James F. Ross

Preface

The same word often has different but related meanings (e. g. catchl fish; catch/ball). Some philosophers have proposed that shades and kinds of such differences can be explained systematically. Aristotle and Aquinas developed theories of analogy to account for such phenomena, theories they applied to the philosophy of language, metaphysics, philosophy of science, philosophy of nature, theology and ethics, and that distinctively characterized their philosophical styles and even their products, such as analogous definitions.

After investigating the analogy doctrines historically, (Ross 1958, 1961, 1962a and b) and applying them to religious discourse (1970b), I began to examine whether the 'classical account', the account developed from Plato and Aristotle through Aquinas and Cajetan (1498), approximates the truth or points toward truths about meaning with similarly comprehensive consequences for contemporary philosophy.

By 1970 it emerged that the analogy phenomena have to be comprehensively redescribed because the classical theory suffers from limitations of scope and perspective (see chapter 1) and is based upon false premises, chiefly: (1) that word meanings are ideas(concepts-,thoughts-) in-the-mind-signified-by-conventionalsounds; (2) that sentence meaning is the molecular sum (syncategorematically computed) of the atomic meanings of the component words; (3) that different meanings for the same word are formally caused by differences in the referents, differences that are reflected in the conceptions, formed by abstraction, that are the meanings of the words; and (4) that only some words are analogical, equivocal, and the like.

In fact, the key assumptions and metaphors of the classical story about analogy were exhausted, as far as fruitful theoretical elaboration is concerned, by the time Cajetan produced De Nominum Analogia in 1498, the last systematic explanation of analogy of meaning since the middle ages.

And so, over the last decade I have undertaken a new investigation, exploiting Aristotle's, Aquinas' and Cajetan's insights into the analogy phenomena and into the profound consequences of such phenomena for philosophy, to identify an explanatory structure inherent in natural language and to trace some of its ramifications for philosophical analysis. At least to a first approximation, I have found an explanation for analogy of meaning, both denominative and proportional, and have identified some very important consequences for philosophy.

I thank the many people who have influenced this work, especially Roderick M. Chisholm, a model of philosophical excellence, who suggested my inquiring into analogy in 1957 and is surely an originating cause.

Felix Alluntis, O.F.M., first introduced me, as anundergraduate, to Duns Scotus in 1951 and emphasized the importance of understanding the analogy debate. Arthur T. Geoghegan, another teacher and friend for thirty years, taught me as a teenager that a good thinker develops positions of his own, enriched with historical perspective - something also encouraged by the late Francis P. Clarke, my friend and predecessor at the University of Pennsylvania. A number of friends helped by their obvious respect for the idea that contemporary thought can be enriched with distillations from medieval achievements, among them my colleagues Charles Kahn and Elizabeth Flower, Norris Clarke, S.J., Mary T. Clark, C.S.S.J., Alan Wolter, O.F.M. and Ronald Santoni and his colleagues at Dennison University.

I particularly thank the generous philosophers who examined my work in their papers - George Mavrodes, William Rowe, David Burrell, L. S. Lyons, Tobias Chapman, D. Griffin, Charles J. Kelley, Kai Nielsen, John Donnelley, James J. Heaney, Norris Clarke, Edwin Curley, Edward Weirenga, Robert Oakes andJohn Hick.

I have had only one discussant in depth on these matters, my friend Margaret Wilson, who astutely noted my lapses from promised explanations and diagnosed some counterintuitive classifications of examples, while steadfastly nourishing my confidence. She recommended that I find some form of publication before I killed interest in the material by baroque elaboration. For her faith that something worthwhile would eventuate, and for other things, I cannot express thanks enough.

Otto Springer, Vartan Gregorian and Claude Welsh (an exemplar of a friend) and David Goddard brought the resources of the University of Pennsylvania to my assistance for leaves of absence when help was indispensible. Similarly, I thank Morton White who supported my 1976 year at The Institute for Advanced Study,* Princeton, where I thought about analogy while inquiring into the relationship between natural rights and natural law (results forthcoming in due time).

I particularly acknowledge Nelson Goodman's Languages of Art, (1968) that made my world partially the product of his art and discourse. Phillip Quinn discussed a version of this material, responding with written suggestions that, along with the remarks of other members of the 1977 NEH Summer Seminar at Brown ' (Faith, Meaning and Religious Knowledge)t led to a substantial rewriting of the book. Chris Bache, a self-styled stowaway at the mentioned Seminar, disputed me persistently and developed the opinion that analogy is the coral reef of dead metaphors. Some of the many contributions of seminar members are mentioned fro.m time to time in text and notes.

Many secretaries helped. Janet Wolff did the first major typing, followed by Sherri Jones. Duane Williams, a philosopher with more important things to do, set up the basic typescript that has since gone through several rewritings. My daughter, Ellen, typed directly from my henscratch in 1977 and collated drafts of the footnotes in six different manuscripts. Diane Stillwell and Lore Silverberg typed the antepenultimate draft. And the penultimate retyping was done by Kathy McClendon, Mrs Ruth Hugo, Marie Cannon, Margaret Rose Steeg, and Sandy Natson. The final typescript was prepared by Sandy Natson. My research at the Institute for Advanced Study was supported by an N.E.H. Grant to the Institute, no.FC 1050

t N.E.H. Grant award no.FA-26537-77-163 to Brown University

Introduction

Why look for a systematic account of the meaning relationships different tokens of the same word have to one another (drop/stitch, drop/friend, drop/hint) when hardly anyone seems to have missed having one for five hundred years?

First, something with comprehensive philosophical consequences may have been overlooked. Both Aristotle's and the medieval Aristotelean writers' accounts of analogy (of diverse but related same-word meaning) affected the whole of those philosophies.

Secondly, the linguistic phenomena are pretty much as the classical writers said they were, though their explanation is not right. Such extensive data fairly beg us to probe for the hard bone of law underlying 'focal meaning', 'family resembling terms', 'systematically ambiguous expressions' and the panorama of verbal proportionality, metaphor, denomination and paronymy, remarked by philosophers of all epochs and persuasions.

Thirdly, there is no competitor for the classical account, despite its false foundations, mentioned in the Preface, and its limitations of scope and perspective, described in chapter 1. We need a competing portrayal, preferably one that construes the superficially chaotic data as the logical outcome of simple linguistic universals.

Fourthly, the analogy phenomena require revisions in the philosophy of language and probably in the philosophy of mind as well. A systematic study of same words refutes the assumptions that sentence meanings are the syncategorematic 'sum' of independent word/morpheme meanings (like a wall assembled out of varied stones) and that an ideal dictionary would have an entry for every meaning of every word in the language. Instead, the analogy phenomena and the 'software' that creates them support 'infinite polysemy' hypotheses (cf. Weinreich 1971: 322, and Lyons 1977: §13.4, 550f) and establish that one expression affects the meaning of others concatenated with it. 2 Further, the analogy phenomena disconfirm that sentences have truth conditions that are their senses, and that knowing its truth condition is necessary for knowing the sense of a sentence, and so require a reassessment of the objectives and capabilities of truth-conditional analysis (see chapter 8).

Moreover, opinions about the relationships of thought to language - like the revived Ockhamist 'language of thought' idea (Fodor 1975) - that suppose meaning to be affixed to inscriptions and sounds by their relationship(s) to something else (whether 'sentence-like mental representations', thoughts, or denoted things) will fail, as the medieval 'concept' theory of meaning did, to explain the analogy phenomena.

Instead, we have to work from the premise that linguistic meaning is intrinsic to acceptable expressions in natural language. It is not some extrinsic relationship of the parts, one by one or all together, to thoughts or other things, but is the contrasting combinatorial acceptability among the distinct words of the language.

Lastly, if analogy of meaning is even substantially as I construe it, there are consequences for what philosophical analysis can do; for instance, a philosopher cannot provide a single truth-conditional analysis of what 'being a cause', 'knowing', 'being thought' (and most of the other things philosophers want to know about) really is. Yet we can refine a practicable relationship between analysis and analysandum and we can find a place in philosophy again for fashioning 'analogous' definitions. So there are important incentives for reopening this old topic.

Although I use the classical account (summarized variously atpp. 17-19, 165, below) to identify and initially classify the data that need an explanation, I do not absorb it into the present theory. That is because it assumes, wrongly, that same-word

meaning relationships do not result from linguistic structure, but from the relationships of thought to things. Yet, without the classical observations and hypotheses, I doubt that I would have found the phenomena coherent enough to demand explaining (after all,

Wittgenstein's identifying family resembling terms did not stir demands for their explanation) or have realized how much rethinking about linguistic meaning philosophers still have to do. So, I frequently mention what Aristotle, Aquinas and Cajetan said, not as

'scholarly boilerplate' but because their perspective upon the issues is still revealing.

I. EXPLANATORY PORTRAYAL

This is a typical explanation, one that (i) conceptually aligns and classifies observed phenomena and (ii) identifies an underlying structure that (iii) logically (or causally) determines the observed phenomena to be as they are (in the relevant respects needing the explanation). As far as it is practicable, I describe the explanatory structure in terms of the broadest explanatory factors known in the area of inquiry

It is not explanation in the sense of finding the (efficient) cause of a happening - e.g. explaining someone's being taken sick or how an accident or fire happened; this is a structural explanation of a kind of happening, like an explanation of what hepatitis is in terms of what the liver is and how it functions.

I call it a 'portrayal' because the explanation's key predicates are metaphorically extended from other realms of discourse and are not yet naturalized citizens in talk about meaning, the way 'attraction' and 'force' became long ago in physics. The metaphor, as Aristotle remarked, 'puts the matter before the eyes' (Rhetoric, III, ch. 2 and ch. 11). It reveals the presence of law where before there appeared only disorder, just as a certain profusion of dots with the right perceptual set causes the faces of the Presidents or various animals to emerge. Thus, 'dominance', 'resistance', 'adaptation', 'contagion', 'intransigence', 'linguistic inertia' and 'linguistic force' are used metaphorically to draw a new picture of analogy.

The most important question for evaluating a new portrayal, I think, is this: What does it make us see in its subject that was not accessible to us before? In the present case it is that analogy of meaning (all the analogy phenomena including mere equivocation) is the consequence of an underlying synchronic structure, composed of (i) linguistic inertia (that words recur in the same meanings if nothing differentiates them) and (ii) linguisticforce (that words resist concatenating into unacceptable expressions by making step-wise meaning-adaptations, comparatively to other occurrences, to avoid doing so.)

The judgment that one portrayal is better than another is notjustified exclusively by the narrower considerations that apply to whether, within, say, a Kantian portrayal of knowledge, one account is 'a better explanation than another2; For, as in painting, comprehensiveness, inventiveness, discipline and an inexhaustible revelation (in the representation) of what is represented have more to do with significance and preferability than item by item truth. 'Science . . . is willing to accept a theory that vastly outreaches its evidential basis if that theory promises to exhibit an underlying order, a system of deep and simple connections among what had previously been a mass of disparate and multifarious facts.' (Carl Hempel 1966: 132.) Besides being as embracing as the classical account, I think this portrayal of analogy meets two conditions for being preferable: (i) the more you understand the portrayal, the more the phenomena appear as portrayed; (Goodman 1968: 33, reports someone's telling Picasso that his portrait of Gertrude Stein did not look like her; Picasso replied 'It will'); (ii) the more extensively this picture is compared with the classical one, the more the phenomena appear to be as portrayed here rather than as in that account.

Some of the layers in this account have involved project-bound decisions about the predicate scheme model (to be explained in chapter 3), about how to characterize predicate modes and about how extensively to use explicit definitions as a means of conceptual alignment; these are all matters that might have been handled differently, or more successfully, without much change in the overall observations or explanation. So, one has to distinguish the details that are incidental, temporary, dispensible and entangling, from the underlying insight into differentiation, its causes, and the systematic identification and distinction of analogy, simple metaphor, mere equivocation, denomination and paronymy. These things can survive a rather considerable reformulation of the whole explanation.

2 THE GENERIC PHENOMENON

Everyone who speaks one of the relevant natural languages (the living Indo-European ones, at least), from the youngest speaking children through the least intelligent persons capable of coheren discourse and the most sophisticated masters of the language, characteristically and automatically uses the same words in differen meanings, sometimes related (see/light, see/point; collect/books, collect/friends, collect/debts, collect/barnacles), and sometimes unre lated (charge/enemy, charge/battery, charge/account).

