Université du Québec à Montréal Speech production in prelingually deaf children with a cochlear implant: Evolution of inter-rater agreement Marie-Eve Bouchard1, Lucie Ménard1, Marie-Thérèse Le Normand2, Henri Cohen1,3 1. University du Quebec à Montreal; 2. INSERM Robert Debré Hospital, Paris, France; 3. CNRS-Paris V FRE 2987 UQÀM Université du Québec à Montréal OBJECTIVES: 1) determine how cochlear implantation can improve upon the articulatory skills of prelingually deaf children, using a sublexical unit (syllable); 2) assess the effect of age at implantation and mode of communication on articulatory skills. METHODS: Two experienced (EXP) and 40 inexperienced (INEXP) judges were asked to identify the vowels or consonants in the 891 stop C-V repertoire extracted from the speech of 12 implanted PDC recorded at 6, 18 and 36 months post-surgery. It was hypothesized that improvement in pronunciation should result in an increase between EXP and INEXP agreements for the syllables identification over time. Perceived vowels were located in three areas of the vocalic space (low central, high anterior, high posterior) whereas consonants were classified according to place of articulation (bilabial, alveolar, velar). RESULTS: ANOVAs with repeated measures showed a clear effect of time for vowel perception, especially low central. More limited progress was observed for consonants. Age at implantation and mode of communication had no effect on performance. DISCUSSION : Overall results suggest that the auditory information provided by the CI supports the acquisition of global speech skills. This increased perception, however, may not be sufficient in constructing or refining the articulatory strategies required to produce more intelligible linguistic units deprived of contextual information. Abstract Speech stimuli -The speech stimuli were taken from a 20 min standardized free play session. -Speech productions were audio- and videorecorded at 6, 18 and 36 months post-surgery, and segmented into syllables extracted from the audio signal using Gold Wave. -All extracted syllables were transcribed by two experienced transcribers. -Only stop syllables were selected. PRAAT was used to verify that segments respected the acoustic criteria of a stopped syllable. -Given that the intended phonological target was not always known, consonants were divided into 3 areas according to their place of articulation: • Area 1: labials (b, p) • Area 2: alveolar (d, t) • Area 3: velar (g, k) -Three areas were also created for the vowels: • Area 1: central open (a) • Area 2: front closed (i, e, y) • Area 3: back closed (u, o). Procedure -Two perception tests were created using PRAAT, for the identification of vowels and consonants, respectively. -Each test had two conditions: the B and NB (or mixed) condition. -20 listeners were given the vowel identification test, and 20 others the consonant test. - Syllables were presented only once. -Judges were asked to select the vowel or consonant they heard. Methods Participants -12 prelingually deaf French children, 8 girls and 4 boys; into 3 age of implantation groups of 4 children • Group 1 : 25 - 33 months • Group 2 : 46 - 50 months • Group 3 : 72 - 78 months -10 children implanted with the the Nucleus 22 device ( SPEAK strategy); 2 with the Clarion CI (CIS strategy). -All subjects enrolled in rehabilitative program that emphasizing use of oral language; communicative support (French Signed Language, Signed French, Cued Speech) was offered to some. -40 normally hearing French adult listeners, inexperienced with the speech produced by deaf individuals, were recruited as judges. Figures 1 and 2 present the most frequently observed non-correspondences in identification between Experienced and Inexperienced listeners over the 3 testing sessions for Vowels and Consonants, respectively. Figure 1. Most frequently observed non- correspondences in identification between EXP and INEXP listeners for vowels at 6, 18 and 36 months post-implantation FRONT HIGH BACK OTHER CENTRAL OPEN . Figure 2. Most frequently observed non- correspondences in identification between EXP and INEXP listeners for consonant at, 6, 18 and 36 months post-implantation MANNER STOPPED PLACE OF ARTICULATION LABIAL ALVEOLAR VELAR V O I C E D U N Main effect of time for both vowels and consonants; no significant interaction F= 8.15, df= 1, p<0.05 Main effect of area — vowels only (Area 1> Area 2 and 3) F=11.843, df= 2, p<0.01 No effect of age at implantation and mode of communication. Discussion By stimulating the auditory nerve, cochlear implant (CI) permits profoundly prelinguistically deaf children (PDC) to gain access to information about spoken language. When they learn to speak and articulate, PDC often rely on visual and tactile/kinesthetic cues which are not as rich as the ones provided by the auditory channel. As a result, PDC present a broad range of abnormal speech behaviors that make it difficult for them to be orally understood by their environment. Most studies designed to speech production in this population have presented judges with speech material integrated into context. Though highly instructive on global speech development, the effect of contextual information on perception cannot be overlooked. Furthermore, these studies do not explain how these implanted PDC make use of the new acoustic feedback to construct or refine their speech production skills and how it impacts on their capacity to produce intelligible speech. Introduction Most speech production studies with PDC have presented EXP or INEXP judges with material integrated into context. The general conclusion of these studies was a continuous improvement in speech clarity as well as an influence of age at implantation and mode of communication. The present results showed that the effect of contextual information on perception cannot be overlooked; when sublexical units are used, more limited progress is observed. In adition, no effect was obtained with respect to the influence of the two clinical parameters studied. Significant improvement was seen for vowels only. Despite their greater visibility, bilabial consonants showed progression comparable to their less visible counterparts, the velar phonemes. Results thus suggest that the auditory information provided by the CI supports the acquisition of global speech skills but that perception may not be sufficient to construct or refine the articulatory strategies required to make linguistic units deprived of contextual information more intelligible to either EXP or INEXP listeners. This result is in accord with others reporting that the more accurately a speaker discriminates a contrast, the more distinctly the speaker produced that contrast (Perkell et al., 2004). Such a relationship between speech production and perception is compatible with a model of speech production in which articulatory movements are planned primarily in auditory space. Results Objectives of the study - Follow the articulatory progress over the first 36 months following auditory restoration in a group of implanted prelinguistically deaf children (IPDC) by comparing the perception of experienced (EXP) and inexperienced (INEXP) judges of speech (syllables) produced; -Examine whether perception of vowels differs from that of consonants; -Assess the influence of age at implantation, mode of communication on the IPDC performance. -ANOVAs with repeated measures: Main effect of the phone class (vowels > consonants): F= 31.53, df= 1, p<0.01)