Language and Linguistics
Correlative Linguistics
Prefaced by a framework of linguistics as a whole




William Haas has written:


            Every science may be said to have its origin in some radical complexity: in a new sense of wonder, about something always taken as obvious – wonder which asks to be transmitted into sense of understanding. Amid the sophisticated complications of contemporary linguistics, it is still vitally important to remain aware of the simple radical problems of the discipline (1960: 121 – 122).


            Haas recognizes two main division of linguistics, each with its radical question serving as a point of departure: linguistic analysis and linguistic comparison. Slightly modifying and adapting Hass’s formulation of the question, I shall further subdivide the latter (linguistic comparison) into two: historical and correlative. So we have.




How do we succeed in understanding one another’s speech?

How do we manage to say AND to grasp an endless succession of new utterances with the help of a limited stock of resources?


            We choose our way through a maze that proceeds from the more general to the more specific patterns. These patterns (which are indeterminate but presumably finite in number – collectively referred to as a system of rules) are what stand between elementary items (which are finite in number – collectively referred to as an inventory) and usable texts (which are denumerably infinite in number – collectively referred to as a corpus). We begin by matching texts, items, and sets of matched sets within the language being analyzed. We examine how a text is reproduced (i.e. rerendered or reexpressed) in the same language system. We examine how a text is used in relation to what it symbolizes and the situations into which it fits.




Why do we fail when we do? How does one make sense of this irrational babel of languages (i.e. sets of matched systems, inventories, corpora)?

The answer can be sought either by finding out how languages come to be what they are, or by finding out whether there is some old bags of tricks that each language draws upon. In either case we begin by comparing languages – texts, items, and rules.




            How does a language reproduce itself from one population to another population of users? What are the patterns of stability, innovation, and diffusion  (collectively called primary or linear phylogeny); and of maintenance over a line of descent divergent descent into a family, and convergent influence over a zone (collectively called secondary or dendroidal phylogeny ) that follow from linear phylogeny; and of contact, acceptance, rejection, maintenance, and displacement of language systems within a language network. (collectively called tertiary or reticular phylogeny) that link up linear and dendroidal phylogeny to the population of language users?


            We begin by seeking out diatopic and diachronic correspondences between texts and then between items, and between systems of rules and, also, by carrying out reconstructions on the basis of such correspondences.




            How does a text reproduce itself from one language to another? What are the recurring traits that characterize natural languages as such and natural languages types? What are the recurring patterns of linear, dendroidal, and reticular phylogeny?


            We begin by seeking out trait correlations – within and across languages, among language traits, and between language traits and traits of users and situations.


            It will be seen that the crucial differences between (1) and (2) and between (2a) and (2b) lie in the different kinds of collations that constitute the methodological starting point in each case – pattern matchings of analysis, correspondences of history, and correlations of universal and typical features.


            Before the differences between the three kinds of collations are explicated, it will be useful to offer a few more definitions. The instantiation of a language or a transition dialect in an individual user is an idiolect. The history of an idiolect is linguistic ontogeny, which is thus distinct from linguistic phylogeny. The instantiation in an individual of a language network or of an intersection of language networks is the linguistic repertory of that individual. Patterns of exposure, acquisition, maintenance, overall shift, and loss of items and rules within a idiolect (a language) and patterns of contact, acceptance, rejection, and displacement of whole languages (whole idiolects) within a network (repertory) take us beyond linguistics proper into psychology, social psychology, ethnology, sociology and cultural history of language.1


            The instantiation of a language in a given situation is language use. Language use has three modes, namely, production, reception, and reproduction. Reproduction has a little of both reception and production in it, and may be within the framework of a single language (the original and the reproduced texts are both from the same language) or across languages or stages or varieties of that same language. Reproduction – whether intralinguistic or translinguistic – may aim primarily aim at recapturing the reproducing user’s reception of the original (e.g. translation as a form of glossing) or at recapturing the production of the original (e.g. ready equivalents in a traveller’s phrase book, translation of a poem as recreation). Translinguistic reproduction may be either transrendition (e.g. of English [1phoust] by Marathi [1poə], or translation (e.g. of English I have two sons by Hindi mere do lke hƐ n). 2 The distinction between the two matches the distinction between rendition and formulation as a aspects of production, between recognition and comprehension as aspects of reception, and between rerendition and reexpression as types of intralinguistic reproduction. Reproduction has an important bearing on linguistic ontogeny and phylogeny.


            Between texts, items, or rules, there may be historical relationship – one may be a successor of the other, or both may be cosuccessors of some third thing. Historical relationship may be based on the descent of the descendent language system from the ancestral language system arising out of childhood transmission from one generation to the next. Alternatively, they may be based on influence arising out of contact between two languages, which may range from bare contact, “intimate” bilingualism. Influence involves a three cornered relationship between the model, the filter, and the replica, which is a successor to both model and the filter. A replica text is reproduction of the model from which it deviates because of some filter. If the replica is being offered as a text of the filter language, we speak of the influence as borrowing from donor model language to the recipient (filter –replica) language. Borrowing on a large scale brings a new descendent of the recipient language into being. If the replica is being offered as a text in the model language, we speak of the influence as mutation or interference in the model through the mutator- filter language, the replica being assigned to a mutant of the model. When a language L1 is consistently a mutator of L2 into a mutant L21, then L2 is the mutated ancestor of the mutated descendent L21. Thus, English as used in India by native speakers from England is a descendent of English acculturated to India – let us call it Indianized English. It borrows from Indian languages by transrendition (e.g. bidi, ahimsa) or by translation (e. g. leaf cigarette, non violence). But English, as used in India by Indians whether as native or as foreign speakers, is a mutated descendent of English, the various Indian languages being its mutators – let us call it Indian English, or more specifically, Hindi, English, Tamil English, All-India English etc. It is characterized by transrendition from English (e.g. resulting in homonymy between state and estate in Hindi English or between eights and Yeats in Tamil English), translation from Indian languages (e. g. communal riot, had gone yesterday), transrendition from Indian languages (e. g. jira for cumin seed), and of course plain mis-expression (e.g. really speaking for speaking truthfully, feel homely for feel at home).                    


            Shared line of descent yields a language chain; codescendent relationship yields a language family; shared influence yields a language zone. The first great task of linguistic prehistory is the reconstruction of earlier states of language. Reconstruction of texts, items, and rules may proceed from a descendent towards an ancestor (internal), from codescendents towards an ancestor (comparative), or from a more remote ancestor and codescendents to a less remote ancestor (reverse). 3 Reconstruction of historical relationships of descent and influence between language systems is the second great task of linguistic prehistory.


            Returning to the three kinds of collations, we may begin by observing that the distinction between correspondences and correlations is reminiscent of the biologist’s distinction between the homologies of comparative anatomy (e.g. resemblance between the human hand and the mammalian foreleg pointing to common origin) and analogies of comparative ecology (e.g. resemblance between the hand and the elephantine trunk pointing to common function). Again, pattern matchings and correlations may be distinguished in terms of the three modes of language use. In pattern matching we are observing the exercise skills of native production, reception and intralinguistic reproduction. In correlating of traits we are concerned with translinguistic reproduction. The items brought together in historical correspondences may or may not be translinguistic reproductions. Sanskrit cakra is hardly a transrendition of English wheel (its historical cognate). While Marathi sətkar ‘act of honoring’ would be an acceptable transendition of Bangla šɔtkara funeral rite’ (a shared loan from Sanskrit satkāra), one could hardly be translation of the other.


            The basic data of historical linguistics are not so much observation of the exercise of the skills of native production, reception, or reproduction, or of translinguistic reproduction, nor, again, the description of languages or languages stages. Rather, they are the observation of successor relationships and of derived cosuccessor relationships between texts, items, and rules4. If historical linguistics is comparative in the sense in which comparative anatomy or comparative physiology or comparative ethology (comparative psychology) is comparative, correlative linguistics is comparative in the sense in which comparative ecology or comparative religion is comparative. For the kind of ambiguity (or is it richness) that the phrase ”comparative linguistics” is thus seen to enjoy, we shall have to find a match, probably in the phrase “comparative literature.”5


            The reason behind this twofold nature of linguistic comparison is, of course, the peculiar way in which languages instantiations (idiolects) within a language community are like members of a biological species within a population (Stevick 1963). Like two species (which are not interfertile – cats and dogs do not interbreed: horses and donkeys do, but the mules do not reproduce themselves) and the unlike two nonlinguistic institution, two languages do not yield a mixed idiolect with two ancestors. A child in a bilingual environment ends up by acquiring two languages and not by acquiring a mixed language, though some of his initial efforts at production look suspiciously like one. Granting that pidginization has been a much more common phylogenetic process than linguists have been disposed to grant (Southworth 1971a, 1971b), 6 a pidgin is unmistakably the mutated descendent of one language. Marathi (Southworth’s example) remains an Indo-Aryan language despite the Dravidian grafting, as much as Finnegan’s wake remains English despite Joyce’s drastic distortions and conflations within and across the boundaries of English. (The possibility of historical comparison does not entail the possibility of comparative reconstruction, which calls for postulations of certain other properties as well. See below, section 2.3, hypotheses [15] and [23].)


