Fig 1.
Schematic illustration of exposure phase trial structure.
The audio icon on the image represents the beginning of the audio; no such icon was visually displayed to participants.
Table 1.
Participant exclusions in Experiment 1.
Some participants were excluded for multiple reasons.
Fig 2.
Transcription performance during test in Experiment 1.
The values plotted in this graph are adjusted to control for the effects of nuisance variables (see text for details). Dots show individual participants. Error bars show 95% confidence intervals based on non-parametric bootstrap over by-subject means.
Fig 3.
The trial structure of the conditions in Experiment 2 (lengths not to scale).
The Absent condition was identical to that of Experiment 1. The analyses presented below compare the four subtitle conditions against the data from the Absent condition of Experiment 1.
Table 2.
Participant exclusions in Experiment 2.
Some participants were excluded for multiple reasons.
Fig 4.
Transcription performance during test in Experiment 2.
The values plotted in this graph are adjusted to control for the effects of control variables (see text for details). Dots show individual participants. Error bars show 95% confidence intervals based on non-parametric bootstrap over by-subject means. Despite the general trend of longer subtitle delays conferring reduced subtitle facilitation, even subtitles delayed 6000 ms after the offset of the word significantly improve participants' transcription accuracy compared to the Absent condition.
Fig 5.
The number of occurrences of each position-specific phone during both the exposure and test phases of Experiments 1 and 2.
The dot size indicates the number of phones for that particular set of values. Panels show the three stimuli lists, of which participants were randomly assigned to one. Select phones are labeled in IPA, with # indicating word boundaries (see Fig 6 for the subtitle benefit for the same phones).
Fig 6.
Benefit concurrent (y-axis) and delayed (x-axis) subtitles in exposure for each position-dependent phone during test.
The benefit shown here is the difference in transcription accuracy for words containing each phone in the subtitled conditions compared to the Absent condition, in empirical log-odds. Point size indicates the number of instances of each phone during exposure and test. Select phones are labeled with # indicating word boundaries. The realization of word-final /d/, for example, is known to be strongly affected by a Spanish accent [55,56], whereas /ɛ/ is not. To convey a sense of the variability in the data, the 95% confidence intervals for three of the more outlying phones are provided.
Fig 7.
Illustration of two accounts of right-context effects.
(a) Phonetic maintenance: Listeners maintain some level of subcategorical phonetic information about a phone and integrate this information with later contextual information. (b) Uncertainty maintenance: Listeners only maintain the barest amount of subcategorical information to integrate with later context: their relatively certainty about the possible categories. Evidence based on offline categorization tasks employed in previous work is compatible is compatible with either account.
Fig 8.
Illustration of how right-contexts might facilitate accent adaptation.
(a) When a listener first hears an ambiguous /d/ in "dent," there are multiple likely inferences about the phoneme category. (b) But after hearing the right-context of "fender," /d/ becomes the most likely category for the sound. (c) If listeners learn to connect elements of the percept of the sound to the phoneme category, it would suggest that they maintained some amount of subcategorical information about the percept until they received the right-context.