11 Louis Hjelmslev

11.00 Hjelmslev's early interest in languages developed into a desire to assure the scientific study of Language. Part of that maturation was a concern for the study of General Grammar at a time when this was regarded as unscientific: his thesis on the universal category of case was rejected by the University of Paris as nonlinguistic. That investigation showed his preference for the clarity of deductive generalizations over what he saw as the increasing dilution of inductive ones. In the meantime, he was sketching out a general linguistic theory that would take syntagmatic relations into account more than the usual focus on word-morphology. In 1928 he published his Principes de Grammaire Generate, and in 1935-7, the semantic study, La Categorie des cas. He mentions no distinct turning point in his thinking, but named Wivel, Sapir, Fortunatov and Saussure as providing confirmation of his own ideas, and Uldall as all but a later twin in his development of them. 11.1 The Autonomy of Linguistics, His Prolegomena record dissatisfaction with the status of Linguistics as an autonomous science, and sketched how autonomy might be approached. The prestige of Logical Positivism, exemplified in the Wiener Kreis and the writings of Rudolf Carnap, is evident. The Prolegomena is generally reckoned the most important of his publications, and his other works as preparatory for and consistent with it. Some read the Prolegomena as a logical development of de Saussure's basic ideas and of what Structuralism implies. Others saw it as a justification for practical techniques they had developed, and a codification of what they had been aiming at. Its perspective helps make approaches like Stratificational and Systemic Grammar more intelligible, and clarifies structural bases of Chomsky's work. Hjelmslev's own views of how his thought compared with de Saussure's are also relevant to evaluating directions in linguistic study. 11.2 Glossematics. Isidore of Seville's 'etymologies' had little to do with the real history of words, but did contribute to mnemonics. How could anyone forget the 'meaning' of cadaver if its etymology were first proposed as CAro DAta VERmibus (flesh given to the worms') The term Glossematics was coined by Hjelmslev and Uldall. Whatever the real reasons for their choice, an Isidoran etymology of it might serve to situate it more memorably in linguistic theory.

340 / GENERAL LINGUISTICS Glosse- in Greek matches the Romance root for tongue in the word language; -modes suggests a study with mathematical abstractness and rigor. It resembles the logistical study of language by Carnap and others, but without their neglect of the dual nature of the linguistic sign (Hjelmslev 1947:76). It also suggests how that degree of abstractness is attained, and its consequent power: computability free of subjective connotations, or even of objective denotations, which Bloomfield found a mark of scientific discourse. Mathematics and logistic deal objectively with quantity undistracted by differences in quality. Judgements about identity or difference in qualities remain clearly and nobly subjective, but once identified for the mathematician, relations among objects of even debatable qualities can be calculated, and all possible permutations computed. Glossematics ignores the peculiarities of concrete languages to establish a calculus of possible linguistic form upon which individual languages draw. For example, if language-as-system were Column A, then a language-asprocess could be illustrated under Columns I to IV. Glossematics, of course, is not as simple as A, and real languages are more complex than I to IV.

Column A is a calculus of all possible combinations for A and B. Glossematics is to be a calculus of all possible combinations in Language. Columns I through IV show actual combinations in a language. All the Columns from A through IV are independent of interpretations that exemplify them (phonetics, semantics, games, decisions, music, heads-or-tails, sentences, etc.). Yet Columns I to IV have the premised nature Hjelmslev proposes for an object to be considered a language (Hjelmslev 1947:78): 1. 2. 3. 4. 5. Content and expression Succession (text) and system Content and expression are linked by commutation Constraints within succession and system Lack of isomorphism between content and expression, and nonexpressive constituents of expressions.

Columns I-IV meet these requirements, as do Hjelmslev's examples from traffic lights, Morse code, or chiming clocks. (1) They have expressions like

11 Louis Hjelmslev / 341 A, +A or -A, and interpretations like head-or-tail, odd-or-even can be their content. (2) Drawn exclusively from (what Firth also called) the System (A, B), successive Structures are A, A B, B, or B A, and 0. (3) These structures contrast as minimal pairs. (4) Dotted lines in Columns I-IV show structures excluded (Constraints), and the absence of is a Constraint upon the System (A, B). (5) Since head or tail, odd or I'm a monkey's uncle are possible contents, there is no isomorphism between content and expression, and the non-expressive constituents of A and are these visual features of how those expressions are drawn or printed(a) the slanted lines of the printed letter A, (b) the horizontal line [-] of A, (c) the straight line of B, and (d) the curved upper and lower elements of B. Glossematics and Column A are also alike in not being falsifiable by applications (like coins standing on edge in heads-or-tails). They are just shown to be inappropriate for that particular interpretation. Here we find a challenging example of how abstract Linguistics can afford to be without being irrelevant, compared to how concrete it can afford to get without getting lost in concrete detail. 11.04 The Prolegomena and the Empirical Principle. This is Hjelmslev's concern: linguistic theory has suffered from premature and irrelevant involvement with uses to which Language can be put. While intimately involved with culture, psychology, logic, and so on, Language is definable without reference to those studies and autonomously investigable apart from them. Language as System is logically prior to, and independent of, Language as Process. It is 'a self-sufficient totality, a structure sui generis.' ( 1961:5-6). Its study should be immanent, not based on derivative or associated factors; it should be exact, with each succeeding step defined on the preceding step; and it should be appropriate, based on, and applicable to, experience. The title, Prolegomena, says this work is not the theory, but a preface concerned with the properties such a theory should have, how the theory ought to proceed, and how immanent study relates language to social phenomena. That such a preface is required is direct criticism of other linguists. Hjelmslev thinks that there is little of value he can extract from the work of his predecessors: he acknowledges resonance between his own thought and what he found later in Saussure, as well as direct indebtedness to contemporaries in the Copenhagen Linguistic Circle, especially to Hans J0rgen Uldall. For a while, he and Hjelmslev formed a complementary unity like the Chesterbelloc, of content and expression. Hjelmslev's forte was a broad linguistic base and IndoEuropean content. Uldall had done phonetics under Daniel Jones, anthropology and Amerindian with Boas, and recorded Maidu texts, one of Malinowski's early interests. 11.05 Practical Problems. Development of Glossematics was impeded during the Second World War, since Uldall's absences made direct collaboration impossible. In preparing to present a version of Glossematics

342 / GENERAL LINGUISTICS at the 10th Anniversary of the Copenhagen Linguistic Circle in 1941, Hjelmslev outlined it (translated 1973), wrote Language (1963 [1970]), then Omkring Sprogteoriens Grundlaeggelse (1943). It was translated into English by Francis Whitfield as the Prolegomena to a Theory of Language (abbreviated here as PTL) in 1953 and revised in 1961. Uldall's algebraic version in Part I of a projected Outline of Glossematics (1957) was an awkward base for Hjelmslev. He preferred his original verbal formulations: his own Part II of the projected Outline did not appear. 11.06 Prediction and Explanation. Hjelmslev mentions a barrier to progress: the a priori assumption of humanists that human acts are not amenable to scientific study, that such acts can only be described, not explained, and only inexactly described at that, because human acts are not recurrent like natural events. On the contrary, he finds it a priori reasonable to assumeand this is the core of his approach-that
...any process can be analyzed into a limited number of elements recurring in various combinations. Then, on the basis of this analysis, it should be possible to order these elements into classes according to their possibilities of combination. And it should be further possible to set up a general and exhaustive calculus of the possible combinations. (PTL .9)

