A Corpus-Based Study on Epistemic Adjectives in Academic English

Veiw figure View Table

The present study concentrated on the use of epistemic adjectives found in the doctoral dissertations produced by native, Turkish-speaking and Spanish-speaking academic authors of English.

2. Background to the Study

Skelton (1988) contends that epistemic comments are widely used in the arts and sciences stressing that they are found in one third to one half of the sentences in the texts. In a study where Finnish-speaking learners of English and English native speakers were compared in terms of epistemic stance devices, Ventola and Maureanen (1990) revealed that the non-native group used relatively a small number of safe expressions of epistemic possibility than the native group. Investigating scientific research papers with respect to modality markers, Gosden (1990) found that over 7% of grammatical subjects consist of these markers that convey writer’s perception of uncertainty. In another study, Gabrielatos and McEnery (2005) analysed MA dissertations written by students at the Department of Linguistics and English Language, Lancaster University and concluded that native speakers of English employed epistemic adjectives, epistemic adverbs and epistemic modals more frequently than the non-native speakers. In his study focusing on stance modals, Kafes (2009) compared three sets corpora including research articles produced by Turkish-speaking learners of English, Spanish-speaking learners of English and native speakers of English with regard to stance modals. He found that non-native speakers of English do not significantly differ in the use of stance modals except may and that there is a statistically significant difference between the native and non-native groups concerning the use of can, could, may, might, would and should. Kanté (2010) investigated head nouns across legal and academic corpora and revealed that head nouns controlling that-clauses such as fact, statement and assumption might be used for different communicative functions and effects concerning text genres.

Biber et al. (1999: 671) propose that the adjectives that control a that complement clause convey stance and that they fall into three major semantic domains: degrees of certainty (e.g. certain, confident, evident); affective phonological states (e.g. annoyed, glad, sad); and evaluation of situations, events, etc. (e.g. appropriate, odd, good, important, advisable). According to them, certainty adjectives indicate the degree of certitude of the proposition presented in the that-clause or they can reflect the writer’s assessment of a position held by some third person. The following statement was taken from Longman Grammar of Spoken and Written English (LGSWE) to exemplify the latter (Biber et al., 1999: 672).

e.g. They are undoubtedly right that it has now become clear that the government will not pay for the expansion it desires (NEWS).

They point out that adjectival predicates controlling extraposed that-clauses signals a stance or attitude towards the proposition conveyed in the that-clause and that they mostly represent the attitude of the speaker/ writer even though they are not overtly attributed to him/ her. Their study revealed that certainty adjectives constituted the majority of adjectival predicates controlling these clauses.

3. Research Design

3.1. Research Questions

This study is intended to provide answers to the following research questions:

1. Do Turkish-speaking academic authors of English and native academic authors of English significantly differ with respect to the use of epistemic adjectives?

2. Do Spanish-speaking academic authors of English and native academic authors of English significantly differ with respect to the use of epistemic adjectives?

3.2. Methodology

The present study is corpus-based in design. It primarily investigated author stance across doctoral dissertations of native and non-native academic authors of English. It includes the analysis of the epistemic adjectives that are frequently reported to occur in the written academic registers in Biber’s study (2006), as illustrated in Table 2.

Table 2. Epistemic Adjectives (Adapted from Biber, 2006: 92)

