Page 17 - Summer 2018

speed of the nonnative speakers, it is still not as fast as that found in native speakers. That is, it appears that while non- native speakers are generally slower than native speakers, to some extent, the listeners expect it and like it that way.
This sort of higher level accommodation to nonnative speakers is quite pervasive, and, although it is not all be- nign, it often can be quite helpful. A good example of this accommodation was found in another, very different sort of study of language learners. This study examined how learn- ers of a second language can perform in expressing sarcasm. Alexander (2011) created a number of different role-play scenarios in which speakers would perform an individual apologizing for various reasons. She also devised a number of other scenarios in which the speakers would role-play an individual not apologizing, producing speech with markers of sarcasm.
For example, in one scenario the speaker was put into a situ- ation where there was an accident in which they spilled hot coffee on a conversation partner who was a stranger to them. In such a case, the clear expectation was that a person saying “I’m really sorry...” is actually apologizing for the accidental damage to the person of the interlocutor. In a contrasting scenario, the damage to the interlocutor is just a nontangible offense taken by a roommate for something that the speaker does not think is reasonably something to give offense. In this case, saying “I’m really sorry ...” is not an actual apology, and the speakers were to convey this.
Alexander (2011) then had both native American speak- ers and nonnative speakers from Thailand, either living in Thailand or America, perform the role-play scenarios. In ad- dition to analyzing the acoustics of the recordings of these sincere and sarcastic apologies, she took clips from the re- cordings of individuals saying “I’m sorry” and the like, pre- sented them to native listeners, and asked them to decide if the speaker was being sincere or sarcastic.
Alexander (2011) used a similar statistical treatment as Park (2008), treating it as a signal-detection task. Here, the signal to be detected is an apology. The sarcastic apologies, then, are considered nonsignals which the listeners are to correct- ly reject as a nonapology.
Figure 6 plots the summary data from the study. The listeners exhibited d' values for the native productions that were above chance. Native speakers could detect the sincerity in these apologetic clips while rejecting the insincere apologies. Also, a second statistic (c; Figure 6, red circles) measures the degree
Figure 6. Blue line, d' values for native listeners of English detect- ing whether an apology is sincere for different speaker groups; red line, measure of the degree to which responses are biased to one response (c).
to which the listeners were biased toward thinking all of the apologies were sincere or insincere. The values of c for the na- tive speakers are very near zero, indicating very little response bias one way or another. This suggests that they were very ac- curately tuned to pick up on the sarcasm of the responses.
The nonnative speakers, however, did not fare so well; d' values were very near zero, indicating that listeners were very bad at being able to detect whether the nonnative speakers were try- ing to be sincere or not. However, here there is a bright, silver lining to the cloud. Analyses of c indicate a very strong bias in the listeners toward saying that the nonnative listeners were sincere. That is to say, the listeners were strongly disposed to treat any apology, sincere or not, as sincere if the nonnative speaker produced it. Note that this was not the case with the native speakers, where the decisions were much more attuned to the marks of insincerity in the recordings.
So, although there may be many demonstrations of people’s negative adjudication of foreign-accented speech, where there is detection that someone is a nonnative speaker, there is also, then, the possibility that a listener can explain di- vergences from what is expected as being due the speaker being nonnative. It is part of how we make allowances for nonnative speakers. In the case of Alexander’s (2011) work, this appears to explain why native listeners were much more willing to take the nonnative apologies at face value. For nonnative speakers, this is probably a good thing; the pro- duction differences due to our nonnative experience can and often are accounted for by the listeners.
Summer 2018 | Acoustics Today | 15