On such assumptions, history could change from a discipline working at the level of mere primitive description to a generalizing science where all events and combinations would be foreseen, and their conditions established. He sees only two options open to the humanities: to remain on the level of poetry, or to combine poetic and scientific treatment as two coordinate forms of description. Linguistic theory aims at testing the thesis that every process has an underlying system. 11.07 The Empirical Principle. This goal leads to his version of the Empirical Principle:
The description shall be free of contradiction (self-consistent), exhaustive, and as simple as possible. The requirement of freedom from contradiction takes precedence over the requirement of exhaustive description. The requirement of exhaustive description takes precedence over the requirement of simplicity. (PTL 11)

Hjelmslev distinguishes this from induction. He defines that as proceeding from component to class: his approach proceeds from class to component, structurally. He says that induction results only in vague reifications incapable of generalization beyond a single language, e.g. constructs like subjunctive mood common to Latin and Greek (Latin distinguishes indicative from subjunctive, Greek opposes both to an optative, etically and emically). 11.08 Text is the sole given. So the Empirical Principle demands that text be taken as a class analyzed into components. These components themselves can be hierarchiesclasses analyzable into components which are classes of

11 Louis Hjelmslev / 343 components, etc.until analysis terminates in what he calls entities. Though recognizing that the term is debatable, he says linguistic method must be deductive, proceeding from class to component, structurally:
Scientific statements must be structural statements in this sense of the word... a scientific statement must always be a statement about relations without involving knowledge or a description of the relata themselves. (Hjelmslev 1947:75, quoting Carnap 1928)

That quasi-mathematical perspective is what he found even in Saussure's 1879 Memoirs. That established him as a linguist, rather than a philologist:
At the outset, Saussure arrived at this view through a consideration of the IndoEuropean vowel system. As early as 1879, the analysis undertaken by Saussure of that system in his famous Memoire had shown him that in some cases the so-called long vowels can be conveniently reduced to combinations of a simple vowel plus a particular unit which by Saussure was symbolized by *A. The advantage of such an analysis over the classical one was that of furnishing a simpler solution, the so-called long vowels being discarded as such from the system, and of revealing a striking analogy between ablaut series which had been considered up till then as radically different. But interpreting, e.g. tithemi : thomos : thews as *dheA : dhoA : *dhA, this ablaut series reveals itself as fundamentally the same as that of derkhomai, dedorka, edrakon, which is equal to derk,: dork,: drkv Thus, *eA is to *oA what *er is to *or, and A plays the same role in the ablaut series as the *r of *drkv This analysis was carried out for internal reasons only, in order to gain a profounder insight into the fundamental system; and it was not based on any evidence available in the languages compared; it was an internal operation within the Indo-european system. Direct evidence for the existence of *A has later on been furnished by Hittite, but not until after Saussure's death. The unit *A has been interpreted, from the phonetic point of view, as a laryngeal. But it is well worth noting that Saussure himself would never have ventured any such phonetic interpretation. To him the *A was not a sound, and he took care not to define it by any phonetic properties, this being immaterial to his argument: his concern was the system only, and in this system, * A was defined by its definite relations to the other units of the system, and by its faculty of taking up definite positions within the syllable. This is expressly stated by Saussure himself, and this is the famous point where he introduces the term phoneme to designate a unit which is not a sound, but which may be represented or manifested by a sound. (1947.71-2)

11.09 Langue as Pure Form. Hjelmslev found more than one sense for langue in Saussure's pioneer work. He
...endeavored to disentangle, as far as it goes, the various layers or strata which can be observed in Saussure's meditations, and to lay bare what to my mind is the entirely new and really profitable idea in his work. This is, if I am not mistaken, the conception of language as a purely relational structure, as a pattern as opposed to the usage (phonetic, semantic, etc.) in which this pattern is accidentally manifested. [Cahiers Ferdinand de Sausswe.(1943:2.29-44)}

He believed others could only grasp Saussure's insights when 'langue is not identified with pure form, but where language is conceived as a form within the substance, and not independent of the substance' (PTL 1947:70). This was a fault he found in the Prague School's definition of Phoneme. Such a view did not agree with the last line of the Cours:

344 / GENERAL LINGUISTICS Linguistics has as its unique and true object la langue envisaged in and for itself. Where de Saussure only spoke of la langue as form generically, Hjelmslev will list its specific logical forms. Where Saussure assumed la langue was a single form unifying two substances (amorphous ranges of experience and sound), he posits two: content-form and expression-form. 11.10 Logical Implication. This perspective identifies part of the unity of Unified Science: independent of the content of particular sciences, it involves an implicit definition of description and explanation, and the relation between them. Hjelmslev prefers an intra-, not an extra-linguistic account. Its procedure relies on formal logic, particularly logical implication. He takes the form of all scientific theorems to be the 'If., then' expression of implication, or transposable into it. Such theorems, he says, assert only that 'if a condition is fulfilled, the truth of a given proposition follows' (PTL 14). 11.11 Valid vs. True. This ignores the distinction between valid and true. That difference is obvious in the rules defining Material Implication in Symbolic Logic. If that's IE, I'm an allophone is a valid2 Material Implication. Strict Implication, where conclusions involve causal connections (in his terms, appropriatemore or less 'verifiable in the real world'), involves factors irrelevant to formal logic's definition of truth-conditions. Formal logic only studies valid connexity among assertions relative to each other. The logician accepts truth-values of individual assertions as presented by a science, or assigns them arbitrarily for computation, as in Columns A-IV. That formally valid conclusions can be nonsense shows that a different point of view is involved. Total population divided by the number of families says 2 parents + 2.5 children are the average family, validly - but inappropriately. 11.12 Arbitrary vs. Appropriate. Hypotheses can be formed on the basis of theorems. But hypotheses (as Hjelmslev uses the term) are subject to verification while theorems are not. He discusses neither axioms nor postulates, since he finds them more primitive than linguistic theory. So glossematic theory is arealistic (as Firth said) by being arbitrary, realistic in its appropriateness (as in Firth's 'renewal of connection'). Its aim is 'self consistent and exhaustive description':
Such a self-consistent and exhaustive description leads to what is usually called a knowledge or comprehension of the object in question. (PTL 15)

Description and explanation are thus identified, or the difference neutralized, on Hjelmslev's premises (cf. 1961:78f., 1947:75, where description of substance presupposes description of form). It was in this sense that it was suggested above that a description is an implicit explanation.

11 Louis Hjelmslev / 345 11.13 Text as System-and-Process. The procedure must be illustrated in analyzing a text. Any text is only an instance of the System-and-Process organization the theory assumes, and any analysis is only an instance of the method the theory dictates. If both are appropriate, we can construct any possible text in the language analyzed, and extrapolate that analysis/synthesis to any language whatever (PTL 1961:16-17). That is why all conceivable possibilities must be foreseen, and why Hjelmslev does not offer either a practical discovery procedure or the theory demanding it, only a prefatory discussion of what the theory will have to be like (PTL 17). 11.14 Principle of the analysis: Hierarchic relations. Column A is logically faultless but cannot not handle contingencies like those which come up when we actually toss real coins. Saussure said the investigator's point of view creates his object; Hjelmslev notes that naive realism assumes there is an object to partition. But for science as he conceives it, there are no things: science discusses (Firth says 'makes statements about') intersections of relations of three principal kinds. These relations are Interdependence (both terms presuppose each other), Determination (only one term presupposes the other) and Constellation (terms are compatible, but neither presupposes the other). Subdesignations for these relations are appropriate in System and Process: Interdependence
(Two CONSTANTS)

Determination Constellation
(ONE CONSTANT, ONE VARIABLE) In (TWO VARIABLES)

Process: a BOTH-AND
function

Solidarity

Selection

Combination

In System: an EITHER/OR
function

Complementarity

Specification

Autonomy

Since parts of a text 'have existence precisely by virtue of these dependencies', Analysis is formally defined 'as description of an object by the uniform dependencies of other objects on it and on each other.' (PTL 28-29) Hjelmslev calls the Aristotelian category Relation a function and a term in a relation, a functive. In a hierarchy (a class of classes), a functive can itself be a function. If it is not, the Analysis terminates in an entity. 11.15 Translating Metalinguistic Terms. Compare this terminology with (a) Bloomfield's free/bound, endo-/exocentric distinctions, (b) Hjelmslev's use of the notions of Constant and Variable, and (c) a Syntactic Structures-type of Chomskyan tree.

