This page was updated by Ralph Kenyon on 2018-01-15 at 03:25 and has been accessed 6194 times at 48 hits per month.

The Paradox of Reductionism

1985 by Ralph E. Kenyon, Jr.. *

October 17, 1985

Quine makes some significant philosophical contributions, but one cannot follow him if one is chopping down all his 'insubstantial' foundations.1 'Hatchet philosophers' won't get to his tapestry, as they will be two busy with the individual fibers. Threads are spun from detached fibers. Cloth is woven from threads, yet there is no bond between individual fibers in the cloth; moreover, A great many individual fibers may be broken and removed without destroying the tapestry. Also, the tapestry does not depend upon any single set of fibers. Using a Hatchet on Quine merely chops some fibers, and even some threads; it can even make some holes in the tapestry, but does not destroy it.

Reductionism, sometimes called "nothing-but-ism", claims that the entire significance of the tapestry is nothing but the direct consequence of the fibers; changing the fibers changes the significance. No error tolerance is built in. The threads are at a level of abstraction above that of the fibers, and the tapestry a level above that; some construe reductionism as denying that the tapestry has any significant apart from its fibers. Quine denies reductionism.

In philosophy, reductionism claims all talk is 'reducible' to talk of stimulation; that is, that talk of things other than stimulation is merely a convenient shorthand for talk of stimulation. On the other hand, some claim that some talk cannot be reduced to talk of stimulation, that is, that some talk has significance apart from the significance of any talk of stimulation into which it might be reduced. Why is there so much currency in the notion that the abstract can be represented solely in terms of stimulation?

Philosophers know well that equivocation in the use of a term, or statement, may produce a fallacious argument. Resolving the fallacy in such an argument usually requires distinguishing among the equivocated uses of the term or statement in question. It seems to me that something akin to that is going on in the currency and denial of reductionism. The 'problem' here is much more subtle; the contrast involves major differences in the world-views of the two claims, differences which are far less apparent because it involves not a single term or statement, but the very fabric of terms and statements. The two implicit world views are in great accord over much of human linguistic activity, but differ greatly in their respective perspectives.

Communication about a new world-view is very difficult. The use of significant terms in the new world-view will differ from the way in which those terms were used in the old world-view. To one not conversant with the new world-view, some statements made from the perspective of the new world-view may seem self-contradictory, false, or otherwise unacceptable. Traditional philosophy is such an old world-view. Quine speaks from a new world-view.

Perhaps the key term distinguishing the new world-view from the old is "levels", in particular, "levels of abstraction". Our senses 'abstract' a limited range of energy from the manifold spectrums in our environment. (We hear vibrations ranging from 10 to 20,000 Hertz; we see electromagnetic vibrations ranging from 4000 to 7000 Angstroms; etc.) From those ranges we abstract a few characteristic bits of information. From thousands of such bits processed in parallel, we abstract certain configurations or patterns. From the patterns we abstract certain learned verbal associations. At still higher levels of abstraction we make inferences about the patterns and, further still, we make judgments about the inferences and patterns. We carry this process of abstracting to higher and higher levels. It is not yet known how neurolinguistic processes transform neurological levels of abstraction into verbal levels of abstraction, but in human processing, many levels of processing occur together.

In the use of words, these many levels come continuously into play. Some words are used in such a way that they apply differently at different levels of abstraction. We could say these words have different 'meanings' at different levels of abstraction, but perform the same or a similar function at all levels of abstraction. Terms which function in the same way at all levels of abstraction, but which have different 'meanings' at different levels of abstraction are called multiordinal2.

"Meaning" is such a multiordinal term. There has been much argument and discussion about it. Arguments have been made that 'meaning' is reference, sense, definition, and etc., but none of these has proven satisfactory. That is partially because each of these applies at only one level of abstraction, or at most a very few. Because "meaning" is multiordinal, its use does not identify a level of abstraction; the level of abstraction must be specified by other, contextual, information. Once the level of abstraction is determined by the context, the answer can be framed in terms of its referent, sense, definition, or etc.

