[This is an electronic version of a paper that I presented at a "Symposium on Ethnoscience," held at the Central States Anthropological Society, St. Louis, Missouri, 28–30 April 1966. The resulting article was published in Anthropological Linguistics, Vol. 8, No. 8 (November 1966), pp. 1–12. For the convenience of the reader, I have filled out incomplete references and corrected minor errors in spelling and punctuation. The article is reproduced here with the permission of the publisher. Instead of revising the original article I have added a postscript to it in which I discuss how I view these questions now.]
This paper is concerned with the theoretical bases of Benjamin Lee Whorf's hypothesis which states that "all observers are not led by the same evidence to the same picture of the universe, unless their linguistic backgrounds are similar, or can in some way be calibrated."[1] In particular I shall try to describe what Whorf believed about the nature of language, and suggest that his basic assumptions about language led him to adopt specific theories about the relation of language to other phenomena: in particular to human thought and to culture.
A useful point of departure is the fact that Whorf was from the first interested most of all in the nature of human thought, in the "connections between ideas,"[2] to use his own expression, in the manner in which associations between ideas are set up, also in the extent to which such associations are common to all of us no matter what our peculiar psychic individualities are. It is interesting to read the first two articles in John Carroll's well-known collection of Whorf's writings, as evidence of his early interest in human psychology. In the second article, for example, he reviewed the psychological theories of the day, and reached the conclusion that none of them was able to yield what he termed a "topology of the inner or mental life." The concluding paragraph of this article is so characteristic of Whorf's approach to the problem that I shall quote it in full: "One fact that stands out to a detached viewpoint, but is not stressed by any of the schools, is the great and perhaps basic importance of the principle we denote by the word 'meaning.' Meaning will be found to be intimately connected with the linguistic: its principle is symbolism, but language is the great symbolism from which other symbolisms take their cue."[3] Thus Whorf was convinced that language offered the key to an understanding of the human mind even before he knew very much about linguistics.
But what was the character of the linguistic theories which Whorf became familiar with when his interest turned to language? Two aspects of those theories are relevant to this question. One is the nature of grammatical units, and the other is the question of form-classes, or parts of speech as they were called by traditional grammarians. Let us first discuss the grammatical units.
The basic grammatical unit recognized at this time was the morpheme, though many linguists did not use the term itself. Language was thought to be (among other things perhaps) a repertory of morphemes. A sentence consisted of a sequence of morphemes. Intermediate between the morpheme and the sentence, however, was the word. A word, like a sentence, was a sequence of one or more morphemes, but the processes by which morphemes combined to form words were different from those by which sentences were formed. Within the word at least two types of morphemes could be distinguished, often called roots and affixes. In an English word such as unthinkable, for example, think was regarded as the root, and the remaining two parts, un- and -able, were affixes.
The problems begin to multiply, however, as soon as we turn our attention to the semantic properties of these three units, the sentence, the word, and the morpheme. Traditionally, meaning was thought to characterize sentences and words, but nothing smaller than the word. This explains why dictionaries were always lists of words. The meanings of words could be described by means of definitions, and the meanings of sentences by inference from the meanings of their constituent words. It seems that nobody explicitly claimed that elements smaller than words have meanings until morpheme theory was formulated in the early 1880s by Baudouin de Courtenay.[4] It is interesting that Baudouin offered two different definitions of the morpheme. He defined the morpheme, on the one hand, as the smallest unit of morphological analysis, and on the other hand as the smallest linguistic unit capable of conveying meaning.[5] Thus the morpheme was from the very start a sort of hybrid creature. However that may be, until linguistic analysis assumed its procedural cast in the 1940s these definitions of the morpheme were the ones which the majority of linguists implicitly subscribed to.[6]
The tenacity of morpheme theory is not altogether surprising since it had much to recommend it. To use our first example over again, it is not absurd to claim that the meaning of the English word unthinkable is in some sense a product of the meanings of its component morphemes un-, think, and -able. One can plausibly argue that even if a dictionary of English did not contain this word, its meaning could be inferred from the meanings of its component morphemes as infallibly as the meanings of whole sentences can be deduced from the meanings of their component words. Semantically and morphologically then the word could be regarded as a derivative function of the morpheme, and it is not surprising that linguists, and especially Americanists, concentrated more and more attention on the pattern of morpheme occurrences in sentences.[7]
But although morphemic analysis functioned reasonably well on the word level, for example in the study of morphological systems, it was not able to account satisfactorily for the meanings of whole sentences. This is because sentences convey meanings for which none of the constituent morphemes can be made responsible. To take one of Whorf's own examples, [8] the English sentence My baby's name is Helen conveys among other things that the baby in question is female. But the meaning 'female' receives no specific morphemic expression.