346 / GENERAL LINGUISTICS (a) In Poor John runs away, Bloomfield would call Poor, John, run and possibly way 'free forms'. Bound forms are -s with runs and a- in away. The constructions Poor John and runs away are endocentric: if John and runs are obligatory heads, Poor and away are respectively optional expansions of them: Poor John as a construction behaves syntactically like its member John, and runs away like its defining member, runs. The construction John runs is exocentric, because its syntactic behavior is autonomous, not like that of either member, John or runs (Bloomfield 1933. Ch. 10). (b) Expressed as a relation between constants and variables, heads and free forms are constants (presupposed but not presupposing); optional expansions and bound forms are variables (presupposing, but not presupposed). The relation between expansions independent of each other like Poor and away (for which Bloomfield has no designation) is between two variables. Exocentric constructions have no Head: the grammatical behavior of John runs is not in parallel distribution with either John nor runs. That relation is a relation between two constants. (c) In a slightly different example, but one that can highlight some problems in Hjelmslev's approach, consider this display of the sentence The Romans destroyed Carthage, where each categorial constituent of Sentence (labelled S1) has its own numerical identification:

In Hjelmslev's terms, S1 (Sentence) is a function, whose functives are NP (Noun phrase) and VP3 (Verb phrase). But (2) and (3) are not only functives of (1), but are themselves functions. The functives of (2) are d6 (determiner) and N7 (Noun), while the functives of function (3) are the functives V4 (Verb) and NP5 (Noun phrase), and (4) and (5) too, are both functives that are functions. While (6) and (7) are functives in the function between the classes (determiner) and (Noun), the functives of (6) and (7) are entities (The and Romans), which are not functions, as this analysis presents them.
2

11 Louis Hjelmslev / 347 (4) as a functive is on a par with (6) or (7), a function whose functive is an entity, destroyed. But NP5 is a function whose functives are also functions on a par with (4) (6) and (7): its functives are the entities 0 and Carthage. Hjelmslev calls the relation of a functive to its function derivation. In that relation, the functive is called a derivate. S is not presented as a derivate, but every other functive is. Porphyry's Tree has Substance as supreme genus, from which other genera and species (functives that are functions) are derived, until derivation terminates in individuals (entities). Others use labels like class, set, or category vs. subclass, subset, or subcategory: these are relations (functions) between terms (functives), terminating in elements, individuals, or entities. Firth's relationship of exponence of a category shifts the criterion from member of to manifested by, a point of view which is one reason for saying morphemes cannot be composed of phonemes. Degrees of derivation can be assigned on such trees. Assigning grade 0 to S1, to which derivates show a dependence, we compute (2) and (3) as both degree 1 derivates; (6) and (7) are degree 2 derivates, as are (4) and (5). But (8) and (9) are degree 3 derivates, compared to (6) and (7). 11.16 'What is a' and 'What functions as a'. Labels for different degrees of derivation, as well as for derivations on different normslike exponence should correlate. Hispanus, for instance, distinguished between Subject-asSubject or Predicate-as-Predicate ('functions') and what-is-Subject or what-isPredicate ('functives') as Hjelmslev distinguishes among function, functive, and entity. Or, as we shall see, Fillmore noted that grammatical Subject and Object are not appropriately identified by immediate derivation of NP2 from S1, compared to the immediate derivation of NP5 from VP3, or NP5,s derivation from (or, as Hjelmslev calls it, dependence on) S1. This kind of tree-display has only two dimensions (like circular figures). To include multidimensional factors in a two-dimensional representation is a challenge. Alternatively, the differences can be incorporated into conventions for reading trees or other representations, which are not explicitly represented in the models, as in columns I-IV. Bloomfield noted the importance of the convention that presented signs in the order of speech, in his account of the development of writing. 11.17 Exclusion. Define system as a finite number of units, in a defmable totality, constrained in a finite number of ways. To paraphrase Hjelmslev, it would seem a priori reasonable to assume that the description of a System like I-IV vs. A, should answer about the Structures derivable within it: 'Given X, what does it demand, exclude or tolerateV. A structural description should do that. But there is no provision for exclusion in Hjelmslev's Interdependence, Determination, or Constellation. It might be argued that it is there implicitly, since exclusion is the opposite of demand. In The Romans destroyed Carthage, The Romans is neither obviously endocentric (determination) nor exocentric (interdependence): with Romans

348 / GENERAL LINGUISTICS as constant and the as variable, it is endocentric or determination; with both the and Romans as constants, it is interdependent. In English, the absence (Firth's exponence') of the category determiner with Carthage could be described as a subcategory proper noun demand for the absence of an article, or exclusion of articles. Hjelmslev is specifying de Saussure's generic/o/ra in logical terms, but he does not deal with a standard logical relation: Constellation does not distinguish exclusive disjunction (where only one option is valid) from inclusive disjunction (where either of two options are valid). In this example, Poor or away are inclusively disjunctive (either can freely occur). No example is given of what Firth's Colligations or Collocations where one form might exclude the other. 11.18 Langue, Text and Context De Saussure insisted on the centrality of la langue (not parole or langage) for a linguistic discussion of language and Hjelmslev presents text as the sole given. Missteps in linguistic study result from the admixture of concerns like logic, message, or nonlinguistic consequences other than System determining text as Process. For some linguists, the Prolegomena clarified the dimly perceived unity of what they had been doing; for others, its degree of abstraction foredoomed Glossematics. A simple example like Poor John ran away suggests that one root of both reactions correlates with willingness to entertain different degrees of abstraction. It also is a function of what is considered relevant in linguistic analysis. Is analysis to terminate in categories, subcategories, or entities? Are there demonstrably linguistic criteria omitted in this analysis? Are Glossematics' limitations in its theory, or in its Porphyrian lineage? Hjelmslev preferred his verbal approach to Uldall's algebraic formulations and did not use the tree representations common in linguistics since Chomsky's 1957 Syntactic Structures. 11.19 Signs and figurae. Signs have been said to represent, or be the exponents of, or to realize content. Sign expressions are readily analyzed into nonsign components like phonemes, clusters, syllables, or phonematic units. What, if anything, is the composition of content? As formed in the Porphyrian Tree, a Western logical tradition defines all but primitives vertically through the genus and species to which they belong, and horizontally through their peculiar way of belonging to a class. Man is defined generically as a rational species of the genus animal. Specific differences proposed as peculiar to man have been the ability to laugh, or as 'the only animal that can blush and needs to* (Mark Twain), or as 'the ungrateful animal (Dostoyevsky). Descriptive (or accidental) definitions include the negative Jeatherless biped" or the positive anthropological description, 'the tool-making animal. Dissecting men phonetically isolates the components [m] [e] and [n]: Hjelmslev calls these expression figurae. Analyzed morphologically, we distinguish root and infix, or base and phonetic modification, with contents

11 Louis Hjelmslev / 349 like man and PLURAL; we can isolate some sense-elements he calls content figurae from the definitions given above such as:

