects per se, but rather for the processing of manipulable objects in general, whether natural or man-made. These findings both support psycholinguistic theories suggesting that certain lexical categories may evolve from, and the act of categorization rely upon, motor-based knowledge of action equivalency, and have important implications for theories of category specificity. Thus, the finding that the processing of vegetables/fruit and articles of clothing give rise to similar activation is difficult to account for should knowledge representations in the brain be truly categorically organized. Instead, the data are compatible with the suggestion that categories differ in the weight they put on different types of knowledge.

Full text

When Action Turns into Words.
Activation of Motor-Based Knowledge during
Categorization of Manipulable Objects
Christian Gerlach, Ian Law, and Olaf B. Paulson
Abstract
&Functional imaging studies have demonstrated that pro-
cessing of man-made objects activate the left ventral premotor
cortex, which is known to be concerned with motor function.
This has led to the suggestion that the comprehension of man-
made objects may rely on motor-based knowledge of object
utilization (action knowledge). Here we show that the left
ventral premotor cortex is activated during categorization of
‘‘both’’ fruit /vegetables and articles of clothing, relative to
animals and nonmanipulable man-made objects. This observa-
tion suggests that action knowledge may not be important for
the processing of man-made objects per se, but rather for the
processing of manipulable objects in general, whether natural
or man-made. These findings both support psycholinguistic
theories suggesting that certain lexical categories may evolve
from, and the act of categorization rely upon, motor-based
knowledge of action equivalency, and have important implica-
tions for theories of category specificity. Thus, the finding that
the processing of vegetables / fruit and articles of clothing give
rise to similar activation is difficult to account for should
knowledge representations in the brain be truly categorically
organized. Instead, the data are compatible with the sugges-
tion that categories differ in the weight they put on different
types of knowledge. &
INTRODUCTION
A central question in cognitive neuroscience concerns
how conceptual knowledge is organized in the brain.
Evidence relevant to this issue comes from studies of
patients with impaired comprehension (semantic knowl-
edge) of man-made objects along with relatively spared
comprehension of natural objects, or vice versa (for a
recent review, see Forde & Humphreys, 1999). The
existence of such category-specific disorders suggests
that knowledge may be categorically organized in the
brain or, at the very least, that different categories are
not processed in the same way. Recently, functional
imaging studies have revealed that the processing of
man-made objects selectively activates the left ventral
premotor cortex (PMv) (Chao & Martin, 2000; Gerlach,
Law, Gade, & Paulson, 2000; Grabowski, Damasio, &
Damasio, 1998; Grafton, Fadiga, Arbib, & Rizzolatti,
1997; Martin, Wiggs, Ungerleider, & Haxby, 1996). This
has led to the suggestion that comprehension of man-
made objects may rely on motor-based knowledge of
object utilization (action knowledge) mediated by the
left PMv (Chao & Martin, 2000), which together with the
left posterior parietal lobe and the left posterior middle
temporal region may form a visuomotor action network
(Devlin et al., 2002).
The suggestion that the PMv may be critically involved
in the comprehension of man-made objects is based
on the following premises: (i) that the PMv is likely
to be the human homologue of the monkey F5 area
(Binkofski et al., 1999; Rizzolatti & Arbib, 1998), which is
involved in mediating goal-directed actions such as
grasping, holding, and manipulation of objects (Rizzo-
latti & Fadiga, 1998); (ii) that patients with category-
specific disorders for man-made objects often have
frontoparietal lesions likely to involve the PMv (Gainotti,
2000). Although both premises are likely to be correct,
the interpretation regarding the role of action knowl-
edge in the comprehension of man-made objects is not
as straightforward as it might appear. This can be
appreciated if one takes into consideration the literature
on apraxia. Thus, whereas it is widely held that apraxia is
associated with lesions of the left frontal and parietal
cortex (Haaland, Harrington, & Knight, 2000), object
comprehension need not be compromised in patients
with apraxia. This has been documented in many studies
in which patients with apraxia have demonstrated pre-
served knowledge for the function of objects that they
cannot utilize properly (Moreaud, Charnallet, & Pellat,
1998; Buxbaum, Schwartz, & Carew, 1997; Buxbaum,
Veramonti, & Schwartz, 2000; Buxbaum, 2001; Heilman,
Maher, Greenw al d, & Rothi, 1997) or the reverse
(Buxbaum et al., 1997; Sirigu, Duhamel, & Poncet,
1991). The finding that knowledge of object functionCopenhagen University Hospital
©2002 Massachusetts Institute of Technology Journal of Cognitive Neurosc ience 14:8, pp. 1230 – 123 9

and knowledge of practical object utilization can disso-
ciate has l ed Buxbaum et al. (2000) to distinguish
between the two types of knowledge in terms of ‘‘what
for’’ and ‘‘how’’ knowledge. Although Buxbaum et al.
associate these two types of knowledge with the ventral
(occipito-temporal) and dorsal (occipito-parietal-frontal)
stream, respectively, they do not explicitly associate
‘‘how’’ knowledge with the left PMv. Rather, they seem
to associate ‘‘how’’ knowledge with the left inferior
parietal lobe (IPL) (Brodmann’s area [BA] 39/40), which
they argue is the mo st critically involved region in
ideomotor apraxia (IMA) (apraxia characterized by spa-
tiotemporal movement errors) (Buxbaum et al., 2000;
Buxbaum, 2001). This immediately raises the question of
what role the left PMv might serve in action processing.
One possibility, which we will return to in the discus-
sion, is that the PMv might represent some high-level
interface between the ‘‘what for’’ and the ‘‘how’’ path-
ways. This would be compatible with the observation
that some neurons in F5, discharge not only when
monkeys grasp objects, but also when monkeys simply
observe objects without grasping them (Murata et al.,
1997). If so, we might expect lesions of the PMv to result
in more ‘‘high-level’’ praxis disorders. Two such disor-
ders are ideational apraxia and conceptual apraxia,
which are characterized by action sequencing errors
and content errors (e.g., use a toothbrush as a shaver)
(Leiguarda & Marsden, 2000). Unfortunately, it is not
clear whether these disorders are distinct neuropsycho-
logical disorders or whether they represent the extreme
end of a continuum of IMA (Buxbaum, 2001; De Renzi,
1989). Perhaps because of this, these disorders have not
firmly been associated with any specific lesion, except
that they usually follow frontal damage (in addition to
damage elsewhere), whereas the critical damage in IMA
seems to be in the parietal cortex (Buxbaum, 2001;
Leiguarda & Marsden, 2000).