Download as

Veiw figure View Table

3.3. Data

Three sets of corpora were built with the collection of 133 doctoral dissertations produced by Turkish-speaking Academic Authors of English (TAEs, henceforth), Spanish-Speaking Academic Authors of English (SAEs, hereafter) and Native Academic Authors of English (NAEs, henceforth) within the field of English Language Teaching, English Language and Literature, Applied Linguistics and Modern Languages [TACE (Turkish Academic Corpus of English): 48; SACE (Spanish Academic Corpus of English): 43; NACE (Native Academic Corpus of English): 45]. Most of the dissertations in NACE were submitted to four different universities in the UK while 11 were submitted to five different universities in the USA, one to the one in Netherlands and one to a university in Australia. The majority of the dissertations in TACE were submitted to seven different universities in Turkey and one to a university in Australia. As for SACE, thirty dissertations were submitted to 14 different universities in Spain, eight to six different universities in the USA and six to five different universities in the UK. All were written between 2005 and 2012 and for practical reasons, the corpora were constructed with the collection of dissertations available online. Assuming that those who supervised the dissertations might have influenced the writing process of the sections Abstract, Introduction, Review of Literature, Methodology and References, they were excluded from the corpora. Instead, data analysis was realised through the sections Findings, Discussion, Conclusion, Pedagogical Implications (Implications to English Language Teaching) and Suggestions for Further Research. Finally, all figures, paraphrases, quotations, tables and titles were excluded from the above-mentioned sections of the dissertations in the corpora. Table 3 shows the size of corpora across which epistemic adjectives were investigated in this study.

Table 3. Corpus Size

Download as

Veiw figure View Table

3.4. Data Analysis

Since our study aimed to reveal whether TAEs, SAEs and NAEs significantly differ from each other in terms of epistemic adjectives, the Contrastive Interlanguage Analysis (CIA) (Granger, 1996) was used as method of data analysis procedure. The following are the six stages of the process in concern.

1. The Analysis of TACE: The corpus including dissertations produced by TAEs was examined with respect to epistemic adjectives.

2. The Analysis of NACE: The corpus including dissertations produced by NAEs was searched with respect to epistemic adjectives.

3. The Analysis of SACE: The corpus including dissertations produced by SAEs was examined with respect to epistemic adjectives.

4. TACE vs NACE: Findings obtained from the analyses of TACE and NACE were compared to see whether they significantly differ in the use of epistemic adjectives.

5. SACE vs NACE: Results of the analysis administered to SACE and NACE were compared to reveal whether they significantly differ in the use of epistemic adjectives.

6. TACE and SACE: Findings obtained from the analyses of TACE and SACE were compared to find out whether they significantly differ concerning the use of epistemic adjectives.

For the first three stages of data analysis, Wordsmith Tools 6.0, a computer software package developed by Scott (2012), “for analysing the lexis of texts and corpora in order to produce frequency lists, to run concordance searches and calculate collocations for particular words, and to find keywords in a text and examine their distribution” (Baker et al. 2006: 169), was used to identify epistemic adjectives in three corpora. For the last three stages, Log Likelihood Statistics, which was previously used in significant studies conducted in corpus linguistics (e.g. Granger and Rayson, 1998; Scott, 2011), was utilized to compare the frequencies of epistemic adjectives measured in each corpus.

4. Results and Discussions

Table 4. Epistemic Adjectives not Found in Three Corpora

Download as

Veiw figure View Table

This section presents findings drawn from the analysis of three corpora including doctoral dissertations produced by TAEs, SAEs and NAEs. Following Biber (2006), a total number of 26 English epistemic adjectives (under two categories as certainty adjectives and likelihood adjectives) were identified over three corpora and their frequencies were computed via WS Tools 6.0. Subsequently, three corpora were analysed into categorical and individual epistemic adjectives to see whether native and non-native academic authors of English significantly differ with respect to the use of these adjectives. Yet, it might be useful to begin with presenting epistemic adjectives not found in three corpora. Table 4 illustrated the items at stake.

4.1. Epistemic Adjectives in Three Corpora

Three corpora were compared in terms of the epistemic adjectives. Figure 1 shows the distribution of epistemic adjective types found across three corpora.

Figure 1. Epistemic adjective types in three corpora

Download as

Veiw figure View Figure

View next figure

As displayed in Figure 1, epistemic adjectives were found in highest variety in NACE; namely, 19 out of 26 types were used in NACE, 17types in TACE and 15 types in SACE. In parallel with this, they were mostly used in NACE (586 times) while they occurred251 and 192 times in TACE and SACE, respectively. Hence, they were underused by TAEs and SAEs in comparison with NAEs, which is illustrated in Table 5.

