My Projects & Presentations

John H. G. Scott1,3, Sadi E. Phillips2, Ryan Z. J. Lim1, Charys B. Russell1, 

Isabelle Darcy2, & Lisa Süßenbach1 

1University of Calgary; 2Indiana University, Bloomington; 3University of Maryland, College Park

In contrast to numerous studies of high-variability phonetic training (HVPT), wherein learners encounter numerous (e.g., 4–20) voices, few studies investigate low-variability phonetic training (LVPT), and fewer investigate the relative benefits of these training regimens for beginners at the earliest stages of foreign language, exposure rather than more advanced learners (e.g., [1,2,3]). However, comparison of different levels of voice variability training interventions has crucial implications for the interaction of speech perception with cognitive load, especially at early stages of exposure, before listening routines for the target language are sufficiently automated to incur reduced cognitive load during speech perception [4,5]. Learners must also learn a novel orthography for familiar and novel speech sound categories. Taken together, the effects of phonetic training on category perception—and mapping speech sounds onto orthographic labels during earliest exposure—warrant investigation to inform training and teaching interventions. 

We investigated category perception and grapheme-phoneme correspondence (GPC) awareness in German by adult native speakers of English during the first semester of exposure. Pre-tests included dictation spelling and oddity tasks. Trainees (nL+H = 19) completed a two-day (2 × 2 hr) aural phonetic training prior to classroom exposure. Training with feedback targeted 4 novel German phones (2 vowels [yː øː], 2 consonants [ç x/χ]) alongside familiar [aː iː] and [t k]. Trainees were assigned to LVPT (nL = 10; 1 voice) or HVPT (nH = 9; 4 voices) conditions. Participants also completed spelling and oddity post-tests (4 voices, including 1–3 not in training, no feedback) at midterm and finals (final post-test included novel spelling targets with familiar GPCs). During the training phase, only the LVPT group exhibited improved accuracy, while the HVPT group saw gains only at the end of the term; both training groups outperformed the untrained group at midterm, but only HVPT at final post-test (Fig. 1). We will discuss oddity and spelling results for specific conditions ([6]) and considerations for potential early training and teaching interventions. 

John H. G. Scott1,3, Sadi E. Phillips2, Ryan Z. J. Lim1, Charys B. Russell1, 

Isabelle Darcy2, & Lisa Süßenbach1 

1University of Calgary; 2Indiana University, Bloomington; 3University of Maryland, College Park

Until adult learners of a second or additional language (L2) establish automated selective perception routines [7] for listening to speech sounds of a target language (TL) that reduce cognitive load during speech perception [4], both aural perception of L2 categories and L2 spelling (i.e., matching TL sounds to orthographic representations) present significant challenges during early stages of exposure that can last for years. Alphabetically literate adults face not only interference from L1 phonetic and phonological categories, but also from any previously learned first-language (L1) grapheme-phoneme correspondences (GPCs), which may neither be fully congruent with TL GPCs nor exhibit the same degree of consistencywithin each language (i.e., transparent vs. opaque orthographies). Both levels of L2 learning may becomplicated by dyslexia [6] (20% of people) as L1 strategies become unavailable or inappropriate. Highand Low-Variability Phonetic Training (HVPT/LVPT) may benefit L2 learners in these areas (e.g., [2,3]).

We investigated category perception and GPC awareness of adult L1 speakers of English during their first semester of university L2 German, targeting consonant contrasts ([ʃ]-[ç], [h]-[ç], [h]-[x], [k]-[x]) and vowel contrasts ([uː]-[yː], [oː]-[øː], [uː]-[oː], [yː]-[øː]), with two controls ([t]-[k], [aː]-[iː]). Pre-tests included dictation spelling (words) and aural oddity tasks (nonwords). Trainees alternately assigned to LVPT (nL = 10; 1 voice, different tokens) or HVPT (nH = 9; 4 voices) conditions completed aural oddity phonetic training with text feedback (2 hr × 2 days) prior to classroom exposure. At midterm and finals, participants completed German dictation spelling (word) and aural oddity (nonword) post-tests without feedback. To investigate generalizability of category perception to novel voices, post-tests featured 4 voices. The first post-test featured novel voices for both training groups (LVPT: 3, HVPT: 1); the secondpost-test featured one voice, novel to both groups. Similarly, the final post-test included novel spelling words with familiar GPC targets to test generalizability of TL GPC awareness beyond familiar words.

Figure 1 displays oddity results from [k]-[x] and [uː]-[yː] conditions known to challenge L1 English learners of German [5]. Accuracy for [uː]-[yː] starts high at pre-test, but declines by midterm, recovering slightly for trainees by semester-end. Meanwhile, [k]-[x] accuracy starts slightly lower and rises during training, with gains lasting until midterm for both training groups, but only the HVPT group maintains through semester-end. For [k]-[x], learners without training exhibit accuracy at midterm similar to that of trainees at pre-test; for [uː]-[yː], their midterm accuracy aligns with trainees’ decline. These results suggest important interactions between perception of consonants versus vowels and the listener’s status as naïve pre-learner or early L2 learner [1]. Overall, reduced accuracy for HVPT versus LVPT during training, global second-day training decline, and decrement for LVPT trainees at midterm post-test with multiple voices all suggest impacts of task complexity on cognitive load for speech processing during early L2 exposure [4,7].

We will present quantitative analyses of aural oddity and dictation spelling results (normalized DamerauLevenshtein distance [8]) for specific conditions, relate these findings to broader issues in L2 category perception and LVPT/HVPT research, and discuss considerations for potential early training and teaching interventions.

Features of prosodic phrasing in English are difficult to acquire even for advanced learners. One notably difficult prosodic feature is "thought-grouping", or pausing to delineate meaningful word groups. Learners may split clauses unpredictably or skip prosodic boundaries, creating “run-on” sentences. Either pausing pattern may impede processing of non-native speech, but it is unclear how much they impact processing difficulty. We used a tone detection task as an indirect measure of how split and run-on sentences impact processing difficulty of L2 speech. Thirty-four native English listeners responded to short tones semi-randomly inserted in sentences spoken with target-like or non-target-like thought-grouping. Listeners also judged each sentence as true or false, ensuring they processed them for meaning (not listening for tones alone). Tone detection reaction times (RTs) were compared in three pausing conditions: Original (pause at clause boundary); Run-on (pause absent); Split (additional pause mid-clause). As predicted, non-target-like pausing increased processing difficulty as evidenced by RTs in Run-on or Split conditions being slower than in the Original condition. No clear difference emerged between Run-on and Split conditions. These findings provide corroborating evidence with a processing task that non-target-like pausing results in increased processing difficulty. We discuss the implications of these findings for language pedagogy.