Action Matters in Language Processing: a Tms Study

/ MirrorBot
IST-2001-35282
Biomimetic multimodal learning in a mirror neuron-based robot
Title: Functional Interaction Of Language And Action Processing: A TMS Study (Deliverable D1)Authors: Friedemann Pulvermüller , Olaf Hauk, Vadim Nikulin and Risto J. Ilmoniemi
Covering period 1.6.2002-1.6.2003

MirrorBot Report 8

Report Version: 1
Report Preparation Date: 1. June 2002
Classification: Restricted
Contract Start Date: 1st June 2002 Duration: Three Years
Project Co-ordinator: Professor Stefan Wermter
Partners: University of Sunderland, Institut National de Recherche en Informatique et en Automatique at Nancy, Universität Ulm, Medical Research Council at Cambridge, Università degli Studi di Parma

0.ABSTRACT2

1.INTRODUCTION3

2.MATERIAL AND METHODS7

3.RESULTS13

4.DISCUSSION17

ACKNOWLEDGEMENT22

6.REFERENCES23

0. Abstract

Transcranial magnetic stimulation (TMS) was applied to motor areas in both hemispheres while right-handed subjects made lexical decisions on words related to actions. Words referring to leg actions (example: "to kick") were compared with words referring to movements involving the arms and hands ("to pick"). Magnetic stimulation was to hand motor areas and to dorsal sites that included the pre-motor representation of the legs. TMS at hand and leg loci influenced the processing of arm and leg words differentially: Stimulation at the leg loci speeded lexical decisions on leg words specifically, whereas hand locus stimulation produced the opposite effect, faster processing of arm words compared with leg words. A separate analysis of the effects of TMS to the dominant left hemisphere confirmed the significant interaction of the factors stimulation site and word group, whereas right-hemispheric stimulation failed to yield significant effects, suggesting that left-hemisphere networks are more relevant for linking actions and words in the brain of right-handers. The results support a distributed account of word processing according to which information about words and the actions they refer to are stored and processed by distributed neuronal ensembles that include mirror neurons with somatotopic localization.

1. Introduction

Word processing in the context of objects and actions leads to correlated activity in different cortical areas. If a word is frequently used when an action is performed by the learner or by a person perceived by the learner, neurons subserving lexical processing and those involved in processing the action become active simultaneously. This correlated activity leads to strengthening of neuronal connections and thus to a linkage of word form and action representations (Pulvermüller, 1999). The resulting word-related neuron ensembles, word webs, may have distinct topographies, depending on where in the cortex the correlated activity occurred.

The somatotopic organization of the motor and pre-motor cortex implies that actions performed with different body parts relate to different topographic patterns of activation in motor, pre-motor and adjacent prefrontal areas. Somatotopic organization has been demonstrated for the primary motor cortex (Penfield and Rasmussen, 1950) and could be revealed for more rostral frontal areas, in particular for pre-motor areas as well (He et al., 1993; Rizzolatti et al., 2001). In the primary motor cortex, the leg representation is to a large extent hidden in the interhemispheric sulcus, but pre-motor representations of the legs are also present on the lateral surface of the frontal lobe, where they are located superior and medial to the hand representation anterior to the precentral gyrus. In both the precentral gyrus and the pre-motor areas, arm and hand movements are represented lateral and inferior to leg movement representations. If action-related information is woven into the cortical neuron webs representing and processing words, one may predict that words referring to different body parts correspond to networks with different cortical distributions. The action-related neurons of a word referring to a leg movement - such as "to kick" - should be located superior and medial to those of a word related to an arm or hand movement - as, for example, "to pick". The semantic difference between subcategories of action words should thus be laid down in the cortical distribution of word-related neuron webs, a hypothesis with clear implications for neurophysiological brain research on language (Pulvermüller, 2001).

