Feature specifications and contrast in vowel harmony: the orthography and phonology of Old Norwegian height harmony
Sandstedt, Jade Jørgen Michael
In this thesis, I provide a new approach to the role of phonological patterning in determining the featural content of phonological relations and the size and shape of sound inventories. The empirical scope of this project has particular focus on vowel harmony and vocalic features with an extended case study of Old Norwegian. Vowel harmony, simply defined, is a process where vowels in a word show systematic correspondence for some feature. Because of its many moving parts and obvious class behaviour, vowel harmony and harmony languages provide one of the best laboratories for exploring the emergence, acquisition, specification, and common patterning of phonological features. In chapter 1 I provide an introduction to Old Norwegian vowel harmony and some unexplained harmony exceptions. This chapter explores parallel phenomena in the typology of harmony languages and the theoretical challenges these patterns pose. In particular, I illustrate that non-harmonising segments display three distinct behaviours with respect to phonological activity and visibility while the core components of popular grammatical and representational approaches to vowel harmony commonly only predict two. I suggest the solution to this problem lies in the representation and definition of phonological contrastivity. Chapter 2 presents the principal components of a new approach to the acquisition and specification of features using a version of Contrastive Hierarchy Theory (Dresher, Piggott & Rice 1994; D. C. Hall 2007; Dresher 2003, 2009) which incorporates emergent and substance-free features and feature-nodes (Iosad 2017a). In this chapter I argue that phonological features, segments, feature classes, and whole sound inventories emerge according to the Correlate Contrastivist Hypothesis which holds that a language’s phonemic inventory is defined by the set of active phonological features required to express the language’s phonological regularities. Drawing insights from Westergaard’s (2009, 2013, 2014) model of micro-cues, I posit that language learners generalise small pieces of abstract linguistic structures (‘micro-cues’) in the form of features and feature co-occurrence restrictions while parsing linguistic input. In the course of language acquisition, these micro-cues accumulate, and the sum of these cues defines a sound inventory. I argue a segment’s feature specifications and the shape of feature classes in a language are determined by a version of the Successive Division Algorithm (Dresher 2009, §7.8; D. C. Hall 2007, §1.2.7; Mackenzie 2013, 2016) which takes an ordered set of representational micro-cues as its input and returns a contrastively specified segment inventory as its output. Finally, this chapter demonstrates how these components combined with the hierarchical organisation of features afforded by the contrastive hierarchy architecture recapitulates all the important insights of feature geometry, providing an economical and principled model of phonological representations which narrowly vary cross-linguistically. In chapter 3 I present a formal model of harmony using a licensing approach, adapted from Iosad (2017a) and Walker (2005), inspired by the recipient-oriented model of Nevins (2010). Using a detailed study of cross-dialectal microvariation in harmony and harmony neutrality in Yoruba (Atlantic-Congo), I demonstrate that this framework makes the right predictions, affording a ternary contrast in the behaviour of non-alternating harmony segments without any necessary additional grammatical mechanisms. A principal assumption of Contrastive Hierarchy Theory is that the hierarchical scope of features is cross-linguistically variable, and this chapter illustrates how variable feature ordering predicts common asymmetries across harmony languages in the presence or absence of required agreement for orthogonal features (so-called ‘parasitic harmony’). Specifically, the contrastive hierarchy derives parasitic harmony languages by nesting harmony feature contrasts within other featural divisions. This chapter closes with an exploration of the predicted typology of non-/parasitic systems and provides explicit diagnostics for identifying true vs. false parasitic harmony. The theoretical chapters present a coherent, limited, and highly predictive model of phonological representations and vowel harmony, but the real value of a theory is whether it can provide new insights on questions which have otherwise resisted explanation. Old Norwegian vowels and vowel harmony represent such an example. Old Norwegian vowel harmony displays remarkably complex patterns, and its analysis is considerably complicated by the philological nature of available evidence. Chapter 4 presents the materials and methods I employ for the automated collection and phonological annotation of Old Norwegian vowel sequences in a corpus of mid-to-late 13th-century manuscripts. The corpus study’s data set is freely available online at http://dx.doi.org/10.17613/ gj6n-js33. Chapter 5 provides a grapho-phonological study of the Old Norwegian vowel inventory and segmental phonological patterns. This corpus study shows that Old Norwegian manuscripts display robust (pre-decay), transitional, and decayed vowel harmony, which provides invaluable empirical evidence for the otherwise poorly documented decay of harmony systems. The rest of the chapter provides a detailed survey of pre-decay Old Norwegian surface harmony patterns and their interaction with other sound processes and sound changes (e.g. umlauts, vowel deletions, and vowel mergers). A major goal of this project has been to develop tangible heuristics for the reconstruction of historical phonological representations on the basis of phonological patterns evidenced in textual source material. Tying together this thesis’ theoretical and empirical components, I show in chapter 6 how the active vocalic features and feature co-occurrence restrictions in Old Norwegian can be discerned according to the Correlate Contrastivist Hypothesis. In turn, the intricate harmony and neutral harmony patterns in Old Norwegian receive a straightforward explanation following these representational generalisations. This case study illustrates how even complex harmony systems such as Old Norwegian can be reduced to simple emergent effects of the categorisation and co-occurrence of features in contrastive feature hierarchies. This chapter concludes with a historical phonological investigation of the implications of this harmony system for the status of other Old Norwegian sound patterns. The main features of this thesis’ theoretical component and useful abstract schemata are provided in chapter 7 to aid in applying this framework to new data. For ease of comparison, I provide an appendix with contrastive hierarchies and summaries of each harmony language cited in this thesis. The unique contribution of Old Norwegian neutral harmony patterns within the typology of vowel harmony languages provides important evidence for the role of feature specifications and contrastivity in phonology. This thesis’ broad typological and narrow empirical studies confirm the descriptive and explanatory adequacy of the proposed framework in providing novel insights on new and old problems regarding the link between phonological representations and phonological patterns.