Effects Of Cross-linguistic Phonetic Similarity On Chinese Learners’ Perception And Production Of Russian Speech Sounds

Cross-linguistic similarity is believed to be one of the decisive factors in the acquisition of second languages(L2).The phonetic and phonological similarity between learners’ native language(L1)and L2 could dictate how L2 learners perceive and produce the target sounds as proposed by the Speech Learning Model(SLM)and the Perceptual Assimilation Model-L2(PAM-L2).That being said,the assessment of cross-linguistic similarity is not without contentions;various approaches had emerged in the extant literature concerning the deduction of L1-L2 perceptual similarity,their speculations could be at times inconsistent with each other.This divergence could severely bias the verification and the interpretation of similarity related theories like SLM and PAM-L2.To date,research concerning the comparison between different similarity assessing methods is still limited,the investigation of which is in absolute need.In the meantime,empirical studies on the learning of Russian speech sounds by Chinese learners are by far scarce,but they are of great importance,since Russian segments like stops are contrasted in a different way from either Mandarin or English,yet their learning difficulties for Chinese learners are still unknown.Besides,the trill sound /r/ is one of the iconic segments in Russian,the learning of which by Chinese learners would certainly contribute much empirical significance.Theoretically,even though PAM-L2 and SLM had been much discussed in the extant literature,the assimilation patterns of the former still need to be stipulated in a finer manner,and the assessment of L1-L2 similarity of the latter is not without contentions,alongside with the unclear definition of learning difficulties.Therefore,both theoretical models still need more clarification and verification.To this end,the current dissertation integrated five commonly adopted methods,namely IPA Comparison(IC),Acoustic Difference(AD),Feature Redeployment(FR),Articulatory Similarity(AS)and Phonological Similarity(PS),to make reasonable predictions on the perceived similarity between Mandarin and Russian in wordinitial stops and liquids by Chinese learners of Russian.PAM-L2 and SLM were used to make hypotheses on the acquisition patterns of the corresponding Russian sounds.Among the five methods,IC and AD are related to SLM;AS patterns with PAM-L2,whereas FR and PS are stemmed from phonology.The five methods in large reflected the primary approaches adopted in previous studies.Taken together,the present dissertation attempts to discuss the validity of different similarity assessing methods via the investigation of the learning of Russian initial stop and liquid contrasts by Chinese learners.The efficacy of PAM-L2 and SLM will be systematically scrutinized based on the unique L1/L2 pairing of Russian and Mandarin,which could furnish novel and valuable findings from both empirical and theoretical standpoints.The research purposes were threefold: 1)to investigate the acquisition of Russian initial stops and liquids by Chinese learners;2)to delineate a sequence of predictive power among the five approaches;3)to test the validity of SLM and PAM-L2,and,if necessary,advance a composite model complementary to the existing L2 phonetic acquisition models.To reach these purposes,three research questions were addressed: 1)Which of the five methods could more accurately capture the perceived cross-linguistic phonetic similarity between Russian and Mandarin in stops and liquids by Chinese learners? 2)How can PAM-L2 be applied to account for the perception of Russian initial stops and liquids by Chinese learners? 3)How can SLM be applied to account for the perception and production of Russian initial stops and liquids by Chinese learners?To answer the research questions,four experiments were incorporated in the dissertation,including two perception experiments and two production experiments.The two perception experiments incorporated two perceptual assimilation tasks(PATs),two identification tests(IDEs)and two discrimination tests(DISs);the two production tests were conducted as sentence reading,and were analyzed via acoustic and global evaluation means.For the first research question,the PATs revealed that both Russian voiced and voiceless stops were deemed as highly similar with Mandarin voiceless unaspirated stops by Chinese learners,confirming the prediction from AS,while overriding IC,AD and FR.Regarding liquids,PAT showed that Russian /r/ was perceived as a “new” sound relative to Mandarin,whereas Russian /l/ was assimilated to Mandarin /l/,also supporting the prediction from AS,and partly in conformity with IC and AD;FR,on the other hand,was negated.Regarding phonological influence,the PATs did not reveal much distinction between Mandarin and non-Mandarin vowel contexts in either stops or liquids,disconfirming the prediction from PS.As such,the predictive power of the five methods could follow a sequence of AS > AD ≥IC > FR ≥ PS(the validity of the five methods is specifically elaborated in Section 7.4).This finding patterns with PAM-L2 which holds an ecological direct-realist view of perceptual assimilation under the framework of Articulatory Phonology.In the meantime,the IPA and acoustic accounts for perceptual similarity posited by SLM were challenged,since they may not always make accurate predictions.The distinction between predicted similarity and actual perceived similarity needs to be treated with discretion.For the second research question regarding PAM-L2,the results from the DISs showed that Russian initial stop contrasts,being assimilated to single categories of Mandarin voiceless unaspirated stops(SC),were poorly discriminated as predicted(around chance level of 50%);Russian initial liquid contrast,on the other hand,was only moderately well discriminated(around 75% accuracy).This