6. Louis Hjelmslev¹

6.1 Hjelmslev purports to offer neither a general survey of language and its types (like Sapir's and Bloomfield's) nor a general theory of linguistics (like Saussure's and Hartmann's), but a preparatory ‘prologue’ to the formulation of any ‘theory of language’. The Prolegomena (PT), his central book, published in Danish in 1943 and in English in 1953, proposes to stipulate in the broadest terms the conceptual layout for any such theory. His Resume (RT), circulated in a few typed copies in 1941-43 and eventually published in 1975, is a technical compilation of terms, symbols, definitions, rules, and notes. His ideas often build on Saussure's, but are, in an ambivalent way, more radical, digging for the roots while trying not to get dirty.

6.2 Like our other theorists, Hjelmslev declares his profound respect for language as a human faculty (2.8, 32; 3.1, 3; 4.2, 10, 82, 5.69; 13.22). ‘Language -- human speech² -- is an inexhaustible abundance of manifold treasures’ and ‘the distinctive mark of the personality’, of ‘home, and of nation’ (PT 3). ‘Language is the instrument’ whereby man ‘forms thought and feeling, mood, aspiration, will and act’. It ‘is inseparable from man’ and ‘all his works’, ‘from the simplest activities’ to the ‘most sublime and intimate moments’ during which the ‘warmth and strength for our daily life’ flows from ‘the hold of memory that language itself gives us’. ‘Language’ is thus ‘a wealth of memories inherited by the individual and the tribe, a vigilant conscience that reminds and warns’. It is ‘the ultimate and deepest foundation of human society’, but also ‘the ultimate, indispensable sustainer of the human individual, his refuge in hours of loneliness, when the mind wrestles with existence and the conflict is resolved in the monologue of the poet and the thinker’. ‘Before the awakening of our consciousness language was echoing about us, ready to close around our first tender seed of thought’ (cf. 3.3). Such praises might portend a mentalistic, phenomenological, or humanistic approach, but Hjelmslev offers nothing of the kind.

6.3 Again like our other theorists, Hjelmslev is stringently critical of ‘conventional linguistics’ (cf. PT 4, 5, 44, 65, 73, 79, 99; 13.4).³ He asserts that ‘the history of linguistic theory cannot be written’, being rendered ‘too discontinuous’ by ‘superficial trends of fashion’ (PT 7) (a view I hope to refute with this volume). In his opinion, ‘linguistics’ was ‘frequently misused as the name for an unsuccessful study of language proceeding from transcendent and irrelevant points of view’ (PT 80, i.r.). ‘Attempts to form a linguistic theory have been discredited’ ‘as empty philosophizing and dilettantism, characterized by apriorism’ and ‘subjective speculation’ (PT 7). ‘Until now, linguistic science’ has ‘remained vague and subjective, metaphysical and aestheticizing’, and relied on ‘a completely anecdotal form of presentation’ (PT 10). In this state of affairs, we might do well to ‘forget the past’ and ‘start from the beginning’ (PT 7). Instead, Hjelmslev elects to work ‘in contrast to previous linguistic science and in conscious reaction against it’, seeking ‘an unambiguous terminology’ ‘in linguistic theory’ (PT 37) (cf. 5.33; 8.40; 13.7, 15, 48).

6.4 Past failings are attributed to several obstacles. One obstacle was the ‘humanistic tradition which, in various dress, has till now predominated in linguistic science’ (PT 8) (cf. 8.36; 12.49). This ‘humanism’ ‘rejects the idea of system’, and ‘denies a priori the existence’ of any ‘integrating constancy’ and ‘the legitimacy of seeking it’ (PT 10, 8). Hence, ‘the humanities’ ‘have neglected their most important task’: ‘establishing’ their ‘studies’ as ‘a systematic, exact, and generalizing science’ (PT 9). The ‘most’ ‘humanistic’ ‘disciplines’, i.e., ‘the study of literature’ and ‘art’, have been ‘historically descriptive rather than systematizing’. They offer the justification that ‘humanistic phenomena are non-recurrent’, and thus ‘cannot, like natural phenomena, be subjected to exact and generalizing treatment’; and that ‘we cannot subject to scientific analysis man's spiritual life’ ‘without killing’ it (PT 8, 10). Their only method is either ‘a discursive form of presentation, in which the phenomena pass by, one by one, without being interpreted through a system’; or a ‘mere description’, ‘nearer to poetry than to exact science’ (PT 8f) (cf. 12.38).

6.5 Another obstacle was the ‘transcendent aim’ and ‘objective’ of many researchers, including ‘philologists’ (PT 6, 10). In this work, ‘the theory of language’ was often ‘confused with the philosophy of language’, including some modern ‘offshoots of medieval philosophy’ (PT 6, 77).⁴ Researchers would seek a ‘universal’ ‘system’, a set of ‘generally valid’ ‘types’, an ‘eternal scheme of ideas’, or a ‘construction of grammar on speculative ontological systems’ (PT 76f) (cf. 13.16ff). Or, they would try to ‘construct’ one ‘grammar on the grammar of another language’, e.g., ‘blindly transferring the Latin categories’ ‘into modern European languages’ (PT 75f; cf. EL1 125) (cf. 2.5; 3.50; 4.4; 5.24; 8.5; 9.25; 12.20f). ‘Such projects are necessarily foredoomed to miscarry’, lacking any ‘possible contact with linguistic reality’ (PT 76f).

6.6 A further and related obstacle was the tendency to treat ‘language, even when it is the object of scientific investigation’, not as ‘an end in itself, but a means’ ‘to a knowledge whose main object lies outside language’ (PT 4) (cf. 12.23). Here too, ‘language is a means to a transcendent knowledge’, ‘not the goal of an immanent knowledge’. For example, ‘language’ ‘was expected to provide the key to the system of human thought, to the nature of the human psyche’ (cf. 3.10ff; 5.69; 6.2; 7.10; 8.24; 12.17ff, 22; 13.10, 14). Or, ‘it was to contribute to a characterization of the nation’, to an ‘understanding of social conditions, and to a reconstruction of prehistorical relations among peoples and nations’ (PT 4f) (cf. 12.19, 91). ‘The main content of conventional linguistics -- linguistic history and the genetic comparison of languages’ -- was a ‘knowledge of social conditions and contacts among peoples’. Such research fails to ‘grasp the totality of language’, and incurs ‘the danger’ of ‘overlooking’ ‘language itself’ (cf. 2.5f). To be sure, ‘it is in the nature of language to be overlooked, to be a means and not an end’; ‘only by artifice’ can we direct a ‘searchlight’ on it (cf. 3.1; 4.2; 12.9; 13.1). ‘This is true’ both ‘in daily life, where language normally does not come to consciousness’, and ‘in scientific research’.

6.7 More recently, ‘science has been led to see in language a series of sounds and expressive gestures, amenable to exact physical and physiological description, and ordered as signs for the phenomenon of consciousness’ (PT 3f) (cf. 4.28, 32; 5.44; 6.54; 8.20, 22). Here, science is restricted to ‘the physical and physiological description of speech sounds’, which ‘easily degenerates into pure physics and pure physiology’ (PT 4). Or, science ‘has sought in language, through psychological and logical interpretations, the fluctuation of the human psyche and the constancy of human thought -- the former in the capricious life and change of language, the latter in its signs’. Here, ‘words and sentences’ are held to be ‘the palpable symbols of concept and judgment respectively’ (cf. 3.32, 36); and ‘the psychological and logical description of signs (words and sentences)’ leads to ‘pure psychology, logic, and ontology’. Either way, ‘the linguistic point of departure is lost from view’. ‘Physical, physiological, psychological, and logical phenomena per se are not language itself, but only disconnected, external facets of it’ (PT 4f) (cf. 13.5).

6.8 To offset all these misconceptions, Hjelmslev offers his ‘prolegomena’ to ‘a linguistic theory that will discover and formulate the premises’ of ‘a real and rational genetic linguistics’, ‘establish its methods, and indicate its paths’ (PT 6). ‘A true linguistics’ ‘cannot be a mere ancillary or derivative science’ (PT 5) (cf. 8.17; 13.9-20). It ‘must attempt to grasp language, not as a conglomerate of non-linguistic’ ‘phenomena, but as a self-sufficient totality, a structure of its own kind’ (PT 5f) (13.22). ‘Only in this way can language in itself be subjected to scientific treatment’. Hjelmslev sees ‘an immanent algebra of language’ as the ‘main task’ of ‘linguistics’ ‘whose solution has been almost completely neglected in all study of language’, apart from ‘a beginning in certain limited areas’ (PT 79f) (cf. 2.82; 5.86; 6.29; 7.40, 7¹⁸; 13.15).⁵ ‘To mark its difference from previous kinds of linguistics’, he proposes to call this ‘algebra’ by the ‘special name’ of ‘glossematics (from “glossa,” “a language”)’ (PT 80).

6.9 By centring linguistics firmly on language and ‘removing’ the ‘provincialism in the formation of concepts’ (PT 6), Hjelmslev expects far-reaching benefits for science at large (cf. 13.21f). Because ‘it is impossible to elaborate a theory of a particular science without an active collaboration with epistemology’, ‘the significance of such a linguistics’ can be ‘measured by its contributions to general epistemology’ (PT 15, 6). Just as Hjelmslev's own ‘presentation’ is ‘forced’ ‘into a more general epistemological setting’, ‘every theory is faced with a methodological requirement whose purport will have to be investigated by epistemology’ (PT 102, 11). Yet ‘such an investigation may, we think, be omitted here’. And the ‘terminological reckoning’ ‘to be made with epistemology’ is postponed for ‘later’, though he hopes that ‘the formal foundation of terms and concepts given here should make possible a bridge to the established usage of epistemology’ (PT 11, 31f). Besides, ‘the science of categories presupposes such a comprehensive and closely coherent apparatus of terms and definitions that its details cannot be described without its being presented completely’, so it cannot ‘be treated in the prolegomena of the theory’ (PT 101).

