THE USE OF HEDGES IN THE SPEECH OF ESL LEARNERS

Abstract

This study investigates the use of hedges (mitigating expressions like I think or sort of) in the speech of learners of English at multiple proficiency levels as well as of native speakers. Hedges are used to moderate the force of an utterance or the certainty of its content and therefore play an important role in interpersonal communication. The results of this study show that learners generally underuse hedges in comparison with native speakers, although learners at the highest proficiency level use hedges at a rate comparable to that of native speakers. The range of hedges is generally similar across levels. In both the rate and the range of hedge use, striking changes appear to take place between the second highest and the highest proficiency level. The results are discussed in terms of their importance for comparing data from different task types and discourse types.

Headnote

This study investigates the use of hedges (mitigating expressions like I think or sort of) in the speech of learners of English at multiple proficiency levels as well as of native speakers. Hedges are used to moderate the force of an utterance or the certainty of its content and therefore play an important role in interpersonal communication. The results of this study show that learners generally underuse hedges in comparison with native speakers, although learners at the highest proficiency level use hedges at a rate comparable to that of native speakers. The range of hedges is generally similar across levels. In both the rate and the range of hedge use, striking changes appear to take place between the second highest and the highest proficiency level. The results are discussed in terms of their importance for comparing data from different task types and discourse types.

Key words: hedges, English as a Second Language, discourse, task type, monologic data

En este estudio se examina el uso de estrategias mitigadoras (partículas de atenuación como I think o sort of) en el discurso de los estudiantes de inglés en varios niveles de competencia, así como de los hablantes nativos. Las estrategias mitigadoras se utilizan para mitigar la fuerza de un acto de habla o de la certeza de su contenido y, por lo tanto, juegan un papel importante en la comunicación interpersonal.

El análisis indica que los estudiantes generalmente utilizan menos estrategias mitigadoras en comparación con los hablantes nativos, aunque los estudiantes de nivel de competencia más alto utilizan estrategias mitigadoras casi frecuentemente como los hablantes nativos. La variedad de estrategias mitigadoras es generalmente similar en todos los niveles. Sin embargo, se observa un cambio dramático en la tasa y en la variedad de uso de estrategias mitigadoras entre los estudiantes de nivel de competencia más alto y los del segundo más alto. Los resultados se discuten en términos de su importancia para la comparación de datos de diferentes tipos de tareas y discursos.

Palabras clave: estrategias mitigadoras, inglés como lengua extranjera, discurso, tipo de tarea, datos monológicos

1. Introduction

Pragmatic competence has been defined as "the ability to communicate your intended message with all its intended nuances in any socio-cultural context ant to interpret the message of your interlocutor as it was intended." (Fraser, 2010: 15) As such, pragmatic competence is essential for successful communication and is a fundamental part of general communicative competence1, which is the goal towards which much language learning and teaching strives. In fact, the research has shown that pragmatic errors can be perceived as more disruptive for communication than grammatical errors (Bardovi-Harlig & Dömyei, 1998).

One aspect of pragmatic competence is the ability to use vague language. Vague language has several possible pragmatic functions; it can be used to mark politeness, to indicate solidarity with one's interlocutor, or to soften a request (Yates, 2010). Vague language is often produced through the use of hedges, which are single- or multi-word expressions used to indicate uncertainty about the propositional content of an utterance or to diminish its impact.2 The ability to use hedges appropriately can cause difficulties for second language learners.3 Unfortunately, like other aspects of pragmatic competence, it does not generally receive enough attention in second language teaching (Fraser, 2010). Learners, even those at high proficiency levels, may know how to make their language more precise but not when or how to make it "strategically vague" (Yates, 2010: 297). This lack of familiarity with vagueness results in L2 speakers whose speech is grammatically correct but who nevertheless do not succeed in accomplishing their communicative goals (Fraser, 2010: 15).

Although the use of vague language poses a challenge for learners, it has received comparatively little attention in research on L2 pragmatics (Yates, 2010). The present study aims to contribute to the research on hedge use by second language learners and to fill several gaps in the previous research. This study examines both the rate and range of hedge use by learners of English at varying proficiency levels and includes comparison data from native speakers. This study also examines whether the type of task being performed affects hedge use.