finding is not completely in support of PAM-L2,since Russian initial /r l/ were perceived as uncategorized and categorized(UC)relative to Mandarin,the discriminability is predicted to be very good as argued by PAM-L2.The reason might be that there was considerable overlap in the UC assimilation of Russian /r l/(around 35% of Russian /r/ were also assimilated to Mandarin /l/),constituting a UC-o scenario which could render lower discriminability than UC-no.The subcategorizations of UC-o and UC-no were used to describe the assimilation of suprasegmental features in previous studies,the evidence provided by the current study suggested that they could be also effective in the segmental level.Alternatively,the supposedly good discrimination was dampened by the high initial similarity of Russian /r l/ being a within-organ contrast.As regards the third research question concerning SLM,the IDEs showed that Russian initial /b d g/ and /l/,being Mandarin “similar”,were well identified,whilst the identification of the Mandarin “identical” /p t k/ and Mandarin “dissimilar” /r/ was relatively low in accuracy.These findings were,to some extent,at odds with SLM in light of the predicted learning difficulty of “identical”,“similar” and “new” L2 sounds.However,the production results were in keeping with SLM in that the Russian initial /b d g/ were indeed poorly produced,in contrast with the well-produced /p t k/.This result implied that Chinese learners did not acquire Russian initial /b d g/,their relatively high identification accuracy could be triggered by the choice bias that when they were uncertain about the choice,they would be inclined to choose “voiced” over “voiceless”,resulting in “false” accuracy.The production of Russian initial /l/ did not manifest conspicuous improvement over the incremental learning experience,while /r/ was produced more precisely by learners with longer Russian learning experience.Notably,despite the progressive learning trajectory,Russian /r/ in this study was still produced less accurately than /l/ as revealed by intelligibility,suggesting that “similar” L2 sounds may benefit from some “head start” associated with L1-L2 proximity.Collectively,these findings imply that the functionality of SLM might be constrained in the production domain,and the learning difficulty might be better outlined as learning trajectories,going beyond the coarse-grained terms of “easy” or “difficult”.All the findings from this dissertation,both supportive and critical of PAM-L2 and SLM,when taken together catalyzed a composite model proposed as the Speech Similarity Hypothesis(SSH)which aims to disentangle the ambiguity of similarity assessment in SLM,and to extend the tenets of PAM-L2 to the realm of production.In comparison with SLM,SSH differentiates the perceived(subjective)similarity from the predicted(objective)similarity,since the two concepts can at times run counter to each other as revealed by the current dissertation.In SSH,the objective similarity of articulatory gestures is believed to be more valid in deducing perceptual similarity.That said,SSH holds that perceived similarity is the key factor shaping the subsequent L2 speech learning,and which shall be directly measured via quantitative approaches like PAT,rather than relying on the objective similarity assessed by methods like IC and AD as proposed by SLM.Moreover,SSH addresses only the production outcome influenced by perceived L1-L2 similarity,bypassing the possible ambivalent findings between perception and production that are unaccountable for SLM.At last,SSH turns to learning trajectories to define acquisition difficulty,making the verification of hypotheses more operable.Collectively,the research significance of the current dissertation is as follows: 1)It comprehensively probed both the perception and production of Russian initial stop and liquids by Chinese learners,filling the research gap in the concerned field.2)It systematically verified the effectiveness of the five similarity assessing methods,and found that AS was more advantageous in terms of deducing perceptual similarity between consonants from different languages,providing valuable reference to the concerned future studies.3)As a new attempt,reaction time(RT)was added in the analyses of the results from PATs,the results showed that it might be an informative parameter regarding the capturing of degree of L1-L2 perceptual similarity.4)It was found that L2 learning experience would have little impact on the perceptual assimilation between L1-L2 sounds.5)The subcategorizations of UC-o and UC-no from suprasegmental assimilations were also found to be necessary in the segmental level,which modified the application of PAM-L2.6)The contrast types of AOs proposed by PAM-AOH could also have an impact on adults’ perception of L2 phonemic contrasts.7)This dissertation manifested that the IC and AD accounts for L1-L2 similarity proposed by SLM could be unreliable in deducing perceptual similarity.Besides,the predictions from SLM could be valid only in production,but not in perception.Also,the learning difficulties noted in SLM might need to be defined in a finer-grained fashion,so that the verification of its predictions could be more applicable.8)Combining the findings from the current dissertation,a new model noted as “Speech Similarity Hypothesis(SSH)” was put forward,which utilizes PATs(as in PAM-L2)to directly measure L1-L2 perceptual similarity,and makes predictions on L2 speech production(as in SLM)via learning trajectories.As such,SSH clarifies the ambiguities in the assessment of L1-L2 similarity and certifies the evaluation of the production outcome,being unclear in SLM.That to say,SSH grounds upon the great predictive power of SLM in L2 speech production,and serves as a complement to the perception-oriented model of PAM-L2.SSH needs further corroboration;its applicability could provide new suggestions for future research.