.The Merrill Linguistic Readers (Fries 1966), widely used for beginning reading, employ common words like'fix' in multiple senses to achieve variety of expression within a beginner's vocabulary: 'l:)addy will fix you', 'Fix my wagon', 'Fix my pants', 'Fix my dinner', 'make a bed', 'make a sandwich', 'make a mistake', 'make time for', 'make a plane', 'make an appointment', and so on. It is easier for a small child to comprehend the same word with multiple meanings than it is for him to distinguish the same number of visible word forms. So the child has already mastered response to some semantic structure by which distinct senses for the same word 18 are determined in context. What is that structure?

Further, the multiple English entries for a single Italian word in an English-Italian dictionary (J. Grassi 1933) illustrate its variety of meanings: 'levare' means 'to raise, lift up, take away, remove, carry, take off, raise up, lift, get out, buy (a large quantity), prohibit, levy.' (In English the words 'raise' and 'lift' pretty much divide the senses of the Italian word between them). And 'dare' means 'to give, hand, grant, permit, commit, appoint, announce, produce, yield, show, tell, stroke, dart, incline to, sell, buy...' Certainly native speakers know what'dare' means, so something determines which sense the word has. What is it?

Other dictionaries - French, Polish, Spanish, German - yield consonant results, as do dictionaries employed in reverse to give multiple foreign-word equivalents for English words. Which meaning a word has in a given occurrence is neither random nor haphazard, and yet what determines which meaning it has cannot be the speaker's intentions about what the words are to mean, either. For who, except perhaps a poet, has embracing intentions as to what ~ individual words are to mean? And who knows how by intention to determine what a word means (except to stipulate or to distinguish)? It cannot be done by intention alone. For a string of words can mean exactly what we intended them not to. We are looking, instead, for a universal linguistic force- some ubiquitous causation - that is present throughout the language (as inertia and gravitation are present throughout the universe) and is variously manifested by identifiable phenomena in all extended discourse.

Aristotle and Aquinas, as I said, recognized that the meaning relationships among same words are regular, classifiable (as mere -equivocation, analogy of proportionality, metaphor, or denomination) and systematically explainable; (their explanation was a theory of concept-generation). But no succeeding mainstream philosophers, except for Locke's interest in denomination (Essay III, III, 171 VI, 9 and III ch.25, 2-7), paid serious attention to accounting for same-word meanings, not even Wittgenstein after he turned up 'family resembling terms' while digging for other things.

The search for what determines which meaning a word has in a given context, and what accounts for its differences from one occurrence to another, gains the prospect of success once we notice that all the regular cases of a word's having diverse meanings have a common feature. (That doesn't include, of course, the initial cases of direct borrowing from another language and of stipulation and misuse - what Boethius, and then Cajetan, called 'equivocation a casu' (by chance) in contrast to 'equivocation a consilio' (by convention.) That common feature is differentiation (contrasting adaptation to context), fit to diversity of linguistic environment. Sections of -chapter 3 illustrate it and show how it is caused by 'dominance' (see below). Here I stress that differentiation has both homonymy and polysemy as outcomes. The classical writers, by failing to identify anything common to mere equivocation and analogy, missed uncovering the structual 'software' that generates the whole range of analogy phenomena and diachronically develops the expressive capacities of the language. (See sections 4 and 5 below.)

3 THE SPECIFIC ANALOGY PHENOMENA

There are five meaning relationships explained in this book. For i3 now, I will provide examples and brief descriptions with the warning that these descriptions skip essential qualifications offered later, and that the 'examples' (e.g. charge/account) are shorthand for examples in complete discourse environments. In general, I ask the reader to interpret the examples as if they were in complete environments that disambiguate them as I construe them.

(a) Mere equivocation Examples: charge/account, charge/enemy; watch (time), watch (duty crew); pen (writing), pen (enclosure); bank (verge), bank (depository). Description: same words (the same from the point of view of spelling and sound) that are unrelated in~ meaning. As will be shown, no near synonym of the one instance is -also a near synonym of the other, unless it, too, occurs merely equivocally.

(b) Analogy of proportionality Examples: collect/pension, collect/ books; create/time, create/trouble; establish (verify), establish (demonstrate). Description: same words, taken in pairs, that differ but are related in meaning. Some near synonyms, (say 'picked up' and 'received' for 'collect', 'make' for 'create', and 'prove' for 'establish'), of the one instance are also near synonyms for the other. Further, the difference in their meanings is marked by a difference in the group of meaning-related other words that can be substituted for the two instances. (I circumscribe the conditions for being meaning-related and for being substitutable later on). For now, it is enough to note that if we listed the near synonyms, the contraries, hyponyms, determinables, determinates, and the like, for each 3a instance, they would differ, even though there would be a definite overlap of common words. I call such a difference a 'predicative' difference and model it (chapter 3) with predicat~e schemes.

(c) Simple metaphor Examples: sow/seeds, sow/dissension; creepl 13 child, creep/disease; ftowlwater, ftow/conversation. Description: same words, taken in pairs, that differ but are related in meaning, that satisfy the analogy conditions in (b) above, and also satisfy an additional condition of asymmetry. Roughly, certain meaningrelated words substitutable for the non-metaphorical occurrence are not substitutable for the metaphorical occurrence, even though other words 'implied' by those very words are substitutable and are part of what is meant.

There is really no single statement adequate to convey this asymmetry until one sees it mapped out with examples (chapter 4). In any case, metaphorical occurrences are probably the most easily identifiable of the analogy phenomena, and meaning asymmetry with non-metaphorical instances is obvious even to persons who have not yet found a way to describe it.

(d) Denominative analogy Examples: brilliant/book, brilliant/writer; playpiano (habit), playpiano (ability); hat (object), hat (picture); plow (tool), plow (activity). Description: same words, taken in pairs, that 3a differ but are related in meaning, and (whether or not they differ in ~ ways already mentioned, predicatively), differ in predicate mode. ~ That is, they differ in so far as what is predicated inheres in what it is predicated of. For instance, 'He plays the piano' can attribute a proclivity, an ability or an occupation, and so forth. Because this sort of differentiation involves the mode of attribution, sometimes it is called 'analogy of attribution', (e. g. by Aquinas, Cajetan and others).

These phenomena are based upon 'relational naming' (denomination), characterizing something by way of relationships it bears to other things: Victorian/manners; brilliant/book, judge, husband. The denominative phenomena are astonishingly varied, ranging from interdefinable pairs, analogous by attribution (brilliant/book; brilliant/man) to 'representative analogy' where the representation gets the differentiated name of what is represented: 'the appellant' (his lawyer), 'the court' (the judge), 'the United States' (its ambassador) and 'the hat' (portion of a picture). (Cf. Plato, Par., 131; Rep., 596a, and Aristotle N.E., 1096a, 17ff; Meta., 1003a 33-1003b, 10; Cat., l a ff).

Comprehended within these four classes of same-word relationships are countless millions of same words, taken in pairs, in the contemporary corpus of the relevant natural language(s). Yet there is still another class of words, as impressively large as these, the morphological variants of one another: paronyms.

(e) Paronyms Examples: healthy/healthfull; explain/explanation; conceallconcealment; smokelsmoker. Description: paronyms are not same words but are variants upon the same root with meaning differences that correlate in a law-like way with the morphological variation. The meaning relationships among paronyms parallel the analogy phenomena among same words and are explained in the same ways.

That is to be expected, because analogy in one language (sana medicina, sana complexio) may be paronymy in a cognate language (healthAll/medicine; healthy/complexion).

The predicates of the analogy theory {is merely equivocal/is analogous by proportionality/is metaphorical/is analogous denominatively} differentiate, by analogy of proportionality, to encompass paronyms. 'Is univocal' drops out of the classification.

That enormously extends the theory's descriptive power, in effect, to all occurrences, taken in pairs, that are not distinct words, and powerfully demonstrates the truth of the theory's key premise: (roughly) that words take on different presuppositions (different affinities and oppositions to other words) depending upon what words you combine with them. For the key words of the analogy theory itself, when applied to paronyms, automatically adjust and no longer presuppose that the things classified are same words, but now presuppose on the contrary that the things classified are morphologically variants of one another. (Chapter 8 shows that differentiation does not, as a rule, generate a disjvfnctive meaning, as one might at first suspect.)

4 SOME BROAD BACKGROUND CONSIDERATIONS

Words differentiate because some words dominate others. One's interchanging contrasting completion expressions (say, the words 'eyes' and 'paint can' printed on little cards) in the same sentence frame (printed with an appropriate blank space) - 'She dropped her. . .' - will show that. No word dominates all other words and all words are dominated some of the time. Some words are dominated in a great variety of ways, especially common words like 'make', completed with 'time/trouble/way/appointment/bed/ money/merry/haste/cake/dinner/love/war'.

Dominance and indifference, illustrated extensively in chapters 2 and 4, are the observable, generic causes of the specific analogy phenomena (sketched in section 3 above) that are modeled and distinguished in chapters 4 and 5. Explanation at greater depth, though more conjectural and metaphorical, postulates that universal forces are manifested by the whole range of analogy phenomena. I describe that briefly, and then explain the principle that linguistic meaning determines the expressive capacity of sentences, that relative meaning (see p. 14 below) is logically antecedent to absolute meaning and that changes in absolute meaning can come about only through changes in relative meaning.

(a) Inertia and linguistirforce A word, recurring in the contemporary corpus of the language, without regard for time order, has the same meaning as any other given token unless something differentiates them. That is linyv.istic inertia.

Nevertheless, the undeniable fact - the one to be explained - is that the same word, far more often than one would initially expect, has different meanings when it recurs. For instance, recall the many meanings of the Italian dare listed above and the corresponding ranges for the English 'give' and 'take'.

A second fact: the meaning differences among distinct words

supervene upon their differences of combinatorial possibility, with the logical consequence that not all grammatically complete and correct strings of words are acceptable in a given environment, and almost every (perhaps every) such string is unacceptable in some environment.

Generally, words resist combining unacceptably in the linguistic environment, untilforced to. That is linguisticforce. In other words grammatical strings will not go together unacceptably (as 'not English') if there is any step-wise adaptation of word meanings (comparatively to their other occurrences in the corpus) which would result in an acceptable utterance and is not prevented by the environment. And those step-wise adaptations are the specific kinds of differentiation described in this book: analogy of proportionality, metaphor, denominative analogy, and mere equivocation. (Paronymy simply replicates those relations for morphological variants of one another.)

For the time being, think of contrasting adaptations as involving the suppression of affinities and oppositions to other words, relations that, if maintained, would make the resulting string unacceptable. Think of it as the suppression of which ones, among other words, are its synonyms, opposites (etc.). For example, 'collect' in 'collect/ pension' and 'collect/garbage' suppresses some affinity in each, for 'received as owed' and 'was paid' are near synonyms of the former and not of the latter. Suppressing affinity and opposition to some words is typically equivalent to acquiring such relations to still other words.

Diachronically, linguistic force has dramatic manifestations in continuously new senses of words as new combinations, new patterns of dominance, happen to occur. The automatic differentiation of the analogy predicates, when I applied the classification to paronyms, illustrates that.

The resistance I speak of is not all of the same kind and strength. Sometimes, when the concatenated meaning is inaccessible - such as 'Night sings posts', 'He lubricated his fallacies'- the string is unacceptable in a given environment and sometimes not, as with 'Cone No.6 doffed its tip in the peephole' (Updike 1980: 191) and'I image December's thorn screwed in a brow of holly' (Dylan Thomas 1934). Unacceptability, then, has various bases, and concomitantly, unacceptability variously based is variously resisted.

Although I do not investigate the bases of unacceptability or the diversity of resistance in this book, I estimate that chiefly and ubiquitously words resist failure to achieve an overall meaning. That happens when a word (or more) is simultaneously dominated to be both duck and rabbit; that is, in one meaning (and not the other) to hook up with what precedes and in another (and not the former) to hook up with what follows: 'Saturday followed Mary' for instance can be environmentally dominated so as to require'followed' to be temporal (and not physical) in order to hook up with 'Saturday' (the day), and to be physical (and not temporal) to hook up with 'Mary'. Such conflicting dominance prevents the overall concatenation that 'Friday followed Crusoe' achieves in Robinson Crusoe.

