Words in the brain's language. -
1. Words in the brain: Where? Why? How? Human language production is caused by neuronal activity and any speech signal necessarily activates neurons in the brains of listeners when being perceived. It is the purpose of language science to specify these processes and their un- derlying mechanisms. However, owing to the enormous complexity of language and the sparsity of our knowledge about brain functioning, neuroscientists, psychologists, and linguists have not attacked this goal directly. Indeed, bio- logical knowledge currently available is still far from mak- ing it possible to spell out the great variety of language phe- nomena in terms of neurons. Nevertheless, it is possible to choose paradigmatic questions about language and to try to find answers for them based on biological principles. I will use this strategy here to approach the problem of language and the brain. The issue I would like to address is that of different vo- cabulary classes. At school, one learns to categorize words into fifty or so lexical categories, such as noun or verb, and one may also be asked to categorize words on the basis of their meaning, according to semantic criteria. Of course it is useful, for didactic purposes, to make a large number of distinctions between classes of words, not only based on their meaning and their function in syntactic structures, but also based on criteria such as their intonation, syllable com- plexity, number of letters or speech sounds, or the fre- quency with which they are used in ordinary language. However, one may wonder whether some of these distinc- tions reflect differences that are biologically real. This would mean that the members of word classes A and B, which can be distinguished on the basis of linguistic or di- dactic criteria, would also be represented differently in the human brain. In psycholinguistics, much effort has been ex- pended to demonstrate processing differences between word classes, for example between the major lexical classes BEHAVIORAL AND BRAIN SCIENCES (1999) 22, 253���336 Printed in the United States of America �� 1999 Cambridge University Press 0140-525X/XX $12.50 253 Words in the brain���s language Friedemann Pulverm��ller Department of Psychology, University of Konstanz, 78434 Konstanz, Germany email@example.com www.clinical-psychology.uni-konstanz.de Abstract: If the cortex is an associative memory, strongly connected cell assemblies will form when neurons in different cortical areas are frequently active at the same time. The cortical distributions of these assemblies must be a consequence of where in the cortex cor- related neuronal activity occurred during learning. An assembly can be considered a functional unit exhibiting activity states such as full activation (���ignition���) after appropriate sensory stimulation (possibly related to perception) and continuous reverberation of excitation within the assembly (a putative memory process). This has implications for cortical topographies and activity dynamics of cell assemblies forming during language acquisition, in particular for those representing words. Cortical topographies of assemblies should be related to aspects of the meaning of the words they represent, and physiological signs of cell assembly ignition should be followed by possible indicators of reverberation. The following postulates are discussed in detail: (1) assemblies representing phonological word forms are strongly lateralized and distributed over perisylvian cortices (2) assemblies representing highly abstract words such as grammatical func- tion words are also strongly lateralized and restricted to these perisylvian regions (3) assemblies representing concrete content words include additional neurons in both hemispheres (4) assemblies representing words referring to visual stimuli include neurons in visual cortices and (5) assemblies representing words referring to actions include neurons in motor cortices. Two main sources of evidence are used to evaluate these proposals: (a) imaging studies focusing on localizing word processing in the brain, based on stimulus-triggered event-related potentials (ERPs), positron emission tomography (PET), and functional magnetic resonance imaging (fMRI), and (b) stud- ies of the temporal dynamics of fast activity changes in the brain, as revealed by high-frequency responses recorded in the electroen- cephalogram (EEG) and magnetoencephalogram (MEG). These data provide evidence for processing differences between words and matched meaningless pseudowords, and between word classes, such as concrete content and abstract function words, and words evok- ing visual or motor associations. There is evidence for early word class-specific spreading of neuronal activity and for equally specific high-frequency responses occurring later. These results support a neurobiological model of language in the Hebbian tradition. Com- peting large-scale neuronal theories of language are discussed in light of the data summarized. Neurobiological perspectives on the prob- lem of serial order of words in syntactic strings are considered in closing. Keywords: associative learning cell assembly cognition cortex ERP EEG fMRI language lexicon MEG PET word category Friedemann Pulverm��ller has an M.A. in Biology, a Ph.D. in Lin- guistics, a Habilitation in Psychology and Medicine, and is now Privat- dozent of Psychology at the Univer- sity of Konstanz, Germany. He is the author of over 80 publications in the area of cognitive neuroscience, in- cluding his recent book on neurobiology of language and the forthcoming ���Neuronal Grammar.��� Among his honors are an early career award from the Society for Psychophysiological Research and a Helmholtz and a Heisenberg Fellowship. His main scientific interest is to spell out language mechanisms in terms of neurons.
called content words (or open-class words, including nouns, verbs, and adjectives) and function words (or closed-class words, including articles, pronouns, auxiliary verbs, conjunc- tions, and so on. Some of these studies will be discussed in sect. 5.). It is good to know that two word groups are dif- ferent however, it is better to know (or to have an idea about) what the actual differences are. A biological ap- proach aims at specifying the difference in terms of neurons and neuronal connections. In recent years, more and more neuropsychological stud- ies have been devoted to the investigation of cortical mech- anisms necessary for word processing, and psychophysio- logical studies have been investigating the brain areas that ���light up��� when words are being produced or compre- hended. Such studies are most welcome because they may contribute to an answer of the ���where��� question, that is, the question of where representations are housed and pro- cesses take place. However, even when questions such as ���Which word classes will be selectively impaired after focal brain lesion in cortical area X?��� or ���Which brain areas will become active when words of class A are being produced or comprehended?��� have been definitely answered, the question of why this is so may still be open. Why are words of class A processed in area X? An explanation of language mechanisms in the brain is only possible if such ���why��� ques- tions can be answered from known biological principles. But even definite and exhaustive answers to ���where��� and ���why��� questions may still not be a satisfactory end point of cognitive neuroscientific research: If it is clear where in the brain particular language units are represented and processed, and if it is clear why this is so, one can still ask how language representations are laid down, and how they are activated when language units are being processed. This target article will certainly not provide complete an- swers to ���where,��� ���why,��� and ���how��� questions related to lan- guage. It will provide preliminary answers to the ���where��� question as far as words of certain classes are concerned it hopes to convince the reader that the ���why��� question can be answered in a few clear cases and it tries to specify some very basic features of cortical representations and the way they become active and maintain their activity. All this is done on the basis of a brain model rooted in Hebb���s con- cept of cell assemblies. In fact, the purpose of this article is not only to discuss the issue of words in the brain, but to make it evident that the Hebbian approach is a powerful tool for cognitive neuroscience that may lead to a biological explanation of our language capacity and of other higher cognitive capacities as well. 2. The Hebbian model, recent modifications and evidence In the late 1940s, Donald Hebb (1949) proposed a neu- ropsychological theory of cortical functioning that can be considered an alternative to both localizationist and holistic approaches. Localizationists would assume that small corti- cal areas are fully capable of performing complex cognitive operations. A localizationist would, for example, propose that an area of a few square centimeters of cortical surface is the locus of word comprehension (Broca 1861 Lichtheim 1885 Wernicke 1874). According to this view, the psycho- logical process (word comprehension) is restricted to one area ��� that is, no other areas are assumed to contribute to this specific process. Only under pathological conditions or during development may there be a shift of the process to another equally narrow area (Luria 1970 1973). In contrast, a holistic approach would imply that the entire cortex ex- hibits equipotentiality with regard to all cognitive operations and that all cortical areas (or even brain parts) can contribute to sufficiently complex processes, such as those involved in language (for discussion, see Freud 1891, Lashley 1950, and, for an overview, Deacon 1989). The Hebbian proposal is in sharp contrast to both of these views. Cell assemblies with defined cortical topogra- phies are assumed to form the neurobiological representa- tions of cognitive elements such as gestalt-like figures or words. This position is radically different from a localiza- tionist approach, because it assumes that neurons in differ- ent cortical areas may be part of the same distributed func- tional unit. The Hebbian viewpoint is also different from the holistic view that ���everything is equally distributed,��� be- cause it implies that the representation of, for example, an image may involve cortical areas entirely different from those contributing to the representation of, say, an odor. Accordingly, the representation of a word would not be re- stricted to a small cortical locus, but would be distributed over well-defined areas, for example over Broca���s, Wer- nicke���s, and some other areas. The Hebbian model is based on three fundamental as- sumptions about cortical functioning, which can be sum- marized as follows: 1. Coactivated neurons become associated. 2. Associations can occur between adjacent or distant neurons that is, the entire cortex is an associative memory. 3. If neurons become associated, they will develop into a functional unit, a cell assembly. Hebb was frequently criticized, because his assumptions were considered too speculative and because some of his colleagues believed that his ideas would not be testable. Therefore, it is necessary to discuss his assumptions in light of evidence presently available. Electrophysiological studies have demonstrated that having cortical neurons frequently active at the same time strengthens their connections. If a neuron, call it L, sends one connection to a second neuron, M, their synapse will strengthen when both are repeatedly active together, so that L will later have a stronger influence on M. Because this effect may last for many hours or days, or even longer, it has been termed long-term potentiation (LTP) (Ahissar et al. 1992 Gustafsson et al. 1987). After this kind of associa- tive learning, connection strength will be a function of the frequency of coincident activity. Table 1 describes this kind of coincidence learning (Palm 1982). One may object to this and similar learning rules that co- incidence learning is only one form of associative learning known to take place between neocortical neurons. If only one of the two neurons is active while the other one remains silent, this could also have an effect on the strength of their connection. In fact, it was shown by electrophysiological ex- periments that activation of presynaptic neuron L alone, while the membrane potential of postsynaptic neuron M is stable (or only slightly depolarizes), leads to a weakening of their synaptic connection (Artola et al. 1990 Artola & Singer 1987 1993 Rauschecker & Singer 1979). Because this re- duction (or depression) of the influence of one neuron on the other is long-lasting, the phenomenon has been called long-term depression (LTD). There is also evidence for Pulverm��ller: Brain���s language 254 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2
