1. The Two Basic Subsystems of Language

Language is made up of two fundamental types of signs, on the basis of which it is subdivided into two systems, the expression system (sounds) and the content system (meanings). The expression system encompasses all of the phonological signs, phonic elements and their combinations, which serve as vehicles for the physical (sensible, perceptible) manifestation of human speech. The content system, a structure constituted by the semantic categories, encompasses all of the grammatical and lexical signs. These content signs enter into oppositions constituted by semantic features, their signata. This ‘twofold articulation’ of language has become a commonplace of structural linguistics, but the precise semeiotic relation between the two planes needs to be emphasized (following Jakobson 1939/1962 and Andersen 1979) in any theory of grammar that aspires to the explanatory understanding of linguistic phenomena. It will in fact be seen that the structure of the expression system ‘presents a clear mirror image of the content system of language’ (Andersen 1979:377).

The expression system—the phonology—is unique among all human semeiotic systems: the signs of this system are identified solely on the basis of their signantia, owing to the fact that their signata are strictly synonymous (Jakobson 1939/1962). All phonological signs that are purely diacritic, i.e., that serve the strictly linguistic ends of a phonology and are differentiated from physiognomic and pragmatic signs (Andersen 1979: 377; more about these below), have the same signatum—’otherness’ or ‘alterity’ (Jakobson 1939/1962). These diacritic signs making up the core of any phonology have the function of differentiating the signantia of the content system. The central place of this function of diacritic signs transpires from the fact that content signs have, for the most part, no signantia of their own apart from those of the diacritic system. With the exception of such cases of direct expression of grammatical meaning as the opposition between declarative and interrogative sentences expressed by a simple difference in intonation contours (rising vs. falling, respectively, in English and many other languages) or the rendering of the opposition exact vs. approximate in Russian by word order (e.g. pjat’čelovek ‘five people’vs. čelovek pjat’‘ca. five people’), meanings can be expressed in language only by recourse to units of sound which have no positive meaning of their own. Each signatum of a phonological signans is definable only as ‘otherness,’ which is to say that its meaning is equal to its diacritic function. There is thus a fundamental asymmetry between content signs and diacritic signs. The former typically have no signantia of their own and must resort to those of the expression system; the latter all have no individual, positively definable signatum and are, therefore, strictly synonymous via their shared signatum of otherness. Content signs form oppositions, which make up the system of meanings in language, strictly on the basis of their signata; diacritic signs, on the other hand, form oppositions, which make up the system of sounds in language, strictly on the basis of their signantia (cf. Andersen MS:20).

2. The System of Phonological Signs

As was mentioned above, the phonology of a natural language contains phonic signs other than the diacritic signs of the expression system proper whose function is the differentiation of the signantia of the signs comprising the content system. Those elements which convey information about a speaker’s age, sex, or specific identity—called physiognomic signs by Bühler (1934:286)—are indexes because they are ‘established by an existential relation between a signans token and the speaker, and not by conventional rules’ (Andersen 1979:377). They are not part of language sensu stricto, being particularized byproducts of the physical realization of speech, or phonetic sinsigns with a purely indexical function and an object (signatum) that are aspects of the utterer of the sign himself.

Besides physiognomic signs, there is a class of phonic elements that, while still not part of the class of diacritic signs, do belong to the expression system of language. These are what have been called pragmatic signs (Andersen 1979:377), although Peirce’s own usage would be better served by the related term practical sign (see chapter 1), i.e., the sort of sign whose interpretant is a Second considered in itself. Such signs are indexical symbols interpretable by reference to aspects of the total communicative system (Andersen 1974:18-9; 54, fn. 1). This system (as delineated in Jakobson 1960:352f., cf. 1980:81-92) comprises the several constituents of the total communicative context in which speech is situated, and the parallel functions defined by these constituents: the (a) emotive, (b) conative, (c) referential, (d) poetic, (e) phatic, and (f) metalingual functions. The phonological signs signal, correspondingly, (a) the speaker’s attitude toward or personal involvement in the speech event, or (b) his attitude towards the addressee and evaluation of the latter’s involvement, or (c) his evaluation of the content of his message; or these signs serve to (d) (re)present the message as an aesthetic object, or (e) establish or maintain communicative contact, or (f) disambiguate or gloss the structure of the code in which the message is communicated (cf. Andersen 1979:377, minus the metalingual function).

The signantia of the diacritic signs that form one part of the bipartite core of the phonological expression system are the familiar (Jakobsonian) distinctive feature terms, such as +nasal, —nasal, +voiced, —voiced, etc. These terms are each signs and as such enter into oppositions—i.e., binary paradigms—and a system of oppositions defined exclusively by their signantia, their sign vehicles. The signantia are realized phonetically as differences between sounds, and the paradigms to which they belong are defined by a coordinated set of audio-perceptual parameters (Andersen 1979:378). The diacritic signs themselves have no positive phonetic values. Owing to the relative and negative character of distinctive feature oppositions, these signs ‘can be realized with appreciably different concrete manifestations in different contexts and by different speakers’ (loc. cit.). The phonetic parameters associated systematically with the diacritic signs that constitute the ultimate units of phonology organize themselves into diacritic categories (Andersen 1978:11, fn. 4), so that, for example, the diacritic signs +voice or —voice are associated with the diacritic category voicing.

Saying that diacritic signs—specifically, each phonological term and its opposite—form binary paradigms means that the relation between them is one of mutual exclusivity: one term is in praesentia while the other is in absentia, and vice versa. This relation is what keeps the two signs distinct from each other. The definition of phonological paradigms via the signantia of diacritic signs is expressed in terms of the latter’s audio-perceptual parameters, and it is this distinctness that is presupposed in qualifying the signatum of all diacritic signs as ‘phonemic otherness.’

The semeiotic relevance of distinctness-within-opposition in phonological paradigms is that the relation between the opposed signs is one of asymmetry. Whether or not the phonetic realization of the two signs of a diacritic paradigm is strictly dichotomous (privative) or is graded, the relation between them is not one of equipollence but of polarity. The signantia of diacritic signs are expressed as polar specifications of the diacritic categories in question, as plus or minus the given opposition. This asymmetry is the primary semeiotic relation holding between the signs; secondarily, and owing to the binariness of the paradigm, there is also a relation of symmetry between them (cf. Andersen 1974:892). Each of the two relations has its consequences for the sound pattern of any language.

Paradigmatic asymmetry is another, more general characterization of the strictly relational content of diacritic signs. One of the two signs in a paradigm is more narrowly defined than the other. The first is the marked, the second the unmarked member of the paradigm. ‘More narrowly defined’ means of greater conceptual complexity, so that the marked sign is conceptually more complex than its unmarked counterpart. In Trubetzkoy’s original formulation, markedness was associated with aspects of phonetic substance such as the presence of a specific articulatory effort connected with the production of the marked term of a phonological opposition. Understood in a semeiotic light as the interpretant of the diacritic sign, markedness can now be properly seen as belonging to linguistic form rather than to substance, of which it is largely independent while being systematically interrelated with it (Andersen 1974:893).