1120 Meaning is Contextual. Essential to Hjelmslev's conception of content is that all meanings are contextual-including the content of the term content itself in his own vs. de Saussure's use of it~and that absence of meaning is not lack of content:
If we think without speaking, the thought is not a linguistic content and not a functive for a sign function. If we speak without thinking, and in the form of series of sounds to which no content can be attached by any listener, such speech is an abracadabra, not a linguistic expression and not a functive for a sign function. Of course, lack of content must not be confused with lack of meaning: an expression may very well have a content which from some point of view (for example, that of normative logic or physicalism) may be characterized as meaningless, but it is a content. (PTL.49)

Content is relatively determined, e.g. John Ciardi: nine is the number of fingers you have after stopping a hole in the dike; Ambrose Bierce: edible is good to eat, and wholesome to digest, as a worm to a toad, a toad to a snake, a snake to a pig a pig to a man, and a man to a worm. These suggest differences among meanings as content, sense or reference, paralleling Saussure's distinctions among signification, value and content. Sense or content here can be the contextualization of a certain number of content figurae which sufficiently overlap for speakers to agree on the referent, even though senses are not identical. Registering an inventory of content figurae in systems and processes should be part of linguistic analysis as Hjelmslev sees it. Some English speakers link exit with its Latin function as a 3rd person, indicative, verb. For others, exit is just like other root-words, so they find no anomaly in a sign that orders Exit right, though they might in Adit right. The content figurae of the verbs loom and forebode are as distinct as those for appear and predict, but share a connotation, ominous, accounting for anomalies many found in A silence looms or A derrick forebodes over the campus in student publications. 1 2 Content: Purport, Schema, and Usage. Once granted contextualiz 11 ation and its implications, several consequences follow in Hjelmslev's approach. (1) A principle of generalization suggests to him the elimination of what he labels variants, like ram, ewe, mare, stallion, man, woman, boy, girl for invariants like he, she, sheep, horse, human being child of which they can be said to be composed. (2) By distinguishing forms of opposition or contextualization from what happens to fill such patterns, he defines a

350 / GENERAL LINGUISTICS linguistic schema as a pattern to be filled, while a linguistic usage focuses on fillers of that formal pattern. (3) Bound expression-level variants (like allophones) are varieties, determined by pattern, not phonetic similarity (unlike allophones). (4) This prepares for acceptance of purport as an unorganized, or as yet unformed, aggregate of what content figurae are actually about or might have connection with. This is like traditional matter as opposed to form: he adopts de Saussure's gaffe of calling it substance. Succinctly:
We thus recognize in the linguistic content, in its process, a specific form, the contentform, which is independent of, and stands in arbitrary relation to, the purport, and forms it into a content-substance.

This parallels distinction of expression purport from its expression-form and expression-substance. If purport is the indefinite range of articulatory possibilities, expression-form involves the number and type of distinctive oppositions, and expression-substance concerns particular articulations realizing it. In a language with three phonemic vowels, the contrastive pattern could be represented as /i/, /a/, and /u/, even though dialects consistently produce them phonetically as [e], schwa and [o]. Syncretisms are also found in content-expression relations, in German, the voiced vs. voiceless contrast in stops is neutralized in word finals: leiden ('suffer') and leiten ('lead') show the contrast, but phonetic [lait] in standard Leid ('sorrow'), Austrian colloquial Leut for Leute (people) or colloquial standard Leit' for Leiter ('ladder') do not. Some Latin masculines contrast -us vs. um for nominative vs. accusative, while neuters have -um for both, by syncretism. 1122 Catalysis. One term in an interdependent relation may be unexpressed. For instance, in a fragment of a Latin manuscript showing the preposition ad, the unattested term can be confidently supplied ad always govern the accusative case, no matter what the lexical form may be. Hjelmslev calls the process catalysis. The analysis of text in general, then, can be understood as a form of catalysis,
...through which the form is encatalyzed to the substance, and the language encatalyzed to the text. (PTL 96)

This is how codes are broken: a consistent principle of deformation conceals normally formed messages. We guess what that principle is and apply it to the text. If correct, we have encatalyzed form to substance. That is, we have imposed or discovered its structure, as in the case of Linear B: once it was assumed to be Greek, its anomalous purport was seen to be structured. So it follows that neither phonetic nor semantic substance can, of themselves, be defining of Language. Hjelmslev gives a taut summary of all this in his novel terminology:

11 Louis Hjelmslev / 351

...if the form is a language, we call it the linguistic schema. The variable in a manifestation (the manifesting) can, in agreement with de Saussure, be called the substance; a substance which manifests a linguistic schema we call a linguistic usage. From these premises, we are led to the formal definition of a semiotic as a hierarchy, any of whose components admits of a further analysis into classes defined by mutual relation, so that any of these classes admits of an analysis into derivates defined by mutual mutation. (PTL 106)

11.23 'Meaning' Systems. Games like chess fit this description. They differ from languages by functioning without the need of two planes (expression [a pawn's shape] and content [a pawn's moves]) which are not isomorphic: when there is a one-to-one correspondence between expression and content, Hjelmslev calls it a symbolic, not a semiouc system. For example, flag semaphores, Morse code, or the language of flowers are symbolic systems; English is a semiotic system. There are further differences: languages involve, not merely a semiotic system, but a connotative semiotic, distinct from the denotative one. In a denotative semiotic, no plane is a semiotic: flag systems and Morse code do not, like languages, provide for the affective use of expression figurae. To say that languages permit the use of denotative expressions for affective purposes is to define what Bloomfield called connotations in another way. 1124 Cognitive Systems. Symbolic systems make only denotative (cognitive) distinction. Morse . . . - - - ... stands for SOS ('save our ship' in English). It cannot connote a stress on our, if one of a convoy pleads for special consideration. Our is not even involved when SOS is verbalized in other languages. Within a semiotic denotative of cognitive distinctions, meaningful elements can be used as meaningless bearers (expression figurae) of connotations absent from the denotative convention. Length is not phonemic in English but can connote strong emotional 'content'. From distinctions like these, the langue-like centrality of Glossematics emerges, since Hjelmslev saw no nonsemiotics that are not components of semiotics, and no object unillumined from its perspective, '... a stand from which all scientific objects can be viewed' (PTL 127). Any sign-system whatever can be translated into the system of ordinary language. (1961.46ff, 108ff; 1970.104). Whitfield's preface to Hjelmslev's Resume of a Theory of Language (1975) diagrams the relationships as:

352 / GENERAL LINGUISTICS

OBJECTS

SEMIOTICS'

DENOTATIVE
SEMIOTICS -NON-SEMIOTICS NON-DENOTATIVE SEMIOTICS

LANGUAGES AND TEXTS

DENOTATIVE SEMIOTICS that are non-Languages and non-texts

C0NN0TATIVE

SEMIOLOGIES

METASEMI OLOGIES META-SCIENTIFIC SEMIOTICS that are SEMIOTICS INTERNAL EXTERNAL S E M I O L O G I E S

not meta-semiologies INTERNAL M E TAs EMI EXTERNAL L I E S

So the degree of abstraction Glossematics requires by its immanent method is defensible (PTL 127):
Linguistic theory is led by an inner necessity to recognize not merely the linguistic system, in its schema and in its usage, in its totality and in its individuality, but also man and human society behind language, and all man's sphere of knowledge through language. At that point linguistic theory has reached its prescribed goal: humanilas et universilas.