6.10 Hjelmslev feels ‘led to regard all science as centred around linguistics’ (PT 78) -- a popular aspiration (2.7f; 5.7, 84; 6.41, 53; 7.8; 8.16, 29; 12.6, 9, 12, 33, 64; 13.21, 59). He ‘supposes that several of the general principles we are led to set up in the initial stages of linguistic theory are valid’ ‘for all science’ (PT 80). His ‘basic premises’ ‘are all of so general a nature that none would seem to be specific to linguistic theory’ (PT 15). He hopes for a ‘universal applicability to sign systems’ or to ‘any structure whose form is analogous to that of a “natural” language’ (PT 102; cf. 6.48-55). ‘Precisely when we restrict ourselves to the pure consideration of “natural” language’, ‘further perspectives’ ‘obtrude themselves with inevitable logical consequence’ (PT 101, i.r.). ‘If the linguist wishes to make clear to himself the object of his own science’, he gets ‘forced into spheres which according to the traditional view are not his’ (PT 101f). For example, ‘the systematics of the study of literature and of general science find their natural place within the framework of linguistic theory’, as do ‘general philosophy of science and formal logic’ (PT 98, 102; cf. 6.54). This grand vision leads to an interesting tension in Hjelmslev's work. On the one hand, he is anxious to demarcate the borders and independence of linguistics and to centre it on language in an ‘immanent’ fashion (PT 19, 108, 127), assigning related issues to ‘the non-linguistic sciences’ (PT 78ff). On the other hand, his ambition to make linguistics the model science keeps him at some distance from language and entrains him in the transcendent theorizing he criticizes.

6.11 The scope is set as wide as possible: ‘a theory’ ‘must enable us to know all conceivable objects of the same premised nature’, and to ‘meet’ ‘any eventuality’ (PT 16). The ‘main task is to determine by definition the structural principle of language, from which can be deduced a general calculus’ (PT 106) (cf. 7.18). Such ‘a general and exhaustive calculus of the possible combinations’ would provide the foundation for ‘a systematic, exact, and generalizing science, in the theory of which all events (possible combinations of elements) are foreseen and the conditions for their realization established’ (PT 9) (cf. 6.11, 30, 33, 36, 38, 50, 63). ‘The linguistic theoretician must’ even ‘foresee all conceivable possibilities’ ‘he himself has not experienced or seen realized’, i.e., ‘those that are virtual in the world of experience, or remain without a “natural” or “actual” manifestation’ (PT 17, 106) (cf. 6.18f, 35; 9.8; 12.55f).

6.12 ‘Linguistic theory’ must also ‘seek a constancy which is not anchored in some “reality” outside language’, but which ‘makes a language a language’ and makes it ‘identical with itself in all its various manifestations’ (PT 8, i.r.) (cf. 4.71; 8.33; 13.57). ‘This constancy’ ‘may then be projected on the “reality” outside language’ -- ‘physical, physiological, psychological, logical, ontological -- so that even in the consideration of that “reality”, language as the central point of reference remains the chief object -- and not as a conglomerate, but as an organized totality with linguistic structure as the dominating principle’ (cf. 6.20, 38; 13.24ff, 57). The essential strategy would be to ‘search for the specific structure of language through an exclusively formal system of premises’ (cf. 13.54). And this search is just what Hjelmslev pursues.

6.13 In such a project, the notion of ‘empiricism’ is given a peculiar interpretation, one whereby Hjelmslev's ‘theory is at once clearly distinguishable from all previous undertakings of linguistic philosophy’ (PT 11) (cf. 7.85). On the one hand, ‘a theory must be capable of yielding, in all its applications, results that agree with so-called (actual or presumed) empirical data’. On the other hand, his ‘empirical principle’ makes no mention of data, stating only that ‘the description shall be free of contradiction (self-consistent), exhaustive, and as simple as possible’ -- ‘freedom from contradiction taking precedence over’ ‘exhaustive description’, and the later over ‘simplicity’. ‘Linguistic theory’ ‘can be judged only’ by these criteria: ‘a theory, in our sense’, ‘says nothing at all about the possibility of its application and relation to empirical data’ (PT 18, 14). ‘It includes no existence postulates’, ‘replacing’ them with ‘theorems in the form of conditions’ (PT 14, 21). Hjelmslev thereby resolves to make ‘linguistic theory’ ‘as unmetaphysical as possible’; it should shun ‘implicit premises’ and should not try to ‘reflect the “nature” of the object’ or rely on the ‘concept’ of ‘“substance” in an ontological sense’ (PT 20, 22, 81; cf. 6.28; 13.26).

6.14 It seems odd to find the existence of objects, the traditional recourse of realism, reckoned under ‘metaphysics’, a term usually applied to the ‘transcendent’, ‘supersensible’, or ‘supernatural’ (Webster's Dictionary). But this move abets Hjelmslev's plan to design theories in purely formal terms. He praises ‘the special advantage’ of avoiding any ‘recourse to sociological presuppositions which the “real” definition’ of ‘terms would necessarily involve’ and which would ‘at best’ ‘complicate’ ‘the apparatus’ and ‘at worst’ ‘involve metaphysical premises’ (PT 20, 89) (cf. 13.16f). For Hjelmslev, the ‘concept of sociological norm’ ‘proves to be dispensable throughout linguistic theory’ (PT 89), though I can't see how such a thing could be ‘proven’ at so preliminary a stage.

6.15 In place of ‘the real definitions for which linguistics has hitherto striven insofar as it has striven for definitions at all’, Hjelmslev recommends ‘giving a strictly formal’ and ‘explicit character to definitions’ and ‘replacing postulates partly by definitions and partly by conditional propositions’ (PT 21). A ‘theory’ ‘consisting of a calculation from the fewest and most general possible premises’ ‘permits the prediction of possibilities, but says nothing about their realization’ (PT 15). A reciprocity is proposed whereby ‘the object determines’ ‘the theory’ and ‘vice-versa’: ‘by virtue of its arbitrary nature the theory is arealistic’ and ‘calculative’; ‘by virtue of its appropriateness, it is realistic’ and ‘empirical’ (PT 15, 17). ‘Arbitrariness’ means here that ‘the theory is independent of any experience'⁶ and only a means for ‘computing the possibilities that follow from its premises’ (PT 14). ‘Appropriateness’, on the other hand, means that ‘the theory introduces certain premises concerning which the theoretician knows from preceding experience that they fulfill the conditions for application to certain empirical data’ (cf. 7.10, 7⁷). Hjelmslev goes on to argue that ‘empirical data can never strengthen or weaken the theory itself, but only its applicability’.

6.16 Hjelmslev stresses that his ‘empirical principle’ is ‘not the same’ as ‘inductivism’, which, in ‘linguistics’, ‘inevitably leads to the abstraction of concepts which are then hypostatized as real’ (PT 11f) (cf. 4.67, 76; 5.17; 7.6f; 8.71; 12.8, 16, 95f; 13.57).⁷ In the ‘inductive’ ‘procedure, linguistics ascends’ from ‘particular to general’, from ‘more limited’ to ‘less’, or ‘from component to class’, e.g., from ‘sounds’ to ‘phonemes’. ‘Induction’ is thus ‘a continued synthesis’, ‘a generalizing, not a specifying method’, and cannot ‘satisfy the empirical principle with its requirement of an exhaustive description’ (PT 31, 12). ‘Induction leads’ ‘not to constancy but to accident’, and to ‘class concepts’ that are not ‘susceptible of general definition’ (PT 12) 7.25, 30; 13.44f).

6.17 In order to ‘clarify our position as opposed to that of previous linguistics’, Hjelmslev asserts: ‘linguistic theory’ is ‘necessarily deductive’; it is ‘a purely deductive system’ used only ‘to compute the possibilities that follow from its premises’, which ‘are of the greatest possible generality’ and thus ‘apply to a large number of empirical data’ (PT 11f, 13f) (cf. 6.17f, 33, 36f, 45, 49, 51f, 62; 12.8; 13.44f). The proper ‘procedure’ is ‘a continued analysis’ ‘progressing from class to components’ in an ‘analytic and specifying, not a synthetic and generalizing movement’ (PT 13, 30; cf. 6.36ff). The ‘object’ of ‘treatment should not be an inductively discovered class’, ‘but a deductively discovered linguistic localized variety of the highest degree’ (PT 84).

6.18 Hjelmslev is optimistic that ‘it is both possible and desirable for linguistic theory to progress by providing new concrete developments that yield an ever closer approximation’ to ‘the ideal set up and formulated in the “empirical principle”‘ (PT 19) (6.13). On that basis, when we ‘imagine several linguistic theories’, ‘one of these must necessarily be the definitive one’ (but cf. 13.3). Yet his standards for deriving and evaluating theories are peculiarly abstract. ‘From certain experiences’, which ‘should be as varied as possible, the linguistic theoretician sets up a calculation of all the conceivable possibilities within certain frames’ (PT 17; cf. 6.11). ‘These frames he constructs arbitrarily: he discovers certain properties present in all the objects that people agree to call languages, in order then to generalize those properties and establish them by definition’ (PT 17f). ‘From that moment the linguistic theoretician has -- arbitrarily, but appropriately -- decreed to which objects his theory can and cannot be applied’. ‘He then sets up, for all objects of the nature premised in the definition, a general calculus, in which all conceivable cases are foreseen’ (6.11). ‘This calculus’, ‘deduced from the established definition independently of all experience, provides the tools for describing or comprehending a given text’ or ‘language’. ‘Linguistic theory, then, sovereignly defines its object by an arbitrary and appropriate strategy of premises; the theory consists of a calculation from the fewest and most general possible premises, of which none that is specific to the theory seems to be of axiomatic nature’ (PT 15; cf. 5.86; 6.15, 22, 44).

6.19 The startling upshot is that ‘linguistic theory cannot be verified (confirmed or invalidated) by reference to any existing texts and languages’ (PT 18) (13.25). ‘Propositions’ and ‘theorems’ ‘will be true or false depending on the definitions chosen for the concepts’ (PT 24). ‘A theorem’, which ‘must have the form of an implication (in the logical sense) or must be susceptible of transposition into such a conditional form’, ‘asserts only that if a condition is fulfilled, the truth of a given proposition follows’ (PT 14). Yet ‘on the basis of a theory and its theorems we may construct hypotheses (including the so-called laws), the fate of which, contrary to that of the theory itself, depends exclusively on verification’. ‘No mention’ is made of ‘axioms or postulates; we leave it to epistemology to decide whether the basic premises explicitly introduced by our linguistic theory need any further axiomatic foundation’, and Hjelmslev hopes ‘the number of axioms’ might be ‘reduced’ ‘to zero’ (PT 15, 21; cf. 6.22, 44).