We investigated the processing of action words using Transcranial Magnetic Stimulation (TMS), a well-established tool for scrutinizing the brain basis of cognition and language. In recent years, TMS pulses delivered to different parts of the left perisylvian language cortex have been shown to interfere with language processing (Epstein, 1998; Bailey et al., 2001). Single-pulse or repetitive TMS (rTMS) of inferior frontal sites can lead to speech arrest and specific temporary articulatory deficits (Pascual-Leone et al., 1991; Michelucci et al., 1994; Epstein et al., 1996; Stewart et al., 2001) and stimulation of superior temporal areas to a facilitation of naming (Mottaghy et al., 1999; Sparing et al., 2001). Language laterality has a clear reflection in the influence of TMS pulses to left and right perisylvian areas; only in subjects with strong language laterality - either to the left or to the right hemisphere - unilateral TMS causes a marked dysfunction of language processing (Knecht et al., 2002). From a linguistic perspective, it is of particular interest that rTMS of the frontal cortex led to deficits in the processing of one specific lexical category, namely verbs (Shapiro et al., 2001). This latter result suggests that language functions, such as the processing of specific lexical categories, can be monitored using TMS. Further, listening to spoken words that include phonemes strongly involving one particular articulator – e.g. [r], which strongly involves the tongue – had a facilitatory effect on the TMS-induced muscle movement of that relevant articulator – in this case the tongue (Fadiga et al., 2002). Together, these results document the importance of TMS as a research tool for exploring very specific brain processes of language. It is, however, also noteworthy that TMS, as any other method in brain science, has its limitations, most of which are due to the fact that the biophysical mechanisms elicited by TMS pulses are not fully understood (Bailey et al., 2001; Ruohonen and Ilmoniemi, 2002). One open question is why TMS sometimes leads to a disruption of processes supported by the stimulated area, whereas, in other cases, facilitatory effects were observed (Maeda et al., 2000). The direction of the effect TMS takes is difficult to predict. More specifically, language functions have been found to be impaired (Shapiro et al., 2001; Knecht et al., 2002) or improved (Mottaghy et al., 1999; Sparing et al., 2001) as a result of magnetic stimulation (Epstein, 1998; Epstein et al., 1999). It appears that stimulation amplitude, duration, frequency and the topography of the magnetic field applied crucially influence the mode (excitation or inhibition) and the magnitude of the effect of TMS (Pascual-Leone et al., 1999; Maeda et al., 2000; Bailey et al., 2001; Stewart et al., 2001; Rijntjes and Weiller, 2002; Ruohonen and Ilmoniemi, 2002).

In this study, we ask whether stimulation of different motor areas has a specific influence on the processing of arm- and leg-related action words in a lexical decision task. If word representations in the perisylvian cortex have specific links to action representations in motor, pre-motor, and possibly prefrontal areas controlling arm and leg movements, stimulation of arm and leg areas should differentially influence the processing of arm and leg words. Generally, stimulation of the hand and leg motor areas should have opposite effects on the processing of hand- and leg-related words. By choosing single-pulse TMS with moderate stimulation intensity, ~90% of the individual motor threshold, we aimed at generating a physiological activation that could facilitate activation in the neuronal networks targeted.

2. Material and Methods

Subjects: 12 subjects (mean age: 26 years, S.D. = 4.8; 5 females) participated in the study. All were native speakers of English and all except for one were monolinguals. Results on the Oldfield Handedness Inventory (Oldfield, 1971) indicated that all subjects were right-handed (mean laterality quotient = 84, S.D. 25). Furthermore, all participants had normal or corrected-to-normal vision and reported no history of neurological illness or drug abuse. Informed consent was obtained from all subjects and they were paid for their participation. The study was approved by the Ethics Committee of the Helsinki University Central Hospital.

Stimuli: The stimuli were 250 letter strings, including 50 arm- and 50 leg-related English words, 50 English distractor words and 100 pseudowords that were in accordance with the phonotactic and orthographic rules of English. The action word groups were matched for word-length (average lengths were 4.5 letters for arm words and 4.6 for leg words) and standardized lexical frequency according to the CELEX database (average word frequency for arm words was 259 and for leg words 263 occurrences per million). A rating study was performed before the experiment. Subjects were asked to rate each word as to whether it reminded them of actions they can perform with their hands/arms or legs, respectively. Ratings were obtained on a seven-point scale. The results shown in Figure 1 demonstrate that there was a clear double dissociation between the two subcategories of action words. Words from one group were consistently reported to elicit strongest arm/hand motor associations, with weak leg associations, whereas the other group was characterized by strong leg (but weak arm) associations. The interaction of Word group and Association type (arm/leg) was highly significant (F (1,49) = 839.00, p < 0.001). This ascertained that arm-related words elicited stronger associations of arm actions than leg actions, whereas the opposite pattern was seen for leg words (F > 100; p < 0.001). Further, ratings of familiarity and imageability of the action words did not
reveal differences between the stimulus groups.

Figure 1: Association ratings of word stimuli used in this study. Words from two groups were judged as to whether they reminded subjects of arm/hand actions (red bars) and of leg/foot actions (blue bars), respectively.

TMS stimulation: The study was performed with the TMS device at the BioMag Laboratory, Helsinki University Central Hospital. The magnetic pulses were biphasic with a duration of 320 µs. A figure-of-eight coil was used in order to produce focal cortical stimulation (Bailey et al., 2001). It consisted of two coplanar circular wings (40-mm diameter) with 15 turns of copper wire in each. The coil was oriented so that the induced current flow was in posterior-anterior direction during the rising phase of the pulse.

