CATHERINE
OUTSTANDING REPLIES
Z. ELGIN
PROBLEMS:
TO ZiF CRITICS
ABSTRACT. Answers set the stage for new questions. Reconfigured terrains require new maps. We ended Reconceptions with the words "constructionatism always has plenty to do". The papers in this volume prove our point.1 They raise issues and disclose avenues that merit further investigation. In what follows, I venture some brief replies that answer objections and indicate areas that deserve further study.
1.
TO G U I D O KI2rNG
Pluralism, Guido Kting reminds us in 'Ontology and the Construction of Systems', is nothing new for G o o d m a n . It has its roots in The Structure o f Appearance, where G o o d m a n rigorously demonstrates that ontologically divergent constructional systems answer to the same pretheoretical commitments. Pluralism pertaining to constructional systems, Kting maintains, may be salutary. Pluralism with respect to worlds, he views with more alarm. H e suggests therefore that we distinguish between constructional and t r u t h m a k e r ontologies. Then we can be pluralists about the one, monists about the other. I ' m not convinced the distinction can be drawn. Constructional systems are not uninterpreted logics. They reflect the truths about the domain we're antecedently committed to. Pluralism results because those truths do not suffice to fix reference uniquely. Constructionalism does not, of course, tell us which truths we are (or ought to be) antecedently committed to. In The Structure o f Appearance, G o o d m a n shows how both phenomenalist and physicalist systems can a c c o m m o d a t e the ways things look to us. KiJng contends, however, that phenomenalism is not a plausible truthmaker. If he's right, phenomenalism's falsity is one of the antecedent commitments an adequate system is required to reflect. Constructionalism can easily a c c o m m o d a t e this. But even if phenomenalism is precluded, pluralism is not. For constructionalism allows for a multiplicity of divergent physicalist systems. Kting seems to think (or perhaps only hope) that if we restrict ourselves to serious, physicalist contenders, pluralism will evaporate. I Synthese 95: 129-140, i993. © 1993 Kluwer Academic Publishers, Printed in the Netherlands.
130
CATHERINE
Z. ELGIN
doubt it. Without a viable analytic/synthetic distinction, we lack necessary and sufficient conditions for the instantiation of our terms. Our commitments then circumscribe but do not determine the identities of their objects. Ontologically divergent systems can do justice to the same commitments. This is not just a limitation on current, admittedly tess than adequate accounts. All theories are underdetermined by evidence. And cognitive desiderata like simplicity, fecundity, and explanatory power take up only part of the slack. Even a theoretically and evidentially ideal theory would admit of ontologically divergent systematizations. We might still insist that exactly one of the alternatives is right, the world supplying the truthmakers that decide among them. 2 But that would require maintaining that the world has its own categories and that truth consists in correspondence to the world's partition of itself. This view, I believe, is unintelligible. Instead, I suggest, a (hypothetically) ideal theory constitutes the commitments a constructional system is supposed to preserve. Any system that does so is ideally adequate. The sentences such a system sanctions are true, its ontological commitments being its truthmakers. Each such system, Gooaman would maintain, constitutes a right worldversion. We neither have nor need any reason to choose among them. Goodman, of course, does not share Kiing's rejection of phenomenalism. His pluralism explains why. Once we deny that there is exactly one way things really are, we need no longer fear that in accepting a phenomenalist system we lose touch with the material world or with the findings of natural science. The existence of irreducibly phenomenalist truths does not preclude the existence of irreducibly physicalist ones. Nor does the rightness of a phenomenalist world-version entail the wrongness of physicalist versions. 2.
