Socio-stylistic aspects of linguistic variation: schooling and monitoring effects

Alveolar stops following palatal glide in Brazilian Portuguese are investigated, with focus on constraints for the choice between occlusive or affricate variants. Data was collected in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria, from a sample belonging to the ‘Falares Sergipanos’ database, and submitted to GoldVarbX analysis. Seven hundred seventy-five variable contexts were identified, in which 550 occurrences corresponded to the occlusive variant. Results show that the phenomenon is constrained by linguistic, social and stylistic factors, and highlight the importance of a detailed control of the context in the analysis of variable phenomena at the phonetic-phonological level.


Introduction 1
Variable realization of alveolar stops following palatal glide in Brazilian Portuguese (biscoito ~ biscoitʃo, endoidar ~ endoidʒar) is constrainded by linguistic, social and stylistic factors and data from a sample collected by a community of practice (in contrast to community of speech) methodology can help explain this variation.
The following sections resume studies on the process of palatalization of alveolar stops following palatal glide, with attention to linguistic varieties of the surroundings of the community under analysis (regions of the States of Bahia, Sergipe, and Alagoas); community and the procedures for data 1 The results presented refer to the development of the projects entitled Mulheres, linguagem e poder: estudos de gênero na sociolinguística brasileira [Women, Language and Power: Gender Studies in Brazilian Sociolinguistics] (CNPq 32/2012), Língua, identidade e cultura: representações sociolinguísticas em Sergipe [Language, Identity and Culture: Sociolinguistic Representations in Sergipe] (CNPq PQ/2011), and uses the database constituted in the scope of the project entitled Da expressividade da língua ao mal na literatura: base de pesquisas interinstitucionais do PPGL/UFS (CAPES/FAPITEC/SE 06/2012) [From the Expressiveness of the Language to the Evil in the Literature: Interinstitutional Researches Base of the PPGL/UFS].collection are described.The results presented evidence that the palatalization phenomenon has a strong interaction with linguistic, social and stylistic factors, emphasizing the importance of a careful control of the context, even in the analysis of a phenomenon at the phonetic-phonological level.