The lesser resistances, varying with the kind of discourse and the community of speakers, include weak resistance to commonplace falsehood, especially to those concatenations that deny the empirical beliefs incorporated into the word meanings that compose the sentence itself. So, 'cats are invisible', does not disambiguate to mean 'too small to see', as in 'atoms are invisible', unless dominant expressions in the environmentforce it to. (See Churchland 1979, on empirical belief elements of word meanings.)

This is a 'weak force', reflecting our collective experience with the world. It is undeniably much weaker than (and gives way to) resistance to 'no hook up at all', but it is still a driving energy modifying dominance. Its single effects are localized, reflecting particular portions of experience, but like the force that deflects neutrinos, it is, in the cosmos of discourse, a starbuilder.

And so, with various resistances to variously based unacceptability, sentences make what sense they can. That trivial truth, like 'apples fall', realizes universal law: inertia and resistance to unacceptability. Its synchronic manifestations are the species of analogy and equivocation, described in this book, and its diachronic manifestation is continuous evolution of expressive capacity.

(b) Intrinsic meaning, expressive capacity and media of thought There simply could not be phenomena of the kinds I have described, which manifest an inertia-resistance structure that generates endless expansion of expressive capacity, and which depend only upon how words happen to be combined, if linguistic meaning were not something intrinsic to the language and distinct from speaker-meaning, referential meaning, ideas in the mind, and sense-meaning (Lewis 1962). Linguistic meaning either is, or is the cause of, contrastive

expressive capacity. I suspect that contrastive expessive capacity is realized by (but is not the same as) the contrasting combinatorial possibilities of distinct expressions, and that linguistic meaning is prior to use the way capacity is prior to function.

Contrasts of linguistic meaning could be described as contrasts of semantic features of distinct expressions (Fodor and Katz 1964; and Katz 1966, 1972) if we could identify a rich enough set of features and if we were to adapt the semantic-feature analysis to deal with infinite polysemy. Even more congenial to my account is Lyons' (1977: 512-70) idea that the lexicon consists of patterns of affinities and oppositions of meaning. For a general account of meaning (and thought) that can accommodate the analogy phenomena has to explain how 'horse' and 'husband' differ within the language, apart from what they stand for and apart from what horse-concept or husband-concept, or idea, someone thinks he has. For if they occurred in an undeciphered language, the symbols would still have contrasting linguistic meaning, even if in our culture we had no use for such words or no corresponding ideas.

Obviously any natural expression has limited expressive capacity. Limits can be altered by altering the component words and by altering the meaning of a component word; for example, by causing 'sober' in'The captain was sober today' to alter, by dominance, from 'not ebullient' to 'not intoxicated'. Difference of words correlates with difference of expressive capacity. 'Snow is white' can't express, as distinct from 'encode', what 'God is one' does (though no one has yet satisfactorily said why). And that difference is independent of what you or I think the words stand for or mean or what 'idea' we think we have. An adequate account of meaning has to explain what it is to have linguistic meaning and how it controls expressive capacity.

Because meaning determines expressive capacity, the words of an analysandum affect the formulation of the analysis (as I explain in chapter 8). Besides, I call attention to the dependence of expressive capacity upon inherent linguistic meaning because thought in words subsists in, and is limited by, the expressive capacity of interacting words.

Expressing oneself in one's natural language(s) is a medium of thought. Some of our thoughts subsist in their verbal expression the way a Wyeth painting subsists in egg tempera on board, having no existence apart from it and yet not being made out of it, as a chair is made out of wood, but being made in it, as joy can be, not merely be exhibited, in a dancer's movement.

Words, of course, are not the only media of thought. There are sketchings, and makings, and doings, and imaginings, and patternings in which we judge, question, assert, doubt, associate and draw conclusions. But only some of those are media for expression. A medium for expression is a means of communication through discriminable, inherent, contrastive possibilities for articulation.

Just how contrasting possibilities within a medium take on expressive actuality, I cannot even guess at present. But I do know that what can be thought in words, like what can be said, is limited to the expressive capacity of the symbols. And, as we now know, the expressive capacity of words can be modulated by new collocations of words - that is, by differentiation. In fact, poets and other creative writers arrange words to exploit evolving expressive capacity with new thoughts.

Philosophical theories that treat thoughts as independent of words and words as independent of one another are doomed to fail to explain the analogy phenomena and doomed to misapprehend what meaning is. The combinatorial differences of different words are the basis for their contrasting relative meanings, and the contrasts among such words are the foundation for the distinct relative meanings of same words. Put simply, the different meanings of the same word correspond to combinations of different words; and the distinct meanings of different words are founded upon the difference between the acceptable combinations of the one and those of the others. And so, for there to be contrasting combinatorial possibilities for two different words, some combinations that fit the grammar have to be unacceptable, either absolutely or environmentally, (otherwise there would be no difference of acceptable combinations and thus no basis for linguistic meaning difference). But unlike artificial languages, expressions in natural languages resist unacceptable new collocations by surrendering affinities and oppositions to other words and acquiring new ones. Differentiation of meaning and evolving expressive capacity are the logical consequence.

The combinatorial differences that I sketched above logically precede (and sometimes determine) application of the symbols to classify (represent) a world (realm) into arrays of referents, just as limited spelling and sound combinations (e.g. 'order counts') logically precede arrangements of letters into words. Because the combination of words can alter the meanings of the combined words (creating new combinatorial possibilities), it can and does alter the realms and ranges of reference.

5 ABSOLUTE AND RELATIVE MEANING

This book is about relations of relative meaning. I do not show how symbols got the first meanings they ever had, or how they came to have the meanings they have now, nor do I trace the differentiations, diachronically, from one meaning to another. We already know that linguistic force (resistance to unacceptability by step-wise suppression of affinities and oppositions to other words that would be unacceptable in the environment) will generate new meanings, both analogous and merely equivocal, as words happen in new configurations. We don't need to trace chains of individual differentiations to prove that. Theories that explain what meaning is and how symbols acquire it, and particularly, how symbols are related both to things in the world and to thought, I call theories of absolute meaning.

My hypotheses are about sameness and difference of same-word meaning and how they come about and how they are related to the combinatorial possibilities of distinct words. I refrain from more than initial and tentative conjectures about absolute meaning (see 4b, above) because I am certain that something is fundamentally wrong with all the theories of absolute meaning I have ever heard of. They do not account for (and usually do not recognize) the analogy phenomena; they provide no account of 'linguistic meaning' - meaning internal to the language; they provide no account of the fact that utterances and inscriptions can express (not merely encode3 thoughts, and no account of the fact that we (often) think in words (as a painter thinks in shapes, contrasts and color) and not merely with words, as with tools. I venture that absolute meaning is to a far greater extent dependent upon antecedent relative meaning (internally acceptable combinatorial patterns) than the reverse, and that the idea that primitive humans endowed grunts and twitters with ranges of referents (and subsequently with rules of combination) is pure superstition.

One can dismiss these broad speculations and still understand my description of the analogy phenomena and appreciate their significance for philosophy. One does not have to be willing to revamp philosophy of language and philosophy of mind to find this material useful.

6 PLAN OF THE BOOK

Limitations of the classical account and the relationship of analogy to verbal proportionalities (like the 'verbal analogies' tests for college students) are explained in chapter 1. Then, differentiation and dominance are explained and illustrated (chapter 2). The predicate scheme model is developed (chapter 3), to allow for a more rigorous statement of dominance claims and to ground the 'near-synonym' shorthand for predicate-scheme relations, (I have already used that shorthand in section 2, above). Chapter 4 distinguishes, by means of predicate schemes, mere equivocation, analogy of proportionality, and simple metaphor and explains how analogy produces clustered meanings of words, how linguistic abstraction differs from metaphor, and what categorial contrast of meaning consists in. An interesting outcome is that a categorial contrast of contexts differentiates all common completion words, but not all of them merely equivocally. (That is important, as Aristotle realized, for metaphysics.)

The many kinds of denominative analogy are explored in the first half of chapter 5 where I speculate about 'inherence' to clarify 'predicate modes'. In the second half, I extend the analogy theory to paronyms, as already mentioned. Then in chapter 6, to support the hypothesis that relationships among same words are adequately encompassed by the five-fold classification already mentioned, I explore figures of speech and figurative discourse. It turns out that figurative or 'heightened' discourse is typically sprinkled with words whose shortest hypothetical sequence of step-wise differentiations (from typical instances outside such heightened discourse) involve looth simple metaphor and another kind of analogy. Thus, I say that figurative discourse is characterized, semantically, by 'double differentiation'.

The account of analogy is applied, in chapters 7 and 8, particularly to explaining the embedding of craftbound discourse in unbound discourse and to a scrutiny of truth-conditional analysis. Certain disputes about the cognitive content of religious and metaphysical discourse can be resolved simply by applying the analogy information. Chapter 7 concentrates upon the cognitivity issue for religious discourse, with collateral attention to legal discourse. The same strategy can be (and should be) adapted to showing meaning continuities between 'theoretical' and 'observation' predicates in science and to displaying the cognitive pathway from discourse about the 'manifest image' of the world to discourse about 'the scientific reality'.

Chapter 8, assessing the objectives and capabilities of truthconditional analysis, rejects the position that to know what a sentence means one has to know its truth conditions. I argue that one cannot fix the sense of expressions under analysis or determine the scope of similar expressions that have to satisfy its truth conditions, without a benchmark3 occurrence in actual discourse.

Concerning whether the analysis has to be substitutable for the analysandum, or mean the same thing, or be merely equivalent to it, and so forth, I conclude that any articulation of what is meant that makes a cognitive advance can be satisfactory, while analyses that state meaning-relevant truth conditions are especially useful. (A condition is meaning-relevant just in case one of the words in the analysandum would have had a different meaning if that condition had not been necessary for the truth of what the analysandum is used to state.) Further, there cannot be a single truth-conditional analysis of 'knowing', 'believing', or practically anything else philosophers want to know about, because such predicates are projected analogously. J.astly, I illustrate analogous projection of predicates from legal discourse and demonstrate that there cannot be disjunctive definitions of analogous words.

When we compare the words used in science, metaphysics and religion with the same words in commonsense talk about the world, we notice how many common words are used analogously (as I explain that relation). There are not equally apt and specific words for us to apply univocally instead of the analogous ones; nor can we invent them. The philosophical significance of that matter has gone too long unexplored.

Chapter One

The limitations of classical analogy theory and the Miller's Analogies transition

The classical theory of analogy of meaning, analogia nominum, explained how same words are partly the same and partly different in meaning, ultimately and fundamentally, by appeal to relevant sameness and difference in the sorts of things to which the words are applied.'

Analogous terms differ from merely equivocal terms (which are also to some degree and in some respects partly the same in meaning) because the sameness and the difference in the analogous meanings are not merely coincidental, not merely haphazard, a casu, but are a consilio, conventional. In a law-like way they reflect metaphysical similarities and differences of the referents, that are preserved as similarities and differences in the concepts abstracted from such things. Those similar but differing concepts are the meanings of the words that are analogous.

The classical account of analogy involved several distinct inquiries, only the first of which will be pursued here. To use medieval Latin terms, there is analogia nominum, 'analogy of names' or 'analogy of meaning', a phenomenon I take to be ubiquitous, indigenous and characteristic in natural languages. It is an obvious characteristic by which natural languages differ from artificial languages and the most evident strength through which natural languages expand to incorporate new ideas and by which the language performs some of our thinking for Us.4 It is the phenomenon investigated and explained in this book.

There is also analogia rationis, 'analogy of reason', analogous thinking or reasoning; and there is analogia entis, 'analogy of being', analogy among things. To varying degrees, Plato, Aristotle, Aquinas, John of St. Thomas and Cajetan explored aspects of these issues as well. (See the extensive citations provided by Lytthens 1952.)

With respect to analogia rationis there are three main lines of inquiry: the logical, the psychological and the epistemological. The former concerns whether reasoning involving analogous terms has any validity, any degree of reliability either deductively or inductively (Bochenski 1948; Palmer 1973). The psychological problems concern how we recognize analogy among words and things, construct analogous reasoning and learn to employ natural language analogously. The epistemological aspects concern the justification of judgments that situations are analogous or that the same predicate applies analogously to contrasting situations, and how, in general, we come to know that things are analogous.

The analogia entis embraces all the ontological or metaphysical issues about whether things are really analogous to one another and if so, what makes them so, and all the other questions about the existence of mind-independent similarities among things, similarities that cannot be reduced to specific identities describable in words that apply univocally to the objects under discussion (Bochenski 1956). Plato, Aristotle and Aquinas (as well as many others) seem to have felt certain that not all similarities are founded upon identities that can be characterized with univocal predicates.