TO R O S E M A R I E
RHEINWALD
Something is grue if and only if it is examined before (future time) t and found to be green or is not examined before t and is blue. 'Grue' then has as determinate and identifiable an extension as 'blue' and 'green'. All emeralds examined to date are grue. Why, Goodman asks, shouldn't we take this as evidence that all emeralds are grue? To do so, Rosemarie Rheinwald replies in 'An Epistemic Solution to Goodman's New Riddle of Induction', we would need to believe
OUTSTANDING
PROBLEMS
131
that at time t emeralds change color. And our best causat account of the world denies that. But the hypothesis that all emeralds are grue requires no color change. Emeralds that started out green remain green. Emeralds that started out blue remain blue. But if the hypothesis is true, all and only the green ones are examined before t. Moreover, if the hypothesis is lawlike, it sustains the counterfactual: If an emerald that is in fact blue had been examined before t, it would have been green. But this does not entail that its having been examined before t would have caused it to be green or to change its color to green. Examination does not affect color. To say that emeralds examined before t are green is not to say that their being examined before t makes them green. Rheinwald takes the grue hypothesis to be stronger and stranger than it actually is. Goodman's hypothesis neither makes nor is committed to any causal claims. It merely proposes projection to an unfamiliar class that the evidence already belongs to. Gruesome alternatives need not depend on position, person, or perspective in any way. Let us define 'schmessure' as follows: schmessure = force/unit area if even; twice force/unit area if odd. If all measurements of force/unit area to date have been even, the evidence conforms to the hypothesis that the magnitude we've been measuring is schmessure. 3 But no more than 'grue' is 'schmessure' projectible. For 'schmessure' competes with 'pressure', and 'pressure' is entrenched. Lack of positional predicates does not affect the issue. Goodman maintains that projectibitity requires entrenchment, But he does not contend, as Rheinwald alleges, that a predicate is projectible if and only if sufficiently entrenched. Besides being entrenched, projectibility requires that the predicate occur in a hypothesis that is supported, unviolated, and unexhausted. Although 'brown' is entrenched, 'All hares are brown' is unprojectible because it is violated by observations of white hares. And although 'empiricist' is entrenched, 'All hares are empiricists' is unprojectible because it is unsupported by observed positive instances. Projectibility then is sensitive to evidence. Rheinwald argues that without the causal connections that she takes to sustain 'green' but not 'grue', language learning would be impossible. I don't see why. As far as I can tell, language learning requires only a
132
CATHERINE
Z.
ELGIN
limited base of intersubjectively discernible constancies. These constancies need not be grounded in laws or even in exceptionless generalizations, so long as they are secure enough to sustain a situation in which teacher and learner respond and take one another to respond to the same thing. It's worth remembering that we have no trouble learning 'grue'. Induction does not occur in a vacuum. Rheinwald is clearly right to emphasize the importance of background information. But she's wrong to insist that belief in the hypothesis under investigation must be part of the relevant background information. One can surely gather evidence for a hypothesis before deciding whether to believe it. And evidence gathered in an effort to discredit a hypothesis can surprise us by confirming it. What motivates Rheinwald, I suspect, is that the hypothesis under investigation seems to fix the parameters on the requirement of variety of evidence. If we didn't think hares' colors changed with the seasons, we would have no reason to look for seasonal variation. This seems wrong. If we even suspected that their colors changed with the seasons, we would have reason to investigate. Belief is not required. Moreover, we often have indirect evidence. Given that other woodland creatures' colors vary with the seasons, we have reason to investigate hares for seasonal variation, whatever we think the result will be. Goodman's solution to the grue paradox turns on recognizing that the relevant background includes information about past investigations and their results. Both 'All emeralds are green' and 'All emeralds are grue' are supported, unviolated, and unexhausted. What favors the former, Goodman claims, is entrenchment. 'Green' and its cognates have been successfully projected far more often than 'grue'. Rheinwald's discussion reminds us that entrenchment alone is not enough. Entrenchment does not insure projectibility nor does projectibility insure inductive validity. Our beliefs about the domain and the methods for investigating it impose additional constraints. In debates about the grue paradox, this is all too often overlooked. 3.