Although it is not possible at present to precisely
distinguish the roles of the PMv and the IPL in action
processing, except perhaps for the vague notion that
the PMv may be involved in action p rocessing at a
higher level than the IPL, it seems reasonable to con-
clude that object comprehension (‘‘what for’’) is not
contingent upon action knowledge (‘‘how’’). In accord-
ance with this, we recently demonstrated that man-
made objects compared with natural objects caused
activation of the left PMv during categorization (decid-
ing whether pictures represented man-made or natural
objects), whereas a comparison between naming of the
same man-made and natural objects was not associated
with activation of the PMv (Gerlach, Law, Gade, &
Paulson, 2002). However, if object comprehension is
not contingent on access to action knowledge, why is
the left PMv activated during the categorization of man-
made objects? To account for this, we have suggested
that action knowledge comprises information regarding
the distinctive actions that apply to objects and that the
act of categorizatio n may be based, in part, on action
equivalence (Gerlach et al., 2000). Although this inter-
pretation is compatible with psycholinguistic evidence
suggesting that categorization begins at the level of
distinctive or characterizing actions (Lakoff, 1987), this
alone does not explain why the activation of the PMv is
significantly stronger for man-made compared with
natural objects. This finding may be accounted for,
however, if we assume that man-made objects, com-
pared with natural objects, are more often grouped
together according to what kind of action applies to
them (e.g., chairs are for sitting, knives for cutting). For
a similar suggestion, see Miller and Joh ns on-Laird
(1976). If this is correct, the PMv may not be involved
in the categorization of man-made objects per se, but
rather may be involved in the categorization of objects
from all kinds of categories as long as these are manip-
ulable. On this account we may expect to find activation
of the PMv during categorization of both manipulable
natural and manipulable man-made objects (e.g., vege-
tables and articles of clothing) but not during catego-
rization of nonmanipulable natural or man-made objects
(e.g., animals and buildings).
To examine this proposal, we measured the regional
cerebral blood flow (rCBF) with positron emission
tomography (PET) in subjects who categorized pictures
of objects. There were four categorizatio n conditions in
which subjects had to decide whether objects were
natural or man-made. The conditions differed in that
they contained a predominance (84%) of either vegeta-
bles/ fruit, animals, articles of clothing, or nonmanipu-
lable man-made objects in the part of the task in which
rCBF was measured (the critical scan window). To
identify brain areas that are selectively activated during
categorization of manipulable objects, regardless of
whether these were natural or man-made, we looked
for areas that were commonly activated by vegetables /
fruit relative to animals and by articles of clothing
relative to nonmanipulable man-made objects. This
was achieved by use of conjunction analysis (Price &
Friston, 1997), in which we identified areas that are
associated with both of the following contrasts: (i)
categorization of vegetables/ fruit compared with cate-
gorization of animals; (ii) categorization of articles of
clothing compared with categorization of nonmanipu-
lable man-made objects, and in which the rCBF did not
differ significantly between the two contrasts. It should
be noted, however, that any activation revealed by this
analysis could, in principle, reflect a motor-priming
effect only. That is, one might expect that merely seeing
manipulable objects would cause activation of motor
areas, even though these areas are unimportant for the
task (the act of categorization). To exclude this possi-
bility we also had sub jects p erform so-called object
decisions on the same pictures of fruit /vegetables and
articles of clothing that were presented in the catego-
rization tasks. In these object-decision tasks the subjects
Gerlach, Law, and Paulson 1231

had to decide whether objects depicted real objects or
nonobjects (see Figure 1B). Evidence from previous PET
studies suggest that this task places greater demands
on shape processing (structural knowledge) than on
semantic processing (Gerlach, Law, Gade, & Paulson,
1999; Gerlach et al., 2000). If the activations revealed by
the conjunction analysis should in fact reflect areas
involved in categorization, rather than areas showing
motor-priming effects only, then we would expect the
rCBF in these areas to be higher during categorization of
manipulable objects than during object decisions on
manipulable objects. Thus, to ensure that the activations
revealed by the conjunction analysis would reflect areas
important for the categorization of manipulable objects,
we only allowed areas (voxels) to enter into the con-
junction analysis if they were activated more during
categorization of manipulable objects than during object
decisions on the same manipulable objects.
Although we were primarily interested in identifying
areas that are equally activated by manipulable man-
made objects and manipulable natural objects, we also
subjected the four categorization tasks to a full two-by-
two factorial analysis. The factors were Category with two
levels (natural vs. man-made objects) and Object Type,
Figure 1. (A) Examples of the stimuli used in the four categorization tasks. (B) Examples of the stimuli used in the object-decision tasks. (C) Three
sections showing the activation associated with categorization of both fruit / vegetables and articles of clothing relative to categorization of animals
and nonmanipulable man-made objects and to object decisions on vegetables/fruit and articles of clothing. (D) This plot shows the effect
size associated with each task in the left PMv where Cat. = categorization, Odt. = object decision, A. = animals, VF = vegetables/fruit,
NM = nonmanipulable man-made objects, and C. = articles of clothing. Values have been mean centered so that the yaxis shows the relative effect
sizes (increases or decreases from the mean).
1232 Journal of Cognitive Neuroscience Volume 14, Number 8

also with two levels (manipulable vs. nonmanipulable
objects). The adoption of a factorial approach also made
it possible to evaluate activation effects that are context
sensitive, that is, activations that reflect the interaction
between Category and Object Type.
RESULTS
Areas Associated with the Conjunction Analysis
The conjunction analysis was associated with increased
rCBF in the left PMv. The peak of this activation was
located on the border between the frontal operculum
(BA 44) and the premotor gyrus (BA 6) (x,y,z=¡42, 8,
22) and was significant at a level corrected for multiple
comparisons (z= 5.48, p< .002). No other areas were
significantly activated (see Figure 1C).
