The Role of the Right Hemisphere
in Resolving Lexical Semantic Ambiguity
Laurie A. Stowe, Albertus A.Wijers, Anne M.J. Paans, Marco Haverkort,
Cees A.J. Broere, Gijsbertus Mulder, Frans Zwarts, and Willem Vaalburg
Dept. of Linguistics
School of Behavioral and Cognitive Neurosciences
Rijksuniversiteit Groningen
Postbus 716
9700 AS Groningen
The Netherlands
Email: [e-mail]
Telephone: [phone]
Telefax: [phone]
Stowe et al
Abstract
Although the right hemisphere is capable of comprehending words, to what extent and
under what circumstances this contributes to normal natural language comprehension
has remained obscure. The positron emission tomography (PET) functional
neuroimaging experiment reported here showed that right hemisphere inferior frontal
gyrus and dorsolateral prefrontal cortex are activated while subjects read temporarily
ambiguous sentences in which the less preferred meaning of the word must finally be
chosen. This result demonstrates that the right hemisphere is indeed involved in normal
sentence comprehension under some circumstances. We relate this evidence to data from
right hemisphere damaged patients which suggests that the right hemisphere is involved
in revising meanings more generally and to neuroimaging evidence that the right frontal
lobe is active in verbal short-term memory tasks.
2
The Right Hemisphere and Semantic Ambiguity
Introduction
Although it is clear that the left hemisphere is generally dominant in language
function, the extent to which the right hemisphere is also normally involved in language
processing remains a source of controversy. It is well known from experiments with
split-brain patients that the right hemisphere s abilities to carry out most aspects of
linguistic processing are limited, at least when it has developed connected with a
language dominant left hemisphere. Zaidel and Peters (1981) discuss split-brain
phonological analysis abilities, Baynes et al, (1995) phonological production, Burgess
and Skodis (1993) morphological analysis, and Zaidel (1977) and Gazzaniga et al (1984)
syntactic processing. All of these are apparently limited.
Nevertheless, the right hemisphere does appear to play some role in language
processing. Data from right brain damaged patients indicate that the right hemisphere is
involved in interpretation of text and discourse (Brownell et al, 1986; McDonald and
Wales, 1986). In addition it appears to be capable of lexical semantic processing to some
extent (Gainotti et al, 1981; Zaidel, 1976), although it does not appear to process lexical
semantics in quite the same way as the left hemisphere. In particular, the right
hemisphere appears to respond differently than the left hemisphere to semantic
ambiguity (Burgess and Simpson, 1988; Deloche et al, 1987; Faust and Gernsbacher,
1996). This difference may allow the right hemisphere to make an important
contribution to comprehension of sentences containing lexical semantic ambiguities.
However, this hypothesis is difficult to test with response time techniques, as the
contribution of the right hemisphere to the response is difficult to establish definitively.
In the experiment reported here, we used positron emission tomography to
identify regions of the brain which are more involved in processing sentences containing
lexical semantic ambiguities than sentences containing no ambiguity. A number of
functional neuroimaging studies have demonstrated that there are large areas of
activation in the right hemisphere during tasks which involve word recognition (e.g.
Price et al, 1996; Stowe et al, In press). However, it is difficult to be certain to what
extent these activations are due to sensory processing and what is due to semantic
processing. However, it is possible to test one aspect of lexical semantic processing,
which will then provide evidence on the role of the right hemisphere in semantic
processing. The results of the experiment suggest that the right hemisphere is actively
involved in resolving lexical ambiguities. In the discussion we will relate these results to
the literature on the linguistic processing abilities of the right hemisphere and discuss the
cognitive function(s) supported by the right hemisphere which may explain its role in
comprehension of semantically ambiguous sentences.
The Contribution of the Right Hemisphere to Lexical Semantics
Evidence that the right hemisphere is capable of recognizing words comes from
several sources. Split brain patients clearly have access to at least some knowledge about
the meaning of words (Gazzaniga et al, 1984; Kutas et al, 1988; Zaidel, 1976, 1977).
Zaidel (1977) estimated that the isolated right hemisphere has a vocabulary that is
similar to that of an 11-year-old child. Electrical stimulation of the right hemisphere
leads to naming errors and substitutions as well as delay and failure to produce words in
left hemisphere dominant epilepsy patients during pre-operative functional mapping
(Andy and Bhatnagar, 1984). Substitution errors in particular suggest that the right
hemisphere is involved in word processing at a higher level than motor function.
However, due to the pathological condition of these patients, some language
3
Stowe et al
reorganization may have taken place, so it is not entirely clear how this evidence bears
on normal language processing.
Right brain damaged patients also show evidence of lexical semantic processing
difficulties. Although some of the studies are liable to alternative interpretations, this
does not appear to be true for all of them. Joanette et al (1988) reported that right brain
damage patients were worse at generating words from a particular semantic class
(semantic fluency) than generating words on the basis of the initial phoneme (phoneme
fluency). This suggests that the right hemisphere is involved in semantic processing but
not phonological processing. However, Goulet et al (1997) argued that this pattern did
not necessarily result from a specific impairment to lexical semantics, as the semantic
fluency tests were typically harder than the phoneme fluency tests for normal controls.
As a general cognitive impairment has been found for right brain damaged patients, this
difference in difficulty could predict more impairment in semantic fluency than in
phonological fluency. Gainotti et al (1981) found that right brain damage impairs word-
picture matching more than a phoneme discrimination task. This remained true even
when general cognitive impairment was parcelled out in the analysis. A possible
alternative explanation of this result is that the patients were suffering from visual
neglect, although the authors attempted to control for this by using vertically centered
pictures from which the subjects had to select one which matched the word. These
alternative explanations do not appear to apply to a study reported by Chiarello and
Church (1986). They compared right brain damaged and left brain damaged patients
with normal controls on a task in which subjects were asked to match words on the basis
of semantic relationship or phonological relationship. Left brain damaged patients found
the phonological matching task harder than the semantic matching task; right brain
damaged patients found it more difficult to match words on the basis of semantic
similarity than phonological similarity. Crucially, the normal control subjects performed
equivalently on both tasks. Visual neglect and general cognitive impairment would thus
presumably have affected both tasks equally.
