Testing oral proficiency: what does pronunciation tell us? DOI:10.5007/2175-8026.2011n60p247

In this study we discuss the role of pronunciation in languagetesting and investigate two features of pronunciation of eightcandidates of the Test of Oral Proficiency in English (TEPOLI)along four bands of the test scale. Deviations in vowel segmentsand in syllable structure are analyzed. The results point to theneed for a global assessment of the candidates’ phonologicalsystems.


Introduction
Teacher development programs should take into consideration what type of language education is needed for the students since they start attending regular school.In Brazil students usually start school at the age of six, in the 1 st grade of Ensino Fundamental. 1 It is from a perspective of a desired language competence for students (Alderson et al., 1995;Hughes, 1989;McNamara, 2000;Scaramucci, 2008Scaramucci, , 2000) ) that criteria can then be established to qualify language teachers, for example, those who are required to work in the various types of Brazilian schooling contexts.
In this paper we focus on the assessment of oral proficiency in foreign languages based on the assumption that it is possible to discuss levels of proficiency for EFL teachers only if we have a clear idea about what characterizes a teacher's profile and what type of language he or she is expected to use as a professional.Although we make reference to data and the reality of EFL teachers in Brazil, we believe this discussion can be of interest for other contexts of foreign language teaching as well.
The National Curricular Parameters (Parâmetros Curriculares Nacionais, henceforth PCN) (Secretaria de Educação Básica, 1998Básica, , 2000) ) for primary and middle school in Brazil encompass a sociocultural view of language and determine that teaching goals-not only concerning foreign language teaching-should be established: according to the students' social, intellectual and professional needs, as well as their expectations (…) Foreign language teaching in regular schools may reveal, however, the relevance for the learning of other languages in Brazilian students' lives.A foreign language, and in this particular historical moment, the English language, gives access to modern science and technology, to intercultural communication, to the world of business and other ways of understanding human life (Secretaria de Educação Básica, 1998, p. 65).
A more recent document, the Curricular Orientations (Secretaria de Educação Básica, 2006, p. 119) states that the proposal for the teaching of foreign languages in middle school shall not be restricted to the market but take into consideration the educational benefits of language learning, for the overall formation of students as individuals and as citizens.But at the same time foreign language teaching should not ignore the job market, and the fact that many student who finish middle school will look for jobs.
The social needs for intercultural communication can be easily seen through the evolution of global relations, favored by the expansion of cultural boundaries and generated from technological developments, especially of the Internet.The present situation indicates that, when the PCN were written, situations of change due to those phenomena were already anticipated.Nowadays many schools in Brazil have equipment for Internet access, even though in some cases the technical conditions may be somehow rudimentary.
Given such a scenario of various needs for foreign languages, we interpret the social needs stated in the PCN as the needs to access information and global communication.In this context the English language has characteristics that make it distinct from other foreign languages in the school curriculum in Brazil, such as Spanish.As a result, English should be taught in regular schools from the perspective of its international status and for multicultural communication.
And "more than reinforcing only momentary social values, the ones that are put forward by the cultural and economical forces of globalization, we understand that the aims of a project for inclusion should create conditions for nationals to establish a dialog with other cultures without having to neglect their own values" (Maturana, 1999, as cited in Secretaria de Educação Básica, 2006, p. 96).In this proposal, the Curricular Orientations for Middle School discuss the teaching of foreign languages not only from an instrumental, but also from a social perspective, which permits one to understand himself and the others, by means of different forms of linguistic and cultural expression.It is essential, to understand such proposal, to highlight the assumption "without having to neglect their own values".This assumption reinforces the maintenance of students own identities and it is compatible with the teaching of English as an international language, which aims at various cultures in which English is not spoken as a mother tongue and does not serve communication needs primarily with native speakers.It is from this perspective that we discuss the assessment of pronunciation concerning EFL teachers.