The following procedure was applied for determining stimulation sites and intensities. The subjects were seated in a comfortable chair below the montage holding the TMS coil. The motor threshold was found separately for each hemisphere. TMS pulses were given every 2 s. EMG responses were recorded through 2 electrodes placed next to the first dorsal interosseous (FDI) muscle of the hand contralateral to the stimulation site. The EMG responses were continuously monitored on a computer screen. Initially, the coil was placed at the C3 or C4 sites of the International 10-20 system of placement of EEG electrodes (Jasper, 1958), which are known to be close to the motor representation of the hand (Lagerlund et al., 1994). An area around C3/C4 was monitored in steps of 5 mm to find the largest elicited muscle response of the FDI. These points were used later as the hand stimulation loci. The motor threshold was determined as the smallest intensity eliciting EMGs with amplitudes of at least 50 µV in 5 out of 10 successive trials.

Ethical guidelines recommended a maximum stimulation intensity for our device and research purpose. Within the range of intensities permitted, it was not possible to determine the leg-areas for each individual subject, i.e., no reliable EMG responses could be measured at a lower limb muscle (M. tibialis anterior). Previous studies also found leg-motor areas to require higher stimulation intensity than the hand motor area (e.g., Chen and Garg, 2000). For targeting the leg areas, we chose that point on the line connecting the hand stimulation locus and the vertex that was one third of this distance away from the vertex.

At the end of the experiment, control trials were added, in which a 2-cm-thick plastic block was placed between the TMS coil and the scalp, so that no relevant TMS of cortical tissue could occur. This condition served to mimic the auditory and somatosensory sensations produced by the mechanical vibration of the coil (Nikouline et al., 1999). Separate control runs were performed for the left and right hemispheres, respectively. The control experiment was run to investigate possible processing differences between word groups in the absence of TMS.

There were two competing requirements for choosing the optimal stimulation intensity. On the one hand, the intensity of the pulse-related click, which varies with TMS intensity, should be kept equal for the stimulation sites, to avoid differences in inter-sensory interference. On the other hand, different motor thresholds for different hemispheres are likely to reflect different geometry of the skull and the brain, or of the excitability of neuronal populations. We decided to find a compromise between these requirements by applying the following procedure: The mean of the left- and right- hemispheric motor thresholds were computed for each subject. The mean between this value and the actual motor threshold was then calculated for each hemisphere, and 90% of this intensity was chosen for stimulation at hand- and leg-loci. The motor threshold for the right hemisphere was on average 9% above that for the left hemisphere.

Stimulation procedure: During the entire experiment, a fixation cross appeared in the middle of the screen, which was occasionally replaced by linguistic stimuli. Word and pseudoword stimuli were tachistoscopically presented, for 100 ms, in pseudo-random order with a stimulus onset asynchrony of 3 s. TMS pulses were applied 150 ms after the onset of each linguistic stimulus. Letter strings were presented in white capital letters on a gray background in the middle of a computer screen at a distance of 2 m from the subjects' eyes. Letter strings subtended a horizontal visual angle smaller than 4 degrees. Subjects were instructed to fixate their eyes on the fixation cross and to respond as quickly and as accurately as possible by a lip movement to words only. The usual response was that subjects articulated the syllable "ba" or performed the corresponding lip movement. A mouth response was chosen to prevent interference of the preparation of the behavioral response with any putative word-elicited activation of hand or leg motor or pre-motor areas.

For each stimulation site, two successive runs, each lasting for about 3 minutes, were performed. The stimulus set was presented 3 times, once when stimulation was to the left-hemispheric loci, once when the right hemisphere was stimulated, and once in the control trials (see below). For stimulation of each hemisphere, the set of stimuli was halved and 25 items from each word category were presented together with 50 pseudowords when TMS was delivered to the arm and leg loci, respectively. The order of the four blocks - where stimulation was to the left hand, left leg, right hand, and right leg areas, respectively - was counterbalanced over subjects. In the control trial performed at the end, 25 stimuli of each word group and 50 pseudowords were presented while the coil was placed over each hemisphere.

Recording of behavioral responses: As mentioned, subjects were instructed to respond by brisk mouth movements. These movements were measured using two EMG electrodes. EMG electrodes were fixed on the left below the lower lip and above the upper lip. The ground electrode was placed on the right cheek. The signal from these electrodes was recorded using bipolar channels of an EEG amplifier. Its input was blocked for 10 ms during and after the pulse by a sample-and-hold technique, thus avoiding the recording channels to be saturated by the relatively high voltages induced by the TMS pulse (Virtanen et al., 1999). The sampling rate was 1450 Hz and the analog passband 0-500 Hz.

Data analysis: An EMG response was determined in the following way. For each participant, the average EMG amplitude was calculated. Since maximal EMG amplitudes varied over individuals, the criterion for overt behavioral responses in single trials was defined as the first point in time exceeding 50% of the maximum average value. The accuracy and average response times of these responses were calculated for each word group and subject, which were then entered into three-way Analysis of Variance (ANOVAs) with the factors Word Group, Stimulation Site, and Hemisphere. Additional ANOVAs were performed to separately investigate effects obtained for stimulation of the left and right hemisphere. One-tailed t-tests were used for Planned Comparison Tests.

3. Results