Two Grades of Evidential Bias Paul M. Churchland Philosophy of Science, Vol. 42, No. 3. (Sep., 1975), pp. 250-259. Stable URL: http://links.jstor.org/sici?sici=0031-8248%28197509%2942%3A3%3C250%3ATGOEB%3E2.0.CO%3B2-7 Philosophy of Science is currently published by The University of Chicago Press.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/ucpress.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Sat May 12 00:24:47 2007
TWO GRADES OF EVIDENTIAL BIAS* PAUL M. CHURCHLAND? University of Manitoba
It is argued herein that there are two distinct ways in which all observailoll vocabularies are prejudiced with respect to theory. An argument based o n the demands of adequate translation IS invoked to show that even the simplest of our observation predicates must dtsplay the first and more obvious grade of bias-irntensionnl bius. It is also argued that any observat~onvocabulary whose predicates are corrigibly applicable must manifest a second and equally serious grade of bias--extensional bias-independently of whatever intensional b ~ a sits predicates may or may not have.
1. Introduction. Is there a theory neutral observation vocabulary? Does the proper use and understanding of even the simplest of our observation predicates involve a commitmei~tto problematic presuppositions? Is it the unavoidable fate of any observation report surreptitiously to beg the question forlagainst various theories? On the answers we give to these questions hangs the survival of any recognizable form of empiricism. Current views on the matter differ widely. Many of us refuse to concede the passing of the palpable, perhaps the central worry being how and whether we can account for the rationality and integrity of the process of theoretical developinent if we deny the existence of a neutral standard of unbiased observation. Others, however, choose to see bright new possibilities opening before us here. At one level, we are faced with the necessity of finding and pursuiilg entirely fresh approaches to the question of what rational ii~tellectualevolution really consists in. And at another level, we are presented with the intriguing possibility of learning and using new and more powerful observation vocabularies to replace our present and relatively feeble modes of conceptual exploitation of sensory stimulation. At the heart of these broader matters, however, is the basic issue of the semantics of observation predicates and the issue of constitutional prejudice itself. Discussion ln the present paper is confined to these. The most obvious variety of evidential bias &ill here be called 'intensional bias'. Explications of this idea run the risk of being instances of it, but the general idea is as follows. A predicate is intensionally biased with respect to theory just in case (a) the proper use and understanding of that predicate requires at least a tacit acceptance of (or, in the case of a predicate not thought to be true of anything, at least a tacit appreciation of) some one or more general assumptions in which that predicate figures, where (b) these assumptions might in principle run afoul of advancing theory and new information. "Theoretical7' predicates are commonly regarded as paradigm instances of what we are here calling intensional bias. One's understanding of (the "meaning" of) a theoretical predicate consists in one's appreciation of the assumptions and principles of the theory in which it figures
* Received May, 1973. t I should like to express my thanks to their helpful criticisms of a n earlier draft.
Leon Ellsworth, Robert Audi, and Steven Savitt for
TWO GRADES OF EVIDENTIXL BIAS
25 1
(degrees of understanding are plainly possible here). ft has for us the specific sense that it does precisely because of the presumed truth of a certain cluster of assumptions and principles containing it. If we are forced to deny wholesale the elements of that cluster, then we must deny that the predicate has any instances. Conversely, to ascribe seriously such a predicate as true of something is to presuppose the truth of at least some of those principles, though not ~iecessarilyall. Hence the bias of even singular ascriptions of such a predicate. The idea that all predicates might be biased in essentially the same way derives a good portion of its plausibility from the increasingly apparent poverty of the analytic/synthetic distinction. The great many "analytic truths," long seen as the incorrigible bearers of stable "meaning," now appear instead as epistemologically ulzdistinguished elements in a decidedly viscous but nonetheless fluid "understanding." Of course, these presumed truths must still be counted as central to our (current) understanding of the predicates which figure in them, but since they have lost their fabled immunity to ref~~tation and modification at the hands of fresh inf'orn~ation,intensional bias is suddenly everywhere. Or almost everywhere. The so-called "phenomenological" predicates such as 'red', 'grey', 'cool', and so on are not obviously instances of the thesis at issue. Indeed, they form the leading examples for the competing idea that the niealaing of at least some predicates is given in and by sensatioiz rather than by some cluster of general assumptions, corrigible or otherwise. Such predicates would therefore be free of intensional bias. It is argued in the next section that this is an illusion. Specifically, tile argument aims to show that, if the intrinsic qualities of our sensations play any role