Language and Linguistics
Ashok R. Kelkar





0. Introductory.* I do not propose to say anything new here so much as to reexamine briefly some of the fundamental properties that have been recurrently predicated of ‘natural languages’ and to follow up some of the implications of what this reexamination reveals.  The three properties we are going to take up, namely, arbitrariness, necessity, and the duality of patterning (respectively in §§ 1, 2, 3), are logically distinct from each other.  It is true that they have frequently been confused with each other in the past, but that is all the more reason for making a special effort to keep them apart in mind.1


While considering arbitrariness and necessity, we can very well assume that Saussure’s analysis of the linguistic sing into the significant and the signifié is an adequate one and not worry too much about the various refinements, elaborations, reservations, and revisions proposed by later thinkers.  With pattern duality, as we shall see, it is another matter.


Having considered the three properties and their interrelations (§ 5).


1. Arbitrariness. Cross-linguistically, the relation between the significant and the signifié is arbitrary.  There is no extra-linguistic reason2 why the given significant should not be correlated with other than its usual signifié in a given language, and vice versa.  We are quite justified in laughing at the English soldier who criticized the French for calling a cabbage a shoe (Fr. Chou /šu/).


A symbolism is non-arbitrary when there is some sort of an appropriateness about it-for example, the geometrical similarity between a map and the original landscape, the similarity of responses that make darkness a symbol of ignorance, the stimulus-response relationship that makes bright red more suitable as a symbol of danger than, say, pale blue.

1.1. Marginal exceptions.  In such phenomena as onomatopoeia, sound symbolism, or phonaesthesis, we have to concede that extra-linguistic reasons can be given and so to that extent the arbitrariness of language symbolism is mitigated.  But even here it is a common-place that these fancied connections between the sound sequence and the sound or other thing referred to vary considerably from language to language.  Thus the element of arbitrariness enters into the picture again.


A more convincing case of linguistic iconism is that of the quotation or citation, when the significant is a specimen of the signifié, namely, some form-token (Archimedes said, Eureka), form-type (the form ‘took’), or set of form-types (the pronoun “thou” is obsolete – referring to thee, thy, and thine as well) in a language.  Even here, minor stipulations have to be made for each language – for example, Marathi employs the masculine singular nominative in citing a variable adjective or adds the so-called ‘inherent’ vowel /ә:/ in naming a consonant.


Tactical features in the significant admit of a subtler iconism.  The overt order of items within a sentence often strikes us ‘natural’.  What is more ‘natural’ and appropriate than that we should speak of *cash and carry or* children and wife, or what words which go together in meaning tend to go together in the temporal speech sequence, or that connectives should come between the items they connect?  Indeed when we come to the largest grammatical unit in size, the sentence, the grammarian gives up and refuses to give any linguistic account of why the sentences in a text occur in the order in which they do.3 Similar considerations will be seen to apply to tactical features other than overt order.


Finally, we could take note of cases of indirect iconism.  In words like cuckoo the significant is iconic of the cry, which in turn is metonymically associated with the signifié proper, the bird.  When we say, The word “man” is spelled with three letters, the significant is iconic of the spoken form which in turn calls forth the signifié proper, namely, the written form.  In the obscure area of intonations, vocal qualifiers, and the like, the limits of the principle of arbitrariness could perhaps be formulated in some such terms: the significant is iconic (not acoustically so much as in respect of articulation) of a gesture which is itself a sign (not necessarily iconic) of the signifié proper, namely, the attitude of the speaker to the addressee and to the matter-in-hand.


Cases of non-arbitrary linguistic symbols other than the iconic ones are not easy to think of.  One can think of the choice of an overt order dictated by what is appropriate according to good manners –ladies and gentlemen! or you and I rather than *gentlemen and ladies! or* I and you.


For all such exceptions to arbitrariness, one can say that to the extent that the extra-linguistic reasons in question are culture-bound in their cogency, they are a less serious threat to the principle.


2. Necessity. Intra-linguistically, the relation between the significant and the signifié is a necessary one-a relation of solidarity or coimplication.4    There is never a linguistic reason why the given significant should be correlated with other than its usual signifié in a given language, and vice versa.5 This is the linguistic or notational analogue to the logical axioms of identity and non-contradiction, A is A and A is not non-A.