That the two interpretants of the signs of the minimal paradigm differ in value, one having the opposite value from the other, transpires from the ontological status of opposition in general. Whereas in mathematics and logic opposition may not entail differences in value of the opposed terms, in human semeiosis it does so without fail: perhaps in consonance with biological and neurophysiological factors (such as brain lateralization), it appears that human beings are unable to integrate signs into paradigms (and syntagms, as will be clear below) except by grading and ranking them (cf. Andersen 1974b:44). All conceptualization necessarily involves evaluation.

It is the evaluative focus of markedness that confirms its independence of substance and its primarily conceptual (formal) focus. The paradigmatic asymmetry of oppositions in the content system, and the syntagmatic asymmetry evident in both the expression and the content systems, are all of a piece; they could not be compared and integrated with phonological markedness were it not for the uniformity inherent in all manifestations of markedness (cf. Greenberg 1966:56f., Andersen 1979:379).

Whereas in grammar and in lexis there can be either asynthetic or synthetic signata—that is to say, a grammatical or lexical morpheme may have one signatum or more than one, respectively—diacritic signs not only have only one signatum (are exclusively asynthetic) but, conversely, can only appear as one of a set of signantia (cf. Andersen 1973:769). That is what is meant by the statement that ‘terms of phonological oppositions do not occur in isolation, but are obligatorily combined into syntagms of various extents’ (Andersen 1974a:892). No signans of a diacritic sign can occur singly; each such sign requires a synthetic domain in which it is joined by other, co-occurrent signantia. These simultaneous syntagms, the smallest simultaneous sign complexes realizable in language, are what is known under the conventional name phoneme.

Contrary to their long-standing (mis)characterization as mere ‘bundles of distinctive features’ (as in Jakobson & Waugh, 25, only the latest in a long series going back to Bloomfield 1933:79), phonemes are organized into hierarchical structures; their immediate constituents—the diacritic signs—enter into relations of subordination (rank order). In this respect phonemes are in full conformity with the chief principles of organization of their isomorphous counterparts in grammar and lexis. However, unlike the latter signs, in whose structure hypotaxis (subordination) and parataxis (coordination) may coexist as possibilities, the constituents of diacritic syntagms may only occur as hypotactic structures in which rank order is determinative in assigning relative importance to any pair or set of signs (Andersen 1979:379).

Phonemes are hierarchical structures, and this means that the entire inventory of phonemes is itself an overarching hierarchic structure. The ranking of diacritic signs in the simultaneous syntagm is the syntagmatic counterpart—the asymmetry—of markedness, since markedness is the asymmetry (as we saw) of paradigmatic relations. Although markedness relations and ranking relations are both universally constitutive principles of organization of all languages, it appears that ‘ranking to a much larger extent than markedness is a purely conceptual operation, independent of linguistic substance, and hence rank relations are determined on a language particular basis’ (Andersen 1979:379; cf. 1973, 1974, 1975, 1978).

The relations which characterize the hierarchy of diacritic signs in phonology—the phonological interpretants—are immanent in the sound system and must be realized in speech in order to attain the level of actual palpability. The set of regulative modes of their systematic realization in speech chains can be expressed in the form of illative transformations, conventionally known as rules. These rules are tantamount to principles of structural organization, a definition of their function which extends beyond the bounds of phonology to embrace all of language. The interpretative nature of such rules or principles transpires from the fact that, as interpretants do generally, they assign a means of representation to an object of representation (cf. Andersen MS: 5, 8 et passim). In this respect such rules conform perfectly to the illative or inferential form of explanatory hypotheses, of the sort characteristically present in scientific theories and all generalizations typically involved in learning-about-the-world. Moreover, they replicate the illative relation holding between the signatum and the signans of a sign. In exactly the same way that signification results from the patterned conjunction of signata and signantia via the mediation of the all-important Third, the interpretant, so all of linguistic content is rendered manifest in the structure of language by series of interpretants that have the form ‘If content A, then expression B.’

The realization of diacritic signantia in speech is the function of an interpretant or ‘realization’ system, which

serves to transform a phonemic representation, couched in the oppositive, negative, and relative terms of diacritic signs into a phonetic representation with sufficiently specific phonetic detail to be realized—though still couched in relative terms. The output of the realization system is merged with (superimposed on or modified by) the expression elements of the content system and of the pragmatic phonic signs. (Andersen 1979:380)

It would be wrong to take this characterization in a way which would have rules ‘apply’ to ‘outputs’ in the mechanistic sense which has permeated contemporary linguistic methodology. In the hermeneutic theory of grammar being set forth here, rules are signs of mediation which relate the ‘underlying’ structure to that of actual speech. No mutational force should be attached to rules of this kind, and the word ‘transform’ must be understood in a strictly relational sense, as a handy synonym for the Peircean ‘translate’—as in his definition of meaning: ‘the translation of a sign into another system of signs’ (4.127).

There are several aspects to the translation of diacritic signs into the directly observable phonetic data of speech. If we adhere to the common usage of linguists and continue to call such translations rules, the primary function of the rules is the translation or transformation of the system of diacritic signs, ‘representations couched in purely oppositive terms’ (Andersen 1979:378), into phonetic representations realized as differences in sound. Implementation rules (as they have been called; see Andersen 1979) of this type fulfill, therefore, a phonetic function: the assignment of phonetic properties to diacritic signs. The ensemble of such properties constitutes many of the physical details of the pronunciation of a given language or dialect, which are recognizable as differences contributing to the differentiation of varieties of speech. These include the so-called ‘low-level’ features (also termed allophonic or ‘redundant’) which despite their lesser significance vis-à-vis the diacritic system have important functions in communication. Among the latter is what Trubetzkoy called the ‘sociative function’ (1958: 46 f., cit. Andersen MS:27), by which allophonic features serve as indexes of immediately contiguous diacritic signs in the speech chain.

Secondarily, there exist two kinds of implementation rules which discharge functions other than the purely phonetic, and these have been called neutralization and variation rules (Andersen 1973:785). Neutralization rules suspend oppositions. They are reductive of communicative information in the sense that they remove the possibility that one member of an opposition of diacritic signs will appear in a particular context. In Russian, for instance, in position before pause and before sonorants or voiceless obstruents (true consonants), the signs +voiced and —voiced which otherwise differentiate, say, rod ‘clan, gender’ from rot ‘mouth,’ cease to be opposed: the two words are realized identically with the final voiceless obstruent t. The suspension of an opposition—a relation on the paradigmatic level—effected by a neutralization is mirrored on the syntagmatic level by the removal of the possibility of contrast between two forms differing only in their manifestation of the two terms of a phonological opposition; this mirroring relationship between opposition and contrast exemplifies the polarity principles of linguistic structure (Jakobson & Halle 1971:15). Neutralization in grammar as a whole (not just phonology) is an important semeiotic process, about which more will be said below. At this point, note should be taken of the fact that it deals exclusively with diacritic signs, i.e., with those signs that have the capability of distinguishing the signantia of content signs.

