Use of Formulaic Sequences in Low to Intermediate Tunisian EFL Learners

The aim of this study is to explore low- intermediate Tunisian EFL learners’ use of formulaic sequences (henceforth FSs), i.e. conventionalized and memorized combinations of lexis and/or grammar, here with a focus on oral production of English as a foreign language. The primary material consists of the informants’ oral performances in three different elicitation tasks, i.e. the role-play, the free talk and the picture story telling. The results of the study show that low-intermediate Tunisian EFL learners use the idiom principle (cf. Sinclair 1991), involving a single choice of retrieving a memorized sequence. The type of task has an effect on the informants’ oral production and on the number and types of formulaic sequences used. It is suggested that the topic, the speaking conditions and the instruction that learners received could be considered as factors influencing the use of FSs. There are many instances of deficient FSs which bear traces of formulaicity from the learners’ L1 and L2. Results show that there is indeed a significant correspondence between knowledge and use of formulaic language and oral proficiency. The more the learner uses formulaic language, the more the oral performance is fluent and accurate and the higher is the grade. 1. Introduction

1.1. Theoretical background
In the last three decades, there has been much interest within the second language teaching community in the phenomenon of formulaic language. Researchers have increasingly focused on actual performance, i.e., language production, and the large store of and access to “formulaic sequence” (hereafter FS), i.e. conventionalized and memorized combinations of lexis and/or grammar, facilitating linguistic decoding and encoding and hence promoting fluency and idiomaticity (Pawley and Syder 1983, Sinclair 1991, Erman and Warren 2000, Wray 2002). In the present study, FSs in Tunisian low-intermediate learners’ use of English as a foreign language (EFL) are analyzed in order to explore how learners deal with them. This is investigated by examining learners’ oral production in the target language as demonstrated in three types of tasks, i.e., the role-play, the free talk, and the picture story telling.

All approaches to the study of formulaic language stress the importance of their functional aspect, i.e., the fact that certain language sequences have conventionalized meanings which are used in certain predictable situations. Pragmatic approaches to the study of formulaic language therefore provide a very important basis for their identification and categorization (Coulmas 1981). However, this perspective needs to be complemented by psycholinguistic approaches, given that all studies invoke processing terminology. Psycholinguistic approaches are also interested in the acquisition of formulaic language. The present study falls within the theoretical framework of psycholinguistics, which includes both a linguistic and a learning perspective, and which sees learning, knowledge and production as closely related. This approach, however, faces considerable methodological difficulties. The definition and identification of FSs are considered the most challenging tasks due to the elusive character of formulaic language. Indeed, what makes identifying FSs even more difficult is the assessment of the extent to which a sequence is conventionalized.

Language is assumed to be either holistically produced, that is memorized and retrieved whole from memory or at least easily accessed, or analytically produced, that is assembled through the use of grammatical rules and lexis. However, the relative nature of formulaic language prevents a dichotomy between analytically and holistically produced language. Moreover, there are extensive issues concerning storage and retrieval of language in general and FSs in particular. Nevertheless, addressing these issues in a comprehensive manner is beyond the scope of the present study. The problem of identifying FSs is brought to a head when dealing with interlanguage, and is most likely one of the reasons that FSs in second language (L2) and foreign language (FL) is an under-researched field.

An attempt to investigate learners’ interlanguage is complicated by the variation that characterizes such language. Many factors motivate research on formulaic language, not only in L1 but also in L2 and FL: the pervasiveness of FSs in L1 (Erman and Warren 2000, Hoey 2005), the processing advantages of formulaic language in L1 and L2 (Conklin and Schmitt 2007), the related economy of facilitating comprehension and production (Wray 2002), speech fluency enhancing by the use FSs (Wood 2006) and the elucidation of L2 learners’ levels of proficiency in the target language (Lewis 2008). It is obvious that FSs are tricky and hard to detect; however, despite their elusive character, they appear to be an essential and frequent feature of language production (Sinclair 1991, Erman and Warren 2000, Wiktorsson 2003). The use of FSs can also reflect and clarify aspects of fluency, idiomaticity and oral proficiency in the EFL learners’ spoken language, as found in the present study. 1.2. Aim of this study

The 47 informants in this study, who are pupils studying in the second year of secondary education in a Tunisian school, were assigned three different types of elicitation tasks, i.e. the role-plays, the free talk and the picture story telling. The primary aim of this study is to explore the use of FSs in low-intermediate EFL learners’ oral production, thereby, providing some insight into the linguistic units stored in their mental lexicon and used in the processing of language.