A symbolism is non-necessitarian when there is either one-to-many or many-to-one or many-to-many correspondence between signifiants and signifiés.  Signs may be either ambiguous or alternating. 


2.1. Marginal exceptions.  No natural language is fully necessitarian, though its users have to proceed as if it is (and the same goes for its students).


When the same signifiant is correlated with different signifies, we have either homonymy (Marathi /palәk/ ‘guardian’, ‘spinach’) or polysemy (Marathi /toṇḍ/ ‘mouth; face’) according as the aberration is more or less serious.  These two are, however, never so numerous in a language as to frustrate the modest optimism of GREENBERG’S observation (1957, p. 35): “Just because you call a dog a ‘dog’, it does not mean that you have to call a cat a ‘cat’.  It is unlikely, however, that you will call it a ‘dog’.”


A less serious threat to the efficiency of a language is the association of more than one signifiants with the same signifié.  Corresponding to polysemy, there is polyonymy illustrated by suppletion and other cases of anomalous non-contrastive alternation where we pretend that we do not have more than one signifiant.  When we run out of reasons to justify this claim, we admit that we have a set of synonyms on our hands.  In the case of synonymy, however, the principle of necessity finally reasserts itself to the extent that the synonymy is imperfect, which it almost always is.  Indeed if a language does come up with perfect synonyms, historical change usually saves the situation by a process of semantic differentiation.6   


3. The duality of patterning.  This is the well-known “double articulation linguistique” (MARTINET, 1949).  At this point we have to give up the simple antithesis of signifiant and signifié and have three terms—the phonetic form plays the signifiant to the grammatical form which in turn plays a similar role to the ultimate signifié.  The utterance is organized and interpretable at two levels—as a phonologic sequence and as a grammatical sequence.  The syntagmatic alignments or “cuts” and the paradigmatic alignments of “sames” and “differents” will differ at the two levels.  The minimum or simplex meaningful forms are independent of each other as to their overt shape.7 If they are not so independent, ether the so-called meaningful minimums are complex or the so-called meaningless minimums (phonemes or components) are not in fact meaningless. The very possibility of setting up two kinds of units will have to be denied.


The principle of dual patterning can be indirectly illuminated by giving an example of a symbolism that is not so patterned— the written mathematical notation.  Here we have to learn the overt shape for each meaningful minimum—“one”, “plus”, “equals”, etc.—independently.  There is no way of reinterpreting 1, +, =, etc., as patternings of units of another ‘lower’ level.


3.1. Marginal exceptions.  The principle of duality may be threatened in either of two ways.  First, there may be an incipient conflation of the two levels into one.  TOGEBY (1951, p. 30) defines ‘morphophonemes’ as solidary or inseparable combinations of a unit of expression and a unit of content—“the phoneme of modulation is always accompanied by a particular morpheme of modulation and is unable to express any other, and the morpheme of modulation likewise cannot be expressed in any other way” (translation mine) 8. Although it may be seriously doubted whether things are so simple with French intonations, which TOGEBY is describing at this point, it is easy to see how this kind of situation can come about with junctures, contours, and the like as opposed to vocoids, non-vocoids, coarticulations, tones, and the like.  Some such consideration seems to have motivated JAKOBSON to accord a special status to such features—they are either configurative or expressive features but not the ordinary distinctive features characterizing vocoids and the like (JAKOBSON and HALLE, 1956, § 2.3).  I do not see why the three have to be mutually exclusive classes of features: I see no difficulty in envisaging the same given features as entering into more than one function—the details varying from language to language.


A more convincing case for a direct hook-up, so to say, between the significant and the signifié can be made out for tactical features as opposed to quotable (or at least isolable) morpheme-shapes.  BLOOMFIELD (1933, pp. 166, 264) tried to establish a distinction between the meaningless taxeme and the meaningful tagmeme to match the phoneme and the morpheme respectively.  It is certainly no accident that this attempt has hitherto remained a mere curiosity in the history of linguistics.  More to our purpose will be his definition of irregularity (1933, p. 274): “any form which a speaker can utter only after he has heard it from other speakers is irregular.”  To the extent that tactical patterns or analogies prevail over irregularities or anomalies, the SEMIOSIS is DIRECT rather than MEDIATED.  To the extent that irregularities detract from the signaling reliability of a tactical pattern, duality breaks in and the grammar stands in need of being rounded off by a lexicon of such anomalies.