1.1    Precursors


              Correlative linguistics has been talked about previously under other names. Some of the terminological anticipations include Hockett’s “contrastive linguistic” coordinate with synchronic and diachronic linguistics (1948); Trager’s ‘contrastive linguistics” inclusive of historical linguistics and coordinate with descriptive linguistics (1952: 6-7 ); Greenberg’s “general linguistics” coordinate with descriptive and historical linguistics (1957a: 86); Halliday’s “comparative descriptive linguistics” inclusive of the theory of translation and the theory of transfer comparison by the side of comparative historical, descriptive, and institutional linguistics as division of general linguistics (Halliday, McIntosh, and Strevens 1964; 15-16, 111-112, 120); Ellis’s “comparative linguistics” (especially its “all-purpose” version) coordinate with descriptive linguistic (1966); and Hymes “syncretic linguistic” coordinate with synchronic, diachronic, and diatopic linguistics (1968:361),


            Now has correlative linguistics so understood been actually practiced so far? It certainly has been. As shall be seen below, such traditional concerns as universal grammar (or its later avatar – language universals), historical universals (e.g. the various hypotheses about “progress” in language), structural or non-historical classification of languages (now being rehabilitated as language typology) and of writing systems (logo-graphic, syllabic, alphabetic), and such innovations as contrastive linguistics and Greenberg’s historical typology (1957b) certainly exemplify, though they do not exhaust, correlative linguistics. A good part of Trevor Hill’s (1958) institutional linguistics, or of translation theory, or geographical and social covariation and dialect studies (cf. Footnote 1), or of theoretical and methodological discussions of archiving and surveying of languages (Kelkar 1969b) exemplify or implify correlative methods. It is about time,  I feel,  that we take stock of the situation and propose at least a tentative but comprehensive frame work for reinterpreting past work, initiating future work, and ultimately stimulating the search for a more adequate and more rigorous frame work. We shall call this frame work correlative linguistics and recognize it as a sub-frame work with in the larger frame work of linguistics sketched above (Section1).


            The presentation that follows is necessarily sketchy and abstract. Too often, perhaps, I have counted on the reader to flesh it out with his own examples and to supply his own footnote documentation.




What is a language trait? It is any fact about a language as such (other than correlation with a user trait or a situational trait) that characterizes it as a semiotic system. Thus, the following are language traits:


a.                   having a retroflex flap, having a distinction affirmative/ negative, having a definite article, having an animate/ inanimate distinction (e.g. who?/ what?, who/ which, man’s/* table’s [in English]), etc;


b.         having a rule or a pattern: having a prohibition on final consonants, having an abundance of heavy nonmedial consonant sequences having a penultimate accent, having a passive construction, having S-V-O order in the surface structure of statements, having double negatives, having a rule in the kinship terminology that, if x is a kin-type K to a married male, x is also K to the latter’s wife, etc.;


c.         having a statistical expectation (e.g. expectation of null frequency, of nonnull frequency, of comparative frequency) about items in some inventory, rules in some system, or texts in some corpus: lacking a retroflex flap, having more polymorphemic words than monomorphemic words in the lexicon (or in a representative subcorpus of occurrent texts), using fusion morphs more often than additive morphs, tolerating homonymy in functors, tolerating homonymy in nonfunctors, using animate-to-inanimate shift less often than using inanimate – to-animate shift, lacking a writing system, etc.


It will be seen that the concept of trait is wide-ranging enough to accommodate the negative traits, quantified traits (e.g. having words with the average length of 2.1 morphemes), and the more sophisticated epistemic traits as proposed by Bazell (1958) (e.g. being more amenable to analytic model A than to analytic model B in phonology, grammar, etc.). But then the following are not, strictly speaking, language traits – rather they are disguised:


a.                   language-user traits : having more than a million users, having more monolingual native users than multilingual native users, having more nonnative users than native users, having no live speakers, having no native users, having no native female literate users, etc. (typically, we shall have to add, “at any given time”).


b.                  Situation-of-use traits:  having been used as a medium for schooling, is use exclusively at homes of native speakers, having been designated as a national language, not used for love letters, in use when close rapport between participants exists, in use when secrecy is desired, etc (typically, some of these do not correlate with whole languages but with specific items, rules, or patterns).


Let us call these extrinsic language traits, to distinguish them from language traits proper or intrinsic language traits. As we shall see, we need to speak of extrinsic language traits too.



2.1      Translinguistic Reproductions


Identifying an extrinsic trait or a purely phonetic intrinsic trait in a language would seem to be easy enough. Even if it is not easy, that would be a headache for demographers, anthropologists, experimental psychologists, or whoever, but not for linguists as such. How do we identify an intrinsic trait, other than a purely phonetic one, within a language? More importantly, how do we know that two languages possess the same trait? If one were to adopt an ad  hoc approach to the establishment of the categories of the linguistic analysis of a language, the identification of intrinsic traits across languages would seem to be an insuperable problem. How could one say that two languages share the phonological unit |r| or the grammatical distinction N/V or A/V or the semantic distinction visible/invisible, if one said that the use of the same symbol or the label in the relation to two languages is no more than a convenience? Indeed, such an impasse probably delayed the resumption of the concern for language universals and language typology in the “descriptive” era of linguistics. In these days linguists delighted in pointing out that the “adjectives” of one language are apt to be translated by the “nouns” (or the “verbs”) of another language, or that the unit /p/ participating in the commutations p:ph:b:bh in one language is not commutation p:b in another language, or that the case of /r/ realized as an apical trill, uvular trill, apical flap, or laminal obsulcate7 in various language is a hopeless one. Equally delighted (or exasperating, according to one’s inclination) was transrendition of both [ph] and [p] in English as [p] by a native user of Marathi, a language which has both sounds in contrast, or the lexicographer’s difficulties in offering  a translation gloss for English have in Hindi or for Hindi juţh a in English.


            A removal of this impasse involves a change in the model underlying analytic linguistics. Indeed one may claim that one’s correlative comparisons cannot be better than one’s linguistic analysis. I should broadly accept this claim, but immediately qualify it in some ways. To begin with, does this claim lead us into postponing any serious correlative comparison until after analyzing all the languages in accordance with some one model deemed to be acceptable, or, at least, until after finding some way of “translating” available descriptions with varied underlying models to some uniform model? Bazell (1958) and, following him, Lyons (1962)8 have already indicated a way out of this impasse, offering thereby to turn adversity into opportunity by proposing a new tool, which I earlier christened “epistemic trait’. A more radical solution, however, would be to accept language users’ translinguistic reproduction (transrenditions as well as translations) as the basic data of correlative linguistics rather than the descriptions of individual languages churned out by some favorite analytic model. Two languages will be deemed to have the same trait, not so much because the same item label or rule formula turns up in the analysis of them, but because texts exemplifying the item or the rule in question in each language are transrendered or translated by texts exemplifying the corresponding item or rule. This proposal would lead us to see that one’s linguistic analysis cannot be better than one’s correlative comparisions.


            A model of linguistic analysis is subject to check of data – oriented correlative comparison to prevent it from ignoring genuine relatively deeper resemblances: cases in point are the salutary effect of Jakobson’s proposal (1966) for a correlative inventory of phonological minima, or of Fillmore’s proposal (1968) for a correlative inventory of grammatical cases. Proposals to economize analytical statements by leaving unstated the appearance of universals or near universals in the language in question (e.g. by marking conventions) also stem from insights arising out of correlative comparison. In general, the search for formal universals is also the search for the foundations of linguistic analysis.

            A model of correlative comparison, on the other hand is subject to the check of the data – oriented linguistic analysis to prevent it from ignoring genuine, relatively deeper differences; cases in point are the salutary reminders that languages can differ profoundly in their handling of the structure and function of syllables or in their handling of word order, grapping, and what Halliday (1970: 43) calls cohesion features.


2.2        Initial and Consequent Collations of correlative linguistics.                        THE SELECTION OF THE DOMAIN. Theoretically, the domain of correlative comparison encompasses every human language – extinct, dead, or living, with perhaps a sideways glance at Esperanto or Rudolf – Carnapese. Taking on this whole domain or language population is obviously not feasible. The languages are not all accessible, let alone the data on all translinguistic reproduction possibilities. Even if one confines oneself to all those that are accessible and not undeciphered, the linguist’s attention span is limited, even if it were to be reinforced by computer memory


Fortunately, all these ambitious undertaking are unnecessary – at least immediately. Years of two-languages-at-a-time comparisons are necessary before a receptable group of testable hypothesis can be assembled, and before at least the major problems of collation and validation in this field are mastered. Refinements of sampling and quantification of correlation will also come to our aid, provided we see the point in exploring the whole gamut from prefect correlation (linguistic impossibilities and necessities) to near zero correlations (mere possibilities). Finally, there may even be some virtue in selecting a subdomain. Such a subdomain, as distinct from maximal domain, may be based on any of the following criteria or any combination of them:


a.                   Membership of a language family (or subfamily) based on shared descent or of a language chain (or subchain) based on shared line of descent.

b.         Membership of a language network (or subnetwork) bases on shared influence in a certain epoch.

c.          Sharing of a certain language-user trait or traits.

d.         Sharing of a certain situation-of-use trait or traits

e.          Belonging to a certain phase of culture history (e.g. feudalism, pre-agricultural or tribal societies enclaved within societies of a later phase, modern industrial societies) – it will be noticed that under (e), there is a merger of the criteria in (c) and (d).


            The overall picture that will emerge after correlative comparison within such a domain may or may not confirm significant generalization, e.g. Indo-European languages are suffixing, Slavic languages palatalizing; Arabic has been borrowing by translation rather than transrendition throughout its long history; Standard Average European (to borrow Whorf’s nomenclature [1941 and 1956b: 138]) favors S-V-O order, have like verbs, subject-predicate cleavage, “meaningless” proper names, and standardization of noncasual languages, and it does not favor clicks (consigned to paraphonology) or tones (confined to accented syllables, and that in very few languages).


2.2.2.      SELECTION OF THE SUBSYSTEM. Just as one can conveniently and profitably confine oneself to a subdomain, one may also confine oneself to a subsystem – say, phonology, graphonomy, syntax, kinship terminology – from which to select the traits for study, Selecting just one trait or a pair is the limiting case and the starting point of study by subsystem.