6.20 When ‘seeking an immanent understanding of language as a self-subsistent, specific structure’, ‘linguistic theory begins by circumscribing the scope of its object’, but without any ‘reduction of the field of vision’ or any ‘elimination of essential factors in the global totality which language is’ (PT 19) (cf. 12.2). ‘It involves only a division of difficulties and a progress of thought from simple to complex, in conformity with Descartes’ rules’. ‘The circumscription’ is ‘justified if it later permits an exhaustive and self-consistent broadening of perspective through a projection of the discovered structure onto the phenomena surrounding it, so that they are satisfactorily explained’, i.e., ‘if after analysis, the global totality -- language in life and actuality -- may again be viewed synthetically and as a whole’, ‘organized around a leading principle’ (PT 19f) (cf. 13.43). ‘Linguistic theory’ is ‘successful’ only when it has done all this, thereby ‘satisfying the empirical principle in its requirement of an exhaustive description; the test may be made by drawing all possible general consequences from the chosen structural principle’ (6.11).

6.21 A choice among ‘several possible methods’ should also follow ‘the simplicity principle’: pick the method that yields ‘the simplest possible description’ via ‘the simplest procedure’ (PT 18) (cf. 6.13). ‘Only by reference’ to ‘this principle’ can we ‘judge linguistic theory and its applications’ or ‘assert that one solution is correct and another incorrect’ (but cf. 13.57). Again, immanence is emphasized: ‘a theory will attain its simplest form by building on no other premises than those necessarily required by its object’ (PT 10).

6.22 The ‘main task’ of ‘linguistic theory’ ‘is to make explicit the specific premises of linguistics as far back as possible’ by ‘setting up’ ‘a system of definitions’ that in turn ‘rest on defined concepts’ (PT 20). As we saw (6.14f), Hjelmslev recommends ‘strictly formal’ ‘definitions’ rather than ‘real’ ones, and ‘hopes to guard against any postulates about the essence of an object’ (PT 20f, 32). Here, ‘it is not a question of trying to exhaust the intensional nature of the objects or even of delimiting them extensionally on all sides, but only of anchoring them relatively in respect to other objects’ (PT 21; cf. 6.25). ‘In addition to the formal definitions’, Hjelmslev would admit ‘operative definitions, whose role is only temporary’; ‘later’, they ‘may be transformed into formal definitions’, or else their ‘definienda do not enter into the system of formal definitions’ (PT 21; for examples, see PT 46, 48, 81, 118). ‘This extensive defining’ should help keep ‘linguistic theory’ free both ‘from specific axioms’ and from ‘implicit premises’ or ‘postulates’ -- perhaps a suitable ‘strategy’ for ‘any science’ (PT 21, cf. PT 15, 21; 6.18f).

6.23 Like Saussure, Hjelmslev grants the ‘evident and fundamental proposition’ ‘that a language is a system of signs’ (PT 43) (cf. 2.8, 21, 25ff, 69; 5.63; 8.54; 12.9ff, 42f, 54, 62-67). ‘Linguistic theory must be able to tell us what meaning can be attributed to this proposition and especially to the word sign’. According to ‘the vague concept bequeathed by tradition’, ‘a “sign”‘ is ‘a sign for something’, and ‘the bearer of a meaning’ (cf. 5.63; 6.47). Such a usage might fit ‘the entities commonly referred to as sentences, clauses, and words’ (PT 43f) (cf. 12.69f, 75f, 78). But problems arise if we ‘try to carry out the analysis as far as possible, in order to test for an exhaustive and maximally simple description’. ‘Words are not the ultimate, irreducible signs’, despite ‘the centring of conventional linguistics around the word’ (but cf. 2.17, 55; 3.31-38, 73; 4.42, 60, 63, 4¹⁴; 5.18, 36, 41, 49, 51-54, 56, 58; 7.70, 7³⁴; 8.47f, 53; 9.75; 12.66, 69; 13.29). ‘Words can be analysed into parts’, such as ‘roots’ or ‘derivational’ and ‘inflectional elements’, that are also ‘bearers of meaning’ (cf. 2.55, 57, 62ff; 3.26, 32, 34, 41ff, 53; 4.50, 59f, 62). So Hjelmslev postulates a further system of ‘minimal’ ‘invariants’ he calls by the ‘purely operative term’ ‘figurae’, which are ‘non-signs’ (PT 65, 46). ‘Through ever new arrangements’ of ‘a handful’ of these ‘figurae’, ‘a legion of signs can be constructed’; otherwise, a ‘language’ ‘would be a tool unusable for its purpose’ (PT 46) (cf. 2.52; 3³). Hence, a ‘language’ is by its ‘external functions’ a ‘sign system’, but by its ‘internal structure’ a ‘system of figurae’ (PT 47). In this sense, ‘the definition of language as a sign system’ proves ‘on closer analysis to be unsatisfactory’.

6.24 This view suspends the problem of determining the size of the set of signs. Since ‘a language’ ‘must always be ready to form new signs’, their ‘number’ must be ‘unrestricted’ in the ‘economy’ of ‘inventory lists’, whereas the ‘number’ of usable ‘non-signs’ ‘is restricted’ (PT 46). ‘To understand the structure of a language’, ‘this principle’ of ‘analysis’ ‘must be extended so as to be valid for all invariants of the language’, ‘irrespective’ of ‘their place in the system’ (PT 65). So far, though, ‘conventional linguistics’ has focused only on ‘figurae of the expression plane’,⁸ whereas ‘an analysis into content-figurae has never been’ ‘even attempted’ (PT 65, 67) (cf. 6.26, 30). ‘This inconsistency has had the most catastrophic consequences’, making the analysis of content seem ‘an insoluble problem’ (PT 67). Thanks to ‘the solidarity between the form of the expression and the form of the content’, ‘the content plane’ can also ‘be resolved’ ‘into components with mutual relations that are smaller than the minimal-sign-contents’ (PT 65, 67) (cf. 6.41, 47f; 13.30). Indeed, the two ‘terms’ ‘are quite arbitrary’: ‘their functional definition provides no justification for calling one, and not the other, of these entities expression’ or ‘content’ (PT 60) (12.31).

6.25 This same solidarity indicates why the ‘popular conception’ of ‘a sign for something’ is ‘untenable’ in view of ‘recent linguistic thinking’ (PT 47). The ‘sign’ is not ‘an expression that points to a content outside the sign itself’, but, according to Weisgerber (1929) and of course Saussure, ‘an entity generated by the connection between an expression and a content’ (cf. 2.25; 12.19). For this connection, Hjelmslev selects the term ‘sign function’; ‘expression and content’ are ‘the functives that contract this function’, where ‘functive’ means the ‘terminal of a function’ it ‘contracts’ (PT 33, 48, i.r.). The concept of ‘function’, ‘adopted’ ‘in a sense that lies midway between the logico-mathematical and the etymological sense’ (PT 33),⁹ occupies the central role in Hjelmslev's ‘theory’, in which ‘only the functions have scientific existence’, and ‘objects’ are purely ‘functives’ (PT 85, 81, 33; cf. 5.20; 6.28). ‘A ‘function’ can also be a ‘functive’ in some higher ‘function’; and ‘a functive that is not a function’ is ‘called an entity’ (PT 33). ‘A constant is ‘a functive whose presence is a necessary condition for the presence’ of its other terminal; ‘a variable’ is ‘a functive whose presence is not necessary’ (PT 35). Hjelmslev introduces a profusion of specific ‘functions’ and ‘functives’, many with colourful names like ‘heteroplane’ and ‘homoplane’ or ‘plerematic’ and ‘cenematic’ (RT 5f, 99, 136), but since he never gives examples, their usefulness is hard to judge (cf. 6.59).

6.26 For every ‘sign’, Hjelmslev emphasizes the ‘solidarity between the sign function and its two functives, expression and content’; these two ‘necessarily presuppose each other’ (PT 48). ‘We understand nothing of the structure of a language if we do not constantly take into first consideration the interplay between the planes’ (PT 75). ‘Except by artificial isolation, there can be no content without expression’, nor ‘an expression without a content’ (PT 49). ‘If we think without speaking, the thought is not a linguistic content’; ‘if we speak without thinking, and in the form of series of sounds to which no content can be attached’, ‘such speech is an abracadabra, not a linguistic expression’. ‘Saussure's “Gedankenexperiment”‘ of ‘trying to consider expression and content each alone’ was therefore pointless (PT 49f). A ‘content’ might appear ‘meaningless’ from the standpoint of ‘normative logic or physicalism’, ‘but it is a content’ (cf. 8⁷; 8¹⁴).

6.27 Nonetheless, ‘a description in accordance with our principles must analyse content and expression separately’ into ‘entities which are not necessarily susceptible of one-to-one matching with entities in the opposite plane’ (PT 46) (cf. 3.22; 5.48, 64; 9.39; 13.55). Though the ‘grammatical method’ of ‘recent times’ ‘starts’ from ‘the expression’ and ‘goes from there to the content’, one could ‘with the same right’ ‘proceed from the content to the expression’ (PT 75). Hjelmslev proposes ‘two disciplines’, each for the ‘study’ of one plane; yet they must be ‘interdependent’, since they cannot ‘be isolated from each other without serious harm’. ‘If we consider’ ‘two or more signs in mutual correlation, we shall always find that there is a relation between a correlation of expression and a correlation of content’ (PT 65f). ‘If such a relation is not present’, then we have ‘not two different signs, but only two different variants of the same sign’.

