Keywords

Attitude has allegedly been the single most researched topic in social psychology (Augoustinos et al. 2006), and language attitudes have also been a frequently found topic in language-related research. Giles and Billings (2004) provide an impressive list of seminal works (Labov 1966; Lambert et al. 1960/1972), large-scale investigation programmes (Riagáin 1997), and, journal special issues (Ryan et al. 1994) and books (Baker 1992) on language attitudes in their comprehensive review, one of the many in the field (e.g. Riagáin 2008; Ryan et al. 1982; Edwards 1999; Garrett et al. 2003).

Sociolinguists chart language shift s and evaluate the effect of language policies in relation to language-attitude changes (Riagáin 1997; Smagulova 2008; Moriarty 2010). When studying education in multilingual settings, language attitudes are often included in the discussion (Garcia 2009), especially in settings where the linguistic hierarchy is more prominent, such as in bi-/multidialectal, creole, heritage language, and minority language situations (Siegel 1999; Riagáin 2008; Li and Hua 2010; Yiakoumetti and Esch 2010). Munstermann (1989) observes that “almost every study on the problems of dialect and education (in the Netherlands) emphasises the importance of teachers’ attitudes towards dialect” (p. 166). In the school setting, the main concerns are that students’ school performance, self-esteem, and language attitudes may be negatively affected if their first dialects/languages are excluded, stigmatised, or marginalised explicitly in the educational system or implicitly by the teacher (Seligman et al. 1972; Papapavlou and Pavlou 2007). In the family setting, the parents’ language attitudes may contribute to those of the children’s (Li et al. 1997; Luykx 2005; McEwan-Fujita 2010).

With so much at stake, however, a “notorious” enigma that haunts general attitude research also affects language-attitude research: the incongruence between language attitudes and behaviours. Researchers constantly find differences between the attitudes deduced from questionnaires and “guised” experiments, and people’s language choices or related decisions in everyday life. Solving this problem is crucially important because the assumption underlying most language-attitude research is that by understanding language attitudes and their relationship to language behaviours, it may be possible to change one by changing the other. For example, by cultivating more positive language attitudes towards the target language, we hope to motivate students to make more effort to learn it (Gardner 1985). By training prospective teachers in schools with lots of dialect-speaking pupils, it is hoped that the teachers may develop more open and positive attitudes towards using dialects in education (Munstermann 1989). Therefore, it seems that “there wouldn’t be much point in studying attitudes if they were not, by and large, predictive of behaviour” (Gass and Seiter 1999, p. 41)

I would argue that the lack of correspondence between language attitude and behaviour has much to do with how we conceptualise language attitude. The language attitudes inferred from questionnaires, experiments, and language use in interaction are all valid but at different levels of specificity. LaPiere (1934), a pioneer in attitude research, argues that attitudes measured by questionnaires are symbolic responses to symbolic attitudes rather than to actual social situations. Such symbolic attitudes do not necessarily predict actual behaviours in real-life situations. Thus, the so-called discrepancy between attitude and behaviour is caused by comparing the more symbolic and “controlled” attitudes with behaviours in more complicated social situations. Therefore, we as researchers should consider what type of attitude is the target of our study, and choose appropriate research methods and contexts accordingly. We must clarify as far as possible the assumptions about language, language attitudes, context, as well as the research procedures and limits. Language attitudes are highly contextualised, situated communicative achievements that should not be interpreted independent of the contexts in which they are constructed/investigated.

1 Researching Language Attitudes

Psychologist Irving Sarnoff (1960) conceptualises attitude as “a disposition to react favourably or unfavourably to a class of objects”, which can be inferred from an individual’s observable responses to the relevant objects “toward which he has an attitude”. This definition considers attitudes as “real” entities in the human mind, which are relatively consistent and durable. “Having an attitude” is similar to having a nose and a hand. The second part of the definition suggests “facial expressions, postures, locomotion, sounds of voices, and verbalisations” (p. 261) as examples of observable responses from which attitudes may be inferred. Most examples in the second part fall into the category of contextualisation cues (Auer 1996; Gumperz 1992) and interactional features of face-to-face discourses. Generally speaking, the cognitivist approaches to attitude mainly focus on the first part of the definition—attitudes as durable mental constructs, while the discursive approaches capitalise on the second part—the social, discursive, and interactional dimensions.

1.1 Attitudes as Mental Constructs

The most frequently used cognitivist language-attitude model is the tripartite model (Cargile et al. 1994; Riagáin 1997; Baker 1992; Ladegaard 2000), also called the ABC model, which views language attitudes as being comprised of affective, cognitive, and behavioural aspects. The cognitive component refers to people’s beliefs about the attitudinal object, for example, the correctness of different language varieties. The affective component refers to how people feel about the attitudinal objects, for instance, whether people approve or disapprove of a certain way of speaking. The behavioural component refers to people’s predispositions to act in certain ways.

With regard to such a model, researching attitudes becomes a matter of developing appropriate techniques of identifying and measuring these components which are readily available in the human mind. One major approach is using self-reporting surveys and interviews, customarily categorised as “direct methods ” in the field. Apart from practical issues, such as asking hypothetical questions or strongly slanted questions (Garrett 2010), the main criticism of direct approaches centres around whether participants’ self-report attitudes and behaviours in situated contexts are manifestations of the same set of underlying dispositions (Riagáin 2008). Social desirability bias (Oppenheim 1992) is often mentioned as the risk. Such kind of bias occurs when participants seek to present socially appropriate views to the researcher rather than what they think or say in private. When the researcher’s questions concern socially, culturally, or politically sensitive issues, social desirability bias is believed to be more influential. Another type of bias is called the “acquiescence biases ” (Oppenheim 1992), that is, participants tend to agree with the questions and give answers they think would please the researchers. Thus, both the formulation of the questions and the identity/personality of the researcher contribute to this bias.

Despite the problems, direct approaches remain the most frequently used methods for researching language attitudes. For example, national language surveys were widely used from the 1970s to the 1990s in Canada, Finland, Ireland, Wales, and the Basque Country (Garrett et al. 2003; Riagáin 2008). Similar policy rationales can be observed in various settings: in post-Soviet countries, such as Kazakhstan (Smagulova 2008), postcolonial states, such as the Malaysian state of Sarawak (Ting 2003) and autonomous regions, such as Catalonia (Woolard and Gahng 1990). It is observed that the burgeoning use of national language-attitude study (by survey and/or experiment) coincides with emerging language issues on the policy agenda of states. As a result, such studies are often connected with language planning, language revitalisation and maintenance, and ideology of nationalism. Historical, social and economic contexts are often addressed at some length, together with the sociolinguistic setting, to link them to language-attitude patterns. However, the links between sociopolitical theories, social contexts, and the empirical data are not always clear-cut (Withers 1994). The inferences are significant, but are often at the level of logical speculation rather than firmly grounded in, and substantiated by, the details of the data. It is often suggested that fuller understanding of language attitudes and language-use pattern depends on ethnographic and qualitative approaches, but very few ethnographic studies of language attitudes have been done so far.