at all in determining the meaning of the simple observation predicates at issue, that role is inessential and exceedingly peripheral. Though of major importance, the issue of intensional bias has dominated recent discussions of empiricism to an extent not entirely deserved, for there is a more straightforward variety of prejudice which deserves at least a comparable share of our attention. Observation vocabularies manifest what I shall call 'extensional bias' in that (a) the range of sensations with which we are endowed does not determine a uniquely appropriate observation vocabulary, (b) possible observation vocabularies can differ radically in the extensional classes into which they divide the observable t+orld, and (c) corresponding descriptions of observable phenomena couched in two such extensioilally orthogonal vocabularies can and will give significantly different degrees of support to one and the same theory. The anonmalies attending the projection of grue-like predicates (see [I], Chapters 3 and 4) may come to mind here, and these are indeed a part of the story of extensional bias, but only part of it. Extensional bias can be found riot only in cases of simple projection, but also in cases involving the confirmation or corroboratio~lof highly general theories. The bias involved is less outrageously apparent, but if we are willing to settle for more subtle examples of extensional bias, there are niuch more "natural" predicates than 'ggre' available to illustrate the poverty of empiricism, as I shall try to show. 2. Intensional Bias. Let us approach this first issue by examining the case of some
252
PAUL M. CIlURCHLAND
differently endowed perceivers. Most will agree that, as presently constituted, we are unable to perceive visually the middle range temperatures of material objects. But it is not dificult to imagine beings who could. Simply imagine a race of me11 wit11 larger eyeballs and/or more highly refractive lenses, and with retinas consisting solely of rods highly sensitive to the radiation which bodies emit in the far infrared. Since the vigor with which any body radiates in the far infrared is a direct function of its temperature, aild since images of these bodies will be formed on the retinas of the kind of eyes described, the people who possess them will be quite prepared, physiologically, to perceive visually the teiilperatures of bodies. The ability to discriminate temperatures would be best exercised in darkness, since in daylight the infrared radiation that bodies emit can be confused with or lost in the infrared radiation they reflect to us from the sun, but this feature of the situation merely corresponds to our own inability to discriminate colors effectively when, conversely, background light is very dim. Let us imagine, then, an independent society of such nocturnal beings speaking a language that is, superficially at least, illdistinguishable from English (but lacking our color vocabulary, including 'black', 'white', and 'gray'), wherein our ordinary temperature vocabulary is learned by the very young as an observation vocabulary for cisual instead of for tactile reports. (To simplify things, assume also that these beings lack a tactile or bodily sense for temperature, as we lack a tactile or bodily sense for color.) They acquire the use of the relevant terms in the same manner we learn the use of our color predicates, and they steadily amass, in various ways, the usual set of general beliefs or assumptions concerning temperature: "a warm thing will warm up a cooler thing, but never the reverse," "fires are hot," "Jf x is hotter than y, and y is hotter than z, then x is hotter than z," "food keeps better in ~l "rubbing things makes them a cold place," "hot things cause p a i ~ i f ~burns," warmer," and so on and so forth. Generally speaking, the set of "tempera,ture beliefs" of any adult member of this society is no more dissimilar to your own set than yours is to, say, your nextdoor neighbor's. A few differences are pretty much standard, of course. "Temperatures can be seen" is held by all of them, but by very few of us, and the reverse holds for "Temperatures can be felt." But given the differences in our sensory modalities and the lack of pressure on the common man to consider the possibility of other sensory modalities, such a failure of perfect correspondence is hardly surprising. Given their linguistic behavior, the special nature of their eyes, and the de facto accuracy of their perceptual claims, the natural position to take is that these people can indeed visually perceive the temperatures of objects, at least under normal (i.e. nocturnal) conditions. As they see it, the visually perceivable world consists not of middle sized and variously colored material objects, but rather of middle sized and variously heated material objects. That is, they can visually perceive hot objects as hot (cold, warm)-they can visually perceive that they are hot (cold, warm). If we accept this conclusion concerning their perceptual capabilities, we should notice that we have done so without benefit of any information concerning the intrinsic qualities of their visual sensations. Should this affect the matter? To make
TWO GRADES OF EVIDEKTJAL BIAS
253