By exploring the quantity of formulaic types (lexical FSs, grammatical FSs, pragmatic FSs) and categories (section 2.5) in the informants’ oral production in the three elicitation tasks, the following research questions are addressed: 1/ To what extent do the Tunisian low- intermediate EFL learners use formulaic types in their oral production while performing each of the three elicitation tasks used, i.e. the role- plays, the free talk and the picture story telling? 2/ Does the type of task have an effect on the learners’ oral performance and on the number and type of FSs used? What are the similarities and differences at the level of distribution, frequency and number of occurrences of FSs between the different elicitation tasks? 3/ To what extent are deficient FSs used, and what is the correspondence between the learners’ use of FSs and their oral proficiency in the target language?

These questions are important for the development of the field of EFL teaching and learning. In fact, by addressing the first question, the oral performances of the participants will be analyzed and the FSs will be detected then classified into types and discussed. The second question seeks to examine the proportions, frequencies and distribution of formulaic categories in the informants’ oral production in each type of task. Moreover, the study explores whether the type of task has an effect on the learners’ oral performance and on the number and types of FSs used. An attempt to account for the differences found between the three tasks will be suggested. Finally, the third question will be devoted to examining the learners’ deficient FSs (so called because they suffer from lexical and/or grammatical deficiencies not matching the native target) used in their oral production in the three elicitation tasks. Furthermore, a brief account of how the learners’ use of FSs relates to oral proficiency will be presented. 2. Methods

2.1. Elicitation tasks
Since the study is interested in formulaic language in oral production, different elicitation tasks – that would show what the learners could do in performing them – were devised. And because one of the questions to be researched is the effect of task on performance, more than one task type was needed. In fact, three different types of task were used in the present research. Finally, as FSs needed to be identified and classified as accurately as possible, there was a vital need for a whole process of judgement that involved a balancing of criteria for identification and classification. The nature of this study requires working with a relatively large number of subjects. It was decided that 47 pupils would participate in this study. The pupils (having an average age of 16) are studying in the second year of secondary education. They belong to 3 distinct classes at a secondary school in Sfax Tunisia. All the participants have been learning English as a foreign language for five years.

They have been taking a compulsory English course for an average of 3 hours a week during these years of studying. Oral proficiency in this study relates to the learners’ speaking ability in English as assessed by the teacher and in accordance with the educational level of their English as an FL. For the purpose of gathering data, a structured approach is adopted, drawing on transcriptions of spoken discourse. It was felt necessary to collect samples of language use by recording the informants’ oral production in each of the three elicitation tasks. A great deal of data was collected in the speaking tests administered in each session of the English course and in the classroom speaking activities carried out during lessons. The classroom was a good and a more or less natural setting where pupils produced oral language. The tasks that were decided upon in the end were thought to give a good picture about how learners could handle FSs in their speech. In other words, whether the subjects use the same proportion and types of FSs in the different tasks they were asked to perform has to be made clear. In looking for ways of creating more varied forms of interaction in the classroom, it was decided to turn to the task of ‘role play’. With this technique, learners were asked to imagine themselves in a situation which could occur outside the classroom.

They were asked to adopt a specific role in this situation. In some cases, they may simply have to act as themselves. In others, they may have to adopt a simulated identity. And more importantly, pupils were asked to behave as if the situation really existed, in accordance with their roles. The type of role- playing which was used to elicit data was controlled by the subjects inside the classroom environment, which was exploited as a social context for foreign language use. The topics of the role-plays were familiar to the participants as they are taught within the second year of secondary education curriculum. The role-plays were transcribed and this resulted in an 8850 word learner corpus. The ‘Free talk’ task was deemed to be a suitable one because it enabled subjects to speak freely without any constraints or pressure as they were allowed to talk about many familiar topics. First, each subject was asked to introduce him/ herself and to talk about his / her hobbies, likes and dislikes.

