The effect of non-conclusive melodic rises on Czech speech sounding French

The effect of non-conclusive melodic rises on Czech speech sounding French Tomáš Bořil, Pavel Šturm, Radek Skarnitzl, Marie Hévrová (Charles University, Prague), Barbara Köpke (Université Toulouse — Jean Jaurès) T. BOŘIL — P. ŠTURM — R. SKARNITZL — M. HÉVROVÁ — B. KÖPKE ABSTRACT This paper is based on astudy of first language attrition in Czechs living in France, which reveals that Czech expatriates in France use prominent rises in non-conclusive intonation patterns in their native language. We manipulated the speech of six expatriates by reducing the fundamental frequency (f0) range of non-conclusive rises in aphrase, and the speech of six Czech control speakers by expanding the f0 range to mimic the French-like prominent rises. The manipulations served, alongside filler items, as the basis for aperception test in which 37 native Czech listeners assessed how much the speakers’ pronunciation manifested marks of along-term stay in France. The results confirmed our hypothesis that expanding the control speakers’ f0 range would increase the perceived effect of French. However, reducing the f0 range in the expatriate group did not yield lower French-effect ratings, most likely due to the presence of other pronunciation features in their speech. KEYWORDS cross-linguistic influence, intonation, L1 attrition, non-conclusive melody, speech perception DOI https://doi.org/10.14712/18059635.2022.1.7 1 INTRODUCTION In today’s world, bilingualism or multilingualism seem to be increasingly frequent, with many people living in acountry where the ambient language is not their native language (L1). As aresult, the number of speakers who use asecond or foreign language (L2) on adaily basis is increasing. It is well known that when there are two or more languages spoken by one person, they affect each other in various ways, and several accounts have been proposed over the past century which attempt to explain and predict the resulting effects (Myles 2013). For along time, the study of crosslinguistic influence (CLI) has involved the L1-to-L2 direction, i.e., the effect of the speaker’s native language on the language that he or she was learning, typically referred to as second language acquisition. In the domain of speech and pronunciation, the almost inevitable result of such interactions between the sound patterns of the two languages is aforeign accent (Moyer 2013). Examining the opposite direction of CLI, where one’s L2 exerts some kind of an influence on their native language, represents acomparatively more recent stream of research and is sometimes labeled first language attrition (although see section 1.1 for amore nuanced approach). L1 attrition concerns anon-pathological decline of L1 skills previously possessed by the speaker OPEN ACCESS 126 LINGUISTICA PRAGENSIA 1/2022 (Köpke and Schmid 2004). More specifically, we can talk about adaptations of sound patterns, lexical representations or grammatical structures of aspeaker’s L1 due to the influence of an L2. Although terms like “cross-linguistic influence” or “language attrition” may be unknown to most ordinary people, the concept of “losing one’s mother tongue” is far from unfamiliar, as it appears in hundreds of hits on the Internet, including newspaper articles, blogs, and websites (cf. Sučková 2020: 2). In addition, many people have witnessed this phenomenon in their family members returning from aprolonged stay in aforeign country or in well-known persons living abroad. In the Czech context, the tennis players Ivan Lendl or Martina Navrátilová can serve as famous examples. The present study is based on research into the attrition of Czech as the L1 in speakers living in L2 French environment (Hévrová, Bořil and Köpke 2020; Hévrová 2021). This research reveals that Czech speakers living in France manifest, among other things, specific forms of melodic behavior, particularly in the non-conclusive patterns, which tend to be marked by aprominent rise in their speech (see also below). The aim of our study is to verify the perceptual salience of this non-conclusive rising melody as amarker of CLI. We will use controlled manipulations of speech melody to help us assess the role of non-conclusive rises in evoking aperceived effect of L2 French in native speakers of Czech who have in fact never spent asignificant amount of time in aFrench environment. In the following two subsections of this paper, we will briefly introduce the study of language attrition and CLI in bilinguals, also summarizing the main findings concerning the pronunciation of Czech speakers living in France (section 1.1), and describe the differences in the melodic patterning of Czech and French which are relevant for our research (section 1.2). In section 1.3 we will address the research questions and hypotheses of the study. Section 2 will detail the experimental procedure and the compilation and administration of the listening task. In sections 3 and 4, we will present and discuss the results, respectively. 