matters interesting, let us suppose finally that, as far as the intrinsic nature of their visual sensations is concerned, the world "looks" to them exactly as it looks to us when it is viewed through a high quality infrared "night scope," one producing an image in black, white, and the intervening shades of gray. (This is in any case highly plausible, given that (a) their retinas contain only rods, and (b) we are supposing their physiology to be entirely human beyond the peripheral respects cited.) That is, on viewing a very hot object they have what we would describe as a sensation of an incandescent white object, and on viewing a very cold object they have what we would describe as a sensation of a black object, and so on. They, of course, describe these sensations quite differently-as sensations of heat, of coldness, and so on. With this assumption we arrive at the heart of the matter. If we embrace the view that the meaning of the simple observation terms at issue is given in sensation, we must insist that their terms, 'cold', 'hot', and 'warm', really mean black, +r.hite, nrzd gray, rather than cold, Izot, and u'arm. But this heterophonic translation is not without serious consequences. If we adopt it, we shall have to count as false all or most of their many background beliefs involving the predicates in question; try substituting 'black', 'white', and 'gray' respectively for 'cold', 'hot', and 'warm' in the sample "temperature" beliefs listed above. And v:e shall have to count as false all of their "observation" judgments involving the relevant terms (save for those accidental cases where, for example, a warm object just happens to be gray), for they certainly cannot see whether objects are black, white, or gray; their sensations are responses to other parameters entirely. To insist on this sensation-guided translation is, I suggest, to make a joke of a perfectly respectable and very powerful sensory modality, and of a simple and appropriate mode of conceptual exploitation which has every virtue we can claim for our own habits of judgment in matters visual. While it is true that, under normal conditions, tlzis is how gray things and only gray things look to standard observers, it is equally true that under (different) norlnal conditions, tlzis is how n w m things and only warm things look to (different) standard observers. But we need not merely plead for sympathetic intuitions here. The crucial consideration is the following. If we insist, by way of this sensation-guided heterophonic translation, on thus making a joke of their beliefs and visual capabilities, we must be prepared to have the very same joke made of our own beliefs and visual capabilities with respect to black, white, and gray. The fact of the matter is that their beliefs, background and observational, involving their 'cold', 'hot', and 'warm', do not match up at all with our beliefs, background and observational, involving our 'black', 'white', and 'gray', and the failure of match is such that if we insist that these two predicate families are genuine translational correlates, then we must insist that at least one of these two sets of beliefs is sy?tematically false. But ~ihiclz? If our infrared cousins adopt the same strategy of translation, as by parity of reasoning they must, they will be as eager and as well prepared to insist that it is our beliefs, background and observational, that must be discounted as systematically false. They will regard our visual judgments on black/white/gray (i.e. "cold"/"hot"/ "warm") things as systematically mistaken; and, in translating our background beliel's, they will find in us silly coti\ictions like "snow is hot," "Africans are colder
254
PAUL M. CHURCHLAND
than Europeans," "a hot shirt shows the dirt more readily than a cold one," and so forth. Their story on us is the image of our story on them; it is as uncomplimentary, and it is as well founded. In short, the proposed translation lands us in an epistemological dilemma ("Will the real owner of the perceptual capability at issue please stand up?') to which there is no resolution. Both of these stories must be rejected as thoughtless and parochial, for there is no relevant asymmetry between our respective cases to sustain the one piece of foolish~iesswhile unmasking the other. Translating their 'cold' as our 'black', and so on, is therefore out of the question. Quite aside from the unwarranted nonsense that translation would make of their background beliefs and perceptual capabilities, there is nothing to locate the relevant kind of nonsense in their case rather than in our own. By contrast, the straightforward homophonic translation of their "temperature" vocabulary has every empirical virtue it is possible for a translation to have. The infrared people judge to be hot, cold, and warm precisely the same particulars that we so judge; the agreement between their background beliefs and our own concerning temperature is effectively complete, with the few deviations yielding to obvious explanation; and the relevant patterns of inference to which they cleave again match our own. The impossibility of the heterophonic translation discussed above is just the impossibility of the thesis that the meaning of the common observation terms at issue is given in sensation. And the independent appeal of the homophonic translation is just the appeal of the view that the meaning of the relevant terms has nothing to do with the intrinsic nature of those sensations which, in us, just happen to prompt their application in singular empirical judgments. It may be objected that there is a tertium quid here, that the wholesale adoptio~t of the homophonic translation is not the only alternative to the particular heterophonic translation rejected above. Perhaps a part of the meaning of the relevant terms is given in sensation, while the remainder is fixed by a cluster of general beliefs. A proponent of this view could concede that, since the infrared people's "temperature" terms are extensionnlly equivalent to our temperature terms, the homophonic translation might be "adopted" (with a broad wink) for practical purposes, while still insisting that our own language contains no genuine translational correlates for the alien terms at issue. 