Парадигмалар назарияси ва унинг фанга таъсири Режа

Recent research on metaphor
This subchapter provides an overview of recent research on metaphor in dis-
course employing a critical cognitive approach. The studies commented on share
the view of metaphor as shaping the way we conceptualize reality, being bound to
a socio-historical context and being partly a cultural product.
The employment of metaphor has been extensively investigated in political dis-
course. George Lakoff (2004) claims that conservative and liberal political views
are based on two different models of a family: a strict father family and a nurtur-
ing parent family, respectively. This stems from the existence of conceptual meta-
phors: NATION IS A FAMILY, GOVERNMENT IS A PARENT and CITIZENS
ARE THE CHILDREN (Musolff 2004, 2). Paul Chilton focuses on the examina-
tion of the employment of metaphor in discourse during and after the Cold War,
pointing out that the political transition discourse following the Cold War was
marked by the metaphor of the COMMON EUROPEAN HOUSE (Chilton and
Ilyin 1993; Chilton 1996). Elena Semino and Michela Masci (1996), while investi-
gating metaphor in the political discourse of Silvio Berlusconi, reveal that Berlus-
coni employs mainly metaphors the source domains of which are FOOTBALL,
WAR and the BIBLE, in an attempt to create a positive image for himself and his
party, and to justify his political actions. The football metaphor takes advantage of
the positive connotations that football as a national sport has for Italians, aiming
to establish national unity. The war metaphor is chosen to establish Berlusconi
as a national leader that is capable of protecting his country, and the Bible meta-
phor is mainly drawn upon to appeal to the voters who are Catholic. Charteris-
Black (2005) studies persuasion in political speeches (such as speeches by Winston
Churchill, Martin Luther King, Jr. and George W. Bush) as performed by a choice
of metaphors in combination with other rhetorical devices.
Another major domain in which research on metaphor has been carried out is
newspaper discourse. Lule (2004) studied metaphors in newspaper articles pub-
lished in the six weeks before the second war in Iraq, finding out that among the
most common and recurring metaphor themes in the reports are: TIMETABLE, 30
Representations of Natural Catastrophes in Newspaper Discourse
GAMES OF SADDAM, PATIENCE OF THE WHITE HOUSE and SELLING THE PLAN. Lule emphasizes that these metaphors, both consciously and unconsciously, structured people’s experience and determined their actions. El Refaie (2001) examines the use of conceptual metaphors in the portrayal of asylum seekers. She points out that the major metaphor themes – WATER, CRIMINALS and an INVADING ARMY – are repeatedly employed in newspaper discourse using conventional lexis and entrenched grammatical patterns, which results in the naturalization of the metaphor themes. El Refaie emphasizes that it is the forms of language through which metaphors are realized that influence the degree to which the metaphor themes come to be viewed as a commonsensical representation of reality, with, for example, compounding playing an important role in the process of naturalization. Santa Ana (1999) reveals that American public discourse dehumanizes immigrants by drawing upon the metaphor IMMIGRANTS ARE ANIMALS, which becomes naturalized as it is routinely employed without drawing attention to itself. The last study to be mentioned is Koller’s (2004) analysis of metaphors describing women managers in business magazines. She reveals that the major metaphor theme employed is the WAR metaphor, which reproduces the dominant paradigm of hegemonic masculinity.
Such studies reveal, on the analysis of concrete linguistic material, the power of
metaphor to shape our conceptualization of certain aspects of reality by construct-
ing a naturalized portrayal that conveys particular ideologies.
What is the Role of Conceptual Analysis in Cognitive Science? Liam C. Kavanagh* ([e-mail]) Christopher L. Suhler* ([e-mail]) Department of Psychology, UC San Diego Department of Philosophy, UC San Diego 9500 Gilman Drive La Jolla, CA 9500 Gilman Drive La Jolla, CA * Both authors contributed equally to this paper. Abstract Cognitive scientists sometimes find themselves embroiled in debates over the precise definitions of high-level concepts in their fields – COGNITION, EMOTION, SENSE, and so on. The idea behind these debates seems to be that achieving a precise definition of these concepts will be a boon to scientific inquiry. We argue that these efforts of conceptual analysis would benefit from greater appreciation of the importance of such high-level concepts in supporting association or semantic priming, as opposed to deduction. In this associative role, they provide the basis for making connections between related concepts, connections that can then be explored by empirical methods, which in turn yield more precise, but often quite novel, concepts. In combination with well-established work in cognitive psychology on the non-classical structure of natural concepts, this perspective suggests that researchers should be cautious about investing substantial time and energy in attempts to precisely define concepts like COGNITION. Keywords: concepts & categories; philosophical issues; philosophy of science Introduction and Background The investigation of concepts has been a central part of intellectual inquiry since at least the time of Socrates. Today, conceptual analysis remains a cornerstone of academic philosophy. Cognitive psychology, too, has taken a keen interest in the meaning of concepts; inquiries here have often taken the approach of first investigating “simpler” and more tangible object concepts, in the hopes of working their way up to a grasp of more abstract concepts. Despite the substantial effort invested in precise accounts of concepts’ meanings, the results have been uneven and at times frustrating. In particular, practitioners of conceptual analysis often doggedly pursue efforts to arrive at precise definitions of high-level concepts, such as COGNITION, EMOTION, SENSE, and so on with less stellar results than are arrived by investigating lowerlevel concepts such as SMILE, SYSTEM, or CIRCUIT. We will make frequent use of COGNITION as an example of a high-level concept, but the same points apply, mutatis mutandis, to other high-level concepts. We also readily admit that the distinction between “high-level” and “low-level” concepts is, itself, not perfectly precise, but do propose the following as a general marker of the distinction: low-level concepts are those which tend to play greater roles in mechanistic accounts of empirical results, while highlevel concepts are those that do not (for example, we rarely, if ever, would explain some empirical result simply by saying that it is “cognitive”). We argue that these efforts of conceptual analysis would benefit from greater appreciation of the importance of such high-level concepts in supporting association or semantic priming, as opposed to deduction. In this associative role, they provide the basis for making connections between related concepts, connections that can then be explored by empirical methods, which in turn yield more precise, but often quite novel, concepts In combination with well-established work in cognitive psychology on the structure of natural concepts, this perspective suggests that researchers should be cautious about investing substantial time and energy in attempts to define concepts like COGNITION. The Allure of Precise Definitions There is something about precise conceptual definitions – definitions which lay out the characteristics that all members of a