1125 STRATIFICATIONAL GRAMMAR. Several themes found in Hjelmslev's Glossematics are developed in Stratificational Grammar. Its principal theoretician is Sydney Lamb, and given his concern for keeping continuity in technical terms, it could have been called Levels Grammar. A stratum is a natural or artificial layer; strata parallel levels above and below them, defining them and being defined by them. Layers of sedimentary rocks form geological strata, tide or temperature ranges are oceanic strata, educational or economic levels are social strata, and the process of formation or the resultant state is stratification. Firth's image of spectral diffusion makes a similar point: light can be examined artificially as discrete bands without being confused about its unitary nature. Stratificational Grammar seeks to isolate elements, but

11 Louis Hjelmslev / 353 reintegrate the contribution and interdependence of linguistically distinct levels of analysis to the integrity of text. 1126 Stratificational Texts. The most accessible sources are Lamb's 1966 Outline of Stratificational Grammar, Lockwood's 1972 Introduction to Stratificational Linguistics and the 1973 Readings in Stratificational Linguistics edited by Makkai and Lockwood. In Readings, John Algeo's Stratificational Grammar (4-11) is a graceful and intelligible overview. Lamb's 1971 Crooked Path of Progress in Cognitive Linguistics accepts Ilah Flemming's 1969 assignment of four stages of adjustment within the theory. In the bibliography of Makkai's Readings, Lamb marks his unpublished 1957 thesis on Northfork Mono Grammar as stage 1 and a 1962 version of the Outline of Stratificational Grammar as stage 2. Stage 3 opens in 1965 with Kinship Terminology and Linguistic Structure, while his 1966 Outline of Stratificational Grammar starts stage 4. Accounts of applied and theoretical developments have appeared in the Forum of the Linguistic Association of Canada and the United States (LACUS), published yearly since 1973. 1127 The Stratificational Approach. Following Algeo's order of presentation, Stratificational Grammar (SG) is said to study language as consisting of a number of strata or levels, and finds other grammars deal with it in simpler terms, e.g. by describing morphemes as consisting of phonemes. In SG, the relation between the morpheme and phoneme levels is not membership, but realization (cp. Firth's exponence). Although recognition of levels entails stratification, other grammars can be inexplicit about how many strata they recognize, what relations within strata are, and how relations between strata are defined. The number of strata required has varied in SG (cf. Lamb 1983:189 ff). In 1979, Lamb and Gleason recognized six strata in three areas: semology: hypersememic sememic grammar, lexemic morphemic phonology: phonetic hypophonemic 1128 'Higher' and 'lower' Strata. The distinction of strata as higher or lower expresses a diffuse agreement that the linguist's task is to relate semantic contents and phonetic signals. So highest involves meaning, and lowest, phonetics. Intermediate levels (just like genera and species) have both upward relations (by being defined) and downward (by defining) relations. But the highest must define without being defined, the lowest must be defined

354 / GENERAL LINGUISTICS without defining: in Hjelmslev's terms, there are functives which are also functions, but they differ from functives which are only entities.3 11.29 Strata! Composition, Distribution and Function. Each stratum is described in terms of its elemental composition, distribution and function. Elemental units of composition are genetically emes, designated by an -erne compound label appropriate to each level on which they function as distinctive features of meaning or sound. In expression, some 12 to 15 hypophonemes (phonic features like plosion, spirancy, or nasality) are mentioned. Content emes are much more numerous, e.g. hypersememes like entity, process, animate, abstract, human, male, female, etc., comparable to Hjelmslev's invariants he and sheep for their variant ram. Kinship terms, or elaborate forms of polite and familiar address in languages show the value of this sort of componential analysis. 1130 Interstratal Relations. Relations between strata can be illustrated in idioms analyzed into sememes, realized on lower strata as phrases (bull in a chinashop, put up with), complex words (refer, housewife) or single morphemes (fox, pup), like the sign in de Saussure's twin-sided definition of it, sememes cannot be divided mto segments without losing all or some of their meaning: so a sememe is the smallest integral unit of meaning. The lexemic stratum handles much of what we saw in Chapter III as surface syntactic relations among free forms. The erne here is a lexeme, the smallest unit that forms syntactic combinations, realized on the next lower stratum as a morpheme (bull, put, with, plural s) or morpheme constructions (housewife) whose internal construction does not involve syntactic relations. Lamb's morpheme compares in size to what had the same label in Chapter 3 above, but is more like a lexically conditioned allomorph (go and the wen of went, or the plural s of hats vs. the en of oxen). His phoneme resembles post-Bloomfield morphophonemes, but is close as well to Firth's polysystemic phonematic unit: the n of an, because prevocalic and never preconsonantal, must be a different Lamb-phoneme from the n of than. Although different in distribution, both n's have the same realization (cp. Firth's exponence) and so are identical on the hypophonemic stratum. A stratum's function defines its emic status; a stratum's composition is its inventory of emes; the distribution of emes is stated in tactic rules. Realization rules connect strata with those higher or lower than themselves. For example, the morpheme blue is related to various lexemes above it such as bluebird, skyblue, blues, and simple blue, as well as to the phonemes below it, e.g. /b/, /1/, /uw/, in the symbols used in Chapter II. 1131 Stratal Realizations. Realization rules are of three kinds: (1) a pattern of alternation, connecting upward to a higher stratum, e.g. accounting for conditioned alternation; (2) a lower alternation pattern, connecting downward to a lower stratum, e.g. accounting for free variation), and (3) the

11 Louis Hjelmslev / 355 sign pattern, connecting downward to a lower stratum. The totality is connected top and bottom to nonlinguistic reality, at the top through thought, and at the bottom through vocal-acoustic phenomena, (cp. Firth's renewal of connection with experience, and the medieval modes of being, signifying, and understanding). 1132 Text The totality of strata is text. A text exists on all strata simultaneously. Each stratum may (but need not) have its own tactic rules, as each may have a different structure, providing different ways of regarding the same text. The semological stratum is like a network where a unit has many connections, e.g. To be safe on the Fourth, don't buy a fifth on the third, a text that few nonAmerican native speakers of English could analyze beyond the obvious grammatical level. Grammatically, text is representable in a constituent tree, phonologically, as a string or chain of elements. But text is all these things at once: network, tree, and string. Text is a complex of relations linking man to bits of experience, its meaning, and to sound waves (or in Pike's more suggestive analogy, can be viewed as particle, wave, or field, like entities in physics). So a pair of texts may be identical except for one stratum: the texts, The sun's rays meet and The sons raise meat differ lexically, and grammatically, but overlap on the phonological level, a phenomenon called homonymy by some {amphiboly by Hispanus following Aristotle). Bloomfieldians dealt with signals, not what was signalled, so the explanation of homonymy, rather than its description, was a problem relegated to some nonlinguistic discipline. Paraphrasing an ancient example, the phrase his picture can have to do with a picture of him, a picture he produced or a picture belonging to him. Stratificational Grammar deals with homonymy as well as synonymy. Expressions like I'll miss you or allmishya and even his arrival, compared to he arrives and for him to arrive, are accounted for explicitly as phonological or grammatically different realizations of the same semological structure. 11.33 Arrangement vs. Process. Strata dispense with the need of process terms which empiricists find dubious in synchronic accounts. In SG, no linguistic item X ever appears, then disappears, is replaced by, or changes into, anything else: a semological unit WX (e.g. become dead) can be a unit x die on one stratum realized as bite the dust on another, unchanged and unchangeable. But Stratification finds nothing intrinsically wrong with process terminology. Linguists are free to create an imaginary time dimension along which they move units, or an imaginary space for storing things, describing only their respective arrangements. Both item-and-arrangement (IA) and item-and-process (IP) ways of talking about linguistic structures are metaphors-one might be preferred over the other, but not on the basis of linguistic reality. Lamb, however, believes his stratal account is more than