The converse of this problem also presents itself. Many morphemes do not contribute to the total meaning of the sentence or word in which they occur. Take the English word capable. It clearly consists morphologically of two parts, cap- and -able. The -able is of course a common English adjectival suffix, recurring in eatable, lovable, and cap- must be the root and possibly recurs in the word capacity. The meaning of the suffix is roughly 'able to be carried out'; lovable, for example, means 'able to be loved.' However, the meaning of the whole word capable is 'able,' which is part of the meaning of the suffix. What then can possible be left for the cap- to mean?
The dilemma is even more embarrassing when the meaning of a morpheme is exactly the same as the meaning of a sequence containing it. Thus the morpheme high and the morpheme sequence heigh-t which contains high, are synonymous. But high is, of course, an adjective and height a noun. One might therefore claim that the final -t of height is a morpheme whose only meaning is the meaning-component present in all nouns. But to do this would involve claiming that all nouns do in fact have a meaning-component in common. This brings me to my second main topic: the problem of word-classes.
In traditional linguistics the parts of speech were a set of word-classes so constructed that every word in a language belonged to one and only one class. The classes were given definitions on the basis of which any arbitrary word could be correctly assigned to its part of speech. Some of these definitions referred to the meaning common to all the members of the class, as for example when a noun was defined as the name of a person, place or thing.[9]
There were also other types of definitions which do not concern us at this point. Suffice it to say that in Whorf's day it had become customary for professional linguists to criticize the semantic definitions of the parts of speech. I refer you to the writings of Henry Sweet, Hermann Paul, Otto Jespersen, and of course Leonard Bloomfield.
But the time-honored notion that the parts of speech had semantic properties lingered on, but in a different form. In his monograph Language, Sapir suggested that the classification of words into parts of speech corresponds to a crude classification of sensory experience. When we use a noun to refer to some element of subjective experience we are thereby representing it as a thing, no matter whether it is in fact a thing. "We speak of the height of a building or the fall of an apple quite as though these ideas were parallel to the roof of a building or the skin of an apple, forgetting that the nouns (height, fall) have not ceased to indicate a quality and an act when we have made them speak with the accent of mere objects."[10] According to Sapir, a naive speaker of English is deceived by the structure of his language into imagining that something similar is conveyed by the two expressions the height of a building and the roof of a building. Sapir went on to suggest that there is no inherent reason why any idea could not be referred to by any part of speech. When, for example, an attribute such as red is in fact referred to in some language by means of an adjective, this does not reflect the universal fact that adjectives are by nature words designed to designate attributes, but merely the fact that the structure of that particular language compels the speaker to represent the notion of red as an attribute. In some other language the grammatical structure might be different, and the concept of red might have to be represented by something other than an adjective. As Sapir himself put the matter: "Just as there are languages that make verbs of the great mass of adjectives, so there are others that make nouns of them."
Now if these modes of representing ideas are meanings, and if meanings are psychological entities (as semantic theorists thought at that time), then it follows that speakers of different languages convey different meanings even when they are referring to the same objective states of affairs, in so far as the grammatical structures of the languages they speak differ from one another.
Now an examination of Whorf's article entitled "Grammatical Categories" will show that he adopted Sapir's notion of word-class meaning. Thus he states: "Certain semantic and grammatical properties are assured in the word by selecting it from a certain class of fixed membership not coterminous with the whole vocabulary."[11] In other words, whenever a speaker selects a word from a certain class he thereby guarantees not only certain grammatical properties, but also certain semantic properties. Thus in selecting the word Helen from the class of feminine proper nouns in the example quoted above the speaker guarantees that the meaning 'female' is conveyed.
However, Whorf's position seems, at least on the surface, to be more complicated than Sapir's. In his article "Grammatical Categories," he places much emphasis on the difference between categories whose presence in a sentence is indicated by some special overt morphemic marker, and categories which betray their presence "by the systematic avoidance of certain morphemes, by lexical selection, by word-order that is also CLASS-ORDER, in general by association with definite linguistic configurations" (Carroll, p. 88). Plurality is an example of an overt category in English because plural nouns are marked as such by a specific ending. Gender, on the other hand, is a covert category since it is signaled by lexical selection (Helen rather than George, for example) and by the use of different pronouns as substitutes for the two types of noun (Helen is replaced by she, while George is replaced by he).
