Declarative Lexical Phonology

¢¹. Representative--  John Coleman's homepage , John Coleman's publications

¢º. The Theory

First of all, John Coleman talks about the theory of declarative phonology. Importantly, in declarative approaches, the well-formedness constraints which constitute the definition of the predicate grammatical characterize the surface structure £r directly, not via transformations  of a deep structure distinct from the surface structure. These well-formedness constraints are of several kinds:


1. The syntactic (constituent structure) rules of the surface grammar. These are usually context free, or no more than mildly context sensitive, and the nodes are usually labelled with feature-structures, rather than atomic symbols.


2. Constraints on feature-structures, such as logical constraints on what features may or may not cooccur, and which features are predictable on the basis of which others. Feature propagation constraints (a form of feature cooccurrence constraint holding between different nodes of the syntactic structure) may be regarded in like manner.


3. Feature-structure defaults. This is a challenging area in declara­tive grammatical research, since defaults may introduce non­monotonicity into the semantics of the grammar formalism, which gives rise to some technical difficulties, though they do not necessarily make the grammar itself (i.e. the syntax) non­monotonic.


4. Constraints on the linear order of sister constituents.


5. The lexical entries of words, containing information about their category, and the categories of the phrases they subcategorize for. Lexical entries may be regarded as highly informative and specific feature-structures which contribute information to the surface representation. Such information is complementary to the more general information contributed by the other grammatical constraints.


6. Rules for productively generating new lexical descriptions from lexical entries, such as productive active-passive morphosyntactic relations. Since these are implications of the form ¡¥if X is well -formed, then Y is well-formed¡¦, these too are simply logical constraints on well-formed surface structures.


7. Principles for the semantic interpretation of well-formed syntactic structures: usually, some form of function for combining the semantic interpretation of the parts of a well-formed syntactic structure to form the semantic interpretation of the larger structure (compositional semantics).


  In the preceding Sections  John Coleman has presented a detailed critique of derivational phonology that is matched only programmatically by the declarative alternative. To give real substance to the declarative approach, he shall make a conclusion by examining several of the rules in Halle and Mohanan¡¦s (henceforth H&M¡¦s) Lexical Phonology of English in some detail, proposing specific declarative alternatives.

  H&M¡¦s rules can be roughly divided into two main groups: morphophonologically motivated rules (i.e. those which accompany a morphological affixation or other category-changing operation), from the stress rules to the Centering Diphthong Rule, and allophonic ¡¥detail¡¦ rules from Diphthongization to the end. All of the rules convert relatively more abstract representations corresponding more closely to lexical forms to relatively more concrete representations corresponding more closely to systematic phonetic forms. Nasal Assimilation, y­Insertion and y-Deletion are somewhat out of place among the morphophonological rules, in that they simply adjust the output of one set of rules in order to match the input to another set of rules without any particular morphophonological consequence. Nevertheless, the approximate divisibility of the rule sequence in this fashion surely cannot be accidental, yet the theory in which H&M¡¦s analysis is cast does not predict or account for this partition. Indeed, given the difficulty of determining the particular rule ordering with maximal exploitation of the potential for feeding, bleeding and parsimony, it is not obvious how it comes about that such a rule ordering can be naturally partitioned in this way. The twofold division is natural in a non-derivational model, however, in which there are two levels: a morphophonological (¡¥syntactic¡¦) level, at which morphophonological alternations are represented, and a phonetic (¡¥semantic¡¦) level, at which non-functional, phonetic details are described.

Cross-cutting the division into lexical strata and the morphophonological/allophonic partition, the rules can be grouped into functionally related blocks on the basis of their effects and structural conditions:


1. Parsing of metrical structure, including syllable structure, e.g. the stress rules and the rules of Prenasal, g-Deletion, n-Deletion, Non-coronal Deletion, and 1-Resyllabification.

2. Phonetic interpretation of metrical structure, e.g. the rules of CiV lengthening, s-Voicing, Shortening rules, Vowel Reduction, Vowel Tensing, and Stem-final Lengthening.

3. Vowel shift, e.g. the rules of Vowel Shift, i-Lowering, Center­ing Diphthong Rule and Diphthongization.

4. Palatal assimilation, e.g. the rules of Velar Softening, i-Lengthening, Spirantization, y-Insertion, Palatalization, y-Vocalization and y-Deletion.

5. Phonetic interpretation of syllable structure, e.g. Nasal Assimilation, æ-Tensing and 1-Velarization.


  H&M¡¦s derivational analysis can be recast non-derivationally, while retaining an analysis which is no less parsimonious and elegant than  theirs, a justified failing of the ¡¥straw man¡¦ non-derivational analyses which Halle and Bromberger put forward. Furthermore, in my non-derivational alternative, the functional relations between what are disparate rules in H&M¡¦s analysis will be more transparent by virtue of the fact that they are cooperative constraints. The strategy of my reanalysis can be summarized as follows:


1. The group of rules which he has labelled ¡¥parsing of metrical structure, including syllable structure¡¦, can be replaced by context-free Metrical Structure Rules. Being context free, these yield the same result however they are applied.


2. The group of rules which he has labelled ¡¥phonetic interpreta­tion of metrical structure¡¦ are akin to redundancy rules, in which the presence of (a) particular feature(s) not present in the lexical representation is predictable on the basis of the metrical structure. These rules can be recast as implicational constraints on feature specification, or in some cases as pro­sodic features supplied by the metrical structure rules.


3. The group of rules which he has listed under the common label as ¡¥Palatal Assimilation¡¦ rules will be replaced by a metrical version of the autosegmental analysis of the spread of a ¡¥high, front¡¦ feature specification from the syllable nucleus to the onset. The variety of phenomena expressed in H&M¡¦s various rules will be shown to be the consequences of palatal assimilation in conjunction with implicational constraints on feature specification and the application of Feature Specifica­tion Defaults.


4. The group of rules which he has labelled ¡¥phonetic interpreta­tion of syllable structure¡¦, are like ¡¥phonetic interpretation of metrical structure¡¦ (of which syllable structure is a part), those in which the presence of (a) particular feature(s) not present in the lexical representation is predictable on the basis of the syllable structure. These rules will be recast as implicational constraints on feature specification, or in same cases as prosodic features supplied by the syllable structure rules.