2.2.3.   SCHEMATA OF CORALATIVE COMPARISON. Given the domain or language population, an inventory of its members, and a repertory of traits likely to occur within domain, one can then establish, measure, and evaluate correlations of the following sort:


a.    Between a trait and a given member language (given the language L, L possesses or lacks the trait T1)

b.   Between a trait and membership of the domain (given the member languages of the domain D, L always or sometimes possesses or lacks the trait T1; thus, T1 may be universal, near-universal, type-yielding or negative universal in respect of D )9.

c.    Between one trait and another and membership of the domain (given the member languages of the domain D, if any language L possesses or lacks T1, then L always or sometimes possesses or lacks T2; thus, T1 and T2 may be compatible or incompatible, may from a syndrome of co occurring and possibly implicating traits, may form a spectrum of competing, possibly conflicting traits, and so on; [a spectrum of traits may be based on either of two considerations: the traits may be varying renditions or expressions associated with a more “abstract” item or they may be varying functions or interpretations of a relatively less “abstract” item]; if L1 possesses T1, L2 possesses a nonidentical T2, and T1 and T2  transrender or translate each other, then L1 and L2 are type-different in respect of the trait-couple T1:T2).

d.                  Between a sequence of a trait and successor trait, on the one hand, and any language sequence in the domain (given that L1 is an ancestor, mutated ancestor, donor, or mutator of L2 in respect of trait sequence within the domain D, if L1 possesses T1, then L2 always or sometimes possesses or lacks T2 as successor of T1; thus T1 à T2 is an ordered sequence of traits; if T1 = T2, then T à T2 is a stable sequence of traits; if T1 à T2 occurs and T2 àT1 does not, it is an irreversible sequence; if L1 àL2 represents stages in a domain made up of ontogenetic chains, then T1 à T2 is an ontogenetic sequence of traits, and so on).


e.            Between a dendroidal or reticular pattern (e.g. a family tree or a network of a certain shape) and a domain of phylogenetic chains (specific dendroidal or reticular phylogenetic patterns and specific linear phylogenetic patterns, as noted in (d) above, may be thought of as intrinsic traits of the domain as such, rather than of individual members)


f.            Between a trait and a extrinsic trait and membership of the domain (given the member languages of the domain D, if any language L possesses or lacks T1, then L always or sometimes possesses or lacks the extrinsic trait ET1; thus, an isogloss is a special subtype of correlations of type (f); the trait of an intrinsic trait may be thought of as an extrinsic trait).


g.                   Between a trait sequence and an extrinsic trait sequence and membership of the domain (given that L1 is an ancestor, mutated ancestor, donor, possesses T1, L2 possesses T2, and T1àT2, then it is always or sometimes or never the case that L1 possesses ET1, L2 possesses ET2, L2 possesses ET2, and ET1à ET2; the sequence may be phylogenetic or ontogenetic).


h.                   Between an intrinsic trait of a subdomain and an extrinsic trait of a domains of the domain D, if any subdomain SD possesses or lacks the domain trait DT1, then SD always or sometimes possesses or lacks the extrinsic domain trait EDT1).


Briefly, there are analytic (a, b. c. f,), ontogenetic (d, g,), and phylogenetic (d, e, g, h) correlations. The ones which involve extrinsic traits, over and above just the membership of a domain (f, g, h), may be called pragmatic correlations. Type (b) correlations may be called conditional analytic correlations. Illustrations of these types of correlation will be provided below (Section 2.3).


2.2.4        EVALUATION OF CORRELATIVE COLLATIONS. Each of these correlations has to be evaluated: this evaluation will not only take into account quantified measurement but will also call for qualitative weighting. Thus, a language with clicks in phonology will be deemed to be more “clicky”, i.e. to be better correlated with the trait of having clicks, than a language with clicks in paraphonology. Locating a trait in five closely related languages is certainly less impressive evidence of its widespread character than locating it in five historically unrelated cases. The membership of subdomain will be deemed to be better correlated with a trait if more “key” members of the subdomain have it than if fewer “key” members have it. Whether a member is a “key” member or not will, presumably, be determined on the basis of the possession of certain extrinsic traits. In determining the trait syndromes, the trait spectra, and the trait sequences with respect to a domain, some traits will probably be deemed to be “key” traits. In general if T1 is subsumable under T2 (e.g. having an alveolar click, and having a click), T2 is more important than T1.


Occasionally, however, even a highly particularized trait may assume a diagnostic value for some purpose – for example, the greasy (s/z) isogloss in American English. An isogloss is, to begin with, merely a correlation between (a) a trait and (b) the extrinsic trait of the language user’s residence or social position or situation of use and (c) membership of a domain of differentiated languages of the same family and network. It is expected that a good many of these isoglosses will turn out to be diatopic historical correspondences based on descent and influence. In other words, correlations are being subjected to evaluative criteria that are essentially historical. It is not difficult to extend the concept of the isogloss to diachronic correspondences and correlations. Thus, we can speak of fascicule of isoglosses marking two stages in the history of a language.


Finally, there are “correspondences” that serve as the basis of internal reconstruction, e.g. d/t nonfinal and t final in German; a/i and o/e as masculine/ feminine markers in different Marathi paradigms, to give a phonological and a grammatical example, respectively. A proposal to call these either correspondences or correlations of isoglosses is attractive enough. Internal reconstructions would then be assimilated to comparative reconstruction. It must be borne in mind, however, that this last move would require a major extension of our notion if linguistic comparison. We shall thereby be recognizing that there is a minor but important overlap between intralinguistic, analytic collations and interlinguistic, comparative collations. I think this is well worth the logical maneuvering called for. We shall sketch the outline of a suggested maneuver below (section 3.1)



2.3              A Sampling of Hypotheses


We now proceed to cite, without approval or disapproval and largely without comment, some examples of hypotheses involving correlations differing in the correlated terms, in conjectural strength, and in evaluative status. Hopefully, these will serve to indicate what correlative linguistics should look like. I must say that I have not always taken the trouble to recast the familiar formulation of the hypothesis into exact conformity to the schemata proposed above (section 2.2.3). We shall begin with some analytic examples (1-11), then offer an ontogenetic example (12), and conclude with linear (13-18,24-27) and dendroidal (19-23,28-34) phylogenetic examples. Some of these are pragmatic and thus involve extrinsic traits (11, 24-27, 28-34).


1.            Both English and French share a trait: my old friend and mon vieil ami are ambiguous in the same way. Both refer either to one who has been a friend for a long time or to a friend who has been alive for a long time,; an old soldier goes the same way but an old hat, an old man, the old wife, the older wife don’t. Marathi and Hindi lack this trait. It will be interesting to explore other Indo-European, Indian, and European languages.

2.                  A number of languages in Negro Africa use the same metaphor – a door is called a “mouth of the house” (Greenberg 1957a:70).

3.                  Some metaphor types predominantly go in one direction: body-part name for artifact or other inanimate object, physical for mental state, spatial for temporal.

4.                  High vowels have narrow phonetic ranges; the range of mid vowels is never narrower than that of high vowels.

5.                  The nasal systems form a trait spectrum: m, m/n, m/n/n, m/n/n/ñ, etc. such that the presence of ŋ implies the presence of m and n, the presence of n implies the presence of m, but not vice versa.

6.                  A voiceless obstruent is never followed by a voiced obstruent in close transition.

7.                  The presence of the number system implies the presence of singular; that of dual implies that of singular; if one member of the number system lacks a marker, it will be the singular number.

8.                  The following traits from a syndrome; S-V-O as statements; V-S-O and v-S-V-O (or S-V-s-O) as questions; and S1-V1-O1-and –S2-O2 as gapping transforms (where v and s stand for dummy verb and subject respectively).

9.                  The following traits form a syndrome: all syllable boundaries equally open as transitions; tone contrasts; and syllables, morphemes, and words invariably or predominantly coterminous.

10.              Every language has a  phonology; only some languages have writing systems.

11.              The “deeper” a trait is (in some determinate sense of “deeper” independent of “universal”) the more nearly it is likely to be. In other words, the deep traits constitute a universal syndrome.

12.              When the adult language has a vowel system of the type i/e/a/o/u, then the following is a common ontogenetic trait sequence: I ~ e/a/o ~ u à i/e/a/o/u.

13.              Tamil has miraculously escaped innovations for centuries.

14.              The following is an irreversible trait sequence: s/h à h, i.e. s merges with h never the reverse.

15.              The Neogrammarian hypothesis (underlying comparative phonological reconstruction): phonemes don’t split except by way of resegmentation, i.e. the trait sequence “lack of contrast à gain of contrast” is a negative universal. (This is at least true of direct or nonmutated descent.)

16.              The internal phonological reconstruction hypothesis: the following is a possible trait sequence;

Rule (a as a in Environment 1) with Rule (b as b in Environment 1) àRule (a as b in Environment 1) with Rule (b as b in Environment 1) but the following is not:

Rule (a as a in Environment 1) with Rule (b as a in Environment 1)à Rule (a as b in Environment 1) with Rule (b as b in Environment 1).

Note that both sequences are compatible with (15) above. A more generalized formulation, perhaps, would be: phonological alterations are not abridged or lost phonologically, but only through analogical leveling.

17.              Loss of contrast and loss of sounds in word-initial position is rare; the same in word-final position is rather common.

18.              The number of genders is always reduced, never increased. When it is decreased the masculine gender is never sacrificed.

19.              Languages “progress” from a low morphemes-per-word ratio to a high one: this was later replaced by a contrary hypothesis.

20.              If the following are borrowed at all, they are almost always borrowed to fill previous gaps in the systems, not to supersede previously existing items: numerals, kin terms, functors, terms for body parts and body functions.

21.              Some languages borrow freely through transrendition; others prefer to borrow through translation.

22.              Languages split but do not merge; when they split, they do so decorously into two languages at a time. In other words, a language cannot have more than one line of ancestry and more than two immediate descendents. (In wave hypothesis, the first part is questioned in so far as possibilities of the following sort are accepted: L1 is ancestral to L2 in respect of T1 but L3 is ancestral to L2 in respect of T2. In a modified family tree hypothesis, such a possibility will be accepted, provided that L1 and L3 are barely separated codescendants of L4. [Cf. South worth 1964 and the nation of transition dialects.]).

23.              The following is a recurring pattern in dendrodial phylogeny: a single branch proliferates, the other atrophy, e. g. Latin against Oscan, Umbrian, etc. in Italic; Proto-New-Indo-Aryan against various non-literary, spoken dialects of Middle Indo-Aryan; Bantu against other branches of Niger-Congo; Classical Arabic against other cognate languages.