356 / GENERAL LINGUISTICS metaphoric; he believes it is an analog of structural differences in the brain involved in speech production and perception. 1134 Text vs. Sentences. Among the advantages claimed for Stratificational Grammar are: (1) SG does not suffer from the limitations of Phrase Structure Grammar which Transformational Grammar was supposed to amend, to make them adequate for dealing with natural languages. So SG is an alternative to TG on that basis; (2) SG aims at text, not sentences alone: it is concerned with anything that has that kind of formal unity, (3) SG claims to be not only a model of the abstract system behind production and comprehension, but a model of the process itself: an analogical model for the production and comprehension of speech; not only how text is produced, but how it is understood. Expressed in Chomskyan terms we will examine shortly, SG claims to model competence as well as performance. In terms used earlier, it should therefore claim to present Language Itself (a) as all options accessible to speaker(s) or hearer(s), (b) as those options speaker(s) chose and those excluded, and (c) as (i) hearer's account of (ii) speaker's choices. Lamb's Crooked Path labels his work cognitive (Hjelmslev's denotative semiotic). So some of the uncertain singulars vs plurals in the last paragraph may not be needed, if one kind of connotative semiotic is excluded, e.g. personal, instead of the institutional connotations Bloomfield (1933.149 ff) discussed. The difference involves what is understood to pertain to competence vs. performance, both expressions ambiguously interpretable as referring either to group or an individual's capacities and activities. 1135 Group vs. Individual Abilities. It also has to do with the number of levels of analysis relevant in a cognitive theory, levels Lamb calls strata4. following Hjelmslev. For example, the words Irish or Italian, Mick or Dago may denote identically, but connote differently, for speaker or hearer. But knowledge that Mick and Dago are pejoratives is part of competence in the English language. A speaker's choice or a hearer's interpretation of either pair may be acts, part of performance. The distinction can be correlated with behavior, but cannot be accounted for in Bloomfieldian linguistics as Lamb presents it. 1136 Strata and Elements. In that approach, fie and vie establish /f/ and /v/ as phonemes. If wife's and wives have the same free morpheme base and differ only in bound morphemes of possession and plurality, how is the otherwise distinctive contrast of /f/ and /v/ to be accounted for? Different answers have been proposed: a morphophonemic level between morphemes and phonemes (Hockett 1961.27-53); distinguish morpheme and morph; speak of morpheme1 aod morpheme2; or call morpheme1 a lexon and a morphologically (lexically) conditioned morpheme2 simply morpheme.

11 Louis Hjelmslev / 357 If the elements of morphemes are phonemes, there are only two levels or strata to describe, and only one source of generalization or explanation. Phonology as a higher level can be said to explain a lower level that phonetics describes: postulating /p/ because of its contrast with /b/ explains why [p], [ph], [pw], [py] etc. are not phonemic differences in English. If a level of morphophonemes is added, there are three strata to describe, and two sources of explanation. As expressions increase in complexity, the need for explicit terms with refined coverage becomes clearer. Saussure's distinction of negative value, positive signification, and positive-negative content demonstrate how vague a single term like meaning is: some expressions as isolates can be described as overt signals of a signification (love), others of covert values (I love vs. you love), and others overtly contrastive (he loves vs. I or you love). These distinctions remind us that Bloomfield's definitions of simple, complex and compound forms concern signals, not what is signalled. They may help clarify Hjelmslev's assertion that lack of meaning (signification?) is not the same as absence of content, since every linguistic element is defined by its value in any structural approach. For instance, one English equivalent to the Latin single, free, complex, form amabatur is he was being loved - four free forms [one simple (he) and three complex (was, be-ing, and lov-ed)]. The signals -f-and he coincide in signification, as do am- and love. But -b- covers both was and -ing, Latin -ur alone correlates with passive, where English uses the discontinuous signals be...-ed. Without a fourth level, where each element of content is represented discretely, the signals can be described explicitly, but not explained explicitly. Anyone competent in the two languages intuits the elementary difference. But Linguistics requires explicit technical terms or rules of interpretation, applicable to the pair, to pick out where contrasts in Language (like Column A) are realized or nonexistent in a language (like Columns I-IV). That is, from one gender System, where English demands an overt expression in Process as he, she, or it, Latin Process tolerates ambiguous expression of the Systemic options in -t-. Saussure could say that amabatur and he was being loved may share signification, but their values differ, so their content does as well; Firth could have applied identical colligations to System, but have different exponence in Structure; Hjelmslev distinguished Interdependence in System as complementarity ('either-or') and Determination in Process as selection, one of four terms central to Bloomfield's account of syntax. (1933.164-169 et passim) 1137 Examples vs. Formalization. One complaint about Hjelmslev's Prolegomena was its lack of examples. Works claiming a basis in his point of view (e.g. Togeby 1951) have been received cautiously by Glossematic adherents and critics alike. But when examples of Stratificational analysis are published, its peculiar notation occasions complaint. This is based mainly on its unfamiliar stratal coverage, the patience needed to master the notation and

358 / GENERAL LINGUISTICS then verbalize it by isolating elements and interpreting their connections above and below strata. The problem is not peculiar to SG, as difficulties with previous notation show. It is one thing to welcome a disciplined treatment of a complex matter, and another to so master its presentational peculiarities as to find it as intelligible as ordinary language. There are few native speakers of a new metalanguage: if novel insights are obscure, or terminology discontinuous with formulations it promises to clarify, it can be rejected as just a new jargon. 1138 Stratificational Notation. Stratificational Grammar developed predecessors' terms but has its own system of notation for making linguistic statements. Three such systems are currently available: Algebraic notation (like Schleicher-Sapir's older one for morphology, Chomsky's newer S -> NP, etc. for syntax); rigorous prose (like Hjelmslev's novel vocabulary, for which most need ordinary language equivalents or examples to judge its relevance); or circuit diagrams such as Lamb uses. The diagrams are not an indispensable part of the theory, but for now, are characteristic of its perception and presentation of linguistic relations, some of which are not treated explicitly by others. So Geoffrey Sampson's 1970 stratificational study of English numerals has advantages as a sample at this stage of our study: it exemplifies some strata and their relations, SG concepts and mechanisms; Bloomfield said numbers denote without connoting, so as a Hjelmslevian denotative semiotic it may eliminate a stratum; it is what Firth could call a restricted language, but its exclusion of a phonological stratum raises doubt about the validity of the notion restricted language; it involves a small well-defined subset of semantic units, while the higher stratum is ill-defined in nature and extent. The subset can be well-defined because its numerical elements are structurally defined independent of any denotation. 1139 English Number Stratification. Sampson presents the restricted language of English cardinal, ordinal, and fraction denominator (reciprocal) expressions as represented on the sememic, lexemic, and morphemic strata of SG. He excludes a hypersememic stratum since ...such a stratum, being composed largely of the speaker's knowledge of the world, should be considered as outside linguistic structure. I therefore regard English as having five strata, from top to bottom, the sememic, lexemic, morphemic, phonemic and hypophonemic strata. (45) Sampson excludes a phonological stratum because he could not deal with numericals at that level without an analysis of English phonology as a whole, and there is none at present (36). His analysis is given in graphic and algebraic form (samples appended), and discussed in more ordinary terms.