Theoretical framework
The status of the English language has changed dramatically in the past years.It has gone beyond the boundaries of a first language and is now spoken by a massive number of non-native speakers, one which exceeds the number of native speakers.If the predictions made by Graddol (2006Graddol ( , 1997) ) are correct, around 2015 there will be two billion people learning English.Adding that number of learners to those who already have some competence in the language, we foresee a panorama in which half of the world's population will be able to communicate in English (Crystal, 1997;Ives, 2006).In this global scenario, it seems non-realistic to consider English language teaching based on the native speaker norms, especially concerning phonological aspects.Conversely, we must consider the social changes through which the world has gone.
Authors such as Jenkins (2000), Simo Bobda (1994), Kachru (1992Kachru ( , 1986) ) and Atechi (n.d.) have challenged the relevance of native speaker models in non-native contexts.They state that, in contexts of English as an international language (henceforth EIL), it does not make sense to adopt the judgement of native speakers for standards of language proficiency because in those contexts English is used for communication between non-natives whose first language may vary considerably.In contexts of EIL the key factor to judge communicative competence and language proficiency is intelligibility, besides strategies of accommodation on the part of speakers.
According to Smith (1992) intelligibility is defined as the recognition and decoding of sounds (segments and suprasegments) in words and statements.In a scale for comprehension, we will find a level above intelligibility called 'comprehensibility' and, at a higher level, 'interpretability' , which concern semantic areas, more precisely in the relations between meaning and form, and discourse, respectively.Atechi (n.d.) states that it is difficult to establish precisely where one category ends and the next category begins, due to the fact that the context, the topic and world knowledge play a role in the understanding of sounds, by means of an inference mechanism.Based on those reasons, Atechi (n.d.) has chosen not to establish a distinction between intelligibility and comprehensibility, and uses the term 'intelligibility' to refer to both of them.
Native speakers make use of contextual information to solve problems of comprehension and/or ambiguity.Non-native speakers tend to depend on segmental and articulatory information (Jenkins, 2000) and this seems to be a strong reason to maintain the distinction between the three categories, namely intelligibility, comprehensibility and interpretability.
Intelligibility has always been investigated by means of native speakers' judgement.Nevertheless, in a panorama where English is an international language it seems useless to insist on intelligibility on the part of a native examiner, as well as it is useless to assume native speakers' phonological features-such as reduction, rhythm and intonation-as parameters.The rationale is that non-native speakers base their judgement much more, if not only, on segmental information (articulation of sounds) to build their auditory comprehension.Thus, in such cases suprasegmental elements have a weak influence on intelligibility.
In order to make our intentions in this discussion clear enough we summarize the construct of proficiency testing based on the desired future use(s) of the target language.To our minds, teachers of English in Brazil ought to be able to equip their students with linguistic-communicative tools so that students can interact with any other non-native (or native) English speaker 2 .Bearing in mind that the opportunities for such communication are a lot more frequent among non-native speakers, given the contexts where English is spoken as lingua franca, students need to be equipped for EIL, to use English as a language of dialog and exchange among the most different peoples.Accordingly, it does not seem reasonable to insist that English teachers' proficiency be judged according to native models, especially when it comes to phonology.
However, teachers are the largest source of information about language form and use students have access to.This means that they work as pronunciation models to students and it is therefore essential to keep in mind that our position, in accordance with Jenkins's (2000), is not to take on just any pronunciation as acceptable, as another variant.Having said this, it seems imperative to uphold a discussion about how to establish parameters to test non-native English teachers' proficiency, since there is still no consensus among researchers about which model to adopt (Atechi,n.d.,p. 20).Jenkins (2000) makes a clear distinction between what constitutes a norm and what constitutes a model.For pronunciation we are not speaking of using Received Pronunciation (henceforth RP) or General American (henceforth GA) as a norm but as "useful classroom models or 'points of reference' to prevent learners from diverging too far in different directions" (p.226).
From the perspective of EIL, information about most nonnative speakers' phonological repertoire can be useful for the development of a model that guides examiners in proficiency tests.Atechi (n.d.) states that despite the fact that each variety presents its own peculiarities that may operate as identity marks there is a large amount of phonological characteristics shared by most of new Englishes (p.34).The author's position comes from works such as Bamgbose's (1998) and Crystal's (1997), which bring valuable information about the general features of non-native Englishes.