The evidence discussed up to this point suggests that the right hemisphere is
capable of processing lexical semantics and that it also makes a contribution beyond that
of the left hemisphere. Otherwise, it would not be expected that right brain damage
would cause a deficit in lexical semantic processing. However, it seems likely that the
right hemisphere does not act as an  inferior left hemisphere in language
comprehension. The right hemisphere appears to process some aspects of semantics
differently than the left hemisphere. For example, the right hemisphere may encode
concrete or imageable words more strongly than abstract words (Zaidel, 1994; Deloche
et al, 1987). A metaphorical meaning of a word may also be more readily activated in the
right hemisphere than the left (Brownell et al, 1984, 1990; Bottini et al, 1994). Idiomatic
meaning may also be accessed partially via the right hemisphere (Myers and Linebaugh,
1981; Van Lancker and Kempler, 1987). The right hemisphere also appears to activate
distant associates of a word for a longer time than the left hemisphere (Koivisto, 1997).
Most important for the current study, the processing of semantically ambiguous
words presented to the left hemisphere differs from the processing of those presented to
the right hemisphere (Burgess and Simpson, 1988; Faust and Gernsbacher, 1996). When
a semantically ambiguous word is presented centrally on a screen (i.e. to both
hemispheres) one meaning is normally quickly chosen. At first, both meanings prime
associated words, but after a short period only one continues to show priming. This
suggests that the other alternative meaning has been dropped. Swinney (1979) discusses
evidence for a decision on the basis of semantic information;, Seidenberg et al (1982)
and Tanenhaus et al (1979) for a decision on the basis of syntactic information, and
4
The Right Hemisphere and Semantic Ambiguity
Simpson et al (1989) for a decision on the basis of frequency. When words are presented
to the left hemisphere alone using the divided visual field paradigm, the same pattern is
seen. However, the right hemisphere apparently maintains both meanings for a longer
time. Burgess and Simpson (1988) showed that both the primary, most frequent meaning
and a secondary, less frequent, meaning of an ambiguous word continue to prime
associated words for a considerable time period when they are presented to the right
hemisphere (in the left visual field). Faust and Gernsbacher (1996) examined the effects
of sentence context on the choice between two meanings of a word. Again, there is a
quick choice on the basis of context in the left hemisphere, but both possibilities remain
active in the right hemisphere.
The difference between the two hemispheres may be due 1) to inhibition of the
less probable meaning in the left hemisphere only or 2) to overall slower processing in
the right hemisphere. These two possibilities have also been discussed as potential
explanations for a similar dissociation between the two hemispheres which has been
found for distant associates (Koivisto, 1997). If the left hemisphere, but not the right,
inhibits an unlikely meaning (as well as other unnecessary semantic information), the
dissociation may be related to attentional control effects. Lambert and Voot (1993)
showed bigger effects of semantic processing of unattended words presented in the left
visual field; in this case, too, the right hemisphere apparently does not suppress
processing as thoroughly as the left, so that semantic processing occurs despite the
instructions to ignore certain stimuli which were given to the subjects.
Whatever explanation is correct, the availability of a second meaning in the right
hemisphere may play a role in ambiguity resolution when the meaning which was
initially chosen turns out to be incorrect. Divided visual field studies, although they are
compatible with this hypothesis, are open to questions with regard to effects of
interhemispheric transfer of information and of motor response. The evidence from
lesion studies relevant to this hypothesis is mixed. Zaidel et al (1995) examined the
effects of anterior temporal lobectomy on the comprehension of syntactically and
semantically ambiguous sentences. They showed that right anterior temporal lobectomy
leads to impaired recognition of a second meaning of a semantically ambiguous
sentence; the effect is as strong as that found for the left hemisphere. On the other hand,
Brownell et al (1990) examined the availability of a second meaning in right hemisphere
damaged patients and found that the impairment was not nearly as large as those they
found for alternative metaphoric meanings. This evidence suggests that the left
hemisphere alone can find both meanings relatively easily. The differences between the
patient groups and experimental paradigms make it difficult to compare these results.
Neuroimaging provides a source of information that may be more definitive.
The Goal of the Experiment
The goal of this experiment, at the highest level, is to investigate the role of the
right hemisphere in language comprehension. As we pointed out in the introduction,
neuroimaging studies have shown that the right hemisphere is activated during language
comprehension, but the nature of the cognitive process involved is difficult to determine.
The resolution of semantic ambiguity provides a specific example of semantic
processing which is susceptible to experimental test.
As we have discussed, in general it appears that a quick choice is made between
potential meanings of an ambiguous word. However, this is not always advantageous. In
a sentence, the more frequent or initially more plausible meaning does not always turn
out to be the correct meaning. If the secondary meaning remains active for a longer time
5
Stowe et al
in the right hemisphere, it may be available there to aid in recovery from the initial
miscomprehension. The goal of the current experiment is to examine to what extent the
right hemisphere is active during the resolution of semantic ambiguities, particularly
when the most likely meaning of the sentence has to be rejected for one which initially
seemed less likely. Finding such an activation therefore provides support for the theory
that the right hemisphere contributes to certain aspects of the processing of semantics in
normal subjects.
Experiment
Subjects
Fourteen volunteers (8 female; 6 male; age range 19-26; mean = 21.6)
participated in this experiment after giving written consent on the basis of written
information about the experiment and PET technique under a procedure cleared with the
University Hospital Medical Ethics Committee and in accordance with the declaration of
Helsinki. All subjects were right handed native speakers of Dutch, had normal or
corrected to normal vision, and had no history of neurological problems. Two additional
subjects were excluded from the analysis because it was not possible to adequately
correct for movement between scans.