Another influential approach to language attitude within the cognitivist tradition is called the “indirect methods ”, which is almost used as a synonym for the matched-guise technique (MGT) (Lambert et al. 1960/1972) and its modified versions (Ladegaard 2000). Proponents of the indirect methods frequently speak of the virtue of accessing people’s more private, covert attitudes that are inaccessible through questionnaires or interviews. MGT was introduced by Lambert and his colleagues (Lambert et al. 1960/1972) in a seminal study, which lay the foundation for the interface between social psychology and sociolinguistics, thus helping to establish the cross-disciplinary field of language-attitude research (Giles and Billings 2004). MGT has inspired an enormous amount of language-attitude research worldwide, for example, in Spain (Woolard and Gahng 1990; Echeverria 2005), Canada (Bourhis et al. 1975), Britain (Garrett et al. 2003), France (Hoare 2001), Hong Kong (Ladegaard 2000), and mainland China (Kalmar et al. 1987). It is arguably the single most influential research technique in the field of language attitudes.

The original MGT study (Lambert et al. 1960/1972) used speech samples produced by English–French bilinguals as the stimuli, and asked Canadian college students to evaluate the speaker’s personality according to his or her voices. The students thought each sample represented one speaker, while in fact the bilingual speakers produced samples in both languages they spoke. Therefore, the voices became “guises” of the speaker’s identity, and students’ evaluation of different speech samples produced by the same speaker could be “matched”. It is intended to investigate how the use of different languages alone may cue stereotypic judgments.

The study was ground-breaking in several respects. First, it was the first time to find out that people (the French-speaking college students) can hold unfavourable attitudes towards their own group. The situation was attributed to powerful social stereotypes of different speech communities, which is in turn explained by linguistically mediated socioeconomic hierarchies. It suggests that the minority or dominated groups in a society may adopt the stereotypical values of the majority or dominating groups, and thus denigrate their own groups—a linguistic “minority group reaction ” observed in many subsequent studies (Edwards 2011). The participants are not necessarily aware of, or willing to admit, such attitudes, and thus MGT can help to tap into such private attitudes. Second, hardly any of the variables from the questionnaire correlated systematically with the MGT results. This discrepancy between attitudes elicited by questionnaires and MGT is consistently mentioned in later research. Third, one of the bilingual speakers in Lambert et al.’s study (1960/1972) spoke Parisian rather than Canadian French and the participants’ evaluation of this French guise observably differed from that of other Canadian French guises. This has inspired subsequent investigations of attitudes towards not only whole languages (Echeverria 2005), but also more finely grained aspects of language use, such as speech accommodation (Bourhis et al. 1975), dialectal varieties (Luhman 1990), and accents (Creber and Giles 1983).

In addition, the traits used in the semantic differential rating scale for the speaker evaluation tests have given rise to many now well-recognised dimensions of language attitudes: status (or competence, superiority), social attractiveness (or solidarity), and dynamism (Garrett 2010). Hence, it reveals that language attitudes are multidimensional, so that people may have positive attitudes towards one dimension and negative attitudes towards others.

The MGT has attracted much criticism since its introduction. The main concerns include: ethical issues regarding deception, exclusion of contextual factors, availability of “ideal” bi-/multilinguals, unavoidable bias in the selection of texts, whether participants are evaluating the intended style features of the test, and the possibility of one style feature covariating with others (Garrett 2010). By addressing these problems, researchers continue to refine and develop the MGT, which significantly contribute to our understanding of language attitudes.

For example, Bourhis et al. (1975) switched the focus from attitudes towards individual style features to attitudes towards the process of speech accommodation. The researchers investigated participants’ evaluations of a female Canadian speaker’s switch between formal Canadian French, Standard (European) French, and informal Canadian French when responding to a European French speaker. There was no need to disguise the number of speakers, and the content of the speech samples was closer to real communicative situations. The participants also indicated if they thought there was a shift of styles in the speech samples. The study found studying such processes of identification is significant for language-attitude research. The communication accommodation theory (Giles et al. 1973) underlying this study foregrounds the dynamic aspect of attitude construction and negotiation in interaction:

Making adaptations as we communicate with others may be (or may be seen as) a behavioural signal of our own attitudes, and these adaptations may themselves also evoke attitudinal responses in our communication partners…communication accommodation theory can also be seen as the implementation of attitudes in discourse. (Garrett 2010, p. 105)

Arguably, this dynamic, interpersonal aspect of language attitudes cannot be fully explored by the MGT.

Garrett et al. study (1999) on attitudes towards Welsh English accents view dialect performance and people’s responses to such performance as holistic rather than a simple combination of single variables. The speech samples approximate naturally occurring discourses as the researchers simply asked the speakers to tell stories in their local English dialects. In this way, the researchers were not able to use “neutral texts” for the experiments, which is a significant divergence from a typical MGT study. It is argued that the search for “neutral texts” is fundamentally in vain because social contexts inevitably leak through situated discourses. Therefore, instead of trying to conceal social contexts embedded in the speech samples, the researchers placed considerable weight on analysing the content. They suggest that dialect features cannot fully explain the students’ and teachers’ preferences. Narrative styles and their different social roles also influence their decision. Cluster analysis and multidimensional scaling (MDS) were used to explore the “structure” of the data. Overall, the study demonstrates that people’s evaluation of discourse performance is multidimensional and conditioned by many social and contextual factors, and some of the dimensions contradict and compete with each other. They argue that social evaluation works in more localised and contextualised ways than has often been assumed by previous MGT studies and future research must address language attitudes as a more complex phenomenon in its situated discursive context. In particular, the ideological property of dialects means the study of language attitudes towards dialects should be “linked coherently to the current emphases in the analysis of discourse” (Garrett et al. 1999, p. 323).

1.2 Attitudes as Discursive Practice

As mentioned above, researchers in the cognitivist tradition suggest conducting ethnographic, qualitative study of language attitudes in relation to discourses, in order to compensate for the limitations of quantitative methods. However, investigating language attitudes in discourses, such as interview texts, is not the same as conceptualising attitudes as discursive practices. The latter requires a complete reconceptualisation of the nature of language and language attitude.

To take the discursive approach to attitude, the emphasis is shifted from considering attitudes as underlying mental constructs to focusing on people’s practices of evaluation in particular settings (Potter 1998). Moreover, the link between the evaluative and the factual features in the discourse is considered inextricable, so that “what we see is the entity constituted in discourse in such a way that the evaluation is part of a description of the object”, and that the discourse works as a package—“a seamless texture of talk” (Potter 1988, p. 65). This discursive approach to attitude does not deny the existence of human cognition, but deliberately moves the analytic focus from the cognitive processes to discursive practices in situated activities.