Occasionally recent philosophers have tried to resolve one or another of these issues; e.g. Bochenski (1948) attempted to provide a model of formal validity for analogous reasonings and inquired provocatively about 'sameness' and 'identity' (1956), and Palmer (1973) attempted to demonstrate the invalidity of all such reasoning about God.s But as long as analogy of meaning has not been sufficiently explained, satisfactory progress on the other issues is unlikely.

The objectives of a contemporary analogy theory, namely, to describe, distinguish and explain analogy, equivocation, metaphor, denomination, paronymy and figurative discourse, and to apply the explanation to the resolution of certain cognitivity disputes and to a more articulate understanding of philosophical analysis, differ very little from those of the classical theory. But the locus of the account, almost exclusively in linguistic regularities and without an 'idea in the mind' model for linguistic meaning, is entirely different.

Aquinas and Aristotle apparently thought that the natural language so closely reflects reality that the final step in explaining any meaning relationship must always lie in some entitative relationship. They did not look to the language for the explanation of the meaning relationships (analogy, mere equivocation and metaphor), but looked through the language, to what they regarded as ultimate.

Without begging the questions as to whether the 'final' explanation for analogy, equivocation and metaphor is to be sought in metaphysics, psychology, or physiology, and without begging questions as to the relationships of analogia nominum to analogia rationis and analogia entis, I provide an initial explanatory pattern that locates these meaning relationships within natural language itself. Four particular limitations of the classical theory are discussed below.

I LIMITATIONS OF CLASSICAL ANALOGY THEORY

The classical theory (1) gravely and implicitly misappraised the universality of differentiation for same words; (2) it provided only an inchoate and misdirected explanation, hardly more than an initial description of the phenomena; (3) it was a hybrid of synchronic and diachronic hypotheses; (4) it had so heavy a metaphysical and psychological bias as not really to be a linguistic theory at all. As Henry Hiz (1967:74) pointed out, criticizing Chomsky, there is an important question that must be answered linguistically: 'How does it happen that an ambiguous sentence ceases to be ambiguous when placed in the context of other sentences?' This book provides a partial answer.

I further agree with Hiz when he says, 'But it should be easier to explain why we assign such-and-such a structure to a sentence by pointing out how this sentence changes the readings of neighboring sentences than by referring to innate universal ideas and mental reality' (1967:74).

(a) The extensiveness of the phenomenon

As Aristotle noticed, things are said to be, no matter of what sort they are; and yet, not all are said to be, in the same sense. 'Some think "being" and "one" mean the same thing, while others solve the argument of Zeno and Parmenides by asserting that "being" and "one" are used in a number of senses.' (Soph. Ref., ch. 2, 182b-25)

For it must be either by an equivocation that we say these are [substances and the things of the other categories] or by adding to and taking from the meaning of iare' (in the way in which that which is not known is said to be known) - the truth being that we use the word neither ambiguously nor in the same sense, but just as we apply the word 'medical' by virtue of reference to one and the same thing, not meaning one and the same thing, nor yet speaking ambiguously; for a patient and an operation and an instrument are called medical neither by an ambiguity nor with a single meaning but with reference to a common end [italics added].

Aristotle also says (at 1003b, 10):

So too, there are many senses in which a thing is said to be, but all refer to one starting point; some things are said to be because they are substances, others because there are affections of substance, others because they are a process towards substance, or destructions or privations or qualities of substance, or productive or generative of substance or of things which are relative to substance, or negations of one of these things or of substance itself. It is for this reason that we say even of non-being that it is non-being. (See Aquinas' explanation, In Met., IV, 1, n.539.)

'To be' differs in meaning according to the category of the things that are said to be and is univocal only for things of the same category, (N.E., 1096a, 10-29.) Aristotle also recognized that other words, 'cause', 'principle', 'good', 'material', 'form', and many others (note his continual attention in Metaphysics to diverse senses of same terms) differ in meaning according to their general discourse contexts. So he inquired into the meaning relationships of same words and distinguished the univocal, the equivocal, and the meaning relationships that are intermediate, the analogous.

Aristotle did not discuss whether intermediate relationships of meaning are the rule or the exception. So, he did not postulate that all noun phrases and verb phrases and modifying expressions exhibit differences of meaning as their contexts are varied and that there must be a general explanation of that fact.

Similarly, Aquinas treated analogy of meaning as being characteristic of religious, scientific and philosophical discourse and, like Aristotle, used many examples from ordinary discourse. But he, too, failed to grasp that the underlying phenomenon is differentiation, and thereby left out the large-scale background against which the other limitations of the classical account would have appeared anomalous. Would they have continued to classify words, e. g. 'sharp', 'obscure', as 'equivocal' (as if all words were not), once they noticed that (a) every word has (or can have) merely equivocal occurrences, (b) every term has analogous occurrences and (c) every sortal has or can have metaphorical occurrences?

The classical analogy theory purported to account for scientific (philosophical) words that differ in meaning but do not differ as 'simple equivocals', e. g. 'to be' in various existential claims, 'thing', 'something', 'one', 'good', 'true', applied to all things no matter of what sort, 'knows', 'understands', 'loves', and'wills'as applied to rational animals and to intelligent beings essentially distinct, (Aquinas, Summa, Ia, 13), the application of'matter', 'form', 'accident', 'substance', etc. to quite different things.

Although the examples were varied, the classical theory was applied to escape theological and metaphysical difficulties that arise, on the one hand from considering that such pairs, applied to categorically contrasting things, mean exactly the same thing and, on the other hand, from considering that such terms are totally unrelated in meaning. The metaphysical and theological applications smothered the analogy theory by confusing and concealing the generic nature and the ubiquity of differentiation. Particularly, the behavior of simple (mere) equivocals was not noticed to be generically related to analogous and metaphorical terms and was not thought to require a similar explanation.

Even if one is asking the right questions, the nature of the phenomenon is still elusive. For instance, asking what is common to all cases of analogy, I replied earlier with full confidence (even after considerable reflective inquiry) 'meaning derivation', something by which analogy seems to differ from mere equivocation. I concluded therefore (Ross 1970a and b) that to construct a new analogy theory one needs primarily to find the regularities underlying derivation and, having found them, to apply such regularities to resolving the cognitivity puzzles over religious and theoretical discourse. Moreover, the idea that analogous meanings were differentiated because of the categorial contrast of their completion expressions (Ross 1970b) seemed persuasive because (a) it was illustrative of the emerging dominance phenomenon, (b) categorial contrasts were so important to Aquinas and Aristotle and (c) the kinds of analogy that most often interest philosophers involve words used for experienced things and in religion, metaphysics and science, for things categorially different. Even by 1971 it still had not become clear that analogy that involves categorial contrasts is only a sub-case of analogy in general and that 'derivative meaning'

is a misleading and inefficient conception of the overall phenomenon of analogy. A few years later, the importance of explaining the generic phenomenon, differentiation, finally emerged and eventually lead to the inertia-resistance structure underlying dominance.

The right way to begin description is not to consider analogy as derived meaning, but rather to consider analogy as a species of a common phenomenon that includes mere equivocation and is characteristic of natural languages and inevitable in the absence of semantic regularities that prevent it: namely, words di~erentiate in suitably contrasting environments. Satisfactory explanation for that must, logically, precede an account of derivative meanings.

Some may think 'arialogy of meaning' should not be treated as a species of differentiation because analogy, at least etymologically and by custom, connotes some sort of derivation ('per prius et posterius'), particularly some relation of similarity among the designata of the terms. But that would not prevent analogy from being a kind of differentiation. Furthermore, similarity of meaning is not the foundational characteristic of analogy of meaning; relatedness of meaning is (see chapter 4): mere equivocation is differentiation without relatedness; analogy is differentiation with relatedness.

(b) Limited explanation

The classical writers counted as 'explanations', hypotheses that were too limited in scope, erroneous in assumption and in some respects plainly irrelevant. They appealed to the ontology of the things referred to, and the heart of their account was an attempt to connect features of the world (similarity of things) with features of words (similarity of meaning) by hypotheses concerning the way in which concepts are formed (similarity of concepts). The classical theorists did not discuss the factors that determine which sense (or meaning) of a merely equivocal word ('pen') belongs to a given occurrence. That is because they considered linguistic meaning to be in the mind and to be the result of abstraction and not to be inherent in the written and spoken words.

My general picture is different. Meaning is inherently in the sentences just as law-like regularity is in nature. Grasping the meaning is a mental or quasi-mental phenomenon; having meaning is an inherent property of well-formed and acceptable expressions. Writing and speech are not encodings of one another or of something in the mind. Words are not signs of meanings (e.g. ideas), they mean (contrastively symbolize and combinatorially have expressive possibilities, actuality and limitations).

The fact that things of different categories differ, and yet are similar in their kinds of being and give rise to similar but differing concepts (as Aristotle and Aquinas supposed), explains why 'to be' and 'exists' differentiate when applied to such things only on the assumption that those similar but related concepts are the meanings. But the theory of concept formation does not account for the fact that the same word differentiates instead of being replaced with a distinct word having a similar meaning.

The classical psychology of concept formation does not explain adequately why we form the concepts that we do form rather than others in the presence of perceivable objects or why we have the words we do have and not others. Apparently it did not occur to the classical writers to investigate why we attach 'similar' concepts to different occurrences of the same word when we healthy is chronologically, psychologically or epistemologically prior to the sense in which his medicine or complexion is healthy. After all, the order of learning can be reversed, and even the order of the definitions can. At every crucial stage, the functioning of language is 'explained' by appeal to non-linguistic phenomena whose role as a meaning determinant is insufficiently justified.

Finally, the traditional theory cannot answer internal questions it generates, for instance: what characterizes the difference between a false statement involving an analogous predicate and a metaphor? What exactly is the difference between simple equivocation and analogy of meaning? How does analogy differ from metaphor? What accounts for the fact that the same word has different meanings when combined in diverse ways with certain other terms (law case/case law)? And if two terms T' and T2, are analogous to one another and T3 is analogous to T2, is it analogous to T and if not, why not?

A defender of the traditional theory might rejoin that I unfairly criticize it after having generalized the phenomenon to be explained far beyond Aristotle's and Aquinas pretensions. They considered cases of derivative meaning, but I have claimed that the phenomenon first to be explained is differentiation, a much larger class of linguistic relationships. I reply that one cannot explain derivation synchronically until we have accounted for differentiation For, we cannot say what 'derivation' consists in, what causes it, what determines the order among derivative meanings or what the meaning relationships among various derivatives might be.

'Healthy' in 'Fido's complexion is healthy' is 'derived' from the sense of'healthy'in'Fido is healthy'. That certainly does not mean, or even entail, that one must learn the one by first learning the other or that the second occurred before the other in someone's experience. The same holds for the statement that 'wise'in 'God is wise'is derived from 'wise' in 'Socrates is wise'. What does 'derived' mean here? Does it mean that we learned to talk about Socrates before God? Of course not. Nor does it mean that no one could learn that one meaning without first learning the other or that the word in one use is definitionally prior to the word in another use, in such a way that the order could not be reversed.

Aquinas' observations (In I Sent., 22, 1, 2, c; Cajetan 1498: ch. 11, no. 123) about psychological and ontological order among concepts are in effect misleading because they conflate the account of analogy often also attach equally 'similar' concepts to distinct words. Yet there must be an explanation and one that relies not merely upon the supposed intentions (or purposes) of speakers or upon the supposed analogies among things. Why is there not a common word for 'pounce', 'bone' and 'spine' (Aristotle, Post. An., 98a, 20f and Cajetan 1498, ch.ll, no. 117), when there is a common word for gravitational, chemically and electrically caused motion- 'attraction'? Do the former have less in common than the latter?

Even if the classical theory did succeed in explaining partial cawyover of meaning by overlap of concepts, that still leaves it unexplained that the same word is used in different but related meanings rather than different words.

The classical theory is inextricable from its commitment that the meaning of a word is an accompanying concept in the mind. The implausibility of that underlying account of meaning makes it impossible to repair the classical theory. Further, the classical theory, especially in Cajetan's exposition, (1498: ch.ll, no. 123), makes assumptions about the priority of meanings over one another that can be supported only on doubtful metaphysical premises, for instance, that 'exists', applied to God, is 'prior' in meaning to 'exists', applied to creatures because the being of God is ontologically prior to that of creatures (Aquinas, In. I Sent., 22,1,2 c.)

