Acquisition of Formulaic Sequences
The Role of Input, Use, Proficiency and Individual Learning
MA Thesis Applied Linguistics
Department of Linguistics
Faculty of Arts
0. Abstract 2
1. Introduction 4
2. Background 6
2.1 Historical review: Behaviorism and Universal Grammar 6 2.2 Usage-Based Theory and Emergent Grammar 7
2.3 The Unified Model 9
2.4 Dynamic Systems Theory 11
2.5 Formulaic sequences (chunks) 12
2.6 Research questions and predictions 17-18
3. Method 19
3.1 Subjects 19
3.2 Materials and Procedures 20
3.3 Design and Analysis 21
4. Results 22
4.1 Development of Proficiency 22
4.2 Chunks Acquisition 27
4.3 Individual Differences in Chunks Acquisition 34
5. Discussion 40
5.1 Summary of Results 40
5.2 Overall Chunks Acquisition Pattern 41 5.3 Individual Development and Individual Differences 42
6. Conclusion 45
7. References 47
Unlike most L2 learners, native speakers seem to share knowledge of conventionalized, preferred ways of expression. These preferred ways of expressions (so-called chunks) make language sound fluent and authentic and they are therefore an important aspect of second language acquisition. It is argued that these chunks need regular use and input before they are successfully stored in the language system. Also, proficiency plays a role in chunk
acquisition; proficiency level needs to go up before more difficult chunks can be acquired. This thesis focuses on the development of chunks in writings of students (n=17) who have several hours of English input a week. Taking into account previous studies on chunking, it was expected there is a specific learning pattern; students will most likely start with using easier, shorter chunks and will acquire longer, more complicated chunks once they have received more input and reached higher proficiency levels. The results of this thesis indicate that there is indeed such a pattern. This paper also aimed to investigate individual learning patterns as opposed to group learning. Results of individual learners indicate that they do not always strictly follow the general acquisition pattern which means that additional factors besides use, input and proficiency play a role in second language learning.
In recent history, much has been written about second language acquisition and over the years, a wide range of theories on the topic of second language learning have emerged (e.g. Krashen, 1988; Ellis, 1997, 2000; Klein, 1986; De Bot et al, 2003; Cook, 1993). Besides Universal Grammar, which is still considered an important and valid theory, also other theories on language development have gained ground, leading to different beliefs on how language learning takes place. As a result, there is quite a variety of ideas about language learning and language processing.
A few of these theories prove to be relevant for this paper, in which the acquisition of formulaic sequences (chunks) is discussed. Previous research related to this topic has indicated that regular use and input are vital for successful storage and retrieval of chunks (e.g. Bybee, 2008; Ellis 2002; Verspoor et al, 2011). Also, it assumed that other factors such as motivation, out of school contact, language attitude and language aptitude add to successful chunk acquisition. The interconnectedness of different variables influence and affecting each other is an important aspect of Dynamic Systems Theory (DST), which is one of the main theories that will be discussed in this thesis. Another main theory is Usage-Based Theory, which holds that language patterns will emerge as speakers use language frequently, tying in with the importance of use and input. Through encountering and using language, language learners will not only learn language word by word, but they will also acquire word
combinations (Wray, 2002; Smiskova, 2011). Basically, this underlines the fact that language learning is exemplar driven: through frequent encounters we save exemplars of encounters. […], whether these encounters are words, phrases or complete sentences (Smiskova, 2011). Whereas native speakers seem to instinctively know how and when to use formulaic
In this thesis the development of formulaic sequences as part of learning English as a second language in Dutch Gymnasium (High School) students (n=17) is researched. This research is related to the OTTO project1 which focuses on differences in acquiring English as a second language on regular and bilingual schools. Many studies have set out to compare high input (bilingual education) and low input (regular education), often concluding there is a difference in proficiency between the two groups (see, for example, Huijbregtse, 2001; Admiraal,
Westhof and de Bot, 2005; Verspoor and Edelenbos, 2008). This study has a different starting point, since the focus is on regular students only. In this thesis I aim to investigate not only if this group of students show a general pattern in chunks acquisition (i.e. are easier chunks learned before more difficult ones, related to usage, input and proficiency as proposed in previous paragraphs), but also to see whether individual students follow this pattern and if not, how they deviate from it. Even though it is assumed that this group of students will overall show a similar learning pattern as a group, individual learning patterns might deviate from this general group pattern because b besides use and input, also personal learning behavior might influence second language learning.