There is evidence that several English segmental features do not comprise a risk to intelligibility except to native speakers, as it is the case of the phonemes /D/ and /T/, for example, but that others are much more problematic, as the vowels.Analyzing data from interactions between non-native English speakers, Jenkins (2000) establishes what she calls the Lingua Franca Core (henceforth LFC).The LFC has been discussed and should still be implemented with subsequent research but by now it gives evidence on the areas and segments that need or need not appear as essential to pronunciation testing.The LFC embodies the following areas: (1) the consonantal inventory; (2) phonetic requirements, such as aspiration; (3) consonant clusters and simplification; (4) vowel sounds; and (5) nuclear stress and fluent pauses (Jenkins, 2000, p. 159).From those areas we are particularly interested, in this article, in the vowel sounds area since, as stated by Atechi (n.d.,p. 35), vowels constitute the bulk of deviations of non-native forms from NE.This deviation comes as a result of a tendency in nonnative Englishes to restructure the sounds of native English to suit their purpose.Compared to consonants, vowels exhibit a very high degree of restructuring.In short when we talk of sound restructuring between native and non-native Englishes, what comes to mind more readily is the restructuring of native vowels than consonant sounds.
According to Platt et al (1984) and Atechi (n.d.), some general tendencies are common to many (if not all) varieties of new Englishes, such as the tendency (a) to shorten vowel sounds; (b) not to make distinction between short and long vowels; (c) to substitute central vowels by front or back vowels; and (d) to shorten diphthongs and/or delete the second element of the diphthong, to name some.
Articulating vowel sounds according to native patterns constitutes an extremely difficult task to non-native speakers, being the articulatory distance between them shorter or even non-existent in non-native varieties, where both long and short vowels are homophonic.The reader could by now assume that such a distinction should be eliminated from the scope of EIL pronunciation elements.Conversely, when not adequately articulated these elements threaten spoken intelligibility.That is why vowel sounds are part of Jenkins's (2000) inventory and are investigated together with syllable restructuring in this paper.
In order to preserve intelligibility, Jenkins (2000, p. 159) states that the contrast short/long vowel sounds and also the use of /3:/ need be retained.Inadequate vowel articulation puts spoken intelligibility ate risk.But in order to understand how vowels operate to a nonnative speaker we need to take into account the systemic nature of vowels.Baptista (2000, p. 21) claims for an analysis of vowels through the perception that these elements constitute a system where each vowel is related to the others according to the articulatory distances among them: there is evidence, however, (...) that listeners perceive each vowel in relation to the speaker's total acoustic vowel space, which they calibrate from information in the rest of the ongoing speech.In light of this evidence, it seems reasonable to expect that learners do not acquire each vowel or vowel contrast in isolation, but rather in relation to the rest of their developing vowel system.If this is true, it makes sense to study the L2 learner's vowels as an integrated system.
The author has developed a longitudinal acoustic-phonetic study to investigate the evolution of this system in Brazilian English students' interlanguage.She shows how the acquisition of the distinction between /I/ and /i:/ is blocked until the vowel (or diphthong) /eI/ is calibrated.It might be the case that the same happens to the distinction between /U/ and /u:/ and the segment /oU/.
Acoustic analyses employed by Flege (1980Flege ( , 1986Flege ( , 1987) ) and Major (1989), as cited in Baptista (2000, p. 31), demonstrate that "speech sounds produced by bilinguals are often intermediate to those found in the two contact languages, although they may be perceived categorically as simply non-native".Such results lead to the consideration that the speakers' ability to perceive acoustic differences in segments in L1 and L2 "allow L2 learners to approximate, but not to achieve accurate production of L2 phones which are similar to those of the L1.This limit on successful production is due to the cognitive process called equivalence classification" (Baptista, 2000, p. 32), one which enables children learning their L1 to acoustically identify different phonemes as belonging to the same category, but one which, on the other hand, promotes a phonological translation among adults learning a L2 or FL.This means they "interpret sounds occurring in a foreign language in terms of sounds found in their native language" (Flege, 1981, p. 448 as cited in Baptista, 2000, p. 32).