And there are also false assumptions about psychological and epistemic priority. For instance, that the sense in which a person is healthy is chronologically, psychologically or epistemologically prior to the sense in which his medicine or complexion is healthy. After all, the order of learning can be reversed, and even the order of the definitions can. At every crucial stage, the functioning of language is 'explained' by appeal to non-linguistic phenomena whose role as a meaning determinant is insufficiently justified.

Finally, the traditional theory cannot answer internal questions it generates, for instance: what characterizes the difference between a false statement involving an analogous predicate and a metaphor? What exactly is the difference between simple equivocation and analogy of meaning? How does analogy differ from metaphor? What accounts for the fact that the same word has different meanings when combined in diverse ways with certain other terms (law case/case law)? And if two terms Ti and T2, are analogous to one another and T3 is analogous to T2, is it analogous to Tiand if not, why not?

A defender of the traditional theory might rejoin that I unfairly criticize it after having generalized the phenomenon to be explained far beyond Aristotle's and Aquinas pretensions. They considered cases of derivative meaning, but I have claimed that the phenomenon first to be explained is differentiation, a much larger class of linguistic relationships. I reply that one cannot explain derivation synchronically until we have accounted for differentiation. For, we cannot say what 'derivation' consists in, what causes it, what determines the order among derivative meanings or what the meaning relationships among various derivatives might be.

'Healthy' in 'Fido's complexion is healthy' is 'derived' from the sense of'healthy'in 'Fido is healthy'. That certainly does not mean, or even entail, that one must learn the one by first learning the other or that the second occurred before the other in someone's experience. The same holds for the statement that 'wise' in 'God is wise' is derived from 'wise'in 'Socrates is wise'. What does 'derived' mean here? Does it mean that we learned to talk about Socrates before God? Of course not. Nor does it mean that no one could learn that one meaning without first learning the other or that the word in one use is definitionally prior to the word in another use, in such a way that the order could not be reversed.

Aquinas' observations (In I Sent., 22, 1, 2, c; Cajetan 1498: ch. 11, no. 123) about psychological and ontological order among concepts are in effect misleading because they conflate the account of analogy of meaning with analogia rationis. 'For the order of the term follows the order of knowledge ... but because we know that power through its effects, we name it from its effect' (C.G., 1,34.) The whole analysis of meaning relatedness as a relation per prius et posterius involves a confusion that persisted from Plato and Aritotle right into our time. There are such meaning relationships, but they are not the explanation of analogy or metaphor; rather, they are a result of analogy of meaning.

(c) Synchronic and diachronic phenomena

The classical theory is not explicit on whether analogy of meaning is, in its explanation, a synchronic or a diachronic phenomenon, though in my opinion its dominant thrust, with small lapses, is synchronic. Those elements that concern the relationship of words to their referents, via concepts reflecting the differences among referents, seem to treat analogy as synchronic, as the result of a constant and time-invariant causal structure. And the observations about psychological and epistemic priority and the order of abstraction seem to posit general causal mechanisms, not, of course, mechanisms controlling the evolution of languages but to explain the origin of and distinction among concepts.

Differentiation is obseruable within a static corpus of discourse; so its basic explanation ought to be by way of a time-invariant structure, just as Aristotle and Aquinas supposed. But because the theory of meaning is different, so will be the locus of explanation: 'Being analogous in meaning' is a relationship that holds pairwise among same words regardless of their temporal relationships, provided only that it is reasonable to regard the pairs as belonging to the same corpus of the language. (There are, of course,diachronic analogy phenomena, but they are results of and manifestations of synchronic linguistic structures.)

(d) Metaphysical bias

The classical theory of analogy was metaphysically biased because crucial elements of the theory were justified merely as applications of metaphysics. In fact, the theological, metaphysical, psychological and epistemological applications of the theory subordinated the elements of the philosophy of language.

Historically, it has been common to answer questions about meaning through metaphysics; Plato, Locke and Berkeley, paradigmatically, did so; and some contemporary nominalists, Goodman (1968), Quine (1960) and even a 'Cartesian' linguist, Chomsky O! (1972), have followed suit. But in formulating a new theory, I do not account for a difference in the meaning of'intelligent' in the sen- ~: tences 'God is intelligent', and 'Socrates is intelligent', by talking about the metaphysical difference between God and Socrates, but rather by talking only about the linguistic differences between 'God' and 'Socrates' and the linguistic environments of the utterances. No doubt the separation of purely linguistic and referential considerations cannot be perfect; issues about acceptability of sentences shade into issues about commonplace falsehood. Yet as far as possible, this form of explanation is to be pursued.

Some philosophers think we cannot escape appealing to properties of referred-to-things to account for differentiation because of the uniqueness of some referential expressions. Mavrodes (1970:747-55), for instance, argued that, on my premise (which he did not share), if 'intelligent' differs in meaning in the two cases above mentioned, 'intelligent' should also differ in meaning depending upon whether the utterance, 'The person St. Francis loved best is intelligent', refers to Francis' mother or to God. But, he argued, if the latter is true, then the theory cannot be developed without dependence upon the metaphysical differences of the referents. He argued, further, that if the term 'intelligent' does not alter its meaning on account of a change of the referent, we will encounter an anomaly because this syllogism, 'The person St. Francis loved best is intelligent; St Francis loved God best; therefore God is intelligent', has a conclusion where 'intelligent' differs in meaning from its use in 'St. Francis' mother was intelligent' but where the difference of meaning is not accounted for by anything.

Mavrodes vividly illustrated the difference between explaining differentiation by appeal to semantic dominance, e.g. by appealing to 'his mother' in the second premise of the one syllogism and to 'God' in the other, and explaining differentiation by appeal to differences in the things referred to. The classical analogy theory characteristically reasoned as Mavrodes illustrated, with metaphysical considerations, e.g. that God is transcendent, unperceivable, not in any space, incomprehensible, etc., while Socrates' intelligence is limited, confined to experiences arising in space and time. There is no need to explain such differences of meaning by invoking hypotheses about differences in the referents, as the sequel will show.

2 MILLER S ANALOGIES: A TRANSITION

(a) Proportionalities and analogy of meaning

In an improbable source, the College Board Tests (Miller's Analogies Test, 1969), I find a convenient transition from classical to contemporary theories of analogy, a transition that illuminates Aristotle's and Aquinas' otherwise puzzling correlation of quasi-mathematical proportionalities with analogy of meaning.

In my first expositions of Aquinas (1958 and 1961) I did not appreciate the connection between analogy of meaning and verbal proportionalities expressed in quasi-matbematical form (a form of expression frequently employed by those writers): 'A is to B as C is to D'. And neither did Mascall (1949), Penido (1931), Emmet (1946) and Garrigou-Lagrange (1947-8, 5th ed.), though all tried to illustrate it. But it is now clear enough, including interchanging expressions to create metaphors, which Aristotle illustrated in Poetics, 1457b, 16-30.

Consider the proportionality: 'Warden is to guard as doctor is to (A) nurse; (B) patient; (C) surgeon; (D) ward?' The test-makers offer 'nurse' as the answer, indicating that one should formulate the relationship in which the first pair stands and then find the answer that will establish the'same relationship' between a second pair and will not hold for any other pair. For instance, 'supervise a person administratively subordinate' would characterize the warden-guard relationship and also indicate 'nurse' as the missing fourth element, while excluding each of the others.

'Supervises' might also characterize the relationship of doctor to patient; but the sense of'supervises' would not be univocal with (the same as) 'supervises' in 'The warden supervises the guards'. Presumably, ceteris paribus, univocally designated common relationships are to be preferred to equivocally designated ones and analogously or metaphorically designated common relationships are to be preferred to merely equivocally designated ones.

Verbal proportionalities (verbal analogies) can be based upon implicit predication that is non-univocal but not merely equivocal (as Aristotle and Aquinas observed): 'Dentist:cavity:

:doctor:disease'. 'Treats', 'cures', and the like, are related in meaning, but not univocal in the corresponding explicit predications. So too with 'Asylum:refugee: :destination:traveler'.

'A horse is to a herd as a mountain is to a range.' The common relation may be 'usually belongs to'. Do the sentences, 'A horse usually belongs to a herd', and 'A mountain usually belongs to a range' employ 'usually belongs to' equivocally? I think so. Does the fact that 'The horse stopped belonging to its herd' is acceptable, nonfiguratively, but 'The mountain stopped belonging to its range' is not, show that the phrase'belongs to'has different contraries in its two occurrences and, thus, is equivocal? I think so. Does the fact that 'usually' has different meanings in the two cases show that 'usually belongs to' has different meanings? Certainly. That is, 'a horse usually belongs to a herd' means 'in nature, apart from husbandry', whereas 'usually' with respect to mountains concerns frequency of distribution geographically. That verbal proportionality turns upon a common relational predicate that occurs equivocally.

The point of the classical quasi-mathematical talk about 'analogy of proper proportionality' was that proportionalities typically presuppose predicates analogous in meaning. Some relationships, say, between man and his activities, are proportionally similar to other relationships, say, between God and his activities ('Socrates is to philosophy as God is to the world': 'understands it') and the common words that especially aptly characterize the relations are non-univocal, as we find in many Miller's Analogies examples (see illustrations below).

Aristotle, Topics, I, 17, 108a, 7-11, said: 'Likeness should be studied, first, in the case of things belonging to different genera, the formula being: "A:B::C:D" (e.g. as sight is in the eye, so is reason in the soul and as calm is in the sea, so is windlessness in the air.)'

The classical explanation why a term in related but different senses (meanings) may be the most apt common predicate presupposed by true proportionality, and why there is no comparably specific univocal term that is apt, is that the relations that are only proportionally similar give rise, via abstraction, to concepts that are not the same but are related to one another by priority-posteriority (per prius et posterius) and that although different, they attach, by their similarity, to one word rather than many. The key step, not satisfactorily explained, is why not distinct words when distinct words are also similar in meaning, e.g. 'saw' and 'viewed'?

The classical writers did not assert that we always have an apt analogous term to apply to a true verbal proportionality; in fact, Aristotle remarked that we do not (Post. An., II, 14, 98a, 20f) and Cajetan (De Nom. Anal., ch. 11, no. 117) agreed: 'unity or diversity of name should be considered accidental' [italics added].

For instance, although a squid's pounce, a bone and a spine do not have one name, they are no less analogously similar than if they had one name. Nor would they be more similar if they had one name. Nevertheless, if they were called bones by a common name in such a way that through lack of words or because of their proportional similarity the name bone were extended to the others, we would believe bones, squid's pounce and spine to be of the same nature and notion. Especially, because, as was explained above, certain proportions flow from things which are proportionally the same as if they had one nature. (See also Cajetan 1509:4.)

No matter how small the vocabulary, provided it has relation terms and permits the formation of verbal proportionalities, there is always some word that is at least as apt as any other within the vocabulary to characterize the common relation in a true proportionality. Yet it is not necessary that there be one word more apt than any other, though for relations that are only proportionally similar, no word applied univocally is as apt and specific as some word applied non-univocally. In fact, the expressions involved may make it impossible for any word to apply univocally because they dominate and differentiate any common word that preserves the truth and the sense of the original verbal proportionality. That is the position I will explain and defend. And that, I think, is also the classical position as to God, metaphysics and theoretical science, in so far as common predicates from the vocabularies for 'ordinary experience' are applied to such subjects.

For all we know at the outset, the reason why there are analogous predicates for certain true verbal proportionalities and not distinct words, may be something like this: relatively to our vocabulary, (i) the contrasts implied by any pair of distinct words are greater than (comparatively to other classifications) the case warrants, and fail to represent the discernible relatedness of the cases; and (ii) the two cases are more like other cases with the same name, e.g. of 'usually belongs to' ('Two white knights usually belong to a chess set'), than they are like anything characterized by a contrasting name that either excludes, or otherwise displaces this name, 'belonging to'; and (iii) the two cases are more like one another than they are like any pair of things aptly characterized only by distinct words and not aptly characterized by meaning-related same words.

So, I suggest we look at the way we classify things under a limited label scheme (vocabulary) without, at the outset, explaining our verbal behavior with hypotheses about the things we classify.