Variation rules, on the other hand, ‘assign allophonic features in complementary distribution’ (Andersen MS:28); they flesh out, as it were, the simultaneous syntagm that constitutes the phoneme with non-phonemic signs, so that ‘different values for a given phonetic parameter occur in complementary distribution’ (Andersen 1979:380). For example, vowels in Russian are specified as (occur with the relative phonetic value of) [+ front] before palatalized consonants, but as [— front] before unpalatalized consonants and before pause; hence the characteristic and completely predictable (non-diacritic) fronting of /e/ in words like šest’ ‘six’ as opposed to šest ‘stake.’ This species of complementation is an important heuristic clue in linguistics—not to speak of the process of language acquisition—since it allows the linguist (and learner) to discover which features are diacritic (phonemic) and which are not. In the Russian example, the complementary distribution of relatively more and less fronted vowels shows that fronting is not phonemic in this language.

The semeiotic nature of the relation between the system of diacritic signs and their manifestation in speech via rules/interpretants was first discovered by Andersen (1966) and is summarized succinctly in the following statement (1979:380):

As they apply to diacritic signs implementation rules do not assign positive phonetic values to individual diacritic signs, but rather render paradigmatic oppositions as relative phonetic differences, which remain constant despite contextual variation. Since these rules transform one kind of relation (phonemic oppositions in the diacritic system) into another kind of relation (phonetic differences in realization), they establish diagrammatic signs.

In terms of the kinds of function that rules may fulfill, and remembering that, in Peirce’s semeiotic scheme, diagram is a species of icon, implementation rules which do more than simply specify or constate allophonic particulars can be said to have an iconic function (Andersen 1966). Of course, the phonetic function of such rules is no less iconic or semeiotic than that of neutralization and variation rules. For if we understand the relation between interpretative rule and diacritic sign to be essentially one of form to substance, then its diagrammatic nature is presupposed in the very notions of these terms: ‘the relation between a form and the stuff formed by it is of necessity diagrammatic’ (Andersen MS:28). When fulfilling no more than a phonetic function, implementation rules are still instruments of mediation or representation between relations of one kind and relations of another, since even in this role they translate phonemic oppositions into phonetic differences. That is to say, rules of this sort still deal with relative terms rather than with phonetic absolutes.

Given Peirce’s tripartition of the relation between a Sign and its Object into Icon, Index, and Symbol, it becomes apparent that the several types of phonological rules discussed so far conform to one or another of these three species of sign. The implementation rules are (as we saw) diagrams, a kind of icon. Rules which deal with signs that make reference to immediately contiguous diacritic signs (Trubetzkoy’s ‘sociative’ function) are themselves indexes. And the ‘pragmatic’ signs discussed earlier—that is to say, the phonic signs which refer to one or another aspect of the total speech situation (speaker, addressee, referential context, message, communicative contact, and code)—are (indexical) symbols (cf. Andersen 1979:377).

The role of symbolic semeiosis in phonology is not limited, however, to pragmatic signs.

Besides implementation rules, a realization system comprises a set of adaptive rules, whose function is to adjust the speaker’s pronunciation to the norm of his speech community in accordance with his status and roles. Unlike implementation rules, which diagram relations in the diacritic system, adaptive rules produce phonic signs which symbolize the speaker’s relation to his speech community. (Andersen 1979:381; cf. 1973:773, 781 f.)

Because all interpretants represent the relation of Sign and Object, giving significance to signification, it is not surprising that implementation rules are diagrammatic. Indeed, were this not so, no learner of a language would have access to the relations defining the diacritic system, and the language would be neither learnable nor perpetuable. With respect to linguistic change, the implementation rules are teleologically consistent with the defining relations of the system, which is to say that changes affecting these rules can be counted as genuinely evolutive whereas adaptive changes reflected by adaptive rules can do no more than symbolize the speaker’s relation to his speech community (Andersen 1973a:785-6). Being coordinate with evolutive change, implementation rules are at the center of the phonological system whereas adaptive rules, lacking the intrinsic structural motivation of implementation rules, are at its periphery and consequently subject to gradual elimination over time. In this respect, adaptive rules are like morphophonemic rules (see chapter 4), that is to say they are relatively unproductive (cf. Andersen 1973a:782). Adaptive rules may differ from morphophonemic ones, however, in representing ‘simultaneous and sequential phonetic feature values which are not provided for in the phonological structure’ (loc. cit.). Their essentially ad hoc nature almost inevitably portends their elimination from the language.

There is a more specific semeiotic conditioning involved in what has been characterized as the iconic function or diagrammatic character of implementation rules, which makes the systematic purport of the interpretant system more palpable (to both linguist and learner alike). Variation rules (as was said) translate the relations of the diacritic system into phonetic differences in their realization, in accordance with the principle of complementation. But the ‘relations’ between diacritic signs, and the particularity (cf. Andersen 1973:769, fn. 5) within individual signs between signantia and signata (despite the latter’s synonymousness), are more than the abstraction associated with this word: they are markedness relations. Hence the diagram produced by phonological interpretants at the syntagmatic level are in fact supplemented by a further diagram, in which equivalence in markedness (the paradigmatic level) is reflected as ‘relations of contiguity in phonetic realizations’ (Andersen 1979:381). Complementary phonetic entities are thus correlated with specific contexts.

Perhaps the most fruitful and important discovery in recent phonological research has been that phenomena of the kind subsumed by variation rules conform to a universally valid semeiotic principle, which has been called markedness assimilation (Andersen 1968:175, 1972:44-5). According to this principle (which, as a matter of fact, extends beyond phonology to embrace all human semeiosis; see below), the normally unmarked value for a given feature occurs in an unmarked (simultaneous or sequential) context, and the normally marked value in the marked context.

In conventional phonological terms, the discovery of markedness assimilation has a direct effect on the understanding of what assimilation means. Since markedness is a formal semeiotic universal, explanations couched in its terms make reference to conceptual entities, rather than to physical (phonetic) factors. Assimilation is thereby 1) taken out of the oftentimes capricious realm of articulatory ‘accomodations’ and the putative propensities of speech organs (à la Martinet 1955); and 2) aligned with completely parallel processes in grammar and lexis, where a physical basis is excluded by the very nature of the matter.

The first examples of markedness assimilation (offered in Andersen 1968) were limited to the distribution of allophonic features in complementary simultaneous contexts. For instance, the Japanese rule (cf. Andersen 1972:44) which assigns a [+ strident] realization (the normally marked value for that feature in abrupt obstruents) to dental stops before [+ diffuse] vowels (i.e., the normally marked value for that feature in vowels) is recognized as a markedness assimilation rule: [č] and [c] are marked for stridency, as are [i] and [u] for diffuseness. The [—strident], i.e. [U strident], realizations occur—complementarily—before vowels which are [—diffuse], i.e., [U diffuse].

The common parlance of linguistics makes it easy to fall into the habit of speaking of such relations between sounds and contexts as mutational, as if there were a ‘change’ in the synchronic grammar of Japanese (restating what was obviously true of historical Japanese) whereby dental stops were permuted to strident affricates before diffuse (high) vowels. But the theory of grammar being sketched here ought to prevent the analyst from falling into such a methodological trap: what is central in the Japanese case is the alternation of the sounds in certain contexts. It is true that one particular alternant may be less narrowly defined, hence basic and unmarked. But the point is that the structural coherence of the phonological system at any stage of its development is reflected precisely in the patterned cooccurrence of units and contexts, in the relational correspondences of interpretants. The distribution of strident affricates before diffuse vowels and dental stops before nondiffuse vowels represents, in a Peircean diagram, the semeiotic equivalence of phonetically disparate units, on the one hand, and their respective coherence with the marked/unmarked semeiotic values of the context, on the other.