In terms of levels of abstraction, 'reference' functions at low levels, while 'definition' functions at verbal levels (more abstract). The question of the 'meaning' of a term does not specify the intended level of abstraction. We must infer the intent of the questioner in order to select which level of abstraction to aim our answer for. It may be a question of reference, or a question of connotation, or a question of definition, or etc. In philosophical investigations where the 'meaning' of "meaning" is sought, the level of abstraction intended is ambiguous, and as this very sentence illustrates, "meaning" is also self-reflexive.

The new world-view resolves this double problem by identifying 'meanings' (loosely) as responses in people to linguistic behavior. The key element is expressed by "purpose" or "intent". In the new world-view people 'make-sense' of talk by inferring the approximate level of abstraction intended by the speaker. Because "purpose" is not normally multiordinal, the speaker's purpose provides the means to select the level of abstraction for framing an answer. What does a person 'mean' by using certain words? He 'means' what he intends. When philosophers seek 'meaning' without regard to any levels of abstraction, they risk equivocation between levels of abstraction. 'Meaning' cannot be obtained independently of any agents, as it is people who make sense of things. What sense is made of words depends upon the perspective or world-view brought to bear on those words. In particular, it makes a great deal of difference between the old and the new world-views.

The paradoxical treatment of reductionism stems from the multi-level perspective as seen from a uni-level perspective.

1. In the new world-view, words do not have 'meaning'. A statement has significance to a person; people make sense of words. Quine argues that we can dispose of the suppositious entities called 'meanings'. We must remember that speech sounds are transmitted, and these sounds impinging upon our ears. We interpret the sounds as 'words' (low levels of abstraction to patterns) and bring our own private experiences to the interpretation of these patterns. Each of us interprets these 'words' in the light of our own experience. (To the German ear "man" makes different sense than to the English ear.) When we make sense of sounds we quite literally construct 'meanings' for them. For Quine it is the subsequent (linguistic) behavior that is relevant in a communication situation.

Behaviorism treats us as stimulus-response mechanisms and eschews seeking to account for responses in terms of internal structure. Behaviorism denies 'meaning' is germane. Modeling creates internal structure in the form of state diagrams (finite state automata) to relate input to output. Modeling replaces 'meaning' with structure.

Actually, 'meaning' is no longer important. In its full context, it is behavioral response (including linguistic behavior) which is now important. Agents have motives for action, and these motives entail satisfaction tests in terms of environmental states. The emphasis on 'meaning' by the old world-view is simply as a means to the end of achieving desired states of affairs.

2. All talk spans various levels of abstraction. Talk itself is highly abstracted from stimulation situations. The mechanism of reference indicates lower levels of abstraction, while synonymy indicates higher levels of abstracting. "Unicorn" abstracts (to the best of our knowledge) primarily from linguistic experience, while "cup" abstracts much more from non-linguistic experience.

Reductionism is an attempt to maintain explanations in terms of a single level of abstraction, the most 'objective'. As such it exemplifies resistance to the multi-level world-view. Quine sought to deny reductionism and to explain the new world-view. The new world-view does allow a place for a certain modified view of reductionism in explaining the correlation between lower level abstractions and higher level abstractions. The old world-view (single level) 'reduces' higher level abstraction to the one chosen level with the attitude that the higher level structure is 'nothing but' the lower level structure. That is, the higher level abstraction has no significance apart from the lower level 'stuff' from which it was abstracted. If there is only one level, then a thing has its status of existence only in terms of that one level. The status of existence cannot differ from level to level (as there is only one level) and that which is abstracted from some 'stuff' must either be that 'stuff' from whence it is abstracted, or be something else. If it is something else, an apparent contradiction results in that the same stuff is two distinct things. Reductionism denies that which is abstracted from the 'stuff' by claiming it is nothing but the 'stuff' (perhaps in a disguised form). In the absence of any notion of levels, this is understandable. On the other hand, those who deny reductionism would claim that that which is abstracted from the 'stuff' has its own existence, and is not 'derived' from the 'stuff'.

Adding the notion of levels allows the co-existence of both levels of abstraction. We do not have to say the tapestry is nothing but its threads. In claiming that more abstract statements can be 'reduced' to statements in terms of immediate experience, the unfortunate connotation of 'nothing but' remains. In the new world-view, it is said that "More abstract statements can be abstracted from statements in terms of immediate experience." This emphasizes that higher and lower levels of abstraction coexist and implies that abstraction is a process engaged in by an agent. "Exist" is also a multiordinal term, and its 'meaning' depends upon the level of abstraction at which it is used. The status of existence of lower and higher levels of abstraction differ by the difference in 'meaning' of "exist" at different levels of existence.