LTD occurring when presynaptic neurons are silent while postsynaptic neurons fire frequently (Tsumoto 1992 Tsumoto & Suda 1979). Therefore, the original idea pro- posed by Hebb needs a slight but important modification: Connection strength is not only modified by coincident ac- tivity, it also changes if only one of two connected neurons is active while the other one is inactive. Table 2 describes this kind of learning, which will be called correlation learn- ing, because after this kind of synaptic modification, the strength of the synaptic connection will include information not only about the frequency of coincident firing of neu- rons, but also about how strong the correlation was between their activations. This formulation is very general. It does not make distinc- tions implied by more precise formulations of synaptic learn- ing rules (Artola & Singer 1993 Bienenstock et al. 1982 Tsumoto 1992), in which, for example, the states called ���ac- tive��� and ���inactive��� above, have been replaced by gradual ac- tivity levels (quantified in terms of the frequency of action potentials or the membrane potential of the postsynaptic neuron). In addition, the above formulations leave open the questions of how the w-values should actually be chosen. Whereas w1 may be assumed to be larger than w2 and w3, the exact values of the variables are unknown. These questions will not be addressed here, because they have been discussed in great detail based on what is known about synaptic dy- namics in the neocortex (Tsumoto 1992) and in light of stor- age properties of artificial associative networks (Palm 1982 Palm & Sommer 1995 Willshaw & Dayan 1990). In the pre- sent context, it is most important to keep in mind that a cor- relation rule, rather than a coincidence rule, is a fundamen- tal principle of synaptic learning in the cortex. It appears uncontroversial that excitatory cortical neu- rons located close to each other are likely to have a synap- tic contact. Although this is not a 100% probability ��� it is actually far below (Braitenberg 1978a Braitenberg & Sch��z 1991) ��� it is evident that adjacent neurons are much more likely to be connected than neurons located far apart, that is, in distant cortical areas (Young et al. 1995). It is clear from neuroanatomical studies, however, that most cortical pyramidal cells have long axons reaching distant areas or subcortical structures, and that connections from one area project to several other areas. In the Macaca, for example, what may be considered the homologues of Broca���s and Wernicke���s areas are not only intensely connected to each other they also exhibit connections to additional premotor, higher visual, and association cortices (Deacon 1992a 1992b Pandya & Vignolo 1971 Pandya & Yeterian 1985). Therefore, if correlated neuronal activity is present in a large number of neurons in different cortical areas, some of these neurons will exhibit direct connections to each other. These neurons will become more strongly associated even if they are located far apart. Thus, although the cortex is not a fully connected associative memory in which every pro- cessing unit is connected to every other one, it still appears to be an associative network well suited to allow for both lo- cal and between-area associative learning (Braitenberg & Sch��z 1991 1998 Fuster 1994 Palm 1982). If neurons in an associative network exhibit correlated activity, they will be a stronger influence on each other. This implies that these neurons will be more likely to act to- gether as a group. Hebb (1949) calls such anatomically and functionally connected neuron groups ���cell assemblies.��� The strong within-assembly connections are likely to have two important functional consequences: (1) If a sufficiently large number of the assembly neurons are stimulated by external input (either through sensory fibers or through cortico-cortical fibers), activity will spread to additional as- sembly members and, finally, the entire assembly will be ac- tive. This explosion-like process has been called ignition of the assembly (Braitenberg 1978b). (2) After an assembly has ignited, activity will not stop immediately (because of fatigue or regulation processes), but the strong connections within the assembly will allow activity for some time. Cell assemblies are sometimes conceptualized as packs of neu- rons without an ordered inner structure. However, accord- ing to Hebb���s (1949) proposal, assembly neurons are con- nected so that ordered spreading and reverberation of neuronal activity can occur. The latter point needs further elaboration: Figure 1 is taken from Hebb���s 1949 book and depicts what the author believed to be a possible inner structure of an assembly. In this diagram, arrows represent subgroups of neurons in- cluded in the assembly. These subgroups would each be- come active at exactly the same point in time. Arrowheads indicate the other subgroups to which a given subgroup would project, and numbers denote a possible activity se- quence. After synchronous activity of the neurons repre- sented by the arrow labeled ���1,��� a wave of excitation will run through the assembly as indicated by the numbers, and ac- tivity will finally cease. Thus, it is evident that in Hebb���s early proposal, a cell assembly was already conceptualized as a highly structured entity. Whereas ignition of the as- sembly may simultaneously involve all assembly neurons, it is also possible to have a wave of excitation circulating and reverberating in the many loops of the assembly. The wave can be described as a spatiotemporal pattern of activity in which many cortical neurons participate. Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 255 Table 1. Associative synaptic learning according to a Hebbian coincidence rule neuron L active inactive active 1w* -- neuron M inactive -- -- *1w indicates an increase in connection strength between neu- rons L and M hyphens indicate no change in connection strength. Table 2. Associative synaptic learning according to a correlation rule neuron L active inactive active 1w1* 2w2 neuron M inactive 2w3 -- *1w1, 2w2, and 2w3 indicate positive or negative changes in connection strength.
The question of whether cell assemblies that represent stimuli and cognitive entities exist in the cortex has long been thought impossible to test by empirical research. As mentioned earlier, this belief was probably one of the main reasons why Hebb���s theory was not generally accepted in the 1940s and 1950s. However, more recent experimental work has provided strong evidence for the Hebbian ideas. Neurophysiological work by Abeles, Aertsen, Gerstein, and their colleagues (Abeles 1982 1991 Abeles et al. 1993 1994 Aertsen et al. 1989 Gerstein et al. 1989) revealed ex- actly timed spatiotemporal firing patterns in cortical neu- rons. The specific neuronal connections these patterns are probably related to were labeled synfire chains by Abeles, because a subpopulation of neurons must synchronously activate the next subpopulation to keep the chain going. It is important to note that spatiotemporal activity patterns ac- tually detected in cortical neurons frequently involve the repeated activation of a given neuron, thus suggesting re- verberations caused by loops in the chain (Abeles et al. 1993). Evidently, the concept of a reverberating synfire chain emerging from recent neurophysiological data comes very close to Hebb���s original proposal summarized in Fig- ure 1. In contrast to the original proposal, it appears more realistic to postulate connections not only between consec- utive subpopulations of neurons, but also connections that skip subgroups and directly link, for example, subgroups 1 and 3 in the example illustration (Fig. 1). Such bypass con- nections may be realized by relatively slowly-conducting cortico-cortical fibers (Miller 1996). Furthermore, Abeles���s findings suggest that the neuron subgroups represented by arrows in Hebb���s diagram overlap, so that a given neuron can be part of, say, subgroups 1 and 7. In summary, after its full activation (ignition), neuronal activity may reverberate in the loops of an assembly. Igni- tion and reverberation may represent important functional states of Hebbian cell assemblies. On the cognitive level, ig- nition may correspond to perception of a meaningful stim- ulus and to activation of its representation. The fact that an object partially hidden behind another one can frequently be identified can be explained by full ignition of a cell as- sembly after stimulation of only some of its neurons (Hebb 1949). Sustained activity of the assembly and reverberation of activity therein may represent an elementary process un- derlying short-term or active memory (Fuster 1989 1995 Fuster & Jervey 1981). The latter view arises from studies that evidence a systematic relationship between the occur- rence of defined spatio-temporal activity patterns in cortex and particular engrams an experimental animal has to keep in active memory (Fuster 1995 Villa & Fuster 1992). Recent neurophysiological work not only revealed well- timed spatiotemporal activity patterns in cortical neurons related to memory processes but another line of research uncovered stimulus-specific synchronization of activity in cortical neurons related to perceptual processes. If an ele- mentary visual stimulus, for example a bar moving in a par- ticular direction, is presented to an experimental animal, numerous neurons in various visual cortices in both hemi- spheres start to synchronize their firing and, in many cases, exhibit coherent rhythmic activity in a relatively high fre- quency range, that is, above 20 Hz (Eckhorn et al. 1988 En- gel et al. 1990 1991b Gray et al. 1989 Kreiter & Singer 1992).1 This provides further evidence that neurons in dif- ferent areas are strongly coupled and can act as a unit. Al- though synchronization phenomena have been observed in subcortical structures and even in the retina (Kirschfeld 1996 Neuenschwander & Singer 1996 Sillito et al. 1994 Steriade et al. 1993), cortico-cortical connections are ap- parently necessary for synchronization of neuron responses in cortex (Engel et al. 1991a Gray et al. 1989 Singer & Gray 1995). Because synchronized responses change with stimulus features, for example the direction in which a bar moves (Eckhorn et al. 1988 Gray et al. 1989 Gray & Singer 1989), the idea receives support that there are stimulus- specific distributed neuron groups. It appears that these neurophysiological data can only be explained if cell as- semblies are assumed that are (a) activated by specific ex- ternal stimuli, (b) distributed over different cortical areas, and (c) connected through cortico-cortical fibers (and pos- sibly additional subcortical connections). These results can be interpreted as evidence for a version of Hebb���s theory according to which cell assemblies must synchronously oscillate at high frequencies when active. However, synchronous oscillations are a special case of well-timed activity (Abeles et al. 1993 Aertsen & Arndt 1993). Therefore, these data are also consistent with the weaker position made explicit by Hebb that cell assemblies generate well-timed activity patterns in their many neu- rons. The latter position would imply that at least a fraction of the activated neurons (e.g., those forming one subgroup represented by an arrow in Fig. 1) exhibit synchronized ac- tivity when the assembly reverberates (see Pulverm��ller et al. 1997 for further discussion). If it is taken into account that most cortico-cortical fibers conduct action potentials with velocities around 5���10 m/s or faster (Aboitiz et al. 1992 Patton 1982), it becomes clear that a wave of activity running through and reverberating within an assembly will lead to rather fast activity changes. Suppose a large-scale physiological recording device (e.g., an electrode recording the local field potential, or even an EEG electrode or an MEG coil) is placed close to a frac- Pulverm��ller: Brain���s language 256 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 Figure 1. Hebb���s (1949) illustration of the inner structure of a cell assembly consisting of several subgroups of neurons. Arrows represent subgroups of neurons that become active at exactly the same time. Numbers indicate the activation sequence following activity of the subgroup labelled 1. An ordered spatiotemporal pattern of activity is produced whenever a wave of excitation runs through the assembly.