Since its promulgation more than a decade ago, it has become clear that markedness assimilation is far from being limited to the status of ‘the effect of a universal constraint on the combination and concatenation of phonetic features’ (Andersen 1972:44). Markedness assimilation is the principle of structure governing a whole spectrum of phenomena in language. In phonology, it extends to diacritic (phonemic) signs as well as to non-diacritic. Thus, in positions of neutralization the realization of a neutralized opposition is determined by the sequential context. Trubetzkoy (1936) had thought that it was normal for the unmarked member of an opposition to appear in positions of neutralization, and he was at a loss to explain the exceptional cases of the appearance of the marked member. For example, in Russian, obstruents are realized as [+ voiced] (the marked value) before voiced obstruents, but as [— voiced] in other positions of neutralization—before voiceless obstruents, and before word boundaries followed by segments unmarked for voicing (vowels, sonorants, and voiceless obstruents) or by pause. In cases of neutralization where the marked value appeared—voiced obstruent before voiced—Trubetzkoy did not perceive the essential parallelism between this apparently aberrant result and what he considered normal because he made a basic distinction between ‘externally’ and ‘internally’ conditioned neutralization, i.e., between assimilation explicable by the sequential context and neutralization (as before pause in the Russian case) not so explicable. The absence of sound associated with pause is, of course, only a special variety of context, and as a ‘zero sign’ is conventionally identified with the unmarked value (Jakobson 1939; but cf. Shapiro 1972:357).

What is important semeiotically about neutralization is the relation between the representative sound (or form) manifested in positions of neutralization and the system of units to which markedness as an interpretant makes reference. Specifically, while a rule of the Russian type which eliminates the possibility of contrast between voiced and voiceless obstruents before pause ‘does not have any positive output’ (Andersen 1979:381), the very fact that a particular diacritic paradigm is neutralized ‘may diagram its rank in the hierarchy of diacritic signs’ (loc. cit.). An example of this is the phenomenon of vowel harmony, of which different types may be said to reflect differences in rank between the diacritic categories involved in vowel systems (Jakobson 1971:108-9).

Positions of neutralization are also linked with variation in the sense that ‘neutralization rules provide contexts in which variation rules may assign diagrammatic representations of the markedness values of diacritic paradigms’ (Andersen 1979:381). More concretely, a variation rule seen against the background of markedness assimilation may tell both the learner and the analyst which category is diacritic or distinctive in a language, which non-diacritic or redundant. A good illustration of this role of markedness assimilation is furnished by the relation between syllable peaks and contiguous obstruents in English, as it bears on the problems of quantity in vowels and the phonological status of tenuis and media consonants, respectively (cf. Andersen 1979: 380-1). In English, syllable peaks are [— long] before tense obstruents but are [+ long] before lax obstruents, before sonorants, and in final position: hence beet is [bit], but bead, bean, and bee are [bi:d bi:n bi:]. The complementary distribution of the short and long vowel realizations is a sign (as was mentioned earlier) of the non-diacritic status of quantity in English. Beyond that, however, the assumption that the relevant tenues and mediae are distinguished by tenseness and not by voicing enables the variation rule to make sense as a case of markedness assimilation, since tense obstruents are marked relative to the unmarked lax obstruents. The shorter realization of /i/ in beet [bit] understood as an abridgement of syllable peaks before the [M tense] obstruent /t/ makes the variation coherent: the narrowly defined (marked) context of a contiguous tense obstruent is congruent with the marked (shorter) syllable peak. If voicing had been assumed to be the diacritic category involved, no markedness assimilation could be appealed to, since voiceless obstruents are [U voice] relative to the [M voice] status of voiced obstruents. An additional piece of evidence in support of this interpretation is the asymmetry of the contexts (Andersen 1979:381): since abridged vowels occur only in one context (that of tense obstruents), whereas unabridged vowels occur in all other contexts (before lax obstruents, before sonorants, and finally), the first context is itself the more narrowly defined of the two, and the coherence of markedness values is complete. Parenthetically, Trubetzkoy’s claim (1969:147) that ‘it is impossible to say whether in English a correlation of tension or a correlation of voice is present’ thus turns out to be groundless. Naturally, the resolution of what he perceived to be irresolvable hinges on the admission of circularity into the coherence of structure and the concomitant methodological devaluation of ‘independent motivation.’ The mutual dependency of the elements of the solution—the shorter realizations of the syllable peaks seen as abridgements rather than the longer ones as prolongations, the role of markedness assimilation, and the status of protensity as the relevant diacritic category—all cohere as an ensemble of conditions informing the phenomena in question. This sort of coherence, where units and contexts are evaluated in tandem, in a mutually dependent manner, is of the essence of linguistic structure, and will be recognized as such time and time again in this book (see especially chapter 3).

Implementation rules are diagrammatic: they make manifest in speech the relations that characterize the phonological structure of a language. But it is obvious that there are also pronunciations which reflect rules lacking the intrinsic, structural motivation of implementation rules. Thus, for example, New Yorkers, for whom beard, bared, and bad are strictly homonymous in spontaneous speech, are nevertheless able (albeit not consistently) to differentiate these lexemes when adherence to traditional norms is demanded (Weinreich, Labov & Herzog 1968:134-5, cit. Andersen 1973:782). Whenever an adaptation of this kind takes place, it is possible to speak of adaptive rules, by which speakers adjust their speech to conform to community norms in accordance with their status and roles. The synchronic goal of such rules appears to be that of ‘ensuring relative uniformity of usage, regardless of differences in phonological structure among the grammars of the speakers’ (Andersen, loc. cit.). In the long run such rules tend to be curtailed and eliminated, but at any given point in the history of a language, every phonology has a number of adaptive rules.

Adaptive rules produce conventional signs: they refer to the extra-linguistic reality of social norms broadly conceived (including style), and are therefore symbols rather than diagrams. Implementation rules differ from adaptive rules, therefore, in semeiotic function. Although adaptive rules are formulated in terms of the phonological structure sensu stricto and are subordinate to the latter, they form an additive system which is subject to elaboration and revision throughout speakers’ lives (Andersen 1973:781).

To recapitulate, the expression system of language consists of two basic subdivisions, the structured core and the adstructure. The latter has just been described. The former is comprised of a set of diacritic signs which form binary paradigms and a set of interpretants (rules) by which the signs are combined into hypotactic syntagms, on the one hand, and are manifested linearly in speech, on the other. In simultaneous combination, markedness values—the interpretants of phonological signs—partly determine the extent of syntagms and are themselves partly determined by general principles of organization. In sequential combination, markedness determines both the distributional particularities of (the variants of) diacritic signs and their coherence with the contexts in which they occur.

