PHONEMIC FLUENCY TEST at least two words are generated in a cluster

PHONOLOGICAL CLUSTER EXAMPLE

same initial letters expected (this is the task) same 2 initial letters spaletta-spagetti; sound-sour

rhyming pairs kapál-kalapál; tree-three

same first and last letters sír-sár-sör;it-fat

homonyms só-show; foul-fowl

SEMANTIC CLUSTER words that share the same meaning or are in the same semantic category

Table 7: Clustering principles 2.4.3.4.2. Switching (shifting)

Troyer (1997) defines switching ability as the of frontal lobe functions that requires mental flexibility. In the dissertation I applied Abwender and colleagues’ (2001) protocol on assessing clustering ability. They distinguish two types of switches (’cluster switching’ and ’hard switching’) to draw the attention to the probably different subserving mental mechanisms. They define cluster switches as the shifts between two adjacent or overlapping clusters and hard switches as the transition between a cluster

74

information processing and mental productivity. In the dissertation this differentiation is adapted in order to gain a more subtle analysis. The two types of switches are coded separately and then the total number of raw switches is also calculated in the dissertation in accordance with the related literature (Troyer, 2000; Abwender, 2001).

Table 8 reports on the general evaluation principles of fluency performance:

Table 8: General evaluation principles of fluency performance 2.4.3.4.3. Additional aspects of fluency performance

To capture the differences in the lexicons of the two participating groups the number of unique and academic lexical items has been calculated. In accordance with Verspoor and colleagues’ article (2012) lexical sophistication is measured by the number of unique words and academic vocabulary that were calculated at the group level.

75

In order to calculate the number of academic words and words that belong to different genres COCA (Corpus of Contemporary American English) and MNSZ2 (Magyar Nemzeti Szövegtár) databases were applied. The number of different word types (abstract and concrete nouns, verbs, adjectives and ’others’) was defined due to our assumption that their frequencies might reveal an imbalance between the different language levels of the two groups. Finally the topics of the generated words were also defined. In the dissertation unique words at the group level are defined as lexical items being mentioned only by one of the groups.

2.4.4. Structured interviews

According to Nádasi, (2000) the application of structured interviews is justified if the researcher aims to reveal participants’ views on a specific topic. Given that, learners at this age might not have explicit knowledge about those cognitive abilities that have been tested, interview questions must have been consciously created to meet our primary aim which was the detection of dissimilarities in CLIL and control group participants’ mindsets. Therefore, we sought to create questions that do not require any special expertise in any topics, but provide enough stimuli to make individuals’ specific approaches detectable. For this reason the topic of L2 learning have been chosen. This way information about the applied L2 methodology through which learners are really taught has been received. We have formulated some of the questions in such a way that they can be approached from different perspectives (e.g. What is the most important in learning English?) We assumed that participants with the highest tests results would provide more specific and diverse answers as the signs of mental flexibility.

Before the compilation of questions, a set of different criteria had been considered in line with Nádasi (2000). First, considering the age of participants (adolescents), questions had to reflect sincere interest while being motivating, straightforward and clearly worded. For this reason, the structured interview comprised open-ended questions in a special order among which core items were placed in the middle section.

The order of questions was influenced by the following principles: sharp shifts between a) topics and b) viewpoints (internal-emotional or objective; imagination or reality) have been made. The application of qualitative interviews has some advantages: due to their exploratory nature, participants can express their opinion with high engagement and their thoughts can be analysed in the original context.

76

Regarding the number of questions (11) and the time constraints (approximately 30 minutes in a 45-minute lesson in regular school time), the written format was preferable, since it contributed to the indirect (self-paced) management of time without external intervention. The questions were written in the Hungarian language for two reasons:

first, to eliminate anxiety and secondly, to avoid restricted language use that might have eventuated in less informative opinions. However, participants were given the opportunity to answer the interview questions in line with their language preference.

The text corpora consisted of two pages on average. The processing of the structured interviews was facilitated by the MAXQDA software.

For processing structured interview data, qualitative content analysis (QCA) with the combination of deductive (concept-driven) and inductive (data-driven) logic was applied (Kuckartz, 2019). During deductive content analysis the setting up of a code list (code frame) preceded the data analysis. This list of codes (categories) is required to be described precisely to ensure the reliability of the coding procedure which is only a part of the preparational phase (Kuckartz, 2019). In the study the reliability of coding was ensured by intracoding during which the author coded the interviews twice following the same coding system. The elapsed time between coding and re-coding was two weeks to eliminate remembrance. During re-coding there was no conceptual change, hence the reliability indicator of coding was 1.

The practice of ’a priori’ coding is justified if the compilation of main codes is based solely on theory and/or the researcher’s own experience (Sántha, 2015). After setting up the code frame, the next step was the inductive discovery of the deep layers of text segments. This time subcodes were searched first, and then a match between the meaning of texts and the subcodes. All text segments that are relevant to the research must be and were coded completely. The analytical process comprised the phases of summarizing, comparing, and contrasting data (Kuckartz, 2019). Both qualitative and quantitative techniques were applied in the data analysis. The qualitative analytical techniques contributed to the exploration of the different opinions of the groups on specific topics. In contrast, the application of quantitative techniques revealed the frequency of certain categories and subcategories.

77

2.4.4.1. Selection criteria for the structured interviews

For the qualitative content analysis 4 groups have been created: a high-performance CLIL group (referred to as CLIL ‘high’), a low-performance CLIL group (CLIL ‘low’), a high-performance control group (control ‘high’) and a low-performance control group (control ‘low’). In each group there are three participants. Participants who met the following criteria have been selected for the ‘high’ groups:

- regarding the D2-R test, participants had to achieve high performance level for attention (at least 110 standard points) coupled with high accuracy (at least 110 standard points). Any errors that may result from the measurement accuracy of the test have also been considered while constructing these scores. Considering the measurement error of the test, these participants achieved at least high or average performance in both parameters;

- regarding fluency tests, participants had to be among the five (positive) outliers listed by SPSS software in as many investigated parameters as possible.

Participants in the ‘low’ groups had to meet the following criteria:

- regarding the D2-R test, participants had to achieve low performance level for attention (maximum 94 standard points) coupled with low accuracy (at least 94 standard points). Any errors that may result from the measurement accuracy of the test have also been taken into account while constructing these scores.

Considering the measurement error of the test, these participants achieved at least low performance in both parameters;

- regarding fluency tests, participants had to be among the five (negative) outliers listed by SPSS software in as many investigated parameters as possible.

78