Change in obstruent laryngeal specifications in English: historical and theoretical phonology

Abstract

Two traditions have arisen from an ongoing debate concerning cross-linguistic laryngeal representations in series of obstruents. The first, ‘traditional’ approach assumes universally identical laryngeal representations: /p, t, k/ are unspecified and /b, d, g/ carry |voice|. The second, Laryngeal Realism (LR), assumes underlyingly different representations between languages: ‘aspiration languages’ have unspecified /b, d, g/, and /p, t, k/ specified for |spread|. ‘Voice languages’ have unspecified /p, t, k/, and /b, d, g/ specified for |voice|. In this thesis, I use historical data in order to determine which of these two traditions is correct. Chapter 1 introduces the thesis subject and places it in the broader context of representational models of theoretical phonology and general historical linguistics. In chapter 2, I discuss the discrepancy between traditional laryngeal features and their cross-linguistic implementation, the basis of the debate outlined above. The two traditions are then discussed in detail. It is shown that evidence for LR is drawn from surface facts in aspiration- and voice languages such as respective presence or absence of aspiration of /p, t, k/, respective absence or presence of voicing in /b, d, g/ and asymmetry in assimilation processes in favour of one of the features. Present-Day English (PDE) is best described in LR when these criteria are taken into account, e.g., [ph]in, [b 0]in, and invariable assimilation to ‘voicelessness’, e.g., cats /t+z/→[ts], sacked /k+d/→[kt]. In the following chapters, I present data from historical laryngeal modifications in English which have never been considered together in this respect before. In Chapter 3, I present new evidence that the laryngeal situation just described for Present-Day English dates back to the very beginning of its recorded history. This is shown in the fact that all laryngeal assimilation throughout the history of English is exclusively assimilation to ‘voicelessness’ or |spread| - as in pre-Old English [pd] > [pt] cepte ‘kept’, [td] > [tt] mette ‘met’, [kd] > [kt] iecte ‘increased’, [fd] > [ft] pyfte ‘puffed’, [sd] > [st] cyste ‘kissed’. LR can easily capture this asymmetry because |spread| is the only active member in the laryngeal opposition. |voice| is unspecified in English and can therefore never partake in phonological processes. Chapters 4, 5 and 6 deal with historical English data traditionally interpreted as ‘voicings’, i.e. addition of |voice|, and ‘devoicings’, i.e. loss of |voice|. Therefore, these data are potentially problematic for LR in that, according to this framework, |voice| is not specified in English. However, I show that LR can unproblematically deal with these phenomena as laryngeal lenition, removal of |spread|, and fortition, addition of |spread|. In fact, some of the lenition processes provide extra back up for LR. Processes in word-initial position, e.g., dialectal [v]ather, and final position, e.g., i[z], knowle[d3], are highly marked when viewed as ‘voicings’. However, when viewed as simple lenitions, as in LR, they are natural processes, which are predicted to be found in languages. Therefore, I show in this thesis that all available data from English historical laryngeal modification support LR, and that LR in its turn sheds an interesting new light on the data. It is superior to traditional accounts in that it can account for otherwise puzzling phenomena such as asymmetric assimilation and initial and final ‘voicings’.