All students participating in this study are Dutch and have Dutch as their mother tongue. The students are about the same age and all have three hours of English a week at school. It is expected they will have additional input outside the classroom as well, through social media and TV. The writings, that spanned over three years, were each assessed on proficiency level which was measured by looking at vocabulary, use of complex words and sentence structure but also through examining use of chunks and error patterns. I will start with discussing and elaborating on several relevant second language acquisition theories and provide additional background theory on chunk development and chunk categorization. Then, a detailed
description of the method of this study will be presented, followed by the results of this study. Subsequently, the results will be explained in the discussion section of this paper and finally the conclusion will be presented.
In order to elaborate on the acquisition of formulaic sequences by second language learners, a few theories need some explanation. First, a historical review with a brief discussion of the Behaviorist Approach and Universal Grammar will be presented. Subsequently, Usage Based Theory, the theory of Emergent Grammar (EG) the Unified Model (UM) will be discussed. These theories are important because, as opposed to Universal Grammar, Usage-Based
Theory, Emergent Grammar and the Unified Model consider input and use as the main factors in language acquisition and highlight frequency as a crucial factor in language learning. Thirdly, Dynamic Systems Theory will be discussed in further detail. This theory regards any system, such as a language, as a complex one in which different subsystems influence and affect each other. Finally, I will elaborate on chunk acquisition and categorization.
2.1 Historical Review: Behaviorist Approach and Universal Grammar (UG)
For many years, it was assumed that language was modular and had independent subsystems, which are important aspects of Universal Grammar (UG). Basically, the theory of Universal Grammar, with Noam Chomsky as its main founder, holds that the basic grammatical principles are innate for all human beings, meaning that we all share the same grammar or language device which helps us learn language (Chomsky, 1965). Chomsky presented his theory in the 1960‟s as a counteract against the behaviorist approach, which held that learning is the result of habit formation. This means that more often a word or phrase is encountered or heard, the more easily it will be stored in the brain. Thus, frequent stimuli leads to storage and thus to production; if something was heard and thus used frequently, it became a habit. As a result, behaviorists did not speak of rules, they spoke of habits.
Furthermore, Chomsky introduced the notions of competence and performance, where competence is the abstract underlying knowledge one has of language and performance, which is the actual speech. So, the underlying structure actually enables people to produce surface structures. Finally, Chomsky distinguished between principles and parameters; principles are universal whereas parameters need to be set. In other words, he argues that language is pre-organized but that the exact parameters are set differently for every language. Today, UG is still a highly respected theory but there are also many linguists who argue that language is not a top-down approach but that language structures emerge through use and frequency.
2.2 Usage-Based Theory (UB) and Emergent Grammar (EG)
One of the theories suggesting language is a bottom-up process (instead of top-down as suggested in UG) is Usage-Based Theory (UB). Usage-based theory, which is a
psycholinguistic approach to language learning, argues that language is learned through experience and use. UB‟s main idea is that learners are able to recognize patterns and
subsequently know how store, re-use and generalize them. Usage-based theories of language consequently analyze how frequency and repetition affect, and ultimately bring about, form in language, and how this knowledge affects language comprehension and production (Bod, Hay, &Jannedy, 2003; Bybee & Hopper, 2001; Ellis, 2002, 2008b; Hoey, 2005; Robinson & Ellis, 2008). It is argued that language learners learn constructions through using language instead of already having these structures stored in the brain already, as part of some innate device.
UB is in line with the emergenist approach, which gained ground in recent years. Emergenist theory, as discussed in Hopper suggests that grammar is not a preprogrammed and abstract thing in the mind, but that it emerges in regular patterns, constructions and expressions; it is epiphenomenal, not generative (as is suggested in UG) (Hopper, 1998).
