| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 565347 | Speech Communication | 2010 | 18 Pages |
Non-rhotic British English speakers and Germans living in England were compared in their use of short- and long-domain r-resonances (cues to an upcoming [ɹɹ]) in read English sentences heard in noise. The sentences comprised 52 pairs differing only in /r/ or /l/ in a minimal-pair target word (mirror, miller). Target words were cross-spliced into a different utterance of the same sentence-base (match) and into a base originally containing the other target word (mismatch), making a four-stimulus set for each sentence-pair. Intelligibility of target and some preceding unspliced words was measured. English listeners were strongly influenced by r-resonances in the sonorant immediately preceding the critical /r/. A median split of the German group showed that those who had lived in southeast England for 3–20 months used the weaker long-domain r-resonances, whereas Germans who had lived in England for 21–105 months ignored all r-resonances, possibly in favour of word frequency. A preliminary study of German speech showed differences in temporal extent and spectral balance (frequency of F3 and higher formants) between English and German r-resonances. The perception and production studies together suggest sophisticated application of exposure-induced changes in acoustic–phonetic and phonological knowledge of L1 to a partially similar sound in L2.
Graphical abstractFigure optionsDownload full-size imageDownload as PowerPoint slideResearch highlights► English native listeners use r-resonance information in unrestricted read speech [82]. ► German and SSBE r-resonances differ spectrally and in temporal distribution [78]. ► Length of stay in the UK influences L2 German listeners’ use of r-resonances [78].
