the modular and formal approach of the previous generation of linguists and cog- nitive scientists and by the development toward applying the central insights of Cognitive Linguistics to more and more areas. From an internalist perspective, concerned with everything that goes on inside the cognizing subject, this broad use of the term makes it natural to understand most forms of response and processing as coming under the label ‘‘cognition.’’ Lakoff and Johnson (1999: 17) even see the amoeba as a ‘‘categorizing’’ entity, be- cause it distinguishes between two sets of objects by either moving toward them or away from them. However, a continuity that generalizes mental phenomena ‘‘downwards’’ to include mechanisms that other authors see as clearly nonmental (e.g., Johnson-Laird 1988: 24) raises a philosophical issue with those natural sci- entists who would like to generalize ‘‘upwards’’ from purely physical, neurological reactions in order to eliminate all reference to mental phenomena from scientific description. If we generalize from the physical level, the ontological dilemma of ‘‘reductionism’’ arises (several kinds of apparently different objects are ‘‘reduced’’ to one kind in the theory). Conversely, if we attribute mental life to objects that do not obviously manifest it, the possibility of using Occam’s razor arises, pruning away unnecessary additions to the world picture. Either way, there is a need to clarify the ontology of the object domain. From an externalist perspective, the strategy of emphasizing continuity means that all sorts of human experience—from bodily movement via everyday interac- tion to religion, ideology, and politics (see Dirven, Hawkins, and Sandikcioglu 2000; Dirven, Frank, and Ilie 2001)—are part of the domain of cognition and conceptualization. The continuity between conceptualizations as they operate in- side and outside the individual head is stressed, for example, by Gibbs (1999). Notwithstanding the virtues offered by this all-encompassing approach, it means that the realm of cognitive objects comes to be somewhat heterogeneous. I will try to show where this may become problematic, beginning with the descriptive prac- tice that takes its point of departure in the individual. The emphasis within Cognitive Linguistics on bodily grounding (see below) means that the internalist perspective is generally given priority as a source of explanation. In this, there is a continuity with first-generation cognitive science. Lakoff and Johnson (1999: chapter 2) outline a picture of cognitive architecture which accords with a rule of thumb among cognitive scientists stating that 95%of all thought is inaccessible to consciousness: ‘‘Our unconscious conceptual knowl- edge functions like a ‘hidden hand’ that shapes how we conceptualize all aspects of our experience’’ (1999: 13). Chomsky is generously said to ‘‘deserve enormous credit’’ (Lakoff and Johnson 1999: 472) for bringing into linguistics the notion of unconscious cognitive structures. In this picture, the explanatory status of ‘‘tacit knowledge’’ thus survived from Chomsky to Cognitive Linguistics, while the spec- ification of what it consists of radically changed. The emphasis on an object that belongs in the mind of the individual with- out being directly accessible raises a philosophical dilemma, which is taken up by Geeraerts (1999). Geeraerts’s own background for raising the issue is his work on 1250 peter harder diachronic semantics (see, e.g., Geeraerts 1997), which shows how empirical pro- cesses of lexical change can be accounted for with the tools of Cognitive Linguistics. He presents the dilemma in the form of a classical dialogue that brings out in counterpoint the contrasting difficulties of the idealist as opposed to the empiri- cal orientation. Although the idealist position is illustrated with quotations from Wierzbicka rather than ‘‘mainstream’’ Cognitive Linguistics, the reliance on in- tuitive analyses is similar enough to make the dilemma relevant. The clear-cut idealist position is criticized because too firm a reliance on intuitively accessible concepts is liable to seal off the results of the investigation from criticism—in effect not living up to the empirical commitments of scientific activity. Wierzbicka’s project derives from Leibniz’s idea of ‘‘the alphabet of human thought’’ and could naturally be described as rationalist in orientation (although in terms of a ratio- nalism quite different from Chomsky’s). The orientation toward conceptualization as taking place in processes of lan- guage use has not been equally emphasized, although in principle the commit- ment to language as part of human interactive experience has never been in doubt. Barlow and Kemmer (2000) explicitly set out to place Cognitive Linguistics in a broader picture of linguistic models that share the goal of basing linguistic de- scription on usage. This brings us back to the externalist perspective and the problem of the heterogeneous domain of cognitive phenomena. Unlike an item in one’s personal mental dictionary, the process whereby naming practices are shaped in a speech community over time does not belong in the internal arena of the cognitive un- conscious of an individual. In a philosophical context, the question is what follows from the fact that the object of description is not exactly the same when one is investigating diachronic prototype semantics as when investigating the hidden hand of ‘‘backstage cognition.’’ The difference in ontological commitments between the two approaches, as pointed out in Geeraerts (1999), cannot be straightforwardly cashed out in terms of decidable empirical issues. Conceptualization is clearly in- volved in both phenomena—but if it is too easily assumed that the same kind of thing is going on in both cases, the external, historical, and social factors con- tributing to the shaping of concepts are at risk of becoming invisible, overshadowed by the emphasis on the role of the individual human body/brain/mind viewed as an integrated system. This issue has also been addressed in a number of contexts by Chris Sinha, whose psychological perspective has reflected an orientation toward the dependence of cognitive structures on cultural and discursive processes (see Sinha 1988, 1999). The polysemy issue raised in the previous section involved two perspectives on the same object, namely, semantic networks in the mind. The present issue is more complex, because it involves two sets of related but different objects. The need for clarification of the interaction between the internal and the external domain of cognitively constituted phenomena should not lead to a classical form of dualism. A reversal to Descartes’s division into separate realms of body and mind would both be against the scientific world picture and against the idea of the embodied mind. cognitive linguistics and philosophy 1251 Exactly how best to avoid this, however, is not a settled issue. Both cognitive science and the philosophy of mind are still faced with the question of how to address the basic problem of accommodating mental phenomena in a nondualist world picture. There are positions available that suggest directions one might take. With the scope of phenomena it wants to accommodate, Cognitive Linguistics finds itself somewhere in the same landscape as the biologically based philosophy of mind and cognition developed by Maturana and his associates (see Maturana and Varela 1980). In it, organism and environment are understood to be mutually determined: the organism can only be understood as an aspect of the larger system it interacts with, and the environment is created by the way the organism itself responds to the ‘‘medium’’ in which it finds itself. This approach stresses the closure that comes about around the fused spheres of inner and outer processes, the so-called au- topoiesis, in which the events that happen and the state of the organism itself are in a form of equilibrium that can only be understood from within (see also Varela, Thompson, and Rosch 1991). This approach has influenced cognitive science in- dependently of Cognitive Linguistics (see Maturana 1975; Winograd and Flores 1986) but is congenial in its broad view of cognition and in emphasizing the embodied nature also of environmentally shaped processes. In continuation of this perspective, Clark (1997) discusses the implications of the human ability to extend cognitive processes into the external domain, including as a paradigm case the dependence on language as a public medium. As a conse- quence of Clark’s perspective, an element of distance is interposed between body and mind, that is, between bodily processes including those of the brain and the mind as a property of the human subject. While this perspective emphasizes the embodied status of action and cognition, it thus renders problematic the extent to which the body ‘‘in the flesh’’ (see Lakoff and Johnson 1999) remains the seat of all forms of mental content. In his final chapter, Clark emphasizes the difference between a person’s mind (which includes his links with external objects) and the brain as a bodily organ. Minds are to some extent socially constituted. One clear implication of the argument above is methodological. To the extent that the processes giving shape to meaning and conceptualization occur in the social sphere rather than in the individual mind, the methods of the social sciences need to be taken into account. One challenge is how to include methods addressing social variability in accounting for conceptual distinctions as part of language use (see Geeraerts 2003). The new theoretical platform proposed by Croft, including the diachronic (2000) as well as the synchronic (2001) dimensions, draws on founda- tional elements that have been extended from natural science (biology) to the social side of language. It will be interesting to see how the understanding of cognition, and of the appropriate way of handling it, will develop in the expanding universe of Cognitive Linguistics. 1252 peter harder 5. The Challenge from Cognitive Linguistics to Philosophy The nature of mental entities, however, is not only a philosophical problem—it is also a problem for philosophy. The new light thrown upon conceptualization by results from Cognitive Linguistics also puts the ball in the other court: how should philosophy respond? This point was raised by Johnson (1992) and followed up in Lakoff and Johnson (1999). Johnson claims that if Cognitive Linguistics is to be worthy of its name, then its results with respect to how the mind works must have implications for philosophical positions on how the mind relates to reality. By claiming that mental function is grounded in the body and imaginatively structured in a way that reflects specifically human experience, Cognitive Linguistics has de- fined a new position in the philosophical landscape. This position contrasts, on the one hand, with a belief in absolute objective foundations of knowledge and, on the other hand, with a deconstructionist rejection of any kind of foundation whatsoever. Of special philosophical interest is the way in which the bodily basis, with force dynamics as a key example (see Talmy 