Session 2B

Language learners can use their L1 or L2 when using their L3. This is known as linguistic transfer. There are two powerful factors affecting transfer into L3. Firstly, there is the psychotypological distance: the language with the perceived smaller typological distance (L1 or L2) influences L3 more strongly (Odlin, 1989; Kellerman, 1983). Secondly, there is the L2 status: L2 always influences L3 more than L1 (Williams & Hammarberg, 1998). It is unclear which factor is more relevant.
This study tests these factors by investigating morphological and syntactic transfer between German, English and Dutch. Morphology and syntax have long been considered immune to transfer (cf. Odlin, 1989). Nevertheless, many instances have been observed in recent studies (cf. Jarvis & Pavlenko, 2008). The typological distance between German and Dutch is smaller than between English and Dutch. I hypothesise that the smaller (psycho)typological distance is the more dominant factor and that in language combinations where the (psycho)typological distance between L1 and L3 is smaller than between L2 and L3, transfer from L1 predominates.
I designed an experiment using Dutch data that can pattern both as the corresponding German and English structure: 
(1) Superlatives:
…dat de wilde appel de natuurlijkste/meest natuurlijke appel is. 
…dass der wilde Apfel der natürlichste/*meist natürliche Apfel ist.
…that the wild apple is the *naturalest/most natural apple.


(2) Word order in verb clusters:
…dat de dame gedanst heeft/heeft gedanst.
…dass die Dame getanzt hat/*hat getanzt.
…that the dame *danced has/has danced.


In a pilot study, 16 German learners of L3 Dutch (L2 English) took part in the experiment (self-paced reading task & grammaticality judgement task). Results show that indeed L1 patterns were preferred in L3 for both structures. At the moment, German learners of L3 Dutch (L2 English), British/American learners of L3 Dutch (L2 German) and native speakers of Dutch (baseline) are tested. The results will be included in the presentation.
Questions
How to analyse and present the data in R in a meaningful way?How can I add to the debate on whether linguistic knowledge of different languages is separated in different memory systems or interconnected in a single memory system?How can I add to the debate on whether transfer is relevant in foreign language teaching and whether or not the psychotypologically closer language (L1 or L2) or always the L2 should play a central role in third language teaching? What didactic principles can be derived for foreign/third language teaching of Dutch?

References
Jarvis, S., & Pavlenko, A. (2008). Crosslinguistic Influence in Language and Cognition. New York: Routledge.
Kellerman, E. (1983). Now you see it, now you don't. In S. Gass & L. Selinker (Eds). Language Transfer in Language Learning (pp.112-134). Rowley, MA: Newbury House
Odlin, T. (1989). Language Transfer. Cambridge, UK: Cambridge University Press
Williams, S. & Hammarberg, B. (1998). Language Switches in L3 Production: Implications for a Polyglot Speaking Model. Applied Linguistics, 19, 295-333

Sounding native in one's L2 remains an almost unachievable goal for many adult learners. Since perceived foreign accent stems not only from phoneme divergences but also from phonetic inaccuracies, the speech of L2 users can be judged as strongly accented despite of being highly intelligible (Munro & Derwing, 1995, Language Learning ; Trofimovich & Isaacs, 2012, Bilingualism: Language and Cognition). Sounding foreign can have consequences for the L2 speakers' status in the target society, and therefore phonetic training methods are needed that would decrease the L1 accent in the learner's L2 speech.
We focus on L1 Russian learners of L2 Czech. The speech patterning of Russian and Czech differ in a number of salient aspects, such as word stress position and realization, role of vowel length, and consonant palatalization, and these account for the characteristic Russian accent in Czech. 
To determine the most salient markers of Russian accent in Czech, 10 Russian learners (5 f, mean age = 19 years, mean LOR = 1.6 years) were recorded reading a Czech passage and this material was used in a subsequent perception test. Native Czech listeners (n = 44) first rated the extent of foreign accent and then listed the speech phenomena which in their opinion revealed the non-Czech origin of the speakers. The mean scalar ratings of accentedness strongly correlated with the total count of deviations that the native judges commented on (ρ = 0.83, p= 0.003). The most frequently noticed deviations pertained to the position and phonetic implementation of word stress (lengthening of stressed syllables and reduction of unstressed ones), and, relatedly, to vowel quantity contrasts. Word stress was previously reported as a predictor of high accentedness rating for other L1-L2 language pairs, namely for L1 English speakers of L2 German (O'Brien, 2014, Language Learning), and for L2 English speakers with various L1s (Kang, 2010, System). Since in our experiment, the Russian learners' non-native implementation of various phonetic cues associated with word stress was particularly salient to native Czech listeners, in a subsequent experiment we examined how naïve Russian listeners perceive Czech word-level stress. 
As a next step towards developing an appropriate L2 speech training method we test how word stress in Czech is perceived by naïve Russian listeners in different contexts (when syllables do and do not differ in duration). We designed a perception test, in which listeners hear two- and three-syllabic non-existing words (e.g. [sapa]-[saːpa]-[sapaː]-[saːpaː]) produced by a native Czech speaker, and their task is to mark a syllable, which they consider to be stressed. Data collection is now underway and the results will be presented at the conference. The findings will help us understand which cues, besides vowel duration, Russian speakers use to determine the position of word stress. This will in turn help us design a teaching method that will enable them to perceive and learn the cues for word stress in Czech.

