| ‘International Journal of Bilingualism’ • 14 •  Number 2 • 2010, 1–25| Celata Volume & Cancila: Phonological attrition

1

Phonological attrition and the perception of geminate consonants in the Lucchese community of San Francisco (CA) Chiara Celata Scuola Normale Superiore, Pisa, Italy

Jessica Cancila Università per Stranieri, Perugia, Italy

Abstract

Key words

Two perceptual tests, designed to investigate the discrimination of the geminate singleton/geminate consonant contrast, were submitted to bilingual consonants English-Lucchese immigrants in the USA, and monolingual Lucchese subjects. The study has two major goals. The first goal is to add a muchphonological needed empirical contribution to the incredibly sparse literature on attrition phonological/phonetic attrition in bilinguals. The second goal is to cast the analysis of phonological attrition for socio-linguistic reasons on a more speech perception general psycholinguistic framework for cross-language speech perception. As it will be shown, the comparison between the performances of different groups of speakers in different perceptual tasks allowed us to verify that phonological attrition moves along patterns of perceptual sensitivity which can be fruitfully compared with those surfacing in L1/L2 phonological acquisition.

1 Introduction

*

Socio-linguistic attrition in immigration communities occurs as a consequence of language shift, that is, when immigrants abandon their primary language (L1) and Address for correspondence

Chiara Celata, Laboratorio di Linguistica, Scuola Normale Superiore, P.zza dei Cavalieri, 7, 56100 PISA (Italy). [email: [email protected]] Acknowledgments

This research was supported by Università per Stranieri di Perugia funding ‘L’Italiano degli stranieri e l’italiano all’estero: Osservatorio Linguistico’ and by Italian Government funding PRIN05/07 to both authors (UR Scuola Normale Superiore, Pisa and UR Università per Stranieri, Perugia). We would like to thank S. Giannini for supporting this research project from its very beginning, S. Scaglione for providing useful information preliminarily to the realization of the experimental project in the USA, R.C. Major, B. Bullock and three anonymous reviewers for helpful comments to an earlier version of this article. The International Journal of Bilingualism Copyright © 2010 the Author/s 2010, ISSN; Vol 14 (2): 1–25; ID no 363058; DOI; 10.1177/1367006910363058 http://Ijb.sagepub.com

2

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

conform to the variety of the majority (Andersen, 1982; Dorian, 1981; Li Wei, 1994; Schmidt, 1985b). The linguistic decay caused by an asymmetric contact normally implies both a functional reduction (what is properly called ‘shifting’) and a structural reduction (‘attrition’) in L1 use (Bullock & Gerfen, 2004; Schmid, Kopke, Keijzer, & Weilemar, 2005). Phonology, like other components of the grammar, can be strongly involved in the process of change under multilingual pressure (Bullock, Toribio, González & Dalola, 2006). In particular, perceptual abilities for sound discrimination may give insights into the actual status of a phonological feature in the mind of the speaker, and provide a window on structural phonological change (Major, 2005). This research is a sample-study. It investigates the perceptual processing of the consonant length distinction by Italian immigrants in an English-speaking community. The goal of the study is two-fold. First, it aims at offering an empirical contribution to the field of L1 phonological/perceptual attrition. As noticed by Bullock & Gerfen (2004): [l]inguistic analyses of first language attrition ... are focused largely either on the sociolinguistic context that propels a language toward shift, attrition, and death or on the postulation of theoretical models of the internal and external factors that lead toward grammatical change, primarily with regard to morphological and syntactic structures [while] [f]ew studies are devoted to the empirical issues raised in documenting first language attrition. (2004: 304) This article aims to contribute to the problem of the dearth of evidence in studies of phonological attrition with both ‘structural detail and quantitative data’ (Bullock & Gerfen, 2004: 318). A corpus of perceptual data is thus analyzed, with reference to the processing of the length contrast in consonants by 15 immigrants speaking Lucchese dialect in San Francisco, California. Lucchese, the dialect spoken in Lucca, which pertains to the group of northern Tuscan varieties, is notoriously affected by a variable process of consonant degemination (see the following section for more details). Two different perceptual experiments were elaborated, on words and on non-words respectively. Three groups of subjects were tested: a first-generation immigrant group, a second-generation immigrant group, and a Lucchese-dialect-speaking group (still living in Italy). Native Italian speakers, from a variety that does not exhibit any degemination process, were included as control group. As we will illustrate later, the multilingual context of the Lucchese-speaking community in San Francisco shows a high level of complexity because of proximity (in structural, typological and socio-cultural terms) between the Lucchese dialect and Standard Italian, the former being the mother-tongue of our subjects, used in particular for informal communication in everyday life, the latter being the language of scholarization and supra-regional communication (as well as of mass media). The second goal of this article is to connect the analysis of phonological attrition with a more general psycholinguistic framework for cross-language speech perception. In fact, we believe that the study of phonological attrition can take advantage of a direct comparison with other fields of empirical phonological research. As a matter of fact, where a large body of experimental evidence exists, the mechanisms responsible for change are more easily demonstrable. This strategy has proved fruitful in this study. As it *

The two authors jointly developed this article. For academic purposes, however, CC takes responsibility for sections 3, 4.2–4.3, 6 and 7, JC for sections 2 and 4.1. Sections 1 and 5 were jointly written. The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

3

will be shown in this article, the comparison between the perceptual abilities of different groups of speakers in different perceptual tasks has allowed us to verify that the attritional process does not advance randomly, but follows instead some clearly detectable patterns of perceptual sensitivity. Moreover, these patterns have to be related to the emergence of different perceptual abilities, also documented for L1/L2 phonological acquisition. This article is organized as follows. Section 2 sketches the phonological differences between the three contact languages under investigation (Lucchese, Standard Italian and American English), with respect to the consonant length feature. Some socio-linguistic background is also given with respect to the Lucchese degemination process. Section 3 presents the empirical hypotheses that we put forward with reference to the perceptual processing of our subjects. The two following sections illustrate the experimental design: participants, materials, methods and results of Experiment 1 (Section 4) and Experiment 2 (Section 5), respectively. Section 6 is devoted to a general discussion of the empirical hypotheses outlined in section 3, with some proposed generalizations based on models of cross-language speech perception in adults (Best, McRoberts & Sithole, 1988; Flege, 1993; Kuhl, 1991; Major, 2001), theories of L1 perceptual development (Repp, 1981; Stager & Werker, 1997; Werker, 1994; Werker & Tees, 1984b, 1999), and results previously gathered in non-native contrast discrimination studies (Celata, 2004; Mora, 2005). Section 7 contains some provisional conclusions.

2 The source of attrition: Lucchese dialect, Standard Italian and American English in contact

The Italian first generation immigrants analyzed in this study spoke central Lucchese as their mother tongue. The central Lucchese dialect is a northern Tuscan variety spoken in the city of Lucca, the south-eastern lowland and the river Serchio valley. As in most central (and southern) varieties of Italian, the singleton vs. geminate contrast is present in the Lucchese dialect for all consonants, with the exclusion of inherently long consonants such as //, //, //, / /, / / (see for example Celata & Kaeppeli, 2003; Payne, 2005). However, in this dialect a degemination process is known primarily to affect the rhotic (e.g. // → [] ‘ground’), and sporadically the other consonants (e.g. // → [] ‘brick’; Ambrosini, 1974; Giannelli, 2000; Pieri, 1890; Rohlfs, 1966). This process in Lucchese is said to be variable in nature, that is, depending on speech style (informal speech more degeminating than formal speech), socio-linguistic background (‘rural’ speech more degeminating than urban speech; speakers with lower educational level more prone to degeminate than speakers with higher educational level) and lexical items (some items always surface as degeminated, others do not). From a perceptual point of view, speakers of Lucchese are well aware of their typical degeminating pronunciation, and they often stigmatize this practice in speaking to non-Lucchese individuals, using the degemination process as a marker of ethnic identity (Graff, Labov & Harris, 1986).1 Like many other phonological behaviors connected to the presence of a regional dialectal variety, degemination is in regression in the production of the youngest speakers of Lucchese, since the interference from SI is 1

