3.2.1
Data acquisition
A female NZE speaker was asked to produce in isolation repetitions of[f6t@],[f6s@]and [f6S@]. These tokens were chosen because they are all non-words with [t], [s] and [S] in intervocalic position, and they involve only one lingual gesture, which is required for palatographic analysis. The speaker was presented with the non-words in orthographic form (fotter,fosser andfosher) consecutively on a computer screen until 100 repetitions were gathered. In between each word, their tongue was painted with a mixture of charcoal powder and olive oil. They read out the words to a microphone, then placed their mouth onto a mirror attached to a palatographic bench so as to form an 45◦ angle with the upper jaw. A photo was taken using a camera attached to the other end of the bench and the speaker was asked to rinse their mouth after the photo was taken. We used an AKG C520 head-mounted microphone. The signal was pre-amplified using an external USB-Pre device. Recordings were done using Audacity with a PC. They were saved as WAV files at 44.1 kHz 16 bit. Photos of the palate were taken using a Canon EOS 1000D camera and a 18-55mm kit lens focusing on the mirror with manual
exposure and aperture settings to maintain consistency across photographs. Photos were saved as JPEG files. The palatographic bench was designed by the Laboratoire Parole et Langage. Table 3.1 presents the realisations of the 99 tokens that were collected. Tokens realised as fricatives were selected for analysis. First we present our acoustic results and then we move on to our articulatory results.
speech sounds realised as a fricative realised as a stop or an affricate
[t] 27 6
[s] 33
[S] 33
Total 99
Table 3.1– Number of realisations
3.2.2
Acoustic results
We measured the duration of the intervocalic consonants on the acoustic signal (the start marker was positioned at the offset of modal voicing for the first vowel and the end marker was positioned at the onset of modal voicing for the second vowel). The duration of the intervocalic consonants is shown on the box and whiskers plots in figure 3.2.
0.1 0.12 0.14 0.16 0.18 0.2 [t] [s] ESH Duration (s)
Figure 3.2 – Box and whiskers plots showing the duration of fricated /t/,/s/and/S/.
We run Wilcoxon non-paired tests for equal medians and found very marked differences in duration: /t/ is shorter than /s/ and /S/ (p < 0.0001) while /s/ and /S/ do not differ in duration (p= 0.058). We now turn to our palatographic results.
3.2.3
Articulatory results
A total of 90 palotagrams were visually inspected. We analysed tokens realised as fricatives (we discarded the 6 /t/ tokens produced as stops or affricates) and we discarded two tokens of /S/ and one token /s/ where the speaker unintentionally made full tongue-palate contact before we took photographs. The first series of palatograms are shown of figure 3.3. All palatograms can be found in appendix A for reference. We observed the following patterns: /t/ tokens are generally produced with a very large fricative channel compared to /s/ and/S/
tokens. Differences in channel between /s/ and /S/ tokens are less evident. The location of the constriction for /t/ and /s/ tokens is very similar and it is slightly more anterior relative
/S/ tokens. Finally the side contacts are relatively thinner for /t/ and /s/ tokens compared to
/S/ tokens. Thus it seems that the fricated /t/ differs mainly between /s/ and/S/ tokens in terms of the size of the channel. /t/ seems closer to /s/ with respect to constriction location and side contacts than it is to /S/.
Figure 3.3 – From left to right: palatograms of fricated [t], [s] and[S]. Series 1: tokens 1 to 15.
3.2.4
Short discussion
It is difficult to draw any definitive conclusions about the articulatory make-up of fricated/t/in NZE since this pilot experiment is limited to one speaker. The gathering of static palatography data is time consuming and recruiting a large number of participants is not an easy task. But this will be made easier in the near future as computational tools for quantitative analyses of palatograms are being currently developed at the Laboratoire Parole et Langage (Legou et al., 2008). An extensive palatographic study presents both theoretical and documentary implications: to further investigate further the mapping between acoustics and articulatory data and to extend documentary work on NZE. Further research could also include a larger set of acoustic measures such as those presented in Jongman et al. (2000) to distinguish places of fricative articulation in American English. They measured spectral peak location, spectral moments (Forrest et al., 1988), locus equations (Sussman and Shore, 1996), normalised RMS amplitude and relative amplitude (Hedrick and Ohde, 1993). Jongman et al. (2000) report that spectral peak location, spectral moments and amplitudinal information prove successful and robust in distinguishing between all places of articulation. Jones and McDougall (2009) used similar measures and concluded that fricated /t/ in Australian English was closer to /S/
than to /s/ acoustically. They also inferred from their acoustic results that /t/ was closer to /S/. Our results show that /t/ is closer to /s/ than to /S/ on articulatory grounds. The articulatory patterns that we observed are similar to the patterns reported by Pandeli et al. (1997) on Irish English slit-t.
According to their acoustic results Jones and McDougall (2009) argued that the main difference between fricated /t/ and /S/ in Australian English is a question of duration. In their view, duration is the most likely candidate to distinguish between these two fricatives. However they did not address directly whether /t/ was perceptually closer to /s/ or to /S/. Our palatographic results show that /t/ is closer to /s/ in NZE in articulatory terms and we will further investigate this finding by means of a perception experiment.