2. Background

This section reviews the findings of previous research relating to hedges and the other variables discussed in this study. After a brief overview of the possible functions of hedges, recent work on hedge use by learners of English is reviewed. In the next two subsections, relevant studies relating to monologic data and task type are summarized. The final subsection discusses Hasselgreen (2004) in detail, whose work serves as a starting point for the current study.

2.1 Classifications and Functions of Hedges

Channell (1994) proposed that vagueness, rather than being something that speakers should avoid, was instead a crucial component of communication. Since Channell's landmark study, various characterizations of hedging can be found in the literature, but there is general agreement that hedging is a "rhetorical strategy that attenuates either the full semantic value of a particular expression...or the full force of a speech act..." (Fraser, 2010: 15).

Some discourse effects that arise from the use of hedges include vagueness, evasion, equivocation, and politeness. One class of hedges, general extenders (expressions like and stuff), has been identified as having a primarily interpersonal function-to mark intersubjectivity between speakers (Overstreet & Yule, 1997; Overstreet, 1999; Overstreet, 2005).

Another type of hedge, vague language referring to quantities, makes utterances less assertive by decreasing their exactness. This in turn makes the language more polite. As a face-saving strategy, vague quantity references are employed in apologies, promises, self-justifications, and giving advice (Ruzaite, 2007). Vague language of this type can also perform an interpersonal function, implying that there is no need for specific references since the speaker and their interlocutor understand each other so well (Drave, 2001).

2.2 Hedges in Recent Research on English as a Second or Foreign Language

Research studies on the use of hedges among learners of English as a second or foreign language are relatively limited compared to the number of studies on hedges in the speech and writing of native speakers of English. Much of the previous research on hedges by ESL learners focuses on second language (L2) writing. An early study by Ventola and Mauranen (1990) on Finnish learners of English indicates a limited range of hedging expressions in academic writing. These findings are corroborated by more recent studies by Hyland and Milton (1997) and Hinkel (2002, 2005) who point to the underuse of hedges in L2 writing compared to that of native speakers, and the general difficulties that L2 writers of academic English have in conveying doubt and uncertainty.

More relevant to the current study is L2 research on hedges in ESL learner speech production. Nikula (1997) analyzes hedging in the conversational speech of Finnish learners of English in comparison to that of native speakers of English. Her findings indicate that learners used hedges at a significantly lower rate than native speakers and with much less variety. Certain hedges that were very common in the speech of native speakers, including sort of and kind of, were almost completely absent in the interviews with ESL speakers. Moreover, Nikula (1997) found qualitative differences between native and non-native speakers: learners tended to overuse I think and use it only at the beginning of utterances while native speakers tended to use this hedge in combination with others at a variety positions throughout utterances.

Yu (2009) builds on some of the findings of Nikula (1997) by examining the pragmatic development of hedging by EFL learners. She analyzes written questionnaires, simulated debates, and oral interviews of 211 Chinese learners from junior high school, high school, and university-level English courses in China. She categorizes hedges into quantificational approximators, performative shields, modal shields, pragmatic-marker hedges, and other syntactic and discoursal hedging strategies (127). Results of her study indicate that learners progress from a system in which learners use only performative I think to an intermediate stage with combinations of intensifiers and mitigators before they progress to an advanced system at the university level in which they exhibit awareness of hedging clusters that draw on all categories of hedges. Moreover, Yu (2009) notes that the results of her study may have been affected by the tasks that learners were asked to perform; she points out that there were significant differences in the frequency and range of hedges between the oral interviews with the teacher and the debate task (222). In addition, she notes the lack of native speaker data for direct comparison with the results of her study (2009: 220).

Another more recent L2 study of hedging is Hasselgreen's (2004) book, which the current study follows most closely in its quantitative analysis. Her investigation of hedges is situated within a larger investigation of what she calls 'smallwords.' Hasselgreen defines 'smallwords' as ''small words and phrases, occurring with high frequency in the spoken language, that help to keep our speech flowing, yet do not contribute es- sentially to the message itself" (2004: 162). She identifies 19 smallwords in her corpus, which is comprised of oral test data from Norwegian students learning English, as well as native speakers of English.

Hasselgreen offers a detailed account of the ways in which smallwords function and situates her discussion of the function of smallwords within the framework of Sperber and Wilson's (1986) relevance theory. She proposes that smallwords might contribute to the success of communication in five ways (2004: 146):

1. they express the communicative intention of the speaker, with respect to what is to be communicated and how it affects the interactional roles of the participants.