Palatalization in Brazilian Portuguese
The palatalization of alveolar stops is a process that occurs in different phonological environments, whose distributional results indicate social and dialectic boundaries in the Portuguese spoken in Brazil.Results of several variationist and dialectological studies are generalized as both ownership and directional indexes of language variation.The affricate realization of the alveolar stops /t/ and /d/ preceded by palatal glide, the 'Bahia's affricates', in the words of Mota (2002), is a feature associated with the talk of the Northeastern region of Brazil as a whole, but it is more expressive in Sergipe, Alagoas and the interior of Bahia.The distribution of frequency considering sociodemographic profiles of broad social strata suggests that the feature is a stereotype, in the terms of Labov (1972), because is predominant in the group of individuals with lower level of education and of older age2 .Moreover, it is a feature perceived consciously, and an object of social evaluation: Palatalized realizations before /i/ in the Northeast are generally considered as prestigious variants, unlike realizations after palatal glide, identified as stereotypes, as can be seen in comments from speakers, like "O pessoal do interior e de poca cultura fala no petcho, o sertanejo costuma falar petcho", [Countryside people with little culture says in the petcho (meaning udders), countrymen say petcho], record of a male speaker, college student […] from Aracaju, after the utterance ['pejtu], in response to the question "Where is milk stored in a cow?" (QSL 80) (MOTA, 2008, p. 6).
The variable realization of alveolar stops in Brazilian Portuguese has emerged as an identity indication, such as shown by Pagotto (2004) and Battisti et al (2007) studies, supporting the thesis that there are social reasons for the choice of forms in the system, and conveying social meanings locally founded.Eckert (2012) proposes variation analysis in the scope of communities of practice, setting up a line of work known in academic circles as the third wave of sociolinguistic studies3.The concept of community of practice comes from the educational universe (WENGER, 1998).Communities of practice are groups of individuals in interaction, all engaged around a focus of interest that leads them to seek learning and the improvement of skills.In our daily life, we walk through different communities of practice -the community of practice of work, of leisure, of study etc.Labov (1972), in the first wave of sociolinguistic studies, showed that linguistic variation was correlated, in large cities, to factors such as social class, sex, ethnicity, age and attention to speech, which also resulted in the development of methods that were suitable for large-scale research to reveal the systematic relationship between linguistic variation and social dynamics.However, this approach does not account for the actual use in a situation of social practice (ECKERT, 2012).
There are some studies on 'Bahia's affricates' in Aracaju, capital of Sergipe State4 ; one is Mota (2008a), which analyzed data from eight speakers distributed by sex, fitting into two schooling ranges (incomplete elementary school and incomplete higher education) and into two age groups (18-30 and 50-65 years old).The methodology of data collection follows the dialectological pattern of the Atlas Linguístico do Brasil [Linguistic Atlas of Brazil] project (ALiB) for the phonetic-phonological and lexical-semantic questionnaires, and themes for semi-structured speeches; given the low recurrence of words that allow the context of the phenomenon, occurrences of the morphosyntactic questionnaire were also included (Table 1).These results support the attribution of the stereotype label to the affricate realization of /t/ and /d/ preceded by palatal glide: the profile of speakers that is more associated with the variant includes being male, less educated and older.
In another study, Mota (2008b) observes the influence of the way the collection was carried out, distinguishing data from the questionnaires -in which the speaker is led punctually and directly to a word in response, configuring a greater monitoring, -and data collected at the end of the application of the questionnaire, when the speaker is led to talk about predefined topics such as milestones of his or her life, work, TV shows etc.The results show difference in behavior, though subtle, given the sample size, suggesting that the feature is not so stigmatized, as careful articulation, expected in answers to the questionnaire, points to a relative weight of 0.47, whereas a less monitored context presents a relative weight of 0.58 (Table 2).Mota (2008b) makes a remark controlling the type of lexical item: the word muito computes 68 affricate occurrences, of which only three are found in the more monitored context (in response to a questionnaire item), which does not allow the correlation between degrees of formality and affricate realization suggested by the author; however, the high frequency of the item is an indicative that deserves a more specific control.
And in a variationist sociolinguistics approach, Souza Neto (2008) study tooks as sample the speech of thirty-six Aracaju speakers stratified by sex, age and family income in minimum wages, in order to assess the quantum of the influence of the preceding phonological context and of the following phonological context, of syllable stressing and of social factors in the variation of alveolar/affricate stops (in both contexts: after /j/ and before /i/).About data collection methodology, the author notes "[…] the interviews took place in very different ways, sometimes with little or no interference from the interviewer, sometimes with long dialogues […]" (SOUZA NETO, 2008, p. 41), and lasted from one minute and twenty seconds, and one hundred twentyseven minutes.For reasons of time, the researcher took the methodological decision of considering only a maximum of 17 minutes for each speaker.
While the study by Mota (2008a) was conducted with 316 occurrences of alveolar stops, without distinguishing voicing, Souza Neto (2008) controlled separately the occurrences regarding the phoneme (/t/ and /d/) following palatal glide.For /d/, 130 occurrences were identified, of which 97 realizations were not affricated (Table 3).The variationist analysis excluded speakers with categorical realization, and the age and income group variables were statistically significant, with converging results in terms of the stigma profile outlined in the study by Mota (2008a).With respect to /t/, 384 occurrences were identified, of which 218 corresponded to the occlusive variant, corroborating the results of Mota (2008a) relative to the stigmatized profile of the affricate variant in the linguistic environment in question; it is worth highlighting that the variable sex/gender -whose results often give clues as to the social evaluation of the phenomenon -has been selected (Table 4).In the study by Souza Neto ( 2008), the schooling variable did not prove significant (it is possible that, in the analysis conducted, income range overlapped level of education, but there is insufficient information to test this hypothesis), and the controlled linguistic variable -stress of the syllable in which the phenomenon occurs -did not prove significant5 .It is also worth highlighting the fact that the average length of the recordings is relatively low, with seventeen minutes of duration, which may be sufficient to ensure the number of occurrences in the sample, but does not allow, for