When informed about the purpose of the study, many of our informants made comments like ‘Ah sì, che noi si dice guera, e come no!’ (‘Oh yes, (the fact) that we say [] (for guerra ‘war’), of course!’). The International Journal of Bilingualism

4

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

progressively more massive; however, when the genuine dialect is used (e.g. in informal, familiar speech, with corregional speakers), degemination does appear, even in the speech of the youngest people.2 In this study, the perceptual status of the consonant length distinction is investigated for a group of Lucchese dialect speaking immigrants in San Francisco. We refer to the detailed socio-linguistic research by Scaglione (2000) for any information concerning the general linguistic and extra-linguistic conditions of our subjects. For general remarks on bilingual mixture in the Italian emigrant communities in the USA, see also Schmid (2005). The subjects of this study were born either in Lucchesia (and subsequently emigrated to California: first generation immigrants, henceforth FG), or in San Francisco from Lucchese parents (second generation immigrants, SG). Those individuals, exposed to three linguistic systems, are actively (as speakers) and/or passively (as hearers) exposed to American English, Lucchese, and Standard Italian. American English (AE) is the language of the host community, where these people live and have successfully been integrated. Lucchese represents the actual mother tongue for FG immigrants, but not for SG ones, who were instead born in California and therefore considered AE as their mother tongue, although they grew up with Lucchese speaking parents. Standard Italian (SI) is the supra-regional variety all immigrants are also familiar with, but they have no productive competence in that language. In fact, they are exposed to SI sporadically and as passive users only, mostly from TV. These three languages behave differently with respect to the length distinction in consonants. Only SI has a phonological distinction between the singleton and the geminate consonants (Albano Leoni & Maturi, 2002; Bertinetto & Loporcaro, 2005; Muljacic, 1969), while AE – like British English – does not possess such distinction (Ladefoged, 1999; Roach, 2000; Wolfram & Schilling-Estes, 1998). We have already illustrated the peculiarity of the Lucchese dialect degemination process.

3 Hypotheses and predictions We have opted for an inductive design study in which the perceptual abilities of two groups of Italian immigrants are tested for the discrimination of singleton and geminate consonant duration, on both words and non-words. In particular, the perceptual tasks were designed in order to test the following empirical hypotheses. Hypothesis 1

As we have already mentioned (see section 2), the rhotic is considered by many authors as the major target of degemination in the Lucchese dialect. As a consequence, it is possible that the discrimination between the singleton and the geminate consonant is 2

As far as Raddoppiamento Sintattico (RS) is concerned (e.g. the post-lexical gemination process typical of most central and southern Italian varieties, as well as of SI), the distributional properties in Lucchese dialect are rather similar to those of other Tuscan varieties (and of SI as well). The only exceptions concern some verbal forms (3rd singular of passato remoto, 1st and 3rd singular of future, monosyllabic forms like ho, so, sto, do ‘I have, I know, I stay, I give’), the pronominal form chi ‘who’ and the preposition da ‘from’, which do not trigger RS in Lucchese dialect, while other forms, in particular the definite article i (plural masculine) and some derived prepositional forms (like dei, nei etc.) do trigger RS in Lucchese (see Giannelli, 2000: 77; Marotta, 1995; Saiu, 2004 for details). The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

5

impaired in the case of /()/ more consistently and more often than in the case of other consonants for those subjects who are supposed to rely on the Lucchese phonological grammar, that is, the Lucchese speakers, primarily, and possibly the FG immigrants, to the extent they do not undergo the effects of the exposure to AE phonology (but see later for an extended hypothesis on this point). In order to test this hypothesis, we have carried out the perceptual experiments using /r()/ as well as /s()/ and /t()/ as target consonants. In this way, the discrimination of rhotics can be directly compared with the discrimination of other coronal consonants (fricative and stop). Hypothesis 2

The three groups of subjects are expected to show a progressive decline in their sensitivity to the singleton-geminate contrast, with the Lucchese subjects still resident in Italy performing overall the best on the discrimination tasks, the FG immigrants showing evidence of some impaired discrimination (likely due to first-language attrition), and SG immigrants performing the worst because of the effective lack of a positive experience in either the Lucchese dialect or SI. In particular, if discrimination turns out to be more difficult for the FG immigrants than for the Lucchese subjects, a first-language attritional process is likely to be at stake (Hypothesis 2a). Moreover, although SG immigrants grew up with Lucchese-speaking parents and have some sporadic contact with both the Lucchese variety and SI (see section 2), they are nevertheless supposed to be dominant in the AE phonological system; as a consequence, we argue that they will not be able to differentiate two items only on the basis of consonant length. For this reason, we expect that their discrimination of the singleton vs. geminate contrast will be nearly uniformly poor (irrespective of the task conditions; Hypothesis 2b). By submitting two discrimination tasks to the subjects, on words and non-words respectively, we will be able to single out the effects of the functional reorganization, as much as it surfaces in adults’ speech perception processes, which is assumed to determine the changes in discriminatory abilities on both the developmental and the experiential dimension (see in particular Werker & Tees, 1999; see also Miller & Eimas, 1994; Repp, 1981; Werker & Tees, 1984b). The hypothesis of functional reorganization was originally based on evidence from the development of the native language phonology in infants, but an occasional foray into second language phonology can be documented as well. In this respect, one can speculate that two different specializations of the human perceptual system can emerge in discriminating two non-native sounds: a linguistic strategy, which ‘probably involves a search for meaning’ (Werker & Tees, 1984b: 1868; see also Miller & Eimas, 1994) and is activated in phonemic and contextual information processing, and a non-phonemic strategy, aimed at discriminating subtle acoustic differences among sounds. This assumption emerged in particular from the observation that, while infants appear to be able to discriminate easily between pairs of sounds not belonging to native phonetic categories, after the first year of age a sharp decline is documented in the discrimination ability, probably because subjects tune to just that phonological information which is relevant to master the native language communicative demand (see e.g. Werker, 1994; Werker & Polka, 1993). If tested under appropriate conditions, however, or if extensive language instruction is given, some sensitivity to phonetic contrasts along non-native acoustic dimensions can surface in the performance of adults as well. The International Journal of Bilingualism

6

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

In a previous study on L2 phonological acquisition (Celata, 2004), we found that during the initial stages of language acquisition, when phonemic (lexical) discrimination is difficult, probably due to the limited exposure to the target language, a good discrimination of a non-native contrast is indeed possible on non-words (acoustic discrimination). On the other hand, experienced L2 learners appeared to be definitely less accurate than the inexperienced ones in the non-words discrimination: they performed better in a task involving the discrimination of real words, than when sensitivity to acoustic cues was requested. An interesting parallel has been documented by Mora (2005) for some Spanish/Catalan bilingual advanced learners of English, who where more successful in perceiving non-native phonemic contrasts in known words than in non-words. We would like to interpret the behavior shown by L2 learners in terms of the functional reorganization assumption sketched earlier. During L2 phonological acquisition two different perceptual abilities appear to emerge at distinct acquisitional stages. The acoustic strategy is preferred by those inexperienced learners who cannot take advantage from the input experience, which necessarily is very poor; on the contrary, as learning advances, a linguistic/contextual strategy is preferred, since at this stage lexical knowledge may play an important role for speech discrimination. Hypothesis 3

We hypothesize therefore that both the FG immigrants and the Lucchese dialect speaking subjects will behave like ‘low-experienced learners’ with respect to a SI consonant feature which is absent from their L1 (gemination). As a consequence, these subjects are expected to perform slightly better on non-words than on words. On the other hand, as mentioned in Hypothesis 2b, the perception of SG immigrants will arguably be influenced by the AE phonological system, therefore their performance is expected to be strongly impaired in both experiments. Two perceptual experiments, on words and on non-words respectively, were designed in order to test subjects’ discrimination of the singleton vs. geminate contrast. They are described in detail in the following sections.