The analytic/synthetic distinction makes reference to meanings of terms, which are of little significance in the new world-view; they can only be justified in terms of abstractions made by lexicographers.

In the old world view, some statements are said to be analytic in virtue of their 'meanings'. The new world-view cannot permit this definition because words do not have 'meanings'. The statements which were considered analytic or synthetic, however, can still be made. The question is, "Can the analytic/synthetic distinction be captured in the new world-view?"

Statements which are logically true are analytic. Statements which are empirically true are synthetic. These two seem to be preserved in the new world view. In both cases there is a high degree of consensual validation concerning these statements. Statements involving the substitutability of terms remain murky. Is a bachelor an unmarried adult male or a never married adult male? Before divorce was common we distinguished between bachelors and widowers (very rare), but now that divorce is common, the term is used more generally. Assuming we can agree on these, how do we argue for the classification of the statement "All bachelors are un-married adult males." We cannot say that it is analytic because of the 'meaning' of the terms in question. We could say that there is general agreement among people that, depending upon the context, "un-married adult male" may be substituted for "bachelor".

One view of reductionism would have it that they are the same. In the old world view, the statement is analytic because "bachelor" and "un-married adult male" have the same 'meaning'. In the new world-view, the 'meaning' depends upon the level of abstraction intended by the speaker. Consider another example, that of the morning star and the evening star.

When we say the morning star and the evening star are the same, we appeal to the lowest level of abstraction, that of reference.

When we say the morning star and the evening star are not the same, we appeal to higher levels of abstraction, which could include context (morning/evening), linguistic form (quote), or implicitly indicate the source from which the abstraction is drawn (idiosyncratic knowledge & experience: he may not have ever learned from his (linguistic) experience to say they are the same). The 'meaning' of "The morning star is not the same as the evening star." is ambiguous because we have no collateral information with which to limit the speaker's intended level of abstraction.

In drawing distinctions with multiordinal terms, that distinction must apply at all levels of abstraction. Since multiordinal terms are used at all levels of abstraction in the same way, but with different 'meanings', drawing the analytic synthetic distinction may not use multiordinal terms. Such a proposed definition provides for distinguishing at low levels of abstraction in one way, while at higher levels of abstraction the distinction is drawn in such a way as to cut across the lower level distinctions. In truth, the analytic synthetic distinction selects certain high level true statements as analytic, and certain true statement which rely on lower levels of abstraction as synthetic. It may well be that the analytic/synthetic distinction is simply a rudimentary awareness of level of abstraction in the 'meaning' of "true".

In the new world-view, one asks what do you 'mean', not what does the sentence mean. In fact, the request is not for any 'meaning' itself, but merely for additional linguistic behavior, and should be more properly expressed as "Use other words please, perhaps with more detail." We all know full well that when we say to someone "What do you mean?", his repeating exactly what he said before is not an acceptable answer. We require that the person provide more detailed information (lower level of abstraction), or a 'simpler' explanation (higher level of abstraction), or to use other words (synonymy at the same level of abstraction).

Technically, 'meanings' are not words; we cannot ask for them. We can only ask for a new expression of the intent of the speaker (which may have changed, as we infer when he says," "Forget it!" -- How many "Gavagai's?" can a person take.) Lexicographers examine the context in which citations of terms appear and formulate synonymous expressions; they create dictionary definitions for words.

. . . citations tend to fall into what may be called "contextual clusters." I hesitate to call these clusters "meanings", because the word meaning has been subject to such muddled controversy. I strongly support the position of the logician Wilard Quine, who has said that we should "continue to turn our backs on the suppositious entities called meanings."3

These definitions provide what is usually asked for when we ask what does it 'mean', but we must recall that dictionary definitions follow people's usage, and hence their intents; that the dictionary definition is an abstraction from many selected usages, each with an individual purpose. To say that a dictionary "gives meanings" begs the question since 'meaning' is not words. According to Quine, words do not have meaning. Therefore, dictionaries cannot give them meaning. It is our reading of the definition, in which we assume that 'meaning' is already functioning for us for other words, that allows us to make sense of the word. As a matter of actual fact, we don't make sense of the word looked up; we associate the sense we made of other words with that word and thenceforth (assuming our memory is good) recall that sense when we encounter the word again. The procedure suggests that making sense of words is 'recursive' in nature, but is greatly complicated by multiordinal terms.