It follows from this that knowing to which gender an English noun belongs is not a question of knowing whether its referent is male or female, but is a result of having classified it in a particular way. The classification, however, is a grammatical one, not a conceptual one. Gender is not the linguistic reflection of a certain natural non-cultural distinction, i.e., not the linguistic recognition of some objective difference which is the same for all observers. To use Whorf's terminology, gender is a covert grammatical category.
At this point in the argument, Whorf could have concluded that linguistic categories have no non-linguistic, e.g., semantic properties. That is to say, he could have claimed that since grammatical categories do not reflect distinctions that are really out there in the non-linguistic world, one should let them be, accepting them for what they are, i.e. linguistic phenomena.
But Whorf believed that language is the key to an understanding of the human mind. We can imagine him reasoning in the following way. It is inconceivable that language could have such pervasive features as covert categories if they have no wider significance whatever. After all, morphemes, the ultimate grammatical constituents, have semantic properties. How can it be that morphemes are the only grammatical features having this feature? It is, of course, impossible to claim that precisely these considerations convinced Whorf to adopt his well-known position, but these propositions hang together in a plausible system and provide a fairly satisfactory way of looking at the relation between grammar and meaning. It is possible, therefore, that such arguments as these provided Whorf with the logical underpinning for his idea that the covert categories "may have profound influence on linguistic behavior" (Carroll, p. 92).
The logical bridge to the Whorf hypothesis, I should like to claim, is then the notion that grammatical categories represent "experience seen in terms of a definite linguistic scheme, not experience that is the same for all observers" (Carroll, p. 92). Whorf, like Sapir, invents a type of experience with which the grammatical categories can be correlated, and proceeds to claim that the nature of these subjective experiences can be explained by their being correlated in this way. Just as Sapir suggested that speakers of English forget that a noun like height indicates a quality and not an object, Whorf claimed that to us an event means "what our language classes as a verb." In fact, Whorf took the argument one step further and denied that the native speaker can ever be reminded of the truth: "And it will be found that it is not possible [emphasis mine] to define 'event, thing, object, relationship,' and so on, from nature, but that to define them always [emphasis mine] involves a circuitous return to the grammatical categories of the definer's language" (Carroll, p. 215).
But this "new principle of relativity," if taken seriously, leads to an unfortunate logical paradox.[12] If a human being's perception of categories such as event and object is indeed relative to the language he speaks, how could he ever be made aware of this fact? Thus, Whorf himself claimed that "in English we divide most of our words into two classes, which have different grammatical and logical properties"[13] and that "our language thus gives us a bipolar division of nature," only to state categorically: "But nature herself is not thus polarized." But a critic might have asked Whorf at this point: "How can you know for sure that nature is not polarized; after all, your knowledge that this is indeed the way nature is must be formulated in the very language which forces you to divide nature in a bipolar fashion?" In other words, if Whorf's hypothesis is true Whorf himself could not have known that it was true, nor could he have communicated that truth to his fellow speakers of English.
My contention that Whorf was not too concerned about the logical trap he had fallen into is borne out by an examination of the same passage I have been quoting from. Thus in trying to demonstrate that nature is not polarized, he adduces the fact that many nouns (e.g. lightning) denote events, rather than objects or things. But if Whorf knew that lightning was an event, how could he claim that an event is by definition "what our language classes as a verb"? Lightning is after all a noun.
Curiously enough, a similar flaw vitiates Sapir's discussion of word-classes in Language. Sapir reasoned as follows: "We imagine [...] that all 'verbs' are inherently concerned with actions as such, that a 'noun' is the name of some definite object or personality that can be pictured by the mind, that all qualities are necessarily expressed by a definite group of words to which we may appropriately apply the term 'adjective.' As soon as we test our vocabulary, we discern that the parts of speech are far from corresponding to so simple an analysis of reality." If the various we's in the passage refer to the same people, namely native speakers of English, one's conclusion can only be that they rid themselves of their delusions very easily, so easily in fact that one wonders whether they were ever truly deceived in the first place. In other words, if the delusion is serious, if speakers of English are profoundly deluded about the nature of reality, no such simple remedy as a cursory scrutiny of the vocabulary of English could cure them. If the delusion is not serious, if the speakers of a language have an occasional tendency to oversimplify the structure of reality on lines suggested by their language, a little quiet meditation will cure them of that, and the matter is not worthy of serious attention.