6.28 As we can see, Hjelmslev's vision of a sign system follows from Saussure's but is elaborated and revised. One major revision concerns ‘Saussure's distinction between form and substance’ (PT 123) (cf. 2.16f). ‘If we maintain Saussure's terminology’, ‘it becomes precisely clear that the substance depends on the form to such a degree’ that it ‘can in no sense be said to have independent existence’ (PT 50). ‘What from one point of view is “substance” is from another point of view “form”, this being connected with the fact that functives denote only’ ‘points of intersection for functions, and that only the functional net of dependences has knowability and scientific existence, while “substance”, in an ontological sense, remains a metaphysical concept’ (PT 81; cf. PT 23; 5.20; 6.13, 25, 44 6¹⁵; 11¹⁵).

6.29 Still, Saussure was ‘correct in distinguishing form and substance’, and in ‘asserting that a language is a form, not a substance’ (PT 54, 23; EL1 30) (2.16), and Hjelmslev too carefully separates substance from the concerns of his projected science. He hopes to cover ‘language in a far broader sense’ ‘precisely because the theory is so constructed that linguistic form is viewed without regard for “the substance”‘ (PT 102). ‘“Substance” cannot in itself be a definiens of a language’ (PT 103, i.r.). So ‘linguistics must be assigned the special task of describing the linguistic form, in order thereby to make possible a projection of it upon the non-linguistic entities’ which ‘provide the substance’ (PT 78f) (cf. 13.54). Hjelmslev's ‘science would be an algebra of language’ whose ‘arbitrarily named entities’ ‘have no natural designation’ and ‘receive a motivated designation only on being confronted with the substance’ (cf. 6.8; 13.15). Concurring with his already cited detachment of theory from reality (6.12, 15), Hjelmslev argues that in his ‘calculus, there is no question of whether the individual structural types are manifested, but only whether they are manifestable’ ‘in any substance whatsoever’ (PT 106). ‘Substance is not a necessary presupposition for linguistic form’, but the ‘form’ is ‘necessary’ ‘for substance’. In any ‘manifestation’, ‘the language form is the constant and the substance the variable’. ‘The substance of both planes can be viewed both as physical entities (sounds in the expression plane, things in the content plane) and as the conception of these entities held by users of the language’ (PT 78).

6.30 Form and substance are then deployed as categories for subdividing the two planes of content and expression. On the side of form, ‘the content-form and the expression-form’ are the ‘two functives’ of ‘the sign function’ (PT 57). On the side of substance, the ‘expression-substance’ is ‘the sound sequence’ and ‘is ordered to an expression-form’; the ‘content-substance’ is the ‘thing’ or ‘thought’ and ‘through the sign, is ordered to a content-form and arranged under it together with various other entities of content-form’ (PT 57f, 78, 50). This account is intended to supplant the old notion ‘that a sign is a sign for something’ ‘outside the sign itself’ (PT 57) (6.23).

6.31 Hjelmslev's elaborated four-part scheme is clouded somewhat by an added notion called ‘mening’ in Danish and translated into English as ‘purport’ (cf. PT 135). In several passages, the term is associated with ‘substance’, and ‘content-purport’ appears where we might expect ‘content-substance’ (PT 76f, 78f, 102f, 111). For instance, ‘purport’ is said to ‘have no possible existence except through being substance’ for a ‘form’: ‘the content-form’ ‘is independent of, and stands in arbitrary relation to, the purport, and forms it into a content-substance’ (PT 54, 52) (cf. 6.15; 13.24). ‘Linguistic form’ ‘lays arbitrary boundaries on a purport-continuum’ that ‘depends exclusively on this structure’ (PT 74). Otherwise, the ‘purport’ ‘exists provisionally as’ ‘an unanalysed entity’; ‘subjected to many different analyses’, it ‘would appear as so many different objects’ (PT 50f). To make his point, Hjelmslev uses metaphors, which are otherwise conspicuously absent in his theory books. ‘Purport’ is an ‘amorphous “thought-mass”‘ ‘formed in quite different patterns’, like a ‘handful of sand’ or a ‘cloud in the heavens’; ‘form’ is ‘projected onto the purport, just as an open net casts its shadow on an undivided surface’ (PT 52, 57) (cf. 2.32; 3.3; 6.2, 57).

6.32 Who is to study ‘purport’ and how is even less clear. At one point, we read that ‘purport is inaccessible to knowledge’, because ‘knowledge’ presupposes ‘an analysis’ (PT 76). Yet elsewhere, the ‘description’ of ‘purport’ is allotted ‘partly to the sphere of physics and partly to that of (social) anthropology’; ‘logical’, ‘psychological’, and ‘phenomenological descriptions’ are suggested as well (PT 51, 77f). Later, a ‘science of linguistic content-purport’ is envisioned as a project for ‘a great number of special sciences outside linguistics’, and ultimately for ‘a collaboration of all the non-linguistic sciences’, because ‘they all, without exception, deal with a linguistic content’ (PT 103, 77f) (cf. 6.54; 13.21). These ‘non-linguistic sciences’ ‘must undertake an analysis of the linguistic purport without considering the linguistic form’, whereas ‘linguistics’ ‘must undertake an analysis of the linguistic form without considering the purport’ (PT 78) (13.54). And ‘since the linguistic formation of purport is arbitrary’, ‘these two descriptions -- the linguistic and the non-linguistic -- must be undertaken independently’ (PT 77; cf. PT 103). Yet this division of labour is redundant if ‘the non-linguistic analysis of the purport’ ‘by the non-linguistic sciences’ will lead ‘to a recognition of a “form” essentially of the same sort as the “linguistic form”‘; or else unworkable if ‘purport can be known only through some formation’ and ‘has no scientific existence apart from it’ (PT 80, 76).

6.33 Instead of distinguishing ‘langue’ and ‘parole’, Hjelmslev draws an analogous division between ‘schema’ and ‘usage’ (PT 81; EL1 72) (cf. 2.20; 13.36).¹⁰ The ‘schema’ is ‘the linguistic hierarchy discovered’ by ‘deduction’ and is ‘the constant’, whereas the ‘usage’ is the ‘non-linguistic hierarchy’ discovered by the ‘analysis of purport’ and is ‘the variable’ (PT 81, 106). While Saussure's ‘langue’ was ‘static’ (CG 81), Hjelmslev's ‘schema’ is not even ‘subjected to the law of life’; if the ‘language dies out’, ‘the schema’ remains an ‘ever-present realizable possibility’ that happens to be ‘latent’ rather than ‘manifested’; only the ‘usage’ can ‘come into being’ and ‘die out’ (EL2 116). This assertion too reflects Hjelmslev's demand that linguistic theory cover ‘all conceivable possibilities’ (6.11, 18, 20, 36, 38, 50, 63).

6.34 A similar division, one Hjelmslev develops in more detail (though without comparing it to Saussure's), falls between ‘system’ and ‘process’ -- ‘concepts’ of ‘great generality’ or even ‘universal character’ (PT 39, 102) (cf. 9.41). ‘For every process there is a corresponding system, by which the process can be analysed and described’ (PT 9). ‘A process and a system’ ‘together contract a function’ ‘in which the system is the constant’ (PT 39). Hjelmslev aligns the pair ‘process’ versus ‘system’ with the pair ‘text’ versus ‘language’ and also with the pair ‘syntagmatic’ versus ‘paradigmatic’ (PT 39, 85, 109, 135), though this latter pair again is not developed in detail (cf. 6.39ff; 13.27).

6.35 ‘The process is the more immediately accessible for observation’ and ‘more “concrete”‘, ‘while the system must be’ ‘“discovered” behind it by means of a procedure and so is only mediately knowable’ (PT 39). But we must not assume that ‘the process can exist without a system’. On the contrary, ‘the existence of a system is a necessary premise for the existence of a process’; ‘the system’ is ‘present behind it’ ‘governing and determining it in its possible development’.¹¹ Conversely (befitting his detachment of theory from reality, 6.12), however, Hjelmslev claims that the ‘existence’ of ‘a system’ ‘does not presuppose the existence of a process’. He ‘imagines’ ‘a language without a text constructed in that language’, and requires ‘linguistic theory’ to ‘foresee’ such a language ‘as a possible system’ (PT 39f). Its ‘textual process is virtual’ rather than ‘realized’ (cf. 6.11, 42, 63; 13.39).

6.36 To ‘test the thesis that a process has an underlying system’, the ‘process can be analysed’ into ‘a limited number of elements recurring in various combinations’ (PT 9f). Hence, ‘linguistic theory prescribes a textual analysis, which leads us to recognize a linguistic form behind the “substance” immediately accessible to observation by the senses, and behind the text a language (system) consisting of categories from whose definitions can be deduced the possible units of the language’ (PT 96). This analysis is a ‘purely formal procedure’ for treating the ‘units of a language’ in terms of ‘figurae for which rules of transformation hold’ (cf. 6.23). The ‘basis’ is in the ‘definitions’, ‘made precise and supplemented’ by ‘rules of a more technical sort’. The Resume presents no less than 201 such rules, which predictably state that ‘one must operate with the lowest possible number of variants’; that ‘in free articulation, all conceivable configurations are to be anticipated’; and so on (RT 20, 40).

6.37 ‘If the linguistic investigator is given anything (we put this in conditional form for epistemological reasons), it is the as yet unanalysed text in its undivided and absolute integrity’ (PT 12) (cf. 2.88; 3.31; 5.5, 15; 8.35, 44; 9.1, 3, 8, 16, 41f, 107, 9¹⁹; 11.1f; 13.31). So ‘linguistic theory starts from the text as its datum’ and ‘object of interest’, and attempts to produce ‘a self-consistent and exhaustive description through an analysis’ (PT 21, 16). ‘To order a system to the process of that text’, ‘the text is regarded as a class analysed into components, then these components as classes analysed into components, and so on until the analysis is exhausted’ (PT 12f; cf. 6.39). ‘This method of procedure’ is a ‘deduction’, and to ‘provide’ it is ‘the aim of linguistic theory’ (PT 13, 16; cf. 6.17, 33; 12.8; 13.44f).