Materials
Ambiguous Dutch sentences were constructed, each of which contained a
semantically ambiguous word (e.g., De ezel staat in de schuur allang te rotten, literally,
The donkey/easel stands in the shed already-long to rot). The ambiguous sentences were
constructed so that the meaning remained ambiguous for at least three words after the
ambiguous word appeared; the final words of the sentence were only sensible if the
initially less preferred meaning was used. Initial meaning preference was determined by
means of a sentence completion task. On average the meaning favoured in the
experimental sentences was produced in 20% of the completions; it was never used more
frequently than the alternative meaning. Initial preference for one interpretation over the
other may be due to the frequency of the alternatives, to the semantic context provided
by sentential context, or to a combination of both these factors.
Control sentences were constructed with similar syntactic structures to those
used in the ambiguous sentences (e.g. De thee was in de winkel toch wel duur, lit; The
tea was in the store actually rather expensive). Although it was impossible to completely
avoid ambiguous words, they were immediately syntactically disambiguated in these
sentences. Ambiguous and unambiguous sentences were each distributed into two lists;
the resulting four lists were matched as closely as possible on length of sentences, word
length (mean word lengthin letters per list: Ambig-L1 = 5.22; Ambig-L2 = 4.78;
Control-L1 = 4.86; Control-L2 =5.16); and frequency for content words (mean
logarithmic word frequency per list: Amb1 = 3.10; Amb2 = 3.02 Control1 = 3.06
Control2 = 3.11).. Additionally the plausibility of the sentences was rated on a scale
from 1 to 5; 1 was impossible, and 5 was completely predictable. The lists were
matched as well as possible on plausibility (Mean plausibility rating per list: Ambig-L1
= 2.8 Ambig-L2 = 2.6 Control-L1 = 3.3 Control-L2 = 3.2). Shifting meaning in the
ambiguous sentences may cause a subjective impression of slightly lessened plausibility,
but the difference was not significant.
Procedure
6
The Right Hemisphere and Semantic Ambiguity
Subjects were placed in a Siemens CTI (Knoxville, Tennessee, USA) 951/31
positron emission tomography camera parallel to and centered 3 cm above the glabella-
inion line (Tokunaga et al, 1977). Subjects read sentences presented one word at a time
(750 msec/word) in the middle of a computer screen suspended approximately 90 cm
from the subject s eyes. A practice list was presented while an attenuation scan was
made, allowing the subject to accustom themselves to the procedure and simultaneously
checking that all aspects of the experimental set-up were functioning. Each of the four
lists described above (two containing ambiguous sentences, two containing
unambiguous controls) was presented during a separate scan. Subjects were informed
via the monitor when each scan was to begin and instructed to read in order to
understand the sentences. No additional task was required of them. During a fifth scan,
the same procedure was followed except that the subjects received no input; they were
simply asked to look at an asterisk in the center of the monitor for the entire period
(passive fixation).
Previous to each list, 1.85 GBq H215O was injected as a bolus into the right
brachial vein followed by 40 ml saline via an automatized injector. Presentation of
sentences began seven seconds after injection. Data acquisition began 23 sec after
injection, by which time the peak in radioactivity was assumed to have reached the brain,
and was continued for 60 sec. Fifteen minutes between injections was allowed for ac-
tivity to decrease to background level. To control for effects of the sequence in which
the scans were made, for example learning or attentional confounds, a different order of
the four lists was used for each subject, with each list coming approximately equally
frequently during the first, second, fourth or fifth scan. The passive fixation scan was
always presented third.
Data Analysis
Images were constructed in which regional blood flow was estimated for each
voxel in the camera field of view, using a correction based on the attenuation scan. This
procedure corrected for decreases in apparent activity due to the tissue through which the
radiation passed. The images were then resampled to create images containing 2 x 2 x
2.4mm voxels. Although the subjects were placed in a head mould to control movement,
this was not adequate to prevent all movement. Therefore, a least mean squares
procedure was used to align all the scans made for each subject (Woods et al, 1992).
Two additional subjects were left out of the analysis, as the residual differences in
comparisons between the images showed that the realignment procedure was not
successful.
The data was further analyzed using the Statistical Parametric Mapping (SPM)
program developed by the Wellcome Institute of Cognitive Neurology, London, UK.
Since there is a great deal of anatomical variability between individuals, it is necessary to
normalize the data from each subject into a stereotactic co-ordinate system (Talairach
and Tournoux, 1988) in order to align the voxels from each subject that are anatomically
comparable. The procedure for doing this combines linear (size and orientation) and
non-linear (warping) components and is completely described by Friston et al (1995a).
This procedure, however, may not always align the data from different subjects
perfectly, since gross anatomical landmarks do not always correlate with
cytoarchitecture (Roland and Zilles, 1998). Therefore, a Gaussian filter (20 mm in the
anterior/posterior and left/right dimensions and 12 mm in the dorsal/ventral dimension)
was applied; this  smears the image out and should make nearly superimposed
activations statistically detectable.
7
Stowe et al
Statistical analyses were carried out at each voxel. First an ANCOVA was used
to parcel out effects of global differences in activity, which may result from either
differences in the dose that was delivered or actual differences in blood flow. Tests were
carried out for increased and decreased blood flow in response to the ambiguous
sentence, which produced a Z-value for each voxel and an uncorrected probability for
the significance of the difference. However, due to the large number of comparisons
which are carried out across the total measured brain volume, this probability
overestimates the actual probability of significance. Therefore a correction of
probability taking into account multiple comparisons was carried out (corrected P). For a
more complete description of these procedures see Friston et al (1995b).
Lastly, a calculation was also made for the likelihood that a cluster of
significantly activated voxels will occur by chance (Probability of extent), using the
procedure developed by Friston et al (1994). We checked for clusters of voxels with a
threshold of uncorrected probability < 0.001. This calculation is based on the fact that,
although voxels may show random differences, contiguous clusters of randomly
activated voxels are unlikely to occur by chance. Differences in regional cerebral blood
flow will be regarded as significant in this article if they reach significance for either
extent or for a single maximal voxel. However, these criteria are quite strict, and
additional areas will be discussed if there is some reason with respect to the point which
is being addressed. Obviously, no firm conclusions will be made on the basis of these
data, however.