111 the present context this alternative is not as interesting as it might seem, for it concedes the fundamental point at issue. If a part of the meaning of the relevant predicates is fixed by clusters of general sentences containing them, then, unless we can breathe life back into the notion of semantic incorrigibility, the case is closed: those predicates are intensionally biased. And aside from this, the third alternative has nothing to recommend it over the homophonic alternative anyway-quite the reverse. I t requires us to deny that the beings with the infrared eyes can perceive the temperatures of objects, and indeed to deny that any beings, no matter what their sensory apparatus, could perceive the temperatures of objects i~nlessthey are subject to precisely the same range of bodily sensatioi~sw ~ t hwhich we iiappen to respond to hot and cold objects. If they
TWO GRADES OF EVIDENTIAL BIAS
255
do not have these sensations, they cannot even have the concept of temperature, let alone perceive whether and to what degree it is instanced in objects; they must be perceiving something else unknown to us. If there is any plausibility in this line of argument, it derives, I suggest, from the contingent fact that relative to us both color and temperature are monomodal properties. To help free our intuitions, let us consider some properties which are bimodal for us. For example, small objects can be felt to be round, and they can also be seen to be round. Consider now two people, one of whom can only see while the other can only feel whether objects are round-the latter is congenitally blind, say, while the former is sighted but congenitally paralyzed or benumbed. Shall we say of them that they can both perceive whether objects are round, though by way of different sensory modalities? Or shall we insist that they must be perceiving different properties, that the ordinary term 'round' is systematically ambiguous, and that their respective 'round's are intertranslatable only with a broad wink? I take it that the former is the plausible alternative. But if it is, then we must ask how the case of the nocturnal beings and ourselves, with respect to 'warm', differs from the case of these two differently equipped humans with respect to 'round'. To draw the relevant parallel differently, let us assume that our nocturnal cousins have the normal tactile or bodily sense for temperature after all. Could they not then both feel and see that objects are warm, just as we can both feel and see that objects are round? Would not temperature then be a bimodal feature for them, just as roundness is for us? Reflecting on considerations such as these, it is tempting to think that, had our sensory equipment been sufficiently rich that all properties observational for us were at least bimodal, the idea of a distinct kind of phenomenological meaning would never have been conceived in the first place. The burden of this discussion has been to illustrate the need to drive a wedge between one's understanding of even our simplest observation predicates, and one's acquired dispositions to make spontaneous singular judgments containing those predicates in response to whatever range of sensations nature has given to one. Having made a prirna facie case for this view, we are in a position to make, and handle, a needed concession. It is quite possible for a belief such as, "In normal conditions, white things and only white things look like this to (or cause this kind of sensation in) normal humans," to comprise an element of some importance in a person's understanding of the term 'white'. But that element is a belief, among other beliefs. It is an element whose importance will vary inversely with the number of other general beliefs he has involving 'white', for the fact it expresses is no more an essential fact about white things than is the fact that, under certain conditions, white things and only white things cause such-and-such sensations in bumblebees, or such-and-so sensations in squid. From a broader point of view, these are all equally humdrum facts about white things, and that broader point of view should be our point of view. If it is possible for different kinds of beings to share with us a common observation vocabulary-our present vocabulary, for example-despite differences in the nature of their sense organs and their sensations, then the thesis that the meaning
of obseivation predicates is given ill sciisation must be rejected outright, and, as we saw, we are left with clusters of beliefs as the deterillillants of understanding. But if so, then the emptiness and/or impotence of the notion of semantic incorrigibility is sufficient to make the case for intensional bias even in the case of such alleged paradigms of innocence as 'black', 'gray', 'cold', and 'warm'. If there is no epistemologically significant analytic/synthetic distinction (and there is not), then intensional bias is indeed everywhere.