(b) Verbal analogies based upon non-univocal predication

We can indicate a proportionality without naming the common relationship, where, were the relationship named with one word, it would be named equivocally (but not merely equivocally): 'Philosophy is to insincerity as love is to slavery.'

Metaphors are sometimes created by substituting one word of a proportionality for another: as 'knowing is to the object known, so is sensation to the object sensed'; and thus, one is sometimes described metaphorically as 'sensing that it is so'. Aristotle pointed out (Poet. 1457b) that metaphors can be created from proportionalities by denomination: 'ifA:B::C:D, then in a sentence D is substituted for B'. (He does not give an explanation of why not all such sentences are acceptable.) 'As old age (D) is to life (C) so is evening (B) to day (A). One will accordingly describe evening (B) as "The old age of the day", (D&A) . . . and old age (D) as the evening or sunset of life (B&C)' Poet. 1457b, 23-25. (Aristotle illustrated the reversibility that is common in metaphors supported by proportionalities but that has never been explained.)

Aristotle also indicated that where A:B::C:D and there is a name for the C:D relation but there is no name for the relation A:B, we transfer the name of the C:D relation to name the A:B relation. That indeed is metaphorical proportionality and it is 'improper' (to use Aquinas' phrase, Ia, 13, 3 ad 3; De Ver., 7, 2c; I In De An., 10, n. 160) because the relation signified by 'CRD' is only by similarity, relatively to the vocabulary, the same as the relation signified by'ARB'. In fact, 'ARB', except as considered to be proportionally similar to 'CRD', is evidently false: that is, underlying this kind of metaphor is a commonplace falsehood. 'The boiling sun sowed flames in all the fields.' That is why it is improper proportionality (Aquinas, In I Sent., 23, 1, 1 and 2c, and 2 ad 1; Ia, 13,6).

In the Miller's Analogies tests, besides cases where the relationship is purely mathematical, e.g. 'double' in '4 is to 2 as is to 4' or merely grammatical, e. g. 'is the plural of', or metalinguistic, e. g. 'is more specific than', 'is spelled backwards', and 'is included in the spelling of', there are many cases where the common relational word would (if expressed.) not be used unvocally For instance:

brunch:lunch::smog:haze. The test-makers think 'is a combinatio of. . .' is the key idea, brunch being a combination of breakfast an lunch and smog being a combination of haze and exhausts. Skippin other difficulties now, 'is a combination of x and y' does not appl to brunch and smog in the same sense; that is, 'is a chemical suspen sion of' may apply in the one case and not in the other. So too, wit the senses of'is opposed to' that might apply in the'philosophy . . slavery' example above.

The analogy of proper proportionality that Aquinas talked about i of this sort: the presupposed common word is not univocal but i not merely equivocal either. (Aristotle noted that such terms as 'i the same as' and 'is commensurable with' differentiate dependin upon the kind of things compared; Phys., VII, ch. 3, 248b, 1-27 an 249a, 1-24 and see Top., II, ch. 17, 108a, 7-10.)

For the Miller's Analogies examples, there is always more than on word to name the same relationship. Apart from a theory of the sor to be offered, it would be difficult to determine that there is no always some univocal common word for the relationship. We ca conjecture, from the way verbal proportionalities were employec by classical philosophers, that in relation to a limited vocabulary there are similarities of relations aptly described with analogou terms for which no univocal term is appropriate. And none can b devised.

In my theory the absence of such univocal predicates is explainec by the fact that if one formulates first the verbal proportionality Socrates: philosophy::Fido:his master, and then creates corresponding sentences that express the relation, 'Socrates loves philosophy and 'Fido loves his master', the mere concatenation of the commo predicate, 'loves', with the contrasting sentence frames, 'Socrates ... philosophy' and 'Fido ... his master', differentiates th common word. There can be no univocal common term of the sam~ level of specificity as an analogous term that aptly names the relatior presupposed.

Another example: Even numbers:odd numbers::elephants:mice Someone proposes: 'Every even number is larger than some odc number (one) just as every elephant is larger than some mouse' Concatenation of'is larger than' with'number' in contrast to 'elep hant' sufficiently differentiates 'is larger than'. One cannot say th same thing by merely substituting a common univocal predicate fo 'is larger than'. And that has nothing to do with the metaphysical relationships of numbers and elephants or with your or my beliefs about them; that is simply a result of the linguistic relations of the words involved. A consonant consequence is that a near synonym of an analogous pair, when substituted, will also be analogous.

The test question is 'Law:citizen:: (a) democracy:communism, (b) weapon:peace, (c) reins:horses, (d) gangster:policeman'? The answer is (c). I suppose such terms as 'guides', 'restrains', or 'directs' might name the common relationship. But if 'restrains' is accepted as more specific and more apt than 'guides', then the common word differentiates in 'Law restrains the citizen' and 'Reins restrain the horse'; it is metaphorical in the first case in comparison to the second. (Why the first case should be metaphorical in relation to the second and not the other way around, and how we can explain metaphor without recourse to talk of 'priority' of occurrences in time and 'literalness' are discussed in chapter 4.)

Here are some other proportionalities where an apt common word in the corresponding explicit relational statements would be metaphorical. (1) The farmer:seeds::sun:flames (sows); (2) Thief:money: :age:beauty (steals); (3) baby:locomotion: :time:worry (creeps); (4) leaf:fire::shy:person: (shrivels, burns); (5) road:mountains::argument:objection (twists, turns, surmounts). Once the pattern is exhibited, one can construct lists at will. And one can then construct further metaphors by denominative interchange (see ch. 4, denominative metaphor): e.g. of A for C, or D for B or vice versa, 'beauty's thief', 'the leaves of loneliness', 'autumn of inquiry'.

My hypothesis is that if the common word does not occur univocally in the sentences articulating a true verbal proportionality, then it differentiates because of the semantic features of the other words involved (or implied). Thus, we travel from the medieval doctrine of analogy of proper proportionality among referents to a contemporary doctrine of semantic contagion as the fundamental feature underlying analogy, metaphor and mere equivocation.

Chapter Two

The genus: meaning differentiation

I MEANING DIFFERENTIATION

Words in English and other Indo-European languages, in suitably contrasting contexts, differentiate. For example, 'She dropped a stitch', 'She dropped her hem-line', 'She dropped her book', 'She dropped a friend', 'She dropped her courses'. The meanings of 'dropped' ar,e appropriate, ptted to the completion words and, so, differ. That is the generic linguistic phenomenon that includes mere equivocation, analogy and metaphor.

Words also adapt in predicate mode (e.g. to whether what is ascribed is ascribed as an action, state, proclivity, ability, etc.: 'Jones plays golf', 'Smith plays the piano', and to whether it is ascribed as a cause, effect, symptom, state, etc.) and to contrasts of ascription: to contrasting implied vantages, e.g. of agent or observer. Predicate mode and ascriptive differentiation are explained in chapter 5.

The meaning of any word, whether it be as common as the nouns, 'chair' and 'ball', or the verbs, 'fix', 'strain' and 'look', or whether it be statistically less common, like 'investigator', 'interlocutor', 'expect', or'anticipate', differentiates relatively to some of its other occurrences. Consider the sets:

(1) The junk men took the chair.

(2) The chairman of the board took the chair.

(3) He dropped the ball on his first catch.

(4) He skipped the ball because of the late hour.

(5) He fixed the leak in the pipe.

(6) He fxed the time of the appointment.

(7) He fixed the race.

(8) He strained the soup before serving it.

(9) He strained his back lifting.

(10) He strained his eyes to catch a glimpse of her.

(11) He looked over the plant.

(12) He looked over the fence. .

(13) The investigator was retained by the police.

(14) The investigator does not necessarily publish his results.

(15) He anticipated the response. (16) He went there because he anticipated her.

Even syncategorematic words differentiate. 'That's the way to town', 'The whale is a mammal', 'The trouble is...', and 'The man died'. Prepositions differentiate (e.g. 'in the house', 'in the book') and articles do as well, 'I have an idea', 'I'll have a cookie', 'I have a brother, too'; 'have' differentiates as well.

Enumerative induction will not show that every word differentiates; but I think experiment will convince fluent speakers that fit to context is evident and universal. The versatility of Latin and French words is perhaps obvious to us because we notice how many distinct English words we need to translate a single foreign word. (See Introduction above.) Yet, that all terms differentiate can be proved only after the phenomenon has been explained.

Fit to context is necessary but not sufficient for differentiation; -there have to be contrasting adjustments (fits) to produce diversity of meaning. The contrast between 'door' and 'eyes' in 'He closed his door' and 'He closed his eyes', or 'stitch', 'friend', 'book', in 'She dropped her (noun phrase)', causes the differentiation of'closed' and 'dropped', when those differing direct objects are substituted (the rest of the sentence frame remaining constant). Otherwise, there would be no linguistic cause of differentiation. "

2 SAME WORDS

We have to adopt a criterion for different occurrences of the same || word. Roughly, they are the same if spelled and sounded the same) more formally:

Def: 11-1: Same terms: The class of allograph/allophone tokens, indiffer- -ently substitutable for one another (on grounds of spelling and sound) in every context in which any one of the class occurs meaningfully, is the class of same-term occurrences, with respect to a given inscription or sound which occurs meaningfully in some given context.

'Substitutable', for purposes of this definition, does not mean either of the things it means later (see senses 2 and 3) but means (1) 'can be put in place of one another without creating a different word or sentence or bringing it about that the meaning of any component -word is different from what it was before the substitution'. It corresponds to Lyons' idea (1966:73) that 'two expressions are in contrast:

if the substitution of one for the other in the same context produces a different word or sentence; otherwise they are in free variation'. Thus all tokens of the same type are freely variant (spelled and sounded'the same').

(2) The second kind of substitutability is 'with preservation of acceptability and of consistency/inconsistency with the original' -that is required for members of paradigmatic sets and, therefore, of predicate schemes in chapters 3 and 4, but not here.

(3) The third is the substitutability of near synonyms, that is substitutability 'to produce a replacement sentence approximately the same in meaning as the original'. This notion is explained in chapter 4.

I understand that a properly spelled word cannot, sometimes, be substituted for an improperly spelled occurrence; that is provided for by the requirement of 'meaningful' occurrence, to eliminate 'material supposition', often called 'mere mention' (see below). And that explains why the formal definition (II-1) is so complicated.

'Look out', and 'look over' in some contexts count as whole terms not made up of the constituent terms 'look' and the preposition 'out' or 'over' functioning adverbially. I use 'same term' and 'same word' interchangeably even though the former expression is broader because expressions like 'in English' and 'come on' and 'look out' count as single terms in some contexts where we would not count them as single words. 'Term' has a formality that makes sentences harder to understand than does 'word', so I frequently use the latter where the former is technically more accurate.

In 'Scrooge was mean' and 'He tried to tell us what "incongruous" and "blithe" mean,' 'mean' (meaning 'parsimonious') and 'mean' (meaning 'to convey') are occurrences of the same word, being allograph-allophones of one another. So also: 'The cat caught the mouse at once', 'The student caught the teacher's likeness in his sketch', 'The man caught the train after lunch', 'The boy caught the fish for lunch'. 'Caught' differentiates across a set of sentence frames of the same surface grammar: noun phrase (composed of definite article and common noun) and verb phrase (third person, singular, past tense verb, with direct object) and adverbial phrase. The verb 'caught' is the same word, having multiple occurrences in these sentences, despite differences of meaning. No significance is to be attached to the suggestion that there is some sort of meaning derivation in the 'caught' cases and no derivation in the 'mean' cases, so that 'caught' is a single word with multiple occurrences while 'mean' is an allograph for two different words. For there is no derivation in the 'caught' cases either. 'Derivation' is a false scent left by the classical theory of analogy.

Tokens of the same word may be spelled differently (misspelled or alternately spelled). The misspellings and alternate spellings are allographs of the standard spelling by a normative 'allograph' criterion that correlates the deviant spellings with a preferred allograph type. The same is true of deviant phonemes. Usually, whatever would count, for a distributional linguist, as a set of sameterm occurrences will be acceptable for the theory I am developing (Lyons 1968). Thus the members of the correlated allograph/ allophone class form an equivalence class distributionally but need not be equiform appearances or indistinguishable sounds.

Because homonyms are phonemically equivalent but do not always have allographic inscriptions, not any two tokens that are phonemically equivalent are occurrences of the same word, only the ones that are both said and spelled the same way: e.g. 'He made it home', uttered in baseball talk and about the efforts of a sick man to reach his home, but not 'sail' and 'sale'.