The notion of Emergent Grammar (EG) is meant to suggest that structure, or
regularity, comes out of discourse and is shaped by discourse in an ongoing process. Grammar is, in this view, simply the name for certain categories of observed
repetitions in discourse. (Hopper 1998:156)
Emergent Grammar there is no such thing as an abstract grammar in the mind, but a network of expressions and constructions as a result of an iterative process: its forms are not fixed templates but emerge out of face-to-face interaction (Hopper, 1998: 156) Emergent structures are always in flux; speakers borrow heavily from previous similar communication experiences and grammar is, in this view, simply the name for certain categories of observed repetitions in discourse. Or, as Bybee argues, grammar is viewed as the cognitive organization of one‟s experience with language (Bybee, 2008: 216).
So, central in Usage-Based Theory and Emergent Grammar Theory is the usage-event. After all, in both UB and EG it is argued that language structures emerge from usage events and that these events shape the linguistic system. Usage-based theories generally, and Cognitive Grammar specifically, posit that language builds up a conventional inventory of units (including units that convey grammatical patterns) that a speaker can draw on and put
together for communication (Barlow and Kemmer, 2000). This inventory of units is based on hearing and using the language and through use become entrenched (Langacker, 2008). In UB meaning is central (grammar subserves meaning); the focus is on language production and not on the underlying language system (as is the case in Chomsky‟s UG).
Another significant aspect of UB (and EG) is the idea that language learners often use lower level schemas when producing language. This means that they usually do not use the more abstract constructions in which they just have to fill a gap, but they use specific learned expressions instead. These fixed expressions are called formulaic sequences or chunks (Bybee, 2008). Learners perceive and produce such units repeatedly and store them as a whole. As a result, these sequences processed as a single unit and retrieved as such. L1 learners are naturally exposed to this process, producing and storing chunks goes
automatically. Still, for both L1 and L2 learners it is crucial they are exposed to such units frequently in order to be able to store and reproduce these units in the future. Brian
MacWhinney states that chunking is necessary for language learning since it enables fluent language production (MacWhinney, 2008). After all, it takes less time and effort to retrieve a chunk from memory than to build a sentence from single words. Formulaic sequences will be discussed in further detail later on.
Whereas Hopper and Langacker address how a language system in general has emerged through time, MacWhinney‟s Unified Model (UM) addresses how such a language system is acquired by an individual. The model fits in with UB, since UM also regards input as the source of learning. The model was introduced by Brian MacWhinney and focuses on both first and second language learning. There are two main reasons why MacWhinney developed a model that focuses on both L1 and L2. MacWhinney argues that many of the tasks involved in first and second language learning are similar. The learning of meanings, the memorizing of patterns and speech patterns are all part of L1 and L2 learning. Furthermore, L2 learning is influenced by L1 by transfer so that it is impossible to exclude the structure of an L1 in an L2 learning model. Basically, he states that in the Unified Model the mechanics of L1 learning are a subset of L2 learning mechanics (MacWhinney, 2008).
Basically, according to this model input is compared by looking and accounting for
similarities and differences. The model argues that self-organizing maps (SOMS), which are storage cells that exist at several language levels and can reflect on the language system (with help from a processor). Basically, these SOMs are a representation of a person‟s linguistic knowledge. SOMs work similar to the brain; the units in the maps are linked to each other. So, when information is encountered, the connection between a sets of neurons is
strengthened. So, the more frequent something is encountered, the stronger the connection to a certain linguistic item is and the easier it can be recognized. If specific information is not encountered regularly, the connection will become weaker. Yet again, this process shows the importance of frequency and regular input.
When discussing the language models, it became clear that frequency is an important factor in language development and language processing. Ellis argues that one of the main factors to drive the L2 acquisition process is input frequency because frequency has an important influence on cognitive representations (Ellis, 2002). A high frequency unit has a strong
Learners learn from usage events (encountering and using a set of words) so they eventually recognize repeated patterns and can build on this (Verspoor et al, 2011). (Young) children but also L2 language learners regularly rely on utterances they often hear, utterances that become or are „clustered together‟. So, frequency is necessary for both detecting and storing language, such as formulaic sequences or words. Therefore, the forms that learners learn best are the ones that appear most frequently in the input; they have a so-called high type frequency (Bybee, 2008). For example, the fact that a noun occurs with an article or that a certain word order occurs are constructions that are often encountered and thus become automatized. Automatization is vital for quick retrieval of words and chunks from memory and therefore for fluent language production and comprehension (MacWhinney, 2008).