3. Extensional Bias. It is prima facie plausible to argue that, although theoretical advances may parse and re-parse the world generally in new and surprising ways, the basic lines along which the observable world can be parsed remain fixed by the natural and unchanging patterns of our sensational responses to the environment, patterns which are rooted in our physiology. Granted, in the course of some unprecedented conceptual revolution the likes of 'warm' and 'red' might be displaced by more muscular and overtly theoretical predicates, teeming with useful and systematic implications, to function as "observation terms" in their stead. But the fundamental lines drawn across the observable world will be extensionally the same; we merely end up holding more adventurous beliefs about the areas between those lines and the relations between them. The way to avoid theoretical prejudice in our observation judgments is therefore to make such judgments in terms of predicates which simply parse the observable world along those anyway inescapable lines, and do so without further commitment on the structure and behavior of the world in its less immediate respects. This line of argument is unacceptable for a variety of reasons, but only one of them needs concern us in this final section. The lines along which we parse the observable world are not a simple reflection of the patterns of our sensational responses to the environment. They are a function of these pl~lswhatever criteria we use for distinguishing, for each observable property 4, between what is 4, and what is deceptively similar to 4, what merely looks 4 (feels 4, smells +), and so on. At this stage, I shall not insist that the administration of these distinctions necessarily involves theoretically problematic assumptions; assume, if you find it plausible, that the background beliefs which guide us in drawing such distinctions are all either semantically incorrigible or semantically irrelevant. My concern here is to point out the real and very considerable latitude we are permitted in parsing even the observable world. The adoption of a different set of observation predicates can involve the adoption of criteria such that the lines drawn between mere appearance and genuine reality, with respect to the new predicates, differ substantially or even radically from the familiar lines which mark such distinctions relative to our present predicates. This just means that any new predicate need not be extensionally equivalent to any of the old predicates-that the family of new predicates may parse the observable world along lines which cut completely through and across the old lines. Realistic possibilities are not difficult to imagine, though the requirement of ready intelligibility limits us here to the less dramatically different among them. Consider the following predicates which could in principle operate in place of our
TWO GRADES OF EVIDENTIAL, BIAS
257
present color vocabulary. 1 characterize them exlensionally in terms of our ow11 predicates, but of course they would not be so grasped by their users. x i s nray r x looks gray, to normal human observers, under conditions of din1
moonlight.
x is w e d r x looks red, to normal human observers, under conditions of dim
moonlight.
And so on for all the other "nolors." Relative to this vocabulary, "normal perceptual conditions" are those of dim moonlight. Under such conditions, of course, the light level is too low for our color vision to be operative and, with the sole exception of self-luminous bodies, everything looks black, white, or intermediate shades of gray to normal human observers. Accordingly, the bulk of common objects are (genuinely) nlack or some shade of nray. True, most things look bright nreen, bright nlue, bright nred, and so on under the illusory conditions of broad daylight, but they are not. Oranges for example, are not norange at all-though they look norange in sunlight-they are really dark nray. Very hot coals, on the other hand, are genuinely norange or nred, since they look norange or nred by dim moonlight. And the same for fireflies, which are genuinely pale nreen. Unripe apples, on the other hand, are nray, and approach nlack as they ripen. The important points here are the following. First, these "nolor" predicates are as genuinely observational (under the relevant conditions) as are the more familiar color predicates. If skin fatally sensitive to sunlight had made us nocturilal beings, this vocabulary might even be the more natural and usef~llof the two. Second, the two predicate families (colors and nolors) are "extensionally orthogonal." The class of nray things, for example, cuts through and across all of the color classes save black and white. It even cuts through gray: dark gray things look black by dim moonlight, and hence are nlack, not nray. The class of nred things is not even a subclass of the class of red things, since a heated cast iron bar which still looks appropriately black in broad daylight can be visibly a dull nred in dim moonlight. Not even 'white' and 'nwhite' match up, for the class of nwhite things includes some self-luminous yellow things (automobile headlights are decidedly yellow in daylight, but look white at night). And last, these nolor predicates certainly suffer no more intensional bias than do our ordinary color predicates. They differ in principle oilly in the conditioils counted as normal for their correct perceptual application. We need not explore this minor example ally further, nor any of the many other more complex and intriguing possibilities. That alternative vocabularies of the relevant kind are possible is all we require to proceed with the following argument. Consider any two observation vocabularies V , and V, such that the categories of the one cut across the categories of the other in highly complex ways. And consider two competing theories Tl and T, in contact with Vl and V, by way of four distinct sets of "bridge laws" or "correspondence rules." Suppose that T, parses the world generally along lines which fit nicely with the observational categories of V,; that is, the correspondence rules which connect expressions in the