3. The System of Content Signs

The other basic subsystem of language structure is the content signs. These are signs whose signata consist of single content elements (semantic features) or of syntagms of content elements. Depending on whether content signs have simple or complex signata (i.e., one or more content elements), they can be divided into asynthetic and synthetic signs, respectively (cf. Andersen MS: 19). The typical content sign has no signans of its own: it relies instead on the signantia of the diacritic (expression) system. Its signata utilize (apart from cases already mentioned above) the signantia of the expression system, and it is the resultant inherent disjunction between content and expression that is overcome in language structure by the presence of an overarching interpretant system—i.e., by the rules that relate content to expression.

At the heart of the content system is the content sign, or morpheme, as it is traditionally called. Like all signs in language, morphemes consist of a signans, a signatum, and an interpretant. The interpretant, here as elsewhere, is the semeiotic relation between signans and signatum, which can be said to be established by a rule of the sort discussed earlier. The great variation observed in morphological systems is due to the manifold ways in which morphemes can differ—‘according to the kinds of signantia, the kinds of signata, and the kinds of signatum-signans relations they comprise’ (Andersen 1980:3).

Essential to any understanding of the systematic character of content signs is the notion of hierarchy, by which basic linguistic signs (whether the focus is on signata or signantia) are distinguished from sign variants (of signantia). The latter, traditionally called allomorphs, make up what is subsumed under morphophonemics. Morpheme alternants, when they are co-variants of absolute (basic) signantia, are semeiotically subordinate to these latter signs, to which they stand in the relation of ‘derived’ or secondary entities. Thus, in the typical Indo-European inflectional paradigm, the hierarchy of its constituents accords pride of place to the form of the stem in the maximally unmarked category, e.g. the nominative singular masculine in substantives and the infinitive in verbs (both of which not unexpectedly function as citation forms in dictionaries). In typical examples of morphophonemic alternation (Andersen 1980:4) which produce patterns of relative signantia, consisting in the modification of a real (non-zero) signans—consonant gradation, apophony, accentual alternations, reduplication, and element inversion—there is always a hierarchy among the alternants reflecting the hierarchy of signata of which these co-variants are the means of expression. In some cases, it is not clear (to the learners of a language, no less than to its analysts) whether an alternation is properly morphophonemic or morphological, whether the modification of the sign vehicle is not itself a signans rather than a co-variant of one.

What is important for the theory of grammar generally, and the content system particularly, is the fact that relations between (sets of) signata and signantia that constitute morphological and morphophonemic patterns have a raison d’etre that goes beyond mere regularity of cooccurrence. While it is true that all linguistic signs are fundamentally symbolic legisigns, the conventional aspect (their symbolicity) may be modified in the direction of iconicity and indexicality. The indexical nature of the so-called shifters (categories of deixis in one form or another) is one example of the attentuation of the arbitrariness of the linguistic sign (cf. Jakobson 1971:130-47). A much more significant matter is iconicity, embodied in the ‘extensive patterns of similarity and difference among the shapes of grammatical morphemes which correspond to the relations of similarity and difference among their meanings’ (Andersen MS:21). A typical example (cf. Jakobson 1971: 167 f.) is the correspondence between the opposition plural vs. singular and the relative number of segments in the case desinences of these two numbers in many Russian substantives: regardless of the specific shape of the desinences, each plural desinence is longer by one segment than the corresponding singular desinence. This distribution is of a piece with the most common of such patterns, the specialization of consonants (the larger phoneme class) as expressions of lexical meaning and vowels (the smaller phoneme class) as expressions of grammatical meaning, in such languages as the Semitic group (Trubetzkoy 1958:270, cit. Andersen MS:23). Correspondences like these between content and expression can be understood as diagrams (relations reflected by relations). The fact that signantia may diagram signata is tantamount to the principle by which a system of content signs is established as a structure. If not for these pervasive mapping relations between signata and signantia, the content system would amount to no more than an inventory of arbitrary signs, a purely additive system (like the alphabet or the Morse code).

The particular example of the longer desinences (case by case) of the Russian singular vis-à-vis the plural is, however, simultaneously characteristic of linguistic iconicity and somewhat misleading as to its true nature. In this easily generalized example the conceptual correspondence is between extent and number: the grammatical category of plurality is understood to be an iconic analogue of the (greater) length of the case desinence, a relation of quantity. This analogy is in alignment with such phenomena as the so-called ‘small vs. large symbolism’ (Jakobson 1971:199) imputed to certain phonological relations (cf. Jakobson & Waugh 1979:177 f.).

There is an inherent difficulty, however, in the ontological validity of comparisons of this kind. Given the asymmetry of the content and expression system (discussed in detail earlier in this chapter), there is no direct way of comparing form and meaning in language, and the only access to their congruence is via the interpretants—the units of mediation relating signantia and signata. There may be a kind of naive or intuitive truth to the notion that greater extent naturally reflects greater number, but the existence of counterexamples tends to undermine its validity. For instance, in Russian (to make the comparison homogeneous) the opposition within verbs of determinate vs. nondeterminate subaspect (in the small class of so-called verbs of motion) is expressed inter alia by shorter vs. longer stems. Now, the determinate is the marked member of this opposition, and the nondeterminate is correspondingly unmarked. The longer stem, therefore, appears consistently as the expression of the unmarked form, the corresponding marked form being expressed by the shorter verb stem. This relation between stem length and grammatical meaning is, however, directly counter to that of desinence length and plural number in the example cited earlier: the marked plural form is longer, the unmarked singular correspondingly shorter.

In the face of this discrepancy, a more scrupulous attention to the varieties of semeiosis encompassed by the class of icons becomes important. It is to be recalled that Peirce divided icons into three subspecies—images, diagrams, and metaphors. More precisely, recognizing that a pure icon can only be a possibility, with an object that is a Firstness, Peirce speaks of a sign that may be iconic, ‘that is, may represent its object mainly by its similarity, no matter what its mode of being’ (2.276). This sort of impure icon he then terms a hypoicon.

Hypoicons may be roughly divided according to the mode of Firstness of which they partake. Those which partake of simple qualities, of First Firstnesses, are images; those which represent the relations, mainly dyadic, or so regarded, of the parts of one thing by analogous relations in their own parts, are diagrams; those which represent the representative character of a representamen by representing a parallelism in something else, are metaphors. (2.277)

What appears with greater clarity in the light of these distinctions is the status of imaginal iconic correspondences in the content system of language vis-à-vis the diagrammatic ones. Here, as everywhere in semeiosis, it is a matter of emphasis or relative prominence of one or another aspect of the sign, rather than a matter of all-or-none. Relations of quantity, such as those embodied in the correspondence between desinence length and the grammatical opposition singular vs. plural, have a pronounced imaginal quotient—amounting to the representation of ‘simple qualities’ and ‘First Firstnesses.’ The same aspect predominates in the linguistic representation of the order of events in real time: Caesar’s acts as reported in the sequence Veni, vidi, vici is a straightforward example of the way in which ‘the temporal order of speech events tends to mirror the order of narrated events in time or in rank’ (Jakobson 1971:350). (This is an instance of a mirror image without the attendant reversal.)