1.1 CROSS-LINGUISTIC INFLUENCE AND ATTRITION IN BILINGUALS In most recent definitions, bilingualism relates to the use, and the necessity of use, of multiple languages by one speaker (Grosjean 2013; Kohnert, Ebert and Pham 2022); both conditions are likely to evolve over the speaker’s lifespan and entail modifications in the balance between the different languages spoken by the speaker. Crucially, bilinguals may not have equal fluency in their languages and may speak with aforeign accent in one of them (Grosjean 2013). Given the huge variation in bilingual profiles, however, it is common to differentiate bilinguals along amultitude of dimensions; for example, early and late bilinguals are distinguished with respect to the age of acquisition (AoA), which is known to be responsible for differences in the brain structure involved in the processing of different languages (e.g. Berken, Chai, Chen, Gracco and Klein 2016; Birdsong 2018). However, it is now well established that the link between the age of onset of L2 learning and the ability to attain anative-like L2 proficiency is complex (see e.g. Birdsong 2018 or Hyltenstam and Abrahamsson 2003 for areview). Specifically, age effects have been shown to differ depending on the linguistic domain, with the sound patterns of the L2 particularly affected. Moreover, OPEN ACCESS T. BOřIL — P. ŠTURM — R. SKARNITZL — M. HÉVROVá — B. KöPKE 127 other factors modulate ultimate attainment in the target language, such as the learner’s motivation (Dörnyei 2010) and willingness to communicate (MacIntyre, Clément, Dörnyei and Noels 1998), the type of memory applied during L2 learning (Baddeley 2012; Morgan-Short, Faretta-Stutenberg, Brill-Schuetz, Carpenter and Wong 2014), or the quality and quantity of L2 input (Moyer 2009). Gradually, it has come to light that the L1, even in late bilinguals, is not as immutable as it was believed to be. Early studies in the 1980s started to show that the speakers’ L2 may affect their L1. It was Sharwood Smith (1983) who introduced the term “cross-linguistic influence” (to replace the negatively perceived term “transfer”) and used it also for the reversed, L2-to-L1 direction. It is interesting to note, however, that anecdotal observations about the loss of the speakers’ L1 were offered decades earlier by Vilém Mathesius, who discussed the speech of Czechs living in Slovakia (Mathesius 1935: 102). In the context of the influence of an L2 on an L1, Pavlenko (2000) distinguishes five types of CLI. When linguistic elements from the L2 are added into the speaker’s L1, Pavlenko talks about borrowing; in our Czech speakers residing in France, one may observe afrequent occurrence of an epenthetic schwa at the ends of words (called “stuck schwa” by Hévrová 2021). Some features of the two linguistic systems may approximate each other, resulting in aunitary system distinct from both the L1 and the L2; this situation was termed convergence and appeared, for instance, in the sound shape of the rhotic sound in some of Hévrová’s speakers. In ashift, L1-typical values or structures adjust to approximate those of the target L2; this may be exemplified by the short close-mid front /ɪ/ of Czech to be pronounced as amore close and peripheral [i] due to the effect of L2 French. Restructuring occurs when L2 elements are incorporated into one’s L1, leading to areanalysis of existing L1 elements (see Ulbrich and Ordin 2014 for an interesting example); as for Czech speakers residing in France, the application of French intonation patterns in Czech speech may be used as an example. It is only the last type that is, in Pavlenko’s (2000) typology, called L1 attrition, and it involves the inability to produce an element of the L1 due to the effect of the L2 and due to decreased L1 use and/or input (De Leeuw 2019; Kříž and Chromý 2020). In the domain of the effects of L2 sound patterns