24.              Grammatical irregularities resist analogical leveling in high frequency items, e.g. widespread suppletion in verbs meaning “go”.

25.              Lexical hypertrophy (snow in Eskimo, horse and date in Arabic, Kinship in Indo-Aryan and Dravidian) correlates with special attention to the referential domain in the non-linguistic culture.

26.              Only dead languages escape change. Languages with only non native users resist change with moderate success. Thus, the noncasual language of law, folksong, and written literature tends to be archaic and conservative. Traits associated with certain other situations of use tend to encourage innovations and their diffusion, which results in a high infant mortality in words and idioms. Examples are slang and occupational jargon.

27.              The lexicostatistic hypothesis of glottochronology: the percentage rate of replacements within the basic vocabulary over a given length of time is constant over long periods. The basic vocabulary consists of those meanings whose expressions resist borrowing and are near-universal categories.

28.              The age and area hypothesis of glottochronology: the diffusion of innovations proceeds at a relatively constant rate geographical area and dispersal over time). This is at least valid for pre-industrial societies.   

29.              The language-branching hypothesis of glottochronology: a language diverges into branches at a relatively constant rate (number of terminal branches and their dissimilarity from each other over time), e.g. the relatively low diversity in American English and Bantu indicate recent colonization.

30.              The following is a recurring pattern in language networks: one of the language systems gains nonnative users from among the native users of other language systems; this is the auxiliary dialect / language phenomenon.

31.              The following is a recurring pattern in language networks: a plethora of unrelated or distantly related languages in a small mountainous area, e. g. the Caucasus, northwestern California, the Nilgiris, the Hindukush Kabul River Valley-Gilgit Zone.

32.              Donorship in borrowing goes with high social or political status or with donorship in some field of non-linguistic culture (e. g. Italian musical terms in English); mutatorship in interference goes with low social or political status.

33.              A mutant language system has fewer contrasts than either the mutated ancestor or the mutator (filter) languages, e. g. Marathi English has fewer intonational patterns than either native English or native Marathi. (This is an example of a “poverty” trait.)

34.              Language contact without bilingualism but with acute need of communication results in pidginization. Creolization may or may not follow. If it does follow, the Creole either sheds its “poverty” traits through innovations or is displaced by its ancestral language or borrows heavily from it.





            By isolating universal, negative universal, and near-universal traits of language systems, language families, and language networks within the maximal domain, we arrive at a far more detailed characterization of natural language systems, as such, and their history. The validation of analytic and historical models is aided. (By relating such traits later to extrinsic traits, especially universal, negative universal, and near-universal extrinsic traits of natural languages, we open the way for comparing the so-called natural languages with other sign systems and cybernetic systems in man and nature, and for seeking explanations for them. This of course takes us outside linguistics proper. Examples of the relevant extrinsic traits are: every native user of every language plays all the roles – rendition, expression, recognition or scanning, comprehension, rerendition, and reexpression; language is useable with eyes closed, hands full, mouthful, and feet in locomotion; every language can be acquired without difficulty before the age of six by any child not deaf, dumb, an idiot, or left to the wolves.)

            By isolating spectra of type-yielding traits of language systems, language families, and language networks within the maximal domain, we come to see the full spectrum of possibilities. This can often serve as a needed cross-check on the excessively bold or excessively timid claims and hopes of analytic universals. Thus, while linguists may differ as to the number of “strata” or “components” in a language, they often seem to agree that the number is the same for all languages. In such matters it may turn out to be the case that what were considered rival analytic models are actually opposed language types.(This is, after all, the point of Bazell’s proposal of what I have called epistemic traits in section 2.)Alternatively, it may sometime turn out that what were considered opposed language types appear opposed just because they are applications of rival analytic models. By isolating syndromes, we arrive at generalization of wider scope and greater relevance. By isolating trait sequences we understand the spectrum of possibilities in language history. This can also serve as cross- check on claims for historical universals. The validation of analytic and historical models of specific systems and their histories is aided. By correlating the typologies of languages and language histories with extrinsic traits, we open the way towards explaining either one or other, or both. This also takes us outside linguistics proper into the extrinsic study of language, for which correlative linguistics will provide a firmer base.

            By working out typologies and universals subdomains, we arrive at a far more detailed characterization of such subdomains and their history. The subdomain may be a phase of cultural history, in which case our understanding of the extrinsic history of languages as a human institution is increased. If the subdomain is a language family or a language network, correlative comparison will make explicit the experienced specialist’s “feel” for what to expect and what not to expect in dealing with a new body of data within the subdomain: the scholarly surrogate for the native speaker’s Sprachgefühl that Romanist, Dravidianist, or Americanist has for his respective domain. (The native speaker’s intuition is limited to a particular language.) The search for successor and cosuccessor relationship in historical linguistics may begin with the search for diatopic and diachronic isoglosses or trait correlations.       


3.1.            Proposals for Revising the Schemata


Finally, the subdomain may be a pair of languages or a small enumerative set of languages as distinct from the domain defined in historical terms or in terms of extrinsic traits considered above (section 2.2.1, items (a), (b); (c),(d),and (e)). This will bring out resemblances as well as differences between the two languages, and throw light on the processes of transrendition, translation, and the bilingual’s receptive and productive skills in general involving the members of the domain. The traditional concerns of error analysis, of the language teacher’s comparison between model and filter language and attempts to anticipate ease of learning and errors, and of the translator’s comparisons and attempts to recommend or warn against certain translation all become relevant at this point. The presentation of the results of correlative comparison between a pair of languages may be either nondirected or directed towards one or the other of the two languages10.


The limiting case of such an arbitrarily enumerated subdomain is of course the subdomain of only one language – considered earlier as type (a) correlation. In such a “domain” there will be “universals”, “negative universals”, “syndromes”, but no type-yielding traits, no spectra, no trait sequence. By working with such a subdomain in relations to larger domains, one can arrive at the typological characterization of the language. The stresses and strains set up by competing traits within a language could be exhibited as quasi-spectra (yielding quasi-isoglosses) internal to a language. Such a presentation would also serve to provide the initial collations for internal reconstruction. By allowing itself the privilege of squinting at negative traits and statistical expectations, it will also illuminate a straight analytic presentation in various ways.


To the list of types (a) to (e) of subdomains presented in section 2.2.1, one more may be added:

(f)                 Belonging to an arbitrary enumerated set – typically of one language or of two or three languages – which may be called an arbitrary domain.

An arbitrary domain with only one member may be called a singulary domain.

            The list of schemata (a) to (h) of correlative collations presented in section 2.2.3, now stands revised. Schema (a) now reads as follows:

(a)                Between a trait and membership of a singulary domain (given the member L of the singulary domain D, L wholly or partially possesses or lacks the trait T1; thus, if L is partially possesses traits T1 and T2 which are in competition, L may be said to possess the quasi-spectrum T1-T2; the sum of the traits and negative traits that L possesses, and excludes universals and negative universals of the maximal domain, may be called the trait profile of L; two trait profiles may be similar in all respects or in respect of some traits)

Schemata (b) to (h) remain as they are expect for the added stipulations that the domain in question may be an arbitrary subdomain but not a singulary arbitrary subdomain, and that the trait in question may be a negative trait or trait syndrome of positive and /or negative traits. (A trait profiles is a special case of a sequence of trait syndromes.)


3.2.            Bearing of Correlative Linguistics on Certain problems of Theory


It is worth noting here that the discussion of both universals and typologies in analytic, ontogenetic, and phylogenetic areas is expected to throw light on certain interrelated notions whose precise content and mutual import have remained rather poorly explored so far. I have in mind the following vaguely intuited insights:

a.                   That in any language system the items and the rules range from the inner code to the outer margin. Thus, paraphonology, graphonomy, and onomastics are barely parts of the system. Paraphonology has presumably grammatical and semologic analogues. If script and orthography are marginal, Morse code, Pitman shorthand, and other surrogates of writing are even more so. Proper names tend to be excluded from dictionaries. “Given” names like William, Rover, Ritz, Philharmonic, One who knows, the age of reason are probably less marginal than passively “inherited” or “borrowed” names like Shakespeare, the Malagasy, India, Swahili; “household” names like Shakespeare, (THE Shakespeare, that is), Einstein, London, the English, the Thames (the one which can be set on fire) are less marginal than more “obscure” ones like Thomas Peacock, Bournemouth, Aurangabad, the Middle Kingdom. Again “native” or unmarked elements in phonology and vocabulary are more central than elements in phonology and vocabulary that are marked “borrowed” or “learned” (chair is of course “native’ or unmarked in this context, no matter what the etymologist has to say­). The notion of “basic” vocabulary is also relevant here. Mixing metaphors further, deep rules have been set against low-level rules. (Compare section 2.3[11]). Formal universals (e.g. all languages have phonemes, morphemes, form classes, and transformations) are deemed to be more important than substantive universals (e.g. all languages have vowels, consonants and intonation; nouns, verbs, and predications; animate/inanimate, special markers for speakers and listener). Moderately slow, deliberate speech – all of which tend to obliterate segmental and prosodic contrasts. Derivation but not inflection can be consigned to small print or left out.

(b)               Linguists in their commendable desire to avoid linguistic ethnocentrism fight shy of “loaded” characterizations like “exotic” or “implausible” or “strange” in describing traits like implosive, dental slit fricatives, voiced aspiration, elongated gender, or the pervasive cleavage between ordinary and honorific in the vocabulary (seen in Javanese and Persian). But it will not be too difficult, and will be probably worthwhile, to replace these characterizations by more carefully and objectively defined distinctions. For example, “displaced” articulations like labiodental, apicolabial, retroflex, dorsopostvelar, uvular, and pharyngeal articulations are less common than nondisplaced ones like bilabial, apicondental, and apicoalveolar, dorsovelar, and glottal articulations.