TO WOLFGANG
HEYDRICH
Philosophy's linguistic turn gained much of its angular momentum from the conviction that meaning relations are rich sources of philosophical insight. By demonstrating the untenability of the analytic/synthetic dis-
OUTSTANDING
PROBLEMS
133
tinction, Goodman, Quine, and White unsettled that conviction. This left a problem: speakers do not ordinarily treat all coextensive terms as interchangeable. H o w do we decide what changes to permit? G o o d m a n takes the answer to lie in likeness of meaning. Even if no two terms are synonymous, two terms may be sufficiently alike in meaning to permit substitution in a given context. Wolfgang Heydrich welcomes Goodman's account. But, he contends in 'A Reconception of Meaning', the principle that leads G o o d m a n to conctude that no two terms are synonymous is too strong. A weaker but still adequate principle does not preclude meaning equivalences. Nothing much hangs on the resolution of this dispute. For equivalences of meaning are bound to be rare - too rare to constitute the fixed framework that the early analytic program required. Heydrich's alternative does not demonstrate that there are any synonyms. It only shows that the explication of the schema P that is not a Q need not commit us to the contrary. I don't find focusing on the schema particularly helpful. For one thing, as G o o d m a n notes, it looks like a trick. 5 For another, my intuitions about descriptions like oculist that is not an eye doctor tend to be unstable. We do better to step back slightly and recall what we're trying to do. Presumably, two terms are synonymous if and only if interchangeable (in non-intensional statements) in fiction and fact. The question is not about the phrase 'oculist that is not an eye doctor' but about whether there could be a (perhaps fictive) oculist-description that is not an eyedoctor-description or an eye-doctor-description that is not an oculistdescription. I think there could. I would be inclined to call someone who fitted cosmetic contact lenses an oculist. But if she neither conducted medical examinations nor prescribed corrective lenses, I would hesitate to call her an eye doctor. And I would be inclined to call someone who treated only afflictions of the eye that did not affect vision an eye doctor, but not an oculist. Heydrich does not, of course, rest his argument on this one example~ So even if I'm right, he need not be dismayed. But driving my example
134
CATHERINE
Z.
ELGIN
is an appreciation of the power of imagination to contrive fictions that sever putatively analytic connections. More than any explication of P that is not a Q, a look at the history of literature should give adherents of analyticity pause. Agreement in primary and in parallel secondary extensions provides an intralinguistic criterion of likeness of meaning. This is not enough. A criterion that applies interlinguistically is clearly wanted. As Heydrich shows, agreement in primary extension and mention-selection constitutes such a criterion. 'Einhorn' mention-selects English as well as German unicorn-descriptions, but neither English nor German centaurdescriptions. Heydrich's masterful deployment of mention-selection strengthens Goodman's theory considerably. It supplies a criterion suitable for evaluating translations. This is a signal achievement. Like Goodman, Heydrich construes likeness of meaning as a continuum. Unlike Goodman, he takes the schema inscription of 'X' to define an end point of the continuum. The various inscriptions of a single word are, he believes, identical in meaning. But the identity the schema secures is syntactic, not semantic. What the various inscriptions of 'aardvark' share is their spelling. And sameness of spelling does not align with sameness of mention-selection. Indeed, syntactic replicas may be mention-selectively ambiguous. Some inscriptions of 'Linus', for example, mention-select mythical-Greekfigure-descriptions; others, contemporary-cartoon-character-descriptions. Some inscriptions of 'green centaur' mention-select unsophisticated-centaur-descriptions; others, bilious-complexioned-centaur-descriptions. 6 Synonymy, the limiting case of likeness of meaning, then consists in exact agreement in primary extension and mentions-selection. Sameness of spelling is not sufficient, even in the limiting case. My disagreements with Heydrich concern points of detail. They do not, I hope, overshadow my enthusiasm for his revision of the criterion of likeness of meaning. He has, I believe, eliminated a long-standing weakness in the Goodmanian position.
OUTSTANDING
4.
