Wendy Herd

An acoustic and perceptual analysis of /t/ and /d/ flaps in American English

Abstract This paper presents an acoustic and perceptual study of alveolar flaps in American English. In the acoustic study, vowel duration differences in disyllabic tokens replicated previous findings in that vowels preceding /d/ were significantly longer than those preceding /t/. Flap frequency was also analyzed based on a method of distinguishing flapped from unflapped stops on a speaker-by-speaker basis. It was discovered that females flapped more often than males and that participants were more likely to flap when they were less aware of the contrast between /t/ and /d/. Contrary to past research, neither word frequency nor morphological complexity affected flap frequency in the present study. In the perceptual study, four naturally produced word pairs were used to manipulate underlying representation (/t/ or /d/), vowel duration preceding the flap, and word frequency. Vowel duration alone did not predict the listeners’ perception of flapped /t/ and /d/; word frequency, where high frequency words were identified correctly more often than low frequency words, and a d-bias, where /d/ flaps were identified correctly more often than /t/ flaps, did prove significant. Unlike previous research, this study uses nonarbitrary values to distinguish flapped from unflapped tokens and draws connections between the acoustic and perceptual results.

Acoustics and perception of emphasis in Urban Jordanian Arabic

Abstract Acoustic and perceptual effects of emphasis, a secondary articulation in the posterior vocal tract, were investigated in Urban Jordanian Arabic. Twelve speakers of Jordanian Arabic recorded both consonants and vowels of monosyllabic minimal CVC pairs containing plain or emphatic consonants in initial and final position to investigate the extent of coarticulatory effects of emphasis. In general, the acoustic correlates of emphasis include a raised F1, lowered F2, and raised F3 in the vowel adjacent to the emphatic consonant, consistent with a narrowing near the uvula. These effects are similar in magnitude for vowels following and preceding emphatic consonants. In addition, the spectral mean of emphatic stops, but not emphatic fricatives, was lower than that of plain consonants. A perception study with cross-spliced natural stimuli explored whether Arabic listeners’ recognition of emphasis is based on information in the target consonant or the rest of the word (vowel+non-target consonant). Results show that the rest of the word contributes significantly more to the perception of emphasis than the target consonant itself. Overall, the acoustic data and perceptual results will address the correlates of emphasis, spread of emphasis, and the asymmetry between stops and fricatives.

Perceptual and production training of intervocalic /d, ɾ, r/ in American English learners of Spanish

This study investigates the effectiveness of three high variability training paradigms in training 42 speakers of American English to correctly perceive and produce Spanish intervocalic /d, r, r/. Since Spanish spirantization and English flapping both affect /d/ intervocalically, the acquisition of the /d/-/r/ contrast proves difficult for English learners of Spanish. The acquisition of the trill /r/ is also problematic because it is a new phoneme for English learners and is articulatorily difficult to produce. Past research reported that high-variability perceptual training improves both perception and production [Bradlow et al., J. Acoust. Soc. Am. 101, 2299-2310 (1997); Wang et al., J. Acoust. Soc. Am. 113, 1033-1043 (2003)] and that production training improves both as well [Hirata, Comp. Assisted Lang. Learning 17, 357-376 (2004)]. However, trainees were able to listen to stimuli during production training, making it unclear whether production training alone transfers to perception. This study systematically controls both training modalities so they can be directly compared and introduces a third training methodology that includes both perception and production. All three training paradigms proved effective. While perception and production trainees primarily made gains in perception, combination trainees made gains in production. The effectiveness of each training modality depended on the nature of the contrast being trained and the modality of the test.

Flapping before a stressed vowel – whatever!

In English, word-internal intervocalic alveolar stops are predominantly flapped when preceding an unstressed vowel (water, charity) and optionally flapped at word boundaries preceding unstressed and stressed vowels (that is, private airplane). In this study, we show that /t/ is flapped in whatever although it has been considered word-internal and preceding a stressed vowel. The data were elicited in a sentence reading task, with four speakers of Appalachian English. The duration of /t/ and the acoustic correlates of stress were examined. A comparison of vowel duration and intensity patterns in whatever versus everwhat (both words are relative pronouns in free relative clauses in this dialect) showed that the second syllable is stressed in whatever. A comparison of /t/ durations showed no significant differences among whatever, watermelons, waterlilies, water buffalo (M = 30.5 ms). These results may be interpreted as: a) whatever is an exception to the word-internal flapping environment, or b) the word-internal flapping environment must be modified to include preceding stressed vowels at morpheme boundaries, or c) whatever consists of two phonological words and falls within the word-final flapping environment. Prosodic and syntactic analyses of free relative clauses consistent with the last interpretation are discussed.