category must have – that continues to attract our curiosity. This is despite the fact that introductory cognitive psychology texts have for several decades stated plainly that natural concepts almost never take such a form. Persistent targets of definition-seeking include concepts such as COGNITION (e.g. Adams & Garrison, 2013), EMOTION (e.g. Mulligan & Scherer, 2012), and SENSE (Keeley, 2002). We believe that there are two major assumptions, often tacit, which together explain the continuing pull of precise conceptual definitions. The first assumption is that precise definitions are out there waiting to be found, if only we look (and think and argue) hard enough – or at least that there’s a good enough chance that such definitions are out there to make the pursuit worthwhile. The second assumption is that obtaining such precise definitions will be a boon to scientific inquiry and understanding . Both of these assumptions, we will argue, turn out to be dubious. This, in turn, casts doubt on whether rigorously delineating the set of referents of terms such as COGNITION is a good use of researchers’ time. Part of the appeal of precise definition is likely attributable to vestigial influence of the classical view 1057 of reasoning and concepts (Smith & Medin, 1981), according to which concepts (and their associated categories) are defined in terms of necessary and sufficient conditions. This was the dominant view of concepts for almost the entirety of western thought, with its origins often traced back to Aristotle (Smith, 1997). Concepts of this kind are the epitome of precision; a classical definition of a concept promises precise demarcation of the boundaries of that concept, and allows precise, deductive inferences. The Classical View of Concepts The familiar tale of the failure of the classical view of concepts hardly needs detailed recounting, and as such we will be exceedingly brief here. The key point is simply that empirical research has established that concepts usually do not have a unified definition (i.e., a set of features common to all and only members of the concept) or sharply demarcated boundaries – Mervis & Rosch, 1981). This historical record provides strong reason to think that any proposal of classical criteria for a given concept will fail to capture all of the phenomena that intuition demands. Thus, in the case at hand, it is likely that a definition of COGNITION will fail to capture all (and only) those phenomena that we regard as cognitive. Furthermore, even if it gets most of those right, it will fail to recognize degree, since on a classical view of concepts something is either part of the concept (i.e., possesses the necessary and sufficient features) or not (i.e., does not possess the necessary and sufficient features), with no gradations or gray areas. Debates over the precise criteria for COGNITION are therefore likely to lead to a great deal of ink being needlessly spilled trying to impose a structure that the actual mental concept likely doesn't have. A given definition will include some desirable things, and even where it gets the basic verdict right will fail to recognize gradations of typicality/atypicality. For instance, proposals which assign cognitive status to cellular information processing will be intuitively inadequate if they fail to recognize that these processes, despite possessing features associated with cognition, are not paradigmatically cognitive.1 The aforementioned considerations are of course not definitive proof that definitions of cognition are undesirable or impossible. It is not clear that there is any way to prove that definition-seeking for any specific natural concept will, as a matter of necessity, fail, despite the rejection of the classical view by 1 These general points also provide strong reason to doubt the attainability of satisfactory conceptual analyses in contemporary philosophy, a point developed in Suhler (in preparation). cognitive psychologists and the dubious record of attempts at classical conceptual analysis. We expect that some will continue to hold the intuition that a definition of COGNITION is possible, just as many philosophers hold the very strong intuition that classical analyses of their most cherished concepts are possible despite decades or even centuries of failure (Suhler, in preparation). Further, even if there is a (satisfactory) precise definition of cognition to be found, history strongly suggests it will be achieved because theoretical and empirical inquiry reveals a similar underlying nature of things we take to be cognitive; it will not, in other words, come about through standard methods of a priori conceptual analysis (reflection on our folk concepts, thought experiments, etc.). We will develop this point later in the paper by examining the history of conceptual evolution in physics. Before that, however, we want to look more specifically at why, despite the vanishing rarity of classically defined concepts, there exist strong intuitions that concepts like COGNITION admit of classical definition. Though intuitions are indeed often valuable things, they are not beyond criticism. The most effective response to the belief that we can precisely define concepts like COGNITION, we feel, is to explain how these intuitions in favor of pursuing precise definitions are driven by factors that do not track the actual probability that such definitions are attainable. There are two such factors we wish to highlight here. The first is saliency. The scientific concepts that are likely to be most salient qua scientific concepts tend to be those drawn from mature sciences dealing with relatively simple phenomena – viz., physics and physical chemistry (see below). The second is confirmation bias – the tendency to notice the examples that fit with one’s preexisting commitments rather than those that do not. In suggesting a definition of a particular concept, one will tend to notice the examples that fit it but be less attentive to those that don’t fit. Similarly, at the level of theories of concepts, adherents of the classical view are likely to seek out those examples that (perhaps) fit the classical structure to the neglect of those that do not – hence the ubiquity of the example of BACHELOR in philosophy papers as a natural concept with a putative classical definition. Such cognitive biases do not change the fact, however, that classical definitions are in fact vanishingly rare – all the more so for high-level concepts like COGNITION. Proposing and critiquing precise definitions of cognition is an appealing topic for academic discussion and publication, since there will always be something to say: extant proposals will always be subject to counterexamples – encompassing exemplars that they shouldn’t include and/or missing 1058 ones that they should. Such proposals and critiques are not, however, likely to be steps toward a fully adequate, precise definition of COGNITION. As discussed, efforts at conceptual analysis are probing mental concepts that likely have a non-classical structure and therefore are not amenable to the sort of precise demarcation being sought. As long as faith persists in science’s need for precise definition, critiques of particular definitions simply perpetuate this misguided process. Therefore we will focus on criticizing the underlying rationale at a process level. The Value of Associations A further legacy of classical view of reasoning is a heavy emphasis of the role of logic and deductive reasoning in thought. Recent work, especially from Bayesian perspectives, strongly contradicts such views, arguing that thought is more statistical than formal-logical (Chater, 2009). High-level concepts likely deviate most strikingly from