Analogic semeiosis of the sort exemplified by linguistic diagrams proper is much less straightforward and cannot be reduced to the mere representation of simple qualities. Indeed, what appears to be typical of diagrammatization in language is a semeiotic mapping involving the interpretants rather than the signans-signatum relation proper. Practically, what this amounts to is the necessary involvement of markedness, of which the most important concomitant is reference to a specific context. To return to the example of the opposition of determinate and nondeterminate subaspect in Russian verbs of motion, the marked stem cooccurs with the marked category, the unmarked stem with the unmarked category. Inasmuch as the marked category is expressed by a stem which is consistently shorter than the stem of the corresponding unmarked category, the length of the stem can be understood as being abridged to mirror the definition, relative restrictedness or narrowed scope, of the marked value. This would place the relations between marked and unmarked values of the stem in perfect alignment with the earlier phonological example, in which abridged (shorter) syllable peaks in English occur before marked tenues (tense obstruents) and unabridged peaks elsewhere (before lax obstruents, before sonorants, and finally). In both examples of congruence between markedness values, units and contexts match: in the phonological case the context is comprised by yet another unit, in the morphological case the context is comprised by a grammatical category.

4. Units, Contexts, and Markedness

Although there is nothing ‘compulsory’ about the iconic congruence between signs and the contexts in which they occur (pace Jakobson 1971:357), it is nonetheless true that language frequently resorts to these patterns of diagrammatization, and the teleology of linguistic change clearly has the establishment of form/meaning diagrams as a goal.

Taking semeiotic congruence to be the raison d’être of both the synchronic status of language facts and the diachronic process characterizing them while they are in statu nascendi, it is necessary to detail the ways in which this congruence is achieved. The principle force of iconicity in language is channeled through markedness assimilation and (as we shall see) its corollary, markedness reversal. The cohesion between linguistic facts is rooted in the interpretant system imparting structure to expression and content signs in their cooccurrence.

While it might be misleading to say that ‘markedness inheres in the relationships between signs and contexts’ (Haiman 1980:529, fn. 12), it is nevertheless important to emphasize the interpretant value of context. The context-sensitive nature of markedness also implies the possible subordination of narrower to broader contexts, in a hierarchical structure. This notion would tally well with the common presupposition concerning the all-encompassing nature of part-whole relationships in language.

Markedness as a species of interpretant is subject to the same qualifications as all interpretants, and none is more pertinent to language structure than its indissoluble link with abduction. As was clear from the extended discussion of Peirce’s notion of interpretant in the preceding chapter, the ramified system of Thirds that bind Sign and Object into significational complexes articulates a graded gamut involving a variety of differential factors. One of these is the degree of fixity or codedness of an interpretant vis-à-vis its antecedent sign(s). The determination of degree of codedness of the interpretant-sign-object relation is an abductive process; as such, it is fundamentally fallible (unlike deduction and induction, given the truth of the premisses).

The fallibility of abduction means, as far as the theory of grammar is concerned, that there is nothing obligatory, necessary, or inevitable about the coherence of language data in the short run. Whereas the overarching teleological thrust of language development tends to direct the cooccurrence of linguistic phenomena into patterns which are iconic, at any given point in the history of a language the process may not yet have run its full course, thereby presenting learner and analyst alike with a mottled picture characterized by variation, discontinuities between grammars of different speakers, and numerous anomalies—in short, with the flux that language always presents side by side with what has become stabilized in it.

With respect to markedness, the non-deterministic character of coherence is nowhere more prominent than in what has been called markedness reversal or markedness dominance (Andersen 1972:45-6). This is the phenomenon whereby a marked context reverses the normal markedness values of the terms of an opposition, including reversal of the relation between the elements of a syntagm (‘rank reversal,’ Andersen 1972:46, fn. 23).

A straightforward example of markedness reversal in phonology is provided by the neutralization in Standard German of the protensity category in obstruents functioning as syllable codas (Andersen 1972:44-5). All obstruents in this marked function (syllable onsets being unmarked) are realized as [+ tense], the marked value. (This illustrates the potentially wide scope of markedness assimilation: here its reference goes beyond simultaneous or contiguous diacritic signs to embrace a more abstract element of the context, the position of a segment within the syllable.) Since protensity is phonemic in German (Trubetzkoy 1958:75), one would expect the unmarked term of the opposition (the lax obstruent) to appear in position of neutralization, but the dominance exercised by the marked value of the supervening context of syllable coda effects a reversal such that ‘the normally unmarked value is evaluated as marked’ (Andersen 1972:45). It is, parenthetically, the failure to perceive the role of the syllable as a context that accounts for the common misstatement of this pronunciation rule of German in American linguistic literature (Anderson, loc. cit., fn. 22): it is almost universally referred to as ‘devoicing’ (in ‘word-final position’) because the expectation that the unmarked member of an opposition appears in position of neutralization is fulfilled only if voicing is assumed to be phonemic, tenues in the European languages being nonphonemically voiceless (the unmarked value of the diacritic category of voice).

There is nothing aprioristically compulsory in this dominance of syllable coda in German, although the accomplished fact of coherence between elements and context may appear to render it so retrospectively (the usual point of view of the analyst). It is possible after all to cite a language—English—where no such neutralization rule obtains, despite the identical status of the category involved (protensity in obstruents).

The generality of the semeiotic principle, together with its optional though teleologically favored nature, can be seen in the following further examples, which affect the content system rather than the expression system. The implementation of grammatical categories is commonly determined by the presence of the marked member of an opposition defining contexts of neutralization; the normally marked member of another opposition is selected.

In English, for instance, in sentences in a marked status, the assertive vs. non-assertive opposition (they do know vs. they know) is neutralized, and the normally marked assertive is used to the exclusion of the non-assertive (Do they know? They do not know.) Similarly, in the marked subjunctive mood, the past vs. present opposition (they knew vs. they know) is neutralized, and the normally marked past tense is used to the exclusion of the present (I wish they knew). Here, too, the number opposition (they were vs. he was) is neutralized, the normally marked number being used to the exclusion of the unmarked number (I wish he were). (Andersen 1972:45, fn. 23)

It is inherently not possible, despite the widespread insistence on the part of linguists that the justification for the effects contexts have on elements be ‘independently motivated’ (thus, for instance, in Haiman 1980:529, fn. 12), to discriminate between contexts that do and those that do not condition markedness reversal. That is to say, the determination of a particular rank order in a given hierarchy of contexts may be largely language-specific and subject as a systematic principle of structure to the very ‘ad-hoc and circular’ identifications that linguists fear as analytically illegitimate (Haiman, loc. cit.).

Regardless of the optional force in language of markedness reversal in marked contexts, it is ‘an essential characteristic of all human semeiotic systems. For a simple example, consider the distinction between formal and casual dress. In the unmarked context of an everyday occasion, formal wear is marked and casual clothes unmarked; but in the marked context of a festive occasion, the markedness values of normal and casual clothes are reversed’ (Andersen, loc. cit.). Furthermore, neither the sign nor the context in which it occurs need have an immutable markedness value, for it is possible that they be mutually defining, as in the example (Haiman, loc. cit.) of ‘a bilingual who speaks French at home and English at the office. Marked linguistic behavior for such a person is the use of English at home or of French at the office.’ Neither language is either marked or unmarked except when used in a particular social context.