6.38 The ‘theory’ must also ‘indicate how any other text of the same premised nature can be understood in the same way’ by ‘furnishing us with tools that can be used on any text’ (PT 16). ‘Obviously, it would be humanly impossible to work through all existing texts’, and ‘futile’ as well ‘since the theory must also cover texts as yet unrealized’ (PT 17). But though it ‘must be content’ with a ‘selection’, ‘linguistic theory’ may draw enough ‘information’ to ‘describe and predict’ ‘any conceivable or theoretically possible texts’ ‘in any language whatsoever’ (PT 16f) (cf. 6.11). ‘This principle of analysis’ must be ‘treated’ by ‘the deepest strata of its definition system’ (PT 21). Such a broad demand is contrasted with ‘the restricted practical and technical attitude’ which ‘demands’ that ‘linguistic theory’ be ‘a sure method for describing a given limited text’ (PT 125) (but cf. 6.61; 7.7; 8.44; 9.1f, 109ff; 13.). Hjelmslev proposes instead ‘an ever broader scientific’ and ‘humanistic attitude, until the idea finally comes to rest in a totality-concept that can scarcely be imagined more absolute’ (PT 125f).

6.39 ‘The whole textual analysis’ ‘consists of a continued partition’, ‘each operation’ being ‘a single minimal partition’ until all ‘partitions’ are ‘exhausted’ (PT 30). At each ‘partition’, we ‘make an inventory of the entities that have the same relations, i.e., that can occupy the same position in a chain’, e.g., ‘all primary’ or ‘secondary clauses’, ‘all words, all syllables, and all parts of syllables’ (PT 41f). For ‘exhaustive description’, ‘we must not omit any stage of analysis that might be expected to give functional return’ (PT 42, 97). ‘The analysis must move from the invariants’ with ‘the greatest extension conceivable’ to those with ‘the least’ and ‘traverse’ ‘as many derivative degrees’ ‘as possible’ in between (PT 97). This ‘analysis differs essentially’ from that in ‘conventional linguistics’, which ‘is very far from having carried the analysis to the end’ (PT 97, 99). A ‘traditional’ analysis ‘is concerned neither with’ ‘very great’ nor ‘very small extension’ (PT 97). ‘The linguist’ would ‘begin with dividing sentences into clauses’ and would ‘refer the treatment of larger parts of the text’ ‘to other sciences -- principally logic and psychology’ (PT 97f). Researchers didn't ask whether any ‘logico-psychological analysis of the larger parts’ ‘had been undertaken’, or whether it had been ‘satisfactory from the linguist's point of view’ (cf. 13.17).

6.40 The ‘size’ of ‘the inventories’ is expected to ‘decrease as the procedure goes on’: ‘unrestricted inventories’ yield to ‘restricted’, which in turn ‘decrease in size’ ‘until all inventories have been restricted’ ‘as much as possible’ (PT 42, 71). Since ‘we cannot know beforehand whether any given stage is the last’, every ‘inventory’ ‘must satisfy our empirical principle’ by being ‘exhaustive and as simple as possible’ (PT 60) (cf. 6.13). ‘This requirement’ applies most of all to ‘the concluding stage’, where we ‘recognize the ultimate entities of which all others’ are ‘constructed’; keeping their ‘number’ ‘as low as possible’ is vital ‘for the simplicity of the solution as a whole’ (cf. 13.26). Here, Hjelmslev invokes ‘the principle of economy’, calling for a ‘procedure’ that gives ‘the simplest possible’ ‘result’ and is ‘suspended if it does not lead to further simplification’; and ‘the principle of reduction’, requiring ‘each operation’ to be ‘continued or repeated until the description is exhausted’ and to ‘register the lowest possible number of objects’ ‘at each stage’ (PT 61).

6.41 The ‘partitioning’ of a ‘linguistic text’ ‘defines’ ‘parts’ according to ‘mutual selection, solidarity, or combination’ (PT 98).¹² The first ‘partition’ is ‘into content line and expression line, which are solidary’ (i.e., ‘interdependent in a process’) (PT 98, 24; cf. 2.27; 4.17; 6.26, 47). We can then ‘analyse the content line’ into such classes as ‘literary genres’ or ‘sciences’ (PT 98; cf. 6.10). ‘At a more advanced stage’, ‘the larger textual parts must be further partitioned into the productions of single authors, works, chapters, paragraphs’, and ‘then’ ‘into sentences and clauses’ (PT 98f).¹³ ‘At this point’, ‘syllogisms will be analysed into premises and conclusions -- obviously a stage’ where ‘formal logic must place an important part of its problems’ (cf. 13.18). ‘In all this is seen a significant broadening of the perspectives, frames, and capacities of linguistic theory, and a basis for a motivated and organized collaboration’ with ‘other disciplines which till now, obviously more or less wrongly, have usually been considered as lying outside the sphere of linguistic science’ (cf. 6.49; 13.9-21).

6.42 ‘In the final operations’, the ‘partition descends to entities of a smaller extension than those’ traditionally ‘viewed as the irreducible invariants’ (PT 99). Both ‘the content plane’ and ‘the expression plane’ are now to be ‘analysed’ into ‘an inventory of taxemes’ (cf. 4.45, 64; 5²⁶). These are ‘virtual elements’ that may (but need not) be ‘manifested by phonemes’ in ‘the expression plane’, but how they are manifested in the content plane is a major mystery in Hjelmslev's outlook (cf. 6.47, 13.30). Finally, ‘the end points of the analysis’ are reached by partitioning ‘the taxemes’ into ‘glossemes’, ‘the minimal forms’ and ‘irreducible invariants’ of ‘glossematics’ (PT 99f, 80; cf. 4.45; 6.8). Hjelmslev mentions here ‘parts of phonemes’, but not a single example of an actual ‘glosseme’ appears in PT or RT, though the latter defines many types of ‘glossemes’, such as ‘median’ and ‘peripheral’, ‘centrifugal’ and ‘centripetal’, ‘primary’ and ‘secondary’, ‘principal’ and ‘accessory’, and so on (RT 100, 179f, 184, 187, 192) (cf. 6.52). These definitions merely refer the term back to ‘taxemes’ (e.g. ‘a median glosseme is a glosseme that enters into a median taxeme’) (RT 179) (cf. 6.59). Thus, although ‘glossemes’ are the ‘highest-degree invariants within a semiotic’ (RT 100), we find no reliable way to tell their particular nature or status. Nor are they anywhere situated in respect to the ‘figurae’ said to compose signs (cf. 6.23, 36).

6.43 The ‘method’ must ‘allow us, under precisely fixed conditions, to identify two entities with each other’ (PT 61).¹⁴ This ‘requirement’ is needed because in each ‘inventory’, we shall ‘observe that in many places in the text we have “one and the same”‘ entity (PT 61f). ‘These specimens’ are the ‘variants, and the entities of which they are specimens’ are ‘invariants’ -- a distinction ‘valid for functives in general’. A prime example ‘in modern linguistics’ is ‘the so-called phonemes’ as the ‘highest-degree invariants of the expression-plane’ (cf. 2.69; 4.29f, 33; 5.42f; 8³⁵; 12.80; 13.26). But neither ‘the London school’ (e.g. Daniel Jones) nor ‘the Prague Circle’ (e.g. N.S. Trubetzkoy) ‘recognized that the prerequisite for an inventory is a textual analysis made on the basis of functions’ (cf. 8.69). Instead, they used a ‘vague “real” definition’ with ‘no useful objective criteria in doubtful cases’, e.g., when they ‘defined’ ‘vowel and consonant’ by ‘physiological or physical premises’ (PT 62f) (cf. 13.26). The London group used ‘position’ to define the ‘phoneme’ and made no ‘appeal to the content’, whereas the Prague group insisted on the ‘distinctive function’ that allowed ‘differentiations of intellectual meaning’ (PT 63ff; cf. 2.70; 7²⁷; 13.26f). Hjelmslev judges the ‘distinctive criterion’ of ‘the Prague Circle’ ‘undoubtedly right’, though he adds gruffly that ‘on all other points strong reservations must be made’ about their ‘theory and practice’ in ‘phonology’.

6.44 The criterion of ‘appropriateness’ stated for ‘the empirical principle’ (6.13, 15) suggests that the ‘basis of analysis may differ for different texts’ (PT 22). Still, ‘the principle of analysis’ is ‘universal’. ‘Naive realism’ might ‘suppose that analysis consisted merely in dividing a given object into parts’, ‘then those again into parts’, ‘and so on’. But to ‘choose between several possible ways of dividing’, the ‘adequate’ ‘analysis’ is the one ‘conducted’ ‘so that it conforms to the mutual dependences between parts’.¹⁵ Hence, ‘the principle of analysis’ is centred on the ‘conclusion’ that ‘the object’ ‘and its parts have existence only by virtue of these dependences’ (PT 22f) (6.25, 28). ‘The objects of naive realism’ are found to be ‘nothing but intersections of bundles of such dependences’ and can be ‘defined and grasped scientifically only in this way’. ‘The recognition’ ‘that a totality does not consist of things but of relationships, and that not substance but only its internal and external relationships have scientific existence’ ‘may be new in linguistic science’ and shows ‘the exclusive relevance of functions for analysis’ (PT 23, 80f; 5.20; 6.25; 12.25). ‘The postulation of objects’ ‘is a superfluous axiom’ and ‘a metaphysical hypothesis from which linguistic science will have to be freed’ (PT 23; cf. PT 81; 6.13ff, 25).

6.45 This line of argument is intended to strengthen the thesis that ‘the principle of analysis must be a recognition’ of ‘dependences’ (PT 28). We may ‘conceive of the parts to which the analysis shall lead as nothing but bundles of lines of dependence’. ‘The basis of analysis’ must therefore ‘be chosen according to what lines of dependence are relevant’ and proper for ‘making the description exhaustive’. ‘The analysis’ proceeds by ‘registering certain dependences between terminals’ ‘we may call parts of the text’, these too ‘having existence precisely by virtue of the dependences’. Both ‘the dependence between the whole’ ‘(the text)’ ‘and the parts’, and the one between ‘the so-called parts’ are ‘characterized’ by ‘uniformity’ (PT 28f). For example, we shall ‘always find the same dependence between a primary clause and a secondary clause’, or ‘between stem and derivational element or between the central and marginal parts of a syllable’. In sum, ‘we can define’ ‘analysis’ ‘formally as description of an object by the uniform dependences of other objects on it and on each other’; ‘the object’ is ‘a class’ and the others are its ‘components’ (cf. 6.17, 33). To fit his concept of ‘a deduction’, Hjelmslev requires that ‘each operation will premise the preceding operations’ (PT 30f). He advocates ‘a special rule of transference’ to ‘prevent a given entity from being analysed at a too early stage’ and to ‘ensure that certain entities under given conditions are transferred unanalysed from stage to stage’ (PT 41). Examples include ‘a sentence’ of ‘one clause, and a clause of only one word’ (cf. 5.51, 53). ‘The Latin imperative “i” (“go”)’, for instance, can be ‘at the same time a sentence, a clause, and a word’ (and a morpheme and a phoneme too).