4 Experiment 1 4.1 Subjects

Fifteen Lucchese-dialect-speaking immigrants living in San Francisco were tested: eight of them were born in Lucchesia and went to the USA as adults (FG immigrants), while seven were born in the USA from Lucchese parents (SG immigrants). The former group was largely homogeneous as far as age was concerned. They were all born in small villages surrounding Borgo a Mozzano (Corsagna, Diecimo) and Capannori (Pieve San Paolo, Santa Margherita, Carraia, San Gennaro). They arrived in San Francisco when they were 18–27 years old. Their educational level was rather varied: conforming to the Italian school system, two of them (FG1, FG6) reached the primary/elementary level, two (FG2, FG8) the middle level, and four (FG3, FG4, FG5 and FG7) the secondary/high The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

7

school level.3 As for their current professional position, the men were factory workers or proprietors of bars/restaurants, while women were housewives or shop assistants. As for literacy in Italian, they all confirmed that they were not used to reading books or newspapers in Italian, but they often watched Italian TV channels. The group of SG immigrants included younger people, who were born in the USA to Lucchese parents. Their educational level was higher with respect to that of FG immigrants, since they had all reached either high school graduation or college. When asked about the proficiency in their parents’ language, they mostly reported that they were nearly able to understand their parents speaking Lucchese to each other, but they could not speak that variety themselves. In addition, 16 Lucchese dialect native speakers still living in the homeland were tested. Twelve of them were over 40 years of age while the remaining 4 were between 29 and 36. All of them were born and raised in Lucca or the south-eastern lowland of Lucca, where they currently live. Their educational level was middle–high (8–16 years of education) with one exception, who only reached the primary level. They spoke the native Lucchese dialect as mother tongue; most of them had a rather high competence in SI as well, even if their speech accent was strongly influenced by the local variety (as usually happens in diglottic territories, such as Tuscany). Socio-linguistic details about FG and SG immigrants and Lucchese speakers are summarized in the Appendix. Six native Italian speakers (central variety) performed the perceptual tests as control group (NIS). 4.2 Materials

In this first experiment (E1) we used Italian real words. We constructed 30 pairs of short sentences/phrases, where pronunciation was identical for the two members of each pair with the exception of the length of the target consonant. The target consonants were // – //, // – // and // – //. The sentences were produced by a native Italian speaker, who spoke a regional variety characterized by the presence of the consonant length distinction and who was explicitly instructed to approximate the Standard Italian pronunciation as much as possible. The following prosodic and lexical variables were controlled across the members of each pair: speech rate, tonal contour, average intensity, relative frequency of items, and plausibility of lexical junctures within the phrase. As for items’ frequency and lexical junctures, we took care to use words with high or very high frequency of occurrence in the Italian lexicon, and placed them in short phrases/ sentences of frequent use and immediate comprehension. As for speech rate, tonal contour and average intensity, the speaker was instructed to produce carefully the list of stimuli by reducing as much as possible rate, tone and intensity variation; several repetitions of the list were made, in order to obtain a sufficiently homogeneous production. Afterwards, the recorded materials were acoustically inspected. In particular, we calculated the average number of syllables per second (in order to control for speech 3

Conforming to the Italian school system, the primary level corresponds to 5 years of education (6–11 years of age); the middle level to 8 years of education (6–14), the secondary level to 12/13 years of education (6–18/19). The International Journal of Bilingualism

8

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

Table 1 Materials of E1. Mean duration (in msec) of singleton consonants (C), geminate consonants (CC), stressed vowels preceding a singleton consonant (Vc) and stressed vowels preceding a geminate consonant (Vcc). In the fourth column, the geminate to singleton ratio is given.

/r/ /s/ /t/

C

CC

35.76 103.7 112.23

80.85 161.8 184.92

CC/C ratio 2.26 1.56 1.65

Vc

Vcc

164.12 125.17 139.41

127.56 102 104.58

rate) and the average intensity value, across the whole set of materials; the global F0 curve was controlled separately in each pair of sentences. Only the sentence pairs that did not exhibit any inconsistent pattern as for tone, intensity and speech rate were retained and used as experimental materials. The whole list of sentences is reported in the Appendix. Three examples, one for each consonant, are reproduced here: Caro lontano [] vs. Carro lontano [], ‘distant relative’ vs. ‘distant carriage’; Casa aperta [ ] vs. Cassa aperta [ ], ‘open house’ vs. ‘open box’; In dote al matrimonio [   ] vs. Indotte al matrimonio [  ], ‘as dowry for the marriage’ vs. ‘induced to marriage’. Since duration is known to vary considerably across consonant classes, prosodic contexts, speech styles, and speakers (cf. Celata & Kaeppeli, 2003; Loporcaro, 1996; Payne, 2005; Pickett, Blumstein, & Burton, 1999), we decided to measure the relevant consonant length in the stimuli. Moreover, the length of the preceding vowel (if stressed) could vary as a consequence of a natural compensative tendency by which a stressed vowel is longer when followed by a singleton consonant, shorter when followed by a geminate consonant (syllabic isochronism; cf. Bertinetto & Vivalda, 1978; Marotta, 1985; Muljacˇic´, 1972). Therefore, we measured the relevant consonants and the preceding vowels, when stressed. In Table 1 the average duration of each segment is shown (in msec), together with the geminate to singleton ratio. It appears straightforward that the [r] – [r] contrast is produced with a significantly larger CC/C ratio than the [s] – [s] and the [t] – [t] contrasts. Moreover, as expected, the presence of a geminate consonant is also correlated to a shortening of the preceding stressed vowel, and this durational difference resulted in being statistically significant for the three consonants. The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

9

4.3 Method

The subjects were asked to listen to a sequence of sentences, randomly presented, one for each pair; half of the repetitions contained the singleton, half the geminate consonant. For each sentence, they had to choose whether the relevant consonant was single or geminate. Two written options, corresponding to the members of the pair (e.g. T vs. TT, conforming to the Italian spelling conventions), were given on a sheet of paper; subjects had to mark the sentence they believed to have been uttered by the speaker. The analysis of results included the calculation of error percentages and an analysis of variance (one-way ANOVA). 4.4 Results of E1

Error percentages on words (E1) are given in Table 2. The results are listed for each subject and each consonant of the test. The performance of NIS is not reported in the table, since, as expected, their discrimination was optimal (mean error 0%). As we can see in the table, the mean values reveal that the Lucchese dialect speaking subjects, as a group, were more able to discriminate the singleton vs. geminate contrast and, among immigrants, the FG subjects performed better than the SG subjects.4 The difference is statistically significant (F(2, 90) = 40.717, p < .001), thus indicating that the three groups of subjects performed quite differently with regard to the consonant length discrimination on words. Most importantly, the difference in performance between the Lucchese dialect speakers and the FG immigrants prevents us from naïvely attributing the imperfect discrimination by FG immigrants to the fact that their L1 (the Lucchese dialect) exhibited the process of consonant degemination mentioned earlier. In fact, if this were the case, the same degree of difficulty should emerge for the Lucchese dialect speakers. On the contrary, socio-linguistic attrition is responsible for shaping the perceptual performance of FG immigrants in E1, thus explaining the higher error rate for this group of subjects with respect to the Lucchese dialect speakers. Other important differences between these two groups of subjects are discussed in the following. Focusing again, for the moment, on the general trend for E1, some striking differences arose among subjects, in particular within the group of FG immigrants. Some of them performed on a par with native speakers (e.g. subjects FG5 and FG7), some performed poorly (e.g. subjects FG1 and FG6). In order to explain this disparity, we tried to relate it to some external factors that could be relevant in perceptual processing. In this respect, we could exclude that age, age of arrival in the USA, and current job position were relevant in 4

Contrary to the general trend observed for SG immigrants, SG5 showed a rather low error score (17.7%) that approximates to the average performance of FG immigrants and Lucchese subjects. Even if it is impossible to give a clear and irrefutable explanation for that, such discrepancy has probably to be viewed as evidence of the fact that, even if SG immigrants are supposed to completely rely on the AE phonological system, they are also known to have been having an early and continuous contact with Lucchese, due to the multilingual and multicultural character of their families. This circumstance may have produced, in some of the subjects, an individual sensitivity to the singleton-geminate contrast, which goes beyond the expected inability to perform the discrimination task. The International Journal of Bilingualism

10

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

Table 2 Error percentages in E1.