on those of the L1, or what is often referred to as phonetic attrition1, most studies have focused on late bilinguals whose L1 or L2 was English (e.g. Bergmann, Nota, Sprenger and Schmid 2016; Chang 2012; Sancier and Fowler 1997; Ulbrich and Ordin 2014). There are only very few studies which address phonetic attrition in contexts where the Czech language acts as either the L1 or L2. In her doctoral dissertation, Sučková (2020) examined Anglophone expatriates living in the Czech Republic; she measured VOT values in the participants’ fortis and lenis plosives, the degree of devoicing of word-final lenis plosives, and the speaker’ realizations of English dental fricatives. The results were not always straightforward. On the one hand, VOT in fortis plosives was found to be shorter in some of the English words, supporting ashift towards L2 Czech values; 1 Note that some researchers also talk about phonetic drift to describe effects of ashorter stay in an L2 environment or even ashort exposure to an L2 on aspeaker’s pronunciation in his or her L1 (Tobin, Nam and Fowler 2017). OPEN ACCESS 128 LINGUISTICA PRAGENSIA 1/2022 on the other hand, however, word-final lenis plosives were actually produced with more voicing and the dental /ð/ more canonically than what speakers in the native English control group produced. Sučková (2020: 128) attributes the latter finding to greater “enunciation effort”. Note that Pavlenko’s (2000) typology does not consider such apossibility, which may be termed dissimilation (see e.g. De Leeuw 2019 or Flege and Bohn 2021). Šreková’s (2020) MA thesis focused on phonetic CLI in the L1 speech of Czech students majoring in English translation studies and reported only insignificant effects of English VOT values on those in Czech. Finally, as was mentioned above, Hévrová’s (2021) doctoral dissertation examined L1 attrition in Czech speakers residing in France. Other studies of L1 attrition with Czech as an L1 or L2 concentrate on other linguistic domains such as the lexicon (Kříž 2020) or morphology (Zajícová 2010). Returning to phonetic attrition, most studies have examined segmental aspects of speech, but the L2-on-L1 effect has also been addressed in prosody, and specifically in the melodic domain. For instance, Mennen (2004) examined L1 Dutch speakers of L2 Greek; although the two languages use apre-nuclear rise in declarative intonation, the phonetic implementation of the rise differs. The study revealed abi-directional CLI in most of the speakers who were examined, with the melodic peak alignment different from either of the native speaker control groups. De Leeuw, Mennen and Scobbie (2012) also investigated tonal alignment of pre-nuclear rises in the L1 of late German-English bilinguals. The authors predicted an assimilation effect concerning both the onset and offset of the rise; overall, however, their analyses confirmed an L2 effect only on the alignment of the onset, which occurred significantly earlier in the German of the bilingual speakers than in that of the monolingual control group. Intonation has also been mentioned by listeners as contributing to anon-native impression of speakers living in an L2 country (see for example Mayr, Sánchez and Mennen 2020). Since the study reported in this paper is motivated by Hévrová’s (2021) findings, we will summarize her research in more detail. Aperceptual experiment revealed that native Czech listeners, students of phonetics, do perceive aFrench accent in the Czech speech of speakers residing in France. The listeners pinpointed several phonetic features which, in their opinion, “flagged” the speakers as sounding “French”. These features included vowel quality (typically /ɛ/ and /ɪ/), the replacement of /ɦ/ by /x/, the realization of /r/ and, crucially for the present study, intonation. Intonation of the Czech speakers affected by French was described as “very rising”, having “aprominent rising pattern” or “rising ends of phrases” (Hévrová 2021: 149). In asubsequent acoustic study, the author identified the presence of CLI in the speakers’ production of unrounded vowels (F1 of [aː] and [ɛː], F2 and F3 of [ɪ] and F3 of [iː]), as well as in phonetic properties of the Czech [r] and [x]. In addition, the acoustic study confirmed ahigher proportion of rising melodic contours in the non-conclusive pattern than in the control group. Finally, the Czech speakers living in the French environment used the stuck schwa (see above) more frequently. It is the rising melody in non-conclusive patterns which provided an impetus for the present study. Hévrová’s results lend support to the common assumption of models like the SLM-r (Flege and Bohn 2021) or PAM-L2 (Best and Tyler 2007) that L1 and L2 sounds of late bilinguals OPEN ACCESS T. BOřIL — P. ŠTURM — R. SKARNITZL — M. HÉVROVá — B. KöPKE 129 exist in acommon phonetic space; Flege and Bohn (2021: 22) refer to a“composite L1–L2 phonetic category”. As such, the sounds can interact with one another, which results in perceptible and measurable changes even in the speakers’ native language. 