@phdthesis{Himmel2021,
  author    = {Himmel, Marie-Christin},
  title     = {Phonetic and phonological variability in the L1 and L2 of late bilinguals: The case of /r/ and /l/},
  doi       = {10.25972/OPUS-23933},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-239335},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2021},
  abstract  = {A large body of research has shown that a late bilingual's L1 and L2 phonetic categories influence each other, yielding deviations from monolingual norms in the phonetics of both languages. Existing models of L2 sound acquisition (e.g., the Speech Learning Model; Flege, 1995, 2007) predict unified phonetic spaces which accommodate both L1 and L2 sound categories. Such connections between an L1 and an L2 are believed to lead to persistent non-nativelikeness in the L2, but also to divergence from the monolingual norm in the L1, as shown in numerous studies (e.g., Bergmann et al., 2016). In this dissertation, I focus on the differences in the sound patterns of a bilingual's languages which do not only emerge in the precise phonetic realizations of L1 sounds but also in language-specific distributional patterns that determine the realization of these sound categories in different phonetic contexts. Previous work in L1 attrition is limited to a small set of phonetic properties (especially VOT, e.g., Flege, 1987), variables beyond L2 transfer which are known to give rise to variable realizations have been neglected. Thus, little is known as to whether bilinguals' realizations of an L1 sound category in different phonetic contexts (e.g., position within a syllable) are subject to change in L1 attrition, and whether such changes arise due to long-term exposure to different distributional patterns of an equivalent L2 category. In this dissertation I address these gaps by exploring L1 attrition in the distributional and phonetic characteristics of liquids to shed light on the contribution of the L2 and the role of general phonetic and phonological variables to the processes that drive change in an L1. I investigate changes to phonetic properties and distributional patterns of rhoticity and /l/-allophony in the L1 of American-German late bilinguals, a language constellation which offers an instructive test case to investigate the causes of L1 attrition as well as the source from which changes due to L1 attrition emerge. Furthermore, changes to liquids can also shed light on the processes which drive sound change, gradience and variability due to various positional and phonetic factors (e.g., preceding vowel, syllable structure) in liquids across many native varieties of English. In particular, I explore the variable realization and distributional patterns of two sounds known to be subject to a considerable degree of gradience and variability, namely English /r/ and /l/, in American English-German late bilinguals. To that end, I present the results of a production study of 12 L2-dominant American English-German late bilinguals as well as a monolingual control group for each language. The speakers performed a variety of production tasks which were aimed to elicit the realization of (non)-rhoticity and /l/-(non-)allophony in both languages of the late bilinguals, English and German which were analyzed auditorily (/r/ only) and acoustically (/r/ and /l/). Although L1 attrition of rhotics and laterals has been investigated previously (e.g., de Leeuw, 2008; Ulbrich \& Ordin, 2014), the effect of contextual variables on L1 attrition and whether such variables also shape L1 attrition remains unexplored. The results of the auditory analyses of postvocalic /r/ revealed that the late bilinguals showed non-convergence with monolingual (non-)rhoticity in both of their languages by vocalizing postvocalic /r/ more frequently in their L1 (English) and failing to entirely suppress rhoticity in their L2 (German) leading up to a higher degree of rhoticity in their L2. While the loss of rhoticity in the bilingual's English was distributed along a spectrum of contextual constraints (e.g., type of pre-rhotic vowel and morpho-phonological environment) known to affect rhoticity in other English varieties, the non-targetlike productions of non-rhoticity (i.e., non-vocalized postvocalic /r/) in their L2, German, were not sensitive to the same contextual constraints. The acoustic analyses of the bilinguals' rhotic productions in English and German differed from the monolinguals in the acoustic correlates of rhoticity in pre-rhotic vowels where they showed reduced anticipatory F3-lowering (i.e., less /r/-colored vowels). I take my results to indicate that the bilinguals operate in two separate phonological grammars which approximate the respective L1 norm but show an increase of variability along constraints already present in each grammar. In contrast, the bilinguals' phonetic system seem shared between the two grammars. This leads to persistent L1-L2-interactions as the two grammars operate within the same phonetic space. Thus, the changes in L1 attrition are induced but not governed by the L2: Change to the L1 reflects constraints underlying the L1 as well as more general laws of phonetics and universal trajectories of language change. The lateral results revealed that just like in postvocalic /r/, the bilinguals showed non-convergence with the monolingual norm regarding the velarization of coda /l/ in both their languages. The changes to English laterals were sensitive to their positional context and more substantial for word-initial laterals than word-final laterals. Similarly, their German laterals were non-convergent with the monolinguals in two ways. Firstly, the bilinguals differed with regard to the acoustic specifications of their laterals, and secondly, the bilinguals failed to suppress the lateral allophony from their L1, leading to a non-targetlike allophonic pattern in their L2 laterals. I interpret the lateral results to lack evidence that the L1 allophonic rule was affected by the presence of an L2; nevertheless, L1 change emerged in the phonetic specifications of laterals. Furthermore, the bilinguals did not establish a nativelike allophonic pattern in their L2, leading to non-convergence in the allophonic distribution as well as the phonetic realization of German laterals. In this way, this dissertation provides evidence for L1 attrition in the distributional and the phonetic properties of liquids in the L1 of late bilinguals. In particular, the study presented in this dissertation provides evidence that L1 attrition is induced by the presence of a similar sound pattern in the L2. The pathway of attrition follows constraints not only underlyingly present in the L1 but also part of the universal laws of phonetics known to shape sound change. To explain these results, I draw from existing constraint grammars in phonological theory (such as Optimality Theory and Harmonic Grammar) to develop my Dynamic Constraints approach which allows the effects of external variables (e.g., L2 acquisition and its effect on the mind), and internal variables such as an increased likelihood of variability due to articulatory differences can be modeled using scaling factors which can interact with each other, the noise within the grammars, and the constraint weight itself. In this way, the model links previous findings on L1 attrition and its connections to diachronic and synchronic variability, offering insights into the links between the individual languages in a bilingual's mind.},
  subject      = {Zweisprachigkeit},
  language  = {en}
}