But it would be a mistake to conclude that Sapir's discussion of the nature of word-classes makes no sense at all. Sapir's remarks make a great deal of sense if one interprets the we's of the first sentence in the last quotation as referring to traditional grammarians as they were imagined to exist by Sapir and his contemporaries. Traditional grammarians are the people who are deluded and delude other people into imagining that a simple relation exists between metaphysical categories and word-classes. The we's of the second sentence, on the other hand, refer to people who have enough perspicacity not to be deluded by such arguments. They are able to realize that not all verbs refer to actions, that not all nouns refer to objects or persons, and that attributes are not necessarily referred to by means of adjectives.
Turning back to Whorf's articles we find that a similar reinterpretation of his discussion lifts it out of the realm of self-referential absurdity. What Whorf realized is that an event is not just anything which is referred to by a verb in English, that there are many nouns which do not refer to objects. Knowing other languages, particularly languages as remote from English as Hopi and Nahuatl, enables one to be even more aware of this fact than one might otherwise be, since the word-classes of different languages do not correspond to one another in any simple fashion.
Whorf's deluded prisoner of language who defines event as anything which his language classifies as a verb is properly a caricature of the traditional grammarian who philosophizes on the basis of the formal features of his own language. Hence, Whorf's hypothesis is really just another attempt to discredit traditional linguistics. The irony of the situation is that Whorf himself (unlike Sapir) was unacquainted with traditional linguistics, and thus did not really know what he was talking about.
The logical bases and historical sources of Whorf's hypothesis are therefore quite complicated. Ultimately they all have to do with that same anti-traditionalist attitude which linguists have adopted for the past hundred and fifty years. One logical component of Whorf's theory, viz. morpheme theory, would seem not to fall into this category. However, there is evidence that it too is part and parcel of the new comparative linguistics of the nineteenth century, i.e., part of the rejection of eighteenth-century universal grammar.
Thus, Theodor Benfey, a mid-nineteenth century historian of linguistics, in commenting favorably on John Horne Tooke's theory that all affixes were originally separate words, declared: "One sees how firmly he was convinced that a word cannot express more than is expressed by the conjunction of its constituent semantic elements, a conviction which the great Indian grammarians had reached in essentially the same form through an analysis of Sanskrit and which modern linguistics has established with regard to the Indo-European languages and many others."[14] Benfey, it should be pointed out, considered Horne Tooke to be the forerunner of modern linguistics, and contrasted him in this respect with the eighteenth-century universal grammarians represented by Horne Tooke's contemporary James Harris. But the most interesting aspect of this quotation is that it contains a clear prefigurement of the notion of the morpheme. What Benfey is stating as an item of contemporary dogma is the idea that the meaning of a word cannot comprise more than the sum of the meanings of its constituent morphemes. The cardinal sin committed by the universal grammarians, according to Benfey, was to read into words meanings not found in any of their constituent morphological parts, a propensity fostered by their lack of any rigorous word etymology.
But by the twentieth century it had become obvious that words, and especially whole sentences, do in fact convey many ideas not contained in any of their constituent morphological parts. Linguists again needed a theory to explain this residue of meanings, and some notion of word-class meaning was again called for. At the same time linguists were still fighting eighteenth-century universalism and were loath to fall back on the very notion of word-class meaning which they had spent so much time attacking. One of the possible logical solutions of this dilemma was the Whorf hypothesis.
Thus Whorf was not really proposing a new principle of relativity. But more than this, he actually had a perfectly definite notion of rationality and universal grammar. Once more a glance at Sapir's monograph Language will show us what to look for. In his discussion of grammatical concepts in chapter 5 of that book he pointed out the following: "In English we have made up our minds that all action must be conceived of in reference to three standard times. If, therefore, we desire to state a proposition that is as true tomorrow as it was yesterday, we have to pretend [emphasis mine] that the present moment may be elongated fore and aft so as to take in all eternity." To this he adds in a footnote: "There are many 'primitive' languages that are more philosophical and distinguish between a true 'present' and a 'customary' or 'general' tense." What Sapir is proposing here is that there is a standard of rationality (or philosophicalness) against which many primitive languages make a better showing than Indo-European languages. Put another way, our notion that Indo-European languages are the embodiment of rationality is a vain conceit: in reality they are in many respects less rational, less philosophical than many non-Indo-European languages.