1.2 NON-CONCLUSIVE INTONATION PATTERNS IN CZECH AND FRENCH It seems to be alinguistic universal that non-conclusive intonation patterns signal asyntactic or semantic boundary in speech and, at the same time, the speaker’s intention to continue speaking. Such patterns alert the listener that the utterance is not yet complete and that continuation of the utterance should be expected (Di Cristo 2016); for that reason, non-conclusive patterns may also be called continuing intonation patterns (ibid.). While it is beyond the scope of this paper to provide asystematic comparison of intonation in Czech and French, it is necessary to at least briefly compare the nonconclusive nuclear melodic patterns in the two languages. Before turning to this comparison, it should be pointed out that acontrastive analysis of Czech and French intonation (Duběda 2012: section 4.3) has identified significantly greater melodic variability in French than in Czech; this is achieved not only by ahigher pitch range, but also by more frequent melodic movements. As for anon-conclusive melody, the classic description of Czech intonation (Daneš 1957) distinguishes one unmarked and four marked phonetic realizations of the non-conclusive phonological pattern; the marked realizations also take into account whether the last pre-nuclear syllable is lower or higher than the onset of the nuclear tone (see also Palková 1994: 313ff). In the most recent systematic treatment of these patterns in Czech, Volín (2008) used k-means cluster analysis to examine nonconclusive intonation patterns produced by 252 speakers in areading of continuous text. Focusing on the nuclear tone along with two preceding syllables, the author identified six realizations of this functional pattern; these are illustrated in Figure 1. Volín has shown that realizations with aslightly rising intonation (appearing in the upper row of the figure) were most frequent. The other phonetic realizations, in the lower row of Fig. 1, involve aflat or falling melodic contour. Crucially, therefore, it is obvious that Czech does not regularly use ahigh rising melodic realization of nonconclusive tones. Figure 1.Aschematic illustration of Czech non-conclusive patterns (adapted from Volín 2008: 177); note that only one pre-nuclear syllable is shown for the sake of comparison with French. The empty circle corresponds to the pre-nuclear syllable, the filled circle to the nuclear syllable, and the solid line to the melodic movement in the tail of the prosodic phrase. OPEN ACCESS 130 LINGUISTICA PRAGENSIA 1/2022 In French, it appears to be precisely the non-conclusive (continuation) intonation patterns which are responsible for the overwhelmingly rising impression (Delattre 1953). Non-conclusive intonation patterns in French were examined in anumber of studies. As summarized in Delais-Roussarie, Post, Avanzi et al. (2015), tonal movements associated with non-final intonational phrases are most typically rising (cf. also Duběda 2012), with the rise manifested usually on the last syllable (Santiago 2019). However, Di Cristo (2016: 218) listed another five possible realizations of nonconclusive patterns which, similarly to those observed in Czech, also include falling movements and aflat melody (see Figure 2). Figure 2.Aschematic illustration of French non-conclusive patterns (adapted from Di Cristo 2016: 218). The empty circle corresponds to the pre-nuclear syllable, the filled circle and the solid line mark the nuclear syllable and its melodic movement. To summarize, non-conclusive intonation patterns are more frequently rising in French than in Czech, with the rise typically occurring on the last syllable and being more prominent. As described in section 1.1, Hévrová (2021) documented assimilation effects in non-conclusive melodic patterns in Czech speakers residing in France in what may be regarded as restructuring (Pavlenko 2000), and her Czech listeners mentioned rising intonation as the “revealing” feature. From the perspective of L2 acquisition, we may expect this feature to be acquired early by Czech speakers living in France due to its semantic weight and perceptual prominence (Pešková 2020). This study aims to provide another perspective on the perceptual salience of rising melody as amarker of French influence in the speech of Czech speakers. 