One important distinction between the two approaches is the view of the relationship between contexts and attitudes. The cognitivist approach seems to suggest that the participants’ “true” attitudes are modified and distorted by social and contextual factors in the process of expression. In contrast, the discursive approach sees the attitudes constructed in discursive practices as legitimate attitudes in their own right and as products of the interactions of all relevant interpersonal, contextual, ideological, and social factors. These attitudes are not simply views or ideas conveyed in communication but “are also components of our own communicative competence that underpin…our moment-to-moment deployment of linguistic, non-verbal and discursive resources to achieve our communication goals” (Garrett 2010, p. 120). In other words, attitudes are both the resources and outcomes of the meaning-making and social positioning processes (Wetherell 2007; Wetherell et al. 1987). It has been found that such attitudes systematically vary from moment to moment, which fundamentally challenges the basis of cognitivist attitude research—the stability and durability of attitudes (Garrett 2010; Riagáin 2008).

Compared with the cognitivist tradition, there have been very few language-attitude studies to date explicitly utilising discursive approach es (except Soukup 2009; Hyrkstedt and Kalaja 1998; Liebscher and Dailey-O’Cain 2009; Giles and Coupland 1991). While the “societal treatment studies ” have dealt with the meaning of discourses, they have generally been concerned with written and published texts rather than engaging with interactions (Garrett 2010). On the other hand, language-attitude data used in the field of language socialisation (Garrett 2007; Ochs 1996), language ideology (Woolard and Schieffelin 1994), interactional sociolinguistics (Gumperz 2008, 1982), and many others have rarely been reviewed in the literature of language attitudes. It is fair to say that the discursive approach to language attitudes is a broad umbrella term for methodologies informed by various theories that share similar commitments: studying social meanings in situated discourses (particularly in interaction) and construing languages as resources for constructing social meanings rather than transparent reflexes of meanings. Specific studies strategically combine several theories and methods relevant to their purposes.

Liebscher and Dailey-O’Cain (2009) differentiate between three different levels of analysis in a discursive approach to language attitudes: content-based approaches, turn-internal semantic and pragmatic approaches, and interactional approaches. They argue that the “privilege” of MGT—accessing language attitudes discretely—is equally possible through such three-level analysis . At the first level, content-based approaches are characterised by the qualitative elements found in direct approaches. Narratives are analysed for their articulated meanings. Researchers usually try to discover themes and patterns in such utterances to supplement quantitative findings (Hoare 2001; Riagáin 1997; Garrett 2010).

The second level of analysis requires the researcher to pay more attention to specific linguistic and rhetorical features within the turn of each speaker. Such analysis may make use of linguistic categories, such as assertions, presuppositions, comparison, contrast, and others (Levinson 1983). Liebscher and Dailey-O’Cain (2009) illustrate the difference between the first and second levels of analysis by citing an excerpt from Dailey-O’Cain’s earlier study (1997). The second level of analysis helped to discover how confident the speakers were in the points they made. Another example comes from McEwan-Fujita’s study (2010) of adult Gaelic learners in the Western Isles of Scotland where the language situation is characterised by language shift and revitalisation. A “dance of disclosure” discursive pattern was found in that study. McEwan-Fujita concludes that by doing so, the speaker “subtly indexed a stance of negative affect” towards the intergenerational language shift, implying her strong and complex affects toward the Gaelic language.

The third level of analysis encompasses the first two levels, but looks beyond individual turns to take account of discourse and interaction features. It draws heavily on analytical notions in interactional sociolinguistics, conversation analysis, and linguistic anthropology, such as contextualisation (Gumperz 1992), footing (Goffman 1981), language preference (Auer 1995), language crossing (Rampton 1995), translanguaging (Gutiérrez et al. 2001), heteroglossia (Ivanov 1999), and language socialisation (Ochs 1996). Thanks to such analytical tools that this level of analysis can convincingly link the microanalysis of interactional texts to wider social and cultural processes, which results in a sense of groundedness that cannot be achieved by either correlational statistical analysis or purely content-based discursive study. The “discursive approach” of the current study refers to the third level of analysis. However, the first two levels are the basis on which the third level is built.

1.3 Previous Research on Chinese Dialects

Since the 1980s, researchers have been investigating language use and language attitudes concerning Chinese dialects in mainland China. The cognitivist approaches are the dominant, if not the only, approaches adopted in the published studies (for example, Zhou 2001; Miao and Li 2006; Kalmar et al. 1987). Most of the work involves Cantonese (and sometimes also Shanghainese) which is considered a major competitor for Putonghua and a hindrance to Putonghua promotion. Comparing these studies among themselves and against research elsewhere, we may note two phenomena concerning the language attitudes towards and the language use of Chinese dialects over the past decades.

The first notable phenomenon is a gradual but rapid language shift from multidialectalism to Putonghua-dominant diglossia . Take the case of Guangzhou for example. In the early 1990s, the government issued several directives to urge more rigorous efforts to promote Putonghua in Guangdong Province (such as Guangdong 1992), indicating dissatisfaction with the Putonghua promotion work there. It is against this backdrop that Leung (1993) conducted a survey in the Cantonese-speaking Pearl River Delta area (centred on Guangzhou) to investigate the language situation and people’s language attitudes. Sixty-two percent reported that the Chinese language class in primary school was conducted in Putonghua. While there is no data from earlier periods, I believe that, that was already a considerable “progress” due to 40 years of Putonghua promotion campaign, but it was seemingly not up to official standards. Moreover, the study showed that Cantonese dominated most situations of communication and socialisation, and it was a strong competitor with Putonghua in the public domains. There was no functional compartmentalisation in these domains at all. However, it is significant that people had realised by then the instrumental value of Putonghua, especially in achieving educational success. Putonghua was preferred as the preschool language at home, although Leung (1993) suggests that this was a “pure” attitudinal choice and hardly any parent executed the plan.

Thirteen years later, another study of secondary school students in Guangzhou (Tang 2006) demonstrated that the language-use pattern in Guangzhou has noticeably changed to a relatively clear diglossic pattern (see Table 3.1, the “*” marks the biggest figure in the same row.).

Table 3.1 Language-use pattern of secondary school students in Guangzhou
Full size table

Apart from the shift from Cantonese to Putonghua in the education and public domains, another notable change is that 19 % of the students reported speaking Putonghua as their first dialect. It indicates that a group of native speakers of (a variety of varieties of) the national standard language have emerged. They are different from previous native speakers of the Mandarin dialects. What they speak may be described as “a supralocal dialect ” (Britain 2010), which has significant implications for sociolinguistic theories but has not yet been researched in the context of Chinese. Moreover, while 19 % of the participants in Tang’s study (2006) claimed to be Putonghua-D1 speakers, another 6 % reported starting to learn Putonghua at home. This means the “attitudinal” choice of the parents in the early 1990s has become reality in the 2000s—parents start to teach Putonghua at home before their children go to school. Another study (Wang and Ladegaard 2008) conducted in 2002 also noted that 15 % of local parents use Putonghua at home with their children, but simply glossed over this phenomenon. Both the “native Putonghua speakers ” group and the use of Putonghua at home for preschool population become even more noticeable phenomena in the current study, which we will address in detail in later chapters.

The second phenomenon in previous language-attitude research on Chinese dialects is the nonconformity to classical contrasts and patterns consistently found elsewhere, namely, the status–solidarity dichotomy and the gender dimension of language attitudes.