Results
The analysis showed significantly increased activation for the semantically
ambiguous sentences relative to the control sentences in the inferior right frontal lobe,
including most of Brodmann's area 47 and portions of Brodmann's areas 44 and 45
(extent 394 voxels at a threshold of P < .001; P(of extent) = .005; maximal Z = 4.36,
corrected P = .028) The location of the maximum voxel in the co-ordinate system of
Talairach and Tournoux was x = 38, y = 40, z = -4 (see Figure 1). An area including
parts of the right posterior superior temporal gyrus and inferior parietal lobule was the
only other area which showed any activation in the ambiguous vs. unambiguous
comparison, although it failed to reach significance (Z = 3.09; uncorrected P = .802).
Unambiguous sentences did not show any areas of increased activation relative to
ambiguous sentences (all corrected voxel or extent P's > 0.7).
________________________________________________________________
Figure 1 about here.
_______________________________________________________________
Four additional comparisons were carried out. These tested for increased or
decreased blood flow relative to the passive fixation condition for the ambiguous and
unambiguous scans respectively. These comparisons provide additional information
relative to whether the activation found in the main comparisons should be interpreted as
increased blood flow during the reading of semantically ambiguous sentences or may
reflect decreased blood flow during the reading of unambiguous controls. When the
scans during which subjects read semantically ambiguous sentences were compared with
the passive fixation scan, there were a number of activations (see Figure 2). Crucially,
there was an activation of the right inferior frontal gyrus which was significant according
to the extent statistic (813 voxels at cut-off Z = 2.33, P (of extent) = .018; maximal voxel
8
The Right Hemisphere and Semantic Ambiguity
x = 44, y = 28, z = 4 Z = 3.27, corrected P = .374). The area of activation included much
of the area which was activated in the comparison of ambiguous sentences with control
sentences, although the largest difference was somewhat different, 12 mm further back
and 8 mm higher.
________________________________________________________________
Figure 2 about here.
________________________________________________________________
________________________________________________________________
Figure 3 about here.
________________________________________________________________
The comparison of scans during which subjects read the control sentences with
the passive fixation scan showed neither increased blood flow nor decreased blood flow
in the area activated in the main comparison (see Figure 3). This remained true even
when the criteria for significance was substantially lowered. Since these comparisons are
planned comparisons looking for evidence for changed blood flow in one particular
region of the brain, the correction for multiple comparisons is not appropriate. This
suggests that the activation seen in the main comparison is due to increased blood flow
during the ambiguous scans as opposed to decreased blood flow in the unambiguous
scans.
Discussion
The major goal of the current experiment was to determine whether the right
hemisphere plays a role in lexical semantic processing during sentence comprehension.
The results show that the right hemisphere is involved in normal language
comprehension. This is important, as some earlier results may reflect abnormal
reorganization of language resulting from pathological conditions such as epilepsy, while
others may depend in large part on the experimental method used: divided visual field
presentation.
More specifically, the right hemisphere appears to play a role in the resolution of
lexical semantic ambiguities, as suggested by divided visual field studies (Burgess and
Simpson, 1988; Faust and Gernsbacher, 1996). These studies suggested that the right
hemisphere maintains a second meaning of a word for a longer time than the left
hemisphere. This information is thus available when revision of the initial interpretation
of a sentence is necessary. This logic inspired the experiment reported here and provides
a straightforward explanation of the results. However, the exact cognitive function which
leads to the activation reported here requires more discussion. What is the frontal lobe of
the right hemisphere doing during the processing of lexically ambiguous sentences?
Function of the Right Frontal Lobe in Lexical Semantic Ambiguity Resolution:
Memory
Our original suggestion, based on divided visual field studies, was that the right
hemisphere maintains a second meaning of an ambiguous word, which is then available
9
Stowe et al
if revision becomes necessary. The revision, we assumed, would take place using those
left hemisphere areas which are involved in building up the original interpretation.
Under the hypothesis that the right hemisphere s major role during the
processing of semantically ambiguous sentences is in maintaining a secondary meaning,
it might seem unexpected that the right hemisphere activation found in this experiment is
located in the frontal lobe. Lexical recognition seems more likely to occur in the right
hemisphere homologue of Wernicke s area in the posterior temporal lobe, including
access of a secondary meaning of a semantically ambiguous word. Andy and Bhatnagar
(1984) showed that stimulation of right temporal and parietal sites could lead to
production of word substitutions. More directly, Zaidel et al (1995) showed that right
anterior temporal lobectomy impairs recognition of the second meaning of sentences
containing lexical semantic ambiguities. In fact, as we noted in the Results section, there
was a trend toward activation in the right posterior temporal lobe during the processing
of semantically ambiguous sentences that failed to reach significance. This area may
play a role in accessing the secondary meaning. On the other hand, the inferior frontal
gyri bilaterally are frequently activated during maintenance and retrieval of word lists
(Buckner et al, 1995, 1996; Nyberg et al, 1996). The right frontal lobe activation in this
experiment is thus quite likely to reflect those processes which keep the second meaning
of the ambiguous lexical item available, as opposed to reflecting the original access of
the meaning.
The right frontal lobe subserves processing in a number of tasks which appear to
make use of some form of (verbal) memory. First, memory for verbally presented
information in text or discourse is impaired due to right hemisphere brain damage
(McDonald and Wales, 1986). Right brain damaged patients had more difficulty than
normals in rejecting false statements or inferences (e.g. The bird was in the cage. The
cage was under the table, followed by a later question: Was the bird on the table?). On
the other hand, the patients could recognise the truth of a question that was close to one
that they literally had heard (e.g. Was the cage on the table?). The increase in accuracy
when the question overlapped the original statement suggests that right brain damaged
patients had trouble retrieving information from memory without sufficient overlap in
the input or alternatively that they could not process information from both sentences
simultaneously. Right brain-damaged patients also have trouble in revising inferences
which are inconsistent with information presented later in the text (Brownell et al, 1986);
this may also be related to a memory deficit. Revising inferences requires the use of
information from two or more sentences simultaneously. These deficits found in right
hemisphere patients could result from a limitation of the amount of verbal (possibly
semantic) information that can be maintained or manipulated.