example, the assessment of how vernacular -according to Labov (1972) and depending on less/more monitoringthe occurrences are.Cristófaro-Silva et al. (2012) analyzed data from Mota and Rollemberg (1997), evidencing that the conditioning segment, the palatal glide /j/, may not occur with realizations of [dodʒu] type, suggesting a process of interaction between palatalization and other processes, such as the deletion of the palatal glide of the falling diphthong (though it is a retention context in other phenomena, in palatalization processes there is a tendency to deletion).Such a suggestion the study reiterates the core premise of the variationist analysis that linguistic factors can influence the variation process, motivating a more refined control, if not in terms of gradualness of the realization of the phenomenonwhich is not always possible due to the acoustic quality of the field data -at least concerning other factors, besides syllable stressing.Considering these achievements on the palatalization process of alveolar stops following palatal glide in Brazilian Portuguese, this work has as objective to discuss a) whether, at a phoneticphonological level of analysis, formality/informality6 is relevant in the conditioning of a variable phenomenon and, if so, how it is possible to measure this gradient (in other words, to what extent the effects of the observer's paradox interfere with the phonetic-phonological level); and b), since the affricate realization in that environment is a complex phenomenon, for interacting with other processes (CRISTÓFARO-SILVA et al., 2012), whether internal factors (tone of the syllable where palatalization occurs, word class) have statistical relevance to the recurrence of the phenomenon7 .
To reach these objectives, the variation in the occlusive realizations /t/ and /d/ preceded by palatal glide is analyzed in a data sample constituted in a community of practice (cf. FREITAG et al., 2012).Aligned with the 3 rd wave of sociolinguistic studies (ECKERT, 2012), the communities of practice approach allows examining an individual's role in variation and linguistic change phenomena, a role that is not defined in an a priori manner, due to broad sociodemographic factors, but that rather depends on relationships established within these social spaces.In communities of practice, leadership, for example, can give to the leader the power to propose innovations, even linguistic ones, as the group led legitimizes and follows him or her, complying with the behaviors he or she adopts.It is also in communities of practice that it can be observed, through ethnographic study, how relations between use of language, style and identity construction take place for each individual.In this approach model, it is possible to investigate more carefully social and stylistic motivations for the choice of sociolinguistic forms, allowing evidencing how language variants can convey locally based social meanings (LABOV, 1972;ECKERT, 2001;2012).
The community of practice of 'Praesidium Mãe da Divina Graça da Legião de Maria' While data of the analysis by Mota (2008a and b) and Souza Neto (2008) come from speakers from urban areas and from the capital, selected according to broad sociodemographic stratification criteria, this work analyzes data coming from a sample of community of practice, belonging to the 'Falares Sergipanos' database (FREITAG, 2013), still in development stage, which aims to store samples of speech communities and communities of practice from six municipalities of Sergipe State8 .
The community of practice analyzed has a religious nature, Praesidium Mãe da Divina Graça da Legião de Maria (Catholic), and is located in the countryside, in Açuzinho village, one of the more than 100 villages in the municipality of Lagarto, in the south-central region of the state of Sergipe (cf. FREITAG et al., 2014).
Legião de Maria is a Catholic association that emerged in Ireland in 1921, having as a goal the glory of God through the sanctification of its members by prayer and active co-operation, under the direction of an ecclesiastical authority in the work of Mary and of the church.Legião de Maria is organized into hierarchical councils: the Praesidia serve the Curia, which serve the Comitia, which are linked to the Regia, which, in turn, provide service to the Senatus, which serve the Concilia.The Latin names refer to the Ancient Rome, since the institution mirrors itself in the organization of the Roman army.In Açuzinho village, Praesidium Rosa Mística, linked to Legião de Maria, in Santa Rita de Cássia church, Açuzinho village -Santa Luzia Colônia Treze parish -Lagarto, Sergipe State, was established on October 2, 1988 with 26 members, of which four were officers -president, vice president, secretary and treasurer.Speaker Edv, already engaged with religious movements, was the first president of the group and took over this function for 10 years; at the time, he was 47 years old and was a farmer.A year later, the group split up, forming the Praesidium Mãe da Divina Graça da Legião de Maria -the community of practice under analysis -currently consisting of 13 members who meet twice a week.In the first meeting, on Mondays, the schedule for volunteer work (visits to the sick, to bereaved families, Bible circles, prayer chains etc.) to be held in the second meeting, scheduled over the week, is defined.
In the Monday meeting, the president of the group performs a spiritual reading, then the secretary reads the minutes; subsequently, visitors and members are welcomed, attendance of all members is taken, and the oral report of each member on the progress of the previous week's work is done; if he or she has done it, he or she should say how it has been done, and if he or she has not done it, he or she should justify why.After the report, the Catena is prayed, which is the hymn of Mary; then there is the lecture, an explanation of the spiritual reading that is done in the beginning of the meeting, and that can be done by any group member appointed by the president, with space for all members to discuss their understanding of the spiritual reading.After that, the treasurer reads the treasury cash report.Then, there is the distribution of the volunteer work for that week, such as consecrating the Eucharist along with the sick, visiting nursing homes, hospitals, prisons, bereaved families etc.The meeting is closed with the final prayer.
In this community, members have a direct relationship with the officers.Although subordinate, they opine on the distribution of work, evaluate the minutes and interact in discussions of the readings.In addition, the environment of the meeting presents a variety of styles: from formal speech, uttered by officers and members, to reading and prayer.
Variety is also found in the group configuration: the president, speaker Ire, 45 years old, plays the lead role in the hierarchy: she is the one who starts and ends the meeting, establishes the order of readings, chooses who will speak at a certain moment, defines where the volunteer work will be carried out, etc.The vice-president, RaA, 64 years old, can substitute the president when necessary, has the role of welcoming visitors and/or guests and take the members attendance.The secretary, RaN, 38 years old, writes and reads the minutes of the meeting and of the work done in the previous week.The treasurer, Vel, 43 years old, is responsible for taking care of the finances, collecting money for expenses such as the church's decoration, xerox copies and printing, group trips for meetings with other councils, registering the values collected and accounting for how much has been spent.