tion of the neurons of the assembly sketched in Figure 1. In this case, a reverberating wave of activity in the assembly will cause rather fast activity changes at the recording de- vice. If the neuronal subpopulations represented by arrows are assumed to be located in different cortical areas sepa- rated, say, by a few centimeters, it will take some hun- dredths of a second for neuronal activity to travel the loop labelled 1-2-3 and for the neurons denoted by the first ar- row (the first and the fourth in the sequence) to become synchronously active for the second time. It follows that synchronous and fast reverberating activity in the assembly is most likely to lead to spectral dynamics in the high fre- quency range (.20 Hz) recorded by the large-scale de- vices.2 If specific dynamics in high-frequency cortical activity are taken as an indicator of reverberating activity in Heb- bian cell assemblies, the question of whether particular cognitive processes are related to high-frequency dynamics becomes particularly relevant for further testing the Heb- bian ideas. It is known from animal experiments that if the receptive fields of two neurons in visual cortices are each stimulated by a moving bar and both stimuli are aligned and move together in the same direction, neuron responses can synchronize their fast rhythmic activity. However, if one neuron is stimulated by a bar moving in a particular direc- tion, while the other is stimulated by a bar moving in the opposite direction, synchrony of rhythmic responses van- ishes (Engel et al. 1991a). This result and similar findings indicate that synchrony of high-frequency neuronal activity reflects gestalt criteria, for example the fact that two objects move together (Singer 1995 Singer & Gray 1995). Consis- tent with this finding in animals, patterns of regularly mov- ing bars have been found to evoke stronger high-frequency electrocortical responses recorded in the EEG compared to irregular bar patterns (Lutzenberger et al. 1995). Fur- ther support for the role of high-frequency cortical activity in cognitive processing comes from studies of electrocorti- cal responses to attended and unattended stimuli (Tiitinen et al. 1993). Most important, gestalt-like figures such as Kanizsa���s triangle have led to stronger high-frequency EEG responses around 30 Hz compared to physically similar stimuli that are not perceived as a coherent gestalt (Tallon et al. 1995 Tallon-Baudry et al. 1996). Thus, dynamics of high-frequency responses appear to be an indicator of the cognitive process of gestalt perception. These results are consistent with the idea that gestalts, such as a coherent bar pattern or a triangle, activate cortical cell assemblies that generate coherent high-frequency responses, while physi- cally similar stimuli that are not perceived as coherent gestalts lack cortical representations and, therefore, evoke desynchronized electrocortical responses. Therefore, the idea that cell assemblies are relevant for cognitive process- ing not only receives support from recordings in animals��� brains, but is consistent with noninvasive recordings of hu- man brain activity using large-scale recording techniques such as EEG. In summary, recent theoretical and empirical research provides support for the existence of Hebbian cell assem- blies and for their importance for cognitive brain processes. It must be noted, however, that, based on experimental and theoretical work, the Hebbian concept and the assumptions connected with it have changed slightly. Some of these modifications are summarized in the following postulates (which are closely related to points (1) to (3) in sect. 2): 18. Simultaneous pre- and postsynaptic activity of corti- cal neurons leads to synaptic strengthening. However, pre- or postsynaptic activity alone leads to synaptic weakening. 28. Associations can occur between adjacent neurons and between cortical neurons located far apart, provided there is a synapse connecting them. The cortex is an asso- ciative memory although it is not fully connected. 38. If synaptic strengthening occurs among many neu- rons, they will develop into an assembly that can ignite and exhibit well-timed reverberatory activity. Future empirical testing of the modified Hebbian frame- work is, of course, necessary, and neuroimaging techniques make it possible to perform such testing, although tech- niques available at present do not allow for localizing each member of a widely distributed neuron set in different cor- tical areas. If an assembly ignites and stays active, signs of activity should be visible in single-cell and multiple-unit re- sponses, local field potentials, and more global electrocor- tical activity, and possibly in metabolic changes in the brain as well. The cortical topography of these activity signs may allow for conclusions concerning assembly topographies. In addition to general signs of activity enhancement ��� en- hanced blood flow, larger event-related potentials, more powerful single-cell responses ��� changes in well-timed high-frequency cortical responses may include information about reverberatory neuronal activity in cell assemblies. It may be appropriate at this point to mention possible theoretical problems of the Hebbian approach, some of which have been summarized in a recent article by Milner (1996). If an ignition takes place, there is danger that activ- ity will spread to additional assemblies and finally to the en- tire cortex or even brain, resulting in overactivity such as that seen during seizures. To avoid this, it is necessary to have a control device regulating the cortical equilibrium of activity. This device has been called ���threshold control mechanism��� (Braitenberg 1978b) and its neuroanatomical substrate has been proposed to be located in the basal gan- glia (Miller & Wickens 1991 Wickens 1993) or, as an alter- native, in the hippocampus (Fuster 1995). Furthermore, if a large number of cell assemblies are built up in the cortex, this may lead to an increase in average connection strength, and, in the worst case, to a clumping together of all assem- blies. This would make it impossible to activate representa- tions individually. However, this problem primarily occurs if a coincidence learning rule is assumed (Table 1). If LTD rules are added (e.g., in the case of correlation-based learn- ing as sketched in Table 2), simultaneous activity of a set of cortical neurons will not only lead to synaptic strengthen- ing between them, but also to a weakening of connections to neurons outside the set (Hetherington & Shapiro 1993 Palm 1990 Willshaw & Dayan 1990). In this case, the prob- lem will occur only if w-parameters (see Table 2) are cho- sen inappropriately. It has also been argued that the cell as- sembly framework is not flexible enough to allow for a representation of complex objects. If a house includes a door and a window, how would the respective representa- tions relate to each other? Here, it is necessary to allow for hierarchical organizations of cell assemblies: One assembly may be a subset of another one. This is also important for the semantic representations of words with similar mean- ings, for example, for hyponyms and hyperonyms. Adjust- ment of the global activation threshold may account for whether the set or its subset is being activated (Braitenberg 1978b). Furthermore, concepts that have features in com- Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 257
mon may be represented in cell assemblies that share some of their neurons. These assemblies will, therefore, not be entirely different neuron sets, but they will overlap. The relations of inclusion and overlap can be realized quite nat- urally within a cell-assembly theory built on the Hebbian notion (Braitenberg 1978b Palm 1982). Therefore, a mod- ified version of the original Hebbian proposal appears to be well suited to provide neurobiological answers to important questions in cognitive science. 3. Cortical distribution of cell assemblies In recent years, the Hebbian idea of distributed assemblies with defined cortical topographies has been incorporated into large-scale neuronal theories of language and other cognitive functions (Abeles 1991 Braitenberg & Sch��z 1991 Damasio 1989a Edelman 1992 Elbert & Rockstroh 1987 Fuster 1995 Gerstein et al. 1989 Mesulam 1990 Miller & Wickens 1991 Palm 1982 Posner & Raichle 1994 Pulverm��ller 1992 Singer 1995 Wickens et al. 1994). At this point, there appears to be a consensus that neurons in distant cortical areas can work together as functional units. However, the Hebbian framework would not only postulate that there are large-scale neuronal networks, it also pro- vides clear-cut criteria for the formation of cell assemblies and, therefore, straightforward predictions on assembly topographies. For assembly formation, Hebb (1949) outlines the fol- lowing scenario (pp. 235f): If a particular object is fre- quently being visually perceived, a set of neurons in visual cortices will repeatedly become active at the same time. Therefore, a cell assembly will form representing the shape of the object. This assembly is distributed over cortical re- gions where simultaneous neuronal activity is evoked by vi- sual stimulation, that is, in primary and higher-order visual cortices in the occipital lobes, for example in Brodmann���s (1909) areas 17, 18, 19, and 20. For convenience, Figure 2 displays a lateral view of the left cortical hemisphere on which the approximate locations of Brodmann���s areas are indicated. If correlated neuronal activity is caused by input through other sensory modalities, or if it is related to motor output, the cortical distribution of the coactivated set of neurons will be different. For example, if motor behavior co-occurs with sensory stimulation, cell assemblies may form including neurons in motor and sensory cortices. To put it in a more general way, the cortical localization of a representation is a function of where in the cortex simulta- neous activity occurred when the representation was ac- quired or learned. Whereas correlated neuronal activity of a connected cor- tical neuron set is a sufficient condition for cell assembly formation, correlated occurrence of sensory stimuli is not. In the most extreme case, when an individual is asleep, cor- related stimuli (e.g., in the somatosensory and acoustic mod- ality) may not cause enough cortical activity to lead to synaptic strengthening. The same may be true in an in- dividual exhibiting very low arousal. Furthermore, the amount of cortical activation caused by a stimulus depends on whether it is being attended (Heinze et al. 1994 Man- gun 1995). Therefore, to make it possible for correlated stimuli to induce synaptic learning, sufficient arousal and attention to these stimuli appear necessary, and synaptic learning may depend on how much attention is being di- rected to relevant stimuli. In the following considerations it will be tacitly assumed that correlated stimuli receive a suf- ficient amount of attention from the learning individual to allow long-lasting changes of synaptic connections to occur. 3.1. Assemblies representing word forms Turning to language, it appears relevant to ask where in the cortex correlated neuronal activity occurs during verbal ac- tivities at early ontogenetic stages, when language learning takes place (Pulverm��ller 1992 Pulverm��ller & Schumann 1994). The infant���s repeated articulations of syllables dur- ing the babbling phase are controlled by neuronal activity in inferior motor, premotor, and prefrontal cortices (Brod- mann areas 4, 6, 44, 45). One may well envisage that one specific synfire chain controls the articulation of a given syl- lable and thus represents its articulatory program (Braiten- berg & Pulverm��ller 1992). In addition to and simultane- ous with cortical activity related to motor programs, specific neurons in the auditory system are stimulated by the sounds produced during articulation (Braitenberg & Sch��z 1992 Fry 1966). These neurons are localized in primary and higher-order auditory cortices (superior temporal lobe Brodmann areas 41, 42, and 22). Furthermore, somatosen- sory self-stimulation during articulatory movements evokes activity in somatosensory cortices (inferior parietal lobe ar- eas 1���3 and 40). Therefore, neuronal activity can be as- sumed to be present almost simultaneously in defined pri- mary and higher-order motor and sensory (auditory and somatosensory) cortices. All of these areas are within the first gyrus surrounding the sylvian fissure, the so-called perisylvian cortex (Bogen & Bogen 1976). Neuroanatomi- cal evidence from monkeys suggests that the perisylvian ar- eas are strongly and reciprocally connected, whereby long- distance connections between areas anterior to motor, adjacent to primary auditory, and posterior to primary so- matosensory cortex are particularly relevant (Deacon 1992a Pandya & Yeterian 1985 Young et al. 1995). Given that necessary long-distance connections are available, it follows by learning rule 19 (see also Table 2) that the coac- tivated neurons in the perisylvian areas develop into cell as- semblies (Braitenberg 1980 Braitenberg & Pulverm��ller Pulverm��ller: Brain���s language 258 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 Figure 2. Lateral view of the left cortical hemisphere. Brod- mann���s (1909) areas are indicated. (Adopted from Pulverm��ller & Preissl 1991.)