Not all deficits in language comprehension and production found after right
hemisphere brain damage can be so easily summed up as memory effects. It has also
been shown that right brain damaged patients tend to be less sensitive to the knowledge
of their partner in a dialogue, which influences their production and comprehension
adversely. Brownell et al s (1997) right brain damaged patients were less able than
normal controls to use the appropriate formal and informal terms of personal address.
The choice of term of address is based on a recognition of the addressee s knowledge of
the individual to which the speaker is referring and on the social relationship of the
listener and the person being referred to. In the same vein, Siegal et al (1996) showed
that right brain damaged patients tend to have problems with  theory of mind tasks, in
which it must be recognised that others do not have the same information as the listener
does. Although some form of memory limitation might be involved in these cases, too, it
is not as obvious an explanation as with the cognitive impairments which we have just
10
The Right Hemisphere and Semantic Ambiguity
discussed. These deficits thus quite probably result from other cognitive functions of the
right hemisphere. If so, it should be possible to find dissociations between these and the
other deficits described above.
Right dorsolateral cortex and the inferior frontal gyrus are activated by a number
of short-term verbal and non-verbal memory tasks. Buckner et al (1996) and Nyberg et
al (1996) describe a number of these tasks and compare the localizations found when
different elements of memory tasks are emphasized: the initial encoding phase,
maintenance over a short period, and retrieval of the information for specific uses later.
The right hemisphere is particularly activated during tasks in which the information
which was presented earlier must be used to carry out a later task. Nyberg et al (1996)
characterize the cognitive function carried out here as retrieval.
Swick and Knight (1996) emphasize that damage to the right inferior frontal
gyrus does not typically impair cued recall, which apparently contradicts the functional
neuroimaging results, as retrieval clearly takes place in this task. However, like the
improved memory performance for some sorts of questions which was described above
for right brain damaged patients, the overlap between input and memory clearly helps
the patients to retrieve the information from memory. Swick and Knight (1996) point
out that the task was not particularly difficult and that the right frontal lobe may regulate
or control recall of information in tasks where more strategic control is useful. Fletcher
et al (1998) demonstrated that dorsolateral frontal cortex is activated during a task in
which a search of memory must be carried out, while it is not activated when words are
simply presented for recognition. This result supports the hypothesis that the cognitive
function of this area has more to do with the regulation or use of information which is
available in memory. It seems likely that the current results, which in some sense can be
characterized as memory-related, depend on some aspect of this cognitive function. We
will come back to the relationship between this function and the current results below.
McDonald (1993) has also attempted to relate right hemisphere language deficits
to frontal lobe functions. She reviewed the right hemisphere lesion literature and noted
that many of the deficits attributed to right hemisphere damage resemble those attributed
to frontal lobe damage bilaterally. She concluded that there is little reason to assume that
the right hemisphere has specialized verbal functions as opposed to the verbal functions
of the frontal lobes in general bilaterally. However, this conclusion appears to be
premature. Buckner et al (1996) and Nyberg et al (1996) both discuss clear dissociations
between the frontal lobes in verbal memory tasks. The results of the current experiment,
added to the results of a previously reported experiment, demonstrate an additional
dissociation between the verbal functions of the right and left frontal lobes. Based on the
current results, the right, but not the left hemisphere appears to support comprehension
of lexical semantic ambiguities. On the other hand, the left, but not the right hemisphere
is activated during the comprehension of syntactic ambiguities (Stowe et al, 1998). This
suggests that the frontal lobes do make quite different contributions to sentence
comprehension.
Function of the Right Inferior Frontal Lobe in Lexical Semantic Ambiguity
Resolution: Revising an Implausible Interpretation
Although there are good reasons to assume that the right frontal lobe is involved
in certain sorts of memory-related operations, the pattern of results obtained in this
experiment suggests that the right frontal lobe also plays a more active role in
comprehension of semantically ambiguous sentences, rather than simply maintaining the
secondary meaning of the ambiguous lexical item. If the right hemisphere simply
11
Stowe et al
maintains the alternative meaning, the left hemisphere should be implicated in revising
the meaning of the sentence as a whole. This is not supported by the results of the current
experiment. There is no significant difference between ambiguous and control conditions
in the left hemisphere, as can be seen in Figure 1.
Null effects always provide a weak argument against the existence of an effect.
Furthermore voxels were only included in Figure 1 if they exceeded a criterion of P <
.001; this is a reasonable cutoff since even most of these voxels were not significant after
correction for multiple comparisons. Examining the comparisons with rest suggests that
there is somewhat more left frontal activation for the ambiguous sentences (Figure 2)
than for the control sentences (Figure 3). Revising  semantic garden path sentences like
those tested in this experiment appears intuitively to be quite simple; the sentences we
used basically only required the substitution of the meaning of a single word in the
revised meaning. It is therefore possible that the revision of the sentence meaning is so
easily accomplished that it causes very little change in blood flow in the left hemisphere.
Syntactically simple sentences do not activate the left frontal lobe very much (Stowe et
al, 1998). The addition of a little extra effort for revising the sentence meaning might be
just enough to become apparent in the comparison with rest, but not enough to become
significant in the direct comparison with unambiguous sentences. On the other hand, the
processing effort was great enough to cause a highly significant right frontal lobe
activation, although we would not have expected the maintenance of a single word to
cause much difficulty.