Secondly, the duality principle may be called into question because of incipient fissions of the two levels into more than two.  Let us look at the grammatical level first.  It has become amply clear by now that BLOOMFIELD’S definition of the morpheme as the minimum meaningful unit proves to be unworkable in practice.  Empty morphs (for instance, some stem-formatives) do not commute with anything, even with their own absence; some morph resemblances (MARTIN, 1952, Ch. 16) (like crash, crush, dash; snide, sneer, side; or see, sight in English, or  /lomb-, lomkә-, okhәmb-, oәmba, hindkә-/, etc., in Marathi9 do not quite add up to morphemes but have an undoubted nuisance value; the meaningfulness of derivative suffixes is often questionable, yet calling them morphemes seems to make sense; unique constituents (like cran- in cran-berry) render their partners redundant—all such ‘aberrant’ cases arouse precisely those ‘suspicions’ that BAZELL takes care to allay (see fn. 7 above) in speaking of typical morphemes.  So if we try to find a way out of this difficulty by suitably redefining ‘morpheme’ and renaming BLOOMFIELD’S meaningful minimum, say as ‘idiom’ (cf. HOCKETT, 1956, 1958), we are face to face with three levels—the phonologic, the morphemic or grammatical, and the semantic.10


At the phonologic end of the semiotic chain, we have JAKOBSON’S treatment (JAKOBSON and HALLE, 1956 and elsewhere) of distinctive features. If we are justified in regarding components (as he seems to) as distinctive primarily of phonemes and only secondarily of utterances (I think the reverse is true), then we have a splitting up of the phonologic level into two—the componential and the phonemic.11


The multiplication of levels entails a multiplication of THRESHOLDS OF LAWFULNESS OR WELL-FORMEDNESS in language.  The phonologically lawful (‘pronounceable’), the grammatically lawful (‘correct’, ‘grammatical’), and the semantically lawful (‘makes sense’, ‘usable’) would then seem to be successively less inclusive classes.12


So long as we have more than one level (whether two or three or more), we can say that the sign-system has mediated semiosis.


4. Interrelations of the three principles.13 Arbitrariness and mediated semiosis are logically independent—that is, one can be present or absent in the presence or absence of the other.  Arbitrary but one-level sign-systems are common enough: traffic lights and written mathematical symbolism (see § 3 above) are good examples.  Non-arbitrary sign-systems with more than one level are also found.  Stick figures are iconic likely ordinary line-drawings, but unlike the latter they are simple combinations of a limited number of isolates.  This two-level patterning is what makes them easily reproducible.  Another example would be a medical system in which all disorders are diagnosed in terms of combinations of a few basic symptoms, so that symptoms reveal disorders, which in turn ‘indicate’ certain remedies.  Here what renders the system non-arbitrary is not iconism but ‘natural’ cause-and –effect relations.


Cross-linguistic arbitrariness and intra-linguistic necessity are mutually independent and not incompatible (as BENVENISTE, 1939, seems to have wrongly thought).


Necessity and mediation are independent too.  If the two levels—the ‘lower’ (phonologic, expression, cinematic) and the ‘higher’ (grammatic, content, plerematic)—are compared as to the operation of necessity (the avoidance of many-to-one or one-to-many semiotic relationships), the lower one on the whole seems to score better.  It is interesting to note, in this connection, that in much historical reasoning, especially in internal reconstruction, there seems to be a tacit assumption that earlier stages tend to have a ‘cleaner’ allophonics and morphophonemics.  Historical changes muddy the water, so to speak, and we are left with highly dissimilar allophones, intersection of phonemes, suppletion, homonymy, and similar other complications and aberrations on our hands.  I think that this assumption needs to be looked into; it is not probably as naive as it sounds.