Turning back to Whorf's articles we discover the same idea expressed repeatedly. Thus in his paper "The Punctual and Segmentative Aspects of Verbs in Hopi," Whorf shows that Hopi verb forms belonging to a certain morphological category refer to vibratory phenomena such as zigzagging, flapping, and quivering. He concludes: "The Hopi actually have a language better equipped to deal with such vibratile phenomena than is our latest scientific terminology." Again, in his paper "A Linguistic Consideration of Thinking in Primitive Communities," he discusses the supposed mystical mentality of primitive people and points out: "Yet many American Indian and African languages abound in finely wrought, beautifully logical [emphasis mine] discriminations about causation, action, result, dynamic or energic quality, directness of experience, etc., all matters of the function of thinking, indeed the quintessence of the rational [emphasis mine]. In this respect they far outdistance the European languages."
It is as if Sapir and Whorf believed that all languages were equally irrational, but some were more irrational than others, namely our own much-vaunted Indo-European languages.[15] That this is also a form of anti-traditionalism is too obvious to need emphasizing. Sapir and Whorf saw themselves in the role of liberators freeing linguists from a prejudice in favor of Indo-European languages and the kind of grammatical theory which had grown up to describe them. But they were liberators, not anarchists. Their professed relativism was in reality only skin-deep.
The conclusions I would recommend drawing from this historical survey of Whorf's and Sapir's linguistic theories are as follows:
1. Like other forms of relativism, linguistic relativism is logically untenable.
2. Sapir's and Whorf's relativism was spurious. They were in reality dogmatic anti-traditionalists.
3. The question of the relation of language to culture, if the question should turn out to be meaningful at all, is unaffected by the preceding conclusions.
Needless to say, the secondary literature bearing on the Sapir-Whorf hypothesis and on Whorf's life and ideas is oceanic. Within that literature, I have found two books especially helpful, namely Helmut Gipper, Gibt es ein sprachliches Relativitätsprinzip? Untersuchungen zur Sapir-Whorf-Hypothese, Conditio humana (Frankfurt am Main: S. Fischer Verlag, 1972),[16] and Penny Lee, The Whorf Theory Complex: A Critical Reconstruction, Amsterdam Studies in the Theory and History of Linguistic Science, Ser. III, Vol. 81 (Amsterdam/Philadephia: John Benjamins, 1996). For a general survey and critique of the Whorf literature see Dell Hymes & John Fought, American Structuralism, Janua Linguarum Series Maior 102 (The Hague: Mouton, 1981), pp. 97-100.
A revealing analysis of some of the antecedents of the Sapir-Whorf hypothesis has been provided by John E. Joseph in his From Whitney to Chomsky: Essays in the History of American Linguistics, Amsterdam Studies in Theory and History of Linguistic Science, series III, vol. 103 (Amsterdam/Philadelphia: John Benjamins, 2002), pp. 71-105. In that connection, one fact that I find significant is that the Argentinian writer Jorge Luis Borges expounded the world-view theory in a charming story, dating from 1940, entitled "Tlön, Uqbar, Orbis Tertius," which has appeared in many collections of his writings. It seems highly unlikely that Borges was influenced by Sapir or by Whorf. The idea must simply have been in the air at that time. Borges's story may be found, for instance, in his Ficciones, translated by Anthony Kerrigan (Grove Press, 1962), pp. 17-36; see p. 23 for the relevant passage.
Another general observation that I have always wanted to make with regard to Whorf's extreme version of the world-view hypothesis is that people who habitually speak more than one language (and both I and my wife fall into that category) switch back and forth between languages completely unaware of the fact that each time that happens they allegedly move from one world view to another. Moreover, the same observation may be made of all people who speak more than one dialect of the same language. This means that most of the human race handles different world views with unconscious ease.