classical structure because their role is to cope with a massive, and poorly understood, world by holding vast numbers of statistically related ideas together into relatively few clusters. Crucially, this associative role does not require – and indeed can be undermined by – precise, exhaustive specification of conceptual content. It is well known that exposure to a particular word or concept brings semantically related concepts sometimes including opposing concepts, into working memory. Thus subsequent thoughts are more likely to involve these related concepts. This priming role is important because our conceptual workspace is limited (e.g. Baars, 1997) and so there are finite number of concepts that have a chance, at any one time, to fit together into an explanation that will in turn fit with reality, producing a “eureka moment”. As such, explanations will only avail themselves to us if a good number of the needed components are “close to mind”. Useful high-level concepts bring to mind other concepts, such that as a whole the contents of the workspace are set up to yield explanations efficiently. Some authors go so far as to propose that a word’s meaning is its statistical relations with other words (Lund, & Burgess, 1997), but even authors such as Barsalou (2008), who see little role for abstract symbol manipulation in cognition, acknowledge this power of words. To see why this is of critical importance for science, consider an example from clinical psychology: what is meant when one asks whether a disorder such as depression is “psychological” or “biological”. We would submit that the questioner is basically wondering whether the explanation (or “the story”) of the disorder would be most usefully constructed using concepts that come to mind when one says “biology” such as HORMONES, and NEUROTRANSMITTERS, or couched in terms of DESIRES, OBSESSIONS, and RELATIONSHIPS, that come to mind when one says “psychology”. In this example, the cognitive utility of the concepts of PSYCHOLOGICAL and BIOLOGICAL lies not in their being explanatory themselves; rather, their value lies in their ability to organize groups of associated terms/concepts that may eventually figure into an explanation. We contend that this associative role of high-level concepts is, in the vast majority of cases, more important than their ability to support deduction, both in scientific and everyday inquiry. That is, these overarching concepts – what we are calling “high-level” concepts for short – are powerful drivers of association, but few play a direct role in mechanistic explanations; for example we seldom make strong, specific empirical claims about particular phenomena simply because “they are psychological/biological”. Though we do not deny that concepts whose form allows for broad deduction are useful, it is not the case, as mentioned, that natural concepts typically have such a structure, nor it is it the case that concepts must be forced into such a form in order to help us to generate explanations. Statistical knowledge must be relied upon in the early stages of investigation – when our investigatory approaches are based on hunches, necessarily entered into with highly imperfect knowledge of eventual results. For instance, EMBODIED COGNITION seems, at this stage, to be a concept that mainly serves to associate a number of disparate approaches motivated by the idea that an organism’s physical form and environment cannot be abstracted away usefully from questions about cognition (Ziemke, 2003); it is, in essence, the broad claim that satisfying explanations of core cognitive phenomena will nontrivially involve bodies and their environs. If we use high-level concepts to help us home in on explanatorily valuable associates, it’s tempting to suggest that we should strive for greater conceptual precision. However, rather than being dictated by an imposition of formal definitions, semantic associations are built up over time by repeated coactivation of concepts, as during reading, conversation, or inner speech. Reflecting this historydependent nature of association, individual researchers will, and should, tend organize their own workspace in a way that is suited to their particular problems. With the many kinds of cognitive science being done, the associations of cognition are bound to be promiscuous. We would simply add that this is not quite so undesirable as is often assumed. Still, it is true that we associate precise ideas with good science. The great exemplars of scientific concepts are those that emerged as precise categories 1059 (ATOM, MOLECULE, ENERGY, etc.). However, these concepts are precise because it is the nature of their referents to lend themselves to precision – they are putative “natural kinds”. These conceptual success stories, coming mainly from physics, were not arrived at by tirelessly interrogating and refining preexisting ancient folk-physical concepts. Rather, precision in physical concepts was achieved through tireless empirical inquiry, and the positing of theories to make sense of these empirical observations. Historical Evidence from Physics Like most special sciences, the origins of physics were in philosophy; more specifically, it evolved out of what was called natural philosophy from antiquity through much of the 19th century. Initially, rather than there being any field called physics, which precisely defined its explananda and proceeded to explain them, there were instead a number of interests in specific natural (as opposed to man-made) phenomena. It was gradually realized, however, that explanations of such phenomena as heat, magnetism, and light could be made in terms of similar processes and entities. In more recent times, physics has become defined as the study of matter and energy (and even this definition may become obsolete), entities that actually would be rather foreign to the ancients from whose work the modern field descends: the Greeks thought that water fire, air, and earth were separate elements, of which all things were admixtures, including minds. What the history of physics suggests is that fields of inquiry may discover what their explananda “actually are” as they move forward (Einstein & Infeld, 1961). The term “physics”, rather than timelessly referencing a clearly defined set of concepts, questions and tools all revealed through analysis of the folk concept PHYSICAL, carries constantly updated statistical information about what ideas, facts, and phenomena have, to date, been found to “go together”. What the ancients more likely had in common with us were very basic immediate sensory experiences, such as the sensations of light, sound and heat, of pressure, and of movement. These sensory experiences were what originally demanded explanation; however, these same entities have not ended up defining a coherent science. Rather, their study produced a cluster of explanatory concepts that continued to co-occur over and over again such that when one turned to be an important part of a particular story, another almost always did as well. At any point in time, then, an attempt at precisely defining the subject matter of the field of physics would have proved descriptive rather than prescriptive. Paradigmatic examples of precise concepts arose quite anew out of a process of discovery, and might even have been closed off from discovery if the content and nature of the field had been fixed more than two millennia ago in the name of trying to precisely define the concept PHYSICAL. Prospects for a Precise Definition of COGNITION The pursuit of the unknown by any research method must be based on a guess at its potential benefit. It is impossible to prove, for example, that experimental work on any given question