Areas Associated with the Two-by-Two Factorial
Analysis
Main Effects
The only main effect that revealed significant activation
was the main effect of manipulable objects. This main
effect was associated with increased rCBF in the left
PMv. The peak of the activation was located on the
border between the frontal operculum (BA 44) and the
premotor gyrus (BA 6) (x,y,z=¡42, 8, 24) (Z= 4.89,
p< .01 corrected for multiple comparisons).
Interactions
No voxels were associated with interaction between
Category and Object Type. However, to see whether
any activation of the left PMv could be established at a
lower threshold, we lowered the threshold to p< .001
uncorrected for multiple comparisons. Although this did
reveal some areas of activation, none of these were
located in the left frontal cortex.
Behavioral Data
The mean correct reaction times (RTs) to the 16 pictures
presented in the critical scan window of the categoriza-
tion t asks were subjected to a two-way analysis of
variance. The factor s were Category with two levels
(natural vs. man-made objects) and Object Type, also
with t wo levels (ma nipulabl e vs. non man ipulable
objects). There was a significant interaction between
Category and Object Type, F(1,11) = 9.99, p< .01. No
other effects were significant. Post hoc analyses (Tukey
HSD tests) revealed a significant difference ( p< .05) in
RT to animals and vegetables/fruit and in RT to animals
and nonmanipulable man-made objects (in both com-
parisons RTs were faster to animals). There was no
significant difference in RT to vegetables/fruit and arti-
cles of clothing or in RT to articles of clothing and
nonmanipulable man-made objects. Analysis of errors
did not reveal any significant difference between the
four categorization tasks (Friedman, p> .1). The mean
correct RTs and standard deviations (SDs) as well as
mean error rates for the four categorization tasks and
the two object-decision tasks are given in Table 1.
DISCUSSION
In accordance with our hypothesis, we found activation
of the left PMv during categorization of manipulable
objects regardless of whether these were natural or man-
made. Moreover, we did not find any areas associated
with an interaction between Categor y (natural vs. man-
made objects) and Object Type (manipulable vs. non-
manipulable objects), which further suggests that the
activations associated with vegetables/fruit and articles
of clothing are rather similar. Given that we did not find
any significant main effect of Object Type in the behav-
ioral data, nor any significant difference in error rate
between the four categorization tasks, it seems unlikely
that the PMv activation should reflect differences in task
difficulty. Accordingly, the PMv activation appears to
reflect task imposed processing differences between
categories.
The findings reported here clearly suggest that the left
PMv activation is related to the categorization of manip-
ulable objects rather than to the categorization of man-
made objects per se.Although this finding is compatible
Table 1. Behavioral Data
Mean Correct RT Mean Error Rate
Categorization of animals 448 (86) 0.2
Categorization of vegetables/fruit 510 (81) 0.5
Categorization of articles of clothing 487 (51) 0.6
Categorization of nonmanipulable man-made objects 515 (94) 0.5
Object decisions on vegetables/fruit 563 (74) 0.9
Object decisions on articles of clothing 587 (70) 1.4
The mean reaction times (msec), standard deviation (in brackets), and mean error rate for the 16 objects presented in the second block of the
categorization tasks and o bject-decision tasks.
Gerlach, Law, and Paulson 1233

with our hypothesis that categorization of manipulable
objects may be based, in part, on access to action
knowledge, it does not readily explain the role of the
PMv in action processing. Thus, and as mentioned in the
Introduction, the PMv is likely to be part of a larger
visuomotor action network, which also includes the left
posterior parietal lobe and the left posterior middle
temporal region (Devlin et al., 2002). The fact that
neither of these regions was found activated in the
present study is intriguing and warrants some explan-
ation. If we consider the role of the dorsal action path-
way, as envisaged by Milner and Goodale (1995), this
pathway is held to terminate in the superior parietal
lobe and to be important for the visual control of goal-
directed actions. The computations performed by this
pathway in isolation are probably sufficient for reaching
objects in space, such as gripping a pencil (Leiguarda &
Marsden, 2000; Milner & Goodale, 1995). However, for
the grip to be really efficient so that the object can be
put to work as a tool for writing, the precise function of
the object should be identified in advance. Although the
dorsal pathway contains cells that are sensitive for
orientation and size, its capacity for shape recognition
appears to be rather limited (Milner & Goodale, 1995).
Accordingly, for efficient action to take place, the praxis
system needs to have access to knowledge of object
function, a product of ventral pathway processing, so
that the appropriate action can be selected. In the
model proposed by Buxbaum (2001), actions, or gesture
engrams (e.g., that a hammer is used with a vertical
oscillating gesture), are presumed to be stored in the
IPL. Thus, the unfolding of an object-directed action
would take place in a collaboration between the supe-
rior and inferior parietal lobe, with the latter area
providing the scheme of the action and the former area
taking part in on-line execution of the scheme (e.g.,
computations regarding the position of body parts with
respect to the object and the position of body parts with
respect to each other). The selection of the correct
gesture engram, however, must be based on a knowl-
edge of which function the object serves. In other
words, there must be an association between the func-
tional knowledge of an object and the gesture engram
that can bring this function about. In line with sugges-
tions made by Tranel, Damasio, and Damasio (1997),
Tranel, Adolphs, Damasio, and Damasio (2001) , and
Damasio (1990) we would speculate that the PMv might
mediate this link and thus act as a convergence zone for
the binding of functional knowledge (‘‘what for’’) and
gesture engrams (‘‘how’’ knowledge).
The division of labor described above seems to be
roughly compatible with the different types of apraxia
described in the literature. Buxbaum (2001), for exam-
ple, distinguishes between two major subtypes of IMA:
(i) dynamic IMA, which is characterized by problems
with computation of spatiomotor information about the
position of body parts with respect to objects and the
position of body parts with respect to each other; (ii)
representational IMA, which is ch aracterized by an
inability to store or access representations of comp lex
movements (gesture engrams). Whereas dynamic IMA is
associated with lesions in the dorsal pathway (including
the superior parietal lobe), representational IMA is held
to follow lesions in the IPL (Buxbaum, 2001). In addition
to these types of apraxia, there are ideational apraxia
and conceptual apraxia. Although these disorders, as
mentioned in the Introduction, are not well defined,
they are usually held to be characterized by faulty use of
objects when a complex sequence of actions must be
organized (Leiguarda & Marsden, 2000; De Renzi, 1989).