1.3 RESEARCH QUESTIONS AND HYPOTHESES In the reported experiment, we will manipulate fundamental frequency (f0; roughly, the acoustic correlate of perceived pitch) in the speech of two groups of Czech speakers. In those living in the Czech Republic, f0 will be changed to mimic the prominent rises, whereas in those who have been living in France for aperiod of time, the rises will be reduced. We hypothesize that both changes will modify the listeners’ perception. Specifically, we predict that 1) introducing more prominent rises into the speech of Czech speakers will evoke an influence of French in the ears of our respondents, and that 2) flattening the rises in the speech of those residing in France will render the impact of the French language less distinct. OPEN ACCESS T. BOřIL — P. ŠTURM — R. SKARNITZL — M. HÉVROVá — B. KöPKE 131 2 METHOD 2.1 SPEAKERS We used recordings of 32 native speakers of Czech who talked freely for several minutes on ageneral topic (usually answering questions like How was your day? What is your typical day?). Half of the speakers had never spent asignificant time (over amonth) in aFrench-speaking country, which should preclude any CLI from the French environment, whereas the other half did, ranging from two months to 28 years (see Table 1). In these speakers, it is reasonable to expect substantial CLI on their Czech pronunciation. The two groups of speakers will be referred to as the Control group (n = 16) and the Expat group (n = 16) for convenience, which is not meant to imply that they necessarily consider themselves as expatriates. Age summaries are provided in Table 1. The recordings of the Control group were obtained in aquiet room at the Czech Technical University using ahead-mounted Beyerdynamic Opus 55 microphone connected to ahigh-quality portable recorder Zoom H1. The recordings of the Expat group were acquired in arecording studio at University Toulouse Jean-Jaurès using aNeumann TLM 49 microphone and the MOTU ULmk3 sound card. Variable Group Min Max Mean SD Age Control 21 49 32.9 9.7 Expat 20 49 31.6 10.2 Length of residence Control — — — — Expat 0.17 28.25 8.1 9.2 Table 1.Age (in years) and Length of residence (in years) for the Expat and Control groups of speakers. 2.2 MATERIAL The material included target and filler stimuli. One target stimulus was selected from each of six Control and six Expat speakers, amounting to 12 stimuli; note that the demographic characteristics of the selected speakers do not substantially differ from those reported in Table 1. The selected portions were meaningful stretches of speech extracted from the larger spontaneous recordings, with the nuclear tone spanning at least two syllables. Their key characteristic was the presence of two rising nuclear non-conclusive contours. The length of the extracts varied from 2.60 to 6.85 seconds due to tempo, but in terms of prosodic structure it was kept approximately the same across speakers, ranging between 6 and 8 stress groups. Filler stimuli were extracted from the recordings of the remaining speakers (9 Control and 9 Expat), and differed from the target stimuli in the structure of the phrases. Specifically, two rising contours were not allowed in asingle phrase, and avariety of contours differing in shape and level were used. The length of the phrases also varied more as compared to the target stimuli. The target stimuli furthermore underwent acoustic manipulation of f0. First, f0 tracks extracted in Praat (Boersma and Weenink 2021) using the default settings OPEN ACCESS 132 LINGUISTICA PRAGENSIA 1/2022 were manually corrected for octave jump errors and saved as PitchTiers. Second, the portions of the recording with the relevant melodic movements were identified and annotated in aTextGrid. For each target contour, we identified in the speech signal and marked in the TextGrid (i) what would be the melodically stable part of the nuclear tone and (ii) the rising part of the tone, corresponding to the last syllable of the tail. Third, the manipulation involved