The degree to which the new world-view is pervading language and thought is sporadic; there are traditional hold-outs and there are perspectives which adopt some aspects of the new world-view, but fail to adopt it fully.

Equivocation is buried in the different levels of abstraction. To say that talk of 'things' (higher level abstractions) can be expressed in terms of talk of direct experience (lower level abstraction) is consistent with the multi-level perspective since both lower and higher level abstractions co-exist. Reductionism becomes simply a choice to focus at lower levels of abstraction, and represents a bias of the speaker. Denying the higher levels of abstraction however, is not consistent with the new world-view; denial of the form "is nothing but" in necessitated by the need to choose only one level of abstraction, and opts for the more objective, lower level as 'reality'. The single level perspective must place multi-level terms in a context of only one level; by such choices, a level is selected as 'real', and all other levels are to be explained in terms of the selected level, all the while failing to admit of levels. Empiricists choose lower levels of abstraction. Platonists choose higher levels of abstraction. Etc.

A rudimentary awareness of levels of abstraction manifests itself in the world view that "sometimes one must take levels into consideration". A better developed view recognizes levels as pervading the majority of linguistic experience.

Denying Reductionism takes two forms. One form, that of the single level perspective, requires that what are perceived as higher level abstractions by the new world-view have the same status of existence as what are perceived as lower level abstractions. In short, 'things' have the same status of existence as direct 'experiences'. Consequently, 'things' cannot be 'reduced' to direct experience.

The second form of denial, from that of the multi-level perspective can seem (from the uni-level perspective) at times to be an affirmation of reductionism, in its recognition that higher level abstractions 'are abstracted from' lower level abstractions; yet it can also seem to be a denial when it eschews identification across levels of abstraction with the prohibition "not nothing-but".

It seems to me that Quine is denying reductionism in the second manner, but is taken to deny it in the first manner by most philosophers. It also seems to me that he fails to fully appreciate the application of the new world-view, however, when he goes on to relate reductionism to the analytic/synthetic distinction. The distinction is unclear when 'meaning' is taken as a single level phenomenon. Across levels of abstraction, the distinction is confused. At the most abstract levels, analyticity of statements in the form of logically necessary truths (e.g. all bachelors are bachelors) is clearly understandable. At the lowest levels of abstraction synthetic statements are also clearly understandable. Their truth depends upon 'fact', which are simply statements at low levels of abstraction; e.g. (I see that) The grass is green. The same process, abstraction, relates lower to higher levels, and synthetic applies to lower levels, while analytic applies to highest levels, yet there is no clear distinctions among levels where one can say that one stops and the other begins. Abstracting from "unmarried adult male" to "bachelor" shows that two levels are involved, and the truth of "all bachelors are bachelors" at the higher level is not the same as the truth of "all unmarried adult males are bachelors". (Aside from the distinction between never married, divorced, and widowered). True in virtue of 'meaning' cannot work for the new world-view.

Quine denies both the analytic/synthetic distinction and reductionism. Unfortunately, when he goes on to identify them as each manifesting the other, he has failed to recognize that both fail on the merits of the change in world-view he has begun to assimilate. It seems to me that he has not fully apprehended the new perspective, or he would recognize that the analytic/synthetic distinction distinguishes among levels in the 'meaning' of "truth", while reductionism relates higher to lower levels of abstraction.


  1. Willard Van Orman Quine, Two Dogmas of Empiricism, From a Logical Point of View, 1953, Harvard University Press.
  2. Alfred Korzybski, Science and Sanity, 1933, The International Non-Aristotelian Library Publishing Company, (4th ed, 1958). p. 14
  3. Allen Walker Read, The Relation of Definitions to Their Contextual Basis, ETC.: A Review of General Semantics, Vol. 39, No. 4 (Winter, 1982), pp. 318-27, (Reprint of a paper delivered befor the Dictionary Society of America June 12, 1981, and published in its Proceedings..)