TO OLIVER
PROBLEMS
135
SCHOLZ
In 'When is a Picture?', Oliver Scholz carefully elucidates the difficulties philosophers confront when trying to talk informatively about pictures. Often philosophers don't. For we lack an appropriate forum for such discussions. Philosophy of language restricts itself to verbal symbols. Aesthetics is indifferent to pictures that are not works of art. But, as Scholz rightly points out, most pictures don't even pretend to be art. It would be wrong to subject news photographs, scientific illustrations, or mug shots to aesthetic standards. We need to know what standards to subject them to, instead. This involves understanding where and how they function, what interests they are supposed to serve. Considerations like Scholz's lead Goodman and me to recommend that epistemology be broadened to comprehend the cognitive contributions of pictures, and that philosophy of language be embedded in a more comprehensive study of symbols of all sorts. Such a reconfiguration of the philosophical domain would enable us to enhance our understanding of pictures. And by fostering comparisons between verbal and pictorial symbols, it would probably engender a deeper understanding of how verbal symbols function as welt. Following Goodman, Scholz argues that the distinction between the pictorial and the verbal should be drawn syntactically. Not surprisingly, I agree. But his argument against a semantic solution suggests that our theory of symbols motivates a further reconception. Semantics, Scholz urges, cannot draw the wanted distinction because it cannot account for non-denoting pictures. As semantics is currently conceived, this is surely right. But if, as Goodman and I maintain, reference is not limited to denotation, semantics has a wider range than is commonly recognized. Exemplification, expression, allusion, and the like fail within its jurisdiction. The study of semantics then should be extended to comprehend reference of all kinds. Although there are non-denoting pictures, I am not sure there are pictures that fail to refer altogether. This would need to be investigated, once a suitable semantics was in hand. Scholz ends his paper with a lucid explication of "Representation Re-presented". The conclusion of that study is that "a symbol functions as a picture only when taken as a character in a full pictorial scheme ''7. This suffices to differentiate the pictorial from the verbal. But it does not distinguish pictures from other analog symbols - maps, diagrams,
136
CATHERINE
Z. ELGIN
charts, and the like. Both the semantic density of exemplificational and denotational fields, and the relative repleteness of the symbols involved, I suspect, figure in that distinction. But before we are in a position to say 'a symbol functions as a picture when and only when.. 2, much work needs to be done. It is a virtue of Scholz's paper that it makes painfully clear how many problems remain to be solved. 5.
TO W O L F G A N G
KI]INNE
Whether, as G o o d m a n and I charge, "the faults of truth are many and grave", 8 depends on what truth is supposed to do. A faulty clock, incapable of telling time, may make an excellent paperweight. Someone who only wanted a paperweight anyway would find no fault with it. Both c o m m o n sense and philosophical lore seem to favor a correspondence theory of truth. It is truth so conceived that we find faulty. Its espoused function is to connect language to the mind-independent world. On such a construal, truth is radically non-epistemic. But if the world is as it is, regardless of our characterization of it, truth may elude us, no matter how good our evidence. Truth as correspondence then invites skepticism. Unfortunate as this may be, it hardly constitutes a refutation. For nothing guarantees that truth can be known. The problem with the correspondence theory lies elsewhere. If truth consists in correspondence to a mind-independent world, it is indifferent not only to evidence we muster but also to boundaries we draw. But surely whether Asteroids are small planets or Asteroids are smaller than planets is true depends on how we choose to classify celestial bodies. To insist that there is a fact of the matter independent of our choices requires holding that the world itself decides whether asteroids belong in the extension of the term 'planet'. What in the world could settle the issue? Systems of classification are human constructs. The truths there are depend on categories we contrive, on boundaries we draw. In the absence of a suitable classificatory system, there would be neither truths nor falsehoods about asteroids. By constructing category schemes then,
OUTSTANDING
PROBLEMS
137