Rumelhart and McClelland‟s activation theory also stresses the importance of frequency in language learning (Rumelhart and McClelland, 1987). This theory mainly focuses on the importance of cues, which are clues that help interpret language learning. Furthermore, they argue that, besides input and frequency, which are the main sources for learning, also reliability, availability and validity help determine the course of acquisition (Rumelhart and McClelland, 1987). Cue availability refers to cue presence in the input: it is about noticing (both consciously and subconsciously) and therefore learning. Cue reliability refers to
whether or not a cue is used in a consistent way. If a cue is encountered constantly in a correct way it is much easier for a language learner to store this piece of language than when he or she encounters different and inconsistent types of cues because it is then harder to store the correct language use. Cue validity is the sum of cue availability and reliability. Verspoor et al argue that a full valid cue is always available (for communication purposes) and used
consistently (Verspoor et al, 2011).
attention to establish the correct form. As a result, receiving incorrect cues could in fact slow down language learning.
2.4 Dynamic Systems Theory
UB describes the language system as a whole, MacWhinney‟s model addresses the factors that play a role in language acquisition, but neither model specifies how changes may occur over time. Dynamic Systems Theory (Thelen & Smith, 1994; van Geert, 1994) regards any system as a complex and dynamic one. Also language is regarded a complex, dynamic system with its different subsystems relating to and influencing each other. This basically means that there is not just one causal factor in acquisition, there are several (Verspoor et al, 2011). Furthermore, Dynamic Systems Theory (DST) regards a system as changing over time. This fits in with the theories as discussed above since language and language use is always in flux. It is also assumed that language learners build on previous experience; once something is acquired learners are able to take the next step and learn new things (Verspoor et al, 2011).
An important aspect of DST is the idea of initial conditions or the initial state. Some dynamic systems highly depend on their initial state and therefore minor differences at the beginning may have dramatic consequences in the long run (de Bot et al, 2005). This is called the „butterfly effect‟, a term introduced by meteorologist Lorenz, who wanted to account for the huge impact small local effects may have on global weather. This effect refers to the
unpredictability of the development of dynamical systems. Basically, it is very difficult to determine the effect of initial conditions on L2 development (Verspoor et al, 2011) meaning that even though initial conditions are similar for a group of participants, individuals might develop in different ways. In line with this idea is the notion of non-linearity; meaning that learning often does not proceed in al linear, clear-cut way.
start at the same proficiency level and do not all develop in the same way. Therefore, it is interesting to see how students develop individually and in what way they deviate from or conform to general learning patterns.
Variability and Individual Differences
As was previously suggested, all language learners have their own way of learning language patterns. As Verspoor et al argue: „the degree and pattern of variability at the individual level can tell us more about the developmental process, not only when we look at errors, but also when different types of constructions appear in a language” (Verspoor et al, 2010). As discussed in the previous paragraph, language learners will not learn a language in a linear way. Instead, they will develop different aspects of a language at different times. So even though a general learning pattern may seem linear, individual patterns will most likely show setbacks, progress and regress (van Dijk et al, 2008). In a more traditional view variability should be seen as noise in the data that should be eliminated or ignored, rather than being considered to be helpful for the developmental process (Verspoor et al, 2011). Yet, instead of looking at a group as a whole and determine whether or not the group was successful, it might be quite informative to analyze learning patterns individually. It could be very interesting to see how individuals develop and see how new forms and constructions appear or disappear, change, or settle (Verspoor et al, 2011: 22).