Table 5. Overall Frequency Distribution of Epistemic Adjectives in Three Corpora

Download as

Veiw figure View Table

As indicated in Table 5, the highest frequency was observed in NACE, which means there is an underuse in the dissertations produced by SAEs and those written by TAEs against NAEs.

Table 6. LL Ratio of Epistemic Adjectives in TACE and NACE

Download as

Veiw figure View Table

The results showed that epistemic adjectives were underused in TACE with -139.71 LL value (p<0.05), indicating a statistically significant difference between two corpora with respect to epistemic adjectives; thus, the underuse in TACE relative to NACE was approved by LL calculation. Subsequently, SACE and NACE were compared in terms of the overall frequency of the items. Table 7 outlines LL ratio of the frequency in concern.

Table 7. LL Ratio of Epistemic Adjectives in SACE and NACE

Download as

Veiw figure View Table

Table 7 suggests that the underuse in SACE against NACE was confirmed by LL results. Finally, TACE and SACE were compared in order to reveal whether they significantly differ with respect to the use of epistemic adjectives. The test results are demonstrated in Table 8.

Table 8. LL Ratio of Epistemic Adjectives in TACE and SACE

Download as

Veiw figure View Table

As indicated in Table 8, epistemic adjectives were relatively more frequented in TACE against SACE, which was approved by a +7.29 LL value. To this point, frequency categories of epistemic adjectives in three corpora as well as the LL results obtained from the analyses of the native and non-native corpora have been provided. However, this does not allow us to claim whether these devices are used to convey the meaning of “certainty” or “likelihood” throughout the dissertations they were found. Therefore, as a further analysis, three corpora were individually compared with each other regarding certainty and likelihood adjectives.

4.2. Certainty and Likelihood Adjectives in Three Corpora

This section is intended to report the results of the analysis administered to three corpora to investigate whether they significantly differ in the use of certainty and likelihood adjectives. Figure 2 illustrates the related results.

Figure 2. Certainty and likelihood adjectives in three corpora

Download as

Veiw figure View Figure

View previous figure

View next figure

As can be seen in Figure 2, certainty adjectives occurred more frequently than likelihood adjectives in the non-native corpora whereas they were much less frequented than likelihood adjectives in the native corpus. Nonetheless, this does not necessarily mean that certainty adjectives were overused in both non-native corpora against the native corpus. Namely, they were mostly used in TACE with a frequency of 185 while they were seen 158 and 139 times in NACE and SACE, respectively. Thus, they were overused by TAEs and underused by SAEs against NAEs.

The most frequented certainty adjective across three corpora is clear, which appeared 81 times in TACE, 49 times in NACE and 47 times in SACE. The following statements are extracted from each corpus to exemplify the item in concern.

Example 1.

[From considering their experiences, it is clear that the Spanish government needs to radically rethink its plans for the English PAU if it does not want to repeat the mistakes and failures of the past.]

Extracted from <SACE-UPV-2012-DLN>

[Table 5.2 makes clear that if GENRE effects were to be estimated from all data in this table, i.e. by comparing data in cells A to D with data in cells E to H, we would not be comparing like with like.]

Extracted from <NACE-UE-2011-RJA>

[Think-aloud protocols made it clear that gifted learners try to use the exact translation of more complicated expressions they want to use, which makes them fail in activities that are supposed to be quite easy for them.]

Extracted from <TACE-CU-2011-DIA>

It is considered useful to visualize certainty adjective types found across three corpora. Figure 3 displays the frequency distribution of the individual items over three corpora.

Figure 3. Certainty adjectives in three corpora

Download as

Veiw figure View Figure

View previous figure

View next figure

Certainty adjective sure, which revealed the second mostly used item following clear in the native corpus, was relatively less frequented in the non-native corpora. In response, true and convinced appeared the second and third mostly used certainty adjective in SACE, respectively while they were found in much lower proportions in NACE and TACE. It is interesting that obvious, apparent and evident appeared among the most frequented items in TACE whereas they were among the least used items across NACE and SACE.