At 22, speaker Mon is the youngest member of this community: a college student, she moves through other communities of practice made up of members of her age group, has greater contact with the urban environment, and, for being a teacher, tends to monitor more her linguistic use.Speaker Edv, in turn, is the only man in the community of practice and the oldest member, being 71 years old.Throughout his life, he worked in the fields to provide for his family, studied until the second grade of elementary school only, and walks with people of his age group, with little contact with the urban area.This heterogeneous and hierarchical composition is a characteristic of communities of practice, because every group of people coming together with a common purpose requires someone that is always ahead to make decisions and stand up so the community can progress before the intended goals.A more realistic composition of society than the homogenized stratification of speech communities.
For data collection, the recordings of four meetings were analyzed, each about 90 minutes long, and seven interviews lasting 50-60 minutes each, totaling approximately 10 hours of recording.The selection criterion for the interviews was speakers being active members in the community (Table 5), manifesting orally in the meetings 9 .Both data collection strategies allow investigating stylistic effects in palatalization realization: the sociolinguistic interview script prioritizes the emergence of the vernacular, with interaction conducted towards a greater focus on subjective topics such as life experience narratives and events reports, which tends to minimize self-monitoring in speech.In addition, the interviews were conducted in the speakers' houses.The recordings of the meetings, in turn, prioritize the speech of the speakers concerning the exercise of their institutional roles and, consequently, present a greater speech self-monitoring.
9 Complying with the guidelines for research involving humans, normatized and regulated in Brazil by Resolution 196/96, the project Falares Sergipanos was submitted for approval of the Research Ethics Committee -REC of the Federal University of Sergipe, which is linked to the National System of Information on Research Ethics -SISNEP, being granted a certificate of compliance with the ethical research guidelines No 0386.0.107.000-11.The recordings of the meetings and the conduction of interviews were a responsibility of Cristiane Conceição de Santana (PIBIC/CNPq-AFE) and Thaís Regina Conceição Andrade (PIBIC/COPES/UFS).The first researcher resides in the municipality, and her grandmother is a former member of this community of practice, which facilitated the contact and minimize the effects of the observer's paradox.In a parallel way, interviews were conducted aimed at collecting information on the constitution of the community, in addition to documental investigation in minutes and record books of the community; said investigation allowed outlining the profile of the community, enabling a historical rescue, and can be seen in Freitag et al. (2014).Palatalization in the community of practice of Mãe da Divina Graça The ten hours of recording of meetings and sociolinguistic interviews were analyzed by means of impressionistic hearing, which implies a slight reduction in the gradient of the variation that could be identified if data was manipulated by software.In practice, this means the computation of affricate and occlusive discrete realizations; even though there is a gradient between them (alveolar occlusive realizations from dental occlusive realizations are not distinguished, for example).
Initially, a ternary dependent variable was expected, with affricate alveolar-palatal contexts following palatal glide /dojdʒo/ ~ /pejtʃo/, monophthongal alveolopalatal affricate (subjacent palatal glide) /dodʒo/ ~ /petʃo/ and alveolar stop /dojdo/ ~ /pejto/.However, the affricate variant with explicit palatal glide was not perceived in the impressionistic hearing analysis.A total of 775 occurrences were found, correlated to social, stylistic and linguistic variables and subjected to statistical treatment by means of the GoldVarbX package (SANKOFF et al., 2005), of which 225 refer to the affricate variant.For the presentation of results, alveolar occlusive realization is taked as rule application (Table 6) 10 .To capture the sensitivity of the phenomenon to the stylistic dimension, the results are analyzed according to the type of data collection: if in sociolinguistics interview, conducted individually and in the interviewee's house, or if in a meeting, in a more formal environment, with different levels: formal speech, reading of texts, and prayer.
In reading and prayer contexts, the occurrence of occlusive variant was categorical, while there was no difference in percentage terms between the 10 The studies of Mota (2008a and b) and the study of Souza Neto (2008) take the very same geographical area, Aracaju, Sergipe, but differ when it comes to the rule scope: whereas Mota (2008a and b) considers the affricate variant as application value, Souza Neto (2008) considers the occlusive realization.The absolute results are converging: in both works, the occlusive variant presents a greater frequency than that of the affricate variant.Mota (2008a) highlights the stigma related to the affricate variant, which justifies her choice for the application of the rule, pattern kept in Mota (2008b). Souza Neto (2008) and Santos (2011) resort to the productivity criterion to define the application value of the rule, choice that is adopted in this work too.
interview situation (less formal) and the meeting speech situation (more formal).
The initial hypothesis that the distribution of variants would be related to the formality /informality of the situation, as shown by Mota (2008b), is not confirmed with global data.Categorical data -reading and prayer -were removed from the analysis, leaving 705 occurrences.The 'degree' (non-diminutive /ʒeito/; /diɾetʃo/ ~ diminutive /ʒetʃiɲo/; /diɾejtiɲo/), the position in the word (stressed syllable /pɾefetʃura/; /cujdado/ ~ unstressed syllable /pɾefetjo/; /dojda/), the word class (verb/apɾovejta/;/detʃava/ ~ numeral / otʃo/; /dezojto/; /otʃenta/ ~ name /lejtuɾa/; /letʃuɾa /; ~ "muito") and voicing (/t/ /nojte /; / notʃe/ ~ /d/ dojdo /; dodʒo/ ) were selected in this order, in the round already without the categorical data (Table 7).The overall data analysis shows that: a) when in stressed syllable, there is a tendency to occlusive realization, with a relative weight of 0.67; b) the occlusive environment /d/ was not productive in the sample, with approximately 3% of the total data; c) the lexical items plays a significant role, as verbs disfavor the occlusive realization, with a relative weight of 0.27; d) the item 'muito' and its variations account for more than half of all occurrences; e) there is overlap between the controlled linguistic variables (only names can be diminutive; all occurrences of 'muito' have the realization of the variable in the unstressed syllable), which compromises the orthogonality of the factors.Such results motivate a new dimensioning of analysis, with the exclusion of occlusive /d/ data and the analysis in two moments: only with the item 'muito' and with the other data.
In the analysis of the behavior of the variable in all 360 occurrences of 'muito' in the sample, only the individual factor proves significant.It is worth highlighting that, when controlled, the latter was Acta Scientiarum.Language and Culture Maringá, v. 37, n. 2, p. 127-136, Apr.-June, 2015 selected as statistically significant in the first place, in all rounds.Social factors will be addressed later.