Usage Based Theory and the Unified Model go together well on several levels because both theories focus on usage, input and frequency. In addition, both theories state there is no clear division between lexicon and grammar and acknowledge that pieces of language can be acquired and retrieved as a whole. Research has provided evidence that language processing is sensitive to formulaicity and collocation and that these prefabricated pieces of language, more commonly known as „‟chunks‟‟, are perceived as single units. As a result, chunks are stored in the memory as being prefabricated and can be easily retrieved and repeated as such, regardless of their length (Smiskova, 2011).
Furthermore, research has convincingly demonstrated that a major characteristic feature of native speaker language (both written and spoken) is in fact formulaic, pre-fabricated
linguistic development (Eyckmans et al., 2007; Pawley and Syder, 1983). If language can be processed in chunks, it will enhance the processing capacity in the brains so that language can be retrieved more quickly and thus understood more quickly. In other words, use of formulaic sequences helps in the second language acquisition process. Learners will eventually
recognize that certain words are often uttered together and they store this information (for example a noun and its modifiers or articles) as a whole. Also, grammar issues such as pluralization or gender can be learned through chunk use so learners get around grammar problems of this kind (Ellis, 1994).
Different types of Chunks
In her research, Hana Smiskova distinguished different types of chunks, namely A) Chunks with a referential function, B) Chunks with a textual function, and C) Chunks with a
communicative function (Smiskova, forthcoming). Chunk type A consists of word and phrase level chunks, such as compounds, lexical collocations and grammatical collocations but also particles and phrasal verbs. Also, chunks such as a idioms, similes, sentence frames, slot fillers and preferred ways of saying things such as „It‟s hard to explain‟ fall under A. Category B, the chunks with a textual function, consists of complex prepositions, complex
conjunctions and linking adverbials. The final category C, chunks with a communicative function, consists of speech act formulae, attitudinal formulae, proverbs, commonplaces and slogans. The chunks were divided into these categories to make a distinction between
difficulty of chunks, according to length and schematicity. Category A chunks are expected to be acquired first, followed by B and finally C.
longer, more difficult and scarce. Given the background on frequency, use, and input, students will probably acquire chunks in a specific order.
Lexical chunks are often short (2-3 words) and quite common because they occur regularly in everyday speech. They are frequently encountered and therefore not really difficult to acquire. Student will probably already use some of these chunks in their early writings. The grammar constructions are already a bit more difficult because of their complexity, length and use. These chunks are slightly longer than the lexical ones but they are also still quite common in everyday language, simply because constructions such as “I‟d like to” or “I think that” are frequently used by speakers. Subsequently, idioms and preferred sentences are expected to be acquired. These chunk types are often are often a sign of higher proficiency because they are less common than lexical chunks and grammar constructions. Textual connectors are less common than lexical chunks and grammatical constructions as well. Therefore they are expected to be acquired at a later stage than the first two categories. Least common and most difficult are the conventionalized phrases; these chunks are quite scarce in speech because of their length and type; not many speakers use slogans ever other sentence. It will thus require a certain level of proficiency and language contact before these chunks can be successfully acquired. The categories including their chunk types are displayed in the table below:
Lexical combinations of 2-3 words
1 compounds sunbathing
2 lexical collocations heavy rain
3 particles afraid of
4 phrasal verbs blow up
5 bi- and trinomials bed and breakfast
7 complement structures avoid –ing; want to (inf); think that (+clause)
8 slot-fillers even warmer than, better and better, as hard as;
X years old
9 sentence structures the sooner we are finished the sooner we can go
Idioms and preferred ways of saying things
10 idioms to spill the beans
11 similes as old as the hills
12 variable idioms think nothing of –ing
13 preferred sentence stems one thing I know for sure is
14 preferred sentences It`s hard to explain
15 complex prepositions apart from
16 complex conjunctions as soon as
17 linking adverbials in other words
18 textual sentence stems another thing is
Conventionalized phrases with a communicative function
19 speech act formulae to cut a long story short
20 attitudinal formulae to be honest
21 proverbs and proverb fragments When in Rome
22 commonplaces it`s a small world
16 Figure 2.1. Table of different types of chunks
The particular question relevant for this study is in what order these chunks are learned. The L2 learners in this study have three hours of English a week besides some expected additional input from social media. It is expected that these learners will initially use the so-called low order chunks, due to little input and low proficiency. However, the more proficient the students become, the more advanced chunks they will start using.