5. Markedness and Derivational Relations

Differences in phonic expression which accompany discrete constituents through a derivational network or family tree are typically compared amongst themselves. Alternatively, differences in expression are juxtaposed directly to corresponding differences in content. Both gambits fail to address the real question: just why are certain specific expressions associated with certain specific contents?

The semeiotic values that at once inform sound and meaning are markedness values. Just as the phonological structure is determined ultimately by the markedness relations obtaining between its sets of oppositions, so grammatical and lexical categories organize themselves into a coherent system through oppositions of grammatical and lexical meaning informed by the evaluative dimension that is markedness. The apparent chasm between expression and content is thus bridged.

Taking the position that patterns of meaning arc determined by semeiotic value makes it possible to meet head on the problem of the coherence of derivational relations (as, indeed, of all linguistic relations). In many languages there is a set of relationships between words that is sometimes referred to (in the typical process-oriented or mutational mode) as a ‘truncation.’ Elements present in what is taken to be the deriving base of a derivative are absent in the latter, and the resultant morphophonemic alternation is characterized as including the deletion of certain elements (segments, suffixes). Truncation can be seen as one of five constituents of an exhaustive typology of formal processes accompanying derivation: (1) augmentation (affixation); (2) reduction (truncation); (3) a combination of augmentation and reduction; (4) morphophonemic alternation (including prosodic features)—with or without augmentation or reduction; (5) change of form class only, i.e., without any overt morphophonemic change. Typically, a relationship such as Russian krépkij ‘strong’ // krepít’ ‘make strong’ is seen (correctly) in directional (i.e., hierarchical) or (incorrectly) in mutational terms and represented as a truncation of the stem-final k of krépkij in krepít’. This kind of analysis centers exclusively on the morphophonemics of the relationship, utterly oblivious to the fact that there is a difference in meaning associated with the difference in form. Nor does it embrace the knowledge that the verb is factitive and transitive and thereby implements the grammatical opposition defined by the conjoined categories factitive/stative and transitive/intransitive. The consistent inclusion of the grammatical categories at stake is important generally; in this particular example all the more so, as there exists a stative form krepčát’ ‘become strong’ in which the k of krépkij alternates (as regularly elsewhere) with č.

An analysis which aspires to explanatory status must come to grips with the formal alternation in the triplet krépkij // krepít’ // krepčát’ associated with the different categories that implement it. This translates into the demand that the earlier question be reiterated here: why a particular expression for a particular content? Historical exegesis alone, while not wholly irrelevant, will not suffice, since what we need to uncover is the precise semeiotic motivation which permits the forms to subsist synchronically as differentiated expressions of coordinated contents. It is just at this point that we have recourse to markedness and its semeiotic implications.

Truncation, understood as shorthand for the relation between forms which differ in part by the presence vs. absence of a given element of expression, involves an unmarking. A term of a derivational correlation which is taken to be primary (the so-called deriving base) vis-à-vis a specific secondary counterpart (its derivative) is marked, and the latter is unmarked. This assignment of markedness values accords well with the notion that the absence of something (signe zéro) is normally unmarked, whereas the presence of that same something is normally marked.

In our example, however, truncation only occurs in the factitive/transitive form, so we must ask: why there and not in the Stative? The answer resides in the markedness values of the opposed categories. Since factitive is unmarked, the presence of a truncated form is congruent semeiotically with the value of the category in which the form is implemented. Stative/intransitive is, conversely, marked; and this marked context dictates a marked means of expression—the č that is phonologically marked for stridency as opposed to the unmarked k. We have here, consequently, an instance of markedness assimilation.

Truncation (and for that matter, all reduction) is normally an unmarking, as has already been asserted. It follows, to be sure, that all augmentation is normally a marking. In an opposition dominated by a marked context, however, the reverse may obtain, in conformity with our second semeiotic universal, markedness reversal. Thus in affective formations (diminutives, hypocoristics, etc.) reduction applies with reversed sign values, so that a Russian or English diminutive like the proper name Miša ‘Mike’ is marked vis-à-vis its neutral counterpart Mixaíl ‘Michael.’ An analysis of reduction in terms of markedness not only reveals the semeiotic coherence of the derivational relations involved but also accounts for the widespread incidence of truncation as a preferred means in forming hypocoristics and diminutives. This observation is tangent with the knowledge that ellipsis, abbreviation, and univerbation are particularly favored processes in the formation of marked (i.e., non-neutral) vocabulary and the marked (social/professional) use of language.

Composition is a particularly revealing testing ground for both the overarching conception of linguistic structure being advanced here and the specific illustrative analysis, in that it subsumes both major types of derivation, lexical and syntactic (cf. chapter 3, section 1). In the face of a [+ sharp] stem-final consonant v’ in Russian krov’ ‘blood,’ it has proven troublesome to interpret the presence of its [-sharp] counterpart v in compounds such as krovopodtëk ‘bruise,’ krovosmešénie ‘incest,’ krovožádnyj ‘bloodthirsty,’ etc. However, given the semeiotic analysis of truncation proffered above, it is apparent that what we have here is yet again an unmarking, i.e., the substitution of an [U sharp] sign for a [M sharp] sign as part of the elaboration of a set of relationships (composition) that is structurally homologous with truncation.

The Russian data can profitably be compared to similar phenomena in Japanese. In Japanese compounding is often accompanied by the substitution of a lenis consonant for a fortis in the second constituent of a binomial, so that, for example, fúufu ‘husband and wife’ + kenka ‘quarrel’ fuses in fuufugénka ‘family quarrel.’ Again, assuming that k is distinctively [M tense] and g is [U tense], the compound shows an unmarked sign over and against the marked sign of the identical constituent when uncompounded. The concomitant alternation in the locus of high pitch is likewise in need of explanation, and the repository of explanantes is, unsurprisingly, the same. Each constituent is liable to a loss of its prosodic characteristics when compounded. Thus fúufu, with high pitch on the first mora of the first syllable, becomes totally low(er)-pitched, while kenka, with high pitch occurring on any following intraphrasal constituent, acquires high pitch on the first syllable. The prosodic features of both constituents have been superseded by the characteristic prosody of the (Sino-Japanese) compound—high pitch on the first mora of the second constituent. The loss of independent prosodic characteristics by both constituents through composition is an unmarking.

The Japanese ‘retraction’ of pitch in the second constituent is structurally akin to the alternation of stress position accompanying univerbation and nominalization in English. Phrases like rènt a cár or lòng beách when substantivized into rént-a-car and Lóng Beach not only display a loss of individual constituent stress but also a simultaneous retraction of primary stress from all syllables other than absolute initial onto the latter. This phenomenon is to be explained, just as retraction in verbal/nominal pairs like permit /pérmit, frequént/fréquent, rejéct/réject, envélop/énvelope, as a markedness assimilation. Initial stress in English is unmarked, non-initial marked. Correspondingly, the category of nominals (substantives, adjectives) is unmarked relative to verbals, which are marked (cf. Jakobson 1980:105). Note that what could be seen here as the iconicity (more precisely, diagrammaticity) of nominal stress when juxtaposed to the unmarked order (Subject Verb Object) of English sentence units is at bottom a markedness assimilation.