The joint consequence of arbitrariness and duality is that every morpheme shape is an irregularity (cf. BLOOMFIELD, 1933, p. 274)—the speaker cannot make it up for himself without introducing a linguistic change, that is, without modifying the threshold of lawfulness.  On the other hand, ideally there is no room for irregularity at supra-morphemic size-levels, no place for supra-morphemic units as entries in a lexicon.  All complex forms should be at the dispensation of the tactical code (the principle of necessity) without falling back upon the lexical listing (the principle of arbitrariness).  The linguist’s uneasiness is greater in accepting tactical homonymy and tactical perfect synonymy than in accepting homonymous morphemes and morpheme alternants—and his instinct is well-founded.14  


A word of clarification about the relation between the ‘lower-level’ units, say, phonemes, and the ‘higher-level’ units, morphemes, is perhaps in order at this point.  Nothing in what we have said so far implies that the morpheme—or more accurately, the morph, its correlate at the ‘lower’ level—has to be ‘longer’ than the phoneme (or at least as long).  Typically, however, morphs are longer than unit phonemes in natural languages and this has led some to the mistaken idea that morphemes (sic!) are in some sense ‘composed of’ phonemes.  By itself a morpheme is a Euclidean point—it has a (linear) position but no magnitude. The distinction between the morpheme and the morph is fundamental.  Again, nothing in what we have said so far implies that the ‘lower-level’ minimal units are few in number, while the ‘higher-level’ minimal units are rather numerous.  It so happens, however, that this is the case with natural languages and plays an important part in their economy.15


5. A fourth property? The reader who has patiently followed us so far must no doubt have been impressed more by the marginal exceptions and deviations than by the principles themselves.  It is a part of the linguistic field-worker’s training to learn to reconcile oneself with the discrepancy between what ‘ought’ or ‘ought not’ to occur and what does actually occur and not impatiently conclude that the information is either stupid or obstinate.  The disappointment of logicians and logical semanticians with natural languages is well-known, though some of them are more graceful about it and speak of the open texture of natural languages.


That indeed seems to be the wiser part.  Rather than blame these inconvenient exceptions and discrepancies entirely on the imperfections of our formulations or the incompleteness of the corpus, we may perhaps do well to dignify them as consequences of a single principle, namely, that all grammars leak, that language is not merely changeful, shifty, subject to dialect variation, but also fundamentally crude, makeshift, messy.16 (Thus crudity, let me repeat, is more than either linguistic change or dialect variation.  These two do not render the postulation of an ideally well-behaved language impossible but merely pose a challenge to our ingenuity.17)


By so recognizing crudity as a fundamental property, we are to that extent rescuing linguistic analysis from the hazards of the linguist’s temperament.  He may have his passion for or horror of tidiness—but he has to accept crudity: his business is to find out just how crude a language is, 18 and how best he can cope with it.


6. Methodological implications.  There is an unfortunate tendency to divorce the discussion of the fundamental properties of language (SAUSSURE’S language) and the discussion of the traits of specific languages (SAUSSURE’S langue).  As a consequence both suffer.  The former gains a reputation (often justifiably so) of being armchair philosophy, debates in vacuo, or just a textual criticism of SAUSSURE.            The latter, aiming to concern itself with the practical business of describing a language (practical, indeed!), degenerates into myopic shoptalk.


What is exactly the status of these predications about language? Are they descriptions of general properties? Assumptions underlying our methods of investigation (discovery), presentation, and validation (evaluation)?  Elements in the definition of language?  Assertions of language universals, inseparable accidents? We cannot even begin to answer these questions here.  We shall have to wait, presumably, till we get around to a comprehensive typology of sign-systems.19


The first three principles (and possibly a few others) are the unspoken assumptions underlying many of the analyst’s intuitions about some of the familiar cruxes of descriptive analysis.  At the minimum, the principles will help us realize why suppletion, homonymy, zero, intersection of phonemes, idioms, intonations, and the like make us uneasy or cautious in the first place, why they are cruxes at all.