2.6 Research questions and predictions
As discussed in previous sections, it is assumed that there is a specific order in which chunks are acquired, due to their length, schematicity and presence in speech. Short, common, easy chunks will most likely be acquired before infrequent, long and difficult ones because they will be encountered and used more regularly. Furthermore, it is expected that the more difficult chunks will be acquired when proficiency level goes up. A general chunks acquisition pattern related to use, input and proficiency is thus expected to be found in the data.
Besides looking at chunks acquisition by the group of students in order to find out what the exact order of chunks acquisition is, also some individual learning patterns will be analyzed. Even though all students might collectively show a pattern in which they, as a group, acquire chunks, it could also be interesting to see how individual learning patterns are similar to or deviate from this general pattern. Analysis of individual learning patterns might provide insight into learning behavior. Individual analysis will, for example, show when a student regresses or has difficulty acquiring certain chunk types. This could lead to further analysis as to why this might be the case. In the section on Dynamic Systems Theory it was argued that language learning is highly complex and that other factors besides use and input, even though vital in language learning, might influence the development of the participants. Students will probably show variability in their chunks production because they do not learn in a linear-like manner. In fact, students may differ substantially from each other how fast and what types of chunks are acquired.
1. Is there a general order in how chunks are acquired (related to input, use, proficiency and chunk type)?
2. Do individuals tend to follow this general order when acquiring chunks? Are they
similar in use of chunks (per level) or do they deviate from the general pattern?
This section elaborates on the subjects, materials, design and procedure of this study including an explanation of how data was gathered and analyzed and concludes with a short description of the design of the study.
The students (n=20) participating in this study are native speakers of Dutch and live in the Netherlands. All participants attend a Gymnasium school2, which is a school for the most intelligent, talented students in the Netherlands. These schools select their students according to their CITO3 scores and motivation. Their average CITO score was 547 which is above average. Only students from the first three grades participated in the study, ranging in age from 12-13 (first year students), 13-14 (second grade students) and 14-15 (third grade students).
Besides the „regular‟ VWO program (that allows them to go to university after graduation) Gymnasium students also take courses in Latin and ancient Greek, meaning that language learning often happens in a more traditional, grammar-based way. The students have three hours of English a week and thus have limited school input. However, it is very likely that most students have extra English input through TV and social media.
Students were required to write a short text (approximately 200 words) in English about a certain topic which was provided by the teacher. The topics include vacation, hobbies and admiring someone4. They wrote these texts in a period of three years (from first grade to third grade), producing 3 to 5 writings a year. The students were to write texts in October,
November and December 2007 which were subsequently labeled writing 1.1, writing 1.2 and writing 1.3. In 2008, they were to write texts in February, June and December, so writings 2.1, 2.2, 2.3. Finally, students were to produce writings in January, March, May, June and
All students attend the same school (in the Netherlands)
The CITO test is a test all students in the last class of elementary school take to decide their level for High School education.
November 2009, writings 3.1, 3.2, 3.3, 3.4, 3.5. In total, 170 writings were analyzed. Students usually only had a few minutes to write whatever they wanted about the given topic. Students wrote the texts in class (in a computer room at school) but participation was voluntary so sometimes texts are missing or taken out because they are too short. All students wrote their texts approximately at the same time. When students missed more than 2 texts, they were excluded from the study because results would not be reliable enough to work with. In the end, data of 17 students was analyzed and used for this thesis.
Proficiency was measured according to the writings that were produced by the students. Furthermore, students had to take a writing test which was administered three times a year, at the start, in the middle and at the end of the year. The writing assignments were specifically designed by the researchers in order to generate enough material to judge the writing
proficiency properly. Also, the writing assignments were designed in such a way they would be appropriate for the learner‟s age and within range of expected proficiency. For example, in their first year students were asked to write about their new school, teachers and friends. Other examples are corresponding with a pen pal or describing what one would do when winning a thousand Euros which are more advanced topics.