We believe it is important that the speaker keeps the distances in height and duration that are distinctive of each vowel since this characteristic will be perceived by the listener and will let him or her judge what segment the speaker is producing.Therefore non-native speakers will hardly ever produce vowel segments of the same quality native ones do.It is fundamental that distinctive distances are kept.
Another feature of non-native English is syllable restructuring.Atechi (n.d.) states that epenthesis-the addition of a vowel sound to restructure syllables-comprises one of the very prominent processes most typical of new Englishes around the world.That phonological strategy tends to cause serious communication problems between non-native speakers of different varieties (Bansal 1969;Tiffen 1974;Smith 1992; all cited in Atechi, n.d.).
We thus suggest that oral proficiency testing in EFL must embrace such considerations so that in a near future we can find clearer patterns concerning the minimal proficiency levels for the English teacher in Brazil, as well as to contribute for this issue in other non-native speaking contexts.We present the design of our study below.

Objective
Our general objective in this article is to contribute to the discussion about oral proficiency testing in foreign language by listing some elements that need be taken into account in proficiency tests.This proposal converges to a specific context of oral testing in English by means of the Test of Oral Proficiency in English (henceforth TEPOLI) 3 .
The TEPOLI is a test conceived to assess the oral proficiency level of Brazilian teachers of English and it has been administered as a research instrument to undergraduate students of Letters 4 since 2002.It consists of a face-to-face interview between an examiner (interlocutor) and two candidates, in which the interlocutor and another non-participant examiner assess the candidates' performance and place them in one of the five bands of the scale (Table 01).Band D is considered a pass level.Since the TEPOLI does not intend to distinguish very low levels of proficiency, its scale does not present bands of a so-called zero level nor very low levels of proficiency.This leads to the fact that all candidates not considered able to perform in a pass level (D) fall into the lowest category (band) of the scale (E).Thus band E could be considered the one which comprises the widest range of proficiency, from a zero level up to a quasi-pass one.
The test starts with a warming-up phase that intends to relieve the tension caused by the expectations about the test.In this stage, the examiner addresses the candidates general questions aiming at briefly establishing a cordial contact.It is also during this phase that the examiner explains the test in general lines.
Next the candidates have to perform two tasks.The first one is based on the description of a picture chosen by the student from within three pictures presented to him or her.The examiner asks the candidate to describe the picture and to make any assumptions in respect to it.When the test is taken in pairs, the candidates also draw comments on their peer's prompt and may ask questions one another about the pictures, simulating teacher talk.The second task comprises a role-play where the candidates have to perform, in turns, as a teacher and as a student, and they must interact with each other.Input is generated from two transcriptions taken from real student speech in EFL classes.It is expected that the candidates focus on problematic aspects in the other candidate's speech (the candidate who plays the role of one of the students in the transcript) by correcting and offering explanations to relevant issues.To finish the test, the candidates speak about their impressions of the test.For this paper, the corpus is composed of transcriptions of the warmingup phase and the first task only.
The TEPOLI has a five-proficiency band scale (A, B, C, D, E), being the lowest one (E) representative of a candidate considered unskilled to teach English, and probably unable to conduct a class in the foreign language.The scale does not intend to distinguish every level of proficiency from a zero level.As a result, every possible level lower than D is comprised in band E. Thus a speaker whose oral proficiency is practically null is also placed in band E.