The status–solidarity dichotomy is one of the most consistent findings in a “standard” vis-à-vis “nonstandard” language situation (Luhman 1990; Ladegaard 2000; Trudgill 1972; Preston 1999). The standard voices are often associated with traits conveying social status and mobility, such as education and ambition, while the non-standard voices are “rewarded” with group solidarity and social attractiveness, which is considered the “covert prestige ” of non-standard varieties. This dichotomy has been found true in bilingual settings too, such as the minority language setting of San Sebastian (Echeverria 2005) and Catalonia (Woolard and Gahng 1990). The solidarity value of non-dominant, stigmatised varieties and their potential for indexing personal authenticity (Woolard 1998) are considered the main reason that people continue speaking them, despite their low instrumental values. When we look at the studies of language attitudes towards Putonghua and Cantonese in Guangzhou, the picture is more complicated (Table 3.2).

Table 3.2 Findings on the solidarity and status of Putonghua and Cantonese in Guangzhou
Full size table

The three MGT studies listed in Table 3.2 span three decades at a 10-year interval approximately. Kalmar et al.’s (1987) study is a preliminary study with too small a sample to calculate the significance level. While this study seems to confirm that Putonghua had higher status than Cantonese, the configuration of the voice samples caused a major concern. Unlike the other two studies, the voice samples in Kalmar et al.’s study consist of standard and Cantonese-accented Putonghua. It is impossible to tell whether the participants recognised the voices as four native Cantonese speakers with different competence in Putonghua or as two native Cantonese and two Putonghua speakers. We have mentioned earlier that the uncertainty about what the participants are evaluating is one of the pitfalls of MGT studies. However, this problem is arguably more acute in this study because the ratings may be meant for Cantonese speakers with different education levels, rather than two distinct groups of Cantonese and Putonghua speakers. Maybe they just did not like a “mixed” speech style. While this may not change the conclusion that Putonghua proficiency was associated with social status and upward mobility, our interpretation of the study would be different.

The results of the second study (Zhou 2001) show variation from the existing findings on the status–solidarity dichotomy . Zhou’s (2001) study consists of two independent groups: Cantonese and Shanghainese. The Putonghua voices are given higher ratings by the Cantonese judges in both status and solidarity traits, while being rated higher by the Shanghainese judges only in solidarity traits. On one hand, as neither the Shanghainese nor Putonghua voices were rated significantly higher in any social status traits, it seems that the Shanghainese speakers consider the social prestige of Shanghainese and Putonghua as more or less equal. On the other hand, Putonghua voices were given high ratings in some solidarity traits by both the Shanghainese and Cantonese judges, which may imply that Putonghua has been associated with certain affective values in addition to its utilitarian roles. Zhou (2001) tentatively explains the nontraditional language-attitude pattern as being a result of decreasing social distance between Putonghua and Cantonese/Shanghai speakers in changing language situations. As Putonghua is increasingly used for “low” (in diglossic terms) functions, such as communication with family and friends, as reported in the Shanghainese data, it is not surprising that it starts to acquire solidarity values. What direction will the language situation in multilingual China go with Putonghua acquiring more affective and solidarity value? More in-depth and up-to-date research is necessary to answer the question.

Findings from Tang’s (2006) study contrasts with the previous findings sharply. The Cantonese guise received significantly higher ratings than the Putonghua guise in all traits. While the non-Cantonese student gave lower ratings to the Cantonese guise compared to their Cantonese peers, most of the ratings for the Cantonese guise were still higher than those for the Putonghua guise. While it is well-known that Cantonese is a strong regional dialect competing with Putonghua, this finding is still extraordinary. One problem with Tang’s study is that only two voice samples were used. The participants could have realised that they were listening to one bidialectal person speaking in Cantonese and Putonghua alternatively rather than listening to two speakers. If so, did the findings mean people find Cantonese speakers more attractive and competent when they speak Cantonese rather than Putonghua? Another possible explanation is that Cantonese is much more than a traditional “low variety” in a diglossic language situation. While it is not the national standard variety, it has overt (not just covert) prestige in the local community, which is why it is associated with social status traits in MGT studies. These are only theoretical speculations. More research is necessary before we can get the answers.

Concerning the gender dimension of language attitudes, previous language-attitude studies elsewhere have often found that women take the lead in using the standard varieties while men tend to report more use of or more positive attitudes towards the nonstandard local varieties (Trudgill 1972; Ladegaard 2000; Ladegaard and Bleses 2003; Labov 1972; Milroy 1987). It is suggested that women tend to speak the socially more prestigious linguistic varieties in order to elevate their social status. In contrast, men over-report their use of the non-standard variety to show positive attitudes, which will gain them solidarity with the local community—the “covert prestige” (Trudgill 1972) of the nonstandard varieties. Most language-attitude studies on Chinese dialects also investigate the gender dimension, but the findings are inconclusive or even contradictory.

Four studies are listed in Table 3.3, three of which have been discussed above in terms of the status–solidarity dichotomy . The additional study is a large-scale survey study of secondary school students’ language attitudes in Guangzhou (Wang and Ladegaard 2008), specifically focusing on the gender dimension.

Table 3.3 Gender differences found in language-attitude studies towards Chinese dialects
Full size table

In the three studies that find gender differences at significance level, the pattern seems to partly confirm previous findings. Taking a closer look, both the Cantonese and the Shanghainese female participants in Zhou’s (2001) study take the lead in giving higher grades to the Putonghua voices, but in terms of solidarity value rather than social status. This conforms to the gender pattern found in previous findings, namely, that female speakers tend to use socially powerful high varieties more often and have more positive attitudes towards them. In Wang and Ladegaard’s (2008) study, the gender difference in reported language use is much smaller than the LG–OG difference. There is no significant gender difference in language preferences , not even in the way they explain their language preferences. Tang’s (2001) study again provided surprising findings: females took the lead in giving higher ratings to the Cantonese guise in social status traits, which is diametrically different from previously found trends. Could the female participants be still doing the same thing—showing preference for socially prestigious varieties, only that it is Cantonese rather than Putonghua in this case? This seems a rather bold speculation. This ambivalent finding on the gender dimension certainly calls for further research.

Apart from findings on the status–solidarity dichotomy and the gender dimension, the potential influence of local cultural factors is worth noting in the study of language attitudes towards Chinese dialects. Tang (2006) found significant difference in the language attitudes towards Putonghua and Cantonese in different regions of Guangzhou, while Miao and Li (2006) found significant difference between Shenzhen and Guangzhou, two cities in the Pearl River Delta Region in Guangdong Province. Historical, cultural, social and demographic differences between these regions seem to have contributed to language-attitude differences. As a post hoc explanation, it is suggested that the social network structure in these communities could have played a role. However, the design of these studies prevents them from substantiating their claim empirically.

Above all, the conflicting trends found in language-attitude towards Chinese dialects in China suggest that language attitudes in this vast country are far from uniform. Discrepancies exist not only between different cities, different home-language groups, but also between different regions of the same city. Individuals in the same group are simultaneously members of differing numbers of other groups and it is possible to enact more than one identity and membership in specific situations, which can further complicate the explanation of their language attitudes.