All writing assignments were holistically scored by a team of evaluators. This team consisted of eight experienced teachers of English, of which three native speakers. Of the other
evaluators, two were native speakers of Dutch and the others were native speakers of Chinese, Portuguese or Spanish. All evaluators live in the Netherlands and are familiar with
Gymnasium schools and education. The texts were assessed by giving scores, where 0 was the lowest and 6 the highest proficiency score. These assessments were then discussed among the teachers in order to reach consensus. During the assessment, the teachers focused on
vocabulary, tense, syntactic complexity and use of authentic expressions in order to successfully scale the writings.
The levels assigned to the texts correspond with the levels used in the Common European Framework of References (CEFR)5. Levels 1 and 2 correspond with level A1 in the CEFR.
Level 3 corresponds with level A2. These levels are really basic, with basic grammatical structures and the easier, basic chunks. Level 4 and 5 correspond with level B1 of the CEFR. Language use starts to become more complex; more complex sentence constructions are used and accordingly more complex chunks are produced. Level 6 relates to level B2 in the CEFR. Students at level 6 start to use authentic, „real‟ and creative language and make us of more complicated, longer chunks. The B2 level is pretty advanced and none of the 17 students managed to produce a level 6.
Each text was analyzed and all chunks were categorized according to the table below, first by the main researcher, and then discussed by a group of three to four other researchers engaged in a similar project. The final categorization was established by consensus. The total number of chunks and total number of errors were determined and also the totals of the different types of errors and chunks were added up. As discussed earlier, seventeen different types of chunks were distinguished. As discussed in the background section, Smiskova indentified the chunks according to length, schematicitiy and difficulty6. Twenty-three chunk types were
distinguished. Chunk type 6 was removed from the table by the researchers because the description of this type was thought to be too vague.
For each writing the total amount of chunks was calculated. In addition, also the number of chunks per category was calculated, since it is assumed that students will use different types of chunks in different stadia of their writings. Also, the total amount of chunks per proficiency level was calculated to investigate what types of chunks were used at which level and how often a specific chunk type was used at a particular level. Since students did not produce the same amount of texts at each proficiency level a chunk/text ratio was calculated by dividing the number of chunks (per chunk type) by the number of texts per level7. For example, if the chunk type “particle” would occur 8 times in 10 level 2 writings, the chunk type ratio for this level for all students was 8 was divided by 10. This was done for all proficiency levels, so that the comparison between the levels and their amount of chunks would be fair. For the
individual students, also a ratio of amount of chunks divided by the number of levels this chunk was produced at was calculated. For example, if student A would use 5 particle chunks
For the table see appendix or background section
at 4 level 4 writings, the ratio would be 5 divided by 4. All this data was saved into an Excel file for the analysis.
3.4 Design and analysis
This study is qualitative, explorative and longitudinal. The group of students produced writings for a period of a three years which were used for this study. All 17 students were in the same class. The individual learners that were analyzed were chosen randomly. No statistics were applied.
The main set up of the study involves the writings produced by the students. For each student, all chunks per type per writing (and thus proficiency level because each writing was scored) were added up in an Excel file. Subsequently, chunk types per level were calculated,
In this section, the results of this study will be presented. This section is dived in three
subsections. First, an outline of the proficiency development of the students will be displayed. Then, the chunks that were used per proficiency level will be listed in order to find out if there is a specific pattern in which chunks are learned. Finally, the learning patterns of six
individual learners are presented.
4.1 Development of Proficiency
As explained in the method section, the writings of all seventeen students were scaled according to proficiency level focusing on vocabulary, tense use, syntactic constructions and authentic language. As the chart below shows, most students begin at proficiency level 1 or 2 and progress from there on.
Figure 4.1. Indication of students‟ beginning levels (the level they started at in the first writing (1.1)
Figure 4.2 Indication of students‟ end level (the level of their final writing (3.4 or 3.5)
As visible in graph all students have progressed; none of them is at level 1 or 2 anymore. All students have landed at levels 3, 4 and 5. None of the students actually makes it to level 6. Most students have progressed to levels 3 and 4, the intermediate proficiency levels. In addition to the beginning and end levels of the students, the graphs below show a detailed view of the proficiency level for each student per writing in order to illuminate the
developmental pattern and variability of the students and their writing proficiency.
