Identifying Factors’ Effects on Degree of Perceived Accent in the L2 Finnish Uzal, M. (2013) University of Helsinki Faculty of Behavioural Sciences Institute of Behavioural Sciences Department of Speech Sciences Subject: General Phonetics 1
It is suggested that the age of onset of second language (L2) acquisition (AO) predicts the degree of perceived accent (DPA) in the L2. However, previous research presents conflicting results as to which one of the factors plays a more significant role in the DPA. AO gender age at the time of testing first language (L1) use L2 use L1 proficiency/ability L2 proficiency /ability years of formal L2 instruction years of education in the L2 setting amount of L2 exposure/LOR 1. Background of The Study 2
The study was guided by two research questions: 1.What is the degree of perceived accent (DPA) in Finnish produced by L2 learners of Finnish (bilingual native Turkish (NT) children, early bilingual NT children born in Finland and bilingual adult L2 learners)? (see Table 1) 2. Which factors (AO, gender, L1 use, L2 use, L1 proficiency, L2 proficiency, years of formal L2 instruction, years of education in Finland, LOR, age at the time of testing) affect DPA in what degree in L2 Finnish for early bilinguals and late bilinguals? Research Questions 3
The study is unique since never before have early bilingual children born in the L2-speaking country been under investigation, which means that it is the first to evaluate whether they can avoid foreign accent detection in their L2. The Significance of The Study 4
5
To answer the second research question, the bilingual NT children and bilingual adult L2 learners (n = 42, 25 females, 17 males) were assigned to two different subgroups according to their AO: early (n = 25, mean AO = 4, age = 7 – 32, mean = 12, 18 of them were born in Finland) and late (n = 17, mean AO = 25, age = , mean = 32) bilinguals. 2. METHODOLOGY 2.1. Participants 6
42 monolingual native Finnish raters rated the speech samples on foreign accentedness. Two rater types (33 naïve raters and nine expert raters) ensured a balance of sensitivity to accents (Moyer, 2007). As a matter of fact, the linear mixed model analysis showed that raters’ experience of foreign accent (naïve vs. expert rater) or contacts to immigrants (how often they hear nonnative speech: 1.never, 2. rarely, 3. often, 4. everyday) which would suggest their degree of familiarity with accented speech did not have an effect on mean DPA ratings (both p values >.05). Mean DPA ratings given by experienced and naïve raters differed very slightly (mean ratings = 2.5 and 2.7.) and the difference was not statistically significant, F(1, 40) =.885, p = Raters 7
There were three types of speech samples: spontaneous speech, five sentences and three word pairs. All participants were given a delayed repetition task, which involved repetition of eight sentences. In the actual rating task five sentences and three word pairs were used. Three word pairs were obtained by cutting each word pair from the remaining three sentences Speech Samples 8
Rater reliability: The intraclass correlation was p= 0.99, F(718, 28720) = 114.5, p < 0.001, which indicated a very high degree of inter- rater agreement. 3. Results (continues) 9
RQ2: When the AO effect was the focus, pairwise LLM analysis showed that the 25 early bilinguals obtained significantly milder DPA ratings than the 17 late bilinguals (means 2.5 vs. 5.6, p <.01, Bonferroni). None of the 17 late bilinguals spoke L2 without a detectable accent, whereas 5 of the 25 early bilinguals passed for native speakers according to Flege et al.’s z < 2 criterion of native-likeness. The most striking finding was that only four of the 18 early bilingual native Turkish children who were born in Finland passed for native speakers, whereas 14 were detected as speaking accented Finnish. The results revealed that for the early bilinguals AO had a significant effect on the DPA (p =.012) followed by L1 and L2 use (both p values.037), whereas for the late bilinguals none of the factors had significant effects (all p values >.05) on the DPA. Results 11
12
The finding of very high rates of accent ratings for both early bilinguals (80%) and early bilingual NT children born in Finland (78 %) is one of the central findings of the present study and it has important implications for the maturational constraints effect in L2 phonological acquisition. It means that early childhood acquisition does not guarantee the development of native accent in L2. Early AO by itself, does not guarantee native accent in the L2. Even being born in the target language speaking country and starting daycare between the ages of 2,5-3 years do not necessarily mean that L2 will be accent-free. Besides, it must be kept in mind that even though these L2 learners’ age of arrival (AOA) was birth, their real age of extensive exposure to Finnish (AO) was postponed until years, when they started daycare (DeKeyser & Larson-Hall, 2005). 4. Discussion 13
This study confirmed previous research in showing that there is a relationship between AO and DPA. However, for all L2 learners AO did not prove out to have very robust effect on the DPA since it lost its significant correlation with mean DPA (r =.85, p.05). Consequently, it is clear that AO alone cannot explain the DPA which means that the DPA cannot be accounted by a single factor. It seems that attainment of native pronunciation in L2 does not occur as a function of a single factor. It might be the interplay of multiple factors such as the interplay between AO and language use factors. 5. Conclusion 14
There were only 42 L2 learners and there were unequal numbers of participants in different subgroups which affects the generalizability of the study’s findings. Limitations of the study 15
Future study directions should find out the real cause combinations of the AO effect on L2 speech acquisition. Longitudinal studies examining whether foreign accents disappear in time for early bilinguals born in the target language speaking country and if so, which factors specifically enable them to improve their foreign accents needs to be studied. Although this study found that for the early bilinguals, AO and language use patterns explained the DPA, none of the factors explained the DPA for the late bilinguals. Thus, factors that affect DPA for late bilinguals should be investigated further. Future study directions 16
Thank you for your attention! Kiitos! Te ṣ ekk ȕ rler! 17
-The modified version of the language background questionnaire developed by Flege and Mackay (2004) was translated into Finnish and was administered to participants. -A female monolingual native Finnish adult model voice was recorded in a soundproof recording studio and it was presented via a computer’s loudspeaker to participants, representing the spoken standard Finnish norm (Karlsson, 1999). -A delayed repetition technique developed by Flege et al. (1995) was used to elicit speech samples. The speech samples were presented in both written form and aurally to reduce the effect of reading ability to bias DPA ratings (Flege et al., 1995). Each sentence was presented once, followed by 6 seconds of delay which consisted of silence. After 6- seconds of delay a beep sound was played after which participants repeated each sentence. The 6-second delay between each sentence to be produced was to minimize the possible effect of direct imitation (Tench, 1996). After this, spontaneous speech was provided by all of the adults Procedure Procedures for collecting speech samples 18
-All raters completed the rating task individually in a soundproof recording studio and a total of 599 speech samples (72 participants x 5 sentences + 3 word pairs + 23 participants x 1 spontaneous speech) were presented over earphones. -A preliminary information form and a short training session were provided before the rating. -A 9-point equal arriving interval scale was used as a rating technique. -A scale from 1 (no foreign accent) to 9 (very strong foreign accent). The rating was done in a single session lasting between 60 to 100 minutes in three separate blocks. The different sample types were broken up so that the rating task of each block contained only word pairs, only sentences, or only spontaneous speeches. Runs within each sampling type, including speech samples and participants, were randomized. Three blocks were provided in randomized orders to balance ordering effects Procedures for rating DPA 19