Such multiplicities and complexities pose problems not only for cognitivist but also discursive approach es to language attitudes. Predominantly, quantitative studies tend to simplify or “tidy-up” the complexities in order to make it feasible. In contrast, qualitative studies, especially when informed by the discursive approaches, aim to reintroduce the complexities. They do not attempt to totalise the language attitudes of individuals or take into account all the complexities. Instead, the researchers modestly and critically examine just a finite number of complexities integratively according to the research purpose, and try to understand how these complexities contribute to the participants’ language-attitude construction activities. By doing so, a discursive study of language attitude may be able to shed light on some of the questions left unanswered by cognitivist approaches.

2 Conceptual Framework of the Current Book

2.1 Basic Assumptions

The current book takes a discursive approach that is informed by Alvesson’s approach to postmodernism (2002). The postmodernist view of language and discourse is the most important assumption underlying the current study. Postmodernists reject a naive “representational view of language” , and so do most social science researchers after the “Linguistic Turn” (Denzin and Lincoln 1998). Language is not seen as transparent or as a vehicle carrying meanings in the way that a train carries passengers. Rather, language consists not only of the meanings of utterances, but also of contexts, relationships, and subjectivities. It is claimed that language “precedes any experience of what is external to it, since experience gains its shape and intelligibility through language” (Alvesson 2002, p. 49). In other words, social realities are always mediated through language—an undeletable “text” (discourse). We are born “into on-going discourses that have a material and continuing presence” (Alvesson 2002), and our experience of the world is structured by the way in which discourses make us attend to it, and as we acquire such discourses more fluently, the discourses more properly speak through us. For such reason, discourse is always given primacy in social research informed by postmodernism, and a social constructionist position is taken when discussing the origin of meanings. In analysing such discourse, a restricted conversation analysis approach that attends only to the immediate interactional contexts rather than including wider social, historical processes is inadequate. The framework of language attitudes in the current study, which includes language ideology , contextualisation , heteroglossia , and language socialisation , is in order (to be delineated in the next section).

Another postmodernist tenet is the rejection of a fixed, coherent, unitary identity for the individual. Instead of subjects, postmodernists talk about subjectivities, or subject positions that the individual occupies, constructs, and shifts in given discursive practices (Silverman 2006; Cheek and Gough 2005; Alvesson 2002). As a result, identities, relationship, and membership should not be assumed but investigated and proved to be relevant. Such identities and membership are fluid and flexible. What is “found” in one context may be inconsistent with, or may even contradict what is “found” in other contexts. Therefore, I try not to “tidy-up” the contradictions and multiplicities to produce a totalising coherent account in this study.

2.2 Empirical Foci and Analytical Concepts

2.2.1 Contextualisation

The notion of contextualisation (Auer 1996; Gumperz 1992) differs from a traditional, narrower sense of “context” in which the relationship between “text” (or the focal events) and “context” is unidirectional. Instead, this relationship is considered reflexive and dialectic. Context is not a collection of material or social facts, but what is jointly and discursively constructed by the participants. Some contexts “might be stated by an objective on-looker … without looking at what takes place in it, but it may also include information not statable before the interaction begins, or independently of it” (Auer 1996, p. 2).

Contextualisation cues, such as prosody, paralinguistic signs, choice of varieties, registers or styles, choice of lexical forms, or formulaic expressions, do not have decontextualised referential meanings. It is through comparison with other possibilities that the meanings and functions of such cues are realised. For instance, in a study of linguistic practices of waiqi (foreign businesses) professionals in Beijing (Zhang 2006), it was found that these professionals selectively combine local and supra-local features in their speeches to construct a cosmopolitan Mandarin style , which differs from both Putonghua or the local Beijing dialect. One participant said in an interview: “…in the business world, when you reach a certain level, you’ll find, that is to say, Greater China integration. When speaking in Mandarin, you’ll meet Taiwanese, Hong Kongers, Singaporeans, and Shanghainese…” The words are underlined by the original author to indicate that they were said in English. The mixing of different languages in speeches was common among the waiqi professionals interviewed.

The second notable aspect of the excerpt is the use of the term “Mandarin”. Here, Mandarin loosely refers to the common speech of communication among Chinese professionals from Beijing, Shanghai, Taiwan, Hong Kong, Singapore, and other places. The varieties of “Mandarin” are different but mutually intelligible. More importantly, the varieties are named differently in Chinese by these people. For example, the variety used in mainland China is called Putonghua, the variety in Taiwan is called Guoyu , and varieties used in overseas countries are called Huayu (“Chinese (ethnicity) language”). Instead of using one of these labels that may be immediately indexical, the participant used the general English term “Mandarin”. This seems to be in line with the emphasis on “integration” in the “business world”. However, choosing the word Mandarin in a different situation may not contextualise the same sense of integration. Therefore, the analysts must establish the relevance of contextual information by demonstrating that it has been brought about in specific interactions. Such an emerging and reflexive approach to context means that the analyst’s shared knowledge with the participants, familiarity with local norms, awareness of potential sources of contexts, and sensitivity to strategies of contextualisation are crucial in reconstructing meanings of the target interactions.

2.2.2 Code-Switching, Translanguaging, and Heteroglossia

Code-switching is a common phenomenon in bilingual communities. It can be broadly defined as the use of two or more linguistic varieties in the same conversation or sentence by bilinguals (Gardner-Chloros 2009). There has been considerable controversy regarding the term “code-switching”. Some differentiate between “code-switching” (mixed use of different languages) and “style-shifting” (mixed use of dialects and registers) (Schilling-Estes 2008). Others question the artificiality of putting boundaries between different linguistic varieties, and propose using terms, such as “heteroglossia” and “translanguaging” (Blackledge and Creese 2010; Garcia 2009).

The notion of translanguaging differs from code-switching on the issue of boundedness of languages, but perhaps more significantly, the study of translanguaging focuses not on the speakers or the languages, but on the process of languaging—on-going language practices (Becker 1991). Garcia (2009) defines translanguaging as “multiple discursive practices in which bilinguals engage to make sense of their bilingual worlds” (p. 45). Translanguaging not only includes, but also goes beyond code-switching to include many other bilingual and multilingual language uses. Such multiple discursive practices are not regarded as appropriating the collection, combination or alternation between several monolingual semiotic systems, but as hybrid language use of one coherent multilingual system. Translanguaging describes specific language use as well as the overall discursive practices of both individuals and multilingual communities. The act of translanguaging is considered transformative in nature. It discursively constructs a space for bilinguals to bring together “different dimensions of their personal history, experience and environment, their attitude, belief and ideology… into one coordinated and meaningful performance, and making (to make) it into a lived experience” (Li 2011a, p. 1223). Hence moments of translanguaging provide valuable analytical foci for the current study.

