• Tidak ada hasil yang ditemukan

R ESEARCH D ESIGN AND M ETHODOLOGY

3.6 Phase Two: Main Fieldwork

3.6.1 Data Collection

3.6.1.1 Method

Using various methods and techniques of investigation we sought to collect material and gain insights in order to formulate answers to our research questions.

3.6.1.1.1 Participants and Locations

At first, the research locations for the study were identified. Thereafter, the main fieldwork was conducted at six locations in Guwahati: (1) Bathoupuri Bodo M.E.

and H.E. School, (2) J. B. Hagzer Bhawan, (3) Bathoupuri, (4) Ganeshpara, (5) Champavati Higher Secondary School, and (6) Indian Institute of Technology Guwahati (see Figure 3.2 and Appendix B). The research was carried out by paying several visits to these locations. Out of the six locations two of them (Bathoupuri and Ganeshpara) received more attention because apart from the main experiments, the norming experiments were also conducted in these locations.

Figure 3.2 Locations for main fieldwork (adjusted from Google MapsTM).

Participants were recruited through consultation that emphasized the need for native Bodo speakers who were bilinguals and used Assamese as their second language. In total, 107 Bodo−Assamese bilinguals took part in the experiments. Each participant took part in more than one experiment with a gap of minimum fifteen days between the experiments. They were presented with the tasks in the environment most convenient for them (participant’s home/workplace, research lab) and the tasks were completed in a quiet environment with only the researcher and the participant present.

Hagzer Bhawan

Bathoupuri Ganeshpara

3.6.1.1.2 Proficiency Measures

A sociolinguistic profile of the informants’ background is necessary to predict bilingual behavior, or to exclude certain informants from participation. One of the most important reasons of asking questions about the sociolinguistics background is to find out how proficient bilinguals are in both languages, whether one of the languages is dominant, and what the language preferences are. To find out about these, four categories of questions were included: (1) language history, (2) language choice, (3) language dominance, and (4) language attitudes. Two measures of language proficiency were administered: Firstly, a self-assessment measure which formed a part of language background questionnaire, and secondly, an objective naming test in both languages.

Language Background Questionnaire.

The questionnaire was initially constructed in English and subsequently translated to Bodo considering that the same was the mother tongue of the participants in the study. This is done in order to ensure eliciting of increased degree of valid responses considering the fact that the participants would be more comfortable in expression using the mother tongue.

Reverse translation from Bodo to English was done in order to check the validity of translated questionnaire (see Figure 3.3, Appendix C (i) and C (ii)).

Measure for self-assessment by participants relating to age of acquisition of L1 and L2 were incorporated in the language background questionnaire in order to ensure that Bodo language was begun to be learnt by the participants first. The query on age of acquisition in the survey was also utilized to classify the participants on the basis of age of acquisition of second language (i.e. Early L2 learners and Late L2 learners). The criterion adopted for making the said classification was learning of Assamese language prior to 7 years1 of age (designated as Early L2 learners) and

1 Seven years of age was selected as the cut-off for Early L2 learning following the convention

post 7 years (designated as Late L2 Learners).The questionnaire also included a 7 point self-assessment scale to map the proficiency of the participants in speaking, reading, writing and comprehension in respect to both languages. This was used to bifurcate the Late L2 learners into high and low Proficient groups. Also, over and above the self-assessment proficiency measures as depicted above, the participants individually performed the Bodo and Assamese versions of the Objective Naming Test (see next section).

Figure 3.3 Participants in Bathoupuri filling out language background questionnaires.

Objective Naming Test.

The objective naming test was an adaptation of the Boston Naming Test (BNT), which is a standardized test of object naming ability (Borod, Goodglass, & Kaplan, 1980; Kaplan, Goodglass, & Weintraub, 1976; 1983).

The test contains 60 line drawings of objects presented individually. In this test, participants are asked to name pictures as quickly as possible, which have grading in naming difficulty embedded, so that relatively familiar objects are put on display prior to less familiar objects. Considering that linguistic and cultural factors have a bearing on naming ability of individuals, a modified version of BNT was developed for the study. While maintaining the core procedural integrity of the original BNT, modifications were made on the itemized content. As such, 18 of the original 60 items were retained for the test, considering the objective validity of the items vis-à- vis the Bodo participants. In order to arrive at the stipulated item level of 60, 42 new items were added and were adjusted in the increasing order of difficulty.

During the course of experiment, this modified version of BNT was administered once in Bodo (L1) and once in Assamese (L2) to each participant, with embedded counterbalance of the order of presentation of language. The individual participants were accorded 15 seconds to name each of the object depicted in the pictures being presented one by one. As far as scoring was concerned, participants were awarded one point credit for correctly naming a picture. In case of objects with multiple names, credit was accorded for any one correct term being recorded by the participant. Further, dialectal variations, if any, were also accepted for the object naming task.