Then, every participant was given a list of written themes that were already studied during the school year. The topics that were included are dreams and ambitions, dream job, fame , good teachers, real friendship, family life, family relationships, family problems, child labour, money and evil, reading books, surfing on the internet, libraries, working mothers, school uniforms, the environment, pollution, health, travelling, holidays. The majority of the subjects proved to be fluent and succeeded to produce extended flows of speech. They were able to elaborate on the different topics they were asked about. The participants’ responses were taped and subsequently transcribed. This resulted in a 9471 word learner corpus from the ‘free talk’ task. The third task is the ‘picture story telling’ task. It was chosen on purpose in order to allow for a comparison between the learners’ performance in the different tasks that were opted for. Having for objective to answer the third research question ‘does the task type have an effect on learners’ oral performance, and on the number and type of FSs used?’, it was decided to choose the ‘picture story telling’ task which is different from the ‘role-play’ and the ‘free talk’ tasks.

What is worth mentioning is that almost all the subjects did not succeed to produce very extended flows of speech compared to the ‘free talk’ task. Participants could not elaborate more on telling the stories and could not make their oral production a richer one like the previous task, which resulted in obtaining just a 2477 word learner corpus from the ‘picture story telling’ task. The subjects’ oral production was recorded during three days in the end of the school year.

A timetable outside classroom time was set up for recordings and participants chose a time convenient to their own personal schedule that would ensure no time pressure on the research task. It was not possible to record the elicited data inside the teaching room due to time constraints and curriculum restrictions. Thus, it was decided to use one adjoining non- teaching room to allow one candidate per session of recording, but ensure privacy for each one during the performance. The data was transcribed in orthographic words, which is the most widely used method of transcription. In fact, since phonetic features such as pronunciation were not targeted for investigation in the present study, it was decided to discard phonetic transcription. 2.2. Definition and defining criteria used to identify FSs

In modern validity theory in educational measurement, a crucial step initially is to define the construct at a conceptual level. This then provides a basis for judging the adequacy of operational measures of the construct. In the case of FSs, Wray (2002: 9) has proposed a definition which is likely to be very influential.

A sequence, continuous or discontinuous, of words or other element, which is, or appears to be, prefabricated: that is, stored and retrieved whole from memory the time of use, rather than being subject to generation or analysis by the language grammar. A point of departure for the present study is Erman and Warren’s definition of a prefab (2000: 31): “A prefab is a combination of at least two words favoured by native speakers in preference to an alternative combination which could have been equivalent had there been no conventionalization”. Conventionalization is a key word that is difficult to define. FSs are identified within the sentence but the topic is considered as a contributing contextual feature.

The word “combinations” can be continuous or discontinuous, and frequent co-occurrences may be indicative of conventionalization. Lexical, syntactic and / or contextual restrictions may occur, making the sequence formulaic, i.e. a habitual way of expressing meaning and hence an additional sign of conventionalization. FSs are identified empirically as the combinations of words that in fact recur most commonly in a given register. Indeed, to qualify as an FS, a word combination must frequently recur in a register. Restricted exchangeability is also used as a criterion to single out FSs, as in Erman and Warren’s study (2000). The reader is reminded that restricted exchangeability means that one word in an FS cannot be exchanged without affecting the function, meaning, and/or idiomaticity of the sequence (2000: 32). Slots generated by FSs are identified, and although the FSs are open in theory, at times contextual features impose restrictions. 2.3 Identifying FSs

FSs are ubiquitous in language (e.g. Erman and Warren, 2000; Wray, 2002; Pawley and Syder, 2000 ) but at the same time they present researchers, especially in the areas of NLP , descriptive linguistics and language acquisition ( see for example Sag et al, 2002; Wray, 2000, 2002 ) with a number of challenges. Two of the most serious challenges are the identification and definition of FSs. These are interdependent and cause a circular problem. As long as we cannot identify and describe the properties of FSs fully, a definition remains only partial and, in turn, without a full definition the identification process is incomplete.