(c)                A similar comment may be offered on another set of epithets usable in relation to trait sequences. Some changes are dubbed “natural” or “plausible” or “progressive” or “economizing” or “balance-restoring;” while others are dubbed “strange” or “degenerative” or “costly” or “imbalance-producing.” This is obviously connected with the nonexotic /exotic dichotomy. A change “restoring” nonexotic patterns is presumably more likely than a change bringing in exotic features. At least sometimes the implied value judgments make obvious sense, for example, hypertrophy of homonyms or synonyms; the dozen senses of hari - or the several dozen synonyms for “water” in Sanskrit are suspect, not expectable in a “living” language. So the alleviation of homonymy and synonymy in functors and contentors is a really accepted explanatory motivation in discussions of linear phylogeny. Again, a change, filing a “hole” in a phonological, grammatical or lexical paradigm, has something “natural” about it.


3.3              Applications of correlative Linguistics

Enough ha been said on this subject. As examples of application from the general to the specific, see the sample hypotheses (Section 2.3). Application from correlative linguistics to other parts of linguistic theory applications from correlative linguistics to the extrinsic study of language, and, finally, applications from theory to practical concerns have all been hinted at in Sections 3, 3.1 (first two paragraphs), and 3.2.



By methodology in linguistics, I understand the procedures that take us from observations of the exercise of the skills of production, reception, and reproduction on the part of native, adherent, and (if need be) other speakers, from observed subcorpora, and from tentative determinations of successor and cosuccessor relationships between texts, items, and rules to maximally validated presentations of analytic systems, of linear and dendroidal phylogenetic processes, analytic and phylogenetic universals, and typologies of subdomains and of the whole domain in every human language, living, dead, or extinct. To the extent that we are prepared to go beyond linguistics proper into behavior-oriented, text-oriented, and system-oriented studies of language, we can also offer explanations that being in matters extrinsic to language as well. (The exercise of receptive skills, it may be noted in passing, also includes the giving of metalinguistic judgments, for example: that two texts or text fragments are or are not reproductions of each other in the same language or across different language systems; that a given text is appropriate in a given meaning; and that two texts are homogeneous in language or are in a successor or cosuccessor relationship.)


Methodological procedures can be grouped under the following phases: (a) collection and storage; (b) collation and collated storage; (c) analysis or prehension of small-scale patterns, wide-ranging hypotheses, and fundamental paradigms or postulates; (d) presentation and storage of collated and validated presentations; and (e) validation (inclusive of relative evaluation). These different phases are not equally amenable to the use of mechanical aids (consider sound recording, photocopying, data processing, and storage and retrieval systems) or of well-defined (especially algorithmic) procedures as opposed to rough-and-ready, heuristic, ad hoc, or intuitive procedures (consider alphabetization, formulation of generative routines, ear-phonetic transcription, “semantic transcription,” assembly-line phonemics à la early Pike, and specified format for formulating hypotheses and presentations).


            We may make the following comments: (1) A prehension or discovery is the least amenable to mechanical aids and well-defined procedures; collection, collation, and storage are the most amenable. (2) The traditional prejudice against mechanical aids and procedures, especially in phases (b) and, (d) is on the wane. (3) Phonology and graphonomy are more amenable, semantics least amenable to such aids and procedures. (4) Areas requiring the handling of large bodies of data have a grater need for such aids and procedures, e. g. grammar more than Phonology, lexicon more than grammar, larger domains more than smaller domains. (5) The prehension and validation of small-scale patterns are more amenable to such aids and procedures; the validation and, even less so, prehension of fundamental paradigms are the least amenable; computer-aided validation of the analysis of specific languages is a possibility.11 (6) The increasing formularization of presentation certainly facilities validation, but then t probably has a ceiling, if language is in some fundamental sense crude, fuzzy, leaky, open-textured, or ill-defined; it is certainly not an accident that most analytic models intended for actual application provide for a dustbin in which to put sweepings from under the rug – call it usage or idiom or performance or high delicacy zone or nipāta or, even, lexicon. (7) As we come to understand language or any subdomain of it better, the intuitive “feel” of the specialist that aids him in prehension and evaluation will become more widely available as a set of objective formulations; this “vulgarization” of the art of linguistic methodology probably has an upper limit.


4.1              The Methodology of Correlative Comparison


When we apply the forgoing wider considerations to correlative comparison as such, we can anticipate rapid advances in the next few years in archiving, i.e. systematic collection, collation, and storage of data, collations, and presentations. Something like linguistic analogue of the Human Relations Area Files is badly needed. This will materially assist the measurement of the strength of analytic correlations. Phylogenetic correlations call for a different kind of archiving – not of systems and corpora, but of rules of successor and cosuccessor relationships between corpora and between systems. Presumably the refinements in the formalizations (of the kind proposed in section 2.2.3 and 3.1) will be of interest to students of comparative ethnography.


Correlative linguistics eminently shows the recursive or cyclical character of the linguistic method. We start with initial collations (L1 has or hasn’t T1, and L1 and L2 both have or haven’t T1); propose a hypothesis and validate the collations with reference to consequent or derived collations; if the hypothesis is valid, the new collations based on it suggest other proposals; and so on. The so-called initial collations are themselves hypotheses based on more primitive observations and are separated from latter by the same scientific leap which resists formalized procedure.


It is not strictly necessary, but still perfectly legitimate, to seek justification for this leap in formulating analytic universals in considerations outside linguistics proper. Consider the following conversation:


By one estimate, there are about four thousand languages spoken today, and there must have been many more in the past, some of which have probably left no trace at all, “How can you verify your universal theory without a knowledge of quite a number of them?”

“It is true that we transformationalists have studied only a handful of languages in a really intensive way, but each new language that we study intensively in the future will support the conclusions that we have already drawn. I am confident of this, because it seems to me that if we assume that any infant can learn any language – that no infant is genetically a speaker of a specific language – then every attribute we postulate in order to explain an infant’s ability to learn one language must be true of any child’s learning of any language, and so must be a universal condition of a universal grammar. Thus, on the basis of the evidence that we have from the study of a few languages we can safely assume that for learning languages there must be a schematism in the mind – a physical mechanism in the brain – that is the same in every human being”.12


While the assumption that “any infant can learn any language”, given an environment of a certain kind, seems to be safe enough, it is clear that this rationalization of the scientific leap in correlative comparison will have a point only if one can formulate a validation procedure for sorting out those traits of a specific language that have to be presupposed by ANY infant’s ability to learn THAT language under certain universal conditions from the traits that are, for one reason or another, not so presupposed.

            By way of concluding, I should like to indicate an epistemologically vulnerable spot. For some, the mention of neogrammarian hypothesis as a sample of phylogenetic correlations may have rung a warning bell. It is not quite clear how far this hypothesis (which is currently under fire) is a truth claim that can be proved, refuted, or replaced by a revised truth claim, and how far it is merely a methodological postulate that, in conjunction with the family tree hypothesis (sample 22 above), makes comparative phonological reconstruction possible. Probably, the same is the case with the claims that transformations are all meaning-preserving or that they all “precede” the lexical pass or the bad odor once associated with context-sensitive rules. Are these truth claims about formal universals or demands for methodological economy?

            Even assuming that we can find some touchstone with which to answer these questions and make the necessary discriminations, there are three further questions: (1) What bearing, if any, has this distinction on the philosophical distinction between categories involved in category mistakes and classes? (2) How far is the concept of validation or evaluation applicable to fundamental paradigms or postulates? (3) What bearing, if any, has this distinction on the alleged possibility of nonautonomous facts in linguistics – facts that won’t brutally stare you in the face, but will be available only to noses sensitized by certain theories?

            Are we perhaps dealing with a three-way distinction among universal traits – genuine truth claims, defining traits, and methodological postulates? To say that the language has AT LEAST two articulations or strata is offering to define the commonest use of the term language; but to say, with Saussure or Bloomfield, or Hjelslev or Hockett or Ross-McCawley-Lakoff, that it has only two, or, with Trager or Chomsky or Lamb or Halliday, that is has at least three, is making a true claim. To say that the claim that is has, eight strata is prima facie dubious is making a methodological demand. Now to which of these three belongs the general conspiracy to agree that all languages have the same number of strata?




            The following is a selected, partially annotated bibliography of works on correlative linguistics. (Not all of the items have been seen by me personally.) The bibliography includes all references cited in the text of this article.

            In addition to the specific items below, I also recommend the following general sources on correlative linguistics: (a) the archiving issue of IJAL 20 (2) (1954); (b) the translation issue of IJAL 20 (4) (1954); (c) the typology papers in IJAL 26 (3) (1960) and IJAL 28(4) (1962); (d) the section on language universals in ICL-9 (1962); (e) the section on typology in ICL-10 (1970); (f) the papers in the plenary sessions on language universals and the section on typology in ICL-11 (1972); and (g) TLP-2 (1966), -4 (1972).





AL                   Anthropological linguistics

ICAES-5          Selected papers of the V international congress of anthropological and ethnological sciences. Philadelphia; University of Pennsylvania press 1960. 

ICL-8               Proceedings of the V111th international congress of linguists. Edited by Eva Sivertsen. Oslo: Oslo University press 1958.

ICL-9               Proceedings of the 1Xth international congress of linguists. Edited by Horace G. Lunt. The Hague; Mouton. 1962.

ICL-10             Actes du xe congress international des linguists. Edited by A.Graur at al. Bucharest: Academiei Republicii Sociaaliste Romania. 1970.

ICL-11             Proceedings of the X1th international of congress of linguists. Edited by Luigi Heilmann. Bologna and Florence. 1972.

IJAL                International journal of American linguistics.

IRAL                International Review of applied linguistics.

Lg                    Language.

TLP                 Travaux linguistique de Prague.






1948        The importance of language universals. Word 4: 168-172.


1958        “The metaphor: a psychological inquiry,” in person, perception and interpersonal                    behavior. Edited by R. Tagiuri and L. Petrullo, 86-94. Stanford; Stanford University Press.


1957    Criteria for phonetic similarity. Lg  33: 538-543. (Attestations of unusual phonetic distance between allophones and between phones in successor relationship.