We examined the possibility that there was a sub-threshold change in activation
in the left hemisphere in greater detail by lowering the cut-off criterion for this
comparison to uncorrected P < 0.05 (see Figure 4). A few isolated voxels in the left
hemisphere showed up under this criterion, but no biologically plausible area of
activation. The right frontal lobe activation became more extensive, including part of the
anterior temporal lobe (cf. Zaidel et al, 1995). The weak activation in the right superior
temporal and parietal lobe (right hemisphere homologue to Wernicke s area) appears
clearly under this criterion. The best interpretation of this data seems to be that resolution
of lexical semantic ambiguity in these sentences is due to right hemisphere mechanisms
only. However, the issue deserves further investigation; more extensive semantic
revisions may prove to activate the left hemisphere as well as the right.
________________________________________________________________
Figure 4 about here.
________________________________________________________________
The hypothesis that revision as well as maintenance of the second meaning
occurs in the right hemisphere is compatible with evidence that right brain damaged
patients in general have difficulties when the meaning of a sentence or text has to be
changed. Schneiderman and Saddy (1988) asked right brain damaged patients to add
words to a sentence (e.g. add white to the sentence The girl looked out the window at the
snow). In sentences like the example just given, where the syntacticstructure had to be
revised and the semantic interpretation expanded, this was possible for these patients.
However, it proved to be difficult when the additional word changed the meaning as well
as the structure of part of a the original sentence. For example, daughter can only be
added to the sentence She saw her leave the house after the word her which changes the
semantic role as well as the syntactic structure assigned to her; such sentences were
12
The Right Hemisphere and Semantic Ambiguity
difficult to revise. This result suggests that the right hemisphere is involved in revising
the sentence s meaning, although not its structure, in this task.
We have already noted that right brain damage frequently leads to deficits in
revising inferences as well. Brownell et al (1986) tested recall of inferences (e.g. Susan
took a pen and paper with her to talk to the star; inference: Susan wants an autograph).
Patients tended to remember these inferences as true, even if the initially plausible
inference was cancelled by a subsequent sentence (e.g., Her article was going to feature
the comments of famous people on nuclear energy). When these sentences were
presented in the opposite order, the patients were generally able to avoid the false
inference. Again it seems to be the revision of the meaning that causes trouble.
Taken together, these results suggest that the right hemisphere is involved in
revising sentential and discourse interpretations. The activation in the current
experiment may well reflect the anatomical region which subserves this process. This
conclusion may seem contradictory to evidence from the isolated right hemisphere
(Gazzaniga et al, 1984; Kutas et al, 1988; Zaidel, 1977) and divided visual field studies
(Faust and Chiarello, 1998; Faust and Kravetz, 1998) which suggests that the right
hemisphere is not itself capable of much sentential processing. However, the right
hemisphere may be capable of far more when it acts on input from the left hemisphere.
The effects of right hemisphere brain damage on discourse and text comprehension
suggest that the right hemisphere has access to the meaning of the sentences which make
up the discourse. The left hemisphere certainly appears to be the primary source of the
meaning of the sentence as a whole, but the propositional content of the sentence is
apparently communicated to the right hemisphere for further discourse-related
interpretation. For lexical semantic ambiguity resolution, this sentential representation
together with the lexical semantics of the secondary meaning, which is apparently
available in the right hemisphere, are presumably sufficient to construct a revised
meaning.
We have been assuming that a second meaning is constructed when the first
becomes implausible (revision). That is not necessarily the case. Non-literal
interpretations appear to be accessed and used immediately in the right hemisphere
(Hirst, LeDoux and Stein, 1984); the right hemisphere may also initially construct an
alternative sentence meaning, possibly on the basis of a partial interpretation derived
from the left hemisphere. It is worth noting here that results from divided visual field
studies show no sentential level semantic priming effects in the right hemisphere, which
argues against this possibility (Faust and Chiarello, 1998; Faust and Kravetz, 1998).
However, these experiments were not designed to directly test the question at issue here.
This can be done by comparing ambiguous sentences which must be revised with
sentences which remain plausible and for which therefore no revision is required. If the
secondary meaning is always calculated rather than being triggered by implausibility,
increased activation should be seen even if the secondary meaning never becomes
necessary. Since revision was required in the current experiment, we cannot decide
which of these possibilities is correct on the basis of our results. A future experiment will
investigate this issue.
The hypothesis that the right inferior frontal lobe supports revisions of semantic
interpretations provides a common explanation for an interesting set of data, although it
is clear that further research is necessary. If this hypothesis is correct, the right
hemisphere frontal lobe subserves a cognitive function which supports some aspects of
semantic processing, including at least revision of discourse inferences and lexical
semantic ambiguity. The two views of the function of the right frontal lobe that we have
discussed, memory and revision, are not necessarily contradictory. The semantic
13
Stowe et al
processing function may be dependent on the memory-related function which we
discussed in the previous section, particularly as it appears to primarily provide
revisions. Indeed, it relates quite clearly to the hypothesis that the right frontal lobe
supports goal-oriented strategies for retrieval of relevant information from memory
(Fletcher et al, 1998; Swick and Knight, 1996) as opposed to pure maintenance or
retrieval. Thus the activation found in the experiment provides a means of refining our
understanding of the cognitive function of the right frontal lobe as well as our
understanding of how the brain processes involved in comprehending semantically
ambiguous sentences.
Conclusion
The right frontal lobe subserves verbal memory-related functions; we have
suggested that this cognitive function is used to support some aspects of semantic
processing. In the current experiment, we have shown that the right inferior frontal and
anterior dorsolateral cortex are activated during the processing of sentences in which a
lexical semantic ambiguity occurs. The meaning was eventually resolved to the less
preferred meaning. We suggest that this area of the right frontal lobe may be involved
more broadly in the revision of an initially favored sentence or discourse interpretation
and that it also plays a role in making inferences about the relationship between
sentences. The circumstances under which language comprehension depends on this
cognitive function remain to be determined.
14
The Right Hemisphere and Semantic Ambiguity
References
Andy OJ, Bhatnagar SC. Right-hemispheric language evidence from cortical
stimulation. Brain Lang 1984; 23: 159-166.