In general, these patients seem to unders tand the
function of objects and the movements themselves are
performed well, but the object selected for the action is
wrong (e.g., as when trying to comb the hair with a
knife) (Leiguarda & Marsden, 2000). Accordingly, these
patients seem to have intact ‘‘what for’’ and ‘‘how’’
knowledge but faulty links between the two knowledge
types. In terms of the model proposed above, such an
impairment would be expected if these patients had
lesions involving the PMv, which we argue provides the
link between functional knowledge and action knowl-
edge. Although it is unknown if it is damage to the PMv
that is responsible for ideational and conceptual apraxia,
these disorders are usually reported following frontal
damage (in addition to damage elsewhere) (Buxbaum,
2001; Leiguarda & Marsden, 2000).
According to the interpretations offered here, the
PMv activation observed reflects that there is tighter
coupling between functional knowledge and action
knowledge for manipulable objects than for nonma-
nipulable objects and that this coupling is essential for
the categorizatio n of manipulable objects. This is per-
haps the most unequivocal converging evidence yet
reported in favor of the suggestion that certain lexical
categories may evolve from, and the act of categoriza-
tion may rely upon, knowledge of action equivalency
(Lakoff, 1987). Accordingly, categories may be based not
only on equivalence between their members in terms of
their intrinsic properties, such as color or shape, but
also on equivalence in terms of their extrinsic proper-
ties, such as how an organism may interact with them.
This is not to say that categorization does not depend
on semantic knowledge in general, but only that the
categorization of manipulable objects may, in addition,
be based on equivalence in terms of strong bonds
between function and action. In fact, given that manip-
ulable objects may be categorized, in part, based on
their function-action coupling, there may be little need
for actually instantiating gesture engram s into actual
object-directed action in order to categorize them,
and this may be the reason why we did not observe
any activation of the IPL. It might be tempting to
suggest that a similar argument could be made with
respect to access to functional knowledge. That is,
1234 Journal of Cognitive Neuroscience Volume 14, Number 8

activation of the inferolateral temporal cortex—a region
usually associated with semantic processing (Gerlach
et al., 2000; Mummery et al., 2000)—may fail to show
up in the contrasts performed because the categoriza-
tion of manipulable objects is based entirely on func-
tion-action couplings. When we consider this suggestion
unlikely, it is because the PMv is conceived as a con-
vergence zone and therefore must get some sort of
input. Granted that this input cannot come from acti-
vation of gesture engrams, as no gestures are per-
formed, it must come from the semantic system where
the identity (and function) of the stimuli is recognized.
Accordingly, the lack of activation of the inferolateral
temporal cortex is unlikely to be a consequence of
manipulable objects being categorized entirely via non-
semantic route. Instead, activation of the inferolateral
temporal cortex may fail to show up because all objects,
be they natural, man-made, manipulable, or nonma-
nipulable, activate semantic knowledge during catego-
rization (Gerl ach et al., 2000). Thus, the lack of
activation of the inferolateral temporal cortex in any of
the contrasts performed could suggest that access to
functional properties is equally important for the com-
prehension of natural, man-made, manipulable, and
nonmanipulable objects, but that these categories may
instead differ in how these functional attributes map
onto other types of knowledge, in this case action
knowledge (for a similar suggestion concerning links
between st ructur e and function, see Tyler, Moss,
Durrant-Peatfield, & Levy, 2000).
While we do acknowledge that conclusions based on
negative evidence are ill-advised, the failure to find
differential activation of the inferolateral temporal cortex
as a function of categor y appears to be problematic for
models of category specificity in which disorders for
man-made objects are thought to arise because these
objects depend more on functional knowledge than do
natural objects (Farah & McClelland, 1991). Of more
direct value for the evaluation of current models of
category specificity is the finding that the categorization
of vegetables/fruit and articles of clothing gave rise to
similar activation. This finding is difficult to account for
in theories that argue that knowledge representations in
the brain are truly categorically organized (Caramazza &
Shelton, 1998). However, this finding, as well as the
finding that the left PMv was not activated more by
nonmanipulable man-made objects than by animals, also
poses a serious problem for theories in which category-
specific disorders for man-made objects are thought to
arise primarily because of impaired action knowledge
(Gainotti, 2000). Instead, the present data are in accord
with theories that suggest that categories are not totally
segregated (Devlin et al., 2001) but may differ in the
weight they put on different forms of knowledge in
particular tasks (Forde & Humphreys, 1999; Gerlach
et al., 1999, 2000; Tranel, Logan, Frank, & Damasio,
1997; Warrington & McCarthy, 1987). This adds more
weight to the notion that category-specific disorders for
natural and man-made objects do not reflect a semantic
system, which is partitioned according to lexical catego-
ries, but that the distinction between natural and man-
made objects is a useful approximation only to some
other underlying factor(s) of division. This suggestion is
also compatible with the fact that the vast majority of
patients with category-specific disorders experience
problems within both domains of knowledge (natural
and man-made), albeit to different degrees, as well as
with the observation that category-specific disorders
are often not clear-cut, with some classes of natural
objects (e.g., body parts) being affected in patients with
category-specific disorders for man-made objects and
vice versa (Barbarotto, Capit ani, & Laiacona, 2001;
Dixon, Piskopos, & Schweizer, 2000).