Subject FG1 FG2 FG3 FG4 FG5 FG6 FG7 FG8 mean s.d. SG1 SG2 SG3 SG4 SG5 SG6 SG7 Mean s.d. LU1 LU2 LU3 LU4 LU5 LU6 LU7 LU8 LU10 LU11 LU14 LU15 LU16 LU17 LU19 LU20 mean s.d.

Mean

R

FG immigrants 50 36 22  9  0 50  0 29 24.5 24.5 SG immigrants 34.7 57 45 43 43.7 58 69.7 83 17.7  8 51.3 57 61.3 41 46.19 49.57 20.37 22.89 Lucchese speakers   3.3  0  0  0   2.7  8 14.7 27 13.7  7  0  0 15.3 22  0  0   5.8  8   2.6  7   8.8 25 13.1 21   8.8  8   8.8 16 41 41   2.9  0   8.75 11.88 11.472 12.25 50 30 11 12  0 46.7   3.3 24.7 22.21 18.91

S

T

50 33 11 10  0 40 10 30 23 17.63

50 21  0 17  0 50  0 15 19.13 20.8

11 50 40 60 29 40 60 41.43 17.54

36 42 33 66 16 57 83 47.57 22.57

10  0  0 10 20  0 10  0 10  0  0 10 20  0 30 10   8.13   9.11

 0  0  0  7 14  0 14  0  0  0  0  7  0  8 50  0   6.25 12.72

Average values (in both rows and column) and standard deviations (in row) have been italicized in order to be differentiated from speaker-specific values.

shaping the FG perceptual performance in E1. On the contrary, we found that the Italian educational background was a significant parameter. In fact, a clear positive correspondence between schooling and perceptual success was found. The two subjects who reached the primary level of education (i.e., FG1 and FG6) appeared to have more perceptual biases with respect to the two subjects with a middle level of education (i.e., FG2 and FG8), who The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

11

in turn performed poorly if compared with the four subjects with a high school level of education (FG3, FG4, FG5 and FG7). Before commenting on this correspondence, let us analyze what happens within the Lucchese speakers’ group, where some degree of variability in the perceptual score was also attested. Some participants performed like NIS (mean error 0%, see LU2, LU6, LU8), while others exhibited a higher error score (see in particular LU4, LU5, LU7, LU15 and, dramatically, LU19). However, in the case of the Lucchese-speakers group, it proved impossible to clearly relate these discrepancies to socio-linguistic variables such as educational level or age, since these factors did not appear to interact with the speakers’ individual performance. Consider, for example, that among the four subjects who made more than 10 per cent of errors (LU4, LU5, LU7, LU15) there were young as well as older people (29, 42, 57 and 62 years old), with either a high level of schooling or a low one, male and female and so on. A very similar pattern of variability can be found for those subjects whose performance was very good or excellent. The only noticeable exception is represented by subject LU19, whose error rate radically differed from the average performance of all other participants (41%). It seems likely that both age (LU19 is 67, thus among the oldest subjects of this group) and, in particular, educational level (LU19 is the only one who only reached the primary level) might explain this result. This is however the only detectable pattern of supposed correlation between socio-linguistic variables and perceptual performance to be found in the data referring to the group of Lucchese speakers. We can then conclude that the perceptual status of the singleton vs. geminate contrast among our sample of native Lucchese speakers is rather blurred and varied across subjects. Recall that the process of consonant degemination in production was said to be highly variable in nature, as far as the Lucchese dialect is concerned (see section 2). The comparison between the Lucchese native speakers group and the FG immigrants reveals therefore that, in the case of the former group, variability (in perception as well as in production) is intrinsic and structural, while in the case of the latter, attrition is influenced by the educational level of individuals in the sense that those subjects who had reached a higher level of schooling were more prone to preserve the perceptual sensitivity to the singleton vs. geminate consonant contrast. As far as the three consonants are concerned ([r()], [s()] and [t()]), we found that subjects’ discrimination did not differ either as a whole (F(2, 90) = 0.436, p >.05), or considering the three groups separately (F(2,60) = 1.001, p >.05 for Lucchese speakers; F(2,45) = 0.997, p >.05 for FG immigrants; F(2,18) = 0.282, p >.05 for SG immigrants).

5 Experiment 2 5.1 Subjects

The subjects of E2 were the same who were mentioned in section 4.1 for E1. 5.2 Materials

This experiment made use of non-word VC(:)V stimuli where vowels were consistently /a/ and the first one was stressed. As in E1, the target consonants were // – //, // – // The International Journal of Bilingualism

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

12

and // – //. Six stimuli were uttered by the same native Italian speaker of E1: [], [], [], [], [], []. These tokens constituted the starting point for the construction of three continua, whose members had consonant length values covering gradually the full range between the singleton and the geminate value. Both the target consonant and the preceding vowel were modified as for duration. The details are given in the Appendix. The total number of tokens for every continuum was six. The geminate to singleton ratio (tokens 6:1) was 3.72 for /r/, 1.79 for /s/, and 2.02 for /t/. Even in the case of E2, therefore, the CC/C ratio for the trill appears to be much more salient than for the other consonants. These circumstances and their consequences will be discussed in sections 6 and 7. 5.3 Method

Subjects had to assign every stimulus of the continuum, randomly presented, to one of the two possible categories (singleton vs. geminate consonant). The three continua were played separately. Similar to the first experiment, subjects had to mark the correct answer on a sheet, where the two possibilities (with singleton and geminate consonant) were given. Every continuum was presented three times, in three different randomizations. The analysis of results included the calculation of mean responses by stimulus and by subjects and the Pearson’s correlation coefficient. 5.4 Results of E2

The results for non-words are given in Figures 1, 2 and 3. On the x-axis the stimuli of the continuum are represented, where 1 corresponds to the original singleton consonant and 6 to the original geminate consonant; positions 2 to 5 are occupied by the synthesized stimuli. On the y-axis scores are assigned to the quality of the response: we attributed the value 1 to the singleton consonant response, the value 2 to the geminate consonant response. The average values among subjects of each group are plotted in the diagram. Responses which are located at intermediate points between 1 and 2 reflect within-group variability in the score assignment. The NIS (control group) performed a clearly categorical discrimination of the phonetic continuum, with edge members being labeled as either singleton (value: 1) or geminate (value: 2), and only one intermediate member with oscillating labels (in the /t/ continuum, there are two intermediate members with oscillating label, but the values are however increasing: 1.5 for stimulus 3, 1.75 for stimulus 4). The category boundaries were situated at the third stimulus of the continuum for // (Fig. 1), the second stimulus for // (Fig. 2) and somewhere between the third and the fourth stimulus for // (Fig. 3). This response pattern constitutes the reference point for evaluating the performance of both immigrants and Lucchese dialect speaking subjects. The latter show a pattern of response that closely resembles that of NIS (in all cases the correlation is significant at p <.01; r =.987 for the /r/ continuum,.951 for /s/, and.963 for /t/). The former, conversely, show a more divergent pattern, especially in the case of SG immigrants and in the discrimination of the [] – [] continuum. The statistical analysis reveals that, in the case of FG immigrants, the /r/ and /t/ response pattern is significantly correlated with that of NIS (r = .991 and.996 respectively, p < .01), but no significant correlation is observed The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

13

Figure 1 Mean responses by stimulus and by group in E2, with [] as experimental consonant. Here and in the following tables, NIS = Native Italian Speakers (control group), NLU = Native Lucchese Speakers, FG = First Generation Immigrants, SG = Second Generation Immigrants.