11 Louis Hjelmslev / 359 He justifies SG in terms of general linguistic interest. He assumes traditional grammar, older linguists, and current consumers want accounts that even-handedly assist us both in producing sentences and in analyzing what others say. He finds Transformational Grammar does not address that problem, and so distinguishes between a generative and a communicational description. The theoretical contrast is neither one of conflict nor contradiction, just disagreement about what is worth doing, commonly called relevance. (7-12) 11.40 Generation and Communication. For communication, descriptions must be reversible; for generation, they need not be. If structure is order and constituency of elements, then different structures can be assigned at different levels by Tactic Rules peculiar to each.4 Sampson suggests that SG does that on levels called strata, while TG assigns it in the Base, then alters or interprets it at levels called components (13). SG notational changes should incorporate more inclusive insights, and Sampson presents a neat account of one such innovation to show how relations between strata can either be upward, downward, both, or neither. Letters can be assigned to the four sides of a diamond, from which lines can be drawn to another stratum (19): Algebra of diamonds: (a) line to higher in Realization (b) line to higher in Tactics (c) line to lower in Tactics (d) line to lower in Realization 11.41 Exclusion. In passing, Hjelmslev notes that he did not treat an exclusion-relation. That is relevant to the empirical accessibility of syntagmatic vs. paradigmatic items, and to the determination of what object linguists might study: I did not aim in the present paper to provide a mechanism for eliminating branches from the decision-tree of choices in applying the rewrite rules, before this tree is complete (which is what is required for a performance model). (31) 11.42 Morphemic, Sememic and Lexemic Strata. Numerical expressions at the morphemic stratum are familiar to us, e.g. five, fifth, fifteenth, nineteen fifty-fifths). Less familiar is the representation of the identical data on the sememic and lexemic strata, of which it is asserted that their emes are the system from which we as speakers choose to form structures, and those in terms of which we as hearers analyze the options others have elected.

360 / GENERAL LINGUISTICS The sememic structures the grammatical system permits are strings of: content units (the integers 1 to 10); operators sum-of, times, power-of, ordinal, reciprocal, and left- and right-brackets (American parentheses). Parentheses lack content, but indicate constituency relations among elements in structures by defining the domain of the operators. Sampson points out three factors that call for comment: the use of portmanteau units, the dubiety of string representation of semantic content, and the fact that trillion is the upper numerical limit of his presentation. The last is easily remedied by rewrite rules; the second is more serious as implying linearity or temporality instead of neutral order-as-dependency in networks; the first seems to favor too much, rather than too little structure, but simplifying it is easy enough. Hundred is a portmanteau for what could be represented as HN/ten times ten/ (superscript HN is read Hypersemon). Units at the top of an alternation pattern are named after the stratum above it, though Sampson excludes that level as outside linguistic structure, since it involves knowledge of the world rather than language, as mentioned above. An often used example of a portmanteau item is French au instead of the particle a and article le. Its occurrence is determined by French grammar, not by facts about the world. What it abbreviates or leaves out is readily recoverable on the basis of linguistic, rather than other kinds of knowledge. This can be appreciated by contrasting abbreviation with ellipsis. An English portmanteau baptize can be sememicallypu/ into water put water on. The difference is abbreviation of facts about the world users may or may not know. Bloomfield describes sentences like / couldn't have helped you, even if I had known her to be a friend of his, and neither could you as involving zero anaphora (1933.249 ff): the zero-forms, supplied on the basis intralinguistic convention alone, could be neither could you have helped me, even if you. had known she was a friend of his. 11.43 Strata Mutually Define and Are Defined by each other. Intermediate strata, e.g. social levels, define those below them and are defined by those above. On the lexemic stratum,
A string of downward sememes has as its realization a string of upward lexemes. The upward sememes ""/power-of, times, sum-of, (,), ten-z/ are determined: they have no lexemic realization, and their occurrence is in all cases fully determined by the semotactics. Several downward sememes have alternate realizations: for instance, "/ten/ is realized as will be realized as zero when used as a multiplier of ten. In all cases choices will be determined by the lexotactics, which accepts some strings of lexemes but not others. (55)

French cousin, cousine or German Vetter, Base might be analyzed as lower lexemic UL realization of higher downward semes DS involving something like /son, daughter, uncle, aunt/, and the latter perhaps into upper semes US for vertical and lateral relationships. The lexotactics of English do not allow the distinctions in cousin. There is no alternation pattern, i.e. no systemic English option from which structures like French cousin, cousine are chosen.

11 Louis Hjelmslev / 361 Similarly*.

In the lexotactics, the lexemes ordinal and reciprocal must still be kept apart, although most instances of either will eventually be realized phonetically as [0], because the reciprocal of two, half, differs from the ordinal second (note however that, e.g. twentysecond is both ordinal and reciprocal); and the reciprocal of four may optionally be quarter rather than fourth)... (55)

11.44 Stratal Representation. Representation on the morphemic stratum is by morpheas. This SG term has already been compared to Firth's phonematic units and American morphophonemes and might be comparable as well to Praguian archiphoneme. Since Firth's unit can be represented directly by ordinary IPA symbols, while phonemes are only represented indirectly by them, it is not the systematic abbreviation phonemes can be taken to be. A Prague School archiphoneme is phonetic, but more importantly, a phonological opposition, often represented by a capital letter to stress its distance from phonetic fact. It stands for a factor common to several articulations opposable in general, but neutralized in a particular place in a language, e.g. nasality in bilabials, alveolars or velars. Morphophonemes have been variously presented as bundles of such factors or oppositions, also by capitals between curly brackets rather than phonemic slanted lines. At Stage II (1971) Lamb said 'a morphon is a component of a morpheme; and the term may be considered a shortening of morph(oph)on(eme). (Readings 21), where morpheme was distinguished from morpheme2 (c.f. Readings 146). Although a complete SG phonology of English has not been forthcoming, the details are not expected to be complicated:

Despite their different backgrounds and points of departure, the linguistic theories of Firth, Hjelmslev and Lamb 'translate' into each other fairly well.
Notes 1. Rudolf Carnap, Der Logische Aufbau der Welt (Berlin, 1928) quoted in Hjelmslev's 'Structural Analysis of Language, Studia Unguistica 1.2.75.1947.

3. Whitfield 1961 appends 108 definitions referring to 'other, explicitly premised definitions' (131-8). Many are mutually defining, i.e. structural definitions: Hjelmslev's Interdependence); some are

362 / GENERAL LINGUISTICS

definitions through genus, species, and specific difference (like rational animal for man : Hjelmslev's determination). That is why he called attention to his 'peculiar' use of terms like deductive and empirical. 4. That the resulting tactics (or arrangements) differ is clear from the non-conformity of morphemic and syllabic structures: basks is a single syllable; it could be analyzed into three morphemes, if is recognized as an old Scandanavian reflexive. Reading Hjelmslev, Louis. 1961. Prolegomena to a Theory of Language, tr. Francis Whitfield. Madison. University of Wisconsin Press, pp 144. _____1970. Language /An Introduction, tr. Francis Whitfield. Madison. University of Wisconsin Press, pp xiii 144 (Sproget) _____1975. Resume of a Theory of Language, ed., tr. Francis Whitfield. Madison. University of Wisconsin Press, pp 280. Whitfield, Francis J. 1954. Glossematics. In Linguistics Today. New York. WORD. pp. 250-258. Stratificational grammar Lamb, Sydney. 1966. Outline of Stratificational Grammar. Revised Edition. Washington, D.C. Georgetown University Press. Lockwood, David G. 1972. Introduction to Stratificational Linguistics. New York. Harcourt, Brace, Jovanovich. pp. 365. bibliog 351-358 [admirable text for self-instruction of linguists] Makkai, Adam and D. Lockwood, eds. 1973. Readings in Stratificational Linguistics. University of Alabama Press, pp. vi-331, bibliog 316-329. [Section I theory (1-116), application: SG with best foot forward: must reading] Supplementary reading Hjelmslev Fischer-Jergensen, Eli. 1943. Review of Hjelmslev's OSG. Nordisk Tidsskrift for Tale og Stemme. 7.81-96. _____1965. Louis Hjelmslev, October 3, 1899-, 1965. ActaLinguistica HafniensiaX.l-33. _____1967. Introduction, 2nd. ed. of Uldall's Outline of Glossematics. Travaux du Cercle Linguistique de Copenhague X. _____????. Trends in Phonological Theory. Copenhagen: Adademisk Forlag. [Chapters 7 (Glossematics) and 10 (Stratificational Theory)] Garvin, Paul. 1954. Review of Hjelmslev's Prolegomena. Lg. 30.1.69-96. Haugen, Einar. 1954. Review of Hjelmslev's Prolegomena. UAL 20.247-51. Hjelmslev, Louis. 1932-37. Rask, Rasmus Kristian (1787-1832). Ausgewuhlte Abhandlungen. Einleitung von Holger Pedersen. Kopenhagen: Levin und Munksgaard. _____1943. Omkring Sprogteoriens grundlaeggelsae of Louis Hjelmslev. Kobenhavn: B. Lunos. _____1949. Recherches Structurales. interventions dans le debat glosseinatique, publiee a occasion du cinquantenaire de M. Louis Hjelmslev. Copenhague: Nordisk Sprogog Kulturvorlag. Travaux du Cercle Linguistique de Copenhague. pp 307. bibliog 305-307. _____1968. Langue et Parole. Cahiers Ferdinand de Saussure 2.29-44.1959. also 69-81. _____1968. ProUgomines a une Thiorie du Langage. tr. du danois par un equipe de linguistes. Avec: Structure fondamentale du langage. tr. de l'anglais par Anne-Marie Leonard. Paris. Editions de Minuit. pp 231. _____1971. Prolegomenos a una theorla del Lenguaje. Version espanol de Jose Luis Diaz. Madrid. Gredos. pp 198. _____1968. Die Sprache. Eine Einfiihrung. tr. Otmar Werner. Darmstadt. Wissenschaftliche Buchgesellschaft. pp 183. _____1976. Principios de gramdtica general, tr. Felix Torre. Madrid. Gredos. pp 383.