we fix the range of truth and falsity. We do not, of course, simply declare our favorite sentences true. A category scheme sets the criteria for the instantiation of its terms. It is then a question of fact whether anything satisfies those criteria. Human involvement in the construction of truths and falsehoods does not guarantee the epistemic accessibility of truth. We might easily contrive categories whose instantiation we could never ascertain - for example, categories requiring absolute precision in measurement. But our involvement insures that epistemic accessibility is within our power to secure - provided we're willing to pay the price. In 'Truth, Rightness, and Permanent Acceptability', Wolfgang Ktinne argues that if it requires identifying truth with permanent acceptability, the price is too high. For then we must reject seemingly valid inferences and hold that whatever our best evidence supports is true. Dummett is evidently willing to accept such conclusions. After a brief flirtation, Goodman decided he was not. Goodman then proposed replacing truth with permanent acceptability. He suggested that the tasks philosophy expects truth to perform are better done by permanent acceptability. The prospect of unknowable truths and falsehoods is no longer ominous. For to demote truth is not to deny it. Evaluation of the feasibility of this proposal would require surveying the tasks in question to see whether permanently acceptable sentences prove more philosophically fruitful than (perhaps unknowable) truths. The answer is not obvious. Our position in Reconceptions is more radical still. In the shift from truth to permanent acceptability, declarative sentences remained preeminent. When we move to rightness, that preeminence is lost. Rightness, as Kfinne illustrates, pertains to symbols of all sorts. Pictures, gestures, and other non-verbal symbols, as well as verbal symbols like questions, instructions, and exhortations do not admit of truth value, but may be right or wrong. Rightness, moreover, is multifaceted. So we cannot simply identify truth with the rightness proper to declarative sentences. Truth often has a bearing on a declarative sentence's rightness. But other factors also come into play and may override the claims of truth. Some falsehoods are right; some truths wrong. Kfinne shows that the rightness of falsehoods and the wrongness of truths can sometimes be explained by truths and falsehoods at another level. This insight deserves further study. Is it generally the case? Or
138
CATHERINE
Z. ELGIN
do other modes of rightness and wrongness at different levels have the same effect? On the reconception Goodman and I advocate, there are a multiplicity of cognitive values and vehicles. Philosophy needs to investigate the relations among them, the enterprises they foster, the means they support, and the ends they promote. Truth is marginalized. Rightness takes center stage. Much work remains to be done. 6.
TO DIRK KOPPELBERG
To believe as many truths as possible and disbelieve as many falsehoods as possible - this is widely held to be our epistemic end. It is, I contend, an unworthy one. If numbers alone count, Watson fares better than Holmes. To give Holmes his due, epistemology must accommodate quality as well as quantity. The truths that Watson believes and the falsehoods he disbelieves are utterly pedestrian. Holmes courts error when he refines categories, entertains options, and appreciates complexities beyond Watson's ken. Even if he ends up with a lower proportion of true beliefs than Watson, the ones he has are epistemically more valuable. Despite the criticisms he raises in 'Should We Replace Knowledge by Understanding?', Dirk Koppelberg evidently agrees. He recognizes that "we would rank Holmes' knowledge about wines much higher than Watson's because Holmes' knowledge will be measured along a much broader and a much more detailed range of relevant alternatives", and that Watson "knows a triviality which Holmes would not find worth knowing". 9 It's not that Holmes knows more, but that Holmes' knowledge is worth more. Koppelberg maintains, however, that I am too quick to grant Watson a numerical advantage. Even if Holmes can't tell whether he's drinking a Margaux or a St. Julien, Koppelberg suggests, surely he knows he's drinking one or the other. We should then credit him with knowledge of the disjunction. Not necessarily. Holmes, let us agree, has the discriminatory capacity to know the disjunction. But knowledge requires belief. And if Holmes is disinclined to clutter his mind with ad hoc faUback positions, he has no incentive to generate the disjunctive belief. The issue is not what he could know, but what he does know. That depends on what he in fact believes.