The scale presents proficiency descriptors of (a) fluency; (b) grammar and syntax; (c) lexicon and syntactic structure; (d) pronunciation; and (e) oral comprehension.In order to be placed in a certain band it is necessary that the candidate's performance in TEPOLI "corresponds to and encompasses all aspects described in a given band" (Consolo, 2004, p. 282)  5 .Candidates who display features of different bands are placed in the lowest one.Because of this criterion, it is possible that the participants selected by bands display pronunciation typical of bands higher than their general band, but never of lower bands.The test has been going through an implementation phase to be part of a proficiency exam aiming to also test candidates in written skills, the Proficiency Exam for Foreign Language Teachers (Exame de Proficiência para Professores de Língua Estrangeira, henceforth EPPLE) 7 and there have been several complete and ongoing studies to validate its proposal (Anchiete, 2007;Baffi-Bonvino, 2007;Fernandes, 2007;Ibrahim, 2007Ibrahim, , 2006;;Rodrigues, 2004Rodrigues, , 2002)).
Similarly, this article intends to contribute particularly to the investigation into the pronunciation descriptor of the test scale, by perceptually identifying and mapping the most frequent deviant production found in the speech of candidates who have already taken the TEPOLI.For that, we will focus precisely on vowel sounds and syllable restructuring.Such a portrait will make it possible for the descriptors to be revised in order to give more useful and practical information to examiners, which will help to increase the test reliability.
The most updated issues, the ones that are still under discussion, relative to the scale pronunciation descriptors, concern the use of expressions that are too generic and little descriptive of the type of language one should expect for each band.It is in search for filling this gap that we intend to contribute to the process of test validation with this article.In order to do so our study is grounded on the following research question: "How can the candidates' production in each band of the scale be described relating phonological deviation, both at the segmental and suprasegmental levels?" At the segmental level we scrutinize deviant vowel sounds, while at the suprasegmental one we restrict ourselves to the observation of deviation caused by syllable restructuring phenomena by means of vowel addition (epenthesis).

Corpus and research methodology
Eight senior undergraduate students from a Letters course in a Brazilian public university are participants in this study.All of them are female and are uniformly separated among the four lower bands of the scale proposed for the TEPOLI.
After revising the existing orthographic transcriptions based on the interviews recorded in audio and video, we started the perceptual analysis of vowel sounds and syllable restructuring.We did not make use of any acoustic analysis software on our investigation and we justify our decision grounded on the final goals of this paper.As it aims to contribute to clearer describe the type of linguistic production-particularly related to pronunciation-to supply more concrete references to examiners, we chose to carry out only a perceptual analysis because this is exactly what examiners do when making decisions on examinees classification into proficiency bands.That is, decisions on each candidate's test result are made in real time, when it is not possible for the examiner to evaluate the produced language by means of any other instruments or have the help of any software for acoustic analysis.This way only information perceptual to the human ear in real time is functional to our proposal.The results have been quantified and provide the basis to the discussion that follows.

Results and discussion
This discussion starts with the analysis of vowel sounds typical of each band in terms of deviations from the models proposed by EIL, particularly when the speaker does not keep the short-long distance necessary for phonological distinction and the intelligibility of the candidate's speech.Throughout this section the reader may find useful to refer to Chart 01, presented further in this section.
The first observation concerns three vowel sound oppositions in English, namely (a) /U/ or /u/; (b) /E/ or /{/; and (c) /I/ or /i/.We have noted that deviation in the first pair (/U/ or /u/) can be found along the whole scale through the bands, despite the low frequency, as illustrated in excerpts 01 and 02:

AY: (…) that is probably a campaign about + a: + a good uh + a good price of + telephone calls + I don't know probably (…)
Excerpt 02- LS-band D [gudi] 099 LS: (…) so I think it was a + a good action here + ok?