6. Further Exemplification and Summary

A theory of grammar informed by Peirce’s semeiotic casts the notion of rule in a fundamentally different light from that of its understanding in contemporary linguistics. Specifically, besides their status as interpretants, in a semeiotic theory of grammar rules cease to be viewed as divorced from the sign relations they implement. Rules are perceived necessarily as representing the relations constituting units and contexts and their hierarchies. (We can then claim to have begun the long trek back to the sane methodological postulates of interwar European structuralism.) This view is tantamount to the establishment of a neostructuralist perspective which redefines ‘rule of grammar’ in a radically different manner.

From at least the time of Saussure, language has been considered akin to (if not a species of) a game, whence the conception of rules of grammar as identical in essence to the rules of a game (like chess). Playing a game involves performance along prescribed, standardized patterns which, in chess for example, are called moves. The rules of chess determine the repertoire of permissible moves; also, they determine the domain of correct moves, violation of which is tantamount to not playing the game. The notion of correctness implies an obligatoriness: in certain game situations a certain move is the only mandated one (cf. Wright 1963:6).

All of the conditions attending games render playing them a type of rule-governed behavior. Rules are prototypical of norms, and the jump to the conception of language as rule-governed behavior seems natural:

The rules of grammar (morphology and syntax) of a natural language are another example of the same main type of norm as the rules of a game. To the moves of a game as patterns correspond the set forms of correct speech. To play or the activity of playing a game corresponds speech or the activity of speaking (and writing) a language. Of a person who does not speak according to the rules of grammar, we say either that he speaks incorrectly or that he does not speak that language. (Wright 1963:6)

On closer scrutiny, however, the conception of language as rule-governed does not appear as monolithic as made out to be. Linguistic rules are typically represented in the form of operations on entities, in which respect they can be termed mutational. The implicational or conditional meaning in the formula X → Y/———Z (read: X goes to Y in the environment Z), appropriate as a notational device for diachronic changes, is also promiscuously transposed into the study of synchrony, where its distortive impact is scarcely perceived. It is distortive primarily because language, which is a system of relations, comes to be represented instead as a system of processes.

When considering pluralization in English, to take a familiar example, a contemporary linguist (like Halle 1964) will formulate as part of his analysis a ‘rule’s → ɨz / [+ strident]———. This notation states that the English plural desinence -s assumes the shape [ɨz] following substantival stems which end in a strident obstruent, i.e., one of the group /s z š ž č ǯ/, as in asses, phases, ashes, garages, watches, judges, respectively. The arrow (→) belies the fact that what is really at stake is not a process but a relation between the alternations of form associated with an invariant grammatical meaning, the plural. The alternant [ɨz] is actually one of a set of variants which also includes [s] and [z], as in locks, fops, chiefs, pins, whims, boys, etc. Given the (well-nigh universal) methodology which ‘sets up’ basic forms such as the English plural morpheme and proceeds to ‘derive’ the actually occurring phonetic variants from it by the application of ‘rules,’ no attention is paid to the question of coherence among the data of the English plural.

Every person learning English must incorporate into his or her knowledge the forms oxen and children. These forms are purely rule-governed in that the addition of the suffix -en to the corresponding singular stems must be learned as a fact of pluralization unconnected with the vast bulk of English words both extant and potential. Indeed, these two exceptions to the ‘rules’ of English pluralization are unusual precisely because they must be learned by rote, just as many rules of behavior or of games must be learned. The functional coherence assumed to exist between the forms of the English plural and the associated grammatical meaning is, in other words, almost totally absent from the two cited exceptions.

Once this dichotomy is clearly perceived the structural core ceases to be characterized by rules in the conventional sense, since these are characteristic . . . of exceptions! This means that the structural core consists rather of linguistic data that cohere in virtue of certain principles of organization which assure the solidarity of the data.

Perhaps it can now be better understood just how crucial markedness is to an explanatory theory of the structural core of language—that part of language where data cohere—aspiring to make sense of grammar. For it is markedness alone which allows a unitary explanation of linguistic facts at all levels of structure.

English (like French and Russian) evinces a case of concomitant neutralization in its compact obstruents: strident compact obstruents are predictably [+ acute], mellow compact obstruents predictably [-acute]. This means that the markedness values for the feature [± acute] will be reversed in the compact obstruents of English, as in Russian (Shapiro 1972b:355). Hence /š ž č ǯ/ are designated [M str], whereas /k g/ are [U str]. As for /s z/, they are marked with respect to themselves; consequently, when stems ending in /s z/ precede the plural desinence which is identical in specification, the stem-final segment and desinential consonant are marked with respect to each other—which is to say that the condition of multiple markedness obtains (cf. Shapiro 1972a). This is also true, however, of /š ž č ǯ/, which are [M comp] and [M str].

The question as to which of the alternants [s], [z], or [ɨz]—or, perhaps, some other entity—is to be taken as the ‘base form’ of the desinence can also legitimately be asked. Before attempting an answer, one must understand that this is a question of morphology (‘relations between linguistic signs’) and not one of morphophonemics (‘relations between the contextual variants of the same linguistic sign(s)’;cf. Andersen 1969a:807). The congruence of phonological form (segments, diacritic signs) and the grammatical meaning of plurality is a morphological one which hinges on the markedness value of the category involved. Since plural is assigned the value [M number], the basic sign (morpheme) will mirror this value by implementing the marked term(s) of a phonological distinctive feature. The relevant feature is tense/lax. Consequently, the English plural desinence, like that of the possessive, is s (pace Andersen 1980:4), /s/ being [+ tense] and thereby [M tense].

With regard to the phonetic realization of the basic plural desinence s, the following congruence obtains: stems terminating in lax segments cooccur with the alternant [z] (pins, whims, bears, drives, bibs, etc.); those terminating in tense segments cooccur with [s] (docks, lots, pips, etc.). Vocalic stems, however, present a problem. In word-final position stressed vowels are always diphthongal, that is to say, only underlying tense vowels, which are [U tense], are permitted. As for unstressed vowels in word-final position, there can be no contrast based solely on the protensity feature; this means that whatever the phonetic realization here, the sign value is [U tense]. It is the unmarked value of word-final vowels that explains not only the [z] of boys, spas, Annas, bellows, etc., but the [z] of [ɨz] as well.

The unitary nature of a linguistic theory grounded in markedness has been illustrated above largely by examples from phonology, morphophonemics, and morphology. This should not obscure the fundamental fact that the structure of syntax and lexis is also systematically informed by markedness (cf. Andersen 1972:45, 1975). All levels of linguistic structure are isomorphous as to the principles of their organization. A thorough-going semeiotic approach to language must ultimately fill in the lacunae that persist in our theoretical understanding of these principles and the different means by which they are implemented in widely divergent languages. An adequate program of theoretical inquiry into language as a system of signs, which the present chapter has sought to prefigure, must set the hermeneutic understanding of immanent structure as its primary goal.

In achieving this end, it will need to address the real problems of linguistic structure subtended by real data, and to attempt to solve them in a uniform fashion which does justice to the principles of isomorphism and iconicity underlying language as semeiotic. It is to this task that the following chapters are devoted.