Zero credit was awarded to the participants in case of wrong responses, non- remembrance of names of objects (nil response) or responses after the stipulated time of 15 seconds. A median spilt was then performed on the Late group’s scores after the summing up of the total scores of the individual participants. The Late High Proficient group of participants comprised of fluent bilingual speakers whose scores were above the median on the task pertaining to Assamese (their L2). Similarly, the Late Low Proficient group was demarcated to comprise of bilingual speakers who had scored below the median in the task pertaining to Assamese. The data obtained from the self-assessment reports were juxtaposed with the results obtained from the objective naming test to map their degree of convergence. Three bilingual groups, namely (i.e., Early High Proficient; Late High Proficient; and Late Low Proficient) were created out of the survey responses.

3.6.1.1.3 Tasks and Paradigms

In relation to bilingual studies, there exist multifarious tasks which are considered to be valid and reliable measures to cognitive performance (Altarriba & Heredia, 2013). The last decade has seen a growing body of research investigating various aspects of L2 learners’ performance of tasks (see, for example, Bygate, Skehan, &

Swain, 2001; Ellis, 2003). One of the most widely used experimental paradigms in psychological and linguistics studies is priming. It originated in first language (L1) perception and production research and have been gaining prominence in second language (L2) processing and acquisition studies since the beginning of early 1980s (McDonough & Trofimovich, 2009). In the context of language use, it refers to the phenomenon which involves the positioning of one word preceding the target word to which the participants are expected to respond. The second word against which the participants are to submit their responses is called TARGET while the word preceding the target is called PRIME, which is usually presented for a very short time frame. The time lapse between the presentation of PRIME and the presentation of TARGET is called Stimulus Onset Asynchrony (SOA). Primes can be a sentence, picture or auditory in nature. However, for the course of the study, focus is on primes which are written words or in the form of pictures. The way in which a prime is related to the target can result either in facilitation or inhibition. For example, previous research shows that words that are connected in some way (either semantically, phonologically, etc.) show facilitative effects, whereas words which are not connected shows inhibitive effects. However, as will be evident in the rest of the thesis, this issue is far more complex. For many decades now, experimental paradigms using primes have gained phenomenal popularity, largely because it assists in developing an understanding of how words are inter-linked or stored in memory and whether certain processes occur consciously or are initiated automatically. In other words, priming experimental paradigm assists psycholinguists to explore representation of lexical forms and semantics in memory.

The priming paradigm with bilingual population involves the presentation of cross- language word pairs, in order to examine the organization and processing mechanisms of the bilingual’s two languages. Another widely adopted version this

paradigm in which the participants are not aware of the prime is the masked priming paradigm (Forster & Davis, 1984). This version has the added advantage that the participants are not aware of the bilingual nature of the task, and thereby, curtails strategic factors which may be evident in the general priming approach (De Groot and Nas, 1991; Davis et al., 2010).

In the present study, a priming research is typically carried out in order to establish generalizations about language representations and processes that provide insight into the organization of mental processes. In order to further investigate the influence of strategic factors, the present study also adopts the masked version of this popular paradigm, in order to examine the extent to which participants rely on controlled, strategic processing in completing the task. To minimize the visibility of prime words, we employed masking techniques which included forward masking (when a pattern mask preceded the prime). In all the tasks involved in our study, we also manipulated the amount of time involved between the presentation of each prime word and target word, that is, the SOA. It is generally acknowledged that a longer SOA results in a ‘deeper’, more elaborate processing of the prime and likely encourages participants to use certain strategies to perform the task. In contrast, a shorter SOA is believed to ensure that participants’ performance on the task is influenced by an automatic, unstoppable activation of information available in the prime and target words (Trofimovich & McDonough, 2011). By manipulating SOA, we thus attempted to determine whether participants activate information automatically or by means of strategies. Thus we used longer SOAs as well as SOAs that were brief, and the prime was either preceded by a mask (masked paradigm) or not (unmasked paradigm).

The present research has focused broadly on a variety of design features of tasks and implementation procedures to explore how these factors impact on such aspects of language use as L1 and L2 comprehension and production. The purpose of using different task was to take into account different phenomenon that different tasks often probe when interpreting the findings. Thus from the psycholinguistic perspective, we attempted to deduce the underlying structure of the bilingual system from patterns of behaviour that Bodo−Assamese bilinguals exhibit over a range of

tasks. This includes tasks such as visual lexical decision, semantic categorization, translation recognition, word naming, word translation, simple picture naming, picture word interference, and primed picture naming (see Table 3.7, Figure 3.4 and 3.5). All these tasks differ in terms of the processing operations involved in performing each task. For example lexical decision and semantic categorization tasks involve comprehension processes. In contrast, word naming and word translation tasks ostensibly include both comprehension and production components.

In interpreting task results, it is important to consider how priming effects depends on access to lexical and semantic/conceptual information in comprehending versus speaking a language. Chapter 4 and 5 provide a comprehensive and accessible overview of priming methodology and its application to the present research.