1992 Braitenberg & Sch��z 1992 Pulverm��ller 1992). Fig- ure 3 represents an attempt to sketch such a perisylvian as- sembly. The individual circles in this diagram are thought to represent local clusters of strongly connected neurons. On the psychological level, the network may be considered the organic counterpart of a syllable frequently produced during babbling, or as the embodiment of the phonological form of a word acquired later during language acquisition. The Hebbian framework implies that different gestalts and word forms have distinct cortical assemblies, because perception of these entities will activate different but pos- sibly overlapping populations of neurons. If a language is not learned through the vocal and auditory modalities, but through the manual and visual modalities (sign languages), cortical localization of cell assemblies representing mean- ingful elements should be different. Because gestures are performed with both head and hands and perceived through the eyes, they are related to neuronal activity far- ther away from the sylvian fissure (more superior motor cortices and occipital visual cortices). Thus, it must be as- sumed that meaningful gestures included in sign languages involve these extra-perisylvian visual, motor, and associa- tion cortices (see Pulverm��ller 1992 for futher discussion). In assuming cell assemblies distributed over perisylvian cortices, the Hebbian perspective is in apparent contrast to older localizationist models according to which motor and acoustic representations of words are stored separately in Broca���s (areas 44 and 45) and Wernicke���s regions (posterior part of area 22), respectively (Geschwind 1970 Lichtheim 1885 Wernicke 1874). The Hebbian view implies that the motor and acoustic representations of a word form are not separate, but that they are strongly connected so that they form a distributed functional unit. For this unit to function properly, both motor and acoustic parts need to be intact. This is important for the explanation of aphasias, in partic- ular of the fact that in the majority of cases these organic language disturbances affect all modalities through which language is being transmitted. Whereas localizationist mod- els have great difficulty explaining this (see, e.g., Lichtheim 1885 for discussion), a cell assembly model can account for the multimodality of most aphasias.3 Furthermore, the as- sumption that word form representations are distributed over inferior frontal and superior temporal areas receives support from imaging studies revealing simultaneous acti- vation of both language areas when words or word-like elements are being perceived (Fiez et al. 1996 Mazoyer et al. 1993 Zatorre et al. 1992). 3.2. Cortical lateralization From the Hebbian viewpoint, localization of language mechanisms is determined by associative learning and by the neuroanatomical and neurophysiological properties of the learning device (the cortex). The cortical loci where si- multaneous activity occurs during motor performance and sensory stimulation follow from the wiring of efferent and afferent cortical connections, which are genetically deter- mined. Genetic factors are also important for the formation of cortico-cortical fiber bundles, which are a necessary con- dition for long-distance association of coactivated neurons located in different areas. Furthermore, a pure association- ist approach may have difficulty explaining why, in most right-handers, the left hemisphere ��� but not the right ��� is necessary for many aspects of language processing. Left hemispheric ���language dominance��� is evident from lesion studies in adults and in infants (Woods 1983) and from psy- chophysiological experiments in young children, demon- strating that stronger language-specific electrocortical ac- tivity can be recorded from the left hemisphere than from the right (Dehaene-Lambertz & Dehaene 1994 Molfese & Betz 1988). Neuroanatomical correlates of language later- ality have been found in the size of perisylvian areas (Gala- burda et al. 1978 1991 Geschwind & Levitsky 1968 Stein- metz et al. 1990) and in size (Hayes & Lewis 1993), ordering (Seldon 1985), and dendritic arborization (Jacobs et al. 1993 Jacobs & Scheibel 1993 Scheibel et al. 1985) of pyramidal cells in the language areas. For differences in size of particular areas, epigenetic processes appear to be very important (Steinmetz et al. 1995). It is well known that differences in cell size and dendritic arborization may be in- fluenced by sensory stimulation and motor output (Dia- mond 1990 Diamond et al. 1967) and, consistent with this view, language laterality has been proposed to be caused by environmental factors, such as lateralized auditory stimula- tion before birth (Previc 1991). Such stimulation may well underlie some of the morphological asymmetries men- tioned. However, there are also arguments for a contribu- tion of genetic factors to language lateralization (Annett 1979). At this point, it therefore appears safer not to dismiss a possible role of genetics here. For the Hebbian frame- work to operate, an anatomical substrate is necessary and this substrate is determined by genetic factors. Neverthe- less, given the brain with its preprogrammed input and out- put pathways, its specific cortico-cortical projections, and its probably genetically determined left-hemispheric pref- erence for language, the Hebbian approach leads to highly specific hypotheses about cortical distribution of language- related processing units. One of these hypotheses concerns the cortical realization of laterality of language. According to localizationists, lan- guage processes take place in only one hemisphere. In con- trast, the Hebbian framework suggests a different view. Al- though genetic and/or environmental factors lead to stronger language-related activation of left perisylvian cor- tex when language is being produced or perceived, articu- Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 259 Figure 3. The cell assembly representing a phonological word form may be distributed over perisylvian areas. Circles represent local neuron clusters and lines represent reciprocal connections between such clusters. The connections are assumed to have strengthened because of correlated activity of neurons during ar- ticulation of the word form.