2. they point to the textual context in which an utterance has relevance.

3. they indicate the cognitive effect of the preceding utterance.

4. they indicate the degree of vagueness or commitment.

5. they indicate the state of success of the communication, acknowledging it or appealing for confirmation or assistance in bringing it about.

Hasselgreen discusses these five tasks both in terms of how they might relate to Schiffrin's (1987) planes of talk as well as how they correspond to the types of skills which she believes second language learners need in order to give the impression of fluent speech. A discussion of these issues is beyond the scope of the present study. For the purposes of the current analysis, we will focus on the fourth function of smallwords that Hasselgreen proposes, namely, indicating the degree of vagueness or commitment.

Hasselgreen (2004) argues that smallwords contribute to communicative competence by allowing the speaker to express various ' macrosignals,' listed above. Each of the five macrosignals can be sent by one or more 'microsignals.' In the case of the fourth macrosignal, indicating the degree of vagueness or commitment, Hasselgreen identifies only one microsignal, namely, "signaling a softening of the message: hedging" (154). In identifying only this single microsignal, she notes that she is following the approach of Nikula (1996) in not distinguishing between epistemic modals like I think and other kinds of vague language. Although it can be argued that the first type of hedge (epistemic modal) represents a lesser degree of commitment to the proposition, and that the second type (vague language like kind of) indicates uncertainty within the proposition, Hasselgreen does not find this distinction meaningful for her analysis. She argues that both types of hedges have the same effect, that of softening the message. Hasselgreen identifies eight hedges in her data: 1 think, like, sort/kind of, a bit, just, or something, not really, and everything that stuff things4.

Hasselgreen's analysis identifies five functions of smallwords. Although hedging is only one of these functions, it comprises a comparatively large proportion of both the types of smallwords and their rate of use. Hasselgreen identifies 19 different smallwords; eight of them are hedges. Furthermore, hedges account for approximately 46% of all of the smallword tokens counted (Hasselgreen, 2004: 205). Hasselgreen's work therefore showed that hedges are the most widely used type of smallwords.

2.3 Monologic vs Dialogic Tasks

Another variable which might affect oral task data concerns the conditions under which the data are gathered. In the area of language testing, oral data are often subdivided into monologic and dialogic, which refers to the presence or absence of an interlocutor. For example, in a dialogic task, such as a conversation, the examinee speaks with another person present; in a monologic task, such as a narrative, the examinee can complete the task without an interlocutor present. Previous research into monologic versus dialogic testing data has indicated that this variable can produce significant changes in various aspects of the language elicited (Shohamy, 1994; O'Loughlin, 2001). Shohamy (1994) compared the ACTFL Oral Proficiency Interview (OPI) to the ACTFL Semi-direct Oral Proficiency Interview (SOPI). The OPI is a half-hour long oral proficiency interview that takes place with a partner (the interviewer). The SOPI consists of a series of questions delivered by computer in which the examinees record their responses without an interlocutor present. Shohamy found that the two tests had very high concurrent validity for the 40 students in the study, and that the language they produced did not differ in terms of error rate. However, she did find marked differences in the type of communicative strategies they used and in the discourse they produced. The examinees employed different discourse strategies depending on whether they had a speaking partner present or not. The OPI testtakers engaged in "deliberation, avoiding answers, sharing, hesitation, self-correction, disagreeing, qualifying, switch to L1 for communicative purposes, and demands for clarification" (112). The SOPI test-takers, however, engaged primarily in the strategies of "paraphrasing, repetition of phrases in eliciting questions, switch to L1, self-correction, and silence when no answer was available" (112). Shohamy's (1994) research indicates that the conditions under which oral proficiency data are gathered (monologic/dialogic) might be an important variable when examining discourse-related phenomena.

2.4 The effect of task type on language production

The relationship between the type of task learners perform and the type of language they produce has received increasing attention in recent years. In a series of studies (Foster and Skehan, 1996; Skehan & Foster, 1997; Skehan & Foster, 1999; Foster & Skehan, 1999) Foster and Skehan have looked at the effects of various task characteristics and task conditions on the fluency, accuracy, and complexity of the subjects' output. Skehan (1992, 1996, 1998) has argued that these three aspects of learner performance enter into competition with each other and that gains in one area will come at the expense of losses in the others.