1988), affects mental domains such as knowledge and reason. In the case of reasoning, we understand the force of physical compulsion as a source domain recruited to conceptualize the force of a logical argument. A similar example is the relationship between knowing and seeing, where the perceptual term ‘‘seeing’’ is recruited to serve as an indicator of the cognitive relationship that is the result of seeing (as when you explain something and then ask, You see?) (see Sweetser 1990). Among other philosophical key areas are the concepts of subjective versus objective, of selfhood, of social relations, and of ethics, all of which are illuminated by being seen in the context of their anchoring in embodied experience. This manifesto was followed by a discussion in Cognitive Linguistics (Gorayska 1993; Johnson 1993; McLure 1993; Sinha 1993), offering different views on exactly what kind of force the findings of Cognitive Linguistics can rightly be claimed to have with respect to philosophy. One possible interpretation of the article is that Cognitive Linguistics has proven Philosophy (with a capital P) wrong, an inter- pretation Johnson (1993: 69) denies, while admitting that his title is perhaps mis- leading in that perspective. Sinha (1993: 53) offers a weaker alternative, namely that Cognitive Linguistics, like any new theory, raises the issue of how its results should be interpreted philosophically, but Johnson’s reply suggests that this may not go far enough; results of Cognitive Linguistics are repeatedly referred to as challeng- ing established philosophical positions, although the exact implications are still to ‘‘emerge dialectically’’ from reflections on the philosophical dimensions of Cog- nitive Linguistics. From a slightly different perspective, the issue is whether Cog- nitive Linguistics is going to provide positive answers to some of the philosophical questions for which Johnson sees these answers as relevant. At this point, Sinha and Johnson seem to find each other in their belief that it is simply too early to tell. cognitive linguistics and philosophy 1253 In Philosophy in the Flesh (Lakoff and Johnson 1999), the themes of Johnson’s article are fully developed in the form of a comprehensive set of claims concerning the consequences that results obtained within Cognitive Linguistics should have for philosophy. The chief targets are, as before, abstract disembodied reason cou- pled with objective facts, on the one hand, and total relativism, on the other. Ac- cording to the most radical claim expressed in the book (Lakoff and Johnson 1999: 14–15), traditional philosophical thought, as we know it, can and should be set aside on the basis of empirical results based on embodied cognition and achieved within Cognitive Linguistics. If we look at this strain in the book, it inscribes itself in the twentieth-century tradition of science-based denunciations of speculative preem- pirical thinking (see section 2 above). As I will argue below, this is problematic for methodological reasons (see section 6). This somewhat hard-hitting level of argumentation, however, is interwo- ven with a rather more cautious strain. In that respect, the point of the book is to show how the nonconventional forms of conceptual organization that cognitive linguists have been pursuing may be used to enrich philosophical thinking about basic issues in human life and thought. This ‘‘weaker’’ stance can be seen as a con- tinuation of the more constructive view on the issue, on whose relevance Johnson (1993) and Sinha (1993) agreed, but which they found it too early to explore thor- oughly. Among the specific problems that belong here is the question of exactly how much of the philosophical understanding of concepts like time and causation can be ascribed to literal content and how much is dependent on metaphorical enrichment recruited from other domains. After Philosophy in the Flesh, a renewed philosophical discussion of Lakoff and Johnson’s position arose in Cognitive Linguistics; and again, it proved difficult to achieve consensus on the basic issues. Rakova (2002) argues that an uncompro- mising belief in the ‘‘experiential’’ or ‘‘embodied’’ status of concepts can only result in extreme empiricism, or loss of powers of abstraction (including the kind that is involved in logical thinking). In their reply, Lakoff and Johnson (2002) point out that the undesirable consequences only follow if traditional assumptions of ana- lytic philosophy are taken for granted. In their view, Rakova’s criticism is invalid if the alternative premises are adopted, according to which (i) the embodied mind transcends a rigid dichotomy between empiricism and rationalism, (ii) embodied conceptualizations can be more or less adequate in relation to a body of empirical data, and (iii) abstraction arises naturally from more concretely embodied con- ceptualizations. The disagreement reflects that there exist two readings of the book. On the strong reading, where Cognitive Linguistic positions are put where philosophy used to be, the issues Rakova raises are valid criticisms. On the weaker reading, where Cognitive Linguistics does not replace but rather grounds disembodied philo- sophical reasoning, Lakoff and Johnson are justified in saying that this poses no threat to the concerns Rakova raises. Either way, these are issues in need of clari- fication, as stressed by Krzeszowski (2002) and Sinha (2002). 