Southern stops: Phonation type differences in Mississippi

While the voicing contrast between American English word-initial stops is often described as relatively uniform across speakers (e.g., voiced segments are produced with short positive VOTs while voiceless segments are produced with long positive VOTs), considerable sociophonetic variation exists. The current study investigates variation in the VOT of voiced and voiceless word-initial stops in pot, bot, tot, dot, cot, got produced in isolation and in carrier sentences. Participants included 40 native English speakers from Mississippi grouped according to their self-identified gender and ethnicity. As previously reported, African American speakers produced significantly more voiced stops with negative VOTs and more fully voiced closures preceding voiced stops than Caucasian American speakers. While speakers did not differ in their production of voiceless VOT when words were read in isolation, African American speakers maintained closure voicing preceding voiceless stops far more often than Caucasian America...

Sociophonetic variation in Mississippi: Gender, ethnicity, and prevoiced plosives

While native English speakers are traditionally reported to produce word-initial voiced plosives with short positive VOTs, recent studies suggest sociophonetic variation exists in the production of these sounds. In separate studies, more prevoicing has been reported for men than women, for African American speakers than Caucasian American speakers, and for southern American English speakers than speakers from other regions. The current study investigates the effects of gender, ethnicity, and context on voicing variation in Mississippi by analyzing word-initial /b, d, g/ as read in sentences by forty native speakers of English grouped according to self-reported gender and ethnicity. A significant effect of ethnicity and an interaction between gender and ethnicity were found. African American speakers produced voiced stops with a larger proportion of closure voicing and produced more fully voiced closures than Caucasian American speakers. While African American men and women produced similarly voiced closur...

Flapping before a stressed vowel: The case of whatever

In English, word-internal intervocalic alveolar stops are predominantly flapped when preceding an unstressed vowel (water, charity) and optionally flapped at word boundaries preceding unstressed and stressed vowels (that is, private airplane). In this study, we show that /t/ is flapped in whatever although it is word-internal and precedes a stressed vowel. The data were elicited in a sentence reading task, with four speakers of Appalachian English. The duration of /t/ and the acoustic correlates of stress were examined. A comparison of vowel duration and amplitude patterns in whatever versus everwhat (both words are relative pronouns in free relative clauses in this dialect, N = 699) showed that the second syllable is stressed in whatever. A comparison of /t/ durations showed no significant differences among whatever, watermelons, waterlilies, water buffalo (N = 533, M = 30.1 ms). These results may be interpreted as: (a) whatever is an exception to the word-internal flapping environment, or (b) the word-i...

Incipient /ay/-Raising in Baton Rouge

Canadian raising targeting the /ay/ diphthong has been reported for a number of dialects of U.S. English. In the United States and elsewhere, however, the incipient—or purely phonetic—stage of diphthong raising, wherein it is triggered only by consonants that are phonetically voiceless, has been notoriously difficult to capture. In such a stage, raising is expected before the /t/ in cite (/saɪt/ → [sʌɪt]) but not before the flapped-/t/ in citing (/saɪtɪŋ/ → [saɪɾɪŋ]). Berkson, Davis, and Strickler (in press) recently discovered incipient phonetic raising in northeastern Indiana, however, and have suggested that extremely short diphthongs immediately preceding a primary stress (as in citation, psychology) may be the very first to undergo raising. We investigate this hypothesis with data from Louisiana. While Yat English spoken near New Orleans has been reported as an /aw/-raising dialect (Carmichael, 2012), we also find /ay/-raising in New Orleans. Furthermore, the question of whether raising is present in...

Region, gender, and within-category variation in American English voiced stops

The two-way voicing contrast in American English stops—particularly in initial position—is often described as a long-lag (e.g., long positive VOT for /p/) versus short-lag (e.g., short positive VOT for /b/) distinction, with less frequent instances of lead voicing (e.g., negative VOT for /b/) attributed to individual variation. Systematic within-category gender and region differences have been reported, however, with more closure voicing found for male than for female speakers (Ryalls, Zipprer, and Baldauff, 1997), and more fully voiced closures for /b/ in female speakers from North Carolina than those from Wisconsin (Jacewicz, Fox, and Lyle, 2009). With this in mind, we investigate the interaction of gender and region in the prevoicing of word-initial voiced stops by comparing the VOTs of male and female speakers from Indiana and Mississippi. Participants were recorded reading three repetitions of a pseudo-randomized list of words including bot, dot, and got. Regional—but no gender—differences were found...

Prevoicing differences in Southern English: Gender and ethnicity effects

Differences in the way VOT is used across languages to maintain stop voicing contrasts have been well-documented, but less research has been focused on VOT variation within voicing categories. For example, native English speakers are generally reported to produce word-initial voiced stops with short positive VOTs, but within category gender and ethnicity differences have been reported in one preliminary study, with male speakers prevoicing stops more than female speakers and with African American speakers prevoicing stops more than Caucasian American speakers (Ryalls, Zipprer, and Baldauff, 1997). For the current study, native speakers of English from Mississippi were recorded reading three repetitions of a pseudo-randomized list of words designed to investigate the effects of gender and ethnicity on the prevoicing of word-initial voiced stops. Participants self-identified their gender and ethnicity in a language background survey completed after recordings. Significant ethnicity, but not gender, differen...