6.46 Since ‘the registration of certain functions’ ‘cannot be reached by a mere mechanical observation of entities that enter into actual texts’, we may have to ‘interpolate certain functives which would in no other way be accessible to knowledge’ (PT 93).¹⁶ This method is called ‘catalysis’ (PT 94). Hjelmslev points here to the ‘incalculable accidents’ and ‘disturbances’ ‘in the exercise of language’ (in ‘parole’), such as when a text is ‘interrupted or incomplete’, and he says that ‘in general’ they could be ‘eliminated’ (cf. 7.12). Yet ‘an exhaustive description’ should ‘register’ ‘the outward relations which the actually observed entities have’, including ‘aposiopesis and abbreviation’, which form ‘a constant and essential part’ in ‘the economy of linguistic usage’. This proviso may seem unexpected for such an abstract approach, and Hjelmslev warns us to ‘take care not to supply more in the text than what there is clear evidence for’ (PT 95).

6.47 But ‘clear evidence’ and ‘outward relations’ might be hard to find for the ‘content plane’. He stresses the ‘solidarity’ of ‘content with expression’ (6.26, 41) presumably because he hopes, like many linguists, to analyse content with the methods available for analysing expression, i.e. form (13.54). He hails the ‘far-reaching’ ‘discovery’ that ‘the two sides (planes) of a language have completely analogous categorical structure’, even though the ‘analysis into content-figurae has never been made’ (PT 101, 67) (cf. 6.24). He ‘predicts with certainty that such an analysis can be carried out’ for both ‘planes’ ‘according to a common principle’, and considers it an ‘inevitable logical consequence’ that the same ‘tests can be applied to the content-plane’ and ‘enable us to register the figurae that compose the sign-contents’ (PT 66f, 70). ‘Experience shows that in all hitherto observed languages, there comes a stage in the analysis of the expression when the entities’ ‘no longer’ appear as ‘bearers of meaning and thus are no longer sign-expressions’ (PT 45) (cf. 6.23). But to decide whether something ‘bears meaning’ (in the sense that changing it also changes the meaning, cf. PT 66, 68f, 70), or what that meaning is, may not be easy. He gives one list of ‘entities of content’, including ‘“man”, “woman”, “boy”, “girl”‘ (PT 70), but these are also words, and he doesn't analyse them further. His 1957 paper on ‘structural semantics’ (EL1 96-112) (the earliest proposal I know of for that field) also lacks lengthy or detailed analysis, and offers as ‘elements of content’ ‘“be” + “1st person” + “singular” + “present” + “indicative”‘ (EL1 111); but these are more notions of grammar than of meaning per se.

6.48 At one point, Hjelmslev remarks that ‘the “meaning”‘ which any ‘minimal entity can be said to bear’ is ‘a purely contextual meaning’ (PT 44f). In ‘the continued analysis’ of a text, ‘there exist no other perceivable meanings than contextual meanings’; ‘any entity’ or ‘sign is defined relatively, not absolutely, and only by its place in the context’. Nor do ‘dictionaries’ ‘yield definitions that can be immediately taken over by a consistently performed analysis’ (PT 71f). ‘So-called lexical meanings in certain signs are nothing but artificially isolated contextual meanings or paraphrases of them’ (PT 45). This argument implies that analysts might have to generate or design, on each new occasion, the units of content they propose to discover; and the results might not apply to other texts or even to other analyses of the same text. Also, determining what the entities of meaning are should get steadily harder as the partitioning proceeds, and would seldom come to a ‘self-consistent, exhaustive, and simple’ conclusion (cf. 6.13, 37, 54). The most restricted inventories (6.40) would be very general meanings like the ‘small closed classes’ ‘“large”::”small”‘ or ‘“long”::”short”‘ (EL1 110), which don't match the content of the most restricted inventories of expression like phonemes or letters. The inverse might hold: the most general meanings might reflect the largest segments of text (cf. 11.19, 32, 49, 65).

6.49 Nonetheless, Hjelmslev remains confident that the ‘method of procedure’ during ‘the whole analysis’ ‘proves to result in great clarity and simplification, and it also casts light on the whole mechanism of language in a fashion hitherto unknown’ (PT 59). Now ‘it will be easy to organize the subsidiary disciplines of linguistics according to a well-founded plan and to escape at last from the old, halting division of linguistics into phonetics, morphology, syntax, lexicography, and semantics’. ‘Logically’, ‘process dependences’ could be ‘registered only in syntax’, ‘i.e., between the words of a sentence but not within the individual word or its parts; hence the preoccupation with grammatical government’ (PT 26f). But ‘the description of a language on the basis of the empirical principle does not contain the possibility of a syntax or a science of parts of speech’ (PT 101) (cf. 13.7). ‘The entities’ of ‘ancient grammar’ ‘will be rediscovered in refined form in far different places within the hierarchy of the units’; and ‘the entities’ of ‘conventional syntax’, such as ‘primary’ and ‘secondary clauses’ or ‘subject and predicate’, ‘are reduced to mere variants’ (PT 101, 84, 73; cf. 2.6; 6.43; 7.4; 8.38; 12.41, 88; 13.7). So ‘the distribution of functives into’ ‘invariants and variants’ ‘eliminates the conventional bifurcation of linguistics into morphology and syntax’, (for once) ‘in agreement with several modern schools’ (PT 73, 26) (cf. 2.55; 5.54; 8.57; 9.31; 11.35; 13.28). In addition, ‘we must not expect any semantics or phonetics’, because they are not ‘deductive’ and ‘formal’ enough to handle ‘non-linguistic “substance”‘ (PT 96; cf. 6.13, 28ff). Against ‘linguistics’, which has ‘neglected’ its ‘main task’ (stated in 6.8) Hjelmslev calls for ‘a description of the categories of expression on a non-phonetic basis’ (PT 79n) (cf. 13.26, 32).

6.50 Although he thus rejects the ‘conventional’ domains, Hjelmslev proposes in return a ‘semiotic’ of such expansive scope that ‘semiotic structure is revealed as a stand from which all scientific objects may be viewed’ (PT 127). This inclusion will end ‘the belief’ fostered by ‘conventional phonetics’ that ‘the expression-substance’ ‘must consist exclusively of sounds’ (PT 103) (cf. 3.18; 4.28ff). Hjelmslev wants to include as well ‘gesture’, ‘sign-language’, and ‘writing’, the latter being ‘a graphic “substance” which is addressed exclusively to the eye and which need not be transposed into a phonetic “substance” in order to be understood’ (PT 103f; but cf. 2.21f; 6.50; 4.37ff; 8.72; 9.42f; 12.83; 13.33). The best illustration is ‘a phonetic or phonemic notation’, or a ‘phonetic orthography’ like ‘the Finnish’ (cf. 2.69; 3.54; 4.38; 6.50; 8.75). The written mode is welcomed as evidence that ‘different systems of expression can correspond to one and the same system of content’, and as another reason why ‘the linguistic theoretician’ must not merely ‘describe’ the ‘present’ ‘system’, but must ‘calculate what expression systems in general are possible’ (PT 105) (cf. 6.11). Moreover, ‘the invention of the alphabet’ is a model for ‘linguistic theory’ of an ‘analysis that leads to entities of the least possible extension and the lowest possible number’ (PT 42f) (cf. 8.75). Hjelmslev dismisses the objection that ‘all these “substances” are “derived”‘ from sound and ‘“artificial”‘ rather than ‘“natural”‘; this ‘opinion is irrelevant’ because, even if ‘“derived”‘, the substance is still ‘a manifestation’, and because we cannot be sure that ‘the discovery of alphabetic writing’, ‘hidden in pre-history’, ‘rests on a phonetic analysis’ -- a ‘diachronic hypothesis’ anyway, and in ‘modern linguistics, diachronic considerations are irrelevant for synchronic description’ (PT 104f) (cf. 2.36ff).

6.51 The framework of Hjelmslev's ‘semiotics’ would be a hierarchy of ‘orders’. The minimal requirement for a ‘semiotic’ is ‘two planes’ that do not ‘have the same structure throughout’, i.e., are not ‘conformal’ (PT 112; cf. 6.26). A ‘test’ could be used ‘for deducing whether or not a given object is a semiotic’. But Hjelmslev ‘leaves it to the experts to decide’ if the ‘symbolic systems of mathematics and logic’, or ‘art’ and ‘music’ ‘are to be defined as semiotics’; they may not be ‘biplanar’, such that we could not ‘encatalyze [i.e., supply from outside] a content-form’ (PT 113; cf. 6.10, 41). ‘Games’, on the other hand -- including ‘chess’, Saussure's favourite model (2.80) -- ‘lie close to’ or ‘on the boundary’ ‘between semiotic and non-semiotic’ (PT 110). Whereas ‘the logical side’ sees ‘a game’ as ‘a transformation system of essentially the same structure as a semiotic’, ‘the linguistic side’ sees ‘a game’ as ‘a system of values analogous to economic values’. Still, to the extent that ‘there exist for the calculus of linguistic theory not interpreted, but only interpretable systems’ (to ‘interpret’ here being to ‘order’ a ‘content-purport’), ‘there is no difference between’ ‘chess’ or ‘pure algebra’ and ‘a language’ (PT 111f).