1254 peter harder 6. The Hermeneutic Perspective: The Cognitive Unconscious versus the Content of Understanding There is one important reason why it is problematic to determine the precise implications of embodiment for the philosophy of mind. The reason lies in the distinction, more significant for philosophical than for linguistic purposes, between conceptualizations as part of the equipment that we carry around in our heads and conceptualizations that we form in the process of understanding something, such as the nature of the world (or the content of a text, see below). Cognitive Linguistics naturally takes its point of departure in the mental equipment, whereas philosophy begins with the task of understanding the world. Beginning at the cognitive lin- guistic end, I will try to show where the path toward the philosophical end raises philosophical issues with implications for linguistics as well. When it comes to mental equipment, a central issue concerns the ontological commitment to the cognitive unconscious, which was carried over from Chomsky but was provided with a different content (see section 2). In spite of the new content, it raises the same methodological problem of access that affected Chomskyan com- petence. A similar problem bears upon the simulation paradigm in cognitive sci- ence in general: however great an achievement it is to create a simulation that works, the simulation in itself does not prove that this is the way things actually work in the simulated object. The discussion in Lakoff and Johnson (1999: 38–42) with regard to successful simulations within the ‘‘Neural Theory of Language’’ project also illustrate the relevance of this issue in a cognitive linguistic context. On general methodological principles, such models must be regarded as attractive sce- narios rather than definitive truths. Only data from the neural system itself (see Lakoff 2003) bear directly on the issue of what constitutes the neural reality un- derlying language. Searle (1992: 152; 1995: 128) has presented a general epistemological argument against totally tacit knowledge. His central point is that we do not know what mental states such as knowledge are, if we set them apart from consciousness— simply because there is no access to mental states (as opposed to physical states) except via subjective consciousness. Searle thus does not want to rule out mental states that, for a variety of possible reasons, happen to be unconscious; the problem arises if we simultaneously claim that these are mental states and that we can never be conscious of them. Chomskyan tacit structures are suspect because they transplant properties that essentially belong to the scientific metalanguage into the brain. Conversely, one might object that attributing conceptualizations from Cognitive Linguistics to a tacit level is problematic for the opposite reason: in doing so, we locate at an inaccessible level of knowledge the kind of phenomena we might intuitively very well have conscious awareness of. Among the things we do that are said to be cognitive linguistics and philosophy 1255 ‘‘inaccessible to conscious awareness and control’’ are ‘‘anticipating where the conversation is going’’ and ‘‘planning what to say in response’’ (Lakoff and Johnson 1999: 10–11). It would be natural to suggest that such processes are some- times conscious and that sometimes they are not. Also, such processes are clearly part of the process of coping with cognitive tasks with very direct ties with con- scious mental content. Anticipation and planning lend themselves to direct comparison with actual outcomes: if things work out differently from what we had planned and anticipated, that is something we tend to be aware of. This being so, however, it is not clear exactly how much hidden determination is to be attributed to the structures that Lakoff and Johnson attribute to the unconscious mind, as compared with the less covert kind of influence that is associated with (more or less) conscious processes of construal and interpretation. I think it would be fair to say that the conceptualizations uncovered by Cog- nitive Linguistics are generally assumed to be unproblematically applicable across the domain affected by the distinction suggested above: they are both part of the equipment we draw on in trying to understand and the result of processes of un- derstanding. Such flexibility, however, means that there are certain claims with respect to those structures that become problematic. Above (section 5), we saw how Lakoff and Johnson (1999: 14–15) invoked the science-based pattern of thinking that seeks to prove traditional thinking wrong by pointing to empirical findings. If conceptualizations are bound up with ongoing processes of understanding, how- ever, it is not clear whether the scientific foundations of the account are strong enough to make that claim compelling. If we look at the way in which Cognitive Linguistics has distanced itself from Generative Grammar, it has moved toward a position that is in many respects closer to that of the humanistic tradition than of modern science. The fact that Cognitive Linguistics is moving into territory associated with the humanities is also reflected in the growing commonality of interest between Cognitive Linguistics and the theory of literature (see, e.g., Stjernfelt 1995; Turner 1995; Freeman, this volume, chapter 45), which is perhaps the domain in the humanities that is most remote from scientific methodology and where hermeneutic practices still reign supreme. This has implications for the philosophical perspective within which Cognitive Linguistics must situate itself and which includes the hermeneutic per- spective, as pointed out repeatedly by Geeraerts (1992, 1993, 1999). Cognitive Lin- guistics has in effect