Heteroglossia is similar to code-switching and translanguaging in many respects, but with different theoretical priorities. The notion of heteroglossia consists of two aspects (Ivanov 1999; Bailey 2007): (1) the simultaneous use of different kinds of forms or signs, and (2) the tension and conflicts between them in one text based on the sociohistorical associations they carry. The first aspect covers not only different languages and dialects, but also styles, registers and other variations in language use. This is because the researchers do not assume discrete boundaries between linguistic varieties on an a priori basis. Instead, “languages or codes can only be understood as distinct objects to the extent to which they are treated as such by social actors” (Bailey 2007, p. 258). Hence, heteroglossia accounts for the language practices of both multilinguals and monolinguals, as monolinguals’ language use is not necessarily monoglossic. The second aspect of heteroglossia points to the inherent social, historical associations of linguistic forms, which are too often forgotten or glossed over. Such associations are in fact part of the contextualisation mentioned earlier. The interpretation of the word “Mandarin” is an example of how social historical factors are embedded in local interactions and demonstrates how contrast may help reveal such underpinnings.

In short, the three terms subscribe to different views of languages and language-use practices and have different analytical foci. Using the three of them in a principled way lends more dimensions to the current study of language attitudes.

2.2.3 Language Preference

Code-switching becomes meaningful in a given speech community, not simply because of the contrast between varieties, but also because of the attitudes towards these varieties (Auer 1996). In other words, code-switching practice may index language attitudes in a reflexive manner. Auer (1988) differentiates between two types of code-switching: discourse-related and participant-related, which are not mutually exclusive. Discourse-related code-switching occurs when some aspects of the discourse are redefined or recontextualised, for example, when there are changes in participant constellation, topic, mode of interaction (such as between topical talk and side remarks) . In contrast, participant-related code-alternation happens as a result of the evaluation of participants’ attributes, such as linguistic competence, ideological considerations, and preference. Hence “language preference” was introduced as one source of motivation for code-switching. In this case, language preference does indicate favourableness but is used in a “more technical, conversation-analytic sense” as “an interactionally visible structure” (Auer 1995, p. 125). It describes the speaker’s tendency to use a particular linguistic variety for whatever reasons. In other words, language preference is the attempt to choose, and the action of choosing, a particular linguistic variety in conversation. Its meaning depends on the wider social, political and cultural contexts of the relevant interaction.

I would argue, however, the explanatory power of language preference as an interactional process and accomplishment goes beyond participant-related code-switching. Consider one instance of discourse-related code-alternation: changing the medium-of-interaction because of the switch between topical talk and side remarks. If one asks why topical talk and side remarks should be carried out in different languages, a possible answer may be institutional constraints. The medium for topical talk is possibly a standard, official variety, while the medium for side remarks may be a nonstandardised local variety. In such a case, the language preference is an institutional language preference performed by an individual abiding by it. In some cases, it seems impossible to distinguish between the larger societal, institutional preferences and the individual preferences, because every individual is continuously being socialised into the wider society, and learning to live with various language preferences is part of that socialisation. Such a state of interwovenness is precisely the reason that a broader, interactional notion of language preference may provide a valuable analytical perspective on language attitudes in interaction.

Language competence and ideological/political considerations are considered the two main factors accounting for people’s language preference (Torras and Gafaranga 2002; Auer 1995). Competence-related preference refers to speakers’ tendency to choose to speak the linguistic varieties in which they have greater competence and feel more comfortable speaking . Ideology-related preference refers to language choices motivated by episode-external factors, such as institutional norms, membership categorisation, sociolinguistic, and political allegiance (Torras and Gafaranga 2002) (the notion of language ideology will be addressed later in the section). Speakers can display, negotiate and resist certain language preferences in interactions to construct specific personae, group membership and alignment suitable for their purposes. Examples can be found in Torras and Gafaranga’s study (2002) of trilingual (Catalan, Castilian and English) service encounters in Barcelona, and in Cashman’s (2005) study of bilingual (English, Spanish) conversations during a card game at a senior citizens’ activity centre in a Latino community in Midwestern USA. While both studies provide detailed analyses of how participants negotiate language preferences, the researchers’ greater knowledge of the participants and the situation in the second study facilitated the interpretation of the motivations for their language preferences. As understanding why and how people negotiate language preferences is important in a study of language attitudes, I would argue that combining fine-grained interaction analysis with an ethnographic approach is essential.

2.2.4 Language Crossing

Language crossing refers to code-switching that involves using a language that is not generally considered to “belong” to the speaker (Rampton 1995). The nature of language crossing makes interactions containing such a phenomenon particularly salient in language attitudes. The existence of sharp social or ethnic boundaries between the languages in question is a prerequisite of the phenomenon, so that the act of language crossing involves a sense of “transgression” and moving across boundaries. If the languages are equally accessible and available to all participants, especially in ideological terms, then language crossing does not happen (Auer 2006) .

In Rampton’s (1995, 1998) study of Indian, Pakistani, African Caribbean, and Anglo descent adolescents in a British urban neighbourhood, the participants crossed into Punjabi, stylised Indian English, and Creole on different occasions. The adolescents were seemingly aware of the social stereotypes and sensitiveness associated with the different ethnic languages. It was observed that language crossing only happened at moments when “routine assumptions about ordinary life seemed to be temporarily relaxed, suspended, or jeopardised” (Rampton 1995, p. 500), such as in ritual abuse, self-talk, games, and performing arts. In doing so, the adolescents foregrounded and problematised ethnicity, momentarily destabilising the related social stereotypes and prejudices. In doing so, language attitudes that were taken for granted by the participants themselves and the wider community might also have been destabilised and problematised. Thus, language crossing may help bilinguals and multilinguals to constitute a shared multilingual community, in which people can learn to like and live with linguistic and social differences.

Having done his fieldwork in the school setting, Rampton (1998) observes that language crossing frequently foregrounded second language learning, because bilingual/multilingual competence is the prerequisite for effective language crossing. Hence, language crossing can be a potential opportunity for second language learning, both in terms of raising language awareness and practising the target languages. Using code-switching as pedagogy in bilingual and multilingual classrooms has been proposed by different scholars (Garcia 2009; Creese and Blackledge 2010; Walt et al. 2001; Martin 2005). In this respect, the notion of translanguaging is of relevance and significance. On one hand, since translanguaging is the norm rather than the exception in interpersonal communication in bilingual communities, it follows that teachers and students should learn and teach through translanguaging. In other words, languaging experiences at school should reflect rather than be alien to their daily languaging practices (Blackledge and Creese 2010). On the other hand, the “translanguaging space” created through and for translanguaging practices is a “safe space” for students to learn and use languages creatively and critically (Li 2011a, 2011b; Lytra and Martin 2010).