will yield enlightening results, but we intuit that it will because of experimentation’s past record of success. In contrast, attempts to find, via conceptual analysis, precise definitions of concepts like EMOTION, SENSE, and COGNITION have so far not proven very successful. This may be because these concepts do not admit of precise definitions, or it may simply be that we have not achieved insights necessary for precise definitions. If precise definitions are “out there” then the experiences of other sciences give us the best means of guessing at how best to find them. Experience in physics (see above) shows that intuitively appealing and precise definitions, when discovered, are extremely useful. But it is also clear that these come after great empirical effort, and are unlikely to correspond neatly to the natural concepts that a science has at its outset. So it might be that cognitive scientific concepts such as INTENTION, while intuitively appealing and arguably as real, psychologically, as HEAT, will not be foundational to the field’s mature theories (for related discussion, see, e.g., Churchland, 1981; Thagard, 1990). It is instructive to relate the above ideas to a recent proposal that the “mark of the cognitive” consists of actions that have reasons (Adams & Garrison, 2013). Recent research shows that humans seem to have very stable tendencies to perceive minds (and reasons/intentions) without much prompting. Seeing an entity as having reasons requires that different neural networks are engaged, and we then think about that entity differently than when we do not see it as motivated by reasons (Epley & Waytz, 2009). Thus, while perceptions of reasons behind actions may not be as experientially primitive as experiences of (say) heat and pressure, they are quite hard to avoid. Nevertheless, despite the importance of our perception of minds and intentions in motivating interest in sciences that might help explain them, it is very possible that intentionality will end up taking a position in mature cognitive science is more analogous to that of heat, rather than that of energy in modern physics. For instance, recent influential proposals posit that human cognition is centrally concerned with explaining primary sensory data and 1060 this drives our neural activity, so that perceived minds, like objects, must be seen as part of our attempt at predicting the world (Friston & Frith, 2015). If so, then pursuit of deep principles of prediction, rather than of intentions, may eventually define our field. However, much like in physics, such sweeping redefinition will come after theories have proven their ability to explain empirical facts. Finally it is worth noting that some commentators have argued that, in general, biology and cognitive science tend to provide us with explanations not via general laws (Bechtel, 2008), but rather by positing specific mechanisms and their interactions. Given that the discovery of law-like relations has driven the emergence of our most precisely defined high-level concepts from physics and physical chemistry (e.g. ELEMENT), sciences that do not produce laws would seem especially unlikely candidates to produce precise high-level definitions. What Role for Conceptual Analysis in Science? The inadequacy of the classical view of concepts and the nature of progress in other sciences (esp. physics) provide reasons to doubt the first key assumption we identified, at the outset, as underlying interest in defining terms like COGNITION: that an adequate, precise definition can be achieved if we think and argue hard enough about it. The associative role of concepts, meanwhile, provides reason to doubt the second key assumption – that precisely defining cognition will be a boon to cognitive scientific inquiry – since association does not require precise, fixed definitions. Before elaborating on these points and why they cast doubt on the value of much conceptual analysis in cognitive science, it’s worth pausing to emphasize that clarity about the meaning of concepts does have its uses. In particular, concepts that describe observed phenomena need to be defined clearly enough to allow interpretation and synthesis of empirical results. Most cognitive scientists have had the experience of going to read up on a particular research topic only to find that the key concept around which that topic is organized – IMITATION, EMPATHY, etc. – is used in a wide variety of ways in different papers and by different research groups. As mentioned briefly above, this variety itself is not necessarily a bad thing, so long as the authors of a given study are clear about how they are using a given concept and how their usage relates to other common usages in the literature. But when the time comes to reconcile and synthesize these results – a function typically performed by review articles – a degree of conceptual clarity is a must. Few things more effectively undermine the utility of a review article than inattention to the different ways in which a key term/concept is used in the body of research supposedly being reviewed. When this occurs, the review becomes little more than a bibliography, with the responsibility of achieving a degree of reconciliation and synthesis having been abdicated. If conceptual clarity is important in these cases, then why not in the case of COGNITION? The reason, as already mentioned, is that unlike concepts such as IMITATION, EMPATHY, and ENERGY, high-level concepts like COGNITION and PHYSICAL do not figure very directly into investigations of specific phenomena. As with the concept PHYSICAL in physics, experiments and theories in cognitive science do not examine and explain cognition qua cognition; rather, they examine and explain more specific phenomena and processes. Consider, by way of elaboration, an analogy to biology. Biology is, literally, the study of life, but biological experiments and theories are almost entirely concerned with more specific questions that in some way or another connect to living things. A precise definition of the concept LIFE is not going to help molecular geneticists or evolutionary theorists do their jobs better, since the concepts and methods that they use in day-to-day inquiry within their subdisciplines are much more specific – and empirically grounded – than that (see, e.g., Crick, 1966). Suggestions of and debate over a precise definition of LIFE are mostly philosophical curiosities, orthogonal to the methods, theories, and concepts that actually animate the daily work of scientific research. It is unclear why the great progress that biological science has made without a precise definition of its eponymous concept should necessarily be denied to a cognitive science that lacks a precise definition of COGNITION. As we have emphasized throughout, the value of overarching disciplinary concepts like COGNITION and LIFE is likely to lie in their ability to organize mental workspaces and suggest connections between various lines of inquiry. Given this, the lack of sharp, pre-defined boundaries on what counts as cognition may actually help produce occasions for associations to be made and new lines of investigation to be opened. The potential restrictiveness of precisely fixing a definition of COGNITION can be seen by applying lessons from the history of physics to the history of cognitive science. As with physics in ancient times, a definition of COGNITION fixed in the early days of cognitive science (say, the late 1960s) would likely have centered upon logical symbol manipulation that goes on “inside the head”. Such a definition would have closed off even connectionism as relevant to cognition, not to mention frameworks of embodiment (Varela, Thompson & Rosch, 1991; 1061 