opened up the territory that was consigned to oblivion by the science police and their allies, including Chomsky. That is essentially the territory that was allocated to the humanities by Dilthey in the nineteenth century when the split between science and humanities was becoming unbridgeable, with the area of Verstehen being opposed to that of Erkl € aarung (see Geeraerts 1999: 184). One implication of this is the need to reserve a place in Cognitive Linguistics for the process of interpretation, that is, of assigning (additional) meaning to input in or- der to understand it (more fully). In this context, it is relevant that Cognitive Linguistics does not place great emphasis on the distinction between coded meaning and utterance meaning. While 1256 peter harder this may—in contrast to a ‘‘wastebasket’’ approach to the role of context—be beneficial, it bypasses the question of what the actual processes are whereby com- plex linguistic utterances are assigned meaning in actual interaction. The tradition of Dilthey and Gadamer, with its emphasis on the historical life of conceptuali- zations and on how actual empirical processes of understanding work, is a nec- essary dimension in the landscape in which Cognitive Linguistics belongs. That is not to suggest that the continuity between conceptual structures viewed as re- sources and conceptual structures viewed as constituting the content of under- standing should be rejected, but we need to recognize the two different jobs they are doing. For example, to succeed in understanding something one has never understood before is not the same thing as calling up a preexisting understanding from one’s subconscious mind. Moving into the area of processes of understanding therefore raises issues that are different from asking about the grounding of ex- isting concepts. Ending up in the neighborhood of the humanities is not a regrettable conse- quence of rejecting Chomskyan formal linguistics (as some might think). As pointed out by Itkonen (1978), in the absence of the experimental verification that under- pins physical models, Chomskyan concepts, too, must be understood as humanistic interpretations of linguistic data and must be judged on that basis. While this is inconvenient for the generative ambition of rising above the vagaries of the hu- manities, there is less reason for Cognitive Linguistics to shy away from being in the company of the humanities; it just needs to be considered what that implies for the conception of cognitive science. The social sciences, in comparison, have lived ex- plicitly with the issue of how to incorporate the dimension of Verstehen since Max Weber (see Weber [1925] 1972). Seeing conceptual analysis in the context of the humanities also opens up the issue of what the implications are of language being a sign system in addition to being a cognitive accomplishment. Linguistic meanings are meanings only because they are tied to an expression, which functions as the vehicle. Again, this is con- genial with Cognitive Linguistics; the symbolic nature of grammar is one of the foundational assumptions of Cognitive Grammar (see Langacker 1987). In terms of the cognitive perspective, however, both signans and signatum are cognitive entities, and the Saussurean insistence on the independence of entities within a sign system from the entities denoted by signs is clearly alien to Cognitive Linguistics (cf. the ‘‘encyclopedic’’ view of meaning). The type of sign that is most obviously congenial with Cognitive Linguistics is perhaps Peirce’s ‘‘icon,’’ since the grounding of signs in experienced reality provides them with an iconic basis. However, in general, the results of Cognitive Linguistics are clearly central also from a semiotic perspective, since the formation and organization of complex signs draw on the whole area of mechanisms of meaning creation that is explored in Cognitive Linguistics from construal to blending. (This relationship has been explored, e.g., in work at the Aarhus Centre for Semiotic Studies, see Brandt 2004). Saussure envisaged a dis- cipline of semiology of which linguistics would only be one part, where all other sign systems would be accommodated as well. The two enterprises are similar in cognitive linguistics and philosophy 1257 terms of breadth of scope; but exactly what the relation would be between such a broad-ranging semiology and the rival umbrella discipline of cognitive science will only become clear once the interface between social processes of understanding and cognitive systems has been satisfactorily mapped. 7. Validity: Grounding, Abstraction, and Deconstruction Above we started with conceptualization seen as mental equipment, from the point of view of Cognitive Linguistics and cognitive science. I will now turn around and view the issue from the philosophical end, that is, that of conceptualization as a way of arriving at true knowledge of the world. This reversal of perspective is necessary in order to understand how matters stand when Cognitive Linguistics is viewed not from its own home ground, but as a potential contribution to a philosophical agenda. That is where the question of validity becomes central. The interesting situation here is that of the human mind at work, trying to un- derstand something. Understanding involvesrecruiting a conceptual model to apply it to the case at hand. But after a conceptual model has been found, two additional questions arise: (i) What follows if I choose one conceptualization rather than an- other? and (ii) How do I know which way of understanding the case at hand is the best one? Cognitive