The most interesting methodological problems, however, are those raised by the crudity principle—especially as they concern the presenting of the results of our investigation for the inspection of others.  Broadly, we may say, if language is messy, fuzzy-edged, crude, let our description of it be answeringly messy, precisely vague, and not inaccurately precise.  CHOMSKY’S suggestion (1957, Ch. 2, fn. 2; Ch. 5, fn. 2) that we recognize degrees of grammaticalness (read: lawfulness at the grammatical level) is thus a step in the right direction (cf. Also HARWOOD, 1955).  The counsel of despair which makes a fetish of attestation is certainly not the wiser way of facing crudity—a form may be lawful and yet remain unattested and vice versa.  (In the latter event, the form may be explained as a slip of the tongue, imperfect mastery of the code, playful flouting of the code, a citation in a metalinguistic discourse, and the like).


Another counsel of despair is that co-existent sub-systems (read: rival or competing thresholds of lawfulness) be recognized at the drop of a hat.20 This device, if not handled too carefully, is apt to run counter to the principle of intra-linguistic necessity.  For, to the extent that one sub-system sanctions a form rejected by a rival sub-system, we are beginning to have linguistic reasons why the given signifié should be correlated with a different significant and vice versa.


7. Conclusion.  To conclude, the necessity and mediation principles tell us why we should not despair of describing languages; the crudity principle (or the crudity caveat, if you like) tells why we should not be too confident (as American linguists generally are, when not at their best) or too ambitious (as the glossematicians, who end up by not writing any grammars at all); the mediation and crudity principles tell us why we human beings do not despair of being able to talk about this multifarious, bewildering, contingent world of ours.








BAZELL, Charles E., 1949.  On the problem of the morpheme.  Archivum linguisticum

1: 1-15.


BENVENISTE, Emile, 1939.  Nature du signe linguistique.  Acta ling. 1: 1. 23-9.


BLOCH, Bernard, 1948.  A set of Postulates for phonemic analysis.  Lang. 24. 3-46.


BLOOMFIELD, Leonard, 1933.  Language. New York: Holt.


BOLINGER, Dwight L., 1948. On defining the morpheme.  Word 4. 18-23.


------, 1950.  Rime, assonance and morpheme analysis.  Word 6. 117-36.


CHOMSKY, Noam, 1957.  Syntactic structures.  The Hague: Mouton.


FIRTH, J. R., 1935.  The technique of semantics. Trans. Of Philol. Soc. 1935, 36-72. 

(Reprinted in Firth, 1957 a.)  


------, 1957 a. Papers in linguistics 1934-1951. London: Oxford University Press.


------, 1957 b. Synopsis of linguistic theory.  Contributed to: Studies in linguistic analysis. 

Oxford: Blackwell.  


FRIES, C. C., and PIKE, K. L.. 1949. Coexistent phonemic systems.  Lang. 25. 29-50.


GARVIN, Paul L., 1957. On the relative tractability of morphological data.  Word 13. 12-



GREENBERG, J. H., 1957. Essays in linguistics. Chicago: University of Chicago Press.


HARWOOD, F. W., 1955. Axiomatic syntax: the construction and evaluation of a

syntactic calculus.  Lang. 31. 409-13.


HOCKETT, C. F., 1956.  Idiom formation.  In: For Roman Jakobson.  The Hague:    

            Mouton, 222-9.


HOCKETT, C. F., 1958.  A course in modern linguistics. New York: Macmillan.


JAKOBSON, Roman, 1936.  Beitrag zur allgemeinen Kausslehre: Gesamtbedeutung der 

            russischen Kasus.  TCLP 6. 240-88.


------, 1953.  Participant in: Lévi-Strauss, C., et al. (edd.), Results of the Conference of

anthropologists and linguists, Ind. Univ. Publ. in Anthr.  And Ling., Memoir 8, see Index.


------, and Halle, Morris, 1956.  Fundamentals of language.  The Hague: Mouton.


JOOS, Martin, (ed.), 1957.  Readings in linguistics……Washington: ACLS.


MARTIN, Samuel E., 1952.  Morphophonemics of standard colloquial Japanese.  Lang.

Diss. 47.


MARTINET, André, 1949. La double articulation linguistique.  In: Reherches

structurales (= TCLC 5), 30-8.


