Word-boundary Effects on Pitch Timingin Spanish

There is evidence from a number of languages that the position of the accented syllable with respect to the word in which it appears has an effect on the realization of the corresponding tonal event. Peaks tend to shift back (to the left) as word edges approach. Word-edge effects have been shown for English (Silverman and Pierrehumbert 1990), Catalan (Estebas 2003, Prieto forthcoming, Prieto and Estebas 2005), Serbian-Croatian (Godjevac 2000) and Spanish (de la Mota 2005, O’Rourke 2005, Prieto et al. 1995, Prieto and Estebas 2005, Simonet and Torreira 2005). For instance, in Prieto et al. (1995), a significant effect of lexical-stress configuration in a three-way comparison of Spanish sequences such as numero rapido, numero nervioso and numero regular was found. The data in Prieto et al. (1995) were collected from the performance of two speakers of Mexican Spanish. In Iberian Romance, the role of word edges on pitch peak alignment is an area of some controversy. Recent research has focused on Castilian Spanish, as spoken in Madrid and Salamanca (Prieto and Estebas 2005, de la Mota 2005), and on Central Catalan (Estebas 2003, Prieto and Estebas 2005, Prieto forthcoming), as spoken in Barcelona. In this section, we will briefly review the findings for the two languages. Differences in the shape of pre-nuclear pitch accents between Castilian Spanish and Central Catalan are yet to be found: pre-nuclear pitch accents consist of a rise with the valley aligned with the stressed syllable’s onset and with the peak in the post-stressed syllable (Hualde 2000). Thus, effects found for Central Catalan can hypothetically be extended to Castilian Spanish and vice versa. Investigations on the modeling of Catalan pre-nuclear accents have found robust word-boundary effects (Estebas 2003). Estebas (2003) showed that F0 peaks in Catalan pre-nuclear accents are consistently aligned with respect to the right edge of the word. In her data, peak delay increases as the distance from the stressed syllable to the word boundary increases, i.e. proparoxytones have a longer peak delay than paroxytones, and paroxytones have a longer peak delay than oxytones. This was established with the use of correlations between the duration of the word and a measure of peak delay. Consequently, Estebas (2003) proposes a labeling of Catalan pre-nuclear accents that consists of a monotonal pitch accent, L*, and a monotonal word-edge tone Hω: L*...Hω. Prieto (forthcoming) and Prieto and Estebas (2005) revisited the effect of word boundaries on the location of F0 peaks and the status of word-edge tones in Catalan pre-nuclear accents. These authors examined the production of minimal pairs that are distinguished only by the position of the word boundary. Instances of such potentially ambiguous utterances are as follows: mirava talles '(she) used to watch carvings' versus mira batalles '(she) watched battles'. The results of the experiments showed that peaks were not strictly anchored at the edge of the word since in all sequences peaks were displaced to the post-accentual syllable. However, a consistent and significant word-position effect was found in that peak delay differed between words with lexical stress in the final syllable – mira – and words with lexical stress in the penultimate syllable – mirava. Prieto (forthcoming) and Prieto and Estebas (2005) reject the existence of word-edge tones due to the fact that all pitch accents, including those in oxytones, show peak delay to the post-accentual syllable; however, they do not offer any explanation for the lexical-stress effects on pitch timing they still find.