Barsalou 2008), extended cognition (Clark, 1997), and metaphor (Lakoff & Johnson, 2008). These new frameworks expand the number of perspectives associated with COGNITION, thereby allowing further connections to be drawn and further lines of experimental inquiry to be opened up. The results of such inquiries, then, in turn, continue to modify our concept of COGNITION, and the process repeats itself. Our argument, then, is that the value of the concept of COGNITION does not lie in any fixed definition, for such a definition will inevitably be beholden to the state of knowledge in the field and to the empirical and theoretical fashions of the day. Rather, concepts like COGNITION evolve with the field(s) in which they are used, providing a basis for associations and connections that generate empirical inquiries, and with them further conceptual modification. If a precise definition of COGNITION is possible at all, it will be achieved through a long, messy process of grappling with empirical reality, not traditional methods of conceptual analysis. Conclusion Both the allure and dangers of precise conceptual definitions are likely to be especially acute when a scientific field is in its early stages of understanding – as cognitive science currently is. Clarity and precision are often frustratingly hard to come by, a problem compounded in the case of cognitive science by the sheer complexity of the phenomena under investigation. Precise definitions are appealing in no small part because they promise such clarity and precision. The danger, however, comes from the near certainty that any definitions we strongly commit to at such an early stage will misguide future inquiry by ensconcing in those definitions all the empirical and theoretical limitations of the time at which they are fixed. For instance, even if our investigations ultimately yield an picture of cognitive systems that is not meaningfully “embodied” or “extended”, the fact that we are so caught up with these ideas – ideas which would have been anathema to most cognitive scientists even 40 years ago – is surely a sign that we still have much to learn about what cognition is.
220 ___________________________________________________________________________ METHODS OF CONCEPTUAL ANALYSIS MILOŠ KOSTEREC, Katedra logiky a metodológie vied FiF UK, Bratislava, SR KOSTEREC, M.: Methods of Conceptual Analysis FILOZOFIA 71, 2016, No. 3, pp. 220-230 This paper describes some of the methods usually grouped under the label of conceptual analysis. It delineates and compares three such methods: constructive method, detection method, and reductive conceptual analysis. For each of these three kinds of conceptual analysis, the problems which motivate its use are specified and the wellknown instances of their application are discussed. Based on the general model of method as an ordered set of instructions, the three types of conceptual analysis differ in specifying the instructions involved in their use. Keywords: Conceptual analysis – Conceptual theory – Language – Method 1. Introduction. Analysis is a term of many uses.1 It is common in everyday speech and is often used by laypeople, but also by scientists and philosophers. The use of the term has a long history and hence an attempt to provide an exhaustive overview of its meaning would be limited. One of the forms of analysis is conceptual analysis (CA), the specifics of which are the subject of this paper. However, the semantic jungle behind the uses of CA is also quite thick. In modern philosophy, the rise of CA is connected with the names of G. E. Moore, Bertrand Russell, Gottlob Frege or Ludwig Wittgenstein These philosophers were certainly not the first to provide CA (see Earl 2005), nor were they the only ones to perform CA (see Beaney 2007), but they explicitly aimed to provide such analyses. The central role of CA in their work gave rise to an entire field in philosophy – the socalled analytic philosophy. In this paper, my aim is to specify the features of CA in its various forms. However, I do not intend to make normative claims; at best, the following study of CA is descriptive and methodological.2 My main claim is that there are several methods of CA and that all of them were formulated (or at least studied) under the label of CA. Below, I specify the differences between them and provide distilled, abstract models of these methods. I discuss three basic forms of CA: constructive, reductive and detection analysis. 3 1 This work was supported by the Slovak Research and Development Agency under the contract No. APVV-0149-12. I would like to thank my colleagues for helpful discussions. 2 For a good overview of philosophical methods, see (Dally 2010). 3 In other words, I do not view linguistic analysis as a method of CA. While the aim of linguistic analysis is to provide insight into how a term is used within a specific field or domain, the aim of conceptual analysis is to examine the place of a concept in the conceptual network of a language or a theory. For a methodological dissection of linguistic analysis, see (Nuopponen 2010a, 2010b). FILOZOFIA Roč. 71, 2016, č. 3 Filozofia 71, 3 221 2. A model of method. In the following, I presuppose the model of method developed by Bielik et al. in (Bielik et al. 2014a, b, c, d),4 which views method as a set of instructions that lead to a specific goal. On the basis of this model, any method which leads to a scientifically relevant goal can be considered scientific. Typically, the use of a method is motivated by some kind of a problem formulated on the basis of a certain theoretical and factual background. This background is subject to change in the application of the method – i.e., it is modified whenever new theoretical or factual knowledge is gained by following the instructions of the given method. Generally speaking, a problem is a question to which no ready-made answer is found in the background. The application of a method should transform the background so that it contains that answer, making the problem a non-problem. Proper modeling of this change of knowledge apparently presupposes the distinction between explicit and implicit knowledge.5 3. The methods of CA. There are various kinds of problems solvable by CA. In the following, I discuss three different methods of CA. All of them aim at gaining better knowledge of the language we use. Thus stated, this objective seems vague. What does acquiring knowledge of a language mean? Should not every competent speaker already know her language? But that requirement seems to be too strong from an epistemic point of view (for discussions, see, e.g., Chalmers 2004 and Jackson 2013). A person can be a competent speaker without knowing all of its parts or having a complete correct theory of that language. My aim here is to specify some of the methods of CA. Following the general model of method, in order to specify a method, one should provide a set of instructions and their ordering. In general, the granularity of any model is governed by the needs of the modeler. For my purposes, the three methods of CA need to be specified in such a degree of detail that displays the differences between them. For each of the three methods, I discuss the kinds of problems in which they are used, as well as the respective kinds of backgrounds. I shall not formalize the proposed methods into graphs, although the general model of method I use allows this. Instead, I shall simply state the set of instructions using ordinary language. In the next three subsections, I proceed using the following template. First, I specify the kind of problem motivating the application of the given form of CA and discuss a wellknown piece of philosophical research dealing with an instance of that kind. I then provide a model of the respective method of CA as an ordered set of instructions. 