Linguistics has its main strength in exploring the models in them- selves, including the mappings that link up different models. It has less to say about what follows from using them in particular cases—aside from those consequences that can be read directly off the models in themselves. In the case of metaphorical mappings, the difference between seeing marriage as a journey and seeing it as a desert island is transparent as far as the conceptualization in itself goes. However, it is not clear exactly what actually happens when the model is applied to the object one is trying to understand. To tackle a very basic aspect of conceptualization as viewed in Cognitive Linguistics, the problem can be illustrated by the procedure of tracing metaphors back into the source domain of bodily experience. In the most radical interpretation, pointing out what model lies behind a given conceptuali- zation of the target domain would mean that the source domain was the ultimate content of understanding of the thing conceptualized (and I assume nobody be- lieves that). Force as applying to arguments is not the same as force used in physical compulsion—otherwise there would be no difference between reasoning about a matter and fighting it out. Lakoff (1993: 216) expresses this by the ‘‘target domain override’’ principle that rules out metaphors destroying the inherent structure of the target. But if a target overrides whatever does not fit, it remains open exactly what follows from applying a given conceptualization to that target. 1258 peter harder Davidson (1978) argues that metaphorical meaning is in fact too indeterminate to qualify as part of the description of language. The idea is that while one can be precise about nonmetaphorical meaning, one cannot be precise about a meta- phorical meaning until the metaphorical mapping is complete and the source domain meaning has become part of the target domain and hence ‘‘flattened’’ into a determinate, truth-conditional meaning. To take a familiar example: Going from London to New York is a journey in a determinate sense, but we do not know exactly what love is a journey means until we have completed the work of map- ping the metaphor onto the target domain. A primitive result of such a mapping would be something like ‘love is an experience where, after the relationship has come into being, you move through a succession of new situations, rather than remaining in the same initial position’—which is again determinate. Until the metaphorically transferred meaning has become literalized in its new domain, Davidson claims, it can mean so many different things that its readings are non- determinate. Against this, Collin and Engstrøm (2001) point out that metaphorical mean- ing as something both nonliteral and determinate can be defended when taking an explicitly process-oriented view of meaning. The whole process of recruiting met- aphorical meaning has a determinable content, even though not all of the po- tential of using the metaphor is determinable at a given stage. This theory is like Davidson’s in that it explicitly goes beyond understanding metaphor purely as a mapping—it insists that metaphor needs to be ‘‘brought to bear’’ and thus to make a difference in understanding in order to make sense. But it retains the emphasis on the mapping as a process rather than to look simply for the products of the mapping as if they might just as well have been literal all the way through. Once the implications of a conceptualization (metaphorical or not) have been determined, we face the question of whether it is a good way of understanding what we are dealing with. This problem reflects one of the really basic issues that radically divide the philosophical tradition from the position outlined by Lakoff and Johnson, namely the fear of ‘‘psychologism,’’ that is, of mistaking the nature of reality for what we take to be reality. Lakoff and Johnson (1999) take up this issue in relation to Frege, to whose ‘‘rabid antipsychologistic bent’’ (468) they trace the failure of analytic philosophy to consider the role of embodied meaning. What Frege and later analytic philosophers failed to realize, they say, was the capacity of embodied reasoning to give rise to shared meanings (440). However, the whole point of philosophy from Plato onwards (with the position of Hume as an ex- ception) has been to establish a way of talking that keeps clear of the traps that ordinary mental impressions and opinions are liable to fall into. Frege’s views were not the more or less accidental cause of this aberration in modern philosophy—it is the whole philosophical enterprise that is at stake. From the point of view of the philosophical tradition, shared and grounded meaning is therefore not enough. Truth is the traditional philosophical criterion, which in a usage-based perspective can be seen as one way of being adequate for the job a conceptualization is recruited to do—in this case, to provide an account of cognitive linguistics and philosophy 1259 . form in the process of understanding something, such as the nature of the world (or the content of a text, see below). Cognitive Linguistics naturally takes its point of departure in the mental. the social side of language. It will be interesting to see how the understanding of cognition, and of the appropriate way of handling it, will develop in the expanding universe of Cognitive Linguistics. 1252. the vehicle. Again, this is con- genial with Cognitive Linguistics; the symbolic nature of grammar is one of the foundational assumptions of Cognitive Grammar (see Langacker 1987). In terms of the