The most striking finding of the study is probably that the likelihood adjectives were preferred by NAEs much more frequently than TACE and SACE. Namely, they were found in NACE with a frequency of 428 while they occurred 66 and 53 times in TACE and SACE, respectively, corresponding to the fact that they were employed by NAEs approximately four times as frequently the total frequencies of the items seen in the non-native groups. Figure 4 provides the frequency distribution of these adjectives in three corpora.

Figure 4. Likelihood adjectives in three corpora

Download as

Veiw figure View Figure

View previous figure

As suggested in Figure 4, one cannot mention a consistency across three corpora concerning the distribution of the likelihood adjectives except likely, which is the most frequently used item in all corpora. It constituted approximately 71% of all likelihood adjectives in NACE, 70% in SACE and 55% in TACE. Nevertheless, it was remarkably underused by TAEs and SAEs against NAEs. Example 2 is intended to illustrate the item in question.

Example 2.

[It is therefore likely that the repetition task which DEAP effectively became for them was in fact a NWR task as they had no lexical representations available to assist them.]

Extracted from <NACE-NU-2006-CS>

[It is likely that the profile of the ELT preparatory students will change in the following years; therefore, a new interest checklist might be applied to review the changing topics of interest.]

Extracted from <TACE-AU-2008-GE>

[Although the exact reasons for its extinction in that area are not known, it is likely that it died out in competition with the dingo.]

Extracted from <SACE-UZ-2012-MPBA>

The second mostly used likelihood adjective in NACE revealed possible, which is the second mostly used item in SACE and third in TACE. However, as in the case of likely, it was underused by both TAEs and SAEs in comparison to NAEs. This time, it was found in NACE (93 times) approximately five times as often they occurred in the non-native corpora (SACE: 11 times; TACE: 9 times). Another underused item in the non-native corpora against the native corpus is unlikely, which appeared 29 times in NACE, 2 times in SACE and none in TACE.

Probable is not only the mostly preferred likelihood adjective in TACE but it also the only one overused by TAEs against NAEs in this category. For instance, it was seen 21 times in TACE, 3 times in NACE and once in SACE. In return, liable revealed the only likelihood adjective that was overused by SAEs against NAEs. Namely, it occurred twice in SACE, once in NACE and none in TACE.

Finally, a log likelihood test was administered to three corpora to see whether the overuse and underuse values in non-native corpora against native corpus are statistically significant and whether TACE and SACE significantly differ with respect to categories of epistemic adjectives. The first test was applied between TACE and NACE and the results are given in Table 9.

Table 9. LL Ratio of Certainty and Likelihood Adjectives in TACE and NACE

Download as

Veiw figure View Table

Table 9 demonstrates that certainty adjectives were slightly overused in TACE against NACE, which was confirmed with a +1.99 LL value. Likelihood adjectives, on the contrary, were significantly underused in TACE in comparison to NACE. Namely, they were preferred 185 times by TAEs while they were employed 158 times by NAEs. It was again approved by the test results (-298.31 LL value). The analysis in concern was repeated between SACE and NACE and the results are exhibited in Table 10.

Table 10. LL Ratio of Certainty and Likelihood Adjectives in SACE and NACE

Download as

Veiw figure View Table

As suggested in Table 10, both certainty and likelihood adjectives were underused in SACE against NACE. Yet, the value measured between the two corpora in terms of certainty adjectives is not so high as the one calculated in terms of likelihood adjectives. Namely, certainty adjectives were found158 times in NACE and 139 times in SACE whereas likelihood adjectives occurred in NACE over 8 times as frequently in SACE. Both findings were confirmed by LL results. As a last step, LL test was conducted to TACE and SACE in order to reveal whether the differences observed between them are statistically significant. The related results can be seen in Table 11.

Table 11. LL Ratio of Certainty and Likelihood Adjectives in TACE and SACE

Download as

Veiw figure View Table