Nevertheless, methods of identification have been developed and used, based on broad criteria, e.g. human judgement, frequency information or semantic and grammatical properties (e.g. idioms, adjective- noun collocations). In its present stage of development, the study of FSs still faces fundamental problems in identifying the units of analysis within a data base or corpus. Wray (2002) gives a comprehensive discussion of the criteria that have been proposed or applied in previous research. According to Erman and Warren’s defining criteria, determining whether a stretch of words constitutes an FS is still difficult. In the following sections some aspects to consider when identifying FSs are discussed and means of identification are accounted for. 2.3.1 Phonological coherence

In the case of spoken language, certain phonological features have been investigated as possible indicators of FSs. These include speech rate, pausing, stress patterns and clarity of articulation. Indeed, in speech production, FSs may be uttered with phonological coherence (Coulmas 1979; Wray 2002), with no internal pausing and a continuous intonation contour. There are certain variables that need to be controlled in the interests of internal validity, such as whether the talk is spontaneous or prepared, what the topic is and the nature of the speaking task to be performed. The judges (the researcher and two other English teachers) used not only the transcription of the elicited data but also listened to the recordings while trying to identify FSs. 2.3.2 Pragmatic analysis

Pragmatic or functional analysis is another criterion which recognizes that FSs have important roles in the performance of speech acts. It also acknowledges that FSs are commonly associated with particular speech events. Thus, the knowledge of the social context in which particular formulaic expressions are used is needed to be able to understand their roles in discourse. What is significant here is that the pragmatic analysis is vital for the identification of FSs in the different role-plays enacted by the subjects while embracing a great number of social situations. 2.3.3 Greater length / complexity than other output

This criterion was used by Wood (2006) in his study that was undertaken to identify the role of FSs in L2 acquisition, particularly in the development of speech fluency. It was thought that such a criterion is significant and relevant in this study. The Judges were able to see and hear the entire output of a particular participant, which helped a lot in applying this criterion. It was decided that, in combination with the other criteria, this was a key marker of formulaicity.

2.3.4 The taxonomy used by Nattinger and DeCarrico (1992)
Another criterion used for the identification of certain FSs is the taxonomy used by Nattinger and DeCarrico (1992) that includes syntactic strings such as ‘NP + Aux + VP’, collocations such as ‘look at, at midnight, different from’ and lexical phrases such as ‘how do you do?’ and ‘for example’, that have pragmatic functions. However, this taxonomy is limited and not necessarily applicable in every case. Wood (2006) has also used this criterion in identifying FSs in his study, stating that if a sequence that was uttered with phonological coherence matched other criteria and filled into a category in this taxonomy, it might be marked as formulaic. Thus, this criterion was used by the researcher and the two other English teachers as a guide to possible formulaicity in our research. 2.4 Means of identifying FSs

Overall none of the criteria outlined in the preceding section is adequate by itself for the identification of FSs. In order to obtain valid results, more than one form of analysis was applied. The concept of triangulation which has come to be an integral part of the qualitative research paradigm is very relevant here. It was decided that using human judgement to identify FSs in the data is better than corpus analysis computer software for many reasons. First, computer corpus analysis, as Wray (2002: 25 – 28) points out, relies on frequency counts which makes it difficult to determine the distribution of some types of FSs. In addition, Moon (1998) states that FSs are quite variable and can be specific to certain genres or types of discourse. Moon has also found that numerous FSs that are familiar to native speakers do not occur at all even in the mega corpora. Wray (2002: 28 – 30) discusses two other limitations of corpus analysis. One is the big discrepancy in the estimates by different researchers of the proportion of the corpus they analyzed which could be considered to consist of FSs. Leaving aside any problems with the reliability of the individual analyses, there are clearly validity issues related to differing theoretical and operational definitions of formulaicity. Other limitations of the corpus analysis have been mentioned by Read and Nation (2004).