2.2.5 Language Ideology

A discursive study of language attitudes is closely related to the study of language ideologies . Both attitude and ideology are evaluative beliefs, but ideology is fundamentally social (Van Dijk 1998). The individual only participates in the ideology and partially shares it as the member of a specific group. Hence, one dimension in the relationship between ideology and attitude is the juxtaposition and connection between the group and the individual. Woolard and Schieffelin (1994, p. 62) capture this dimension in saying that “the intrapersonal attitude can be recast as a socially-derived intellectualized or behavioural ideology”. Another dimension implied in the same statement is that ideology often exists in the form of taken-for-granted, rationalised and common-sense “knowledge” rather than “subjective” evaluation (attitude). Yet, since it concerns what is normal, appropriate and desirable, the evaluative dimension is inherent, and thus ideology is always variable and contestable. As a specific type of ideology, language ideology is the social, evaluative belief system related to languages and language practices, which affects the individual and social treatment of languages and interpretation of linguistic behaviours (Mcgroarty 2010). Given its interest in socially shared conceptions, the research on language ideology is generally discourse-based, whether through quantitative approaches, such as corpus research, or qualitative discourse analysis (Mcgroarty 2010; Woolard and Schieffelin 1994; Kroskrity 2004).

Despite sharing many themes, however, language-ideology research has generally overlooked the large body of quantitative socialpsychological research on language attitudes since the 1960s (Mcgroarty 2010). One explanation may be that it is difficult to incorporate a reductionist notion of language attitude into the contextualised study of language ideology. When language attitude is mentioned, it is often simplified to mean positive or negative ratings obtained from a questionnaire or a MGT experiment . Taking a discursive approach , however, language attitude is inseparable from language ideology. Language attitude is the evaluation of beliefs about language and language practice, and language ideology constitutes those beliefs. The process of construction and expression of language attitudes is an active appropriation, contestation and reconstruction of socially shared language ideologies, based on one’s social position and relationships. In other words, by performing language attitudes, the individual is constantly participating in language ideology. Language attitude and language ideology are not two different levels of the delicious mille-feuille (cake), but different forms of egg white—depending on how one beats it, the egg white peaks in different forms and may be used for different purposes. They may be usefully distinguished to certain extent, but at some stages, they become indistinguishable. It is argued that language ideologies mediate between social structures and forms of talk (Mcgroarty 2010), and I would add that language attitudes mediate between language ideologies and forms of talk.

2.3 Research Design: Linguistic Ethnography

The purpose of the book is to investigate the perceptions of mother tongues and other tongues, regional and linguistic identities, in contemporary China where massive internal migration, modernisation and widespread multilingualism are changing the old meanings of language, identity and the relationship between the two. Language attitude is an analytical tool for us to slice through the issues to look for possible answers. The main research question guiding the investigation is: “How are language attitudes discursively and interactionally constructed in primary school communities, and with what linguistic and educational consequences?” The question breaks down to four more specific sub questions that structured the field investigation and data analysis.

  1. 1.

    What are the participants’ perceptions of the language situation in Guangzhou? How do they relate to dialects in their linguistic repertoires within such a sociolinguistic context?

  2. 2.

    What are the teachers’ usual language practices at school? How do they regard language policies in relation to their own language practices? What are their perceptions of the relationship between dialects and education?

  3. 3.

    What are the students’ usual language practices at home and school? How are they socialised to make such language choices? How do they justify their choices?

  4. 4.

    How do the participants negotiate language choices, index language attitudes and identities in interaction? What kinds of competencies and needs can be identified in the process? What are the implications for education in similar contexts?

Discourse is regarded as the “site” in which attitudes are constructed as well as the medium through which attitudes are mediated and represented. It calls for a discursively and interactionally based, linguistically sophisticated, open-ended research programme, which would not only look at the local communicative contexts but would also be sensitive to larger social and cultural influences. Linguistic ethnography (Maybin and Tusting 2011; Rampton 2007), which synthesises tools and insights from ethnography of communication (Hymes 1996), interactional sociolinguistics (Gumperz 1992, 2008), and ethnomethodological perspectives (Goffman 1986; Garfinkel 1967; Sacks and Jefferson 1992), seemed to be best placed to fulfil the needs.

Linguistic ethnography is a relatively recent methodological development that stems from linguistic anthropological studies in the USA (Duranti 1997; Wortham and Rymes 2003) and has taken shape mainly in the UK and other parts of Europe during the past decade (Maybin and Tusting 2011). It is closely related to the growth of sociolinguistics during the 1960s and the 1970s, which assumes a view of language different from that of structural linguistics or Chomskian cognitivism (Duranti 2003). Language is seen as constitutive of cultural experiences in various speech communities, which is a social phenomenon in itself that has to be studied in context. Situated language use rather than grammatical systems is the focus of the studies. The theoretical development of the ethnography of communication (Hymes 1964; Gumperz and Hymes 1972) struck a key note which still rings true today: “It is not linguistics, but ethnography—not language, but communication—which must provide the frame of reference within which the place of language in culture and society is to be described” (Hymes 1964, p. 3).

Linguistic ethnography is also linked to the revival of social constructionism and the growing influence of poststructuralist social theories (Duranti 2003) in the 1980s and the 1990s. Language is seen as an interactional achievement “saturated with indexical and ideological values” (Maybin and Tusting 2011, p. 516). There is a strong desire within this paradigm to connect macro-level analysis of social structure with the micro-level textual analysis of everyday communicative moments. Researchers attempt to capture the fluid, temporal negotiation of identities, institutions and communities by analysing linguistic practices, so as to shed light on the reproduction and transformation of persons, institutions and society at large (Maybin and Tusting 2011). Socially and historically laden notions, such as heteroglossia (Ivanov 1999) and language ideology (Woolard and Schieffelin 1994), are often invoked. When analysing discourse data, studies guided by this paradigm often conduct fine-grained, moment-to-moment analysis according to linguistic theories, while linking it to social-historical processes (e.g. Rampton 2006b).

The methodology of linguistic ethnography claims to “open up” linguistics through its ethnographic commitment to participation and contextualised description of social and cultural processes, while “tying down” ethnography by drawing on the well-developed, focused means in linguistics for studying communication. The two elements of linguistic ethnography may be seen as complementary, as well as pulling in different directions. For example, it is not easy to reconcile the often emic perspective in ethnography (Cohen and Court 2003; Hammersley and Atkinson 2007) and the etic framework in linguistics. The link between textual analysis of local communication and wider social and institutional processes could also be problematic. There is no quick remedy for these tensions. It all depends on the researchers’ choices of linguistic theories and ways of conducting ethnography, and more importantly, how they reconcile the potentially different ontological and epistemological assumptions underlying what they have chosen. Nevertheless, the methodology of linguistic ethnography may be adapted for the purpose of researching a wide range of topics involving the interaction between language, culture and society, especially those emphasising language-in-interaction .

2.4 Fieldwork and Data Collection

The current book draws data from the ethnographic fieldwork done in Guangzhou from August 2009 to August 2010 for my doctoral research project, when I was reading for a Ph.D. in Education at the University of Cambridge. The fieldwork was conducted in two primary school communities in Guangzhou, one in an urban district and the other in a suburban district. Over a period of two semesters (September 2009–January 2010, March 2010–June 2010), I regularly stayed at one school for a whole week and went to the other school in the following week. During my stay at the school, I divided my time between the classroom, the teacher’s office and the playground, and the major school events, extracurricular activities and staff meetings that I was allowed to attend. After school, I paid home visits to key student informants and talked to their families. The original project also targeted language attitudes but also included other issues such as multilingual competence (Li 2011b) and language teaching. This book focuses on the data on language attitudes and identities.