The selection orujder of the factors in the sample with the other data (300 occurrences of names, numerals and verbs) followed the selection order with all data: diminutive word, segment position and lexical item (Table 8).The results show that: a) the diminutive disfavors the occlusive realization, with a relative weight of 0.11; b) stressing, due to salience, operates in the conditioning of the occlusive variant; and c) there seems to be a tendency to exemplary phonological representation, not of the numeral class, but of the oito-radical.
Statistical data suggests a variable bias: while the result for position points that occlusive realization is disfavored in unstressed syllables (with a relative weight of 0.36), the result for control of the 'degree' points that diminutive words disfavor occlusive realization (with a relative weight of 0.11); in all diminutives, the variable occurs in the stressed position (/ʒetʃiɲo/; /diɾejtiɲo/).A statistical round was carried out then, with the exclusion of data on diminutive.The selection order of linguistic factors remained -position and word class -, with subtle changes in relative weights, reinforcing the correlation between stressing and non-palatalized form (Table 9).In this round, however, the type of collection was selected as a statistically significant factor, with 0.72 of rule application (occlusive variant) for meeting speeches -more formal and more subject to self-monitoring -and 0.44 of rule application in the sociolinguistic interview, a more informal and less monitored context.
The interaction between the type of collection and the 'degree' of the name is evidenced by cross-tabulation: diminutive occurrences are polarized in the collection in the form of sociolinguistic interview -a more enabling environment for less formal stylistic uses, as diminutives are (Table 10).In the formal speech situation, collected during the meeting, when subjects of religious nature are addressed, there are only four occurrences of diminutive; however, out of the four occurrences, three have the affricate variant, indicating the possibility of existing an assimilatory process of the palatal nasal feature that is characteristic of the diminutive in Portuguese, deserving a more accurate research with a greater volume of data.The results in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria point to evidences that the control of formality/informality, a stylistic dimension, is relevant in conditioning the affricate realization of alveolar stops preceded by palatal glide, as it can interact, like in the diminutive effect, with linguistic behaviors.
Given that the realization of alveolo-palatal affricate following palatal glide is a complex phenomenon, it is possible that there is evidence of the interaction between processes, such as the tendency to the occlusive realization of the variable in the stressed syllable, different behavior depending on word classes (tendency to affricates in verbs and occlusives in names), and also the effect of the diminutive (marked by nasal palatal or nasalized palatal glide).The statistical relevance of these variables also interacts with social factors, as shown below.Social relevance of a phonological feature in the community of practice of Mãe da Divina Graça In variationist analyzes that control individuals as a factor, if there is indicative of social relevance of the phenomenon, this is usually the first social variable to be selected as significative statistically, since this factor overlaps other controlled social factors such as sex/gender, education, age group etc. (FREITAG, 2005).Individual results tend to reflect a tendency to larger grouping; for this reason, in communities of practice, the control of the individual as a factor is essential, as the sample distribution is not homogenized, and there is an even stronger interaction between social factors.
The distribution of all sociodemographic factors in the community is heterogeneous, with accumulation of social roles: they are 7 women and one man; a young girl, three individuals in the 40-50 years old age group, and three over 60 years old; one of them studied until the 2 nd grade, two until the 4 th grade, one until the 8 th grade, one until school, and three have higher education, four officers and four members.The only man is the only one that has only the 2 nd grade of elementary school and is the oldest one of the group.As for income, the elderly (over 60 years old) are INSS (Brazilian social security) pensioners, receiving the minimum value; the other speakers, who are still active in the labor market for their ages, are either municipal public servants (working as teachers or administrative assistants) or 'stay-at-home' women.The only feature in the community that allows the grouping of behavior patterns is schooling, a complex variable involving both the role of the school and its coercive effect on non-standard variants, as well as the access to cultural goods and professional growth, with the expansion of social networks and contact with writing.
In the community, there is one single male speaker (Edv).Speakers Ire, Ran and Mau present a tendency to occlusive realization in their speech and are the speakers with higher education (degree in Education and Geography); RaA and Edv, in turn, with relative weights that indicate the disadvantage of the occlusive variant, have the lowest level of education in the community: 4 th and 2 nd grades of elementary school (Table 11).Speaker Vel attended until 8 th grade and presents a neutral statistical behavior, at times inclined to occlusive realization (in the word 'muito'), at times inclined to affricate realization (other lexical items).Speaker Mon, a teacher, with higher education, has a categorical behavior (only one occurrence of affricate variant, in the word 'muito', in sociolinguistic interview situation).
Considering the data on affricates in occlusive environment /t/ and in the item 'muito', schooling and collection type are the factors selected, in this order.
The data distribution points to a scale of frequencies of the individuals by schooling ranges, which is repeated in the sample with the data on 'muito': the higher the schooling, the lower the tendency to using the affricate variant (Table 12).This result converges with what Mota (2008a) points out.The constitution of the sample, however, for being heterogeneous, imposes a relativization of these results: the strong disadvantage of the occlusive variant associated with the ES 2 nd grade, for example, refers to a specific speaker, as well as the proximity of the behavior of the variant in high school (one single speaker) and higher education (three speakers).
Safeguarding all diatopic and diachronic dimensions, the results show stability, and support the hypothesis that the affricate variant is a stereotype.The analysis, in addition to motivating the reflection on the influence of the control of formality/informality in the prediction of a variable phenomena at the phonetic-phonological level, contributes also to the discussion about the social value of variation, as it evidences, in this community of practice, the existence of social and stylistic reasons for the choice of forms, thus corroborating the premise that linguistic variables can convey locally based social meanings (LABOV, 1972(LABOV, , 1994(LABOV, , 2001;;ECKERT, 2001ECKERT, , 2012)).