In addition to the studies by Foster and Skehan, a number of different studies of second language learners also indicate that different task types result in significant differences in the language that learners produce. Although these studies do not follow Foster and Skehan's particular language-processing approach to tasks, they offer convincing empirical evidence that different tasks do indeed elicit different kinds of language from learners.

Newton & Kennedy's (1996) results indicated that learners used more conjunctions in a shared information task than in a split information task. They attribute this result to the different discourse modes required by the two types of tasks. The shared information task constitutes a descriptive discourse mode, while the split information task is basically persuasive. Newton and Kennedy conclude that it is possible to select tasks in order to elicit particular language structures. Tarone and Parrish (1988) investigated task-related variation in the use of English articles. They found that a narrative task increased the accuracy of article use and attribute this to the fact that successful narration requires more careful attention to the marking of given and new information. Geeslin (2006) concluded that two different tasks provided a different number and type of opportunities to use the two types of Spanish copula, and that the same rules were applied differently by the learners in each task. Wigglesworth (2001) considered scores of ESL learners on five different tasks and found that the presence of more structure on the task resulted in better scores for the subjects on an oral proficiency test. Duff (1986) showed differences in number of turns, words per turn, words per c-unit, and syntactic complexity in interactions between non-native pairs on different tasks. Bygate (1999b) examined differences in grammatical complexity on argumentation and narrative tasks. The results indicated that the narrative task elicited significantly more words per T-unit, but there was not a significant difference in the use of subordination. A further analysis showed that the narrative task produced data that were more 'nominal', while the argumentation task had more 'verbal' output. The analysis revealed that this was due in part to a very frequent use of formulaic expressions such as "I think."

It is evident from the summaries of task-based research above that the type of task that learners are asked to perform can lead to great variability in the results. Task type appears to have a significant impact on the features of language produced (such as relative clauses, article use, or nominal vs. verbal output). The findings to date indicate that task type is a rich area for further research.

3. Research Questions

The current study brings together two different strands of second language acquisition research that have largely been investigated separately: the use of hedges, and the effects of task type. This research should therefore make a contribution to the understanding of the development of this aspect of pragmatic competence among learners of English. The only previous study that has included both of these areas of research is Hasselgreen (2004). However, the data used in the present paper also differ from Hasselgreen's in several important ways, and they represent an expansion of previous research.

Hasselgreen's subjects were native speakers of Norwegian or English. The present paper analyzes data from native speakers of Chinese, Korean, and English. Hasselgreen's data come from a language proficiency test conducted in pairs, while the data in the present paper come from monologic oral tasks. Finally, the previous research looked at two ESL groups and one native speaker group; this study compares a wider range of learners at four different proficiency levels.

These differences in the current study versus previous research may be important factors in the understanding of hedge use by L2 learners.

In terms of different L1 backgrounds, Yates (2010) has noted the highly culture-specific and sometimes unconscious conventions involved in many discourse functions. Therefore the L1 background of the participants could be an especially important variable when investigating pragmatic functions.

This present paper examines the following research questions:

1. How does the rate and range of hedge differ for ESL learners at different proficiency levels, and how does this compare to the use of native speakers?

2. Does the type of task being performed affect the rate or range of hedge use?

4. Methodology

The data used in this study come from a semi-direct (computer-administered) test of oral proficiency. The Test of Oral Proficiency ( hereafter TOP) is used at an American university in order to certify graduate students proficiency in English for positions as teaching assistants. The TOP consists of oral tasks that examinees are asked to perform while their responses are recorded by computer.

The examinees have three minutes of planning time for each item and are allowed to take notes in order to plan what they are going to say. They cannot see or hear the prompt with the information while they are answering, but they can view the instructions about what they are expected to do in the task. Each response is limited to two minutes; that is, the tape cuts the examinee off after two minutes.

This study examines data from four tasks on the TOP: News, Personal, Passing Information, and Telephone. In the 'news' task, the exami- nees are asked to give their opinion about a news item they have read. In the 'personal task', the examinees give a response to an open-ended audio question about their personal experience, such as how they learned English, or who their favorite teacher is. In the 'passing information' task, the examinees relate some information that they have read to someone who has no knowledge of it, such as describing a job notice to someone they think might like to apply for the job. In the 'telephone' task, the examinees leave relay a message to one of their officemates by voicemail.