Helmut Gipper relates the Sapir-Whorf hypothesis to a long tradition in German linguistics, exemplified by such figures as Wilhelm von Humboldt (1767-1835) and Leo Weisgerber (1899-1985). He also takes on the formidable Marxist philosopher Adam Schaff, author of Język a poznanie [Language and Cognition], (Warsaw 1964, German translation Sprache und Erkenntnis, 1964). Schaff was opposed to Sapir's and Whorf's ideas for ideological reasons, which Gipper discusses at some length. What makes Gipper's contribution to the Whorf debate especially valuable is that he himself went on field trips to the Hopi reservation in northern Arizona in 1967 and again in 1969 in order to test Whorf's assertions about the structure of the Hopi language. To his book, Gipper also appends (pp. 297-304) a short but penetrating essay by Andrea Stahlschmidt entitled "Die Zeit im Hopi-Verbum."
Penny Lee's monograph is valuable in that her exegesis of Whorf's ideas is to a large extent based on unpublished manuscripts that she consulted in the Yale University archives and elsewhere. She interprets Whorf's theories positively, and her attempt to systematize them conveys a clearer picture than Whorf did himself. This is especially true of his inchoate theory of "configurative linguistics," based on his reading of the Gestalt psychologists (see especially The Whorf Theory Complex, pp. 128-136.). It is unfortunate, of course, that the primary sources on which to base an interpretation of Worf's articles are so fragmentary and elusive. Had he lived longer it seems likely that he would have eventually laid out his theoretical vision more comprehensively and perhaps more comprehensibly. Since he was a writer with a rather opaque style attempts like Penny Lee's are to be welcomed.
Both Gipper and Lee deplore the fact that so much of the literature on the so-called Whorf hypothesis has been not only unsympathetic but also badly informed. The hostility directed at Whorf's theories from so many different points of view would surely warrant a complete monograph in its own right! John Joseph makes some perceptive comments on contemporary misinterpretations of Whorf's ideas in his From Whitney to Chomsky, pp. 181-196. In my view, the irony of the situation is that Whorf took aim at traditional "Indo-European" grammatical theory only to be attacked subsequently by a group of professional linguists who were as opposed to that theory as he himself had been! Whorf died in 1941, and from then on his influence on later developments in linguistics faded. His fate seems to have been that of a typical outsider. If he had lived longer one suspects that American linguistics might have developed differently. It is interesting to note that Sapir's influence likewise faded after his death in 1939, though less dramatically, and unlike Whorf Sapir was definitely an insider.
In my 1966 article I suggested that part of the reason for Whorf's hypothesizing cross-linguistic relativity of world-views lay in his rejection of what he took to be a grammatical theory based too exclusively on Indo-European languages. In his 1972 book, Gipper took me to task for that suggestion, and I think that the way he expressed his criticism is worth examining. What he actually said reads as follows (p. 80):
"Whorfs Relativitätsgedanke ist also weniger absolut, weniger radikal gemeint, als es auf den ersten Blick der Fall zu sein scheint. Ob es sich indessen nur um einen bloß 'hauttiefen' Relativismus handelt, der vor allem als Attacke gegen den Dogmatismus und die Vorurteile der traditionellen indoeuropäischen Sprachwissenschaft zu verstehen wäre, wie Keith Percival es in einem ungedruckten Paper annimmt, dürfte trotzdem zweifelhaft sein."
[Rough English translation: 'Therefore, Whorf's idea of relativity is less absolute, less radical than it seems at first glance. However, whether what we are dealing with here is a merely "skin-deep" relativism that may primarily be interpreted as an attack on the dogmatism and prejudices of traditional Indo-European linguistics, as suggested by Keith Percival in an unpublished paper, is nevertheless most probably questionable.']
While Gipper characterizes my position with admirable clarity, he marshals no evidence, either factual or otherwise, that might cast doubt on it. I agree, of course, that even if my suggestion hit the nail on the head the factor that I focused attention on is only part of an extremely complicated picture. The immediate question is surely this: can my suggestion be shown not to be true in any respect whatever? Or let me express this a little differently, was Whorf's attitude to traditional grammar irrelevant to the issue of how his world-view theory originated and took shape? This seems to me to be implausible. After all, some sort of world-view theory had existed for a long time, as writers like Gipper have pointed out. Both Sapir and Boas had expressed similar views. For someone like Benjamin Lee Whorf it was especially important to be able to account for the full range of meanings that utterances convey. The limitations of a purely formal combinatory analysis of sentences were especially irksome to someone like him, and this may well have been a potent factor that predisposed him to invoke and elaborate a world-view theory as an explanatory device.