Patañjali.  Mahābhāya (‘The Great Commentary’ on Pānini, ed. by  KIELHORN). [My

translation is based on the English tr. in P.S. Subrahmanya SASTRI, Lectures, vol. 1, 1960 and the Marathi tr. by MM. Vasudevshastri ABHYANKAR, Vol. 1, 1938.]


SAPIR, Edward, 1921.  Language.  New York: Harcourt, Brace.


SAUSSURE, Ferdinand de, 1916.  Cours de linguistique gnrale, 1st ed. Paris: Payot.


SCHRECKER, Paul, 1948.  Work and history: An Essay on the structure of civilization. 

Princeton: Princeton University Press.


TOGEBY, Knud, 1951.  Structure immanente de la langue franiçase (= TCLC 6). 

Copenhagen: Nordisk Sprog- go Kulturforlag.


WELLS, Rulon S., 3d, 1947.  De SAUSSURE’S system of linguistics.  Word 3. 1-31.  (Reprinted in Joos (ed.), 1957, pp. 1-31.)




            This was published in Indian Linguistics 25: 83-92, 1964 (published 1965).












*I wish to thank here Professors C.F. Hockett, F.W. Householder, jR., Sol Saporta, Bernard Bloch, Rulon S. Wells, 3d, and M.A. Mehendale who read earlier versions of this paper and offered useful comments. Bibliographical details of the references are given at the end.

1.  See further fn. 13 and § 4 below

2.  That is to say, no earthly reason. (We are not concerned here with the possibility of divine reasons of the     sort given in religious myths about the creation of language).

3.  The ancient Indian grammarians of Sanskrit thought fit to announce this refusal much earlier – when the word (pada) was reached.  Patañjali bluntly tell us (Mahābhāya 1. 1.3 apud Pānini 1.1.1, vārtt. 7, Kielhorn’s ed., vol. 1, p.39 ii. 16-19): “The restrictions of usage (proyoganiyama) are not taken up here.  What then?  Words are formed and formed endlessly here.  They are put together in any way thought proper-as in, āhara pātram [fetch (imper.) vessel accus. Sg.)] or pātram āhara.”

4.  While Saussure was the first to formulate clearly the principle of arbitrariness, Benveniste (1939) is I think, the first to counterpose the principle of necessity.  A failure to see the compatibility between the two is probably the source of the ancient Western dispute between thései (convention, arbitrariness) and phúsei (convenance, necessity).

5.  Compare BAZELL’S observation (1949, p. 9) that morphemes are typically unambiguous signs—“signs which are neither homonymous nor synonymous with others.”  Cf. also BLOOMFIELD (1933, 9. 145): “each linguistic form has a constant and specific meaning……. [The presence of synonyms and homonyms] shows, of course that our basic assumption is true only within limits, even though its general truth is presupposed not only in linguistic study, but by all our actual use of language.” (Italics mine).

6.  Similar remedial processes counteracting homonymy have also been noted.  The plethora of perfect or near-perfect synonyms and the presence of some overblown polysemies in classical Sanskrit are obviously tied up with artificial conventions—classical Sanskrit is not quite a ‘natural language’ enjoying the benefits of childhood transmission and the wear and tear of everyday living.  So far as the relation between historical linguistics and the principle of necessity is concerned, it is interesting to observe that, while people have felt the necessity of finding the causes of phonetic change and semantic change producing polysemy, the presence of analogical change and differentiation of synonyms is taken for granted.

            7.  Compare the first half of the GREENBERG quotation is § 2.1 above.  Cf. also BAZELL (1949, p. 9), who points out that morphemes are typically “not suspect either of being divisible into signs or being only parts of signs.”  BOLINGER (1948, 1950) is full of precisely such suspicions.

8.  This use of the term morphophonéme has of course nothing to do with the

usual American use of morphophoneme.

            9.  Meaning respectively ‘hang (intr.) as a part of’, ‘hang (intr,) by attachment’, ‘(of person or limb) swing and exert pull by weight’, ‘plumb-level’, ‘(of liquid in a container) be disturbed by the irregular movements of the container’.