Figure 2 Mean responses by stimulus and by group in E2, with [s] as experimental consonant.

in the case of /s/ (r =.792, p >.05). As for SG immigrants, no correlation proved to be significant, even if it approached significance in the case of /r/ (r =.814 for /r/, p =.059; r = .641 for /s/, p > .05; r = .683 for /t/, p > .05).

6 General discussion Our first expectation (cf. Hypothesis 1) found little if any support in the data. Participants’ performance on real words (E1) showed no significant differences in the contrast discrimination on [r()] as opposed to [t()] and [s()], either between or within subjects’ groups. As for individual responses, only one subject among FG immigrants (i.e. FG3) showed a strongly impaired discrimination of the [] – [] contrast in words, and a more successful discrimination of the other contrasts. Among SG immigrants, the same was true for The International Journal of Bilingualism

14

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

Figure 3 Mean responses by stimulus and by group in E2, with [] as experimental consonant.

SG1, SG3 and SG4, but the reverse pattern was observed for SG5 and SG7, where the rhotic is better discriminated than the fricative and the stop. Similarly, some of the Lucchese subjects showed a higher error rate in the case of [r()] discrimination in words, with respect to [t()] and [s()] discrimination, but others exhibited a higher error rate in the case of [s()]. As for non-words (E2), the [] – [] contrast was perfectly discriminated by both Lucchese dialect speakers and FG immigrants but less successfully by SG immigrants, who however failed to discriminate between the other two contrasts as well. Thus, the preference for the // degemination which is attested in production for the Lucchese dialect (recall the data presented in section 2), cannot be directly mapped onto the perceptual processing of our subjects, either for the groups of immigrants or – which is most relevant – for the Lucchese dialect speakers’ group. On the contrary, when a discrimination impairment arises, it surfaces in the perceptual tasks no matter what kind of consonant is involved. The second expectation, testing the globally impaired discrimination of immigrants with respect to Lucchese speakers, and of SG immigrants with respect to the other groups of subjects (cf. Hypotheses 2a and 2b), has been clearly confirmed by our data. As for E1, FG immigrants showed a higher error rate with respect to the Lucchese dialect speakers; in E2, on the other hand, the Lucchese speakers’ perceptual performance mirrored that of the control group, while the results obtained by the two groups of immigrants were quite distant from it. Socio-linguistic attrition is therefore likely to be responsible for shaping the perceptual performance of FG immigrants. As far as the second part of the expectation is concerned (cf. Hypothesis 2b), we saw that the error percentage in E1 was significantly higher in SG immigrants with respect to the other groups. Similarly, the response pattern in E2 was straightforwardly divergent from that of the NIS. This is especially remarkable in comparison with the performance of FG immigrants on the same task. In particular, FG immigrants were able to perform a robust categorical discrimination of the phonetic continuum, although there was also some degree of variation with respect to the NIS mean response (especially in the case of the [s]-[s] contrast). SG immigrants, on the contrary, demonstrated a far from comparable type The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

15

of categorical perception (no significant correlation with NIS pattern of response): the judgment values for the edge members often fluctuated from 1.3 to 1.6 in a flat distribution all along the continuum, and the labeling was occasionally inconsistent with the general trend. We can therefore conclude that SG immigrants appear to be strongly impaired in the perception of the consonant length feature, regardless of the task condition. This circumstance is probably to be attributed to the fact that SG immigrants, though exposed to the Lucchese variety spoken by parents at home, are nevertheless relying on the AE phonological system. The expectation about their insensitivity to the singleton vs. geminate contrast received therefore a strong support. As far as Hypothesis 3 is concerned, we believe that there is evidence in support of the claim that both the Lucchese dialect native speakers and the FG immigrants perform significantly better than SG ones on non-word discrimination (E2). As shown in Table 2, FG immigrants can successfully identify the length feature even in the case of real word pairs since the average number of errors for that group is significantly lower than that of the SG immigrants. However, bear in mind the individual differences that we remarked earlier. Only two of the FG subjects showed a high level of discrimination of the singleton vs. geminate contrast (i.e. FG5 and FG7); the other six subjects showed an error rate as high as that of many SG subjects. A successful discrimination of the length contrast in E1 cannot be attributed to FG immigrants as a group, since their performance was clearly not comparable to the NIS control group. We therefore believe that the most important discrepancy between SG and FG immigrants lies in the discrimination on non-word pairs (E2), where FG immigrants, and not the SG immigrants, showed a striking consistency with the control group response pattern, in at least two out of three consonant categories. In this respect, the perceptual performance showed by the FG immigrants appears to be more similar to that of the Lucchese dialect native speakers. The high incidence of cross-subject variation, which emerged in E1, is worthy of further discussion. In this regard, we must note that attritional phenomena, like acquisitional phenomena, have different consequences for individual speakers. This is arguably due to the intrinsically variable nature of subjects’ linguistic competence in attritional conditions. Cook (1989), for example, believes that the two most conspicuous phenomena of language attrition are ‘(1) structural (and stylistic) simplifications and (2) dramatic increases of variability due to the incongruent and idiosyncratic “change”’ (1989: 235). Several examples of this variability for phonology (Campbell & Muntzel, 1989; Cook, 1989; Schmidt, 1985b; see also Dressler, 1972) as well as morphology (e.g. Schmidt, 1985a) can be found in the attrition literature. Simplification at the individual level increases indeed the variability and the complexity of the total system. In the case of our study, we might suppose that the inherently variable phonological competence, as documented for phonological attrition in speech production, can also be a source of perceptual confusion. In addition, as said in section 5 for E1, the external variable of education and schooling may explain some aspects of this variability, at least as far as the group of FG immigrants is concerned. We finally consider two particular aspects of this study in more detail. First, we further discuss the results obtained by SG immigrants in E2, by taking into account one of the most powerful theoretical notions elaborated within the framework of cross-language speech perception research, that is the notion of perceived phonetic distance. Second, The International Journal of Bilingualism

16

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

we will discuss the empirical consequences of the relationship between phonological experience and task variable in perceptual processing. With respect to the first point, we have demonstrated that the SG immigrants’ performance on non-words was strongly impaired. However, it bears noting that SG immigrants seemed to perform slightly better on the []-[] continuum, than on the other continua (see Fig. 1). We hypothesize that such a result may be accounted for by the Speech Learning Model and the notion of perceived distance (cf. Flege, 1987; 1993; Guion, Flege, Akahane-Yamada, & Pruitt, 2000). The basic assumption in this model is that discrimination of foreign sounds applies with different degrees of success, according to the possibility of setting up a comparison with the native language categories. Discrimination is successful when a non-native sound is rated as either completely different or wholly comparable with respect to some known sound, since both disjunction and overlap are easy operations to carry out by a non-native hearer. On the other hand, when some degree of similarity is involved in cross-linguistic comparison, discrimination is generally more difficult. The relevant condition in the perceptual activity of the hearer, then, is represented by the fine-grained mapping of the new category onto a known one. The theory predicts that a phonetic contrast involving a known or a new sound will be detected and learned more easily than one involving a similar sound. Other models which have been elaborated within the field of cross-language speech perception are devoted to explain learners’ perceptual performances by means of a hierarchy of possible perceptual relations among sounds, on the basis of some kind of perceived phonetic distance (auditory, acoustic, articulatory); see the Perceptual Assimilation Model (Best, 1995; Best et al., 1988), or the Perceptual Magnet Effect as theorized in Kuhl (1991). Different notions of markedness, however, in production as well as perception, are conceived of as relevant factors in many psycholinguistic models of first and second-language acquisition (see e.g. the Ontogeny and Phylogeny Model, Major, 2001). In the case of our SG subjects, the paradigm of the equivalence classification suggests that the []-[] contrast involves some inherent features that make the discrimination easier compared to the []-[] and []-[] contrasts. We believe that a detailed account of the phonetic properties of these sounds can confirm such prediction. In fact, even if the basic property distinguishing the first from the second member of each contrast is length, the three conditions provide AE speakers with different degrees of perceptual difficulty. The necessary premise is that SG immigrants differ from all other groups of subjects for fully relying on an AE phonological grammar. In the case of the []-[] contrast, the first element of the opposition should sound like a completely known phonetic unit to the SG subjects, since the voiceless sibilant basically shares the same articulatory and acoustic features in AE (Ladefoged & Maddieson, 1996: 146), SI and Lucchese pronunciations. The geminate consonant [], on the contrary, can be classified merely as a similar sound, as the length feature is absent from the AE phonology. The []-[] contrast, then, involves a known vs. a similar sound. In the case of the []-[] contrast, a second feature becomes relevant in addition to the length, and this is the point of articulation: laminal dental or alveodental in SI and Lucchese, apical alveolar in AE (Ladefoged & Maddieson, 1996: 41). It is then necessary to consider the []-[] contrast as an opposition between two similar sounds. The case of the [] – [] contrast is entirely The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