11 Louis Hjelmslev / 363

_____1972. La Caligorie des cos; etude de grammaire generate. 2. Verbessert und mit den Korrekturen des Autors versehen. Auflage der Ausgabe Kopenhagens 1935-37. Mit einer bibliographischen Notiz von Eli Fischer Jergeneen. Munchen. Fink. Text franzosisch. _____1978. La Categoria de los casos: estudio de gramatica general, tr. Felix Pifiero Torre. Madrid. Gredos. pp. 345. bibliography 335-41. ____ 1976. Sistema Unguisaco cambio linguistico de Louis Hjelmslev, tr. Berta Pallares de R. Arias of Sprogsystem og Sprogforandring. Madrid. Gredos. pp 262. ____ 1959. Essais Linguistiques. Copenhague, Nordisk Sprog og Kultursvorlag. pp 271. bibliog 251 -81. Japanese translation. ____ 1971. Essais Linguistiques. Ratier Nourelle ed. Paris. Editions de Minuit. pp 284. ______ . 1972. Ensayos linguisticos. tr. Elena Bombon, Felix Torre. Madrid: Gredos. pp 358. Lamb, Sydney. 1966. Epilegomenato atheory of language. Romance Philology 19.531-73. [Review of Hjelmslev: contemporary with Lamb's Outline] Malmberg, Bertil. 1964. New Trends in Linguistics. Stockholm, pp 140-157. Martinet, Andre. 1946. Au sujet des fondementsde la theorfe linguistique de Louis Hjelmslev. Bulletin de la Sociiti Linguistique de Paris. 42.19-42. Sechehaye, Albert. 1909. Programmes et methodes de la Umguistique thiorique. Paris. Siertsema, Berta. 1954, 1965. A Study of Glossematics: critical survey of its fundamental concepts. 2nd edition, the Hague: Nijhoff. bibliog. 270-284. [readable preChomskian perspective] SpangHanssen. 1962. Glossematics. In Trends In European and American Linguistics. pp. 129-164. Togeby, Knud. 1949. Linguistics in Denmark: 1940-1948. Symposium .2.226-237. ____ 1951 Structure immanente de la langue francaise. Travaux du Cercle Linguistique de Copenhague VI. [Reviews: A. Martinet. Word 9.78-82.1953; M. Fowler, Lg. 29.165175.1953] ____ 1965. Louis Hjelmslev, Oktober 3, 1899-Maj 30 1965. Kobenhavns Vniversitets Festskrift November 1965.159-68. Trabant, Jurgen. 1970. Zur Semiologie des titeraristhen Kunstwerkes: Glossematik und Literaturtheorie. Munich: Fink. ____ 1975. Semiologia de la obra titeraria. version espanol de Jose Saez. Madrid. Gredos. pp 370. Uldall, Hans Jergen. 1957. Outline of Glossematics; a study in the methodology of the humanities, with special reference to linguistics. Part I: General Theory. Travaux de Cercle Linguistique de Copenhague. X,.1957. 2nd ed Copenhagen: Nordisj Sprog og Kulturvorlag 1967. pp xxii 87. TLCC. Whitfield, Francis J. 1956. Linguistic usage and glossematic analysis. For Roman Jakobson. The Hague. Mouton. pp 670-675. Stratifications! grammar Algeo, John. 1973. Stratificational Grammar. Journal of English Linguistics 3.1-7 and in Makkai, Lockwood. Chafe, Wallace. 1968. Review of Lamb 1966. Language 44.593-603. Fleming, Uah. 1969. Stratificational theory: an annotated bibliography. Journal of English Linguistics 3.37-65. [Stratified account of progress: readable on its own] Gleason, Henry A. 1964. The organization of language: a stratificational view. GURP 17.75-95. ____ 1968. Contrastive Analysis in Discourse Structure. QURP 21.39-63 and in Makkai and Lockwood, 258.276) Hajicova, E., and P. Pitha. 1968. Review of Lamb 1966. Prague Bulletin of Mathematical Linguistics. 8.71-76. [mathematically pointed; a quick: read, long thought] Hockett, Charles. 1958. A Course in Modem Linguistics. New York. Macmillan. ____ 1966. Language, mathematics and linguistics. The Hague. Mouton Series Minor 60. ____ 1968. Review of Lamb 1966. UAL 34.145-153. [Umb's work a noble scientific failure] Huddleston, Rodney. 1968. Review of Lamb 1966. Lingua 22.362-73. Lamb, Sydney. 1965. Kinship Structure and Linguistic Structure. American Anthropologist

364 / GENERAL LINGUISTICS

Pt. 2.37-64. (In Makkai and Lockwood, 229-257) ___ 1971. The Crooked Path of Progress in Cognitive Linguistics. Washington, D.C. GURP 24.99-123. (In Makkai and Lockwood 1973.12-33) ____ 1983. On Determining the Number of Strata in Linguistic Structure. Ninth LACUS Forum 1982. 189-203. Columbia, South Carolina. Hornbeam Press. Palmer, F.R. 1968. Review of Lamb 1966. Journal of Linguistics 4.287-95. Pike, Kenneth L. 1967. Language in Relation to A Unified Theory of the Structure of Human Behavior. Second edition. The Hague. Mouton. ____ and Evelyn G. Pike. 1983. Text and Tagmeme. Norwood: Ablex. Sampson, Geoffrey. 1970. Stratificational Grammar: a Definition and an Example. Janua Linguarum Series Minor 88. The Hague. Mouton. ____ 1980. Schools in Linguistics. London. Hutchinson, pp. 283. bibliog. 259-273. [Chapter 7: Relational Grammar: Hjelmslev, Lamb, Reich, pp. 166-186]. White, John. 1969. Stratificational grammar: a new theory of language. College Communication and Composition 20.191-197.

11 Louis Hjelmslev / 365

366 / GENERAL LINGUISTICS

Appendix 2. Part of Sampson's Algebraic Representetioa

THE NUMERAL SYSTEM