Hungarian is a syllable-timed language with fixed stress on the first syllable [1] and native speakers that exhibit stress 'deafness' [2]. In European Portuguese (EP), however, stress is an important feature for lexical distinctions. Other than being variable and contrastive, EP stress induces vowel quality changes [3].
The present study compares the effectiveness of training for EP stress acquisition by Hungarian learners in two conditions: a) a training focused on suprasegmental cues exclusively, and b) a training focused on suprasegmental cues combined with segmental cues (reduced vowels). We predicted that the latter training yields more learning gains, as there are more acoustic cues, which make stress more readily accessible to listeners.
To test this hypothesis, three groups of Hungarian learners completed three training programs, each consisting of six sessions of discrimination tasks. The training programs were equivalent in design and duration, differing only in the stimuli. In Group Stress (n=13), stimuli consisted of nonwords contrasting in stress only (e.g., ['bikulu]/[bi'kulu]). In Group Stress+Vowel (n=11), stimuli contrasted in stress but contained reduced vowels in unstressed position (e.g., ['zazɨ]/[zɐ'zɛ]). A Control group (n=11) received training in EP vowels, inserted in [zV] monosyllabic nonwords. The 35 participants were learners of EP (aged 18–33), with no previous studies in Portuguese. All the participants completed a pre-test and a post-test, before and after the training, respectively. The post-test included the same stimuli as the pre-test, as well as novel nonwords produced by novel speakers to assess generalization in learning. The pre-test and post-test consisted of discrimination tasks with CVCVCV non-words containing only [i] and [u]. Training effect was analyzed comparing the error rates of the pre-test, the repeted trials of the post-test and the novel trials of the post-test, using a linear mixed effect model.
A significant Training × Intervention group interaction effect was found (F(4, 70)=4.84, p< .05), with pairwise comparisons showing that post-test scores were significantly higher than pre-test scores only in Group Stress (p< .05). No significant differences were found between the scores of the pre-test and the novel trials in the post-test, in either of the groups. 
These results do not confirm our hypothesis and suggest that a focus on suprasegmental features alone is effective in aiding Hungarian learners of EP in stress acquisition. In addition, the results of the generalization trials indicate that a six-session training may not be sufficient to develop abstract representations of stress that can be transferred to novel items.


[1] Markó, A. (2017). Hangtan. In A. Imrényi, N. Kugler, M. Ladányi, A. Markó, Sz. Tátrai, & G. Tolcsvai Nagy (Eds.), Nyelvtan (pp.75–206). Budapest: Osiris. 
[2] Peperkamp, S., Vendelin, I., & Dupoux, E. (2010). Perception of predictable stress: A cross-linguistic investigation. Journal of Phonetics, 38(3): 422–430.
[3] Pereira, I. (2020). Acento de palavra. In E. Raposo, M. Nascimento, M. Mota, L. Segura, A. Mendes, & A. Andrade (Eds.) Gramática do Português. Vol. III (pp. 3399–3425). Lisboa: Fundação Calouste Gulbenkian.