Baynes K, Wessinger CM, Fendrich R, Gazzaniga MS. The emergence of the capacity to
name left visual field stimuli in a callosotomy patient: Implications for functional
plasticity. Neuropsychologia 1995; 33: 1225-1242.
Beeman M. Semantic processing in the right hemisphere may contribute to drawing
inferences from discourse. Brain Lang 1993; 44: 80-120.
Bottini G, Corcoran R, Sterzi R, Paulesu E, Schenone P, Scarpa P, Frackowiak RSJ,
Frith CD. The role of the right hemisphere in the interpretation of figurative
aspects of language: A positron emission tomography activation study. Brain
1994; 117: 1241-1253.
Brownell HH, Potter HH, Bihrle AM. (Inference deficits in right brain-damaged
patients. Brain Lang 1986; 27: 310-321.
Brownell HH, Potter HH, Michelow D. Sensitivity to lexical denotation and
connotation in brain-damaged patients: A double dissociation? Brain Lang 1984;
22: 253-265.
Brownell HH, Simpson TL, Bihrle AM, Potter HH, Gardner H. Appreciation of
metaphoric alternative word meanings by left and right brain damaged patients.
Neuropsychologia 1990; 28: 375-83.
Brownell HH, Pincus D, Blum A, Rehak A, Winner E. The effects of right hemisphere
brain damage on patients' use of terms of personal reference. Brain Lang 1997;
57: 60-79.
Buckner RL, Petersen SE, Ojemann JG, Miezin FM, Squire LR, Raichle ME.
Functional anatomical studies of explicit and implicit memory retrieval tasks. J
Neurosci 1995; 15: 12-29.
Buckner RL, Raichle ME, Miezin FM, Peterson SE. Functional anatomic studies of memory
retrieval for auditory words and visual pictures. J Neurosci 1996; 16: 6219-6235.
Burgess C, Simpson GB. Cerebral hemispheric mechanisms in the retrieval of
ambiguous word meanings. Brain Lang 1988; 33: 86-103.
Burgess C, Skodis J. Lexical representation and morpho-syntactic parallelism in the left
hemisphere. Brain Lang 1993; 44: 129-138.
Chiarello C, Church KL. Lexical judgments after right- or left-hemisphere injury.
Neuropsychologia 1986; 24: 624-630.
Deloche G, Seron X, Scius G, Segui J. Right hemisphere language processing: lateral
difference with imageable and nonimageable ambiguous words. Brain Lang
1987; 30: 197-205.
Faust M, Chiarello C. Constraints on sentence priming in the cerebral hemispheres:
Effects of intervening words in sentences and lists. Brain Lang 1998; 63, 219-
236.
Faust M, Kravetz S. Levels of sentence constraint and lexical decision in the two
hemispheres. Brain Lang 1998; 62: 149-162.
Faust ME, Gernsbacher MA. Cerebral mechanisms for suppression of inappropriate
information during sentence comprehension. Brain Lang 1996; 53: 234-259.
Fletcher PC, Shallice T, Frith CD, Frackowiak RSJ, Dolan RJ. The functional roles of
prefrontal cortex in episodic memory. II. Retrieval. Brain 1998; 121: 1249-1256.
Friston KJ, Ashburner J, Frith CD, Poline J-B, Heather JB, Frackowiak RSJ. Spatial
registration and normalisation of images. Hum Brain Map 1995a; 2: 165-189.
15
Stowe et al
Friston KJ, Holmes AP, Worsley KJ, Poline J-B, Frith CD, Frackowiak RSJ. Statistical
parametric mapping and functional imaging: A general linear approach. Hum
Brain Map 1995b; 2: 189-210.
Friston KJ, Worsley KJ, Frackowiak RSJ, Mazziotta JC, Evans AC. Assessing the
significance of focal activations using their spatial extent. Hum Brain Map 1994;
1: 214-220.
Gainotti G, Caltagirone C, Miceli G, Masullo C. Selective semantic-lexical impairment
of language comprehension in right-brain-damaged patients. Brain Lang 1981;
13: 201-211
Gazzaniga MS, Smylie CS, Baynes K. Profiles of right hemisphere language and speech
following brain bisection. Brain Lang 1984; 22: 206-220.
Goulet P, Joanette Y, Sabourin L, Giroux F. Word fluency after a right-hemisphere
lesion. Neuropsychologia 1997; 35: 1565-1570.
Hirst W, LeDoux J, Stein S. Constraints on the processing of indirect speech acts:
evidence from aphasiology. Brain Lang 1984; 23: 26-33.
Joanette Y, Goulet P, Le Dorze G. Impaired word naming in right-brain-damaged right-
handers: Error-types and time-course analyses. Brain Lang 1988; 34: 54-64.
Koivisto M. Time course of semantic activation in the cerebral hemispheres.
Neuropsychologia 1997; 35: 497-504.
Kutas M, Hillyard SA, Gazzaniga MS. Processing of semantic anomaly by right and left
hemispheres of commissurotomy patients: Evidence from event-related brain
potentials. Brain 1988; 111: 553-576.
Lambert A, Voot N. A left visual field bias for semantic encoding of unattended words.
Neuropsychologia 1993; 31: 67-73.
McDonald S. Viewing the brain sideways? Frontal versus right hemisphere explanations
of non-aphasic language disorders. Aphasiology 1993; 7: 535-549.
McDonald S, Wales R. An investigation of the ability to process inferences in language
following right hemisphere brain damage. Brain Lang 1986; 29: 68-80.
Myers PS, Linebaugh CW. Comprehension of idiomatic expressions by right-
hemisphere-damaged adults. In: Brookshire RH, editor, Clinical aphasiology:
Conference proceedings, 1981. Minneapolis: BRK; 1981.
Nyberg L, Cabeze R, Tulving E. PET studies of encoding and retrieval: The HERA
model. Psychonom Bull Rev 1996; 3: 135-148.
Price CJ, Wise RJS, Warburton EA, Moore CJ, Howard D, Patterson K, Frackowiak
RSJ, Friston KJ. 1996. Hearing and saying: The functional neuroanatomy of
auditory word processing. Brain 1996; 119: 919-931.