Conclusion
The results of the analysis of alveolar stops preceded by subjacent palatal glide, the input values obtained in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria do not differ from those obtained by Souza Neto (2008) in the speech of Aracaju people, with sociolinguistic methodology.The results obtained also corroborate the schooling polarization reported by Mota (2008a) in her study of the speech of Aracaju people, with ALiB data.However, the analysis presented herein highlights how much the control of the context, be it linguistic, social or stylistic, is important in the analysis of variable phenomena at the phoneticphonological level.It is possible to refine even more the analysis of variation in the realization of alveolar stops preceded by palatal glide with a more often stylistic control, like the influence of the type of discursive sequence/type of text and discourse topic, for example, in the occurrence of the phenomenon.
The analysis of the linguistic context involved in the phenomenon shows issues that deserve a closer look, following the suggestion made by Cristófaro-Silva et al. (2012, p. 85), that […] what is traditionally referred to as alveolar stops palatalization phenomenon involves, in fact, complex and interconnected relations between various sound patterns.Specifically, usage-based phonology (BYBEE, 2003) approach can be productive to clarify the tendency to the affricate variant in diminutives, a context of interaction with nasalization and monophthongization, and the behavior of the oitoradical, which, unlike names, tends to block the affricate realization, regardless of stylistic or social context.