They pointed out that the quantitative evidence supplied by the software needs to be evaluated by the application of human judgement to determine which of the word sequences are formulaic – and if a classification system is involved , which ones fit in which categories. Concordance software, Read and Nation argue, can automatically locate only contiguous sequences. They state that in order to locate non-contiguous ones, it is necessary for the researcher to enter in the search request either a contiguous subpart of the whole sequence or at least one key lexical component of it. This of course assumes the whole sequence is already known to be formulaic (2004: 31). In the same connection, Danielsson (2007) notes that the identification of FSs poses a problem for corpus linguistics and computational linguistics alike; whereas recurring sequences of words can be identified easily, such sequences are unlikely to coincide exactly with what a human researcher would accept as a unit of meaning in a language.

Unless this problem is addressed, computational linguistics in particular will find it difficult to embrace insights into language that prioritize meaning rather than structure (e.g. Sinclair 1991, 2004). For the present study, the specific nature of the type of speech elicited and the relatively small number of samples from each participant meant that frequency alone could not suffice as a criterion for identifying FSs. Indeed, some FSs may be used only once or idiosyncratically in such a situation. The background literature on FSs has been studied by the judges. Thus a common sense of how to apply the judgement criteria has been built. Moreover, a definition of what is meant by an FS was carefully formulated in advance and the investigator communicated the definition to the other judges, who then attempted to replicate the researcher’s identification of the formulaic units. In order to assess and identify and to have a high degree of inter-rater reliability, the judges analyzed the data base and an FS was accepted as formulaic only when all of the three judges identified it as such.

Although some limitations of using human judgement have been pointed out by some researchers, including Wray (2002: 23), it was decided that this was the best means of identifying FS. To address the limitations, the methodology used in the present study takes a number of measures. One concern is that the judges may cause inconsistent judgement as the result of fatigue or alterations in judgement thresholds over time. This issue was addressed in the present study by listening to recordings and reading the transcripts, providing two parallel sources of data for judgement, and by allowing the judges several weeks to work independently on the judging and identification. In this way the pressure of time limits and variations in focus in the judging task were reduced. The second concern is that decisions may be based on intuitions if the appropriate knowledge is not at the surface level of awareness. In response to this concern, the judges read the most salient literature on criteria for identifying FSs and these criteria were used in identifying items as FSs. Features of the recorded speech such as speed and volume changes were also used as guides.

These procedures helped to avoid reliance on intuition and kept the available knowledge about formulaicity at the surface level of consciousness during judgement. The most compelling reason for using human judgement was the fact that this was a corpus of learners’ spoken language and the act of listening to speech and noting intonation and pause patterns cannot be done by machine. Human judgement was required if all the factors relevant to formulaicity in speech were to be determined. However, whatever measures are taken to identify FSs, it is not possible to claim that all FSs have been identified, nor that all of the FSs identified are indeed FSs. As mentioned in the introduction, the relative nature inherent in FSs prevents such a claim.

It is not possible to claim that all FSs have been retrieved whole from memory. Some may have been analytically created and happened to exactly coincide with FSs which could have been retrieved from memory. There is so far no way of accessing what actually goes on in the mind. Only indirect observations can be made about the process of formulaic language production. Erman (2007) has shown that pauses in speech occur predominantly in analytically produced language, not in FSs. Furthermore, Conklin and Schmitt’s study (2007) shows that FSs are read and thus processed faster than non-formulaic language. 2.5 Types and categories of FSs

Most of Erman and Warren’s (2000) and Wiktorsson’s (2003) types and categories of FSs have been adhered to, thus enabling subsequent comparisons with their studies. These are presented and exemplified in this section, starting with lexical FSs, and continuing with grammatical FSs and pragmatic FSs. The examples which are provided to illustrate the different types and categories of FSs are taken from the informants’ oral production in the three tasks. 2.5.1 Lexical FSs

Lexical FSs are characterized by the presence of at least one content word, and denote entities, properties, events, states and situations of different kinds. They can be divided into either notional categories or syntactic phrases and clauses, as exemplified in Erman and Warren’s study (2000). Wiktorsson (2003) chose the latter type, which is also the categorization utilized in this study. Formulaic phrase structures and clauses thus reflect the topic as the learners transform their thoughts and ideas into language. In accordance with the above studies, proper names and titles are excluded. All names and titles are presumably memorized and holistically processed regardless of the number of words (cf. Jackendoff 1995).