Two trained raters, who are also instructors in the university's English program, rated the exams. When the two raters disagreed, the exam was sent to a third rater to break the tie. The exams were given scores from 2-6. A set of descriptors for each level can be found in Appendix A. For the purposes of the program, a score of 5 or 6 is considered sufficient for the examinee to be certified to teach undergraduates as a teaching assistant. If an examinee receives a score of 3 or 4, they must enroll in a course in oral English for teaching assistants. Scores of 2 or 6 are relatively rare. Two's are generally given only when the examinee is clearly overwhelmed by the demands of the task and gives little or no response; no data from examinees from this level are included in the current study. Sixes are given primarily to examinees whose second language proficiency approaches native or near-native competence.

Data from 47 subjects at different proficiency levels were analyzed. There were 37 non-native examinees and 10 native speakers who took the exam for comparative and research purposes. All of the nonnative speaker examinees came from either a Chinese or Korean language background. The data for each level were evenly split between examinees with a Chinese or Korean L1 background; that is, there were five examinees with L1 Chinese and five examinees with L1 Korean in each group of ten. The group of seven examinees at Level 6 was made up of three native Chinese speakers and four native Korean speakers. This mixture of L1 backgrounds was chosen in order to balance the data against transfer effects from any particular first language background.

After the test responses were transcribed, the author of this article and a colleague coded the data for hedges using the list provided in Hasselgreen (2004). The hedges coded were I think, like, sort/kind of, a bit, just, or something, not really, and everything/that/stuff things. Both coders read through the entire data set in order to identify the hedges. This allowed for the identification of slight variations in wording; for example, or/'and whatever was included as a hedge. Although it did not appear on Hasselgreen's initial list, the coders considered this use of whatever to be similar to the use of the hedges or something and and stuff. This was the only major modification made to the list. Reading through the data also allowed the researchers to exclude non-hedge uses of the expressions. For example, just in a temporal sense as in the sentence I had just gotten home when the phone rang was not counted as a hedge. In the case of a disagreement, the coders discussed the discrepancy and came to an agreement. Interrater reliability was .89.

5. Results

Results from this study are reported below in the order of the research questions posed above. Discussion of the statistical analysis is provided thereafter.

The first research question addresses the issues of the range and frequency of hedges in the native and non-native speaker groups. General results are provided in Table 1, where the frequency is presented as in Hasselgreen (2004), as the number of hedges per 10,000 words:

The results show that ESL learners generally use hedges at a lower rate than native speakers. However, this is not the case for the highest ESL proficiency group, Level 6, who used hedges at a higher rate than native speakers.

The distribution of the hedges used at each level can be seen in Figures 1 through 5. In each instance, the rate of use of each expression is calculated as a ratio of that expression relative to the total number of hedges used at that level.

The charts show that I think was the most frequently used hedge at all levels except for Level 6, which has just as the most frequent hedge. However, the ESL learners at Levels 3, 4, and 5 had a higher proportion for I think (74-81%) than the Level 6 and native speaker groups (31 and 56%, respectively). The proportion of just rises steadily from Levels 3 through 6, with Level 6 and the native speaker group having nearly iden- tical rates. All groups used 5 or 6 of the 8 possible hedges. The proportion of hedge use accounted for by all other hedges (after I think and just) does not show a consistent pattern.

The second research question addresses the issue of task type effects on the rate of hedges for each level. As Table 2 below indicates, the number of hedges varies by level and task.

Levels 3, 5, and the native speakers used hedges at the highest rate on the News task and the lowest rate on the Personal task, with rates in between for the Passing Information and News tasks. This pattern does not hold for the Level 6 and Level 4 groups, however. The Level 6 group used the highest number of hedges on the Passing Information task and the fewest on the Telephone task. The Level 4 group used the highest number of hedges on the News task and the fewest hedges on the Telephone task.

6. Discussion

The results of data analysis indicate that learners generally use hedges at a lower rate than native speakers, except for the highest nonnative proficiency group, Level 6. There appears to be a dramatic change in hedge use that takes place between the Level 5 and the Level 6 group; overall hedge use more than doubled between these two levels. The Level 6 group even uses more hedges than the native speaker group. The difference may be due to the Level 6 group's use of like, which was used at a rate of 25 (per 10,000 words) by the Level 6 group and at a rate of only 2 for the native speakers. The question remains why native speakers appear to avoid the use of like in the data collected. I can only speculate that native speakers might have reacted to the artificiality and formality of the language testing situation and may have therefore consciously or unconsciously used less formal language.