1968    Universals in linguistic theory. New York: Holt, Rinehart and Winston.


1949   Syntactic relations and linguistic typology. Cahiers Ferdinand de Saussure 8: 5-20


1958   Linguistic typology: an inaugural lecture. London; School of Oriental and African Studies, University of London. (Reprinted 1966 in Five inaugural lectures: language and language learning. Edited by Peter Strevens, 27-49. London: Oxford University Press.)



1966   Componential analysis of general vocabulary: the semantic structure of a set of verbs in English, Hindi and Japanese. IJAL 32 (2), part 2



1954   La classification des langues. Conferences de l’Institute de Linguistique de l’Universitė de Paris 11:32-50


1966   Problėmes de linguistique gėnėrale. Paris; Gallimard.



1969   Basic color terms: their universality and evolution. Berkeley and Los Angles: University of California Press.



1971   Problems of typology and genetic linguistics viewed in a generative framework. Janua Linguarum, Series Minor 106. the Hague: Mouton.



1911   “Introduction,” in Handbook of American Indian languages. Edited by Franz Boas, 1-83. Washington, D.C.: Smithsonian Institute.



1966   Transformulation: structural translation. Aeta Linguistica Hafniensia 9 (2): 130-144.



1970   Psycholinguistics: selected papers. New York : The Free Press.



1960    “The pronouns of power and solidarity,” in style in language. Edited by Thomas A.Sebeok, 354-376. Cambridge, Mass: M.I.T. Press.



1966    “Language universals in anthropological prospective,” in universals in language, second edition. Edited by Joseph H. Greenberg. Cambridge, Mass: M.I.T. Press.



1967     A linguistic theory of translation. Languages and language learning series. London: Oxford University press.



1964   Transformational generative grammar and constrastive analysis. Language   Learning 14: 147-160.



1970    The discovery of universals in multilingualism. Monograph series on languages    and linguistics 23:13-22.



1958    General linguistics and comparative philology. Lingua  7:134-174.

1966    Towards a general comparative linguistics. The Hague: Mouton.



1970    Two models of socially significant linguistic variation. Lg 46: 551-563,



1964    “Baby talk in six languages”, in The ethnography of communication. Edited by John J. Gumperz and Dell Hymes, 103-114. American Anthropologist 66 (6), part2.



1960    The phonemes of Bengali. Lg  36:22-59. (includes typological profile of Bangla phonology.)



1968    “The case for case,” in Universals of linguistic theory. Edited by Emmon Bach and Robert T. Harms, 1-88. New York ; Holt,Rinehart and Winston.


1971        “Towards a theory of deixis,” in Proceedings of the pacific conference on contrastive linguistics, 219-242. working papers in linguistics 3. Honolulu: University of Hawii.



1901        Die Klassifikation der Sprachen. Marburg

1905    Die Aufgabe und gliederung der sprachwissenschaft. Halle.

1936        Die Haupttypen des Sprachbaues. Aus Natur und Geisteswelt, third edition, Berlin : B.G. Teubner. (Originally published 1901.)



1953    The category of person in linguistics. Berlin: de Gruyter. (Reviewed by Householder 1955; Hymes 1955.)



1940            Interrogatif et indèfini: un problème de grammaire comparèe et de linguistique gènèrale. Paris.

1944            Systèmes de dèictiques. Acta linguistica 4:111-129.

1948        De la linguistique comme science de loi. Lingua 1:25-33.



1949            Standard average European and Czech. Studia linguistica 3: 65-85.

            (Typological profiles of the domain and of Czech.)



1956            Comparison of linguistic systems. IJAL 22:273-274



1954    A quantitative approach to the morphological typology of language,” in Methods and perspectives in anthropology. Edited by R.F. Spencer, 192-220. Minneapolis: University Press. (Reprinted 1960 in IJAL 26:  178-194. critiques by Kroeber 1960a; Saporta 1957; Winter 1967; cf, also Pierce 1962.)


1956    The measurement of linguistic diversity. Lg 32: 109-115. (Extrinsic correlation of a language network trait; cf. Lieberson 1964.)


1957a  Essay in linguistics.Viking  Fund Publications in Anthropology 24. NewYork: Wenner-Gren Foundation. (see chapter 8 on “Order of affixing: a study in general linguistics.”)

1957b  The nature and use of linguistic typologies. IJAL 23: 68-77.


1962        Is the vowel-consonant dichotomy universal? Word 18: 73-81.


1966a  “Some universals of grammar with particular reference to the order of meaningful elements,” in Universals in language, second edition. Edited by Joseph H. Greenberg. Cambridge, Mass.: M.I.T Press.

1966b  “Language Universals,”in Theorectical foundations. Current Trends in Linguistics 

              3.  Edited by Thomas a. Sebeok, 61-112.The Hague: Mouton. (Discusses marked                 vs, unmarked, kinship terminology, word association.)

1966c  Synchronic and diachronic universals in phonology. Lg 42: 108-117


1969    “Some methods of dynamic comparison in linguistics,” in Substance and structures of language. Edited by Jaan Puhvel, 147-203. Berkeley and Los Angeles: university of California Press.



1966        Universals in language, second edition. Cambridge, Mass.: M.I.T press. (originally published 1963.)



1959        General linguistics in university studies. Universities Quarterly 118-142.



1970        “Language structure and language function,” in New horizons in linguistics. Edited by John Lyons, 140-165. Harmondsworth: Pelican.



1964    The linguistic sciences and language teaching. London: Longmans. (see chapter 5 on comparison and translation.)



1965    A bibliography of contrastive linguistics. Washington, D.C.: Center for Applied Linguistics.



1954            Transfer grammar, IJAL 20:259-270.

1970    Papers in structural and transformational linguistics. Dordrecht: Reidel.



1956-1957   Untersuchungen zer allgemeinen Grammatik, three volumes. Heidelberg.


1962    “Zur Erforschung von Sprachtypen: Methoden and Anwendungen, 11,” in Fachtagung fÜr indogermanische und allgemeine Sprachwissenschaft, 31-55. innsbrucker Beiträge zur Kulturwissenschaft 15. Innsbruck.




1939            Mèthode pour obtenir les lois concretes en linguistique gènèrale. Bulletin de la sociètè linguistic de paris 40: 70-74.




1950    The analysis of linguistic borrowing. Lg 26: 210-231

1956        Bilingualism in the Americas: a bibliography and research guide. Publications of the American Dialect Society 26. University, Ala.: University of Alabama Press (Discusses types of changes through influence.)



1941            “Incorporation as a type of language structure,” in Humanistic studies in honor of John Calvin Metcalf, 65-79. Charlottesville, Va :University of Virginia Press.


Hill, Trevor      

1958            Institutional linguistics. Orbis 7:441-445.



1928        Principles de grammaire gėnėrale. Det. Kgl. Danske videnkabernes Selskab. Copenhagen:Host.


1935-1937             La catėgorie des cas: ėtude de grammarie gėnėrale,two volumes. Acta jutlandica 7(1); 9(2).


1948            Implicationsof Bloomfield’s Algonquian studies. Lg 24: 117-131.

1955    A manual of phonology.IJAL Supplement 21(4), part 1. (See Chapter 2 for a typology of phonological systems.)

1958    A course in modern linguistics. New York: Macmillan. (See chapters 20-31 and 33 on analytical grammatical and morphophonological spectra; Chapter 64 on analytical universals that define language.)

1959            “Animal’languages’ and human languages”, in The evolution of man’s capacity for culture. Edited by J.N. Spuhler, 32-39. Detriot: Wayne State University Press.(Discusses analytical universals that define language.)

1960    The orgin of speech. Scientific American (September) : 5-10.




1970    The linguistic cycle. Language sciences13: 1-7. (Discussess phylogentic grammatical traits.)



1966            Language typology; nineteenth and twentieth century views. Washington,D.C.: Georgetown University press.



1955    Review of The category of person in linguistics by P. Forchheimer. Lg  31:93-99.

1960    First thoughts on syntactic indices. IJAL  26:195-197.

1970            Linguistic speculations. London: Cambridge University Press.



1970            Linguistic variability and intellectual development. Translated by G.C. Buck, F.Raven, and A.Raven Miami Linguistic Series 9. Coral Gables: University of Miami Press. (Originally published 1836.) (On extrinsic correlations with grammatical traits.)



1955    Review of The category of person in linguistics by P. Forchheimer. IJAL 21: 294-300.

1956a  Review of “Comparasion des systemes phonologiques des langues caucasiennes et americaines” by Tadeusz Milewski. IJAL 22:281-287.

1956b  Review of Inductively arrived at models for cross-genetic comparisons of American Indian languages by Carl F.Voegelin. Lg 32:585-602.

1960            Discussion of the symposium on translation between languages and culture. Al 2:81-85.

1961    On typology of cognitive styles in languages, with examples from Chinookan. Al 3 (1):22-54 .

1968            “Linguistics:the field,”in International encyclopedia of social sciences, volume nine. Edited by David l.Sills, 351-371. New York:Mcmillian.


HYMES, DELL. H.,editor

1964            Language in culture and society. New York: Harper and Row. (See Chapters 8-11 and 12-18, as well as chapter 67.)



1955            Athapaskan numeral system. IJAL 21:26-45.



1942            Kindersprache, Aphasie und allgemeine Lautgetsetze. Uppsala Universitets Arsskrift 9. Uppsala. (Reprinted in Jakobson 1962.)


1957            Shifters, verbal categories, and Russian verb. Russian language project. Cambridge, Mass.: Harvard University Press.


1958            “Typological studies and their contribution to historical comparative linguistics,” in ICL-8:17-25,33-35. (Reprinted in Jakobson 1962.)


1961    “Why ‘mama’ and ‘papa’?” in Perspectives in psychological theory. Edited by B.Kaplan and S.Wapner, 124-134. New York.