It can though be observed that the second pair analyzed (/E/ or /{/) has a different distribution depending on the band (Chart 01).In accordance with the general tendencies of new Englishes, the candidates' deviations are characterized by shortening long vowels sounds and lengthening short ones, in a type of substitution with the medial vowel sound-typical of their mother tongue-as illustrated in excerpts 03 and 04: 151 SB: (…) uh + you have some problems with + eh intonation like you said + (…) uh when you say about the classes that you teach + eh is not was + (it's not) were 155 L: (…) and after I think it was ok + the pictures + I think that we stayed (that kind of thing + oh) we need to talk eh fas/ faster + to think fast and talk +(…) + I think that this part of the roleplay we have more time to think (…) While LS and SB (bands D and E, respectively) display high frequency of deviant articulation of these elements (11 cases or more), TS and LC (bands D and E, respectively) have considerably lower frequency when compared to their above-mentioned peers (3 to 5 cases), which approximates their performance to JB's (band C).The results might seem inconsistent when we consider that P (band C) displays more deviations (6 to 10 cases) in these segments than some candidates of lower bands.Further ahead we shall resume this discussion to show how important these data are to fair and precise assessment of oral proficiency.
When analyzing the last pair here investigated (/I/ or /i/) we have noticed that there are no differences between the bands, as illustrated in excerpts 05 to 07.Articulatory deviations in these phonemes are of high frequency (11 cases or more) regardless of the proficiency level in TEPOLI: This way, the phonological pair /I/ or /i/ comprises the highest source of articulatory deviations in vowel sounds for the participants in this study.We also have evidence to believe that this might be the case for most of Brazilian EFL students, as our experience as language teachers tells us so.Such data are relevant to the discussion we shall bring ahead in this text.By now it is sufficient to state that it is not possible to observe a continuum in the evolution of every vowel sound as an indicative of proficiency levels.
Another vowel sound that is difficult for the participants is /V/, frequently over-nasalized due to the approximation or substitution for the Portuguese nasal phoneme /6/, as illustrated in excerpts 08 to 10.Such a deviation can be observed throughout the bands of the scale, but is more frequent and salient in the lowest ones (D and E): This phenomenon is typically caused by the influence of the mother tongue on English pronunciation and it is worth mentioning that it can operate as an identity marker for Brazilians, and it does not tend to threaten intelligibility as much as the exchange of the above-discussed pairs.This is due to the fact that in English there is no sound so nasalized as /6/ to make opposition to /V/ and thus it is comprehensible that we find such a phenomenon in every band.Firstly because we know that teachers work primarily on those sounds that are different in the first and foreign languages, those ones which have fewer distinctive traits and because of that are potential segments to cause comprehension problems.Secondly because the TEPOLI, as a test for Brazilian English speakers, needs to respect this identity and take into consideration communicative and sociodiscursive constructs.It is not required that the candidate eliminates every phonological sign of his or her mother tongue.Identity marks not causing comprehension problems can perfectly and peacefully coexist with high levels of proficiency.Besides, deviations in the sounds dealt with so far do not offer high risk to intelligibility in EIL contexts.
Even the phonological pairs mentioned above do not always cause comprehension problems when distance between them is not kept, or because there is no lexical pair to cause misunderstanding, or because the context is enough to clarify any ambivalence.Anyway, to insist on a perfect articulation of each individual segment is illusory.The literature has already shown how hard it is for adults to acquire the phonology of a second or foreign language and almost the majority of Brazilians start learning English at school age, a time when their phonological system is completely formed by the features of their mother tongue.The fact that deviation in telling /I/ from /i/ is frequent and similar throughout the scale gives us a sign that this segment may also function as an identity mark.Judging by these considerations, it is difficult to decide between what is and what is not acceptable for each band, phonologically speaking.Nevertheless data from this study can shed some more light on this issue.
Results reveal that the inventory of deviations in vowel sounds is vaster in the lowest bands of the scale than in the highest ones.We mean that a band E candidate displays deviations in a larger amount of segments than a band B or C candidate does.Although the frequency is not always high for each segment-and we understand that this happens because we are dealing with spontaneous and not controlled speech -, data provide a broader view of how the vowel system evolves as a whole, confirming Baptista's (2000)  The results above work as good indication to examiners that a candidate should not be punished for displaying articulatory deviations in a particular segment even if with a high frequency.See for example that in band B there is a high frequency of deviations in the pair /I/ or /i/.Conversely, this may be a phonological fossilization 8 while for all the rest of the system the candidate may display features considerably close to acceptable pronunciation.The distribution of deviations in relation to the entire vowel inventory is a better sign of proficiency.