3.1 Constructive analysis. The problem motivating constructive analysis is the lack of an explicit relation among terms or concepts of a language within our conceptual the- 4 For another example of the application of this model, see (Halas 2015). 5 For a study of the difference between the explicit and the implicit use of terms see, e.g., (Glavaničová 2015). 222 ory of that language. Constructive analysis aims to broaden our conceptual theory, either by postulating a new relation or stating that some already known relation holds among previously unrelated parts of the language. Constructive analysis thus enables one to introduce new terms or concepts which were lacking in the initial explicit conceptual theory. Clearly, certain definitions (namely, prescriptive ones) will surely be a part of constructive analysis; another example is explication. A correct constructive analysis should be based on the initial explicit conceptual theory. I view a constructive analysis as correct if the resulting change in the conceptual theory leaves the relations of the initial theory intact. In other words, the new conceptual theory which results from a correct constructive analysis has the initial conceptual theory as its subpart. A constructive analysis should always be correct. A constructive analysis is coherent if the change of the conceptual theory is made using material already present in the initial conceptual theory. Therefore, a coherent constructive analysis either does not postulate any new concepts or terms in the theory, or it does not postulate a new relation. A constructive analysis need not always be coherent. To sum up, a correct analysis is enabled by the relations within the initial conceptual theory, while a coherent analysis is done within the limits of a correct analysis. The enrichment of the initial conceptual theory can be based on present intuitions or on the discovery of certain relations which hold implicitly in it. For example, there could be an implicit relation among terms or concepts which is unknown to the speakers: the speaker may be unaware of the fact that the relation of entailment holds among some propositions, although she may understand those propositions perfectly well. Can there be an incorrect constructive analysis? It would be an enrichment of the old conceptual theory which does not respect the initial conceptual theory. For example, let us presuppose that a definiendum and its definiens should be equivalent. Now, assume a conceptual network with a relation of equivalence among some of its terms. Here, a constructive analysis, if it is to be correct, must not introduce a relation of defining between terms that are not equivalent, for it would not respect the relations of the initial conceptual theory. 3.1.1 Case study: Russell. Constructive analysis explicitly modifies our conceptual theory of language by stipulating new relations which were not explicit before. One of the best known and widely discussed examples of such practice is Russell’s On Denoting (Russell 1905). According to Soames, this paper laid the foundations of modern-era analytic philosophy (Soames 2003, 127). In it, Russell proposes a theory of the meaning of propositions containing denoting terms. He famously proposes the theory of definite descriptions, i.e., of denoting terms containing the definite article as the main quantifier (such as the present king of France, the centre of our solar system, etc.). I shall not discuss the correctness of this theory, for my aim here is methodological. Russell proposes a new theory according to which definite descriptions do not have a meaning of their own. This can be viewed as stating a constraint on the conceptual theory of language. According to that theory, language does not contain such denoting terms as Filozofia 71, 3 223 self-contained meaningful language terms, although it may contain such denoting terms as parts of complex terms. The meaning of these terms is then specified in a well-known way. Russell’s analysis does not enrich the initial conceptual theory with a new concept. Rather, it specifies a whole new relation which, according to Russell’s theory, exists in the language. It is important to note how Russell presents his theory. First, he specifies the general background: “Thus a phrase is denoting solely in virtue of its form. We may distinguish three cases: (1) A phrase may be denoting, and yet not denote anything; … (2) A phrase may denote one definite object; … (3) A phrase may denote ambiguously; …” (Russell 1905, 479) He then formulates the problem: “The interpretation of such phrases is a matter of considerable difficulty; indeed it is very hard to frame any theory not susceptible to formal refutation” (Russell 1905, 479). He proceeds by stating a simple theory for less problematic denoting terms (a, no, some). Up to this point, Russell can still be seen as merely stating the background for the important move that comes next: the formulation of the theory of definite descriptions. From the systematic point of view, constructive analysis could end here. But Russell continues by comparing his theory of definite descriptions with other theories. He does so by stating three problems and comparing how well-equipped each of the theories is to resolve these problems. 3.1.2 A model of constructive analysis. Using the example and the characterization of constructive conceptual analysis above, one can now specify the method as an ordered set of instructions: 1. Specify the initial conceptual background CB! 2. Formulate the conceptual problem P! 3. State the new conceptual relation R! 4. Formulate tests T of the conceptual relation R within CB! 5. Elaborate the new relation R by tests T respecting CB! 6. If the relation R succeeds in tests, declare it a part of CB! 224 The above set of instructions may seem simple. However, that impression merely indicates that the method of constructive CA is indeed a complex method – i.e., one which has other methods as its parts. Instruction 3 really is a placeholder for some such method, for example, the method of definition or explication. We can therefore specify the difference between constructive CA and the method of defining. In at least some forms of the method of defining6 we examine our conceptual system and state the relation between the definiendum and the definiens. In constructive CA, the method of defining is used in a wider context, in order to solve a specific kind of problem, and it is tested on test cases. 