lation of a word form is probably controlled by bi-hemi- spheric activity in motor regions, and acoustic perception of the word certainly leads to activation of bilateral auditory cortices. Because neurons in both hemispheres are coacti- vated when a word form is being produced or perceived, the cell assembly representing the word form should be dis- tributed over bilateral perisylvian cortices (Mohr et al. 1994b Pulverm��ller & Mohr 1996 Pulverm��ller & Sch��nle 1993). However, if the left hemisphere���s neurons are more likely to respond to language stimuli and to control pre- cisely timed articulations, cell assemblies representing word forms would be gradually lateralized to the left in the following sense: They include a large number of neurons in the left hemisphere and a smaller number of neurons in the right. According to this view, a lateralized cell assembly is not restricted to one hemisphere, but a greater percentage of its neurons would be in the ���dominant��� hemisphere and a smaller percentage in the ���nondominant��� hemisphere (Pulverm��ller & Mohr 1996). What would be the cause of this lateralization? Given that genetically programmed differences in the hemi- spheres��� anatomical and physiological properties are the cause of lateralization of cognitive functions, it becomes im- portant to develop ideas about how left/right differences in the ���hardware��� could influence the ���software.��� Based on an extensive and profound review of neuroanatomical and neurophysiological asymmetries, Robert Miller (1987 1996) recently proposed that axonal conduction times in the left hemisphere are slightly slower, on average, than those in the right hemisphere. According to Miller, this may lead to a bias in favor of the left hemisphere for storing short time delays, such as are important for distinguishing be- tween certain phonemes (Liberman et al. 1967). For ex- ample, the probability of finding a neuron that responds specifically to a [p], but does not respond to a [b], may be greater in the left hemisphere than in the right, because neurons with slowly conducting axons that could be used as delay lines for hardwiring the long (.50 msec) voice onset time of the voiceless stop consonant would be more com- mon in the left hemisphere. The availability of axons with particular conduction times may also be relevant for at- tributing additional distinctive features to acoustic input (Sussman 1988 1989). If neurons sensitive to certain pho- netic features have a higher probability of being housed in the left hemisphere, the neuron ensemble representing a phonological word form should finally be lateralized to the left. Although Miller���s theory of cortical lateralization needs further support by empirical data, it clearly illustrates how hemispheric specialization at the cognitive and functional levels may arise from basic neuroanatomical and physio- logical differences between the hemispheres. 3.3. Word categories Associative learning may not only be relevant for the corti- cal representation of word forms, it may also play an im- portant role in the acquisition of word meanings. When the meaning of a concrete content word is being acquired, the learner may be exposed to stimuli of various modalities re- lated to the word���s meaning, or the learner may perform ac- tions to which the word refers. Although such stimulus and response contingencies are certainly not sufficient for full acquisition of word meanings (Gleitman & Wanner 1982 Landau & Gleitman 1985) ��� they would not, for example, allow the learner to distinguish between the morning and the evening star (Frege 1980) ��� they may nevertheless have important brain-internal consequences. From the Hebbian viewpoint, it is relevant that neurons related to a word form become active together with neurons related to perceptions and actions reflecting aspects of its meaning. If this coacti- vation happens frequently, it will change the assembly rep- resenting the word. Coactivated neurons in motor, visual, and other cortices and the perisylvian assembly represent- ing the word form will develop into a higher-order assem- bly. A content word may thus be laid down in the cortex as an assembly including a phonological (perisylvian) and a se- mantic (mainly extra-perisylvian) part (Pulverm��ller 1992). After such an assembly has formed, the phonological sig- nal will be sufficient for igniting the entire ensemble, in- cluding the semantic representation and, vice versa, the as- sembly may also become ignited by input only to its semantic part.4 Thus, frequent co-occurrence and correla- tion of word form and meaning-related stimuli is only nec- essary at some point during the acquisition process. Later on, the strong connections within the higher-order assem- bly guarantee ignition of the entire assembly when part of it is being activated and, thus, they guarantee a high corre- lation of activity of all assembly parts, and, consequently, the endurance of the assembly. When phonological word forms become meaningful, quite different cortical processes may take place, depend- ing on what kind of information is being laid down in the associative network. Hebbian associationist logic suggests that cortical representations differ radically between words of different vocabulary types. In the following paragraphs, a few such differences will be discussed. 3.3.1. Content and function words. Neurons activated by stimuli related to the meaning of most concrete content words (nouns, adjectives, and verbs) are likely to be housed in both hemispheres. For example, the visual perceptions of objects that can be referred to as ���mouse��� will probably activate equal numbers of left- and right-hemispheric neu- rons because a corresponding visual stimulus is equally likely to be perceived in the right and left visual half-fields, and, in many cases, will be at fixation so that half of it is pro- jected to the left visual field (right hemisphere) and the other half to the right visual field (left hemisphere). There- fore, if word form representations are strongly lateralized to the left, the assemblies representing content words (word form plus meaning) will be less strongly lateralized. Assemblies with different degrees of laterality are sketched in Figure 4. In contrast to content words with concrete and well- imaginable meaning, function words such as pronouns, aux- iliary verbs, conjunctions, and articles serve primarily a grammatical purpose. Many of them contribute signifi- cantly to the meaning of sentences, for example, ���and,��� ���or,��� ���not,��� and ���if.��� However, their meanings cannot be ex- plained based on objects or actions to which the words re- fer. Rather, their meaning appears to be a more complex function of their use (Wittgenstein 1967) and can only be learned in highly variable linguistic and nonlinguistic con- texts. Evidently, the correlation between the occurrence of a particular function word and certain stimuli or actions is low. Therefore, there is no reason why the perisylvian as- sembly representing the word form should incorporate ad- ditional neurons. If this is correct, assemblies representing Pulverm��ller: Brain���s language 260 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2
function words remain limited to the perisylvian cortex and strongly left-lateralized in typical right-handers. Note that this argument depends on the formulation of the cortical learning rule. If coincidence of neuronal activ- ity was the factor causing synaptic modification, function words should have widely distributed cell assemblies be- cause these words occur in a multitude of stimulus constel- lations and, in addition, they occur much more frequently than most content words (Francis & Kucera 1982 Ortmann 1975). When a function word (e.g., the article ���the���) is be- ing learned, it may be used with various content words (���the cat,��� ���the dog,��� ���the horse���) and, if there is a systematic re- lationship between the use of the content words and the oc- currence of nonlinguistic stimuli (e.g., animal pictures), there will be a strong coincidence between the occurrences of each of these nonlinguistic stimuli and the word form. If only coincidence learning took place, cell assemblies rep- resenting function words should include even more neu- rons in visual cortices than most content word assemblies, because the assembly representing the function word would incorporate all neurons related to coincident visual nonlinguistic stimuli. However, because connections weaken if only pre- or only postsynaptic neurons fire (Table 2), the relatively infrequent co-occurrence of the function word with each of the visual stimuli will guarantee that its as- sembly does not become associated with representations of either visual stimulus. Correlation of neuronal activity is im- portant for synaptic strengthening in the cortex, and this implies that function words are represented in cell assem- blies restricted to perisylvian areas, or, at least, that they do not include large numbers of neurons outside. 3.3.2. Abstract content words. One may argue that the postulated difference in semantic meaning between con- tent and function words does not apply for all members of these vocabulary classes. Rather, it appears that there is a continuum of meaning complexity between the ���simple��� concrete content words that have clearly defined entities they can refer to (so-called referents), more abstract items that may or may not be used to refer to objects and actions, and function words that cannot be used to refer to objects. It is therefore inappropriate to make a binary distinction between vocabulary classes based on semantic criteria. If semantic criteria are crucial for intracortical representa- tion, the suggested gradual differences in the correlation between word form and meaning-related stimuli or actions should be reflected in gradual differences in cortical later- alization and how assemblies are distributed. An abstract content word, such as ���philosophy,��� may therefore have an assembly somewhat in-between typical content and func- tion word assemblies: It may exhibit an intermediate degree of laterality consisting mainly of perisylvian neurons, but in- cluding a few neuron clusters outside perisylvian areas. Among the abstract content words are words referring to emotional states, for example ���anger��� and ���joy.��� For these words, it is not difficult to find characteristic visual stimuli related to their meaning ��� for example, angry or joyful faces. In addition, there are characteristic meaning-related patterns of muscle activity ��� namely, the contraction of the respective facial muscles ��� and autonomic nervous system activity (Ekman et al. 1983 Levenson et al. 1990). It should therefore be noted that, although these words do not refer to objects and actions in the sense in which the word ���house��� refers to an object, the likely co-occurrence of pat- terns of muscle contractions with the word forms may nev- ertheless lead to the formation of widely distributed corti- cal cell assemblies representing these words. In addition to cortical neurons added to the word form representations during learning, it has been proposed that these assemblies acquire additional links to subcortical neurons in structures of the limbic system related to emotional states (Pulver- m��ller & Schumann 1994). ���Emotion words��� may there- fore be represented by a cortical assembly plus a limbic as- sembly-tail. The amygdala and the frontal septum may be the most important structures for linking the cortical as- sembly to its subcortical tail (Schumann 1990 1997). These considerations should make it clear that the de- gree of abstractness of an item is not the only factor influ- encing assembly topographies. According to the present proposal, the important criterion is the strength of the cor- relation between the occurrences of a given word form and a class of nonlinguistic stimuli or actions. In the clear cases, this likelihood is related to abstractness, but there are ex- ceptions. 3.3.3. Action words, perception words, and other word classes. Content words are used to refer to odors, tastes, somatic sensations, sounds, visual perceptions, and motor activities. During language learning, word forms are fre- quently produced when stimuli to which the words refer are perceived or actions to which they refer are carried out by the infant. If the cortex is an associative memory, the modal- ities and processing channels through which meaning- related information is being transmitted must be impor- tant for formation of cortical assemblies. This has inspired models of word processing in the brain postulating dis- tinct corical representations for word classes that can be distinguished based on semantic criteria (Warrington & McCarthy 1987 Warrington & Shallice 1984). If the modality through which meaning-related informa- tion is transmitted determines the cortical distribution of cell assemblies, a fundamental distinction between action and perception words can be made. Action words would re- fer to movements of one���s own body and would thus be used frequently when such actions are being performed. In this case, a perisylvian assembly representing the word form would become linked to neurons in motor, premotor, and Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 261 Figure 4. Cell assemblies relevant for cognitive processing may be distributed over both hemispheres and may be lateralized to different degrees. Whereas for cell assemblies representing phonological word forms and grammatical function words a high degree of laterality appears likely (right), an assembly represent- ing a concrete content word may exhibit a reduced degree of lat- erality (left). (Adopted from Pulverm��ller & Mohr 1996.)
