Brain Signatures of Meaning Access in Action Word Recognition Friedemann Pulvermu ¨ller 1 , Yury Shtyrov 1 , and Risto Ilmoniemi 2 Abstract & The brain basis of action words may be neuron ensembles binding language- and action-related information that are dispersed over both language- and action-related cortical areas. This predicts fast spreading of neuronal activity from language areas to specific sensorimotor areas when action words seman- tically related to different parts of the body are being perceived. To test this, fast neurophysiological imaging was applied to reveal spatiotemporal activity patterns elicited by words with different action-related meaning. Spoken words referring to actions involving the face or leg were presented while subjects engaged in a distraction task and their brain activity was re- corded using high-density magnetoencephalography. Shortly after the words could be recognized as unique lexical items, objective source localization using minimum norm current estimates revealed activation in superior temporal (130 msec) and inferior frontocentral areas (142–146 msec). Face-word stimuli activated inferior frontocentral areas more strongly than leg words, whereas the reverse was found at superior central sites (170 msec), thus reflecting the cortical somatoto- py of motor actions signified by the words. Significant corre- lations were found between local source strengths in the frontocentral cortex calculated for all participants and their semantic ratings of the stimulus words, thus further establish- ing a close relationship between word meaning access and neurophysiology. These results show that meaning access in action word recognition is an early automatic process re- flected by spatiotemporal signatures of word-evoked activity. Word-related distributed neuronal assemblies with specific cor- tical topographies can explain the observed spatiotemporal dynamics reflecting word meaning access. & INTRODUCTION Neuroscientific principles apply to all brain functions, language included. The Hebbian principle of correlation learning implies that neurons that frequently fire togeth- er become more strongly connected with each other and therefore wire together. The principle of topograph- ical projections holds true in the central nervous system as well, implying, for instance, that leg, arm, and face representations are aligned in the motor cortex. If words refer to actions performed with the leg, arm, or face, the neurons processing the word form and those processing the referent action should frequently fire together and thus wire together, therefore resulting in word-related networks reflecting the somatotopy of the sensorimotor cortex (Figure 1). This position implies that aspects of the semantic access to the meaning of action words, their reference to different parts of the body, are re- flected in the brain response, whenever these words are being perceived, by fast and topographically specific activation of areas in frontocentral cortex. Neuroimaging studies have shown that word mean- ing is processed in dedicated cortical areas, including the temporal (Scott & Johnsrude, 2003; Price, 2000) and frontocentral areas (Martin & Chao, 2001; Pulvermu ¨ller, 2001; Martin, Wiggs, Ungerleider, & Haxby, 1996; Pul- vermu ¨ller, Preissl, Lutzenberger, & Birbaumer, 1996). Recently, functional magnetic resonance imaging (fMRI) showed meaning-related activation in the frontocentral motor system during word reading. Words that refer to actions performed with the face, arms, or legs, such as lick, pick, and kick, elicited somatotopic activation of the motor and premotor cortex that specifically reflected the body parts contributing to word-related actions (Figure 1; de Lafuente & Romo, 2004; Hauk, Johnsrude, & Pulvermu ¨ller, 2004). Because of the low temporal resolution of fMRI, however, it is unclear at which precise point in time meaning-related cortical activation first occurs and, hence, whether this fMRI activation reflects immediate word comprehension and meaning access or later secondary thought processes provoked by the words instead. Here, we used magneto- encephalography (MEG) and objective source localiza- tion (minimum norm current estimates [MNCEs]) to determine the locus of meaning access in spoken ac- tion word processing in both space and time. We hypoth- esized that meaning access is mediated by associative networks connecting word form (in perisylvian language areas) and semantic representation (in frontocentral 1 MRC Cognition and Brain Sciences Unit, Cambridge, UK, 2 Helsinki University Central Hospital and Nexstim Inc., Helsinki, Finland D 2005 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 17:6, pp. 884–892