Phonological Theories Session 3 (version SS2006) Phonological Theories Distinctive Features – SPE and Feature Geometry Bibliography (works cited in Session 5). Clements, G. N. (1985) The geometry of phonological features. Phonology Yearbook 2, pp. 225-252. Chomsky, N. & Halle, M. (1968) The Sound Pattern of English. Fischer-Jörgensen, E. (1975) Trends in Phonological Theory. Copenhagen, Akademisk Forlag. Spencer, A. (1996) Phonology. London: Blackwell

Post-Jakobsonian Features (SPE) Syntactic component Chomsky and Halle (1968) derived the phonological structure from the morphological structure. Phonemes (as units in the observable surface form) were no longer required. Base rules & lexicon Semantically interpreted Deep Structures Deep structure Semantic component The Structuralist principle of separate dscriptions for different levels of structure was abandoned completely within Generative Linguistics. The whole approach was geared to providing a descriptive framework (at least) and an explanatory framework (if possible) for the human‘s ability to produce an infinite number of different „sentences“ from a finite number of grammatical elements. The generation of the morpho-syntactic structure was placed in the centre of the model. An underlying („deep“) structure was related to the meaning of a sentences. This was not the surface form that was produced (or heard). The Surface Structure was still abstract but it allowed an interpretation at (an abstract) sound level. The so-called „formatives“ (the surface morphology) had a phonological definition. Before an actual (phonetic) realisation could be defined, however, the phonological rules (which describe the way the phonological form is realised) had to be applied. Transformational rules Phonetically interpreted Surface Structures Surface Structures Phonological component

Post-Jakobsonian Features (cont.) • The features were more dependent on the articulatory configuration, so more differentiated features were required (e.g. not rounded, pharyngealised and retroflex as [flat]) • All phonological features were strictly binary. (the distinction between underlying and surface forms allowed „phonetic features“ to take on continuous values) • Focus very much on inherent (segmental) features. Only stress was theoretically developed to any degree. The criticism of Jakobson‘s attempts to reduce the number of features to a minmum was well known. The feature [flat], which had three different possible articulatory bases, was clearly unsatisfactory. The articulatory orientation of the features thus brought in a greater differentiation and increased the number of features. The abstract, underlying structure, with its direct 1-to-1 relationship to the morpho-logical structure provided a solution to the bi-uniqueness problem of the Structuralists. It also allowed a separation of a categorical (and binary) differentiation of phonological feature properties from the gradual (non-categorical) values of properties at the phonetic level. The adherence to the binarity principle remained – and was just as intuitively accepted as in Jakobson‘s works. - Prosodic features are defined as stress, pitch and length. But even stress, which is dealt with in SPE (the cyclic stress generation rules) is basically a segmental property, a property of the vowel, not of the syllable (i.e. the syllable had no theoretical status. It had been overlooked because the phonological structure was „derived“ from the morphology; it had no inherent structure of its own..

SPE Inherent Features 1 • Features defined along four dimensions (compared to the three by JFH): - Major class features. - Cavity features. - Manner features. - Source features • Apart from the first dimension, these reflect the articulatory, production perspective rather than the acoustic/perceptual. (functional definition) (articulatory definition) The JFH bias was towards the acoustic basis of the features; hence sonority and tonality groupings (only protensity implied an articulatory, production orientation in the definition). SPE features are much more articulation orientated with the exception of the functionally important „major class features“ which separated vowels from consonants, and sonorants from obstruents. - Sub-grouping vs. „ordering“ of features: It should be understood thought, that although these feature groups defined natural classes of sounds, no intrinsic ordering of the features, no dependency among the features was postulated or implied. However, de facto there are dependencies, both universal (articulatorily determined) and language dependent (structurally determined). These are expressed by redundancy rules and either a 0 or a gap is left in the feature matrix.

Comparison of Inherent Features 1 The use of [±vocalic] and [±consonantal] by FJH was criticised, but the SPE major class features are not very much better (cf. Fischer-Jörgensen, p.228). [+vocalic] is defined as having a constriction no narrower than for [i u] and vocal folds in a position for spontaneous voicing. This is a double characterisation (glottal and supra-glottal) which is unfortunate for an assumed binary feature. Clearly the glottal part of the specification excludes [h] and [?] and the supra-glottal part excludes [w], [j] and [l], but this is not a convincing advance on the FJH definition (unobstructed vocal tract and formants). Note that 6 SPE features correspond to the 4 JFH features: compact/diffuse and the 3 tonality features. [± lateral] has been added. How was /l/ dealt with in the JFH features?

Comparison of Inherent Features 2 Not much has been changed in the manner features (except for an articulatory differentiation of affricates ([- instantaneous release]) from other [+strident] consonants. Strident still remains as a (non-articulatorily defined) feature to distinguish /s/ from /T/, for example, and /X/ from /x/.

Unordered vs. ordered features • The features are subclassified according to function (major class features) or production properties (cavity, manner and source features), but there are no dependencies between the features. • Nick Clements (1985) presented a grouping of features which took the link between features and their articulators into account: The “ordered” view of features is known as “Feature Geometry ” • Some features are regarded as independent of a particular articulator (e.g. consonantal, sonorant, approximant) This point has already been mentioned above, in connection with the specification of redundancies in the feature matrices. There are still clear groupings of different classes of features (cp. major-class, cavity, manner and source features). This is represented in the hierarchy of features as a „root dependency“ for the features consonantal, sonorant, approximant. and a „manner-dependency“ for manner features. All other features are dependent on the place of articulation/articulator. E.g.: Voiced is dependent on larynx activity Nasal is dependent on velar activity Round is dependent on lip activity High is dependent on tongue-dorsum activity ATR is dependent on tongue-root activity (however, some place it as a dependent of the dorsum) • Other features are dependent on a specific area of the production system (e.g. voiced, vs. nasal) • Other features are clearly dependent on a specific articulator (e.g. round, high, ATR)

Basic Geometry for Features (after Clements 1985, cf. Spencer 155 ff.) This „tree“ shows the basic articulatory „geometry“. The ROOT defines the segmental base, and links the particular segment to the rest of the language structure (via a C or V segmental tier, which defines the function of the sound within a syllabic structure). This will be dealt with later in the sessions on autosegmental phonology. The two main „branches“ of the tree are formed by the laryngeal vs. supra-laryngeal distinction. The latter branch is then subdivided into place and manner. Note that these sub-categories of the segment are very traditionally phonetic!

Feature Geometry (all) It may have attracted your attention that there are no ± values for the features specified here. The binarity principle has not been (generally) thrown overboard but it is not necessary as a basic condition of sound specification. The „dependencies“ that exist between the articulators (or articulations) and the features means that a negative specification often becomes dispensible. Non-terminal features are all unary. LABIAL applies to lip articulations, but the [round] feature, which is (of course) dependent on the lips, is an added feature dimension for vowels (and glides). Compare also CORONAL and DORSAL, but also LARYNGEAL.

Feature Geometry representation of /t/ Discuss the

Alternative Feature Geometry scheme (Halle 1992)

Exercise (written) Specify a SPE and a JFH matrix for the word “Standlicht” and compare. Make notes of any problems, queries or objections you have! 2. Draw a series of feature-geometry trees for the same word. (use the Clements arrangement) Does the feature-geometry scheme by Halle 1992 contain any theoretical or practical differences when compared to Clements/Spencer?