Final Consonant Clusters in Middle and Early Modern English

We provide the first comprehensive account of the emergence and development of word final consonant clusters in the Middle and Early Modern English periods. It produces a diachronically layered database derived from corpus data and enriched with phonological and morphological analyses. Consonant clusters are typologically speaking rare, and count as marked, since they face greater articulatory or perceptual obstacles than CV or VC sequences, particularly so in the prosodically weak coda position. Yet, they are historically stable in some languages, English being one of them. It is often assumed that phonological markedness constraints can be overridden for morphological purposes, such as the signalling of morphological compositionality. The Middle and Early Modern English periods are especially suitable for testing hypotheses about phonological markedness and phonology-morphology interaction, since in those periods English coda-phonotactics underwent a radical restructuring through the loss of unstressed vowels in final syllables, i.e. schwa loss. This led to a drastic increase in the number of coda clusters both in lexical items like /nd/ in find as well as via morphology as in signed. The database is extracted from the Penn-Helsinki Parsed Corpus of Middle (and Early Modern) English and enriched with phonological and morphological information. In order to cope with the fact that written texts reflect phonological representations only indirectly, all word forms that could potentially have been realized with final consonant clusters at any point during the periods under investigation, i.e. both before or after their actual attestation, are sampled. The database allows us to construct a systematic survey of coda cluster populations attested during different phases of the Middle English period, and to relate their development to a large number of phonological and morphological parameters. It also allows us to investigate the dynamics of coda cluster evolution through methods such as the following ones. First, we exploit the possibility of simulating the effects of schwa loss on early corpus data: we derive virtual populations of post-schwa-loss clusters from pre-schwa loss texts, and compare them to the coda-cluster populations attested in actual post-schwa-loss texts. These virtual data sets can be interpreted as reflecting the outcome of a development in which the language was affected exclusively by schwa loss, but by no other changes. This facilitates the exploration of hypotheses about the effects of schwa loss and potential therapeutic responses. Second, mathematical methods from evolutionary population dynamics enable us to test and refine hypotheses based on interdependencies of typeoken frequencies. Cluster types are modelled as interacting populations of linguistic items and the historical stability of these populations will depend on the ratio of lexical and morphotactic clusters.

The project has prepared textual data from the Middle and Early Modern English periods for describing the evolution of word final consonant clusters in rigorously quantitative terms. About 380.000 word tokens from six centuries (1150-1750) were retrieved from the Penn- Helsinki Parsed Corpora of Middle English (PPCME2, Kroch and Taylor 2000) and Early Modern English (PPCEME, Kroch et al. 2004), analysed morpho-syntactically as well as phonologically,andmadeaccessible inanonlinedatabase (ECCE, https://ecce.acdh.oeaw.ac.at/), designed in cooperation with the Austrian Centre for Digital Humanities of the Austrian Academy of Science. Identifying and interpreting patterns in the database, the project has yielded new insights on the question how complex sound patterns emerge and get established in phonological systems. Final consonant clusters represent typical examples of complex sound patterns: they are difficult to articulate and to perceive, and cross-linguistically rare. The specific conditions under which they emerge and become (historically) stable are not fully understood. Key findings of the project include the following. (a) Final consonant clusters that reflect and signal morphological complexity (e.g. /md/ in past tense seem+ed) are both more complex and historically more stable than clusters within simple morphemes (e.g. sing /sing/ > /si/). (b) Word forms that end in clusters are the more stable the better their morphological structure can be inferred from shape of their codas, i.e. there is a negative correlation between the morphological ambiguity and the historical stability of word forms, resulting in statistically significant trends towards a complementary distribution of clusters between simple word forms and complex ones (e.g. final /lt/ has become increasingly typical of simple word forms like salt, melt, or kilt, while /ld/ has become more typical of complex ones like called, felled, or killed). Both findings help to explain why morphologically indicative sound patterns can become historically stable even if they are dispreferred on articulatory or perceptual grounds. More spectacularly, they suggest that a selectional pressure in favour of morphologically indicative word forms, and against ambiguous ones, may represent a universal metacondition on the evolution of natural languages. A third finding is (c) that clusters that signal complexity and are highly frequent will not only be stable themselves, but may stabilize homophonous patterns in simple word forms as well. This yields a dynamic that conflicts with the one mentioned under (b) above. The database which the project has produced represents a highly innovative enterprise in methodological terms as well. It is the first diachronic database in which word forms are pre- analysed not only morpho-syntactically, but also phonologically. It makes it possible to address questions in historical phonology in rigorously quantitative terms.