Length in strange places

Qualitative enhancement of sonorant quantity in Celtic and beyond

 Share on Twitter

(Poster presentations)

Present-day Celtic languages are characterized by a typologically relatively unusual contrast between “lenis” and “fortis” coronal sonorants. It is especially developed in the Gaelic languages. Here, “fortis” nn ll rr contrast with “lenis” l n r in postvocalic position (Irish geal ‘bright’ vs. geall ‘promise’). Furthermore, they participate in the initial mutation system, where the “fortis” sonorants generally occur in the unmutated (“radical”) grade, and “lenis” ones in the “lenited” grade, which historically corresponds to postvocalic contexts. In the Brythonic languages, the most notable correspondent of this pattern is found in Welsh, where ll acts as the “fortis” counterpart of : they contrast postvocalically, and is generally the radical consonant in mutation.

Historically, the “fortis” sonorants word-medially are long and the “lenis” ones are short (Stifter 2023). This circumstance is still reflected in their synchronic behaviour: both Irish (Ní Chiosáin 1991) and Scottish Gaelic (Morrison 2020) show patterns that are insightfully analysed if the “fortis” sonorants are contrastively moraic. However, quantity is rarely implicated in their surface realization (but see Lewin 2023). Instead, “fortis” and “lenis” sonorants are distinguished by properties such as place of articulation (Carnie 2002; Musil 2017, 2019; Nance & Kirkham 2020).

The behaviour of Celtic sonorants is unusual. Word-medially, “fortis” sonorants behave as the marked category, as they need to be specified as mora-bearing. This reflects their historical status as the marked member of the length opposition. Word-initially, the situation is reversed: “fortis” sonorants occur in the elsewhere context, while “lenis” ones are produced by specified mutation triggers. Prima facie, this suggests a lengthening or fortition of sonorants in a strong (word-initial, non-intervocalic) position. Although not entirely unexpected (Ségéral & Scheer 2008; Easterday & Bybee 2019), it is somewhat odd to see this process involve a contrastive feature.

In this paper, I discuss the historical origins of this pattern and draw some parallels to similar phenomena elsewhere. Specifically, I argue that the spread of the pattern to initial position did not represent a lengthening, but rather a change involving place features.

The behaviour of sonorants in initial mutation across Celtic follows an implicational hierarchy. The same hierarchy is observed in several aspects of Romance historical phonology. In particular, it determines the preservation of original length contrasts in Western Romance (Brandão de Carvalho 1989). Remarkably, it is also found in the reflexes of initial sonorants in Ibero-Romance. Whilst the distinction between Latin [l n r] and [ll nn rr] is reflected in terms of place and manner (Spanish [l n ɾ] vs. [ʎ ɲ r]) word-medially, initially we find the “long” reflexes for [r] in Spanish, [r l] in Catalan, and [r l n] in Asturian (Cravens 2002).

I suggest that the spread of “quantity” to word-initial position can be understood as a consequence of the transphonologization of quantity to place and its interaction with the life cycle of intervocalic sandhi processes. Both Celtic and Ibero-Romance can be reconstructed with postvocalic lenition crossing word boundaries. I suggest that once original suprasegmental (quantity) distinctions in sonorants became available as subsegmental (featural) representational entities, it was possible for them to be incorporated into the pattern of sandhi lenition.

I show how the proposed account captures the markedness paradox in Celtic by divorcing the place distinctions from distinctive quantity (moraicity). This also accounts for the fact that mutation patterns persist even when earlier word-medial quantity is lost, as in Brythonic. Furthermore, the Celtic parallel can shed light on the Ibero-Romance patterns without recourse to a poorly evidenced process of sandhi gemination (parallel to Italian raddoppiamento) proposed by Cravens (2002).

← A metrical approach to ternarity in Northern European accentual contrasts

|  All presentations  |

The moraic mess that is Danish →