Negations are included in the lexical sequence when they add something more than just the negative feature: ‘not so bad’ meaning quite good, while not care about is only a negation of care about and thus excluded. In accordance with the previous studies, the subject is excluded from the FS, except in clausal structures. Examples from the learners’ data are provided below. Clausal structures are minimally made up of a subject and a predicate. They are mainly statements and comments about the topic of the task. E.g.: A friend in need is a friend indeed, when I was younger, I thought about it, it’s nice to see you, I can’t help it, as you like, I couldn’t stop myself from thinking, I don’t know, that’s unbearable, please tell me, nice to meet you, oh my God, for God’s sake, what’s the matter, take it easy. Verb phrases can be either open or closed.

Open verb phrases contain verbs generating slots for partly or completely lexically unspecified objects and complements (X: NP, Xv: VP). E.g.: dream of X, listen to X, chat with X, search for X, know about X, care for X, write about X, have a connection to X, try to Xv, want to Xv, go down to X, use X as Y, be associated with X, lack in X, play a part in X, have an effect on X, manage to Xv, wish to Xv, contribute X to Y, be accused of X, be addicted to X, be deprived of X. Closed verb phrases are either intransitive phrasal verbs or verb phrases with lexically specified objects or complements. E.g.: surf the internet, earn money, share feelings, waste time, achieve dreams, wake up, stand up, sit down, grow up, listen to music, leave home, make efforts, fall in love, go home, feel better, come late, feel down, have fun, resolve problems, spend time, work hard, face problems, fail exams, give advice, realize a dream, ask for help, enjoy rights. The notion of open and closed verb phrases is used in the identification of FSs but, in contrast to those in Wiktorsson’s study, they are not analyzed separately due to the limited scope of this study. Noun phrases can, just like verb phrases, be either open or closed.

Open noun phrases are phrases only partly lexically specified. At least one slot is open to be filled with appropriate lexical material. E.g.: choice of X, feeling of X, effect on X, death of X, lack of X, expenses of X, cause of X, opinion about X, condition of X, source of X, key to X, definition of X, discussion of X, relationship with X, issue of X, benefit of X, means of X, Closed noun phrases are FSs where the constituents are lexically specified. E.g.: hard work, pocket money, school uniform, baby sitter, hard conditions, child labour, news channels, bottle of milk, question and answer, big business, huge profit, dream job, global warming, ozone layer, love story, home work, heart disease, police station, foster home, girl friend, lungs cancer, family income, health hazards, house work, friends and colleagues, potential abuse, serious problem. Like open and closed verb phrases, open and closed noun phrases are used in the identification of FSs but not analyzed separately; In counting the proportion and the frequency of FSs, only the whole categories of noun phrases and of verb phrases are considered without focusing on the subcategories of open and closed verb/noun phrases. Prepositional phrases have prepositions as heads and noun phrases as complements.

