Instruments

3.2.3.1 The opening questionnaire

In the opening series of questions, the items were concerned with the following set of constructs.

The Cronbach‟s Alpha values are indicated in advance in case of multi-item scales in order to highlight the degree of reliability for measuring the given construct, and sample items are provided to illustrate the concept of the scale.

77

● The extent of exposure in reference to the types of exposure (e.g. exposure through films, internet), Cronbach‟s Alpha: .787;

E.g., I often hear English on TV.

● The extent of exposure in reference to the varieties of English learners are exposed to, Cronbach‟s Alpha: .728;

E.g., I often hear Hungarian English.

● Awareness of English accents and phonology, Cronbach‟s Alpha: .806;

E.g., I can guess where a speaker is from based on their pronunciation.

● Tolerance of ambiguity: focusing on general meaning, coping with not understanding everything, Cronbach‟s Alpha: .771;

E.g., It does not bother me if I do not understand every word when I hear English.

● Benefits of contact with speakers of English, Cronbach‟s Alpha: .574, Adopted from Illés and Csizér (2010);

E.g., It is good to speak to foreigners because I can get to know their pronunciation.

● General attitudes towards phonological variation in Hungarian and in English;

E.g., It is acceptable if people have different pronunciation in Hungarian.

● Biographical questions: age, gender, proficiency in English.

78 3.2.3.2 The comprehension tasks

The aim of the comprehension tasks was to measure the effect that the accent exerts on the comprehension of spoken English. To this end, a comprehension exercise based on a standardised B1 level language exam, the Cambridge Preliminary English Test (PET), was used, which was adapted to suit the purposes of the present research. This relatively low-level exercise was considered to be ideal for the comprehension task since this way the proficiency of the participants would not be challenged. This was necessary because the intention of the task was to measure the effect of the accents in isolation from other factors as far as possible.

An almost identical script was recorded four times with different speakers, by varying target words in a way that the information in the previous texts could not help students with the subsequent target items. The reason for opting for identical texts was to eliminate the potential effects of the different content, that is, to avoid certain learners having an advantage over the others because they were more familiar with a given topic and could thus make better use of their schemata in understanding the excerpt and predicting the target words better. This also made it possible to gain comparable results from the four excerpts. The transcript of the comprehension tasks can be found in Appendices B1–B4.

The speakers in the audio recordings included four narrators: a Hungarian, an Egyptian Arabic, a GA and an RP speaker of English, corresponding to Kachru‟s (1992) classification of World Englishes, the Expanding, the Outer and the Inner Circle respectively. The four speakers were selected from a number of candidates not featuring in the final study including French, German, Pakistani, Scottish, Cockney and inner London accents. Reasons for exclusion included

79

problems with reading the script, conscious effort to conceal accent, or the time-limits of the exercise.

It is assumed that due to the fact that the text was below the level of the learners‟

proficiency, the order in which the accents were played did not have a major influence on the results by making the later recordings easier to understand. This assumption was confirmed by the results showing that the participants had a lower average score on the second accent compared to the first one (cf. Section 3.3.1); therefore, it can be assumed that an increased familiarity with the task did not make the comprehension of different accents easier, contributing to the validity of the measurement.

As for the socio-demographic factors of the speakers, they were all middle-aged educated males, sufficiently competent in English to make the recordings sound authentic. The personal variables of the speakers such as age, gender and education were intended to be kept uniform so that these would not influence the results of the questions related to the speaker. It has to be noted, however, that the RP accent was taken from a professional recording, made specifically for ELT purposes, while the other accents were recordings of non-professionals with the use of sound recording techniques falling short of studio quality. Although these technicalities might have influenced the difficulty of the listening materials, it could be argued that the professional recording represented classroom English, which learners were exposed to in ELT, while the recordings of non-professionals retained the authenticity of English outside the classroom.

During the task, the learners were asked to fill in the missing pieces of information during a single hearing of the text. The texts contained 14 gaps each, which was necessary for two reasons. Firstly, the 14 gaps could ensure a sufficient number of variables in the texts so that no word in a text could be used to provide missing information in a following exercise. Secondly,

80

this number allowed for phonological diversity in the target words and so the speakers‟

pronunciation could exert a strong influence on comprehension. The target words included numbers and basic words which the participants were expected to be familiar with. At the very end of the session, the learners were requested to correct their tasks based on the answer key, which was read aloud by the administrator of the data collection, and indicate their scores. This substantially reduced the time needed to process the data and also provided participants with the opportunity to gain access to their results, making the data collection a more meaningful experience for learners. The process of self-correction was strictly controlled and closely monitored in order to prevent inaccuracies, and the corrections made by the learners were revised subsequently.

3.2.3.3 The questionnaire related to the accents

After each comprehension task, the participants are asked to answer three short open-ended questions based on the particular accent (cf. Appendix C). The open-ended questions provided room for the respondents to express their opinion concerning the accent and the speaker immediately after listening. The learners‟ free associations were expected to provide valuable insights into how they relate to speech varieties. This was followed by a set of Likert scales, which focused on the following areas:

● Perceived correctness, likeability and comprehensibility of the speaker‟s English;

81

● Social proximity to the speaker, referring to „layers of acceptance‟, asking whether the respondent would accept the speaker as an acquaintance, fellow learner, colleague or English teacher;

● Multi-item scales related to stereotypes pertaining to the speaker based on his accent, such as education, intelligence or friendliness; Cronbach's Alpha by accent: Hungarian: .855, Egyptian: .851, American: .917, RP: .902;

● Familiarity with the particular accent.

3.2.3.4 The format of the questionnaires

The items in both the general and the accent-specific attitude questionnaire contained five-point Likert scales on which learners were asked to indicate the extent to which they agreed with statements. The five-point scale was considered ideal as Hungarian learners are familiar with this rating from the five-mark assessment system used in the Hungarian public education system. The five points in the scale also allowed for a sufficient range of distribution among the answers, yet it keept the differences on the scale meaningful for learners.

The questions were laid out in a grid format in order to facilitate the completion of the questionnaire, making the process of responding more reader-friendly and thus fluent with the unified format. It was assumed that as learners got used to the process of understanding the statements, which were designed to be unambiguous and easy to process, and then ticking the appropriate box, their answers would become more and more spontaneous. This was intended to contribute to eliciting unselfconscious answers related to attitudes, minimising the effect of the social desirability bias, somewhat easing the problem of the observer's paradox. The format and

82

the layout was also designed facilitate the timely completion of the questionnaire, which was necessary because of the time constraints of the data collection sessions, the 45-minute school periods, and the relatively large number of questions. These features of the questionnaire were expected to contribute positively to the validity of the research.