Reasons for choosing these two primary school communities are threefold. First, previous studies on language attitudes in Guangzhou indicate systematic attitudinal differences between different regions: the traditional city centre, the new city centre and the suburban districts (Tang 2006). Therefore, the factor of region and social network was taken into account in the sampling of the current study. The urban school, Grand-Estate School, situated in the new city centre, while the Sandwood School is located in a suburban district.

Second, primary school students were chosen as it seems that attitudes formed in the early periods of life tend to be more persistent (Augoustinos et al. 2006). A recent language attitude study (Bokhorst-Heng and Caleon 2009) note that Singaporean primary school children’s language attitudes towards their “mother tongues” and code-switching practices are equally positive. The finding coincides with the trend noted in recent linguistic ethnographic studies (Rampton 2006a; Li 2011a; Blackledge et al. 2008) that young bilinguals discursively create a translanguaging space, in which they are comfortable with their flexible bilingual identities distinct from the more fixed and absolutist linguistic identities of the older generations. It is worth investigating whether similar phenomena also exist in China.

The sampling further narrowed down to Year 5 primary school students (as of the 2009–2010 school year), as they belong to the first cohort of students who received their formal education after the implementation of the Language Law 2001 , which established the status of Putonghua. No study so far has investigated how school experience with dialects may have changed since then and with what consequences for language attitudes and language education. Moreover, the primary school stage in China is an important language socialisation period which accustoms children to acquiring basic literacy skills in SMC(Standard Modern Chinese). These skills are very much taken for granted beyond this period. The current period of primary school education in China typically lasts for 6 years. Year 5 students are not yet under direct pressure from the secondary school entry examinations but are supposed to have acquired sufficient skills to express themselves in Putonghua and through written compositions (Ministry of Education 2000).

An informal survey was conducted in both schools for the purpose of socialising with the students rather than generating structured statistical data. The survey included questions concerning their parents’ places of origin, their place of birth, length of stay in Guangzhou, first dialect, other dialects that they spoke, and their self-assessment of Putonghua and Cantonese proficiency. It was usually done in the classroom during breaks or after school. No effort was made to speak to each student individually. The students could see others’ answers on the form if they wished. As a result, the process of completing the survey sometimes became a group interaction or competition. Some insisted on awarding themselves higher grades for linguistic proficiency, claiming that they were better speakers than someone else was. Some were questioned by others about their self-evaluation and it was the first time for some to find out about their classmates’ multidialectal background and competence.

Each key student participant (see their profile in Appendix 1) was interviewed individually at least twice. The first interview was loosely structured and based on the following themes related to their ethnolinguistic background, language use, and language learning experiences. The subsequent interviews were based on language tasks that I assigned to them or comments on what happened in school or at home. Interviews with parents were not always possible since some worked until late. Whenever feasible, the interviews were conducted at home with parents, or in some cases, the grandparents. The focus was on what role they had played in the language socialisation of their children. Interviews with teachers (see their profile in Appendix 2) were conducted individually in school, usually in their office when no one else was present. The focus was on their learning experience, language use, perception of dialects, and the current language policy.

Apart from individual interviews, I also conducted multiple focus group discussions with students in both schools. I engaged the members of the focus group as “ethnographers” of their own language living by assigning them several language-related tasks (White et al. 2008; Dressman 2006; Heath 1983). The tasks were designed to be “experimental” and educational, directing the students’ attention to sociolinguistic phenomena that they might not have noticed otherwise. There was usually a two-to-three-week gap between the two sessions. The participants were interviewed individually about their experiences in carrying out the tasks before the focus group sessions. During the sessions, I often brought up similar topics as prompts for discussion.

I also collected various forms of written texts and documents during the fieldwork, as school is one of the institutional settings where self-documentation and the consumption of documents is a daily matter (Hammersley and Atkinson 2007). These documents were not taken as facts but as representations that had been constructed for certain audiences and certain purposes. While the focus of the current study is on face-to-face interaction and spoken discourse, the documentary materials, especially policy-related documents helped to situate the immediate contexts of the school within the wider social and political contexts and discourse (Hornberger and Johnson 2007).

2.5 The Roles of the Ethnographer

As this is a linguistic ethnographic study, my autobiographical and sociolinguistic profile has significant impact on what I have access to, what I am able to see in the field, and how I represent what I see in the research text. In Chap. 11 we will have a more detailed discussion on the impact of reflexivity and positionality on this research, which will be illustrated by specific field examples. In this section, I will briefly talk about myself and explain how some of my social identities influenced my access to field sites and relationships.

In relation to the Sandwood community, I am a member by traditional standard. The families of both my parents’ belong to one of those major residential lineages in the township. My parents moved away from the villages when I was 2 years old. Over the years, I have maintained contact with the community through my extended family. My grandfather was once the head of River Village in Sandwood Township, and he helped secure access to the school for me. While I had heard about the school from time to time, my first visit there was during the prepilot study. The headmaster, who was a native Cantonese speaker but not a member of the community, seemed very welcoming to me, saying that it was an honour to have a “celebrity” from the village to stay at the school. This “theme of celebrity” recurred at the beginning of the fieldwork, as some teachers and parents knew about me through my relatives. The benefit was that it was relatively easier to start a conversation and establish trust. The disadvantage was that very much was taken for granted and not explained to me. Hence from time to time, I had to emphasise, if not exaggerate, the fact that I knew very little about village life since I had not grown up there. In this respect, my Cantonese accent worked to my advantage. Since I had not grown up in the township, I had forgotten how to speak with the local accent and distinct local repertoires. Miss Chun once commented, “You don’t talk like a local, but I know because they told me you are from Sandwood”. So my way of talking set me apart from the local community because it did not match the local standard.

On the other hand, access to the Grand-Estate School was gained through my relatives’ business network, and I was introduced to the vice headmistress, Miss Gao, over dinner. The early preparations, such as arranging a time to meet and deciding on which class to follow, were considerably more difficult than at Sandwood School. Moreover, Miss Gao seemingly wanted me to maintain a low profile, having only notified the head teacher of the class that I would follow. Others only knew that there was a new “intern” in the school. I was never invited to any staff meeting or the like. I did not have my own desk in the classroom or the teacher’s office at Grand-Estate School, so my status as an outside observer was more physically conspicuous there. In contrast, in Sandwood School, I had exactly the same desk and chair in the office as other teachers did, the only difference being that I used my own computer. In the classroom I sat where the students were seated. These physical arrangements affected my participation not so much in the sense of what activities I took part in. Rather, what matters is the “sides” or “groups” which I seemed to be part of, in other words, being an outsider, or taking sides among the insiders.

The readers may decide, based on the information given above, what my positions are when I write certain descriptions, analysis, and comments, and how those positions may have influenced the text and the readers’ interpretation of the text.