They often function as adverbials, typically denoting time in the learners’ speech. E.g.: in the past, for a long time, after a while; on the radio, in modern times, to this day, at some point, by the time, for years, in the beginning, at first, in the future, in particular, in front of, in one’s opinion, in reality, on one’s part, by oneself, in danger, on earth. Adjective phrases have adjectives as heads, intensifiers as modifiers and may have complements. E.g.: very good, terribly boring, too extreme, so sorry, partially blind, very important, too late, so cruel, too young, very interesting, really good, necessary for, harmful for, important for, essential for, equal with, aware of, kind with, conscious of, fond of, responsible for, close to, afraid of . Adverb phrases have adverbs as heads. They constitute a very small group in the informants’ texts. E.g.: almost always, probably not, very carefully ,all the time, all day long, very well, all over the world, one day. Since there are few adverb phrases in the informants’ texts, in accordance with Wiktorsson’s study, adverb phrases and adjective phrases have been conflated. 2.5.2 Grammatical FSs

Grammatical FSs mainly, but not exclusively, feature function words. These words typically belong to a functional category, primarily with an intra-clausal function. Consequently they structure or modify information at phrase level and within the clause. Aspect-forming structures signal the writer’s view of the event; lexically specified but excluding the progressive and simple tense distinction. E.g.: start to, begin to, continue to, used to. Determiners start with a determiner and exhibit restricted exchangeability. E.g.: the same, the other, their own, the first, a kind of, the next, certain types of, the rest of, the second, the only. Intensifiers are adverbs (or adjectives) amplifying or reducing the meaning of other words in the sentence. E.g.: not so, not at all, not that, much more, very much, all the more, not a single (one), so much. Introductors: existentials with there + be and non-referential it +be. E.g.: there are, there was, it was, it is. Links are an integral part of the clause and connect phrases (intraclausal).

Some sequences can function both as grammatical and as pragmatic FSs, connecting clauses, for example: E.g.: like … or, because of, more … than, for example, and after, such as, from … to, so … that, between … and, as against. Mood-forming structures with mainly modal auxiliaries indicate necessity, probability, possibility, etc. E.g.: might be, should do, had to, would rather, seem to, would be, can be, have to. Proforms are general expressions used instead of lexically specified items. E.g.: many people, and so on, he or she, et cetera, all that, most men, some people, all these, those who, it all. Quantifiers partly overlap with the subcategory ‘Determiners’ but form a separate group (Erman and Warren 2000). E.g.: a lot of, one of, a little, most of, a number of, a group of, a couple of, the majority of. Tense-forming: ‘have been’ is included for its multiple usage in the simple and continuous perfect and past perfect as well as passive constructions. In addition, ‘be going to’ is included in accordance with the previous studies. E.g.: have been, be going to. 2.5.3 Pragmatic FSs

Like grammatical FSs, pragmatic FSs constitute a functional category, often multi-functional, which means that the predominant function of the FS in context determines the category. Function is superimposed on the FS. Pragmatic sequences typically structure information beyond phrase and clause level. They characteristically operate on an inter-clausal level, at times outside the syntactic structure, monitoring the information, introducing new clauses or sentences, and linking sentences. Pragmatic clausal structures override clauses within the lexical type because pragmatic expressions are vital functions in communication, as strategies, and as a means of self-preservation (Wray 2002). Most of them are restricted to spoken language and some have functions which could be indicated by punctuation. The following categories are suggested, matching Erman and Warren’s categories (2000): Text monitors structure the text and assist in the progression of the text.

They are inter-clausal, and relate and link one clause to another or one part of the speech to another, signalling a relationship or change in the speech. In this category are to be found lexicalized phrases such as conjuncts, for example ‘on the other hand’ and sentence adverbials, such as ‘in general’. Another group of text monitors are discourse markers, combinations of, for instance, conjunctions and adverbs. *Discourse markers are text- oriented markers which are used as indicators of various kinds of transitions in discourse, for instance making boundaries between topics, between modes of speech. E.g.: and then, and finally, and of course, you know, I mean, as I said, I don’t know *Turn regulators are text monitors marking transitions between contributions made by different speakers. They also serve to regulate who speaks and who listens, and above all, to keep the channel open between speakers and listeners. E.g.: you see, you know, well I think, well you know, and you, what about you. *Repair markers are text monitors functioning in the editing of the discourse. The speaker may wish to change the wording, repair a mistake, or give more adequate information related to the previous discourse. Whatever the reason, this usually brings about a break in the encoding of the message, or a deviation from the ongoing line of argumentation.