Table 1 .
Affricate realization of /t/ and /d/ in contexts preceded by glide /j/ in Aracaju, Sergipe, according to the speaker's age group, schooling and sex, based on ALiB data.Adapted from Mota (2008a).

Table 3 .
Occlusive realization of /d/ in contexts preceded by glide /j/ in Aracaju, Sergipe State, according to statistically significant factors.Adapted from Souza Neto (2008).

Table 4 .
Occlusive realization of /t/ in contexts preceded by glide /j/ in Aracaju, Sergipe, according to the factors selected as statistically significant.Adapted from Souza Neto (2008).

Table 5 .
Distribution of speakers in the sample.

Table 6 .
Occlusive realization in contexts preceded by glide /j/ according to the type of data collection in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria.Data surveyed by this study.

Table 7 .
Occlusive realization in contexts preceded by glide /j/ according to the selection order of linguistic factors in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria.Data surveyed by this study.

Table 8 .
Occlusive realization of /t/ in contexts preceded by glide /j/ according to the selection order of the linguistic factors in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria.Data surveyed by the study.

Table 9 .
Occlusive realization of /t/ in contexts preceded by glide /j/ with exclusion of diminutives, according to the selection order of linguistic factors in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria.Data surveyed by this study.

Table 10 .
Occlusive realization of /t/ in contexts preceded by glide /j/ according to the type of collection and to the 'degree' of the name in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria.Data surveyed by this study.

Table 11 .
Occlusive realization in contexts preceded by glide /j/ in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria according to individuals.Data surveyed by this study.

Table 12 .
Occlusive realization in contexts preceded by glide /j/ according to the individuals' schooling in the community of practice of Praesidium Mãe da Divina Graça da Legião de Maria.Data surveyed by this study.