1962            Selected Writings. I. Phonological studies. The Hague: Mouton.

1966            “Implications of language universals for linguistics,” in Universals in language, second edition. Edited by Joseph H. Greenberg. Cambridge, Mass.: M.I.T. Press.

1967    To honor Raman jakobson, three volumes. The Hague: Mouton.



1956            Fundamental of language. The Hague: Mouton. (See Chapter 4 on phonemic patterning and ontogenetic phonological universals.)



1894            Progress in language. London: Sonnenschein.

1922            Language: its nature, development and origin. London: G.Allen and Unwin. (Surveys history of typological studies.)

1924    The philosophy of grammar. London: G.Allen and Unwin; New York: Holt.

1941            Efficiency in linguistic change. Copenhagen: Munksgaard.



1967    On a generative grammar of the Balkan languages. Foundations of Language 3: 117-123.



1969a            “Language: linguistics: the applications,” in Language and society. Transactions of the Indian Institute of Advanced Study 8, 76-88. Simla: The Institute. (Excerpted 1969 in Language Sciences 4:11-13.)

1969b  The scope of a linguistic survey. Pàkha sanjam (Patalia) 1 (O1): 5-12.



1966            Translation on basis for a contrastive linguistic analysis. IRAL 4:175-182.


1964            Relatedness between grammatical systems Lg 40:1-20.



1953            “Universals categories of culture,” in Anthropology today. Edited by Alfred L. Kroeber, 507-523. Chicago: University of Chicago Press. (Includes linguistic universals.)



1959    A quantitative typology of languages. Language and speech 2:72-85. (Analytic Phonological types.)




1907    The numeral system of California. American Anthropologist 9:690. (Discusses their phylogenetic instability.)

1916        Arapaho dialects. University of. California Publications in American Anthropology and Ethnology 12: 71-138. (See especially 9off. On universals in relation to linguistic analysis.)

1960a  On typological indices I: ranking of languages. IJAL 26: 171-177. (Critique of             Greenberg 1954.)

1960b            Statistics, Indo-European, and taxanomy. Lg 36: 1-21. (Extracted in Hymes [e.d.] 1964: 654-659.)



1968            “Unique types and typological Universals,” in Pratidānam: Indian, Iranian, and Indo-European studies presented to F. B. J. Kupier on his 60th birthday, Edited by J. c. Heestermann et al., 68-88. The Hague: Mouton



1965    The evolution of grammatical categories. Diogenes 51 (Fall): 55-71.



1954            Morfolgičeskaja klassifikatsija jazykov. Moscow: Izdatel’stvo Moskovskogo Universitet. (German translation 1956 as Die Morphologische klassofikation der sprachen. Halle: Max Niemayer.)



1957            Linguistics across cultures. Ann Arbor, Mich.: University of Michigan Press.



1960    “Language, evolution, and purposive behaviour,” in Culture in history. Edited by 

        Stanely A. Daimond, 869-893. Newyork: Columbia University Press. (On  

        language universals.)



1863    The “translation-paradigm”: a technique for contrastive syntax. IRAL 3: 221-225.



1964    An extension of Green berg’s linguistic diversity measures. Lg  40: 526-531. (Cf. Greenberg 1956.)



1953    Oneida verb morphology. New Haven: Yale University Press.



Phonemic and non-phonemic phonology: some typological reflections. IJAL 28: 127-133.



1962            Etymology and general linguistics. Word 18: 1-2, 198-219.



1962    La notion de “force” et les changements phonétiques. Studia Linguistica 16: 38-       44.



1968    Le langage. Encyclopédie de la pléide 25. Paris: Gallimard. (On language             classification.)



1965    The inflectional component of a word-and-paradigm grammar. Journal of             Linguistics 1: 139-171.

1970    “Recent developments in Morphology,” in New horizons in linguistics. Edited by             John Lyons, 96-114. Harmondsworth: Pelican.



1972    Colour and colour terminology. Journal of Linguistics 8: 21-33.



1970    English as a VSO language. Lg  46: 286-299.



1953    Modes of translation. Durham university journal  45.



1971    John is easy to please: encounters with the written and the spoken word.   New York: Farrar, Straus and Giroux; London: Secker and Warburg. 



1926            Linguistique historique et linguistique générale, volume one, second edition.             Paris: Champion. (originally published 1921.)

1948            Linguistique historique et linguistique générale, volume two, second edition.             Paris: Champion. (Originally published 1936.)

1948            “Introduction à  la classification des langues,” in Linguistique historique et             linguistique générale, volume two, second edition, 53-69. Paris: Champion.



1950            Typology of languages. Journal of the Acoustaical Society of America  22: 698-     701. (Discusses the phonological shapes of words.)



1950            Sprachtypologische Untersuchungen I. Allgemeine Einführung and Theorie der             Wortbildung. Studia Linguistica 4: 54-93.



1950a            Podstawy teoretyczne typologii jezyków. Bulletin de la société polonaise de             linguistique 10: 122-140. (Discusses the theorectical basis of language typology.)

1950b  La structure de la phrase dans les langues indigenes de l’Amérique du Nord.   Lingua Posnanensis 3: 248-267.

1951    The conception of the word in the languages of North American natives. Linga             posnanensis 3: 248-267.

1953            Typologia syntaktyczna jezyków Amerykanskich [Syntactic typology of             American (Indian) languages]. Bulletin de la Société polonaise de linguistique12:             1-24

1954            Phonological typology of American Indian languages. Lingua Posnanensis 4:             229-276.

1955            Comparison des systèmes phonologiques des langues causiennes et américanes.             Linguia Posnanensis 5: 136-165. (Reviewed by Hymes 1956.)

1957    Le problème des lois en linguistique générale. Lingua Posnanensis 6: 120-136.



1961    A statistical morpho-syntactic typology study of Colorado (chibcha). IJAL 27:       298-307.



1963    Les problèmes théoriques de la traduction. Paris: Gallimard.



1959    Cross-language parallels in parental kin terms. AL I (9): 1-5.



1954            Suggestions on the archiving of linguistic material. IJAL 20: 111-115.



1945            Linguistics and ethnology in translation-problems. Word1: 194-208. (Reprinted in          Hymes [ed.] 1964: 90-97.)

1964            Toward a science of translating: With a special reference to principles and             procedures involved in Bible translating. Leiden: E.J. Brill. (reviewed by             waterman 1966.)

1969    Science of translation. Lg 45: 483-498.



1969    The theory and practice of translation. Helps for Translators 8. Leiden: E.J. Brill.



1963            Limitationas of morphological processes: a note. Lingua 12: 220-225. (Cf.             Uhlenbeck 1962.)



1960    The cross-cultural generality of visual-verbal synesthetic tendencies. Behavioral             Science 5: 146-149.

1966            “Language universals and psycholinguistics,” in Universals in language, second             edition. Edited by Joseph H. Greenberg. Cambridge, Mass.: M.I.T. Press.



1949     Ist eine allgemeine Sprachwissenschaft auf empirischer Grudlage mőglich? 

       Archiv Orietalni 17: 236-238.



1962            Possible electronic computation of typological indices for linguistic structures.             IJAL  28: 215-226. 



1948    Tone languages. Ann Arbor: University of Michigan Press.



1970    “The method of universal grammar,” in Method and theory in linguistics. Edited by Paul Garvin, 113-131. The Hague: Mouton. (Illustrated with reflexivization in             English and Mohawk.)



1960 “Toward a phonological typology of the Indian linguistic area,” in Linguistics in

     South Asia. Edited by Thomas A. Sebeok, 543-577. Current Trends in Linguistics?   

     5. The Hague: Mouton



1958    What is general linguistics?  Lingua 1: 8-24.



1952    Noun and verb in universal grammar. Lg  28: 289-298.

1959    In defence of WP. Transactions of the philological Society, 116-144.



1955    Zum problem der Fragenmelodie. Lingua 5: 87-108.



1929    Die nominate Klassifikations-Systeme in den Sprachen der Erde. Mődling bei             Wein.



1911    The history and varieties of human speech. The popular Science Monthly 79: 45-       67.

1923            Language. New York: Harcourt, Brace. (see chapter 6 on analytic grammatical             syndromes.)

1935            “Language,” in Encyclopedia of social sciences, volume nine. Edited by E. R. A.             Seligman, 155-159. New York : Macmillan. 

1946            American Indian grammatical categories. Word 2: 103-112. (Reprinted in Hymes             [ed.] 1964: 101-107.) 

1949            Selected writings of Edward Sapir in language, culture, and personality. Edited by David G. Mandelbaum. Berkeley and Los Angeles: University of             California  Press.         



1932    The expression of the ending-point relation in English, French, and German. Lg             Monograph 10. 



1957    Methodological considerations regarding a statistical approach to typologies.             IJAL  23: 107-113. (Especially phonological typologies.)



1966            Historische sprachvergleichung und ihre typologische Ergänzung. Zeitschrift für             Deutchen Morgenländischen Gesellschaft  116: 8-22.



1926    Die sprachfamilien und sprachenkreise der Erde. Heidelberg: Carl Winter. (Some             grammatical syndromes established and correlated with extrinsic culture-            historical traits.)



1942            Versuch einer allgemeinen Sprachtypologie. Buletinul Institutului de Filologie             Romînă “Alexandru Philippida” 9: 60-71.



1971             “Developmental psycholinguistics,” in A survey of linguistic science. Edited by        William O. Dingwall, 298-410. College Park: University of Maryland Press.   (About ontogenetic universals.)



1964    Family-tree diagrams. Lg  40: 557-565.

1971a            “Detecting prior creolization: an analysis of the historical origins of Marathi,” in    

             Pidginization and creolization of languages. Edited by Dell H. Hymes, 177-195.             London: Cambridge University Press.

1971b  “Some sociolinguistic implications for historical linguistics and genetic             classification of language,” in Proceedings of the first All-India Conference of             Linguists, 85-95. Poona: The linguistic Society of India at Deccan College.



1958            “Typological and statistical aspects of distribution as a criterion in linguistic             analysis,” in ICL-8: 182-194.