258
PAUL M. CHURCHLAND
theoretical vocabulary of T1 with expressions from V1 are clean, natural, and simple. Suppose contrariwise that T, parses the world generally along lines which cut through, across, and variously around the categories of V,; that is, the correspondence rules connecting T2 with V1 are complex, inelegant, unnatural, and cumbersome. And suppose finally that in all other epistemologically relevant respects (testability, internal simplicity, corroboration to date, and so on) TI and T2 are on a par. Consider this situation from the point of view of someone whose observation vocabulary is V,. The choice between TIand T, is clear. Considerations of overall simplicity demand that TI, with its simpler correspondence rules and smoother fit with prior categories, be chosen over T,. In the circumstances described, a choice for T, would be irrational. But if we turn now to the second observation vocabulary V,, we see that its relations to T1 and T, respectively might well be symmetrical to those enjoyed by V,, but simply reversed. Given that V1 and V, are extensionally orthogonal, T, might enjoy a very simple fit with V,, despite its complex and awkward fit with V,, while T1 enjoys only a very complex and cumbersome fit with V,. In that case, for someone whose observation vocabulary is V,, considerations of overall simplicity demand that he choose T, rather than TI. A choice for T, would be irrational. Vl and V, are therefore not free of epistemic prejudice, however free of intensional bias they may be. Although T1 and T, are equally consistent with honest observation reports (to date) in either vocabulary, considerations of overall simplicity render them differentially consilient with these alternative ways of parsing the observable world. Moreover, this effect-not to be treated lightly in the first place-can feed on itself and multiply. In the long run, new theories will not only have to forge a peace with the categories of the operative observation vocabulary, but also with the categories of whatever previously adopted theories have already forged such a peace. The initial bias can therefore become increasingly entrenched and more difficult to transcend. In sum, so long as overall simplicity is a criterion relevant to epistemic choices, the use of any given observation vocabulary will involve a nontrivial bias in favor of countless theories and against countless others. Any observation vocabulary will fail of theoretical neutrality just because it parses the world along certain lines rather than along other lines. The preceding argument is of a piece with the spirit of the last two chapters of Nelson Goodman's [I]. But I hope it will be found, at least by those still unimpressed by 'grue', a more palatable illustration of both the fact and the universality of extensional bias. The class of extensionally unusual observation vocabularies at issue above is generated by relatively innocent means: variation in what counts as normal perceptual conditions. As a result, it is not even counterintuitive that the terms in such vocabularies might figure in genuine natural laws. The nolors, to take one example, plainly do (e.g. nlack things are better radiators than nray things, nred things emit their own light, and so on). And an added bonus is that the many vocabularies generated by this stratagem are incontestably observation vocabularies. Of course, we pay a price for this degree of "familiarity": simple projections of
TWO GRADES OF EVIDENTIAL BIAS
259
regularities comprehend in terms of vocabularies whose extensional oddness is secured in the fashion described need not produce anything untoward. But our concern here is with inductive matters more broadly conceived. The extensional bias of a predica.te may not show itself in what would be counted as outrageous projections of that predicate, but the bias is still there and will show itself in considerations of simplicity, if in nothing else. Without going too far wrong, one could describe the goal of theoretical science as the identification of the important natural classes into which nature divides itself and the specification of the general relations which hold between them. But as we have seen from the preceding two sections, these two matters are to a significant extent already settled in any observation vocabulary-its extensional bias consisting in its commitments on the first score, and its intensional bias consisting in its network of commitments on the second. We can put this fact into proper focus by conceding that an observation vocabulary is itself just another theoretical vocabulary, one distinguished from others only in that it is (by way of its user's acquired dispositions to couch sensation prompted singular judgments in its terms) the current vocabulary of "first response" to the causal impingements of the environment. To be sure, the methodological consequences of such view are as serious as they are yet unclear. The burden of the present paper is that we have every reason to pursue them, and no excuse not to. REFERENCE
[I] Goodman, N. Fact, Fiction, and Forecast. Indianapolis: Bobbs-Merrill, 1965.