prefrontal cortices related to motor programs. Perception words, whose meaning can best be explained using proto- typical stimuli, would consist of a perisylvian assembly plus neurons in posterior cortex. In many cases, visual stimuli are involved and the respective word category may there- fore be labelled vision words. Assemblies representing words of this category would be distributed over perisylvian and visual cortices in parietal, temporal, and/or occipital lobes. Figure 5 presents sketches of the assembly types pos- tulated for action and vision words. Examples of words whose meanings are related to the visual modality are con- crete nouns with well-imaginable referents, such as animal names. The best examples of action words are in the cate- gory of action verbs. This model draws too simple a picture of the relation be- tween word forms and their meanings, because it does not explain homonymy (Bierwisch 1982 Miller 1991). If a phonological word form has two exclusive meanings ��� if it can, for example, be used as a noun with one meaning or as a verb with another meaning (the/to beat) ��� a mechanism must be assumed that realizes the exclusive-or relationship between the two meanings. As suggested earlier, homo- nyms could be represented by overlapping cell assemblies, that is, by two content word assemblies sharing one peri- sylvian phonological part. Inhibition between the semantic assembly parts is unlikely to be wired in cortex, because the percentage of cortical inhibitory neurons is low and these neurons are usually small (Braitenberg & Sch��z 1991). In- tracortical inhibitors would therefore be unlikely candi- dates for mediating inhibition between cortical areas ��� for example, between assembly parts in frontal and occipital lobes. However, such mutual inhibition between overlap- ping assemblies could be realized by striatal connections (Miller & Wickens 1991). Accordingly, homonymic content words may be realized as widely distributed assemblies sharing their perisylvian part while inhibiting each other through striatal connections. This wiring would allow the perisylvian word form representation to become active to- gether with only one of its ���semantic��� assembly parts (see Pulverm��ller 1992 for further discussion).5 The argument made above for action and visually-related words can be extended to words referring to stimuli per- ceived through other modalities. For those, additional word categories ��� odor, taste, pain, touch, and sound words ��� can be postulated. Members of these word classes should be represented in assemblies with specific cortical topogra- phies. For example, whereas an assembly representing a pain or touch word may include substantial numbers of neurons in somatosensory cortices, sound words may have exceptionally high numbers of neurons in bilateral auditory cortices included in their assemblies. Again, it must be stressed that neurons responding to stimuli of various modalities and neurons controlling body movements and actions are located in both hemispheres. It is for this reason that cell assemblies representing these words are assumed to be distributed over both hemispheres and to be less strongly lateralized compared to assemblies representing function words (Pulverm��ller & Mohr 1996). The definition of action words is particularly delicate be- cause not all action-related associations involve the motor modality. Here it is important to distinguish movements that are performed by the subject���s own body from move- ments that are only perceived visually. ���To fly��� or ���the plane,��� for example, are words that are frequently heard by children when they perceive certain moving visual stimuli. Although a relation of visual stimuli to the motor modality can hardly be denied ��� because perception of visual stim- uli is usually accompanied by eye movements related to neuronal activity in frontal eye fields ��� this eye movement- related neuronal activity is probably not very stimulus-spe- cific (similar saccades are made when different objects are looked at). Therefore, the correlation between visual input patterns and the occurrence of the word forms ���fly��� or ���plane��� may be highest and these words may thus be orga- nized in assemblies including a significant number of neu- rons in visual cortices responding to specific moving con- tours. These words should therefore be classified not as action words but as visually-related words of a certain kind (as words referring to visually perceived movements). On the other hand, action words as defined above, that is, words usually referring to movements of one���s own body, may include movement detectors in visual cortices in their assemblies. Many body movements are visually perceived when they are performed, suggesting that sensory-motor assemblies are established for representing these actions ��� an idea for which there is ample support from recent stud- ies (Fadiga et al. 1995 Gallese et al. 1996 Rizzolatti et al. Pulverm��ller: Brain���s language 262 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 Figure 5. Whereas words eliciting strong visual associations (���vi- sion words���) may be organized in assemblies distributed over peri- sylvian and additional visual cortices, words that remind one of movements of one���s own body (���action words���) may be organized in assemblies distributed over perisylvian and additional motor cortices. Many (but not all) concrete nouns are vision words and many verbs are action words.
1996). These considerations indicate that Figure 5 draws too crude a picture of cell assemblies representing action words. Such assemblies can include additional neurons in visual cortices primarily processing movement information ��� many of which are probably located in the posterior part of the middle temporal gyrus (Watson et al 1993 Zeki et al. 1991). A similar point can be made for somatosensory stim- ulations caused by body movements, suggesting that neu- rons in parietal cortices may be added to the assembly rep- resenting an action word, as well. Further word class-distinctions can be made based on the cortical areas active during meaning-related motor ac- tivity. Different kinds of action words can be distinguished considering the muscles most relevant for performing the actions (to chew, to write, to kick), the complexity of the movement (to knock, to write), and the number of muscles involved (to nod, to embrace). These factors may ���shift��� the neurons in frontal lobes added to the perisylvian assembly in the inferior/posterior (mouth/hand/foot representation) or anterior/posterior direction (complex/simple move- ments), or enlarge/reduce their cortical distribution (many/ a few muscles involved in movement). Similar, more fine-grained distinctions are desirable for visually-related words. Some vision words refer to static ob- jects (house), others to moving objects (train), some refer to colors or colored objects (iguana), others to objects lack- ing colors (penguin). Furthermore, some visual stimuli are very simple (line), others are more complex (square, cube, house, town, megalopolis). This suggests that different sets of neurons are being added to the assembly when contin- gencies between words and different kinds of visual stimuli are being learned. The assembly of a word used to refer to colors or colored objects may include neurons maximally responding to color, and, as discussed above, neurons sen- sitive to moving visual stimuli may be included in the as- semblies representing words referring to such stimuli. Re- cently, cortical processing streams have been discovered in temporal lobes that are primarily concerned with move- ment or color information from the visual input (Corbetta et al. 1990 Watson et al. 1993 Zeki et al. 1991). If move- ment-detecting cells are more frequent in one area, for ex- ample in the posterior middle temporal gyrus, and neurons in primary and secondary visual cortex that respond to color preferentially project to other areas, for example in the in- ferior temporal lobe, this would suggest that words refer- ring to colors or colored objects are realized as assemblies including additional neurons in color areas (e.g., in the in- ferior temporal gyrus), and that words referring to visually perceived movements have assemblies that comprise addi- tional neurons in visual movement areas (in the middle temporal gyrus). It is important to stress that (1) word types defined in this way6 do not necessarily have a congruent lexical category most ��� but not all ��� verbs are action words, and there may be action words from other lexical categories and (2) it is not always clear from theoretical consideration to which category a particular word should be assigned. Most con- crete content words probably exhibit a high correlation with stimuli of more than one modality, and their presentation may therefore remind subjects of multimodal stimuli. Whereas verbs referring to body movements are probably action words, and concrete nouns (such as animal names) are almost certainly related to vision, other word groups ��� for example, nouns referring to tools ��� probably lead to both visual and motor associations. Therefore, when evalu- ating the present ideas about word class-differences related to word meaning in neuroscientific experiments, it is most important to assess quantitatively semantic associations elicited by word stimuli. 4. Cortical activation during word processing: Predictions and methodological remarks Cognitive brain theories lead to empirical predictions in psychophysiological studies. Testing such predictions is not trivial, however. In the case of language, it is particularly dif- ficult to design experiments and interpret their results be- cause there are so many possible confounds to which, for example, a physiological processing difference between two stimulus words could be attributed. Furthermore, the sub- traction logic used in many imaging studies of cognitive processes has frequently been criticized, and one may pre- fer designs that could prove more useful in testing precise predictions on cognitive processes of comparable complex- ities. After summarizing selected predictions derived from the Hebbian model (sect. 4.1), the subtraction logic underlying many imaging studies will be contrasted to what will be called the double dissociation approach to neuroimaging (sect. 4.2), and, finally, methodological issues specific to the investigation of word processing will be addressed (sect. 4.3). 4.1. Predictions about where and how Hebbian logic suggests that content and function words, and words referring to actions and perceptions, have dif- ferent neurobiological counterparts. The cell assemblies representing these lexical elements may differ with regard to their laterality and cortical topography. Whereas all as- semblies representing words are assumed to include a strongly lateralized perisylvian part, neurons outside peri- sylvian language areas (and in both hemispheres) would only be added to the assembly if words refer to actions and perceivable objects. If assembly topographies are a function of semantic word properties, signs of cortical activity should differ when these different assemblies are being activated.7 Based on these ideas, one would expect: 1. function words to evoke strongly left-lateralized signs of cortical activity restricted to perisylvian cortices, 2. content words to evoke less lateralized signs of corti- cal activity in perisylvian areas and outside, 3. action words to evoke additional activity signs in mo- tor cortices of frontal lobes,8 and 4. visually-related words to evoke additional activity signs in visual cortices of occipital and inferior temporal lobes. These are some of the predictions obvious from the above considerations (sect. 3) that relate to the where question. When the assumptions leading to these predictions were discussed in section 3, the why question was traced back, in each case, to a Hebbian learning rule postulating that cor- related neuronal activity is the driving force of assembly for- mation. With regard to the how question, it is important to recall that cell assemblies were assumed to exhibit two functional states, namely, ignition (or full activation) and re- verberation (or sustained partial activity). When outlining Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 263