In terms of the range of hedges used, all learner groups and the native speaker group favored the same two hedges. The hedges I think and just were the most frequently used by all groups. All groups also used a similar general range of hedges, between 5 and 6 out of the 8 possible. Again, the group that exhibited the most differences from the others was the Level 6 group, and the hedge that most distinguished it from the other groups was its use of like.

It is important to consider the results of this study in comparison with the results of Hasselgreen (2004), both because this study relies on her classification of hedges and because these data differs in L1 background and in its monologic nature. When we compare the rates of use per 10,000 words, we find that the highest rates of hedge use in this study (156 and 133) are well below the rate of hedge use Hasselgreen found for her native speaker group (205). This suggests that more hedges may be used on dialogic tasks. It is also possible that this difference is due to another confounding factor, such as L1 background.

In terms of the range of hedge use, Hasselgreen's data showed the same two most frequent hedges, I think and just, as this data. However, in her data, the two ESL groups used I think the most frequently, while for the native speaker group, just was the most frequently used hedge. Since this difference is evident in the native speaker control group, however, it more likely due to the monologic nature of the tasks being performed, or how they were perceived, than to L1 background. The high frequency of I think in the data are also in line with the results of Yu (2009), who found that learners were highly dependent on the use of I think as a hedge but that this dependence gradually started to decrease at higher proficiency levels.

The results for the frequency of hedge use on different tasks show that some tasks generally elicit more hedging than others, although this effect was not seen equally at all levels. The News task, in which the subjects give their opinion on a news item, produced the greatest number of hedges at four of the five proficiency levels. Since the News task asks the respondents to give their opinion, a high rate of use is, perhaps, not surprising. However, it is interesting to note that the rate of hedge use on this task is often more than double that of the Personal or Telephone tasks. These two tasks generally had the lowest rates of use. The generally low rate on the Telephone task can probably be attributed to the fairly simple nature of the message to be passed on. However, it should be noted that the native speakers were an exception to this trend. From my own experience in looking at the data, my impression is that the native speakers were more eager to explain why they were calling their officemates to relay the message. It may be the case that they focused on the possible imposition their call could create rather than on the content of the message to pass on. The Personal task also showed a relatively low rate of hedge use; however, it is difficult to draw many conclusions from this since the Personal task was highly idiosyncratic. When asked to describe one's first day in a different country, some of the respondents engaged primarily in description and others in narration. This type of task might benefit from further qualitative analysis.

Yu (2009) found significant differences in hedge use on a debate task compared with an interview. Yu attributes the differences in rate to a case of misunderstanding by the learners, who thought that they were supposed to be assertive in the debate task and therefore used fewer hedges. Yu argues that the ESL learners misinterpreted the debate task and con- siders this to be an indication of the deficiency of their pragmatic development. However, given the variation in hedge use by task type found in this study, another explanation is possible. The two different tasks (debate and interview) in Yu's analysis may by their very nature lead to the use of a different rate and range of hedges. If this is the case, then the learners did not misunderstand the task, but instead reacted to task demands in an appropriate way. Research would need to be conducted on these task types that includes native speaker comparison data in order to explore this possibility further.

7. Conclusions

This study examined the use of hedges by learners and native speakers of English performing various monologic testing tasks. The results showed that hedge use increases with proficiency level and that ESL learners do attain native-like rates of hedge use. In fact, the highest ESL proficiency group used hedges at a rate higher than that of native speakers. The range of hedges used by learners was generally similar to that of native speakers in the sense that I think and just were the two most frequently used hedges at all levels. However, the proportion of hedges accounted for by just rose with proficieicy level while the proportion of I think decreased. The present paper also discussed the effects of discourse mode (monologic or dialogic) and task type on the use of hedges. Compared with previous research on dialogic tasks, the rate of hedge use was lower on the monologic tasks in this study. The study found that different tasks elicited markedly different rates of hedge use. The results highlight the importance of considering the variables of proficiency level, discourse mode, and task type when investigating hedge use, and by extension, other discourse pragmatic features.

This study has several important limitations that should be taken into account and that also suggest possibilities for further research. First of all, as discussed above, the data for this project come from monologic tasks that were part of an oral proficiency test. The results are therefore not necessarily generalizable to other contexts or other tasks. As was shown above, the differences between monologic and dialogic data, as well as different task types, are not trivial. Future research that considers the use of hedges or virtually any other aspect of learner language should highlight what type of data and tasks are being analyzed and compare them to others. At the very least, more research studies should at least note the limitations to generalizability based on the context and task type of the data.