17

different. In the case of singleton rhotic approximant, a displaced point of articulation characterizes the AE phoneme (which is usually postalveolar, and may present a retroflex tongue configuration; Westbury, Hashi, & Lindstrom, 1998, and references therein) with respect to the Italian corresponding one (which is generally alveolar, and is more often realized as a flap or tap than as an approximant). The Italian intervocalic geminate [], on the contrary, is likely to sound completely new to our subjects since, in addition to the displaced point of articulation and the presence of the length feature, its manner of articulation (trill) distinguishes it from the native (singleton) counterpart. We hypothesize, then, that this contrast involves a similar vs. a new sound. Thus, this condition would be the relevant factor for the improved discrimination by SG immigrants. They hardly perceive the length feature in non-words, but if a new sound is involved, discrimination is improved. One could object that, in our materials, the CC/C ratio for the trill was largely more salient than for the sibilant and the stop, and this would be enough for explaining why SG immigrants’ performance was better with trills than with sibilants and stops. However, we cannot agree with this view. If this were the case, indeed, the same preference for the trill should be expected for all groups of subjects, and not just for SG subjects. We conclude, therefore, that the paradigm of the equivalence classification appears to have a strong explanatory power for the specificity of SG immigrant’s performance in E2. We come now to the second point mentioned earlier, that is, the relationship between perceptual processing and task variables. As already seen, FG immigrants and, partially, the Lucchese dialect speakers, perform much like the NIS in the discrimination of a singleton vs. geminate consonant on non-words (E2). Such finding suggests that an important correlation has to be set between two apparently independent variables: subjects’ linguistic – phonological – skills (that is, the degree of familiarity with some phonetic features of a language that is not, or no longer, one’s first language), and the properties of speech processing strategies requested by the specific task. In the acquisition of Italian affricates by Greek- and Portuguese-speaking learners (Celata, 2004), linguistic experience played a relevant role in a non-native sound identification task which made use of L2 words (low-experienced learners exhibited higher error rates than high-experienced learners), while a non-native sound discrimination test in auditory simplified conditions (non-words) was better performed by low-experienced learners than by high-experienced ones. In order to account for such discrepancy, we hypothesized the existence of a two-way perceptual development in second-language learners. At the initial stage, when the experience with the target language is limited, phonetic processing is prone to follow a purely acoustic-discriminatory strategy; as a consequence, high sensitivity to small variations in the phonetic quality of the stimulus is possible. At a later stage, when a broader experience of the target language has developed, the acoustic discrimination strategy is replaced by an essentially phonemic awareness of sounds; a large-scale interference in processing then takes place, in which semantic and morpho-syntactic processing strategies are activated, producing a general inhibition of the discrimination of fine-grained acoustic differences. Similar results found by Mora (2005) with advanced learners of English have already been cited (see section 2). The hypothesis of the distinction between an acoustic and a linguistic processing of sounds can be related to theories of speech perception in young children vs. adults (see in particular Best et al., 1988; Werker & Tees, 1984a, 1984b, 1999). It is well known that The International Journal of Bilingualism

18

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

Table 3 Perceptual abilities in second-language acquisition (from Celata 2004) and first-language attrition.

Subjects

Word discrimination

Non-word discrimination

NIS L2 experienced learners L2 low-experienced learners Lucchese speakers FG immigrants SG immigrants

+ + -/+ -

+ + + + -

infants under 10–12 months are able to discriminate tokens of the same phonetic category, minimally contrasting for acoustic or articulatory features, while adults appear to have lost such perceptual sensitivity. The basic assumption is that the progressive impairment does not reflect loss of neural plasticity, but rather a change in speech processing strategies, as a consequence of the introduction of a new representation level, where segmental, rhythmical, statistical and possibly other properties of the signal need to be coordinated for full understanding of the speech unit. This functional reorganization principle explains both the decline in infants’ ability to discriminate non-native phones and the decrease in the amount of detail used by infants as they move from speech perception to word learning. We believe that this paradigm may beneficially inspire the study of socio-linguistic attrition as well. As far as this study is concerned, the performance of FG immigrants can be viewed as similar to that of low-experienced second-language learners: sound discrimination is globally unsuccessful on words, but successful on non-words. The FG immigrants of our study behaved, thus, as low-experienced subjects with respect to the length distinction in consonants: prominent phonetic cues could be retrieved in an enhanced auditory context (non-words) but not in the linguistic processing of real words. Moreover, since some familiarity with the length feature has to be postulated for FG immigrants, we can state that their perception does not rely on an AE-like system of oppositions, in contrast with what we argued for SG immigrants, but rather on a more Lucchese-like system, where the consonant length distinction can be overtly realized. The convergence between acquisitional and attritional data can be summarized as in Table 3. Note that the L2 low-experienced learners studied in Celata (2004) and the FG immigrants of this study performed similarly. Moreover, note that a progressive regression in perceptual abilities has to be postulated for FG immigrants with respect to Lucchese dialect speakers, and for SG immigrants with respect to FG ones.

7 Summary and conclusions The present study offered a perceptual investigation of the singleton vs. geminate consonant contrast in two populations of Lucchese speakers, either immigrated to the USA or resident in the homeland. It however suffered from some inherent limits and overwhelming difficulties. First of all, the objective dearth of data regarding the phonological, The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

19

distributional, and socio-linguistic status of the consonant degemination process in Lucchese prevented us from relating our perceptual findings to a comprehensive and conventional phonological framework for this dialectal area. As a consequence, many assumptions were based on empirical evidence, rather than on established knowledge. Moreover, we are aware of the fact that the modest number of subjects who participated in the experiments does not allow us to give a strong support to any incidental generalization suggested by the obtained results. We therefore believe that further investigation is needed, on both descriptive and experimental grounds. However, within the limits of the present study, the analysis of the singleton vs. geminate contrast discrimination in the Lucchese dialect speaking community of San Francisco has allowed us to bring to light some positive evidence of phonological-perceptual attrition in immigrants with respect to the speakers of the homeland. Evidence was found of a progressive impairment in the perception of the length feature in three groups of subjects, most likely due to socio-linguistic attrition: first generation immigrants showed an impaired discrimination of the singleton vs. geminate contrast, if compared with Lucchese speakers still living in Lucchesia; on the other hand, they were more accurate than second generation immigrants. In particular, the discrimination of pairs of words opposing a singleton to a geminate consonant and the identification of a geminate vs. singleton consonant in non-words revealed that first generation immigrants could perform, on the whole, more successfully than second generation immigrants. We have then argued that the perceptual behavior of SG immigrants was based on an AE-like phonological system, where no length distinctions in intervocalic consonants are lexically represented. Moreover, as far as the identification task on non-words was concerned, SG immigrants have shown some degree of sensitivity to variations in the length feature only in the case of the []-[] contrast, where a completely new sound – following the equivalence classification paradigm – was involved. Moreover, we have argued that the attritional process is shaped by patterns of perceptual sensitivity, which in some way mirror the emergence of different perceptual abilities in L2 phonological acquisition. As for the parallelism between FG immigrants’ and L2 low-experienced learners’ sensitivity to consonant duration in non-words, we have hypothesized that there is a stage in both language acquisition and loss, in which the acoustic properties of a non-native signal can be rather easily detected if the access is not influenced by contextual information, normally present indeed in communicative exchange. In both second-language acquisition and attrition, this stage has to be located at the beginning of the process of contact with a new sound system. This condition can also be related to the phenomena of perceptual sensitivity to non-native contrasts in young infants. Several directions are open for future research. First, the body of primary phonological data in attritional situations (for both perception and production) needs to be incremented in order to clearly identify the patterns of language variation, shift and death. Secondly, the study of perception of sounds in a foreign language, in both acquisitional and attritional contexts, has not fully explored the challenging proposals emanating from first language speech perception research about the functional reorganization hypothesis for speech processing. To this aim, both the variables of linguistic experience and of task demand should be tested simultaneously as relevant factors that shape the complex process of L2 sound categorization. The International Journal of Bilingualism