Roland PE, Zilles K. Structural divisions and functional fields in the human cerebral cortex.
Brain Res Rev 1998; 26: 87-105.
Schneiderman EI, Saddy D. Linguistic deficit resulting form right hemisphere damage.
Brain Lang 1988; 34: 38-53.
Seidenberg MS, Tanenhaus MK, Leiman JM, Bienkowski M. Automatic access of the
meanings of ambiguous words in context: Some limitations of knowledge-based
processing. Cognitive Psychol 1982; 14: 489-5.
Siegal M, Carrington J, Radel M. Theory of mind and pragmatic understanding
following right hemisphere damage. Brain Lang 1996; 53: 40-50.
Simpson GB, Peterson R, Casteel MA, Burgess C. Lexical and sentence context effects
in word recognition. J Exp Psychol: Lear Mem Cognition 1989; 15: 88-97.
Stowe LA, Broere CAJ, Paans AMJ, Wijers AA, Mulder G, Vaalburg W, Zwarts F.
Localizing cognitive components of a complex task: sentence processing and
working memory. NeuroReport 1998; 9: 2995-2999.
16
The Right Hemisphere and Semantic Ambiguity
Stowe LA, Paans AMJ, Wijers AA, Zwarts F, Mulder G, Vaalburg W. Sentence
comprehension and word repetition: A positron emission tomography investi-
gation. In press Psychophysiology.
Swick D, Knight RT. Is prefrontal cortex involved in cued recall? A
neuropsychological test of PET findings. Neuropsychologia 1996; 34: 1019-
1028.
Swinney DA. Lexical access during sentence comprehension: (Re)consideration of
context effects. J Verbal Learning Verbal Behav 1979; 18: 645-659.
Talairach J, Tournoux P. Co-Planar Stereotaxic atlas of the human brain. New York,
NY: Thieme Medical Publishers; 1988.
Tanenhaus MK, Leiman MJ, Seidenberg MS. Evidence for multiple stages in the
processing of ambiguous words in syntactic contexts. J Verbal Learning Verbal
Behav 1979; 18: 427-440.
Tokunaga A, Takase M, Otani K. The glabella-inion line as a baseline for CT scanning
of the brain. Neuroradiol 1977; 14: 67-71.
Van Lanckner DR, Kempler D. Comprehension of familiar phrases by left- but not by
right-hemisphere damaged patients. Brain Lang 1987; 32: 265-277.
Woods RP, Cherry SR, Mazziotta JC. A rapid automated algorithm for accurately
aligning and reslicing positron emission tomography images. J Comput Assist
Tomogr 1992; 16: 620-633.
Zaidel DW. Worlds apart: Pictorial semantics in the left and right cerebral hemispheres.
Current Directions Psycholog Sci 1994; 3: 5-8.
Zaidel DW, Zaidel E, Oxbury DM, Oxbury JM. The interpretation of sentence
ambiguity in patients with unilateral focal brain surgery. Brain Lang 1995; 51:
458-468.
Zaidel E. Auditory vocabulary of the right hemisphere following brain bisection or
hemidecortication. Cortex 1976; 12: 191-211.
Zaidel E. Unilateral auditory language comprehension on the Token test following
cerebral commissurotomy and hemispherectomy. Neuropsychologia 1977; 15:
1-18.
Zaidel E, Peters AM. Phonological encoding and ideographic reading by the
disconnected right hemisphere. Brain Lang 1981; 14: 205-234.
17
Stowe et al
Correspondence should be sent to Dr. Laurie A. Stowe, Dept. of Linguistics, School of
Behavioral and Cognitive Neurosciences, Rijksuniversiteit Groningen, Postbus 716,
9700 AS Groningen, The Netherlands; email [e-mail]
Acknowledgments
This research was supported by a PIONIER grant from the Dutch Organization for
Scientific Research. Thanks to Remi Schmeits, Johan Streurman, Marcia Zwartjes and
the Radiopharmoceutical group at the PET Center for technical support. Thanks to
Michael Tanenhaus, Monique Lamers, Rienk Withaar and Roelien Bastiaanse for
valuable comments on an earlier version of this paper.
18
The Right Hemisphere and Semantic Ambiguity
Figure Legends
Figure 1 The area of significant activation in the right inferior frontal lobe is
shown in yellow and white superimposed over a standard MRI Note that the right
hemisphere is displayed on the left side of the MRI image. The three slices are centred
on the voxel showing maximal activation and show 1) a sagittal plane through the brain
from back to front and above to below (upper left) 38 mm right of the plane between the
hemispheres; 2) a coronal plane from side to side and above to below (upper right) 40
mm anterior to the anterior commissure; and 3) a transverse plane from back to front
and side to side (below left) 4 mm below the anterior posterior commissure line.
Figure 2 Activated voxels (cut-off criterion P < .005) for the comparison of
ambiguous sentences with passive fixation superimposed on a  transparent or glass
brain. Upper left shows the voxels as seen from the side of the head, upper right shows
the voxels as seen from the front and the lower left shows the voxels as seen from above.
The combination allows the identification of the activated area in all three dimensions.
Crucially, this comparison showed activation in the right frontal lobe.
Figure 3 Activated voxels (cut-off criterion uncorrected P < .005) for the
comparison of unambiguous sentences with passive fixation superimposed on a glass
brain. Crucially, this comparison showed no activation in the right frontal lobe;
conversely there was no indication of decreased activation in this area.
Figure 4 Activated voxels (cut-off criterion uncorrected P < .05) for the
comparison of ambiguous sentences with unambiguous sentences superimposed on a
glass brain. Crucially, this comparison shows essentially no activation in the left
hemisphere. At this cut-off, on the other hand, an activation appears in the right
hemisphere  Wernicke s area as well as the frontal lobe.
19
Stowe et al
Figure 2:
20
The Right Hemisphere and Semantic Ambiguity
Figure 3:
21
Stowe et al
Figure 4:
22