His profound conviction that a purely formal analysis of the utterances proferred by informants does not yield a complete understanding of the language in question is reflected in his published writings. For instance, in the revealing article "Gestalt Technique of Stem Composition in Shawnee" (see Carroll, pp. 160-172, and my remarks in footnote 12 of this article), he makes the following remark: "Our problem is to determine how different languages segregate different essentials out of the same situation. This is often a crucial question in the description of a language, and it must not be supposed that it has been answered by an account of the formal rules for combination into sentences of the lexemes and other morphemes that represent the language's segregation of essentials out of situations" (Carroll, p. 162).
Significantly, it is at this point that Whorf introduces his idea that recent findings of configurative or Gestalt psychology afford a culture- and language-independent method of investigating how speakers initially perceive the phenomena on which their respective languages base a segmentation of reality. Neither traditional grammar, with its outmoded terminology, nor common sense, even when it is quasi-scientific, would be suitable for achieving this aim. As Whorf expresses it himself: "This cannot be done by describing the situation in terms of subject-predicate, actor-action, attribute-head, etc., for any scientific use of such terms contemplates that they shall have a variable meaning as defined for each language, including the possibility that for some languages their meaning shall be nil. Neither can it be done wholly by familiar terms from the common-sense type to the quasiscientific, as by trying to break up the situation into 'things, objects, actions, substances, entities, events.' Cautious use of such terms may be helpful, perhaps unavoidable, but it must be remembered that in their ranges of meaning they are but the creatures of modern Indo-European languages and their subsidiary jargons, and reflect the typical modes of segmenting experience in these tongues" (Carroll, p. 162).
In this connection, one should perhaps recall another component in Whorf's personal make-up. As is well known, he initially approached the problem of semantics by steeping himself in a two-volume work by the eccentric Biblical philologist Antoine Fabre d'Olivet (1768-1825), who elaborated a far-reaching analysis of the Hebrew triliteral root based on the claim that each letter composing the root conveys a separate meaning (see Carroll, p. 8). Later, Whorf revealed in a letter that "the reason I am interested in every phase of linguistics is that I am primarily interested in something that might be styled 'intra-atomic' linguistics" (Lee, p. 7). At the same time, he could not help but sense that what he was interested in transcended conventional linguistics. At one point, he confesses that what he was after were "relationships uniting hundreds of root words that have always been supposed to be entirely separate and unrelated and indeed are so in the sense of ordinary linguistics" (italics mine) (Lee, p. 3). This kind of statement gives one a vivid sense of the extent to which Whorf himself realized that his point of departure lay outside the intellectual mainstream. He mistrusted not only traditional Indo-European grammatical theory, but also common sense based on scientific or pseudo-scientific reasoning, and even linguistics itself, the field in which he was preparing to immerse himself.
In these circumstances, it is hardly surprising that even after almost a decade of initiation into academic linguistics the attitude to his ideas of his closest colleagues in the field tended to be unenthusiastic. It is, for instance, interesting to read the sections of Lee's monograph that deal with Whorf's relations with the Yale linguist George L. Trager. From 1937 to 1938, the two of them planned to team-teach a course in the Yale department of Anthropology, which ended up being taught by Whorf alone after Trager succeeded in obtaining funds for a study tour to Germany. The extensive written report on the theorizing that went into their year-long course on "configurative" linguistics was apparently drafted by Whorf alone, without Trager's co-operation, because, as Lee plausibly surmises, Trager lost interest in the project (see Lee, pp. 251-280, esp. p. 253fn.).
To end in a philosophical vein, one valuable point that Gipper makes in his book Gibt es ein sprachliches Relativitätsprinzip? (see especially the introductory chapter to that book) is that one should distinguish carefully between relativism as a philosophical position and the theories of relativity that Einstein propounded in the twentieth century. Outside physics, relativity has been confused with the vague general notion that everything is relative, and this confusion has had unfortunate consequences.
In fact, the main idea behind Einstein's theory of relativity was not that everything is relative. Rather, Einstein argued that certain things that had been previously thought to be absolute were in reality relative. However, Einstein's new theories continued to be firmly based on a premiss of non-relativity. According to his theory, natural laws are not dependent on some arbitrary frame of reference, but are invariant and hence not relative. The main thesis of Einstein's relativity theory was merely that there are quantities that classical Newtonian physics had assumed were absolute which in reality are relative, namely relative to observers (as for instance simultaneity).
Unless one admits that one is just speaking metaphorically it is absurd to claim that a principle of relativity in Einstein's sense could be a fundamental feature of any linguistic theory.