A further challenge in all task-based research is the variability brought in by the participant. Duff (1993) encountered the problem that the participant in her study often interpreted the task in some other way, and did something entirely unexpected. I have also encountered this situation, although rarely, in my own data. In the current investigation, the test subjects are asked to talk about their experience learning English in one task, and they almost all do just that. But I found one sample in which the subject instead spoke in abstract terms about why learning English might be useful. It is therefore not always clear that the examinee understands the task that they are asked to perform or that they understand it in the same way as the tester. This is a consideration that further research could examine explicitly using a questionnaire or a think-aloud protocol.

A further possible area for further research is the marked increase in hedge use that was found between the Level 5 and Level 6 proficiency groups, where the rate of hedge use more than doubled. Further research should explore whether this finding can be replicated or whether it is a result of some unknown confounding factor.

Footnote

1 For a detailed discussion of the development of different models of communicative competence, see Gilmore (2011).

2 For a historical overview of research on hedging over the last few decades, see Fraser (2010). For a variety of recent approaches to hedging in the writing and speech of native speakers, see Kaltenboeck et al (2010).

3 See Wishnoff (2000), Hinkel (2005), or Yu (2009) for an overview of research on hedging in ESL and EFL studies. Yu (2009) also includes discussion of studies on the use of hedges among native speakers of English.

4 Hasselgreen notes that these smallwords were combined into one group since they are both formally and functionally very similar.

References

References

Bardovi-Harlig, K., & Dömyei, Z. (1998). Do language learners recognize pragmatic violations? Pragmatic versus grammatical awareness in instructed L2 learning. TESOL Quarterly, 32, 233-259. http://dx.doi.org/10.2307/3587583

Bygate, M. (1999). Quality of language and purpose of task: Patterns of learners' language on two oral communication tasks. Language Teaching Research, 3, 1852-14.

Channell, J. (1994). Vague language. Oxford: Oxford University Press.

Drave, N. (2001). Vaguely speaking: A corpus approach to vague language in intercultural conversations. Language and Computers, 36, 25-40.

Duff, P. (1986). Another look at interlanguage talk: Taking task to task. In R. Day (Ed.), Talking to learn: Conversation in second language acquisition (pp. 147-181). Rowley, MA: Newbury House.

Duff, P. A. (1993). Tasks and interlanguage performance: An SLA research perspective. In G. Crooks & S. M. Gass (Eds.), Tasks and language learning: Integrating theory and practice (pp. 57-95). Clevedon: Multilingual Matters.

Foster, P, & Skelian, P. (1996). The influence of planning and task type on second language performance. Studies in Second Language Acquisition, 18, 299-323. http://dx.doi.org/10.1017/S0272263100015047

Foster, P, & Skelian P (1999). The influence of source of planning and focus of planning on task-based performance. Language Teaching Research, 3, 215-247.

Fraser, B. (2010). Pragmatic competence: The case of hedging. New approaches to hedging (pp. 15-34). Bingley, UK: Emerald Group Publishing.

Geeslin, K. L. (2006). Task design, discourse context and variation in second language data elicitation. In C. A. Klee & T. L. Face (Eds.), Selected proceedings of the 7th conference on the acquisition of Spanish and Portuguese as first and second languages (pp. 74-85). Somerville, MA: Cascadilla Proceedings Project.

Gilmore, A. (2011). "I prefer not text": Developing Japanese learners' communicative competence with authentic materials. Language Learning, 61, 786-819. http://dx.doi.org/10.1111/j.1467-9922.2011.00634.x

Hasselgreen, A. (2004). Testing the spoken English of young Norwegians: A study of test validity and the role of "smallwords" in contributing to pupils ' fluency. Cambridge: Cambridge University Press.

Hinkel, E. (2002). Second language writers' text: Linguistic and rhetorical features. Lawrence Erlbaum Associates Mahwah. Retrieved from http://www. tesl-ej.org/wordpress/issues/volume6/ej21/ej21r10/

Hinkel, E. (2005). Hedging, inflating, and persuading in L2 academic writing. Applied Language Learning, 15, 29-54.