If the left PMv does play a mediating role, our results
suggest that functional knowledge and action knowledge
may be more tightly coupled for some categories of
natural and m an-made objects than for others. This
observation is intriguing because it may relate to the
more fine-grained category-specific impairments that
have been reported. Th ese fi ne-grained d isorder s
have concerned impaired processing of manipulable
man-made objects concurrently with relatively spared
processing of natural and nonmanipulable man-made
objects ( Warrington & McCarthy, 1987), and impaired
(Hart, Berndt, & Caramazza , 1985) or prese rved
(Caramazza & Shelton, 1998; Hart & Gordon, 1992)
processing of fruit/ vegetables in cases with category-
specific impairments for natural objects. However, even
though there is some similarity between the activation
pattern reported here and the fine-grained category
specific disorders mentioned, the present data does
not allow us to account for these fine-grained disorders
simply in terms of impaired or spared links between
functional knowledge and action knowledge. To appre-
ciate this one needs only to consider t he patient
reported by Warrington and McCarthy (1987). Even
though this patient was more impaired at recognizing
manipulable compared with nonmanipulable man-made
objects, one of her preserved categories happened to
concern vegetables / fruit. This points to a restriction of
the present study. We have shown that processing of
both vegetables / fruit and articles of clothing causes
activation of the left PMv in a task where objects are
classified based on equivalence between their members
(categorization). However, this does not necessarily
imply that function-action couplings are equally impor-
tant for discrim inating between objects within the
respective categories. Thus, one might suspect that
function-action couplings associated with manipulable
man-made objects are more distinctive than function-
action couplings associated with vegetables /fruit and
that function-action coupling therefore might be more
diagnostic for differentiating b etween manipula ble
man-made object s than between veget ables / fru it.
Gerlach, Law, and Paulson 1235

Accordingly, before a principled account of the more
fine-grained category-specific disorders can be offered,
which takes into consideration the differential weighting
of function-action couplings on different categories, we
need to further examine the role of function-action
couplings in object comprehension.
METHODS
Subjects
Twelve right-handed healthy volunteers (6 females)
participated in this study. Informed written consent
was obtained according to the Declaration of Helsinki
II and the study was approved by the lo cal ethics
committee of Copenhagen ( J.nr. (KF) 01-194 / 97).
PET Scanning
PET scans were obtained with an eighteen-ring GE-
Advance scanner (Genera l Electric Medical Systems,
Milwaukee, WI) operating in 3-D acquisition mode,
producing 35 image slices with an interslice distance
of 4.25 mm. The total axial field of view was 15.2 cm
with an approximate in-plane resolution of 5 mm. The
technical specifications have been described elsewhere
(DeGrado et al., 1994).
Each subject received nine intravenous bolus injec-
tions of 200 MBq (5.7 mCi) of H
2
15
O with an interscan
interval of 8 – 10 min. The isotope was administered in
an antecubital intravenous catheter over 20 sec by an
automatic injection device followed by 10 ml of physio-
logical saline for flushing. Head movements were limited
by head holders constructed by thermally molded foam.
Before the activation sessions a 10-min transmission
scan was performed for attenuation correction. Images
were reconstructed using a 4.0-mm Hanning filter trans-
axially and an 8.5-mm Ramp filter axially. The resulting
distribution images of time-integrated counts were used
as indirect measurements of the regional neural activity.
MRI Scanning
For accurate anatomical localization of activated foci
structural MRI scanning was performed on every subject
with a 1.5 T Vision scanner (Siemens, Erlangen, Germany)
using a 3-D magnetization prepared rapid-acquisition
gradient-echo sequence (TR /TE/ TI = 11/4/100 msec, flip
angle 158). The images were acquired in the sagittal plane
with an in-plane resolution of 0.98 mm and a slice thick-
ness of 1.0 mm. The number of planes was 170 and the
in-plane matrix dimensions were 256 £256.
Image Analysis
For all subjects the complete brain volume was sampled.
Image analysis was performed using Statistical Paramet-
ric Mapping software (SPM-99, Wellcome Department
of Cognitive Neurology, London, UK). All intrasubject
images were aligned on a voxel-by-voxel basis using a
3-D automated six-parameter rigid body transformation
and the anatomical MRI scans were coregistered to the
individual averages of the nine aligned PET scans. The
average PET scans and corresponding anatomical MRI
scans were subsequently transformed into the standard
stereotactic atlas of Talairach and Tournoux (1988) using
the PET template defined by the Montreal Neurological
Institute (Friston, Ashburner, et al., 1995). The stereo-
tactically normalized images consisted of 68 planes of
2£2£2 mm voxels. Before statistical analysis, images
were filtered with a 16-mm isotropic gaussian filter to
increase the signal-to-noise ratio and to accommodate
residual variability in morphological and topographical
anatomy that was not accounted for by the stereotactic
normalization process (Friston, 1994). Differences in
global activity were removed by proportional normal-
ization of global brain counts to a value of 50.
Tests of the null hypothesis, which rejects regionally
specific condition activation effects, were performed
comparing conditions on a voxel-by-voxel basis. The
resulting set of voxel values constituted a statistical
parametric map of the tstatistic, SPM{t}. A transforma-
tion of values from the SPM{t} into the unit Gaussian
distribution using a probability integral transform
allowed changes to be reported in Zscores (SPM{Z}).
Significantly activated areas were determined based on
the change in a single voxel at a threshold of p< .05
after correction for multiple, nonindependent compar-
isons. The voxel significance threshold was estimated
using the theory of Gaussian fields (Friston, Frith,
Liddle, & Frackowiak, 1991; Friston, Worsley, Poline,
Frith, & Frackowiak, 1995). The resulting foci were then
characterized in terms of peak Zscores above this level.
Cognitive Tasks
The experiment consisted of the following nine tasks:
one pattern-discrimination task (where subjects had to
decide whether gratings were horizontal or vertical), four
object-decision tasks (which differed in that the real
objects would belong to either the category of vegeta-
bles / fruit, animals, articles of clothing, or nonmanipu-
lable man-made objects), and four categorization tasks
(which differed in that the objects presented in the
critical scan window would come predominantly from
the category of either vegetables/ fruit, animals, articles
of clothing, or nonmanipulable man-made objects). The
order of tasks was randomized across subjects with the
only constraint being that half of the subjects first
performed the object-decision tasks while the other half
first performed the categorization tasks. In all tasks, the
subjects were encouraged to respond as fast and as
accurately as possible. Before the actual experiments
started, the subjects performed a practice version of each
1236 Journal of Cognitive Neuroscience Volume 14, Number 8

task while in the scanner. Stimuli used in these practice
versions were not used in the actual experiments.
In the present paper, only results from a subset of
these tasks will be reported. This subset includes two
of the object-decision tasks and the four categoriza-
tion tasks.
