Phonetic implementations of Seoul Korean stops have been examined mostly in phrase-initial positions, primarily focusing on the tonogenesis-like sound change which involves a shift in the primary acoustic property for differentiating aspirated and lax categories. Word-medial stops have been much less discussed, except in the context of inter-sonorant voicing of lax stops. To address this gap, the current study provides a comprehensive analysis of aspirated, lax, and tense stops in Seoul Korean across three prosodic positions, Intonational Phrase (IP)-initial, Accentual Phrase (AP)-initial (inter-sonorant), and word-medial (also inter-sonorant), considering various acoustic properties, namely, stop burst duration, closure duration, post-stop f0 (fundamental frequency), F1, H1*-H2*, and voicing during closure. Based on experimental and corpus data, we report that Seoul Korean stops show distinct acoustic patterns in phrase-initial versus word-medial positions. In IP- and AP-initial positions, burst duration and f0 are the primary acoustic properties that distinguish the three categories, with H1*-H2* further differentiating tense from non-tense stops. In word-medial position, however, closure duration, voicing during closure, and burst duration emerge as the main distinguishing properties, though some f0 differences between lax and aspirated stops are observed. This f0 difference is especially noticeable in non-high vowel contexts, suggesting that the tonogenesis-like change may be spreading beyond phrase-initial positions. Overall, the acoustic implementation of the Seoul Korean three-way laryngeal contrast is highly dependent on the prosodic position of the stops, with this positional variation being most noticeable for lax stops.

Jasmine was an American seventh grader whose first language (L1) is Punjabi and second language (L2) is English. A good speller, she tied for first place at her regional spelling bee. She maintains that she then correctly spelled the tie-breaking word: tomography. After requesting an instant video replay of her word, the judges declared Jasmine's spelling incorrect: <d-o-m-o-g-r-a-p-h-y>. We examine three questions: Is Jasmine's presumed-high word-recognition ability consistent with such an error? Did her L1 interfere with her L2 pronunciation? Did an L1-influenced pronunciation cause the judges to perceive her word-initial /t/ as [d]? After inquiring of Jasmine about the spelling event itself, we examined the considerable research on Indian English, spelling, speech production, and speech perception; we also audio-recorded Jasmine's spoken language at various style levels to assess the degree to which Punjabi may be influencing her English pronunciation. The first question remains without a confirmed answer, though our assumption is that it is ‘no’. Work on Indian English indicates a probable ‘yes’ to the latter two questions. Spectrographic analysis of Jasmine's word-initial English [th]-sounds demonstrates the shorter voice onset times associated with the nonaspirated /t/-sounds of Punjabi and Indian English. It is probable that Jasmine pronounced the ‘t’ [th] of tomography with little or no aspiration, and that the judges heard the near-unaspirated [t] as a [d]. Jasmine's experience demonstrates that the Scripps National Spelling Bee's sole reliance on dictionary pronunciations of Standard American English might well be adjusted to accommodate contestants who speak a nonstandard, or a standard but non-American, variety of English.