replacing the original f0 values with new f0 values in the rPraat package (Bořil and Skarnitzl 2016). For (i), the median f0 of this stretch of speech was computed (f0 *), and all values within the interval were replaced by this median value. For (ii), the values were determined according to the following formula: where x is asequence from 0 to 1 in 100 discrete equidistant steps and d is atarget f0 step in semitones (ST). The target values were based on the results of Hévrová (2021), where the effect typically included an interval of one octave (12 ST) for the rising portion in the Expat speakers, and an interval of half an octave (6 ST) in the Control speakers. However, since having two subsequent nuclear rising contours of exactly the same span would be rather unusual and might sound unnatural, we created two types of jump size: afull span (d = 12 ST for manipulating Controls to sound like Expats, and d = 6 ST for manipulating Expats to sound like Controls) and alimited span (d = 9 ST for the former and d = 4 ST for the latter manipulation type). The stimuli were balanced as to whether the full span was present in the first or second nuclear tone. In the manipulated version, the values from the original PitchTier were simply replaced by these new values. The manipulation itself was done in Praat using the PSOLA (Pitch-Synchronous Overlap-And-Add) resynthesis method. The audio file was converted to aManipulation object, the correct PitchTier loaded, and the resynthesized sound saved as anew audio file. This procedure was applied to both manipulated and unmanipulated target stimuli and to fillers as well. In other words, even stimuli that were not changed by the manipulation process were resynthesized anew, ensuring that all audio files maintain the same sound quality. Moreover, all resulting stimuli were examined auditorily by the authors as acheck for potential artefacts resulting from the manipulation. An example of an unmanipulated and manipulated version is presented in Figure 3. In total, 24 targets were created (2 × 6 speakers, two versions of asingle phrase from each) and 36 fillers (2 × 9 speakers, two different phrases from each). The perceptual test (see section 2.4) thus included 60 stimuli in total. The prevalence of fillers was meant to ensure that the melodic rises of the target stimuli were not too conspicuous.     =+−     *2 12 00 12 1 d ff x OPEN ACCESS T. BOřIL — P. ŠTURM — R. SKARNITZL — M. HÉVROVá — B. KöPKE 133 2.3 LISTENERS In total, 54 listeners participated in the perceptual testing. However, eight were disqualified because they had left the experiment before finishing it, four were non-native speakers of Czech, three more did not indicate what their mother tongue was, and two people were excluded because all their answers were identical (they did not detect any influence of French at all in any of the speakers). The results presented in section 3 will thus be based on the 37 listeners who provided valid data. The listener group under analysis consisted of 24 females, ten males and three people who did not state their sex. The mean age was 26.8 years (SD = 8.8, 75% range = 20–28, max = 67). Eight of the 37 listeners spoke French at the B2 level or better. Only two listeners spent more than amonth in aFrench-speaking country. Due to online data collection and the snowballing technique employed (see the following section), further background is not known about the participants. However, most of the listeners are expected to be students of English and American Studies at Charles University and their friends or family. 2.4 PERCEPTUAL TESTING Aperceptual experiment was constructed from the materials described in Section 2.2. The 60 stimuli were allocated to six blocks of ten items of afixed structure. The first and last item always contained afiller sentence, balanced as to whether an Expat or Control speaker started (terminated) the block. The remaining eight items— four targets and four fillers— appeared in arandomized order for each listener. The four targets included (i) an unmanipulated Expat speaker, (ii) an unmanipulated Control speaker, (iii) amanipulated Expat speaker, and (iv) amanipulated Control speaker. Figure 3.An example of f0 contours of an Expat speaker before and after manipulation, in the sentence Vpoledne jdu vyzvednout mého syna, obědvá doma ‘at noon Ipick up my son, he has lunch at home’; the underlined words correspond to the manipulated intervals. The unmanipulated