In the two object-decision tasks, the subjects were
presented with pictures that represented either real
objects or nonobjects (see Figure 1B). In these tasks
the subjects were instructed to press the ‘‘real object’’
key (index finger), on a serial response box placed in
front of their right hand, if the picture represented a real
object and the ‘‘nonobject’’ key (middle finger) if it
represented a nonobject. Evidence from previous PET
studies suggests that this task places greater demands
on shape processing (structural knowledge) than on
semantic processing (Gerlach et al., 1999, 2000). In the
four categorization tasks the subjects were presented
with pictures from four different categories of natural
and man-made objects (see Figure 1A) and had to press
the ‘‘natural’’ key (index finger) if the picture repre-
sented a natu ral ob ject and the ‘‘man-m ade’’ key
(middle finger) if the picture represented a man-made
object. Evidence from a previous PET study suggests that
this type of task requires more semantic processing than
object-decision tasks (Gerlach et al., 2000).
The two object-decision tasks differed in that they
contained a predominance (84%) of either real vegeta-
bles/ fruit or real articles of clothing in the part of the
task that was presented in the critical scan window
(cf. the section on design). The categorization tasks
differed in that they contained a predominance (84%)
of either vegetables/ fruit, animals, articles of clothing, or
nonmanipulable man-made objects in the part of the
task that was presented in the critical scan window
(cf. the section on design).
Design
Seventy stimuli were presented in each task. All stimuli
were presented on a white background on a PC mon-
itor hanging 60 cm in front of the subjects. They
subtended between 38and 58of visual angle and were
presented in the center of gaze. Each stimulus was
displayed for 180 msec, with an interstimulus interval
of 1320 msec, making each task last 1 min and 45 sec.
All tasks were initiated approximately 1 min and 15 sec
prior to isotope arrival to the brain and continued
during the first 30 sec of acquisition corresponding to
the delivery of radiotracer to the brain. From the point
of task offset, the subjects viewed a blank screen for the
next 60 sec, yielding a total acquisition time of 90 sec.
By reducing isotope washout and improving counting
statistics this protocol optimizes the signal-to-noise ratio
from activated regions (Cherry, Woods, Doshi, Banerjee,
& Mazzio tta, 1995; Hur tig et al., 1994; Silbersweig
et al., 1993).
Each object-decision task consisted of line drawings
of 35 real objects and 35 nonobjects. However, the
presentation was blocked in two so that the first block
consisted of 19 real objects (either fruit /vegetables or
articles of clothing) + 32 nonobjects, whereas the
second block consisted of 16 real objects (either
fruit / vegetables or articles of clothing) + 3 nonobjects.
Each categorization task also comprised 70 line draw-
ings (35 natural and 35 man-made objects). These tasks
were blocked in the same way as the object-decision
tasks. Accordingly, in the categorization task for ani-
mals, the first block consisted of 19 animals +32 man-
made objects, whereas the second block consisted of
16 animals + 3 man-made objects. The categorizatio n
tasks for fruit / vegetables, articles of clothing, and non-
manipulable man-made objects were arranged similarly.
In all tasks, the order of the pictures (real vs. non-
object / natural vs. man-made) was randomized within
each block. The 51 items presented in the first block
within each task were unique in that they did not
appear in any of the other tasks. This was also true for
3 of the items presented in the second block within
each task. However, the remaining 16 items presented
in the second block within each task appeared twice,
once in an object-decision task and once in a catego-
rization task. Thus, as an example, the same 16 animals
would serve as real objects in an object-decision task
and as natural objects in a categorization task (appear-
ing in the second block of each task together with
either 3 unique nonobjects or 3 unique man-made
objects depending on the task).
In all tasks the two blocks were presented sequen-
tially, but arranged so that the first block would be
initiated approximately 45 sec before injection and last
until the bolus was estimated to reach the brain. The
second block was displayed in the actual uptake phase
of the tracer and ended before wash out was likely
to begin (the critical scan window). Because of this
arrangement, the activation seen during the six tasks
should primarily reflect structural or semantic process-
ing of either animals, fruit / vegetables, articles of cloth-
ing, or nonmanipulable man-made objects depending
on the particular task. Thus, this blocked design allowed
us to compare rCBF across categori es, while task
requirements were kept constant across a particular
type of task (object-decision or categorization).
Stimuli
The nonobjects (see Figure 1B) were selected mainly
from the set made by Lloyd Jones and Humphreys
(1997). These nonobjects are chimeric line drawings of
closed figures constructed by exchanging single parts
belonging to objects from the same category. Because
these nonobjects are composed of parts of objects from
the same category, they could be considered either
nonsense fruit / vegetables or nonsense clothing. One
Gerlach, Law, and Paulson 1237

set of nonsense fruit/vegetables was used in the object-
decision task together with real fruit / vegetables, and one
set of nonsense clothing was used in the object-decision
task together with real articles of clothing. The line
drawings of real objects presented in the first block of
each task were selected from various sources but mainly
from the standardized set of Snodgrass and Vanderwart
(1980). Care was taken to insure that the pictures looked
similar overall, regardless of source. The four sets of
pictures of real objects (fruit / vegetables, animals, articles
of clothing and nonmanipulable man-made objects)
used in the second block of the object-decision tasks
and the categorization tasks were all selected from the
pool of Snodgrass and Vanderwart and were matched
with respect to familiarity, visual complexity, and image
agreement, so that they did not differ significantly along
any of these dimensions (Kruskal-Wallis, p> .1).
Data Analysis
To identify brain areas selectively activated during cate-
gorization of manipulable objects regardless of category
(natural or man-made) we looked for areas that were
commonly activated by vegetables/fruit relative to ani-
mals and by articles of clothing relative to nonmanipu-
lable man-made objects. This was done by use of
conjunction analysis (Price & Friston, 1997). To ensure
that the areas associated with this analysis were more
activated during categorizatio n of vegetables /fruit and
articles of clothing than during object decisions on the
same items, we masked the conjunction with the simple
contrasts between categorization of vegetables/ fruit
versus object decisions on vegetables / fruit and catego-
rization of articles of clothing versus object decisions on
articles of cloth ing. The threshold for these masks was
set at p< .001 uncorrected for multiple comparisons.