20

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

References AGOSTINIANI, L. & GIANNELLI, L. (1990). Considerazioni per un’analisi del parlato toscano. Studi Linguistici Italiani, 25, 219–237. ALBANO LEONI, F. & MATURI, P. (2002). Manuale di fonetica. Roma: Carocci. AMBROSINI, R. (1974). Sul dialetto lucchese. Rivista di Archeologia, Storia, Economia e Costume, 2(3–4), 31–35. ANDERSEN, R. W. (1982). Determining the linguistic attributes of language attrition. In R. Lambert & B. Freed (Eds.), The Loss of Language Skills (pp. 83–118). Rowley: Newbury House Publishers. BERTINETTO, P. M. & LOPORCARO, M. (2005). The sound pattern of Standard Italian, as compared with the varieties spoken in Florence, Milan and Rome. Journal of the International Phonetic Association, 35(2), 131–151. BERTINETTO, P. M. & VIVALDA, E. (1978). Recherches sur la perception des oppositions de quantité en italien. Journal of Italian Linguistics, 3, 97–116. BEST, C. (1995). A direct realist perspective on cross-language speech perception. In W. Strange (Ed.), Speech perception and linguistic experience: Theoretical and methodological issues in cross-language speech research (pp. 167–200). Timonium, MD: York Press. BEST, C. T., MCROBERTS, G. W. & SITHOLE, N. M. (1988). Examination of Perceptual Reorganization for Nonnative Speech Contrasts: Zulu Clicks Discrimination by EnglishSpeaking Adults and Infants. Journal of Experimental Psychology: Human Perception and Performance, 14, 345–360. BULLOCK, B. E. & GERFEN, C. (2004). Frenchville French: A case study in phonological attrition. International Journal of Bilingualism, 8, 303–320. BULLOCK, B. E., TORIBIO, A. J., GONZÁLEZ, V., & DALOLA, A. (2006). Language Dominance and Performance Outcomes in Bilingual Pronunciation. In M. Grantham O’Brien, C. Shea & J. Archibald (Eds.), Proceedings of the 8th Generative Approaches to Second Language Acquisition Conference (GASLA 2006) (pp. 9–16). Somerville, MA: Cascadilla Proceedings Project. CAMPBELL, L. & MUNTZEL, M. (1989). The structral consequences of language death. In N. Dorian (Ed.), Investigating Obsolescence: Studies in language contraction and death (pp. 181–196). Cambridge: Cambridge University Press. CELATA, C. (2004). Acquisizione e mutamento di categorie fonologiche. Le affricate in italiano. Milano: Franco Angeli. CELATA, C. & KAEPPELI, B. (2003). Affricazione e rafforzamento in italiano. Alcuni dati sperimentali. Quaderni del Laboratorio di Linguistica della SNS, 4. COOK, E.-D. (1989). Is phonology going haywire in dying languages? Phonological variations in Chipewyan and Sarcee. Language in Society, 18, 235–255. DORIAN, N. (1981). Language death: The life cycle of a Scottish Gaelic dialect. Philadelphia: University of Pennsylvania Press. DRESSLER, U. W. (1972). On the phonology of language death. Proceedings of the Chigaco Linguistics Society, 8, 448–449. FLEGE, J. E. (1987). The Production of ‘New’ and ‘Similar’ Phones in a Foreign Language. Evidence for the Effect of Equivalence Classification. Journal of Phonetics, 15, 47–65. FLEGE, J. E. (1993). Production and perception of a novel, second-language phonetic contrast. Journal of Acoustical Society of America, 93, 1589–1608. GIANNELLI, L. (2000). Toscana. Pisa: Pacini. GRAFF, D., LABOV, W. & HARRIS, W. A. (1986). Testing listeners’ reaction to phonological markers of ethnic identity: A new method for sociolinguistic research. In D. Sankoff (Ed.), Diversity and diachrony (pp. 45–58). Amsterdam and Philadelphia, PA: Benjamins. GUION, S. G., FLEGE, J. E., AKAHANE-YAMADA, R. & PRUITT, J. C. (2000). An investigation of current models of second language speech perception: The case of Japanese The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

21

adults’ perception of English consonants. Journal of the Acoustical Society of America, 107, 2711–2724. KUHL, P. K. (1991). Human adults and human infants show a ‘perceptual magnet effect’ for the prototypes of speech categories, monkeys do not. Perception and Psychophysics, 50, 93–107. LADEFOGED, P. (1999). ‘American English’. In Handbook of the International Phonetic Association (pp. 41–44). Cambridge: Cambridge University Press. LADEFOGED, P. & MADDIESON, I. (1996). The Sounds of the World’s Languages. Oxford: Blackwell. LI, WEI (1994). Three Generations Two Languages One Family: Language choice and language shift in a Chinese community in Britain. Clevedon: Multilingual Matters. LOPORCARO, M. (1996). On the analysis of geminate in Standard Italian and Italian dialects. In B. Hurch & R. Rhodes (Eds.), Natural phonology: The state of the art. Papers from the Bern Workshop on Natural Phonology, September 1989 (pp. 153–187). Berlin, New York & Amsterdam: Mouton de Gruyter. MAJOR, R. C. (2001) Foreign accent: The ontogeny and phylogeny of second language phonology. Mahwah, NJ: Lawrence Erlbaum. MAJOR, R. C. (2005). First Language Attrition Research. Paper presented at the 2nd International Conference on First Language Attrition, Vrije Universiteit Amsterdam 17–20 August 2005. MAROTTA, G. (1985) Modelli e misure ritmiche. La durata vocalica in italiano. Bologna: Zanichelli. MAROTTA, G. (1995). ‘Apocope nel parlato di Toscana’. Studi Italiani di Linguistica Teorica e Applicata, 24, 297–322. MILLER, J. L. & EIMAS, P. (1994). Observations on speech perception, its development, and the search for a meaning. In J. C. Goodman & H. C. Nusbaum (Eds.), The Development of Speech Perception: The Transition from Speech Sounds to Spoken Words (pp. 37–56). Cambridge, MA: MIT Press. MORA, J. C. (2005). Lexical knowledge effects on the discrimination of non-native phonemic contrasts in words and nonwords by Spanish/Catalan bilingual learners of English. ISCA Workshop on Plasticity in Speech Perception (PSP 2005). London, 15–17 June 2005. MULJAC˘IC´, Ž. (1969). Fonologia generale e fonologia della lingua italiana. Bologna: Il Mulino. MULJAC˘IC´, Ž. (1972). Fonologia della lingua italiana. Bologna: Il Mulino. PAYNE, E. M. (2005). Phonetic variation in Italian consonant gemination. Journal of the International Phonetic Association, 35, 153–181. PICKETT, E. R., BLUMSTEIN, S. & BURTON, M. (1999). Effects of speaking rate on the singleton/geminate contrast in Italian. Phonetica, 56, 135–157. PIERI, S. (1890). Fonetica del dialetto lucchese. Archivio Glottologico Italiano, 12, 107–126. REPP, B. (1981). Two strategies in fricative discrimination. Perception and Psychophysics, 30, 217–227. ROACH, P. (2000). English Phonetics and Phonology. Cambridge: Cambridge University Press. ROHLFS, G. (1966). Grammatica storica della lingua italiana e dei suoi dialetti. Fonetica. Torino: Einaudi. SAIU, E. (2004). Sull’origine d’una peculiarità del raddoppiamento fonosintattico in area lucchese. L’Italia Dialettale, 65, 123–130. SCAGLIONE, S. (2000). Attrition. Mutamenti sociolinguistici nel lucchese di San Francisco. Milano: Franco Angeli. SCHMID, M. S., KOPKE, B., KEIJZER, M., & WEILEMAR, L. (Eds.), (2004) First language attrition: Interdisciplinary perspectives on methodological issues. Amsterdam: Benjamins. SCHMID, S. (2005). Code-switching and Italian abroad. Reflections on language contact and bilingual mixture. Italian Journal of Linguistics, 17(1), 113–155. SCHMIDT, A. (1985a). The fate of ergativity in dying Dyirbal. Language, 61, 378–397. SCHMIDT, A. (1985b). Young people’s Dyirbal. An example of language death from Australia. Cambridge: Cambridge University Press. The International Journal of Bilingualism