empirical tests of the cell assembly framework and its ap- plication to language, one may not only be interested in testing predictions about assembly topographies, but one may also want to think about how to distinguish and detect possible physiological signs of ignition and reverberation. As detailed in section 2, ignition may be reflected in a sud- den spreading of neuronal activity shortly after stimulation, and reverberation would follow ignition and could become visible in high-frequency brain responses. Therefore, the following additional predictions are possible: 5. shortly after stimulation, signs of cell assembly igni- tion are simultaneously present at the cortical loci where the assembly is located, and 6. after a longer delay, signs of reverberation emerge in the same areas. It is not possible to deduce the exact point in time when these putative physiological processes take place. However, because words are recognized rather quickly ��� for example, lexical decisions, that is, judgments on letter strings ac- cording to whether they are real words or not, can be made as early as !s second after the onset of written stimuli ��� it is clear that the postulated physiological process of cell as- sembly activation must take place during the first few hun- dreds of milliseconds after the stimulus has been pre- sented. Although numerous additional predictions can be de- rived from the discussion in section 3, sections 5 and 6 will focus on hypotheses 1���6. These hypotheses will be discussed based on the results of psychophysiological and neuroimaging experiments. 4.2. Subtractions versus double dissociations in psychophysiology In psychophysiology, numerous neuroimaging techniques are available for investigating higher cognitive processes. Activity of large neuron ensembles can be visualized using electrophysiological recording techniques, such as electro- encephalography (EEG) and magnetoencephalography (MEG). These techniques provide exact information about temporal dynamics of electrophysiological activation and deactivation processes that occur in the millisecond range. They also allow for localization of sources, although such lo- calization is usually much less precise than imaging of brain metabolism. Metabolic imaging techniques with high spa- tial resolution, such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI), are extremely valuable for localizing brain structures that maximally become active, thereby increasing their meta- bolic rates during cognitive tasks. However, the metabolic methods give only a rough picture of temporal dynamics of brain processes, and it is therefore important to use both electrophysiological and metabolic imaging techniques when investigating brain processes of cognitive functions. It is necessary to recall that important information about where, why, and how cognitive processes take place in the human brain was obtained before modern imaging tech- niques were available. Most of these studies used the indi- viduals��� behavior as the dependent measure. In addition, studies of neurological patients with focal lesions can an- swer the question of which brain structures are necessary for particular cognitive operations (Jackson 1878 1879). Studies of healthy individuals in whom stimulus infor- mation reaches only one hemisphere ��� for example, using the technique of lateralized tachistoscopic presentation of visual stimuli ��� can provide important insights into the hemispheres��� roles in language processing (Hellige 1993 Pulverm��ller & Mohr 1996). Together with such neuropsychological evidence, modern neuroimaging and psychophysiological data can provide even stronger conclu- sions about language mechanisms in the human brain (Pos- ner & Raichle 1994). In recent years, a large number of imaging studies of word processing have been carried out, many of which are relevant for evaluating the Hebbian model outlined above. When interpreting these results, it is necessary to consider basic methodological issues. Giving an overview of all pos- sible methodological problems that may become relevant is outside the scope of the present article (see, e.g., Posner & Raichle 1995 and comments therein). Rather, two impor- tant points will be mentioned briefly, the so-called subtrac- tion logic and the question of stimulus matching, which are both crucial for investigating word class-differences. Various dependent measures recorded by large-scale imaging techniques are usually interpreted as signs of cor- tical activity. However, the exact mechanisms by which an increase in cortical activation (i.e., the frequency of excita- tory postsynaptic potentials in a set of neurons) may lead to an increase in the CO2 concentration in numerous blood vessels, to an increase in intracellular glucose levels, to an enhancement of biomagnetic signals, or to a more positive or negative event-related brain potential are not sufficiently understood to make quantitative predictions possible. For example, one may predict that higher glucose metabolism or event-related potential amplitudes are present in or close to the inferior prefrontal cortex during processing of a given word class, but quantification of the expected difference, in terms of microvolts, for example, would not be possible. Ul- timately, even the rationale underlying the more/less logic may be flawed, because an increase in biomagnetic activity or enhancement of cortical metabolism may be caused by the activation of inhibitory neurons (Mitzdorf 1985 Posner & Raichle 1995). Nevertheless, at least in the cortex, exci- tatory neurons represent the majority (85% of cortical neurons are excitatory), and they are, on average, much larger than inhibitory neurons (Braitenberg & Sch��z 1998). Furthermore, their function is probably to control excita- tory activity in cortex, rather than to process more specific information. It is therefore possible, but not likely, that an enhancement of large-scale measures of cortical activity ex- clusively reflects inhibitory processes on the neuronal level. (This may be more likely for structures with high percent- ages of inhibitory neurons, such as the striatum.) Therefore, in the majority of cases, it appears reasonable to use large- scale neuroimaging measures to draw conclusions on activ- ity changes in large numbers of excitatory neurons in the cortex. The logic underlying all imaging work is that a dependent measure indicates a difference in brain activity between two conditions. In most cases, a critical condition is com- pared to a baseline or control condition. In the simplest case, looking at an empty computer screen or at a fixation cross may be compared to reading words or to making lex- ical decisions on these stimuli. Using a more complex de- sign, the task of silently reading a word may be compared to the generation of a verb that somehow relates to the meaning of a displayed word. If an area of cortex is found to ���light up��� in such an experiment, one can conclude that Pulverm��ller: Brain���s language 264 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2
the perceptual, cognitive, or motor operations induced by the two conditions differ with regard to neuronal activity in this particular area. Unfortunately, however, in many experiments there are several differences between critical and control conditions. For example, the tasks of looking at an empty screen and of making lexical decisions about words appearing on the screen differ with regard to several aspects: (1) perceptions ��� either a word or nothing is being perceived (2) higher cognitive processes ��� the stimulus has to be classified as a real word or as a meaningless element, or nothing has to be done and (3) motor activities ��� a button press is either re- quired or not. In addition, silently reading a noun (e.g., cow) and silently generating a word that refers to an activ- ity related to the object to which the noun refers (e.g., to milk, to buy) involve quite different cognitive processes. Al- though identical words may be displayed in the two condi- tions and no overt response may be required, the two con- ditions differ because only one of them requires strong attention and involves search processes, semantic infer- ences, repeated lexical access, and so on (see also the dis- cussion in Posner & Raichle 1995). Finally, another differ- ence between the reading and the generation tasks is that only in the latter are verbs involved (but nouns are being read in both conditions). Given that an area is found to ���light up��� in the generation condition if compared to the reading condition, it is not clear which of the many differ- ent cognitive processes relates to the difference in brain ac- tivity. The difference may even be used to evaluate predic- tion 3 (sect. 4.1) because action verbs are relevant in only one of the conditions, but, of course, if the prediction is met, the experimental result would not provide strong sup- port for it because of the many confounds. A solution to the problem may lie in a more careful se- lection of the conditions and stimuli that are being com- pared. If, for example, silently reading words is compared to reading random letter strings made up of the same let- ters, one may argue that in this case the critical and control conditions differ only with regard to well-defined linguistic processes, such as word form identification and processing of semantic information. However, the objection can be raised that processing of words is not even necessary under such conditions because random letter strings can fre- quently be distinguished from real words merely by looking at the first three letters of the items and deciding whether these letters can be combined according to the phonologi- cal or orthographic rules of the language from which the real words are taken. Thus, word processing could be avoided by experiment participants in these conditions. To allow conclusions on processes specific for words, even more similarity between the stimulus classes should be re- quired. For example, only letter strings that are in accord with the phonological rules of the language could be al- lowed as pseudowords, and lexical decisions could be re- quired so that experiment participants would be forced to attend to and process the stimuli. In this case, a neuro- imaging difference between conditions could be attributed to the difference between word and pseudoword process- ing, although from a psycholinguistic perspective these pro- cesses may differ under various aspects (including word form identification, semantic processes, and the use of a ���time out��� strategy for rejecting pseudowords Grainger & Jacobs 1996 Jacobs & Grainger 1994 Mohr et al. 1994b). Nevertheless, a difference in brain activity between these conditions would allow stronger conclusions on the cortical processes induced by the words. In many cases, two conditions are being compared in which condition 1 is considered to induce a subset of the processes induced in condition 2. The subtraction of the brain responses would then be interpreted as reflecting the psychological processes that condition 2 exhibits but condition 1 lacks. Subtractions can be performed repeat- edly, so that a hierarchy of conditions corresponds to a set of subtractions (Posner & Raichle 1995). However, the principal problems remain, namely, (I) that a difference in more than one psychological process may be attributed to each pair of conditions, making it difficult to attribute a physiological contrast to one of them, and (II) that statisti- cal criteria for the comparison of two conditions are diffi- cult to choose if multiple pairs of physiological data are compared. If many comparisons are being made (when data from tens of channels or thousands of voxels are con- trasted), the likelihood of a difference occurring by chance is high. On the other hand, if critical significance levels are adjusted to reduce the likelihood of significant results (e.g., by following Bonferoni logic), an actual difference between brain responses in two conditions may be masked because the too rigid statistical criterion is almost impossible to reach (Wise et al. 1991). The only way to avoid problem (I) appears to be to choose maximally similar experimental conditions. To in- vestigate word class-specific processes, a good option ap- pears to be a comparison of two psycholinguistically similar stimulus classes while the experimental task is kept constant in conditions 1 and 2. To reduce the risk of obtaining by- chance results with standard significance criteria (II), more risky predictions can be derived and tested. One way to do this is to predict interactions between topographical vari- ables and stimulus classes, rather than only more or less ac- tivity at a not-yet-specified locus. In the best case, condition 1 and condition 2 would induce quite similar cognitive pro- cesses, but condition 1 would induce a process not induced by 2, and, conversely, condition 2 would induce a specific process not induced by 1. Based on theoretical predictions, processing of stimuli of class 1 in the task chosen may then be assumed to activate a set A of cortical loci not activated by class 2, whereas stimuli of class 2 processed in the same task would be assumed to activate a different set B of areas not activated by 1. (Of course there may be additional areas C activated by both classes.) The brain areas activated by the two conditions or stimulus types would be distinct, and each set of areas would include loci not included in the other. This can be called a physiological double dissocia- tion. The prediction to be tested by analysis of variance would be that direct comparison of the two activity patterns leads to a significant interaction of the task variable with the topography variable. It is unlikely that such a prediction is being verified by chance in a neuroimaging experiment, in particular if the loci where differences are actually found have been specified before the experiment based on theo- retical considerations. The rationale underlying this is very similar to the logic used in neuropsychology, where double dissociations are taken as strong evidence for processing differences (Shallice 1988 1989), although the dependent measure is behavioral in neuropsychology, but physiologi- cal in psychophysiology. In summary, one perspective on overcoming some of the problems of a simple subtraction logic in neuroimaging ex- Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 265