The reasons why we adopted this conjunction approach
in addition to the regular factorial approach were the
following: (i) the conjunction analysis is more sensitive
than a regular, main-effect analysis because it tests for
the conjoined probability that two independent con-
trasts will be associated with activation of the same area;
(ii) as o pposed to a regular, m ai n-effect analysis,
conjunction analysis discounts areas associated with
interactions.
Acknowledgments
This work was supported by a grant to the first author from the
Danish Medical Research Council and the Danish Research
Council for the Humanities. Karin Stahr and the staff at the PET
center at Rigshospitalet, Copenhagen, are acknowledged for
their participation. The Danish Research Centre for Magnetic
Resonance, Hvidovre Hospital, Denmark, is acknowledged for
its participation in the acquisition of structural MRI scans. The
John and Birthe Meyer Foundation is gratefully acknowledged
for the donation of the cyclotron and PET scanner. Finally,
helpful comments by two reviewers are acknowledged.
Reprint requests should be sent to Christian Gerlach, Neuro-
biology Research Unit, N9201, The National University
Hospital, Rigshospitalet, Blegdamsvej 9, 2100 Copenhagen Ø,
Denmark, or via e-mail: gerlach@pet.rh.dk.
REFERENCES
Barbarotto, R., Capitani, E., & Laiacona, M. (2001). Living
musical instruments and inanimate body parts?
Neuropsychologia, 39, 406 – 414.
Binkofski, F., Buccino, G., Posse, S., Seitz, R. J., Rizzolatti, G.,
& Freund, H. (1999). A fronto-parietal circuit for object
manipulation in man: Evidence from an fMRI-study.
European Journal of Neuroscience, 11, 3276 – 3286.
Buxbaum, L. J. (2001). Ideomotor apraxia: A call to action.
Neurocase, 7, 445 – 458.
Buxbaum, L. J., Schwartz, M. F., & Carew, T. G. (1997).
The role of semantic memory in object use. Cognitive
Neuropsychology, 14, 219 – 254.
Buxbaum, L. J., Veramonti, T., & Schwartz, M. F. (2000).
Function and manipulation tool knowledge in apraxia:
Knowing ‘‘what for’’ but not ‘‘how.’’ Neurocase, 6, 83 – 97.
Caramazza, A., & Shelton, J. R. (1998). Domain-specific
knowledge systems in the brain the animate-inanimate
distinction. Journal of Cognitive Neuroscience, 10, 1 – 34.
Chao, L. L., & Martin, A. (2000). Representation of manipulable
man-made objects in the dorsal stream. Neuroimage, 12,
478 – 484.
Cherry, S. R., Woods, R. P., Doshi, N. K., Banerjee, P. K., &
Mazziotta, J. C. (1995). Improved signal-to-noise in PET
activation studies using switched paradigms. Journal of
Nuclear Medicine, 36, 307 – 314.
Damasio, A. R. (1990). Category-related recognition defects as
a clue to the neural substrates of knowledge. Trends in
Neurosciences, 13, 95 –98.
De Renzi, E. (1989). Apraxia. In F. Boller & J. Grafman (Eds.),
Handbook of neuropsychology (vol. 2, pp. 245 – 263).
Amsterdam: Elsevier.
DeGrado, T. R., Turkington, T. G., Williams, J. J., Stearns, C. W.,
Hoffman, J. M., & Coleman, R. E. (1994). Performance
characteristics of a whole-body PET scanner. Journal of
Nuclear Medicine, 35, 1398 – 1406.
Devlin, J. T., Moore, C. J., Mummery, C. J., Gorno Tempini,
M. L., Philliphs, J. A., Noppeney, U., Frackowiak, R. S. J.,
Friston, K. J., & Price, C. J. (2002). Anatomic constraints on
cognitive theories of category specificity. Neuroimage, 15,
675 – 685.
Devlin, J. T., Russell, R. P., Davis, M. H., Price, C. J., Moss, H. E.,
Fadili, M. J., & Tyler, L. K. (2001). Is there an anatomical basis
for category-specificity? Semantic memory studies in PET
and fMRI. Neuropsychologia, 40, 54 – 75.
Dixon, M. J., Piskopos, M., & Schweizer, T. A. (2000). Musical
instrument naming impairments: The crucial exception to
the living / nonliving dichotomy in category-specifi c agnosia.
Brain and Cognition, 43, 158 – 164.
Farah, M. J., & McClelland, J. L. (1991). A computational model
of semantic memory impairment: Modality specificity and
emergent category specificity. Journal of Experimental
Psychology: General, 120, 339 – 357.
Forde, E. M. E., & Humphreys, G. W. (1999). Category-specific
recognition impairments: A review of important case studies
and influential theories. Aphasiology, 13, 169 – 193.
Friston, K. J. (1994). Statistical parametric mapping. In R. W.
Thatcher, M. Hallett, T. Zeffiro, E. R. John, & M. Huerta
(Eds.), Functional neuroimaging: Technical foundations.
(pp. 79 – 93). San Diego: Academic Press.
Friston, K. J., Ashburner, J., Frith, C. D., Poline, J. B., Heather,
1238 Journal of Cognitive Neuroscience Volume 14, Number 8

J. D., & Frackowiak, R. S. (1995). Spatial registration and
normalization of images. Human Brain Mapping, 3,
165 – 189.
Friston, K. J., Frith, C. D., Liddle, P. F., & Frackowiak, R. S.
(1991). Comparing functional (PET) images: The assessment
of significant change. Journal of Cerebral Blood Flow and
Metabolism, 11, 690 – 699.
Friston, K. J., Worsley, K. J., Poline, J. B., Frith, C. D., &
Frackowiak, R. S. (1995). Statistical parametric maps in
functional imaging: A general linear approach. Human
Brain Mapping, 2, 189 – 210.
Gainotti, G. (2000). What the locus of brain lesion tells
us about the nature of the cognitive defect underlying
category-specific disorders: A review. Cortex, 36,
539 – 559.
Gerlach, C., Law, I., Gade, A., & Paulson, O. B. (1999).
Perceptual differentiation and category effects in
normal object recognition: A PET study. Brain, 122,
2159 – 2170.