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

22

STAGER, C. L. & WERKER, J. F. (1997). Infants listen for more phonetic detail in speech perception than in word learning tasks. Nature, 388, 381–382. WERKER, J. F. (1994). Cross-language speech perception: Developmental change does not involve loss. In J. C. Goodman & H. C. Nusbaum (Eds.), The development of speech perception: The transition from speech sounds to spoken words (pp. 93–120). Cambridge, MA: MIT Press. WERKER, J. F. & POLKA, L. (1993). The ontogeny and developmental significance of language-specific phonetic perception. In B. de Boysson-Bardies, S. de Schonen, P. Jusczyk, P. MacNeilage, & J. Morton (Eds.), Developmental neurocognition: Speech and face processing in the first year of life (pp. 275–288). Dordrecht: Kluwer Academic Publishers. WERKER, J. F. & TEES, R. C. (1984a). Phonemic and phonetic factors in adult cross-language speech perception. Journal of the Acoustical Society of America, 75, 1866–1876. WERKER, J. F. & TEES, R. C. (1984b). Cross-language speech perception: Evidence for perceptual reorganization during the first year of life. Infant Behaviour and Development, 7, 49–63. WERKER, J. F. & TEES, R. C. (1999). Influences on infant speech processing: Toward a New Synthesis. Annual Review of Psychology, 50, 509–535. WESTBURY, J. R., HASHI, M. & LINDSTROM, M. J. (1998). Differences among speakers in lingual articulation for American English /r/. Speech Communication, 26, 203–226. WOLFRAM, W. & SCHILLING-ESTES, N. (1998). American English: Dialects and variation. Malden, MA: Basil Blackwell.

Appendix A FG immigrants: socio-linguistic profile.

Subject (gender)

Place of birth

FG1 (m) FG2 (m) FG3 (f) FG4 (m) FG5 (m) FG6 (f) FG7 (m) FG8 (m)

Corsagna Pieve S. Paolo Santa Margherita Pieve S. Paolo Carraia Diecimo Corsagna Corsagna

Age

AoA*

Years of education

55 73 72 58 61 54 57 61

27 20 18 23 23 19 24 22

 5  8 12 12 12  5 12  8

Current occupation factory worker restaurant keeper housewife factory worker factory worker shop assistant factory worker restaurant keeper

* AoA = Age of arrival in the USA.

Appendix B SG immigrants: socio-linguistic profile.

Subject (gender)

Place of birth

Parents’ place of birth

SG1 (f) SG2 (f) SG3 (m) SG4 (f) SG5 (f) SG6 (m) SG7 (m)

San Francisco San Francisco San Francisco Chicago San Francisco San Francisco San Francisco

Corsagna Corsagna P. S.Paolo/S.Margherita Carraia/Diecimo San Gennaro San Gennaro/Carraia Corsagna

The International Journal of Bilingualism

Age 36 24 35 33 30 34 35

Years of education 12 12 15 15 12 15 12

Current occupation secretary student restaurant keeper shop assistant housewife barman factory worker

Celata & Cancila: Phonological attrition

23

Appendix C Lucchese subjects: socio-linguistic profile.

Subject (gender)

Place of birth

Place of residence

LU1 (m) LU2 (f) LU3 (m) LU4 (f) LU5 (m) LU6 (f) LU7 (f) LU8 (m) LU10 (m) LU11 (m) LU14 (f) LU15 (m) LU16 (f) LU17 (f) LU19 (m) LU20 (f)

Porcari Porcari Lucca Lucca Porcari Porcari Porcari Capannori San Ginese San Ginese Lucca Pescia Porcari Lucca Porcari Porcari

Porcari Montecarlo Porcari Montecarlo Porcari Porcari Porcari Porcari San Ginese San Ginese Porcari Porcari Porcari Porcari Porcari Porcari

Age

Years of education

55 56 30 29 62 53 57 57 69 65 36 42 67 31 67 68

12 12 16 16  8  8  8  8  5 12  8 12  8  8  5 12

Current occupation dealer teacher student student contractor housewife housewife mason nurse (retired) postman (retired) housewife office-worker office-worker (retired) factory worker factory worker housewife

The International Journal of Bilingualism

24

INTERNATIONAL JOURNAL OF BILINGUALISM 14 (2)

Appendix D Materials of E1.

Bara pesante – Barra pesante Faro nascosto – Farro nascosto Vero maschio – Verro maschio Era lontano – Erra lontano Sera tiepida – Serra tiepida A dure prove – Addurre prove Lunga aringa – Lunga arringa Mora cinese – Morra cinese I rilevanti casi – Irrilevanti casi Pori troppo grandi – Porri troppo grandi Ho speso poco denaro – Ho spesso poco denaro Mese avanti – Messe avanti Casa aperta – Cassa aperta La sicura alla porta – L’assicura alla porta La segna nel calendario – L’assegna nel calendario Acceso al porto – Accesso al porto Pera senza niente – Per assenza niente Maso di montagna – Maso di montagna Fate bene – Fatte bene Grata forte – Gratta forte L’amata sofferente – La matta sofferente Sete di successo – Sette di successo Loto tutte le sere – Lotto tutte le sere Rimase senza moto – Rimase senza motto In dote al matrimonio – Indotte al matrimonio Trota nell’acqua – Trotta nell’acqua Rito in piedi – Ritto in piedi Composto a strato – Composto astratto Tuta nera – Tutta nera

The International Journal of Bilingualism

Celata & Cancila: Phonological attrition

25

Appendix E Materials of E2. 1. Stimuli of the [r] – [r] continuum

V1 C

Stimulus 1 []

Stimulus 2

Stimulus 3

Stimulus 4

Stimulus 5

Stimulus 6 []

221 msec 47 msec

199 msec 73 msec

177 msec 99 msec

155 msec 124 msec

133 msec 150 msec

111 msec 175 msec

2. Stimuli of the [] – [] continuum

V1 C

Stimulus 1 []

Stimulus 2

Stimulus 3

Stimulus 4

Stimulus 5

Stimulus 6 []

152 msec 147 msec

136 msec 164 msec

120 msec 181 msec

105 msec 198 msec

89 msec 216 msec

74 msec 263 msec

3. Stimuli of the [t] – [t] continuum

V1 C

Stimulus 2

Stimulus 3

Stimulus 4

Stimulus 5

Stimulus 6 []

142 msec 130 msec

129 msec 152 msec

116 msec 174 msec

103 msec 196 msec

90 msec 218 msec

The International Journal of Bilingualism