version has two steep-rising nuclear tones that are reduced to 6 ST and 4 ST in the manipulated version. OPEN ACCESS 140 LINGUISTICA PRAGENSIA 1/2022 L1 attrition of speech? The case of Spaniards in the United Kingdom. Languages 5, 41. Mennen, I. (2004) Bi-directional interference in the intonation of Dutch speakers of Greek. Journal of Phonetics 32, 543–563. Morgan-Short, K., M.Faretta-Stutenberg, K.Brill-Schuetz, H.Carpenter and P.Wong (2014) Declarative and procedural memory as individual differences in second language acquisition. Bilingualism: Language and Cognition 17/1, 56–72. Moyer, A. (2009) Input as acritical means to an end: Quantity and quality of experience in L2 phonological attainment. In: Piske, T. and M.Young-Scholten (eds) Input Matters in SLA, 159–174. Bristol: Multilingual Matters. Moyer, A. (2013) Foreign Accent: The Phenomenon of Non-native Speech. Cambridge: Cambridge University Press. Myles, F. (2013) Theoretical approaches. In: Herschensohn, J. and M.Young-Scholten (eds) The Cambridge Handbook of Second Language Acquisition, 46–70. Cambridge: Cambridge University Press. Palková, Z. (1994) Fonetika afonologie češtiny. Praha: Karolinum. Pavlenko, A. (2000) L2 influence on L1 in late bilingualism. Issues in Applied Linguistics 11/2, 175–205. Pešková, A. (2020) Towards the developmental L2 intonation hypothesis: Evidence from L2 Italian and L2 Spanish. In: Proceedings of Going Romance, 34. R Core Team (2020) R: Alanguage and environment for statistical computing (version 4.0.3). Vienna: R Foundation for Statistical Computing. Retrieved from https://www.rproject.org. Sancier, M.L. and C.A.Fowler (1997) Gestural drift in abilingual speaker of Brazilian Portuguese and English. Journal of Phonetics 25/4, 421–436. Santiago, F. (2019) Théorie, recherche et didactique de la prosodie et de l’intonation en L2: nouvelles perspectives. Recherches en didactique des langues et des cultures: Les cahiers de l’Acedle 16/1, 1–13. Sharwood Smith, M. (1983) Cross-linguistic aspects of second language acquisition. Applied Linguistics 4/3, 192–199. Šreková, T. (2020). Cross-linguistic influence of L2 on L1 in speech of late bilinguals. Olomouc: Palacký University Faculty of Arts (MA thesis). Stoet, G. (2010) PsyToolkit— Asoftware package for programming psychological experiments using Linux. Behavior Research Methods 42/4, 1096–1104. Stoet, G. (2017) PsyToolkit: Anovel web-based method for running online questionnaires and reaction-time experiments. Teaching of Psychology 44/1, 24–31. Sučková, M. (2020) Phonetic and lexical features of first language attrition: Research into the speech of Anglophone expatriates in the Czech Republic. Brno: Masaryk University Faculty of Arts (PhD dissertation). Tobin, S., H.Nam and C.Fowler (2017) Phonetic drift in Spanish-English bilinguals: Experiment and aself-organizing model. Journal of Phonetics 65, 45–59. Ulbrich, C. and M.Ordin (2014) Can L2-English influence L1-German? The case of post-vocalic /r/. Journal of Phonetics 45, 26–42. Volín, J. (2008) Variabilita neukončujících melodií ve světle shlukové analýzy. Acta Universitatis Carolinae— Philologica 2007/2, 173–179. Volín, J., K.Poesová and L.Weingartová (2015) Speech melody properties in English, Czech and Czech English: Reference and interference. Research in Language 13/1, 107–123. Zajícová, L. (2010) Contact and deterioration in nominal morphology of the Czech immigrant in Paraguay. Revista Internacional de Linguistica Iberoamericana 8/1, 171–184. OPEN ACCESS T. BOřIL — P. ŠTURM — R. SKARNITZL — M. HÉVROVá — B. KöPKE 141 Tomáš Bořil Institute of Phonetics Faculty of Arts, Charles University nám. Jana Palacha 2, Praha 1, 116 38, Czech Republic ORCID: 0000-0002-5839-313X tomas[email protected].cz Pavel Šturm Institute of Phonetics Faculty of Arts, Charles University nám. Jana Palacha 2, Praha 1, 116 38, Czech Republic ORCID: 0000-0001-5521-029X [email protected] Radek Skarnitzl Institute of Phonetics Faculty of Arts, Charles University nám. Jana Palacha 2, Praha 1, 116 38, Czech Republic ORCID: 0000-0002-5682-1962 [email protected] Marie Hévrová Institute of Phonetics Faculty of Arts, Charles University nám. Jana Palacha 2, Praha 1, 116 38, Czech Republic ORCID: 0000-0002-8213-6360 [email protected] Barbara Köpke Laboratoire de NeuroPsychoLinguistique Université Toulouse— Jean Jaurès 5, allée Antonio Machado, 31058 Toulouse Cedex 9, France ORCID: 0000-0003-1634-3457 barbara.kopk[email protected] OPEN ACCESS