Gerlach, C., Law, I., Gade, A., & Paulson, O. B. (2000).
Categorization and category effects in normal object
recognition: A PET study. Neuropsychologia, 38, 1693 – 1703.
Gerlach, C., Law, I., Gade, A., & Paulson, O. B. (2002).
The role of action knowledge in the comprehension of
artefacts—a PET study. Neuroimage, 15, 143 – 152.
Grabowski, T. J., Damasio, H., & Damasio, A. R. (1998).
Premotor and prefrontal correlates of category-related
lexical retrieval. Neuroimage, 7, 232 – 243.
Grafton, S. T., Fadiga, L., Arbib, M. A., & Rizzolatti, G. (1997).
Premotor cortex activation during observation and naming
of familiar tools. Neuroimage, 6, 231 – 236.
Haaland, K. Y., Harrington, D. L., & Knight, R. T. (2000).
Neural representations of skilled movement. Brain, 123,
2306 – 2313.
Hart, J., Berndt, R. S., & Caramazza, A. (1985). Category-specific
naming deficit following cerebral infarction. Nature, 316,
439 – 440.
Hart, J., & Gordon, B. (1992). Neural subsystems for object
knowledge. Nature, 359, 60 – 64.
Heilman, K. M., Maher, L. M., Greenwald, M. L., & Rothi, L. J.
(1997). Conceptual apraxia from lateralized lesions.
Neurology, 49, 457 – 464.
Hurtig, R. R., Hichwa, R. D., O’Leary, D. S., Boles Ponto, L. L.,
Narayana, S., Watkins, G. L., & Andreasen, N. C. (1994).
Effects of timing and duration of cognitive activation in
[
15
O]water PET studies. Journal of Cerebral Blood Flow
and Metabolism, 14, 423 – 430.
Lakoff, G. (1987). Women, fire, and dangerous things: What
categories reveal about the mind. Chicago: University of
Chicago Press.
Leiguarda, R. C., & Marsden, C. D. (2000). Limb apraxias:
Higher-order disorders of sensorimotor integration. Brain,
123, 860 – 879.
Lloyd Jones, T. J., & Humphreys, G. W. (1997). Perceptual
differentiation as a source of category effects in object
processing: Evidence from naming and object decision.
Memory and Cognition, 25, 18 – 35.
Martin, A., Wiggs, C. L., Ungerleider, L. G., & Haxby, J. V.
(1996). Neural correlates of category-specific knowledge.
Nature, 379, 649 – 652.
Miller, G. A., & Johnson-Laird, P. N. (1976). Language and
perception. Cambridge, MA: Harvard University Press.
Milner, A. D., & Goodale, M. A. (1995). The visual brain in
action. Oxford, UK: Oxford University Press.
Moreaud, O., Charnallet, A., & Pellat, J. (1998). Identification
without manipulation: A study of the relations between
object use and semantic memory. Neuropsychologia, 36,
1295 – 1301.
Mummery, C. J., Patterson, K., Price, C. J., Ashburner, J.,
Frackowiak, R. S. J., & Hodges, J. R. (2000). A voxel-based
morphometry study of semantic dementia: Relationship
between temporal lobe atrophy and semantic memory.
Annals of Neurology, 47, 36 – 45.
Murata, A., Fadiga, L., Fogassi, L., Gallese, V., Raos, V., &
Rizzolatti, G. (1997). Object representation in the ventral
premotor cortex (area F5) of the monkey. Journal of
Neurophysiology, 78, 2226 – 2230.
Price, C. J., & Friston, K. J. (1997). Cognitive conjunction:
A new approach to brain activation experiments.
Neuroimage, 5, 261 – 270.
Rizzolatti, G., & Arbib, M. A. (1998). Language within our grasp.
Trends in Neurosciences, 21, 188 – 194.
Rizzolatti, G., & Fadiga, L. (1998). Grasping objects and
grasping action meanings: The dual role of monkey
rostroventral premotor cortex (areas F5). In Sensory
guidance of movements, Novartis Foundation Symposium
218 (pp. 81 – 103). Chichester, UK: Wiley.
Silbersweig, D. A., Stern, E., Frith, C. D., Cahill, C., Schnorr, L.,
Grootoonk, S., Spinks, T., Clark, J., Frackowiak, R., & Jones,
T. (1993). Detection of thirty-second cognitive activations
in single subjects with positron emission tomography:
A new low-dose H
2
(15)O regional cerebral blood flow
three-dimensional imaging technique. Journal of Cerebral
Blood Flow and Metabolism, 13, 617 – 629.
Sirigu, A., Duhamel, J. R., & Poncet, M. (1991). The role of
sensorimotor experience in object recognition. A case of
multimodal agnosia. Brain, 114, 2555 – 2573.
Snodgrass, J. G., & Vanderwart, M. A. (1980). A standardized
set of 260 pictures: Norms for name agreement, image
agreement, familiarity, and visual complexity. Journal of
Experimental Psychology: Human Learning and Memory,
6, 167 – 182.
Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic
atlas of the human brain. 3-dimensional proportional
system: An approach to cerebral imaging. Stuttgart,
Germany: Thieme.
Tranel, D., Adolphs, R., Damasio, H., & Damasio, A. R. (2001).
A neural basis for the retrieval of words for actions.
Cognitive Neuropsychology, 18, 655 – 674.
Tranel, D., Damasio, H., & Damasio, A. R. (1997). A neural
basis for the retrieval of conceptual knowledge.
Neuropsychologia, 35, 1319 – 1327.
Tranel, D., Logan, C. G., Frank, R. J., & Damasio, A. R.
(1997). Explaining category-related effects in the
retrieval of conceptual and lexical knowledge for concrete
entities: Operationalization and analysis of factors.
Neuropsychologia, 35, 1329 – 1339.
Tyler, L. K., Moss, H. E., Durrant-Peatfield, M. R., & Levy, J. P.
(2000). Conceptual structure and the structure of concepts:
A distributed account of category-specific deficits. Brain and
Language, 75, 195 – 231.
Warrington, E. K., & McCarthy, R. A. (1987). Categories of
knowledge. Further fractionations and an attempted
integration. Brain, 110, 1273 – 1296.
Gerlach, Law, and Paulson 1239