Hyland, K" & Milton, J. (1997). Qualification and certainty in L1 and L2 students' writing. Journal of Second Language Writing, 6, 183-205. http:// dx.doi.org/10.1016/S 1060-3743(97)90033-3

Kaltenböck, G., Mihatsch, W., & Schneider, S. (2010). New approaches to hedging. Bingley, UK: Emerald Group Publishing.

Newton, J., & Kennedy, G. (1996). Effects of communication tasks on the grammatical relations marked by second language learners. System, 24, 309-322. http://dx.doi.org/10.1016/0346-251X(96)00024-3 lX(96)00024-3

Nikula, T. (1996). Pragmatic force modifiers: A study in interlanguage pragmatics. University of Jyväskylä.

Nikula, T. (1997). Interlanguage view on hedging. In R. Markkanen & H. Schroder (Eds.), Hedging and discourse: Approaches to the analysis of a pragmatic phenomenon in academic texts (pp. 188-207). Berlin: Walter de Gruyter.

O'Loughlin, K. J. (2001). The equivalence of direct and semi-direct speaking tests. Cambridge: Cambridge University Press.

Overstreet, M. (2005). And stuff und so: Investigating pragmatic expressions in English and German. Journal of Pragmatics, 37, 1845-1864. http://dx.doi. org/10.1016/j.pragma.2005.02.015

Overstreet, M. (1999). Whales, candlelight, and stuff like that: General extenders in English discourse. Oxford: Oxford University Press.

Overstreet, M" & Yule, G. (1997). On being inexplicit and stuff in contemporary American English. Journal of English Linguistics, 25, 250-258. http://dx.doi. org/10.1177/007542429702500307

Ruzaite, J. (2007). Vague references to quantities as a face-saving strategy in teacher-student interaction. Lodz Papers in Pragmatics, 3, 157-178. http:// dx.doi.org/10.2478/v10016-007-0011-x

Schiffrin, D. (1987). Discourse markers. Cambridge: Cambridge University Press. http://dx.doi.org/10.1017/CBO9780511611841

Shohamy, E. (1994). The validity of direct versus semi-direct oral tests. Language Testing, IE 99-123. http://dx.doi.org/10.1177/026553229401100202

Skehan, P. (1992). Strategies in second language acquisition. Thames Valley University Working Papers in English Language Teaching, 1, 178-208.

Skehan, P. (1996). A framework for the implementation of task-based instruction. Applied Linguistics, 17, 38-62. http://dx.doi.org/10.1093/applin/17.1.38

Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford University Press.

Skehan, P. (2003). Task-based instruction. Language Teaching, 36, 1-14. http:// dx.doi.org/10.1017/S026144480200188X

Skehan, P, & Foster, P. (1997). The influence of planning and post-task activities on accuracy and complexity in task-based learning. Language Teaching Research, 1, 16-33.

Skehan, P, & Foster, P. (1999). The influence of task structure and processing conditions on narrative retellings. Language Learning, 49, 93-120. http:// dx.doi.org/10.1111/1467-9922.00071

Sperber, D., & Wilson, D. (1986). Relevance: Communication and cognition. Harvard University Press.

Tarone, E., & Parrish, B. (1988). Task-related variation in interlanguage: The case of articles. Language Learning, 38, 21-43. http://dx.doi. org/10.1111/j. 1467-1770.1988.tb00400.x

Ventola, E., & Mauranen, A. (1990). Tutkijat ja englanniksi kirjoittaminen (= Researchers and writing in English). Helsinki: Yliopistopaino.

Wigglesworth, G. (2001). Influences on performance in task-based oral assessments. In M. By gate, P. Skehan, & M. Swain (Eds.), Researching pedagogic tasks: Second language learning, teaching, and testing (pp. 186-209). Harlow, UK: Pearson Education.

Wishnoff, R. J. (2000). Hedging your bets: L2 learners' acquisition of pragmatic devices in academic writing and computer-mediated discourse. Second Language Studies, 19, 119-148.

Yates, L. (2010). Pragmatic challenges for second language learners. In A. Trosborg (Ed.), Pragmatics across languages and cultures (pp. 287-308). Berlin: Walter de Gruyter.

Yu, S. (2009). The pragmatic development of hedging in EFL learners. (Doctoral dissertation). Retrieved from http://dspace.cityu.edu.hk/handle/2031/5731

First version received: June 2013.

Final version accepted: September 2013.