6.52 ‘To establish a simple model situation’, Hjelmslev had ‘proceeded on the tacit assumption that the datum is a text composed in one definite semiotic, not in a mixture of two or more’ (PT 115). Yet ‘any text’ ‘not of so small extension that it fails to yield a sufficient basis for deducing a system generalizable to other texts, usually contains derivates that rest on different systems’, among which he names ‘styles’, ‘media’, ‘tones’, ‘idioms’, ‘vernaculars’, and ‘physiognamies’ (individual speaking styles), alongside ‘national’ and ‘regional languages’ (PT 115f). As types of ‘style’ he enumerates ‘belletristic’, ‘slang’, ‘jargon’, ‘colloquial’, ‘lecture’, ‘pulpit’, ‘chancery’, and so on. The ‘members’ of these classes and their ‘combinations’ are called ‘connotators’. A ‘connotator’, Hjelmslev explains, is ‘an indicator which is found, under certain conditions, in both planes of the semiotic’ and thus can ‘never’ ‘be referred unambiguously to one definite plane’ (PT 118). He ‘views the connotators as content for which the denotative semiotics are expression’; this pair of ‘content’ and ‘expression’ therefore constitute ‘a connotative semiotics’ (PT 119).

6.53 Hjelmslev thus places a ‘denotative semiotic’, that is, an ordinary semiotic ‘none of whose planes is a semiotic’, alongside a ‘connotative semiotic’, that is, ‘a non-scientific semiotic one or more of whose planes’ is ‘a semiotic’ (PT 137f).¹⁷ The next higher order is a ‘metasemiotic’, that is, ‘a scientific semiotic one or more of whose planes’ is a ‘semiotic’.¹⁸ Next comes a ‘semiology’, being a ‘metasemiotics’ with a non-scientific’ (i.e. ‘connotative’) ‘semiotic as an object’; and a ‘metasemiology’ as a ‘meta-scientific semiotic’ with at least one ‘semiology’ for ‘an object’. Within this multi-order apparatus, ‘all those entities’ ‘provisionally eliminated as non-semiotic elements are reintroduced as necessary components into semiotic structures of a higher order’ (PT 127). Ultimately, ‘we find no non-semiotics that are not components of semiotics, and in the final instance, no object that is not illuminated from the key position of linguistic theory’ (cf. 2.8; 6.9f, 41; 12.9; 13.21).

6.54 ‘The metasemiology of denotative semiotics’ will ‘treat the objects of phonetics and semantics in a reinterpreted form’ (PT 125). ‘The metasemiotic of connotative semiotics’ will treat ‘sociological linguistics and Saussurean external linguistics’, including ‘geographical’, ‘historical, political’, ‘social, sacral’, and ‘psychological content-purports that are attached to nation’, ‘region’, ‘style’, ‘personality’, ‘mood, etc’. ‘Many special sciences’, notably ‘sociology, ethnology, and psychology’, are invited to ‘make their contribution here’ (provided they don't mind working within ‘non-scientific semiotics’). Moreover, ‘metasemiology’ can provide the ‘description of substance’ excluded from linguistic theory’ -- by ‘undertaking a self-consistent, exhaustive, and simplest possible analysis of the things’ of ‘content’ and of ‘the sounds’ of ‘expression’ (PT 124, i.r.; cf. 6.29f; 13.24). Sounding unexpectedly like Pike, Hjelmslev says this can be done ‘on a completely physical basis’ (cf. 5.27). Perhaps ‘the analysable continuum’ of ‘zones’ in the ‘phonetico-physiological sphere of movement’ could be studied with ‘a sufficiently sensitive experimental-phonetic registration’ (PT 54, 82) (cf. 4.28, 4¹⁰; 7.20; 8³³). But how ‘the continuum’ of ‘zones of purport’ for ‘the system of content’ can be studied on a ‘physical basis’ is hard to conceive; the ‘colour spectrum’ Hjelmslev uses as an example (PT 52f) is too orderly to be representative (cf. 4.22; 5.68; 7.31, 71).

6.55 ‘Usually, a metasemiotic’ is ‘wholly or partly identical with its object semiotic’ (PT 121). ‘Thus the linguist who describes a language’ ‘uses that language in the description’; the same holds for the ‘semiologist who describes a semiotic’ (cf. 13.48). ‘It follows that metasemiology’ ‘must in very great part repeat the proper results of semiology’, a prospect in conflict with ‘the simplicity principle’. So ‘metasemiology’ should be restricted to dealing not with ‘the language’, but with the ‘modifications’ or ‘additions’ entailed in the ‘terminology’ and ‘special jargon’ of ‘semiology’ (PT 121). ‘The task of metasemiology’ is ‘to subject the minimal signs of semiology, whose content is identical with the ulimate content- and expression-variants of the object semiotic (language), to a relational analysis’ through ‘the same procedure’ as ‘textual analysis’ (PT 123). The ‘terms for’ ‘glosseme-variations’ would be a major concern here (PT 122; but cf. 6.42).

6.56 During his discussion of semiotics, Hjemslev pays tribute to some predecessors who evidently influenced his thinking quite profoundly. Alongside ‘a semiotic whose expression plane is a semiotic’, he places ‘the logistic’ of ‘the Polish logicians’ like Alfred Tarski (1935) as a ‘metalanguage’ or ‘metasemiotic’ whose ‘content plane’ would be ‘a semiotic’; and declares that ‘linguistics itself must be’ just ‘such a metasemiotic’ (PT 119f, 109). ‘The logical theory of signs’ is derived from ‘the metamathematics of [David] Hilbert [1928a, 1928b], whose idea was to consider the system of mathematical symbols as a system of expression-figurae with complete disregard for their content, and to describe its transformation rules’ ‘without considering possible interpretations’. ‘This method is carried over by the Polish logicians into their “metalogic” and is brought to its conclusion by [Rudolf] Carnap [1934, 1939] in a sign theory where, in principle, any semiotic is considered as a mere expression system without regard for its content’ (PT 110f) (cf. Jorgensen 1937) (cf. 6.60, 64; 12.36; 13.17).

6.57 Another expansion of scope follows from the thesis that ‘the object of the linguist’ is not ‘the individual language alone’, ‘but the whole class of languages’, which ‘explain and cast light on each other’. So Hjelmslev calls for a ‘typology whose categories are individual languages, or rather, the individual language types’ (PT 106) (cf. 2.20; 3.47-54; 4.62; 7.19f). ‘It is impossible to draw a boundary between the study of the individual linguistic type and the general typology of languages’ (PT 126). ‘The individual type is a special case within that typology’, and ‘exists only by virtue of the function that connects it with others’ (cf. 6.25, 44). This proviso supports the thesis that ‘in the calculative typology of linguistic theory all linguistic schemata are foreseen; they constitute a system with correlations between the individual members’ (cf. 6.33). We can explore ‘differences between languages’ due to ‘different realizations’ not of ‘substance’ but of ‘a principle of formation’ applied to ‘an identical but amorphous purport’ (PT 77; cf. PT 56f; 6.31). ‘On the basis of the arbitrary relation between form and substance’, the ‘same entity of linguistic form may be manifested by quite different substance-forms as one passes from one language to another’ (PT 97, 103; cf. 2.28ff). Or, we may look into ‘contacts between languages’: either ‘loan-contacts’ or ‘genetic linguistic relationships’ which ‘produce linguistic families’ (PT 126) (cf. 2.42, 76; 4.73; 5⁴⁷).

6.58 An appealing prospect for so broad a framework is to apply it to science itself, and Hjelmslev foresees this opportunity (cf. 12.12f; 13.48). ‘Under the analysis of the sciences linguistic theory must come to contain within itself its own definition’ (PT 98) (cf. 13.36). So the terms applied to the language could be turned back on the theory as well (cf. 8.33; 9.27). For instance, ‘the distinction between process’ versus ‘system’ (6.34) might be pictured as that between the ‘both-and’ or ‘conjunction’ ‘in the process or text’, versus the ‘either-or’ or ‘disjunction’ ‘in the system’ (PT 36). Or, ‘the concept of syncretism’ ‘reached from internal linguistic premises’ might help us ‘attach a scientific meaning to the word ‘concept’ itself, and might ‘cast light’ on ‘the general problem of the relation between class and component’ (PT 92f). Or again, ‘an analysis of logical conclusion’ as a ‘linguistic operation’ could treat it as ‘a premised proposition’ wherein the ‘syncretism which appears as an implication’ is ‘resolved’ (cf. 6.10, 41, 49).

6.59 I hope my survey has conveyed some of the breadth and variety of Hjelmslev's theoretical concerns. In a lecture where he identifies himself as a ‘linguistic theoretician’, he remarks that such persons have ‘very abstract aims’ and ‘overwhelm their audience with definitions and with terminology’ (EL2 103; cf. 7.89). The remark was certainly apt; in RT, he presents formal definitions of no less than 454 terms, only a small fraction of which I have mentioned. Many are brittle neologisms scarcely found elsewhere in linguistics, such as ‘ambifundamental exponent’ or ‘heterosubtagmatic sum’ (RT 177, 198) (cf. 6.42). Even the most familiar terms receive unwonted definitions. A ‘word’ is a ‘sign of the lowest power, defined by the permutation of the glossematies’ (‘extrinsic units’) ‘entering into it’; a ‘noun’ is ‘a plerematic syntagmateme’; a ‘verb’ is ‘a nexus-conjunction’; an ‘adjective’ is ‘a syntagmateme whose characteristic is a greatest-conglomerate of intense characters’ (sounds to me like a linguistics department); an ‘adverb’ is ‘a pseudotheme that is not a connective and that does not include converted taxemes or converted varieties of ambifundamental taxemes’; and so on (RT 202, 99, 206f, 209) (cf. 13.7). Significantly, ‘phrase’, ‘clause’, and ‘sentence’ are not defined at all, nor are ‘meaning’, ‘reference’, and the like.

6.60 Managing so vast an apparatus would be a considerable task. The definitions interlock and cross-refer in such meticulous ways that we would have to either memorize them all or keep looking them up. Nearly every one is accompanied by a formal symbol, but almost none by an example (cf. 6.25, 42). We are again reminded how ‘the naming’ of ‘the “algebraic” entities’ ‘is arbitrary’ in that they ‘do not at all involve the manifestation’ (PT 97; cf. 6.15, 18, 29). But the motto that ‘all terminology is arbitrary’ (PT 58) would certainly need qualifying as soon as we confront a manifestation and try to assign it to one category rather than another (13.27). To the extent that Hjelmslev's ‘algebra’ of terms and symbols is indeed free of all manifestation, it comes close to being no ‘semiotic’ at all -- since the step of ‘encatalyzing a content’ is always deferred -- but ‘a symbolic system’ like those propounded in ‘metamathematics’ and ‘logic’ (cf. PT 110, 113; 6.56). And although Hjelmslev wants to ‘reckon with the possibility of certain sciences not being semiotics’ ‘but symbolic systems’ (PT 120n), it is hard to imagine linguistics being such a one.