With regard to syllable structure, we have noticed that its reconstruction happens by mean of epenthesis (a vowel sound addition) since this strategy is highly employed by Brazilian EFL learners.Our data about consonant deletion are irrelevant.
We have also observed that the participants of this study switch between the vowels sounds /i/ and /@/ as their choice to restructure syllables.It is interesting though to find that the candidates classified in the lowest bands of the scale tend to employ the phoneme /i/ for that purpose more frequently, whereas those ones classified in the highest bands make more use of /@/.We realize that this phenomenon bears relation to the proficiency level since /@/ is not an existing phoneme in Brazilian Portuguese.Its use shows that this element of English phonological system has already been acquired by the participant.Nonetheless, when reconstructing syllable structure by means of /i/, the speaker is still turning to a phoneme existent in his or her mother tongue repertoire.Besides, candidates placed in the lowest bands make use of a broader scope of phonemes to reconstruct syllables, such as /i/, /e/, /u/, and /@/.Excerpts 11 to 14 illustrate cases of syllable reconstruction along the scale: In our data the typical position for vowel addition is word ending because in Portuguese open syllables (CV) are preferred in final word position, except for some special cases where certain specific phonemes are allowed, such as /m/, /r/, /s/ and /z/.Cases of within syllable reconstruction are scarce in the highest bands and when they happen only the sound /@/ is added, as illustrated in excerpts 15 to 17:

Conclusion
It is inevitable that an examiner pays attention to phonological deviations when assessing performance in oral tests.That is why we have discussed, in this paper, the role of pronunciation to assess a speaker's level of oral proficiency, based on data from the TEPOLI test.
We have learned that the assessment of pronunciation should be part of a comprehensive judgement of language skills and be conducted with care, and that any judgement based only on single items should be avoided.The phonological system handled by a given speaker must be understood as a whole, bearing in mind the scope of phonemes in which deviations may occur rather than the high frequency of deviations within the same segment, since these deviations might represent a trace of the speaker's identity or a single fossilized item.
The assessment of pronunciation in a context of EIL should not demand that speakers produce all the segments in accordance with standard patterns of English as a first language.However, such assessment must be serious enough to avoid deviations in pronunciation that could risk intelligibility of native and non-native English speakers.
Future studies could focus on the other areas proposed by Jenkins (2000) in the LFC so as to contribute more specifically with the improvement of the rating scales for the TEPOLI test.We believe that this study is not conclusive and hope that it can motivate further discussions concerning the assessment of pronunciation and its role in oral language proficiency.Notes 1. Ensino Fundamental comprises the first nine years of regular school in Brazil.
2. To this point it is also reasonable to state that not only the linguisticcommunicative ability is necessary but also the social interaction ability that allows the student to recognize him or herself and to act as a global citizen.That is, the social inclusion in the world of globalization.
4. Letters courses comprise both pedagogical education and instruction in the foreign language.
5. "(…) que seu desempenho no TEPOLI corresponda, com relação a to-"(…) que seu desempenho no TEPOLI corresponda, com relação a todos os aspectos explicitados pelos descritores, na referida faixa." 6.The term 'interference' is understood as influences from the phonological system from the candidates' mother tongue (Portuguese) on some sounds when they attempt to produce oral English.8.The concept of fossilization was first introduced by Selinker (1972) and since then it has undergone several interpretations.Here, phonological fossilization is understood as the phenomenon where a given phonological form or rule becomes (permanently or temporarily) established in the interlanguage of a learner in a way which is deviant from the target language and which continues to be produced regardless of further exposure to the target language.
Table 01 presents the TEPOLI proficiency scale:

Table 01 -The TEPOLI proficiency bands
Excerpt 05-LL-band B [bit] [spIk] ['piktSur] ok + ah + I agree with with eh JB + eh + is different because the woman + eh[1]like + like us because + the the the: + the clothes + because the hair + and in my opinion eh + this picture is more eh more beautiful(…)