OUTSTANDING
PROBLEMS
139
If belief causes difficulties, Koppelberg suggests, we might substitute discriminative capacity as the fundamental epistemic notion. This would improve Holmes' epistemic standing and would allow for ascriptions of knowledge where we are loathe to ascribe beliefs. I agree that the capacity to discriminate is epistemically important. Other capacities are, too - for example, the capacities to synthesize, interpolate, extrapolate, organize, interpret, even the capacity to judiciously overlook. The epistemic utility of all of these is predicated on their being effectively deployed. A Watson endowed with the capacity to draw an immense number of tedious, trivial distinctions would hardly merit epistemic esteem. To expiain why Holmes' knowledge is in general preferable to Watson's, I urged, we need to introduce considerations of epistemic quality. To explain why, despite its failure to generate knowledge, Holmes' wine-tasting experience is epistemically richer than Watson's, we need to do more. We need, Goodman and I contend, to shift epistemology's focus from knowledge to understanding. In the options he entertains, the considerations he adduces, the features he attends to, the comparisons he makes, Holmes exhibits a deeper understanding of wines than Watson. Epistemology should, we maintain, provide an account of what Holmes has and Watson lacks, how it functions, and why it is cognitively valuable even when it does not eventuate in knowledge. Koppelberg retains 'knowledge' as the term for epistemic achievement under the shift in focus he envisages. Goodman and i advocate an adjustment in terminology to accompany our change in view. There is, I believe, no fact of the matter as to which usage is correct. (Conceivably both are, since the modification Koppelberg entertains is less drastic than ours.) What Goodman and I call understanding has important affinities with what has long counted as knowledge. Equally important are the differences between the two states. But the connection of knowledge to true belief is deeply ingrained in philosophicai tradition. So to retain the term while attempting to sever the connection seems to us strategically unwise. Such a policy would invite misunderstandings that a terminological revision can forestall. Over the past quarter century, the analysis of knowledge has absorbed enormous amounts of talented philosophers' time and attention. If the mere agglomeration of knowledge were the ultimate epistemic end, determining exactly what qualifies as knowledge would doubtless be a good thing. But if, as I argue, indiscriminate acquisition of knowl-
140
CATHERINE Z. ELGIN
edge inhibits the realization of other epistemic ends, the importance of analyzing knowledge wanes. Epistemology would do better to explore the full range of epistemic values, interests, and goals to see what combinations of cognitive excellences are worth having. That is the reconception of epistemology's task that Goodman and I envision. NOTES 1 The papers in this volume were first presented at a Colloquium on Reconceptions at the Center for Interdisciplinary Research (ZiF) of the University of Bielefeld, organized by Peter Bieri and Oliver Scholz in 1991. Nelson Goodman and I are immensely grateful to the center for its hospitality and to the participants for their careful attention to our work. a Cf. Bas van Fraassen: 1980, Scientific Image, Clarendon Press, Oxford. 3 I am indebted to Robert Schwartz for this example. 4 Nelson Goodman: 1972, 'On Likeness of Meaning' and 'On Some Differences About Meaning', in his Problems and Projects, Hackett, Indianapolis, pp. 221-30 and 231-38; W. V. O. Quine: 1980, 'Two Dogmas of Empiricism', in his From a Logical Point of View, Harvard University Press, Cambridge, Massachusetts, pp. 20-46; Morton White: 1973, 'The Analytic and the Synthetic: An Untenable Dualism', Pragmatism and the American Mind, Oxford University Press, Oxford, pp. 121-37. s Goodman, 'On Some Differences About Meaning', pp. 237-38. 6 Israel Scheffler: 1979, Beyond the Letter, Routledge, London, pp. 32-36. 7 Nelson Goodman and Catherine Z. Elgin: 1988. Reconceptions, Hackett, Indianapolis, p. 130. 8 Ibid., p. 154. 9 Dirk Koppelberg: 1993, 'Should We Replace Knowledge by Understanding?', Synthese 95(1). 10 Farmcrest Avenue Lexington, MA 02173 U.S.A.