6.61 He claims that the ‘names’ of the ‘entities’ are also ‘appropriate’ because they help us to ‘order the information concerning the manifestation in the simplest possible way’ (PT 97). But the claim is premature until we have a reasonable corpus of demonstrations, and his network of terms and rules could hardly be applied in any simple way. He himself is plainly reluctant to venture beyond the preparatory stage. A paragraph was ‘added to the Prolegomena’ in 1960 ‘as a warning’ ‘not to confuse the theory’ with any ‘application’ or ‘practical method (procedure)’ (RT xiii; PT 17). Yet the fact that ‘no practical “discovery procedure”‘ is ‘set forth’ does not impair Hjelmslev's confidence that his ‘theory will lead to a procedure’ (PT 17) (cf. 7.7). Nowhere in his two volumes on theory nor in his two volumes of essays does he actually analyse or describe a text in any detail. Aside from isolated words and phrases, he brings up only a handful of sentences or utterances (‘enonces’) (PT 50f, 56, 94; EL1 66, 156, 158ff, 172, 177, 199, 247; EL2 249), and none of these is treated in any remotely exhaustive manner.

6.62 Hjelmslev seeks ‘the object of science’ in ‘the registration of cohesions’; ‘science always seeks to comprehend objects as consequences of a reason or as effects of a cause’ (PT 83f) (cf. 13.11). Only when ‘the analysis is exhausted’ must ‘clarification by reasons and causes’ ‘give way to a purely statistical description’ (PT 125). He believes this to be ‘the final situation’ of ‘deductive phonetics’ and ‘physics’, the latter perhaps being the non-causal quantum theory prominently developed at his own university in Copenhagen (cf. 12.59). But his own apparatus of rules and definitions makes no provision I can see for assigning reasons or causes to entities. The indeterminacy of quantum phenomena might well be analogous to that of the content plane in general (cf. Beaugrande 1989a; Yates & Beaugrande 1990).

6.63 Still, no one could fail to be impressed by the range and rigour of Hjelmslev's thought within the bounds he sets. His proposals are put forth only in the anticipation of a beginning and their abstractness and difficulty helps make us appreciate the vast domains to be covered. His ‘test’ for the ‘success’ of a ‘linguistic theory’, namely to ‘draw all possible consequences from the chosen structural principle’ (PT 20; 6.11, 18, 20, 33, 36, 38), promises to keep researchers busy for a long time. Equally vast is the utopian prospect of the ‘unrestricted text’ ‘capable of being prolonged through constant addition of further parts’, the grandest instance being an entire ‘living language taken as text’ (PT 42; cf. PT 45, 83) -- whereupon the dualism of ‘process’ and ‘system’ (6.34) would yield to total unity. Finally, ‘the general typology’ for ‘the whole class of languages’, including even ‘virtual’ ones (PT 126, 106; 6.11, 35, 57) is another imposing challenge.

6.64 In a 1948 lecture, Hjelmslev quotes a letter from Bally, ‘the successor of Saussure’, saying: ‘You pursue with constancy the ideal formulated by F. de Saussure in the final sentence of his Cours’, namely that ‘the true and unique object of linguistics is language studied in and for itself’ (EL1 31; CG 232) (2.9). Also cited with warm approval is Carnap's motto that ‘all scientific statements must be’ ‘about relations without involving a knowledge or a description of the relata themselves’ (EL1 32; cf. 6.44; 6.56). Within this constellation of allegiances, Hjelmslev was certainly consistent and, in his way, quite radical in ‘seeking an immanent understanding of language as a self-subsistent, specific structure’ and ‘seeking a constancy inside language, not outside it’ (PT 19) (13.25). ‘A temporary restriction’ is needed to ‘elicit from language itself its secret’ (PT 127). Ultimately, however, ‘immanence and transcendence are joined in higher unity’. Then, ‘linguistic theory’ can ‘reach its prescribed goal’ by ‘recognizing not merely the linguistic system in its schema and its usage, in its totality and its individuality, but also man and human society behind language, and all man's sphere of knowledge through language’.

NOTES ON HJELMSLEV

¹ The key for Hjelmslev citations is: EL1: Essais linguistiques (1970); EL2: Essais linguistiques II (1973); PT: Prolegomena to a Theory of Language (1969 [1943]); and RT: Resume of a Theory of Language (1975 [1941-42]). Most of this material was translated from the Danish by Francis J. Whitfield. French sources are cited in my own translation.

² Though usually in agreement with Saussure, Hjelmslev does not distinguish between ‘speech’ and ‘language’ (cf. 2.20). But he makes two analogous distinctions: between ‘usage’ and ‘schema’, and between ‘process’ and ‘system’ (6.33f).

³ Hjelmslev's censure includes even the phonology of the schools in London (Daniel Jones) and Prague (N.S. Trubetzkoy), who were neither humanistic nor philosophical (cf. 6.43).

⁴ Even Bloomfield (1926) and Buhler (1933, 1934) are included for having proposed a ‘system of axioms’ for ‘transcendent kinds of linguistics’ (PT 6, 6n). In fact, however, Bloomfield rebuked the ‘philosophical’ trends in language study (4.4ff, 19, 38, 51, 72; 6.13; 13.16).

⁵ Contributors to this ‘beginning’ are named: Saussure (1879), Sechehaye (1908), Bloomfield (1933), Trager (1939), Vogt (1942), Bjerrum (1944), and Kurylowicz (1949), along with Hjelmslev himself and his collaborators Uldall (1936) and Togeby (1951) (PT 79nf). Hjelmslev gives several statements of ‘the main task’ (compare 6.11, 22).

⁶ Hjelmslev stipulates that ‘there is no experience before one has described the object by application of the chosen method’; ‘only after the method has been thoroughly tested can experience be obtained’ (EL2 103). But how could ‘experience’ and ‘theory’ then be ‘independent’? And how can we invent a method before having any experience of the ‘object’ to be described (cf. 7.28)?

⁷ Hjelmslev warns that he is using ‘induction’ in ‘a quite different meaning’ from ‘logical argument’, but is using ‘deduction’ in the usual ‘sense’ of “logical conclusion”‘ (PT 32).

⁸ Hjelmslev thinks the ‘invention of alphabetic writing’ was an early result of an ‘analysis into expression-figurae’ (PT 67; cf. 6.48; 8.71).

⁹ The ‘etymological meaning of the word “function” is its “real” definition’, but Hjelmslev ‘avoids’ ‘introducing it into the definition system, because it is based on more premises than the given formal definition and turns out to be reducible to it’ (PT 34). Elsewhere, logic is criticized for ‘neglecting’ ‘the results of the linguistic approach to language’ and thus attaining a ‘sign concept’ ‘unmistakably inferior to that of Saussure’ by not understanding that ‘the linguistic sign is two-sided’ (EL1 33).

¹⁰ In another paper, Hjelmslev proposes to divide the ‘langue’ side into three: ‘schema’ (‘pure language form’), ‘norm’ (‘material language form’), and ‘usage’ (‘the ensemble of habits’) (EL1 72). He says the ‘parole’ side is ‘as complex as that of the langue’, but he declines to ‘conduct an analogous analysis’ (EL1 79).

¹¹ ‘Determine’ is an action performed by a ‘variable’ on a ‘constant’ (PT 35). Since Hjelmslev says ‘the system is the constant’, his statement that ‘the process determines the system’ (PT 39, i.r.) makes more sense than this one here.

¹² These three terms are part of a scheme created because in ‘some cases’ ‘the difference between process and system is only a difference in point of view’ (PT 25). ‘Interdependence between terms in a process’ is ‘solidarity’, and one ‘in a system’ is ‘complementarity’ (e.g. between ‘vowel and consonant’) (PT 24ff, n, 41). ‘Determination’ ‘in a process’ is ‘selection’ (some ‘have long been known under the name of government’, 4.66) and ‘in a system’ ‘specification’. ‘Constellations in a process are ‘combinations’ and ‘in a system’ ‘autonomies’.

¹³ This passage seems to count ‘clauses’ and ‘sentences’ as units of ‘content’, quite unlike the usual explicit practice in linguistics.

¹⁴ Instead of stating these ‘precisely fixed conditions’, Hjelmslev says ‘the problem of identity’ can ‘be dismissed’ ‘as an unnecessary complication’ (PT 61n). He refers us to Saussure, who raised the problem (e.g. CG 43, 91, 107f, 161, 181, 186) but certainly didn't solve it.

¹⁵ This conclusion is said to hold even when ‘the analysis’ is seen ‘from the point of view of a metaphysical theory of knowledge’, though elsewhere, the ‘metaphysical’ view is claimed to rely on ‘substance’ and ignore ‘the functional net of dependences’ (PT 22, 81).

¹⁶ To limit the bookkeeping to interpolations of less than a whole sign, Hjelmslev juggles his idea of ‘function’: if ‘the encatalyzed entity’ is of ‘content’, it ‘has the expression zero, and if it is’ ‘of expression’, it ‘has the content zero’ (PT 96). The silent ‘“-d” in French “grand”‘, which becomes audible in ‘“grand homme”‘, is used as evidence that ‘latency is an overlapping with zero’ (PT 93) (cf. 2²⁶; 4³; 5¹²).

¹⁷ ‘Non-scientific’ means here that the ‘semiotic’ ‘is not an operation’, an ‘operation’ being ‘a description’ ‘in agreement with the empirical principle’ (PT 120, 131, 31, 138; RT 14). Even so, it is hardly a tactful term.

¹⁸ Elsewhere, though, the ‘metasemiotic’ is allowed to have only a ‘content plane’ as its ‘semiotic’ (PT 114, 119). ‘The Polish logicians’ are cited for having prepared the way to such a construction (PT 119; cf. 6.56).