E.g.: I mean, you know, well you know, what I mean is, what I want to say is. Metalinguistic monitors, broadly speaking, convey the writer’s attitude to what is being said. They may have a downtoning/upgrading function, be a reflection on linguistic form, convey an attitude to the truth value and typically function as sentence launchers. *Metalinguistic comments are used by the text producer to reflect on what is linguistically expressed, perhaps by clarifying something. E.g.: what I mean, I’d like to briefly comment on, my understanding of. *Hedges are used to indicate that an expression did not exactly match the target and can be used as a safeguard against being misunderstood, thus protecting the speaker. An additional function may be to tone down an assertion. E.g.: in some way, kind of, something of. *Epistemological comments are similar to hedges and phrasal/clausal expressions of certainty and uncertainty about the information given and thus convey the speaker’s stance towards the truth value of what is expressed, i.e. his/her belief, attitude, certainty, uncertainty, reservation, denial, etc.

E.g.: I don’t know why, I think it was, I don’t really know, to my knowledge, this may or may not be, I agree with , I disagree with, I approve of, I disapprove of, I’m sure that. Some epistemological signals start the sentence and thus function as a launching device, supporting the development of the speech. At the same time they are used by the learner to assert, question and deny thoughts and ideas. The subject is typically I and the verb is typically in the present tense followed by that. Using this category could also be a strategy of getting the sentence started and thus something said. They are similar to Biber et al.’s “utterance launchers” (1999). In the present study, they are called “sentence launchers” with reference to Lewis’s (2008) study. E.g.: I think that, I am convinced that, I suppose that, I believe that *Personal comments are not always necessary for the speech or the text but predominantly express the speaker’s or writer’s personal opinion about and reaction to the topic/subject matter. Thus they express the speaker’s or writer’s likes and dislikes as well as preferences.

These comments appear to be important features, especially in the learners’ production (cf. protection of self in Wray 2002). The subject is typically ‘I’. Another similar way of commenting on the events is when the speaker or writer explicitly refers to his/her mental processes, thus signalling a personal opinion or the impact of the event. E.g. I think, I know, I remember, I will never forget. *Attitudinal markers are responses rather than metalinguistic in nature, notably ‘my dear’. Social monitors involve the reader in the text and support the interaction in communication. These pragmatic structures are frequent in the subjects’ oral production and they include numerous subcategories. *Interactives are classified as social monitors in that they elicit audience involvement by calling for action on the part of the addressee, e.g., confirmation of a previous claim.

The most obvious instances include such elicitors as: you see what I mean, you know, you see, the latter two usually with a rising tone. *Feedback signals serve to confirm audience involvement by ensuring that the channel is open between speaker and addressee(s). They are referred to as back-channel signals. E.g.: well no, I see, I suppose so. *Hesitation markers have a social function because they are signals to the addressee that the speaker wishes to continue inspite of some encoding difficulties. Some instances include ‘you know’ and ‘I mean’ which are frequently used to stall for time. *Responses are like feedback signals in that they confirm audience participation, but unlike feedback in that they constitute a turn and are answers to a question. E.g.: oh no, well yes, yes I see, yes I understand. *Performative routines have a very different function from the pragmatic prefabs just mentioned in that they elicit or constitute actual performances like greeting, thanking, leave-taking, making offers. E.g.: good morning, good afternoon, thank you, good luck, how are you, excuse me, I’m sorry, what about. 2.5.4 Deficient FSs

In this section, deficient FSs will be in focus. Both deficient but accepted FSs and those which have been rejected are accounted for and illustrated with examples taken from the informants’ oral production. 2.5.4.1 Accepted FSs

In the present study, FSs suffering from mistakes in concord, tense and aspect are included in accordance with Wiktorsson’s decision of allowing one grammatical error, providing the meaning is unaffected. Including the above categories implies that unaffected meaning is used quite liberally. Even minor deviations in the language may have an effect on the listener or reader. Although considered as somewhat deficient, these FSs are included and kept separate for further analysis and discussion. 2.5.4.2 Unaccepted FSs