1959            Probability and structural classification in language description. Copenhagen.



1850    Die Klassifikation der sparchen, dargestellt als die Entwicklung der Sparchidee.             Berlin: Dümmler.

1860            Charakteristik der hauptsächlichsten Typen des Sprachbaues. Abriss der             Sprachwissenschaft 2. Berlin: Dümmler.



1963    The biological model and historical linguistics. Lg.  39: 159-169.



1962    “An outline of linguistic typology for describing multilingualism,” in Study of the             role of second languages in Asia, Africa and Latin America. Edited by Frank A.             Rice, 15-25. Washington, D.C.: Center for Applied Linguistics.



1951            Structure immanente de la langue franchise. Travaux de Cercle Lingistique de             Copenhague 6. Copenhagen: Nordisk Sprog-og Kultur. (On grammatical types.)



1965            Structure translation. The Hague: Mouton.



1941    “The theory of accentual systems,” in Language, culture, and personality. Edited by Leslie Spier, 131-145. Manasha, Wisc.: Sapir Memorial Publication Fund.

1952    The field of linguistics. Washington, D.c.: Foreign Service Institute. (Originally             published 1949.)

1961    The typology of paralanguage. AL 3 (1): 17-21.



1929    Zur allegmeinen Theorie der phonologischen Vokalsysteme. TCLP  1: 39-67.

1931    Die phonologischen Systeme. TCLP  4: 96-116.

1939            Grundzüge der phonologie. TCLP 7. (French translation 1949 as Principles die             phonologie by Jean Cantineau. Paris: Klincksieck.) (Analytic phonological             typologies.)



1962    Limitations of morphological process: Some preliminary remarks. Lingua 11:             426-432 (Cf. Odendal 1963.)



1953            Descriptive semantics and linguistic typology. Word 9: 225-240.

1960    The study of the so-called exotic languages and general linguistics. Lingua 9:             417-434.



1969    Somn: sleep: an exercise in the use of descriptive linguistic techniques in literary             translation. Babel 15 (1): 4-14; 15 (2)



1965            Structurnaja tipologija jazykov. Moscow: Nauka. (Translated 1968 as Principles             of structural typology. The Hague: Mouton.)



1962            Théorie et pratique de l’interprétation, avec application particulière à l’anglais et             au français. Munich: Max Hüber.



1952    Review of “La structure de la phrase dans les langues indigènes de l’Amérique du             Nord” by Tadeusz Milewski. Lg  28: 405-410.

 1954            Inductively arrived at models for cross-genetic comparisons of American Indian             languages. University of California Publications in Linguistics 10. Berkeley and             Los Angeles: University of California Press. (reviewed by Hymes 1956.) (On             phonological correlations.)

1955    On developing new typologies and revising old ones. Southwest Journal of             Anthropology 11: 255-260.

1956    Linear Phonemes and additive components (SGC). Word 12: 429-443.

1957    Six Statements for a phonemic inventory. IJAL  23: 78-84.

1958    Review of Syntactic structures by Noam Chomsky. IJAL  24: 229-231.

1960            “Subsystem typology in linguistics,” in ICAES-5: 202-206.

1961            Typology of density ranges II: contrastive and non-contrastive syntax. IJAL 27:       287-297.

1962            Methods for typologizing directly and by distinctive features (in reference to Uto-            Aztecan and Kiowa-Tanoan vowel systems). Lingua 11: 469-487.

1967            “Subsystems within systems in cultural and linguistic typologies,” in To honor             Roman Jakobson, 592-599. The Hague: Mouton.



1960            Typology of density ranges I: Introduction, IJAL 26: 198-205.



1956    The scope of whole system (‘distinctive feature’) and subsystem typologies. Word   12: 445-454.



1966    Review of Toward a science of translating by Eugene A. Nida. Lg 42: 93-105.



1953            Languages in contact. New York: Linguistic Circle of New York at Columbia             University. (Reprinted 1968. The Hague: Mouton.) (On types of changes through             influence.)

1966    On the scientific structure of language,” in Universals in language, second             edition.            Edited by Joseph H. Greenberg. Cambridge, Mass.: M.I.T. Press.



1954            Archiving and language typology, IJAL 20:101—107.



1938    Some verbal categories in Hopi. Lg. 14: 275-286. (Reprinted in Whorf 1956b:             112-124.)

1941    “The relation of habitual thought and behavior to language,” in Language,             culture, and personality. Edited by Leslie Spier, 75-93. Menasha, wisc.: Sapir             Memorial Publication Fund. (Reprinted in Whorf 1956b: 134-159.)

1945            Grammatical categories. Lg 21: 1-11. (Reprinted in Whorf 1956b: 87-101.) (On             analytic grammatical spectra.)

1956a            “Language: plan and conception of arrangement,” in Language, thought and             reality. Edited by John B. Carroll, 125-133. New York: Wiley. 

1956b            Language, thought and reality. Edited by John B. Carroll. New York: Wiley.



1967    Some basic difficulties in the application of quantifying techniques to             morphological typology,” in ICL-10: 545-549. (Critique of Greenberg 1954.)



1959            Subsytem typologies and area linguistics . AL 1 (7): 1-88.



1957             Frequencies and inventories of phonemes in nine languages. IJAL 23: 85-93.





            This was presented in absentia at the IXth International Congress of Anthrological Sciences at Chicago, Fall 1973 and published in Sd Tax [ed.] World Anthropology, volume Approaches to language: Anthropological issues, ed. William C. Mc Cormack, Stephen A. Warm, The Hague: Mouton, 1978, p151-188.



1 In current usage, the terms institutional linguistics or, more fashionably, socio-linguistics have come to lump together indiscriminately these various behavior-oriented extrinsic studies of language as well as that part of correlative linguistics which is concerned with the correlation between intrinsic traits of languages and the traits of users and situations.


 2  Catford’s proposal (1967: 23, 56-61) to regard the former as phonological translation is a brilliant insight. I have tried to incorporate it without disturbing the conventional meaning of the term translation.


3 Reconstructions, whether internal, comparative, or reverse, provide limited access to prior states of languages or language families. Reconstruction text fragments from historically related text fragments, and phonological items from historically related phonological items, is primary. Reconstruction of nonphonological items, of analytic rules and rule system, and of texts on the basis of the primary reconstruction is derivative and much less certain. Prelanguages and protolanguages are, at best, relatively short descent subchains. It is misleading to think of them as states of languages and to hope of making a secondary reconstruction of fables in them. Thus, a reconstructed item or rule assignable to pre-A may conceivably antedate a reconstructed item or rule assignable to proto-A-B, where A and B are a pair of languages in codescencent relationship. (I owe this last point to Gordon H. Fairbanks[personal communication])

4 It would seem that at least the present generation of historical linguists of the transformational-generative persuasion are guilty of harping on descent relationships between rules and between systems at the expense of those between text fragments and items. The formulation of succession or cosuccession rules of the latter kind (pà p or p, b à b or Juppiter dyaus-pitar-pointing to the existence of the compound in the ancestral language or, to take an example involving reexpression rather than rerendition, digged à dug) is logically prior to the formulation of the succession or cosuccession rules of the former kind (Rule X à Rule Y à null Rule or Rules X followed by Y à Rules X followed by Z followed by Y or Rules X followed by Y à Rules Y followed by X or three-gender paradigm two gender paradigm). The Neogrammarian slate cannot (and need not) be wiped clean!

5 We are of course talking of the twofold sense proposed here. Traditionally comparative linguistics has been usually been confined to comparative reconstruction in the methodology of linguistic prehistory.

6 One consequence of this is that the excessive emphasis so far in historical linguistics on divergence at the expense of convergence and on succession by descent at the expense of succession by influence has to be corrected.

7 An  obsulcate like [ ɹ ] or [] is a transverse-groove fricative or approximant, while a sulcate like [s] or [š ] is a fricative or approximant with a longitudinal short or long groove.

8 Cf. also Lounsbury (1953; 11-24), Robins (1959:137ff.), Matthews (1965:141-142; 1970: 110) for similar pointers towards reading a spectrum of languages types into a spectrum of language models.

9 In current usage, the term language universal trait in the maximal domain, or even in a subdomain, but also in the sense of any prefect correlation within a linguistic domain (i.e. any correlation of types [b] to [h] that has “always” or “never” in it). This is sometimes confusing and leads to awkward collocations like “regional” or “conditional universal”. By confining the term language universal to prefect type (b) correlation in the maximal domain, one can speak of a prefect correlation within a stated linguistic domain when one needs a more inclusive term.

10 The name transfer grammar (Harris 1954) or transfer comparison (Halliday, Mc-Intosh, and Strevens 1964; 120) has been suggested for a directed presentation of correlative comparison between a pair of languages. It will consist of transrendition rules (French to English: the vowel of French pur is somewhat like ew in English pew) and translation rules (English to French: English goes and is going are both usually rendered by the present of aller in French). An excessive reliance on such rules, often in an over-simplified version, was typical of a good deal of traditional second language teaching (in the present case, for teaching French renditions and expression skills to English-knowing learners). Transfer comparison may of course bring out resemblances as well (English adjectives normally translate as French adjectives).

11 The recent strictures on the search for mechanical” discovery” procedures apply primarily to phase (c) and to wide-ranging hypotheses and, even more obviously to fundamental paradigms, where they are quite valid. It is open to question, however, whether the posing of the following dilemma is valid or not – any proposed discovery procedure is either mechanical but invalid or nonmechanical but “uninteresting”(Cf. the exchange between William Haas and Chomsky reported in ICL-9: 994-998.) It is an irony that a distrust of mechanical “discovery” procedures has often been accompanied by a rather naïve faith in mechanical “evaluation” or validation procedures (called evaluation metric of the linear size of the symbol inventory, of a specific rule, and of whole description of a language), which of course presuppose a highly formalized presentation.

12 The interlocutors reported are Ved Mehta and Noam Chomsky, respectively (Mehta 1971: 210-211). The point of view attributed to Chomsky may or may not have been correctly reported, but in any case it is a possible point of view.