Sometimes we say that tokens that are spelled the same way and are pronounced the same way are, nevertheless, two different words or two different terms, as 'direct'in 'direct delivery to one's home' and 'direct an orchestra'. But that concept of idifferent terms' and 'different words', otherwise perfectly acceptable, requires, for its application, a comparison of both the syntactic role and the contextually determined meanings of the pair. It involves a distinction needed for some theoretical purposes but inefficient here.

The criterion I employ does not suppose that we do not know the meanings of the words. But it is independent of a comparison of the particular meanings, allowing us to establish that we have sameword occurrences on sound and spelling alone, before making any judgment about their comparative meanings.

We do not need to compare the meanings, e.g. of homonyms, to one another in order to determine the orthographic or orthophonic type of each token and, thus, to determine whether they are tokens of the same word. In 'The fifth sentence on the page contains the symbol "means"', 'means' occurs in the latter without a meaning, since it occurs, as the medievals said, in 'material supposition', merely mentioned but not used. But'means' still belongs to a class of allophone-allograph tokens each of which is substitutable for any other in any context of meaning occurrence.

Discourse context: there is a discourse context for a word whenever it occurs (physically), as used, within an uttered or written complete sentential expression, and there is a distinct occurrence of a word as often as it occurs within a complete sentential expression.

'Complete sentential context' is an idealization. Most 'sentences' in colloquial speech are incomplete or grammatically incorrect (or both) and questions, commands etc., are perhaps as common as declarations. Further, declarations have no monopoly on statements. Nevertheless, let the declarations be representative, and let an utterance corsut as complete when a corresponding complete expression can, with reasonable certitude, be constructed from the context given. That way, we can count such a context as 'complete' for our purposes, as well. I regard complete sentences as 'contexts' and 'broader units of discourse', including gestures, references and other utterances on the same or related subjects, as the 'environment'. (Sometimes other writers regard the sentence as the environment and the discourse background as the context.)

3 DIFFERENCE OF MEANING FOR SAME WORDS

Same words have the same meaning unless the opposite is both caused and indicated. This is the inertia principle. There would be nothing gained from premising that same words can differ in meaning when nothing linguistic indicates it or causes it.

Indicators of difference of sentence meaning are not alone sufficient to indicate difference of word meaning because, when sentences involving a common word differ, the explanation of the difference of meaning could lie in something other than a difference of meaning in the common terms, e.g. in contrasting surrounding terms or in the wider discourse environment, including gestures and prior references or assumptions (see Haack 1968). And, obviously, a pair of sentences with a univocal common word can differ in meaning: 'I saw the cat'; 'I saw the dog.'

Sometimes sentences with the very same words (in the same order) differ in meaning: (a) 'That's my view' and (b) 'That's my view.' Can we tell whether the common word 'view' has the same or different meanings? Not by anything explicitly given within the sentences.

To determine whether there is a difference of sentence meaning (as would be indicated by a difference of entailments), we need already to know whether 'view' differs, (Haack 1968: 64, and Lehrer 1974b). Suppose (a) is employed in an environment that indicates that one is expressing one's controverted opinion. Suppose (b) is uttered with a gesture indicating the panorama visible from one's lawn. The two could as well be reversed. Whatever determines the particular meaning is outside the sentence itself and is environmental rather than sentential or contextual.

One cannot in this case, therefore, determine whether the two sentences differ in entailments, in paraphrases, in which objections or qualifications would be semantically appropriate (e.g. 'Oh, but you have a better one in back') until one has already determined whether the two tokens of the word 'view' have the same meaning.

That is frequently true: applying tests to indicate differences of meaning between sentences presupposes tbat we already know whether words, occurring in both, have or do not have the same meaning. So, tests for difference of sentence meaning may (and sometimes will) beg the question as to whether common terms have the same meaning.

What, then, can be employed to test for difference of meaning for individual words in sentence frames that differ in other respects? For instance, (a) 'Smith stopped payment on his check before it had cleared the bank'; (b) 'Jones stoppedpayment on his car after the third month.' These involve different senses of'stopped payment' (e.g. the contrast of stopping paying (b), and stopping heing paid (a) ), but that difference in the particular expression, 'stopped payment', will not necessarily be located by the differing entailments, paraphrases, grammatical transformations, objections, or qualifications appropriate to the two sentences. Those differences might just as well be attributed to the other features in which the sentences or environments differ.

Tests for sameness of meaning for different words, tests for synonymy, cannot be effective for my purpose, either, because they all involve the substitution of one expression for the other and a subsequent application of tests for difference of sentence-meaning to the resulting sentences. Lyons' earlier account (1968: 428) -'two (or more) items are synonymous if the sentences which result from the substitution of one for the other have the same meaning' - has the incongruous result that merely equivocal words are always synonymous! (That has an important consequence, see below.)

Substitution cannot be effective for same words, as is illustrated with simple equivocals, 'pen' and 'pen'. For every pair of same words is substitutable, salva veritate, regardless of differentiation. That is one of the most significant pieces of evidence that the dominance hypothesis (which lies at the foundation of the present theory) has to be true. Substitution tests work only on the condition that the meaning attached to the substituted inscription resists adaptation to the full extent that it differs from the original word. Though such resistance may obtain in the case of inscriptions for different words, it never obtains in the case of inscriptions for the same word.

If we took the token, 'caught', from the sentence 'He caught a fish' and physically transferred it (by interchanging the very printing) to the sentence 'He caught a cold', putting it in place of that occurrence of 'caught', the resulting sentence would have exactly the same meaning as did the former. This is an important indication of the law-likeness of the adaptation regularities being investigated: the generalizations to be offered support appropriate counterfactual statements.

Tests for sameness of meaning for same words have to be carried out through comparisons of their meanings. That does not, as I mentioned, beg the question as to whether we know the meanings to be the same or not; for one thing, even if they are different, we may not have noticed or characterized the difference. (Aristotle, as did Wittgenstein later, remarked (in Top. 1, 18, 108a, 17-36) upon the treachery of unanalysed terms.)

Many of my claims about analogy depend upon the premise that an n-tuple of occurrences of a particular word differ in meaning. Sometimes the truth of such assertions is not apparent and there ought to be a way of resolving doubts, at least for enough cases to support the rest of the theory, as, for instance, about the Miller's Analogies examples in chapter 1. Moreover, many philosophers, Duns Scotus, Ockham, Quine, Cartwright and Mavrodes for instance, have expressed doubt that'exists' differs in meaning when applied to numbers, sets, Socrates and God; and some people have said that 'caught' does not differ in meaning in the examples mentioned in section 2, above, in this chapter.

Meaning-difference tests will help to resolve such disagreements, though, in the long run, whether a given test, even adjusted to our experience, is adequate to reveal difference of same-word meaning depends upon whether it identifies such differences compatibly with our intuitions - not necessarily with all our individual intuitions, but rather with the general pattern of our intuitions, which will, no doubt, become adjusted themselves after we find tests that are serviceable: a kind of reflective equilibrium reached through the feedback adjustments of both tests and intuitions. So the tests I begin with are no more refined than the linguistic intuitions (insights) I use to adjust them.

Aristotle remarked, Topics (1, 15, 107a, 14), 'the same term does not bear the same meaning in all its applications' and provided us with an initial battery of tests for difference of meaning for same words, in Topics (1, 15, 106a-107b), offering fifteen observations upon marks of difference of meaning.' I present them here, with some alterations to apply to the kinds of cases I am discussing, to serve as patterns upon which further tests may be invented as they are needed.

4 ARISTOTLE S TESTS FOR DIFFERENCES OF MEANING AMONG SAME-TERM OCCURRENCES

(a) Disparity of contraries If there are distinct contraries for a word, then the word bears several meanings: 'dull' and 'flat' are distinct contraries of 'sharp'. Also, 'fine' as applied to a picture may have 'ugly' as its contrary, but as applied to a house, 'ramshackle'; so, 'fine' is an equivocal term. (Aristotle apparently did not realize that all words are, or can be used equivocally.) Some of Aristotle's examples, of course, depend upon oppositions peculiar to classical Greek; but there are in each case modern counterparts: e.g. 'true' which has 'false' and 'out of plumb' and 'off pitch' as contraries.

(b) Homonymous contraries Even when the contraries bear the same name, e. g. 'blurry' as contrary to 'clear', used both for color and for sound, they may be homonymous, and therefore different. So, 'clear' is equivocal.

Aristotle suggested that the difference in the sensory mode by which the judgments of obscurity (blurriness) are warranted indicates a difference of meaning. Similarly, with 'sharp' and 'dull' in regard to flavors and solid edges. 'Blurry' applied to visual images and to thoughts, also manifests a difference in its appropriate contraries: 'well-focused', 'intelligible', etc. Although Aristotle's interpretation of this test relies upon the epistemic disparity in the conditions for applying the common contrary, 'blurry', nowadays we can restrict our interpretation to semantic factors, e.g. to disparity of the near synonyms or further contraries (e. g. 'blurry'/'clear').

(c) Asymmetry of contraries If one token has a contrary while another has none, they differ in meaning. "'To love", also, used in the frame of mind has "hate" as its contrary, while used of the physical activity (kissing) it has none: clearly, therefore, "to love" is an equivocal term' (Top., 106b, 2).

(d) Difference of intermediates (106b, 3) If one token (and its contrary) has intermediates and another has none or if both have intermediates but not the same ones, the tokens differ in meaning. For instance, in one occurrence, 'conscious' may take an intermediate, 'semi-conscious' ('The prisoner was conscious'), and in another may not ('Human beings are by nature conscious').

An example of different intermediates: between 'white' and 'black', used to classify persons, there are different intermediates than there are between 'white' and 'black' classifying colors.

(e) More intermediates Similarly, if one token, in relation to its contrary, has more intermediates than another, the tokens differ in meaning: e.g. between 'white' and 'black' for colors there are many shades, whereas between 'white' and 'black' for people there are only a few: 'yellow', 'brown', etc. In the vocabulary of the model to be used later, despite the sameness of the color words, the words belong to different predicate (classificatory) schemes or lexical systems (as some would say). For example, 'black' may have the same 'sense' meaning when its lingMistic meaning is different.

'Intermediate' is ambiguous and used equivocally in the above examples. In sense 1, there is an intermediate, K, between L and M, just in case anything to which one of the three predicates may apply is such that only one may apply to it. In sense 2, the applicability of the first predicate excludes one but not the other of the remaining two: 'alive', 'half alive' and 'dead'. In a scheme for primary colors 'orange', 'yellow', 'green' and 'blue' are intermediate between 'red' and 'violet', whereas in a color scheme for people, 'black' and 'white' are included, and there are intermediates like 'red', 'yellow', 'brown' and 'olive', but not 'orange' and 'purple'. Those words that belong to both schemes have different 'sense' meanings and different linguistic meanings. (See C. I. Lewis' seminal 'Linguistic Meaning and Sense Meaning', 1962.)

But what of'black' and 'white' applied to cows? There are different intermediates from 'black' and 'white' applied to people. 'Black' applied to cows, as David Schrader pointed out, does not differ in 'sense' meaning (cf. Lewis 1962) from 'black' applied to automobiles, and yet 'purple' is an intermediate for cars and not for cows. That indicates that 'black' is equivocal, has different linguistic meanings in the two cases, although the kind of 'color experience' needed to apply the predicate is the same. That casts doubt upon Aristotle's test because we would not ordinarily say 'black' differs in meaning in those two cases.

The difficulty here is created by the fine line between commonplace falsehood and unacceptability of the sentence. In the context of classification by racial color, is 'Some people are purple' an unacceptable sentence orjust obviously false? Is 'Some cows are purple' a commonplace falsehood, or is it unacceptable? I think we should interpret Aristotle's test as follows: when intermediates for the one occurrence are unacceptable in the environment of the other (see chapter 3 for a more precise statement of the condition), we have a difference of meaning; when the inapplicability of intermediates is due to commonplace falsehood but not unacceptability of the sentences, the test has not revealed a meaning difference; and where we are not sure whether there is unacceptability or commonplace falsehood, the test fails to yield a usable result.

