Germanic obstruent lenition : some mutual implications of theoretical and historical phonology
This thesis seeks to reconcile aspects of philology with aspects of linguistics and to illustrate the advances in understanding that can be achieved in this way. The principal empirical focus is provided by four sets of 'classic' data from Germanic historical phonology and the theoretical background is provided by recent developments in non-linear models of phonology. The mutual implications of these are addressed to provide both a new understanding of the data and a coherent theoretical understanding of the types of phonological process that they exemplify. The data sets are chosen partly because of their iconic status in historical phonology and partly also because they all, at times, have been described as examples of a general process-type: 'lenition'. The background is provided by chapters one and two. Chapter one introduces the way in which the relationship between historical and synchronic phonology is understood in the thesis. Chapter two introduces the classic sets of data, namely: (i) the 'Germanic Consonant Shift', (ii) the 'High German Consonant Shift', (iii) the 'English Initial Fricative Voicing', (iv) the 'Inner-German Consonant Weakening'. Chapter three problematises these sets of data in two main ways. The first of these is through an investigation of how they fit with recent advances in the understanding of phonological structure and phonological processes which have been developed in phonological traditions that see sub segmental units as privative (eg, Dependency Phonology, Government Phonology, Articulatory Phonology). Particular emphasis is placed on the units needed to account for laryngeal specifications (traditionally described as 'aspirated', 'glottalised', 'voiced' and 'voiceless') and a position is defended whereby three privative units are required: [spread], [voice] and [constricted]. The first two of these are shown to be used differently in languages to account for the two traditional categories of 'voiced' and 'voiceless'. The second problematising factor is a discussion of how the data sets fit with the concept of obstruent 'lenition', a notion which has a clear, if problematic, place in both historical and synchronic phonology. Lenition trajectories of the type 'stop > affricate > fricative > glottal' and 'voiceless > voiced' are discussed in the light of the data introduced in chapter two and the theoretical discussion of sub segmental structure developed in the first half of chapter three. Further data from Dutch, Spanish, Dravidian, Indo-Iranian, Greek, Bantu, Celtic and Liverpool English is discussed and a synthesis is proposed which takes into account both attested historical data and a theoretical phonological understanding. A notion of 'lenition inhibition' is developed to explain some of the exceptions (which are frequent in lenitions, but are rarely discussed) whereby the sharing of autosegmental phonological units gives a segment 'strength' in certain environments. Chapter four revisits the data introduced in chapter two and provides a reinterpretation in accordance with the understanding of 'lenition' processes developed in chapter three. The data in (i) and (ii) are essentially shown to be dependent on the presence of a [spread] laryngeal specification and to be qualitatively different from the data in (iii) and (iv), which are shown to be unifiable as the loss of [spread]. Additionally, the patterns of exceptions to these processes are examined in light of the understanding of lenition inhibition developed in chapter three. In chapter five, general conclusions are drawn. It is argued that not all lenition processes are unifiable as a single process type. It is further shown that previous discussions of historical 'lenition' processes have missed generalisations, both in terms of the necessary conditions for their diachronic innovation and in terms of the environments which can be seen to partially inhibit their introduction. It is also shown how a detailed understanding of the diachronic data which is discussed here can make a contribution to theoretical models of phonology.