periments is offered by a double dissociation approach to psychophysiology. In this approach, physiological signs in- duced by maximally similar tasks ��� or even patterns of brain activation caused by matched stimuli in the same task ��� are being compared, and the prediction would be that class 1 of stimuli activates cortical loci A more strongly than class 2, whereas class 2 induces stronger activity signs than class 1 at distinct loci B. With regard to the present discussion, classes 1 and 2 may represent different word categories ��� for example, action and visually-related words ��� and loci A and B would then be large sets of cortical areas ��� for ex- ample, motor versus visual cortices. 4.3. Word properties affecting brain processes Given that comparable stimulus materials are used in an imaging experiment on processing differences between word classes, the expectation would be that defined corti- cal areas ���light up��� when members of a given word class are being processed (see predictions 1���4). But what would ���comparable��� mean in this case? Behavioral studies in which response times and accuracies of responses were measured precisely have clearly shown that various proper- ties of stimuli influence information processing in the brain, and many of the results from behavioral studies could be confirmed by psychophysiological experiments. Imaging techniques with good spatial resolution have only been used for a few years and, therefore, many methodological studies on the influence of stimulus properties have not yet been performed using these techniques. When evaluating imaging studies of word processing, it is essential to keep in mind the stimulus properties for which behavioral and ear- lier psychophysiological studies have demonstrated strong effects on brain processes. Words can vary on various scales. The naive observation that long words are more difficult to read than short ones is paralleled in the observation that words of different length elicit different electrocortical responses measured in the EEG. This appears to be the case regardless of whether the items are presented acoustically (Woodward et al. 1990) or visually (Kaufman 1994). A second important factor influ- encing behavioral and physiological responses to words is whether they are common or exceptional. In contrast to pic- tures or real objects for which it is difficult to estimate whether they are frequently or rarely being perceived, the frequency of words can be exactly determined based on the evaluation of large corpora of spoken or written text. Word frequency is well known to have a strong influence on re- sponse times and accuracies of word processing (see, e.g., Bradley 1978 Mohr et al. 1996). In addition, word fre- quency has a strong influence on cortical potentials evoked by word presentation (Polich & Donchin 1988 Rugg 1990 Rugg & Doyle 1992). Because certain word classes exhibit enormous differences in word frequencies, this variable may affect the outcome of studies of word class-differences. For example, whereas most function words are in the high- est frequency range, only a small percentage of the content words can be found in this high range, and most content words are used only rarely. Thus, word frequency is a likely confounding factor of experimental results about differ- ences between word classes. Additional possible confounds of word category differ- ences are related to psychological processes induced by the stimuli. Some words are more arousing than others: The word ���spider��� may lead to much more pronounced brain activity in an arachnophobic patient compared to ���beetle,��� and normal individuals may exhibit similar differences in brain responses. That event-related potentials reliably dif- fer between more or less arousing words has been shown by numerous studies (Chapman et al. 1980 Johnston et al. 1986 Naumann et al. 1992 Williamson et al. 1991), and there is also evidence that a variable called ���valence,��� that is, the degree to which the stimulus is evaluated as positive or negative, can have an effect on event-related potentials. Therefore, there is some reason to believe that what has been called the ���affective meaning��� of words (Osgood et al. 1975) can influence the brain processes these stimuli in- duce. Stimulus matching for the variables��� valence and arousal therefore appears desirable ��� except, of course, if the role of these variables in word processing is the subject of the experiment. Another variable strongly affecting behavioral and phys- iological responses to word stimuli is the context in which they are being presented. There are different types of con- text effects. They can be elicited not only if words are pre- sented in well-formed or ill-formed sentences, but also when words are presented one by one. If a word occurs twice in the same experiment, event-related potentials are usually more positive-going for the second occurrence (see, e.g., Rugg 1985 Smith & Halgren 1987). The repetition ef- fect appears to be quite complex and can interact with other variables, for example word frequency (Rugg 1990). There- fore, if a physiological difference is observed between words of different frequencies that are repeatedly pre- sented in the same experiment, it cannot be decided to which variable the difference should be attributed. Context effects can also occur between different words that are semantically related (semantic priming). Presenta- tion of a prime word changes electrocortical signs of activ- ity elicited by a subsequently presented target that is se- mantically related to the prime (Holcomb & Neville 1990 Nobre & McCarthy 1994 Rugg 1985). Similar priming ef- fects may also occur when a word is being presented in sen- tence context. A pronounced negative deflection is seen when meaningful words appear at the end of a sentence where they are highly uncommon (Kutas & Hillyard 1980a), and different brain waves have been identified that may indicate different forms of syntactic or semantic viola- tions (Neville et al. 1991 Osterhout & Holcomb 1992). Al- though there are several different effects of sentence con- text on word-evoked potentials, at least one of these effects appears to be quite similar to the effect induced by seman- tic priming (Van Petten 1993). Most importantly, context effects are not necessarily the same for all word classes (Besson et al. 1992). As mentioned above for the effects of word frequency and word repetition, sentence context ef- fects may vary between word classes as well. Event-related potentials elicited by content words are attenuated by a sen- tence context, provided that semantic and syntactic restric- tions are met by the sentence. In contrast, function words also show attenuation of event-related potentials when pre- sented in semantically deviant strings that still preserve some basic sentence-like structure (Van Petten & Kutas 1991). If words are presented in sentences or in sentence- like word strings, it may well be that not only the effect of a stimulus word is seen in the neurophysiological response, but a complex blend of the effects of the critical word, its preceding words, and their semantic and syntactic rela- Pulverm��ller: Brain���s language 266 BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2
tions. The various context effects may therefore either arti- ficially produce word class-difference, or they may mask real processing differences between word classes. When brain processes distinguishing between word classes are investigated, it appears necessary to keep in mind these effects of word length, word frequency, emo- tional (arousal and valence) properties of the stimuli, as well as those of word repetition, priming, and syntactic and se- mantic sentence context. These properties of word stimuli and strings may confound results of any imaging study in- vestigating differences in brain activity evoked by two word groups. Only if such confounds are excluded can a strong conclusion on differences between lexical or semantic word categories be drawn.9 5. Brain activity during word processing: Where? In this section, studies on the cortical areas activated dur- ing word processing will be discussed. The main question will be whether there is evidence for or against predictions 1���4. Studies on differences between content and function words will be dealt with in section 5.1, and section 5.2 will be concerned with action and visually-related words and re- lated categories. 5.1. Content and function words Neuropsychological work clearly indicates that different brain areas are necessary for processing content and func- tion words. Whereas aphasic patients with anomia have dif- ficulty finding content words (Benson 1979), for patients with agrammatic aphasia function words are more difficult to produce (Caramazza & Berndt 1985 Pick 1913). In ad- dition, aspects of agrammatics��� deficit in language compre- hension can be explained based on the assumption that they have a selective deficit in processing these lexical items (Pulverm��ller 1995a). Lesions within the entirety of the perisylvian region can be the cause of the agrammatic lan- guage disturbance (Vanier & Caplan 1990). In contrast, le- sions at various cortical sites outside left-hemispheric peri- sylvian cortices can lead to selective impairment in using or comprehending word categories included in the content word vocabulary (see the discussion in sect. 5.2). If function word representations are assumed to be restricted to peri- sylvian cortices (see Fig. 3), and content word representa- tions are assumed to be more widely distributed (see ex- amples in Fig. 5), a perisylvian lesion will destroy a large percentage of neurons included in function word repre- sentations, but will only remove a smaller part of the rep- resentations of content words. In contrast, lesions outside the perisylvian region will only affect representations of content words. Thus, different cortical distributions of cell assemblies representing content and function words can account for the double dissociation in processing content and function words in specific aphasic impairments such as agrammatism and anomia (Pulverm��ller 1995a Pulver- m��ller & Preissl 1991). In addition, evidence from behavioral experiments in healthy individuals using lateralized tachistoscopic presen- tation have provided further support for processing differ- ences between content and function words. It is well known that words presented either in the left visual hemifield (and, thus, to the right hemisphere) or in the right visual hemi- field (to the left hemisphere) of right-handed individuals exhibit a processing advantage after presentation in the right visual field (���right visual field advantage��� see, e.g., Bradley 1978). In behavioral experiments, these effects can be quantified exactly in terms of response times and accu- racies. A frequently applied paradigm is lexical decision, where words and matched meaningless pseudowords are presented in random order and study participants have to indicate whether an item is a legal word or not. In lexical decision experiments, the ���right visual field advantage��� has been found to be stronger for function words compared to content words matched for word frequency and length (Chiarello & Nuding 1987 Mohr et al. 1994b). For func- tion words, direct stimulation of the left hemisphere leads to faster or more accurate responses compared to stimula- tion of the right hemisphere. This is consistent with the idea that cell assemblies representing function words are strongly lateralized to the left (sect. 3.3.1). The weaker or even absent right visual field advantage for content words supports the idea that cell assemblies underlying content word processing are less lateralized (Mohr et al. 1994b). Several studies investigating event-related potentials (ERPs) have been conducted in search of differential brain activity induced by content and function words. Garnsey���s (1985) early experiment revealed a fine-grained word class- difference in event-related potentials uncovered by princi- pal component analysis. Neville et al. (1992) presented con- tent and function words in sentence context and had subjects indicate whether the sentences made sense or not. Words of the two classes were not matched for word length or frequency. These authors reported a left-lateralized component evoked by function words which peaked at 280 msec after stimulus onset, whereas a peak more symmetri- cal over the hemispheres was evoked by content words at 350 msec. A similar result was obtained by Nobre and Mc- Carthy (1994), who used stimuli matched for word length but not for word frequency. These authors presented words one by one and their subjects studied the sequence while trying to detect words of a particular semantic class. Again, a left-lateralized negative peak followed function word pre- sentation (latency: 288 msec), whereas content words led to an enhanced negativity (latency: 364 msec) that was more symmetrical over the hemispheres. Gevins et al. (1995) used a cued two-stimulus paradigm and asked subjects to indicate whether two stimuli were similar according to phonological, syntactic, or semantic criteria. These authors reported a lateralized positivity (latency: 445 msec) elicited by function words which was most pronounced over left frontal regions, whereas content words failed to elicit a late lateralized component. These authors did not report stim- ulus lengths or frequencies, however, and it is therefore not possible to exclude the most likely confounds. In an exper- iment comparing brain responses to content and function words matched for word frequency and word length (Pul- verm��ller et al. 1995a) while study participants had to make speeded lexical decisions, a negative-going wave that peaked around 160 msec after the onset of visual stimuli re- vealed a significant interaction of the word class and hemi- sphere factors. The peak in the event-related potential was equally visible over both hemispheres after presentation of content words, but it was pronounced over the left hemi- sphere and reduced over the right when function words were processed. Mean event-related potentials obtained between 150 and 300 msec after stimulus onset also re- Pulverm��ller: Brain���s language BEHAVIORAL AND BRAIN SCIENCES (1999) 22:2 267