Greek PhoneticsThe State of the Art

abstract phonological representation (in other words, they are not, as it is often assumed, a type of transcription system for intonation). The tones of an intonational representation are phonetically realized as tonal targets , that is as specific points in the F0 contour, defined along two dimensions: their scaling —their value in Hz or any other unit of measuring F0 or pitch—and their alignment —their temporal occurrence relative to specific segmental landmarks; e.g., boundary tones are typically aligned with the last vowel before the boundary they are associated with. It is this alignment of a H% boundary tone that produces the rise on the last (unstressed) vowel of [meˈseona] “Middle Ages” and [siˈnaðelfi] “colleagues” in Figure 9. Both the scaling and alignment of targets are considered to be stable. However, stability is to be interpreted somewhat liberally, as with all other phonetic realizations. First, the exact scaling of tones depends partly on non-linguistic factors, such as the physical characteristics of the speaker, and paralinguistic factors such as her state of excitement or surprise at the moment an utterance take places. Tonal alignment on the other hand, is less affected by such external factors; rather it is linguistic context that affects it the most, such as the position of the tone in the utterance, the number and position of other tones, the phonological weight of the syllable the tone is associated with, speaking rate and so on (among many, Arvaniti et al. 1998, 2000; Fougeron & Jun 1998, Ladd et al. 1999; Ladd, Mennen & Schepman 2000). When all these variables are kept constant, both scaling and alignment are remarkably stable.