3.6.1.1.4 Stimuli

In addition to different types of tasks used in the study, careful considerations were given to all stimulus characteristics that may have a bearing on the outcome of the study. This step was crucial because the quality of the study depends on the quality of the materials used. In the present study, all tasks included at least two types of prime words: words that are related to the target (either semantically or phonologically) and those that are not. Moreover, the related primes were related to the target in different ways. For example, priming effects have been investigated for prime-target word pairs that are translation equivalents (e.g. आखाइ [akhai]−1' [hat] ‘hand−hand’), semantically related (e.g. आखाइ [akhai]− [anguli]

‘hand−finger’), associatively related (e.g. आखाइ [akhai]− [kharu]

‘hand−bangle’), phonologically related (e.g. आखाइ [akhai]−2 [akhoi]

‘hand−puffed rice’), or phonologically related to the translation equivalent (e.g.

आखाइ [akhai]−1 [har] ‘hand−rate’). By manipulating the types of relationships, we attempted to determine how Bodo−Assamese bilinguals represent and use various kinds of information in their memory.

81

Table 3.7 Overview of Experiments in the Present Research

Experiment Task Process Task

language

Critical item Script Exp. 1A, 1B, 1C, & 1D Unmasked & masked translation

priming Comprehension L2 & L1 Cognates & non-cognates Present in task Exp. 2A, 2B, 2C, & 2D Unmasked & masked semantic

priming Comprehension L2 & L1 Cognates & non-cognates Present in task Exp. 3A, 3B, 3C, &3D Unmasked & masked

associative priming Comprehension L2 & L1 Cognates & non-cognates Present in task Exp. 4A, 4B, 4C, & 4D Unmasked & masked

phonological priming Comprehension L2 & L1 Cognates & non-cognates Present in task Exp. 5A & 5B Masked semantic categorization Comprehension L2 & L1 Cognates & non-cognates Present in

task Exp. 6A & 6B Translation recognition Comprehension L2 & L1 Cognates & non-cognates Present in

task

Exp. 7A & 7B Word naming Comprehension &

production L2 & L1 Cognates & non-cognates Present in task Exp. 8A & 8B Word translation Comprehension &

production L2 & L1 Cognates & non-cognates Present in task Exp. 9A & 9B Picture naming Production L2 & L1 Cognates & non-cognates Absent in

task Exp. 10A, 10B, & 10C Picture-word interference Production L2, L1, &

mixed Cognates & non-cognates Present in task Exp. 11A, 11B, & 11C Primed picture naming Production L2, L1, &

mixed Cognates & non-cognates Present in task

Figure 3.4 Photograph of a participant performing the visual lexical decision task.

Figure 3.5 Photograph of a participant performing the simple picture naming task.

In addition to that, cognate status of target words and pictures was manipulated to examine its influence on the outcome. To make sure that the word pairs that were identified as related were indeed related in meaning and, conversely, that the word pairs identified as unrelated were indeed unrelated, we carried out our own word norming study prior to conducting our experiments (see Appendix E (i), (ii), (iii), (iv), and (v)). The nature of the processing of these word types across Bodo−Assamese bilinguals will be discussed in light of various experiments that examine the processing of those words across a bilingual’s languages.

3.6.1.1.5 Variables

In all experiments, variables were manipulated to determine their effects on data patterns. The independent variable was the one which we surmised would be the cause of the intended result, while the dependent variable was the one which was measured to understand the impact of the independent variable. In the following section we discussion each type of variable separately.

Dependent variables.

The dependent variables in the all the experiments were the response latencies (including correct and incorrect responses). In the calculation of reaction times, we considered only reaction times for correct trials. Reaction time was measured as the amount of time between the onset of the target word/picture and the participant’s response to it. Percentage of error was measured as the proportion (or percentage) of correct trials to each type of target in a condition to the total number of each type of target presented in that condition.

Independent variables.

We manipulated several independent variables in all the tasks. One such set of variables includes the Prime Type, i.e., the various meaning and form relationship between prime and target words. Cognate Status was another independent variable that we manipulated throughout the experiments to explore the role of cross-linguistic overlap. Moreover, the variable of Bilingual

Group was used to examine the role of age of acquisition and proficiency of the bilinguals.

3.6.1.1.6 Instrumentation

The data collection took place in computer-collected format. During experimentations, the protocol set for language usage was kept consistent with the purpose of the study.

Laptop.

An Acer Aspire 5920 laptop computer with a 15 inch viewing screen was used as a tool for stimulus presentation, collection of responses from participant, storage of the data emanating from experiments conducted therewith and analysis of the raw data so obtained to generate meaningful output. During the experiments, participants were seated in front of the laptop computer at a distance of about 60 cm from the screen.

Software Program.

All the experiments were run using the E-prime Software Package (Version 2.0, Schneider, Eschmann, & Zuccolotto, 2002) on Windows XP.

Microphone.

In a production experiment, a voice key and a microphone are often used to record the participant’s vocal reaction times (spoken responses). A voice key uses a bit of dedicated hardware, essentially a microphone pre-amp, which detects the onset of a vocal response, and sends out a signal when it occurs. For the present study, a Sennheiser microphone was used for registering the vocal reaction time of the participants in the production experiments. The microphone was plugged into the sound card of the stimulus computer and on each trial, the vocal responses were recorded as .wav files. Later on, the waveforms for each trial was examined in Audacity (a sound editor), by putting markers on the start of the speech manually, and calculating vocal reaction times relative to the start of the trial. This process was reasonably time-consuming, but very reliable and precise.