Marina Tzakosta from U. of Crete is visiting the department, next Tuesday, Nov 3.
Marina got her PhD recently from Leiden. Her Promotor (Dutch nomencl.) was
Vincent van Heuven, her supervisor was Jeroen van de Weijer and the external
reader was John J. McCarthy. In general, Marina's work is on prosody and the
acquisition of phonology.
Title and abstract of her talk @ 1:30, next Tuesday, Nov 3 are as follows ....
"Exploring the representation of complex segments: the case of Greek affricates"
ABSTRACT (examples and references can be found here [pdf])
The internal composition of consonant clusters has been dealt with primarily by means of sonority (cf. Clements, 1988, 1990, 1992). Sonority distance seems to be responsible for cluster well-formedness (Steriade 1982). More specifically, the bigger the distance among cluster members, the better-formed the cluster. Therefore, [cl] is better formed than [fθ], because the sonority distance holding among cluster members is bigger for [cl] than for [fθ].
Based on recent experimental findings on the synthesis of Greek consonant clusters we display that consonantal strings range on a scale-like manner with respect to their internal coherence (Tzakosta & Vis 2009a, b, c). More specifically, O(bstruent)L(iquid) clusters, though the best formed, exhibit the least coherent phonological structure. This is supported by the fact that OL sequences are frequently prone to epenthesis or deletion, phonological processes frequently attested in language development data (1a-c). On the other hand, O(obstruent)O(bstruent) sequences display high coherence compared to OL sequences. This is shown by the fact that, first, OO sequences are marked by a smaller sonority distance compared to OL clusters and, second, when OO sequences, like [pt] and [kt], are repaired, they are not subject to epenthesis on deletion, but, rather, to fusion, as shown in the examples in (2a-f). Fusion presupposes accurate perception of the whole cluster. We assume that coherent phonological representations facilitate accurate cluster perception.
Finally, Greek affricates exhibit extensive internal coherence; affricates [ts] and [dz] seem to be rarely prone to repair strategies, rather, they are accurately produced from very early in language development, as illustrated in (3a). It is interesting that affricates replace other consonant clusters, mostly O+/s/ ones, something that adds to the argumentation for the coherence of the former. Such data maintain that Greek affricates are complex segments rather than consonant clusters.
In between coherent OO clusters and affricates lie consonantal sequences consisting of O+/s/, most popular of which are the stop + /s/ sequences [ps] and [ks]. [ps] and [ks] are considered to behave like complex segments on the same line with affricates (Tzakosta & Vis 2009a, b, Tzakosta 2009). Additional experimental results show that, first, [ts] exhibits different degrees of coherence compared to [ps] and [ks]. [ts] displays higher rates of degemination compared to [ps] and [ks]. This makes [ts] similar to single segments like [s]. Moreover, [ts] is not prone to voice assimilation, a fact that supports the claim that [s] in [ts] is not a [sibilant]. Our claims are supported by a great variety of data from synchronic and diachronic phonotactics, experimental and language acquisition data.