(f) Difference of contradictory opposites Where the contradictory opposite of a token is a different word or a homonym of the contradictory opposite of another token, the word has different meanings. For instance, Aristotle indicated that the contradictory opposite o 'to see', namely, 'to fail to see', can mean 'to fail to possess the power of sight' and 'to fail to put that power into active use'. Thus, 'Bats do not see' may mean that they do not have the power of sight or that they do not exercise it.

'Having the power to F' and 'exercising the power to F' are different meanings of any predicate, F, which is capable of such differentiation. Needless to say, there are other meanings for'to fail to see' such as 'to overlook'. So, if a pair of occurrences of 'to see’ differs in what is the appropriate contradictory opposite, or if a common contradictory (e.g. 'to fail to see') is homonymous, then they differ in meaning.

(g) Homonymous privatives If the words that denote the privation (or presence) of a certain state bear more than one meaning, then a common word naming the state (or its privation) does also. Thus, if 'to lack senset has more than one meaning, then'to have sense' also does. So also, if'to fail to read' has more than one sense, so does 'to read'. Similarly, if in two sentences 'makes sense' occurs (e.g. 'Crusoe made sense out of the signals' and 'That sentence makes sense') where the appropriate privation would in the one case be 'failed to make sense of' ('found no meaning in') and in the other 'contained no meaning inherently', then 'make sense' is equivocal Aristotle's example is that 'to be wanting in sense' as applied to soul and to body are different things, and therefore, 'to have sense' as applied to both is equivocal.

(h) Different meanings in inflectedforms Here Aristotle shifted ground somewhat: if an inflected form of a word has more than one meaning, then the word has more than one meaning; his example is 'justly' and 'just'. For instance, 'justly' may mean 'appropriately' or it may mean 'fairly'; 'just' is similarly equivocal.

Aristotle said there is a meaning of the uninflected word corresponding to each distinct meaning of the inflected word. We have an English counterpart of that in 'to smoke'. Just as there is an 'active present' sense, in which we might say 'the fireplace is smoking', so there is, corresponding to the inflected form, a dispositional sense in which we might say 'Jones smokes' or 'Jones is a smoker' and there is also a proclivity sense in which we say a flawed fireplace smokes even when it is not actually smoking and even though we do not (though we might, according to the theory to be developed) say that it is a smoker (as a repairman might classify it).

(i) Differences in predicates signifed As Aristotle stated it, if different occurrences of the same word differ in the predicates they signify, then they differ in meaning. He illustrated with 'good', which sig nifies 'productive of pleasure' in some applications to food and i others, merely, 'not harmful to health', and, in the case of medicine, 'productive of health'.

Frequently, just what is at issue is whether different occurrences of the same word signify the same predicates or not. In those situations, other considerations settle whether the significations differ, perhaps the other tests Aristotle provided.

As I classify the material in Topics, 1, 15, there are two sub-tests under 9 (107a, 3-36) that can be applied in some of the disputed cases: 107a, 1 31 and 107a, 31-35. If the genus of the objects -denoted by one occurrence is different from and not a subalternate to the genus of the obj ects denoted by another, the common word is equivocal. Thus, if'a fish' and 'a thief' and 'a cold' are the respective 'catches' (designated by the cognate accusative of'to catch') and they belong to a genera different from one another, genera not subalternate to one another, then 'catch' differs in meaning when applied to them.

Aristotle used a common name for different sorts of things - 'donkey' as a name for an animal and for an engine - and argued the equivocity of the name from the fact that the genus of engines was neither the same as nor subalternate to that of animals. His test can be generalized: where a transitive verb allows different sorts of thing-names to be substituted for its cognate accusative ('the catch'), if these diverse sorts are not species of one genus or species subordinated to one another or ordered as determinates to determinable, then the common verb also has different meanings. Thus, in 'Hejumped the fence', 'He jumped the king', 'Hejumped the trem- ii bling victim', and 'He jumped bail', there are different kinds of differences of meaning, and no pair of the occurrences of'jumped' is univocal.

As Topics, 107a, 33-6 indicates, the same considerations, if applicable to the contraries of a pair of same words, will mark the equivocation of the latter as well.

(j) Difference of contextual def nition Aristotle noted the value of contextual definitions: if the definition for a word in one combination of words, when compared with the definition appropriate to it in another combination, is found to differ, then the word differs in meaning in its two occurrences. Thus, 'clear body' and 'clear note', when contextually defined, with the notions of'body' and 'note' later abstracted, do not yield the same definition: 'a . . . of such and such color' and 'a ... easy to hear'; so, 'clear' differs in the two occurrences.

If'clean record' and'clean floort are compared, we get something like: 'a record that does not indicate crimes or reprehensible behavior' and 'a floor without untoward indications of stains or dirt'. Then, dropping the terms 'record' and'floor', we are left with defnition frames that are clearly different, exhibiting a difference of, meaning between the two occurrences of'clean'.

Comparison of contextual definitional frames is a serviceable test, in fact one of the most easily applied and most illuminating of all the tests. It can be applied, as I usually do, simply by listing near synonyms, opposites, determinables, determinates, hyponyms, etc., for the one token and asking whether the same expressions are both co-applicable with and stand in the same semantic relation to the other.

(k) Equivocation withir contextual de.finitions Aristotle recognized that in any of these definitional processes, the first stage may yield another term common to both occurrences; e.g. 'healthy' may yield 'what is related commensurably to health'. To use a contemporary example, in the cases of'sound ship' and'sound melon' and 'sound argument', analysis or contextual definition might yield'x that does not exhibit or contain a defect' (Mavrodes 1970: 747). But the mere presence of common words in a pair of definitions does not sufficiently indicate sameness of meaning because the common word itself may be equivocal. Aristotle said, 'Often in the actual definitions, as well, ambiguity creeps in unawares, and for this reason the definitions also should be examined' (107b, 7).

In general, the mere fact that a pair of predicates seem to be determinates of a common determinable (e. g. 'unfit for its purpose') does not show that the determinable is applicable univocally to both designates. Nor does it follow that because a given determinable has the same set of determinates in two contexts, that either it or the determinates apply univocally in both. (This has important applications to philosophical analysis, see chapter 8, section 2 (c), p. 189, below. )

In fact, one cannot, practically, establish univocity of a pair of words in differing sentence frames; one can at best (by getting negative results from tests like these) render equivocation improbable For this reason, I suggest that same words are univocal unless equivocation is indicated (linguistic inertia).

(1) Difference of comparatives This can also be generalized. Aristotle suggested that a word differs in meaning if the one instance admits of a comparative and the other does not, or if, while both admit of comparison, they do not admit of (non-figurative) comparison to one another. This seems especially useful. A sharp flavor and a sharp note are not non-figuratively comparable in that the flavor might be sharper than the note. A fast car is not faster than a fast color. Long hair is not longer than a long book. Sometimes the affirmative comparison is acceptable when the negative or privative is not, and a meaning difference is also thereby indicated.

Having the same comparatives will not, however, assure univocity because the comparisons are sometimes possible just because of the differences of meaning. For instance, one man may be more intelligent than another and any man is more intelligent than any dog and some dog is more intelligent than some man; and the reason the three may be true is that 'intelligent' has three (at least) different meanings (that are revealed by some of the other tests). But Aristotle's test, as he stated it, is sufficient in the cases where comparisons are blocked either because one token does not take a comparative ('The patient was alive') while another does ('This class was alive, more alive than most'), or because, despite the possibility of comparatives, the one is not to be compared non-figuratively to the other ('dull day', 'dull knife').

(rn) Generic opposition of the realms of applicability In one occurrence the word is the differentia of a genus that is different from and not subalternated to the genus of the denotata of the other. 'Sharp' differentiates note from note and, likewise, one pin from another, whereas 'pin' and 'note' do not denote things of the same or of subalternated genera. As a result, non-figurative comparison fails (even though both occurrences take comparatives); e.g. 'That note is sharper than a pin' is hyperbolic emphasis. The reverse is harder to provide an environment for, because the effect of hyperbole is more likely to be blunted: 'That pin is sharper than a note'.

(n) Difference in the class-partitioning properties Aristotle suggested that a word differs in meaning if the differentia under the word in the one case is not the same as that in the other. Thus, if the common word is 'color' and the ranges of application are bodies and tunes, then if the differentia of the ranges are different (shades for color and modality in melodies), 'color' differs in meaning. 'Crooked' applied to paths and to moral behavior has different 'differentia'.

(o) Difference in predicable status Finally, if in one case the word is the species and in another the differentia, the tokens differ in meaning because 'the species is never the differentia of anything', (Top., I, 15, 107b, 33). Thus, 'the rational' sometimes means 'that by which men are differentiated from beasts' and sometimes means 'what is reasonable to believe'. So too, 'rational' as the differentia ofthe genus 'animal' and 'rational' as a species of numbers differs in meaning.

Aristotle's interest in tests for equivocity arose not in discussions of analogy but in providing practical helps for scientific thinking in the Topics. Nevertheless, his tests help to disclose meaning differences in cases where competent speakers of the language are in doubt.

Aristotle's tests are, as I have emphasized, not the only ones we can employ. (Nor do we have to rely upon any one of them alone.) For instance, if modifiers appropriate to one occurrence are not appropriate to another, that often marks a difference in meaning. So also, when the same modifiers are applicable but differ in meaning: 'He jumped the fence swiffly' can mean that he moved quickly in doing so. But 'He swiftly jumped bail' does not require that he move at all but only that he jump bail soon; so 'jumped' differs in the two cases.

In a doubtful case, a congruence of two or more of the various tests that disclose asymmetry of contraries, antonyms, modifiers, qualifiers, inflections, comparison, genera of objects denoted, and of contextual definitions, will probably be sufficient to nudge intuition to recognize a difference of meaning. Furthermore, one can apply even the congruence of these tests cautiously, holding that in general, though perhaps not in every case, such a congruence reveals equivocation, but only when one's educated and adjusted verbal 'ear' is not offended by the result.

5 SEMANTIC CONTAGION

Semantic contagion consists of the dominance of one word (or group of words) over another (or group of others). It is observable as a process, relationship or event, depending upon how the examples are arranged. I prefer the relational description because it does not

involve the misleading suggestion of charlges of meaning. That is, differentiation is a synchronic relation, although, of course, it is manifested diachronically, too, as evolving meaning.

Consider the sentence frame: 'He cancelled the (N.P.)', completed with'check' and completed with'appointment'. 'Cancelled' adjusts its meaning to the completion noun. So too, with 'He wrote for a (N.P.)', completed as: 'He wrote for a living' and 'He wrote for an appointment'. The reverse occurs as well: 'He (V.P.) for a living', (i) 'He wrote for a living'; (ii) 'He married for a living'. (That, of course, is ambiguous between 'as an occupation' and 'to avoid having to work'.) Adjustment, (in the sense of'fit'), occurs in the single concatenation of words, the meaning of'for a living' in (i) is not altered in (ii), or vice versa, but rather what 'for a living' means in (i) and (ii) is (within a range of possibilities) settled by what words it is combined with.

The 'completion' word (in a pair of sentence frames) sometimes adapts to the frame elements already present and sometimes the frame words (or one of them) adapt to the completion word, depending upon which dominates.

The distinction between completion words and frame words is entirely relative to what one supposes one is given to be completed. If one is given something wanting a word or two to make a sentence, it is a frame; if one perceives oneself as given words wanting a structure as well as words to make a sentence, one needs a sentence frame. Although it is a matter of gestalt which one sees oneself as having and, consequently, lacking, the dominance relations are not arbitrary, as my examples will illustrate.

I want to counter the impression that the completion words and the frame elements 'already have some context-neutral kernel of meaning' that adjusts or adapts when they are combined. Outside a context a word has no signification at all (nor any denotation) but has (in its other contexts of occurrence) partly delimited possibilities or ranges of signification that are adapted, fitted, to context. The various adaptation results are the species of differentiation (analogy, metaphor, denomination and mere equivocation). J. Lyons (1977: 567), made a similar observation: 'Furthermore polysemy- the product of metaphorical creativity - is essential to the functioning of languages as flexible and efficient semantic systems.' Lyons was wrong that polysemy is the product of metaphorical creativity alone, or even chiefly. Other kinds of differentiation (proportionality and denomination) are even more common in discourse, as the 'clusters' described in chapter 4 will illustrate.

Lyons did not fully appreciate how words interact through dominance to make room for new ideas. Potentially infinite polysemy is Mi the feature upon which the indefinite expressive capacity of language rests and by which it can make room for endless new=ideas. (See section c, following.)

I disavow Lyons' e