3.2 Detection analysis. It is common practice in philosophy to question a declaration that a certain relation holds among concepts of a given language. As Williamson put it, “‘Philosophical questions are more conceptual in nature than those of other disciplines’: that can easily pass for a statement of the obvious” (Williamson 2007, 48). In the following, I differentiate between the explicit and the implicit conceptual theory of a language. I do not presume that rational agents know all the logical consequences of their explicit knowledge.7 This opens up the plane for many questions and, incidentally, is also one way of dealing with the Paradox of Analysis (see, e.g, Fumerton 1983 and Ackermann 1992). For example, one can ask whether some terms are equivalent if we consider some other terms equivalent. This kind of reasoning is common in solving conceptual or mathematical equations (Eagle 2006). Usually, we have some knowledge about relations among terms at our disposal, which can be used, e. g., in substitutions when solving equations. When doing philosophy, one can proceed in a similar way. Using our knowledge, we question the existence of implicit conceptual facts. Simply put, the problem for detection analysis is whether some conceptual relation exists within our conceptual network. When using this method, we ask neither about things we already know, nor about the existence of explicit relations. We ask whether – given our explicit conceptual theory – a given relation could hold implicitly. The difference between constructive analysis and detection analysis is in the role of conceptual theory in the method. As we have seen above, constructive analysis results in a change of our explicit conceptual theory. In detection analysis, the explicit conceptual theory is studied, but not modified. However, the results of detection analysis may, in a next step, motivate constructive analysis. For example, we may discover that our conceptual network contains two regions of equivalent terms which are not connected by an explicit equivalence relation. We can then take the constructive step and change our explicit conceptual theory accordingly. Another difference between constructive and detection CA is that a correct detection CA can lead to a negative result. We can simply detect that our conceptual network does not entail a given conceptual relation or that it precludes it. In contrast, constructive CA fails if it does not result in a new relation. 6 See (Zouhar 2014) and (Zouhar 2015a,b). 7 For a discussion of the rational agent see, e.g., Chapter 6 in (Jago 2014). Filozofia 71, 3 225 The role of intuitions in detection CA is substantial, but they are not indispensable. Intuitions can be used to detect whether a relation among terms is possible. But we can also find the answer in knowledge that is already available or proceed by combining both of these sources of evidence. 3.2.1 Case study: Gettier. The aim of detection analysis is to find out whether some fact holds in our implicit conceptual theory. The implicit conceptual theory is the explicit conceptual theory closed under logical laws. Detection analysis can draw on any of a variety of known logical facts about the relations among concepts, like the fact that if the term A is equivalent to the term B and the term B is equivalent to the term C, then A is equivalent to C. Generally, detection analysis questions whether we can use some terms according to our known use of other terms. Gettier’s short article (Gettier 1966) is a well known and influential work. As before, my aim here is not to question the correctness of Gettier’s analysis. Rather, my focus shall be methodological. Although Gettier’s paper only spans three pages, it contains more than one case of detection analysis. Gettier’s analysis sets off with the title: Is Justified True Belief Knowledge? This is an explicit statement of the problem in the form of a question, namely, a question about the existence of a semantic relation among concepts. Gettier then presents two cases (Smith and Jones have applied for a job, Jones owns a Ford) and models the respective situations. Using these cases, he presents counterexamples to the relation in question. He concludes that the relation cannot hold universally. From the perspective of this paper, it is important to highlight Gettier’s use of logical constraints to broaden the initial conceptual theory.8 In other words, if the implicit conceptual theory is closed under logical laws, the explicit conceptual theory can be broadened while abiding by those logical laws. Gettier specifies the facts that can be described using terms of the language determined by the conceptual theory under investigation. We can thus view such facts as possible models of a conceptual theory, depicting referential relations among concepts. Finally, Gettier asks the intuitions whether the relation in question holds among the concepts in the models presented. 3.2.2 A model of detection analysis. Using Gettier’s example and following the presuppositions specified in Section 2 above, we can now state the method of detection analysis as a set of ordered instructions: 1. Specify the conceptual theory T! 2. Specify the conceptual problem P according to the theory T! 3. Specify the set of logical constraints S! 4. Respecting the logical constraints specified in S, broaden the conceptual theory T into T0! 8 For example, “… for any proposition P, if S is justified in believing P, and P entails Q, and S deduces Q from P and accepts Q as a result of this deduction, then S is justified in believing Q” (Gettier 1966, 121). 226 5. For an undetermined number of times, repeat steps 6 and 7! 6. If the broadened conceptual theory T0 provides a counterexample, return a negative result! 7. If the broadened conceptual theory T0 entails the relation, return a positive result! Although this method is specified using more instructions than the previous one, it is still a scheme rather than a full-blown specification. The conceptual problem could, for example, be the (supposed) existence of a conceptual relation between at least two concepts. Steps 3 and 4 can be filled with various particular techniques (e.g., by proving theorems). In step 6, intuitions may play their evidential role; for instance, Gettier presents counterexamples and interrogates the intuitions. The scheme above also entails the possibility that the method of detection CA returns a negative result. The difference between constructive and detection CA is apparent. Both methods begin by stating the background. But while constructive analysis broadens and changes our conceptual theory without having to abide by the logical constraints of the initial implicit conceptual theory, detection CA must remain firmly within such constraints. 3.3 Reductive analysis. The problem motivating reductive CA is whether some theory or language is reducible to another theory or language. For example, the question could be whether the former is only a notational variant of (a part of) the latter. This, of course, is not the only possible conceptual relation between two theories or languages. We can ask this question both about theories of languages and about theories in languages. If we focus on the latter, we can use our theories of those languages. In the former case, we cannot do this, but I consider this to be less common. Therefore, I shall focus on the kind of reductive CA in which we use our knowledge about languages (or their parts) to reason about the possibility of reduction between them. When analysing reducibility between two theories, one typically proceeds in either of the following two ways which differ from the point of view of the conceptual relation in question. One can either ask whether one language is equivalent to a part of the other or one can seek to find out whether one language is merely a notational variant of a part of the other. The equivalence in question need not be a one-to-one correspondence. More commonly, a number of simple terms in one language are reduced to a number of complex terms of another language. Science often redefines ordinary terms such as weight, color, well-being, knowledge etc. A well known philosophical reduction is the one between knowledge and justified true belief. The main difference from both constructive and detection CA is that while in both of these we study the relations among parts of a language, in reductive CA we study the relations among two or more conceptual networks. The problem of reductive CA is solved by finding out whether the relation in question holds among the given conceptual networks. One could test the existence of such a relation by trying to find counterexamples based on facts and intuitions about these conceptual networks, or one could attempt to prove the existence of the relation on the basis of logical constraints