            10.  Translating this revision into HJELMSLEV’S terms will yield the following scheme:

                        expression-substance (phonetics)

                        expression-form (phonology)

                        content-form1 (grammar)

                        content-form2 (semology)

                        content-substance (semantics)

            Presumably, content-form1 (grammar) will be Janus-faced and phonology and semology will be substances to grammar but forms to phonetics and semantics.

            11.  I think, however, that components constitute an alphabet for writing phonologic forms and not merely phonemes.  In that case two allophones may differ componentially on occasion and so intersection of phonemes may be tolerated.  The difference between the component and the phoneme will be merely one of size-level like that between the morpheme and the word.  Size-levels are in some sense less fundamental than levels of semiotic mediation.  The recent devaluation of the morphology-syntax division is justified.  Unfortunately the term level has become a vogue word and been subjected to careless use.

            12.  So that colourless green ideas sleep furiously as also sincerity admires John (CHOMSKY, 1957, § 2.3 and fn. 7 to § 5.4) will cross the grammatical threshold but fail at the semantic threshold.  This tallies well with the often-heard informant response : “Sure, you could say that, there is nothing against it, but we just don’t, it doesn’t mean anything.”    

            The term canonical could now be defined as morphophonemically lawful or well-formed-provided we keep in mind the essentially secondary status of morphophonemics as compared to the three primary levels.

            The term well-formed is borrowed from modern logic.  For the term lawfulness  I am indebted to Paul SCHRECKER’S Work and history (1948, see Index).  Linguistic lawfulness has closer affinities to lawfulness in the everyday sense than to conformity to ‘natural law’—something we have lost sight of in decrying prescriptive grammar.  Morris SWADESH’S glottochronologic hypothesis, if substantiated, would be a true example of a natural law and be applicable to parole and not langue.

            13.  The ancient dispute between anamolists and analogists rests on an incomplete analysis.  The three principles present three antitheses:

(i)                 arbitrary : motivated (whether iconically or otherwise)

(ii)                necessary : asymmetrical

(iii)              mediated semiosis : immediate semiosis

It is true that there is some degree of polarization here between ‘arbitrary’, ‘asymmetrical’, and ‘mediated’ on the one hand (all defeat analogy) and ‘motivated’, ‘necessary’, and ‘immediate’ on the other (all promote analogy).  But this cannot be simplified into a single antithesis.  We have to find our way to a more adequate formulation.  For instance, the relation between the third antithesis and analogy is probably to be formulated as follows: mediated semiosis arises in order to mitigate anomaly, while perfect analogy actually renders any further mediation unnecessary.

            14.  But the uneasiness should not be tantamount to refusal.  JAKOBSON’S attempts (1936) to force all the disparate uses of a case into a single formula (the Gesamtbedeutung) seem to stem from a mistaken belief that we should keep grammar at least (tactics and marker morphemes) clean of asymmetrical hook-ups between the significant and the signifié.  That is the privilege of the speakers of the language and not of the linguistic analyst.

15.  Two-level sign-systems that depart from this picture are not, however, difficult to find.  In Morse code, for instance, both kinds of units are limited in number.

16.    Cf. SAPIR, 1921, Ch. 2, p. 39, Firth, 1935, pp. 70-1, footnote.

17. Consider in this connection BLOCH’S definition of the ‘idiolect’: ‘The totality of the possible uttearances of one speaker at one time in using a language to interact with one other speaker is an idiolect” (1948, § 1.7).  SAUSSURE’S uneasiness in reconciling diachrony and synchrony is also understandable (cf. WELLS’ 1947, §§ 32-4).  The fact is that there is a sort of built in ethnocentrism in language that regards what is spoken here and now as alone right.

18.  Whether languages differ in the degree of crudity is a moot question, Cf. GARVIN, 1957.

19.   Cf. HOCKETT, 1958, § 64.3

20.  Cf. FIRTH (1957a, passim; 1957b) on ‘polysystematic’ analysis; FRIES and PIKE, 1949; JACKOBSON on ‘code-switching’ as a routine fact of language (1953, pp. 16-7, 36 ff.).