behind these conceptual networks. Filozofia 71, 3 227 3.3.1 Case study: Jackson. Modern debates about the method and role of CA were revived by the work of Frank Jackson (Jackson 1998). Jackson argues for an important role of CA in any efforts to reduce one theory to another: CA is used to locate the meaning of terms of one theory in terms of another theory. There does not seem to be a single unique method to do that. For example, authors working within the so-called Canberra Plan try to provide such reductions using Ramsey-sentences. Here, I focus on the part of Chapter 4 in Jackson’s work in which he deals with the term color. The theory which Jackson subjects to reduction is the so-called ‘folk theory’, i.e., the conceptual theory of ordinary common sense – a theory of language of laypeople. It need not be specified fully and explicitly and indeed it usually is not. Rather, it is determined by everyday intuitions which play a crucial role in the reduction:9 the new theory must respect the intuitions of folk theory. Jackson begins the chapter by summarising the knowledge behind the special theory into which he wants to reduce the term color: “We know that objects have dispositions to look one or another color, that they have dispositions to modify incident and transmitted light in ways that underlie their dispositions to look one or another color, that they have physical properties that are responsible for both these dispositions, and that subjects have experiences as of things looking one or another color. We also know that this list includes all the possibly relevant properties.” (Jackson 1998, 87) He then states the problem: “…we have words for the listed properties – I used them in giving the list. But these words are not color names as such; they are rather terms for dispositions to look colored and affect light, for the physical property bases of these dispositions, and for certain perceptual experiences. Color thus presents a classic example of the location problem” (Jackson 1998, 87). He then presents his solution to the problem, which he goes on to advocate afterwards. From our perspective, it is important to note that Jackson does not define the term color in folk theory. He explicitly conditions the validity of his proposal on the statements of a special theory: “We will see, how this fact, when combined with what science tells us, forces us to identify colors with certain physical properties” (Jackson 1998, 88). Hence, he reduces a simple term from folk theory into a complex term of a special theory. The important relation between these two theories is that they both must respect the same intuitions. The relevance of a theory is measured by the support it finds in folk intuitions. From a methodological point of view Jackson determines the folk theory by stating some non-problematic intuitive facts: “In order to address that question, we need to start with what we find most obvious about color. …We can sum this up by saying that some such clause as: ‘red’ is the property of an object putatively presented in visual experience when that object looks red, is a subject-determining platitude for red. Let’s call this platitude, and the corresponding platitudes for yellow, green, and so on, the prime intuition about color.” (Jackson 1998, 89) 9 For a discussion of the role of intuitions, see e.g., (Nolan 2009). 228 Then he determines the special theory (just enough for his purposes) and elaborates the special theory with the aim of finding a concept which would play the role of the concept color in folk theory. He does this in a simple argument: “We can spell the argument out thus: Pr. 1 Yellowness is the property of objects putatively presented to subjects when those objects look yellow. (Prime intuition) Pr. 2 The property of objects putatively presented to subjects when the objects look yellow is at least a normal cause of their looking yellow. (Conceptual truth about presentation) Pr. 3 The only causes (normal or otherwise) of objects’ looking yellow are complexes of physical qualities. (Empirical truth) Conc. Yellowness is a complex of the physical qualities of objects. And likewise for all the colors.” (Jackson 1998, 93) Jackson simply analyses the special theory and by reasoning within logical constraints he states the concept which plays the role of color in the special theory. In doing so, he solves the location problem for color, at least by his standards. 3.3.2 A model of reductive analysis. Using the above specification of the method and Jackson’s example, I now propose an ordered set of instructions which represents the method of reductive CA: 1. Specify the (part of) theory T in language L to be reduced! 2. Specify the (part of) theory T0 in language L0 into which the theory T will be reduced! 3. State the relation R between the theories T and T0 which shall be respected! 4. State the reduction relation TT0. 5. Test the reduction relation TT0 using the knowledge base with respect to the relation R! 6. If the test is positive, declare the reduction TT0 between (a part of) theories T and T0! The specification of reductive CA in this form enables a comparison with constructive CA and detection CA. While detection CA studies the features of a single conceptual network, reductive CA generally studies the relation between two conceptual networks. As we have seen in the previous subsection, these conceptual networks need not be fully and explicitly specified. The main difference with respect to constructive CA is that reductive CA must not be arbitrary. In other words, it should abide by the role of the concepts of the theory that is to be reduced. On the other hand, a common feature of reductive and constructive CA is that a relation is stated if the analysis is successful. 4. Conclusion. I have presented three different methods of CA. Generally speaking, the method of conceptual analysis is used to study and modify the explicit conceptual theory of some language. It is usually carried out in the form of research into its conceptual network. Filozofia 71, 3 229 The problem motivating constructive CA is the lack of a relation among concepts in the explicit conceptual theory of a language. Constructive CA is used to modify the explicit conceptual theory so that the problem is solved within a (possibly enriched) conceptual theory. This type of analysis studies pre-existing relations in a conceptual theory and proposes a new relation, which is then tested. The problem motivating detection CA is the possibility of the existence of a conceptual relation in the implicit conceptual theory. The implicit conceptual theory is provided by our knowledge of the explicit conceptual theory of the language in question. Detection CA is used to analyse and broaden our knowledge of the implicit conceptual theory. It abides by the specified logical constraints as well as the initial intuitions of competent speakers. The problem is solved when the existence of the studied conceptual relation is found or proved possible or impossible within the implicit conceptual theory. The problem leading up to a reductive CA is the existence of a conceptual relation among different languages. Provided our knowledge of explicit conceptual networks of those languages, we study the possible relations among those networks. The problem is solved when the existence of such a relation is proved or shown to be impossible. All of the methods of CA studied begin with the collection of knowledge about the initial conceptual systems. The researcher then modifies her knowledge either by using intuitions while respecting logical constraints or by providing constructive steps which do not have a negative effect on the correctness of the conceptual theories studied.