Figure 4.3 Proficiency scores per writing for students 1-4 (beginning level 1)8
Figure 4.4 Proficiency scores per writing for students 5-8 (beginning level 1)
Figure 4.6 Proficiency scores per writing for students 14-17 (beginning level 3 and 4)
These four graphs represent the proficiency development for all students. They also indicate that all students progress from first to final writing. Yet, the graphs also show the variability in proficiency scores along the way and thus the differences in development between the learners. None of the students show linear development from one level to another. Also, none of the students seems to settle at a proficiency certain level. Some students seem to move back and forth between levels for a while before they move on to the next proficiency level
whereas others make quite some progress, moving from a very low level to quite a high level in the end.
4.2 Chunks Acquisition (per level)
Besides use, input and chunks difficulty, also proficiency is related to chunk acquisition (easier, more common chunks are acquired at lower proficiency levels, more difficult, scarce ones at higher proficiency levels). Below, there is an overview of what chunks are acquired at which level (totals of all 17 students), to see if there is indeed a specific order or pattern for chunks acquisition
Level 1 Level 2 Level 3 Level 4 Level 5 Level 6
26 phrasal verbs particles bi- and trinomials complement structures preferred sentence stems slot fillers phrasal verbs particles bi- and trinomials complement structures preferred sentences stem slot fillers variable idioms complex prepositions adverbials speech act formulae phrasal verbs particles bi- and trinomials complement structures preferred sentences stem slot fillers variable idioms complex prepositions adverbials speech act formulae sentence structures idioms textual sentence stems phrasal verbs particles bi- and trinomials complement structures preferred sentences stem slot fillers variable idioms complex prepositions adverbials speech act formulae sentence structures idioms textual sentence stems similes complex conjunctions attitudinal formulae proverbs commonplaces phrasal verbs particles bi- and trinomials complement structures preferred sentences stem slot fillers variable idioms complex prepositions adverbials speech act formulae sentence structures idioms textual sentence stems similes complex conjunctions attitudinal formulae proverbs commonplaces Figure 4.7 Chunk types per proficiency level
The graph shows a clear increase of variety of chunks with every level. Also, almost all chunk types are used at proficiency levels 4 and 5 whereas only a selection of chunks is used at the lower proficiency level. This selection indeed includes easy, more common chunk types whereas the more difficult ones are present at the higher proficiency levels.
section, a chunk/level ratio was calculated because students did not produce the same amount of writings per level.
Level 1 Level 2 Level 3 level 4 Level 5
Lexical combinations 2 Particles 1,61 Particles 1,88 Particles 2,07 Particles 4,71
Particles 1,8 Lexical combinations 1,43 Complement structures 1,85 Lexical combinations 1,9 Lexical combinations 3,28 Compounds 0,6 Complement structures 1,23 Lexical combinations 1,81 Complement structures 1,88 Compounds 0,85 Complement structures 0,6 Slot fillers 0,95 Slot fillers 0,91 Slot fillers 1,02 Bi trinominals 0,71
Slot fillers 0,3 Compounds 0,82 Compounds 0,58 Compounds 0,69 Complement structures 0,71
Preferred sentences 0,2 Preferred sentence stems 0,3 Preferred sentence stems 0,32 Preferred sentences 0,45 Complex prepositions 0,71 Preferred sentence stems 0,1 Phrasal Verbs 0,28 Preferred sentences 0,54 Preferred sentence stems 0,33 Phrasal Verbs 0,57 Variable idioms 0,18 Linking adverbials 0,26 Phrasal Verbs 0,28 Preferred sentence stems 0,57 Preferred sentences 0,1 Textual sentence stems 0,17 Linking adverbials 0,21 Slot fillers 0,42 Linking adverbials 0,1 Bi trinominals 0,05 Attitudinal formulae 0,19 Preferred sentences 0,42 Bi trinominals 0,07 Variable idioms 0,05 Variable idioms 0,17 Sentence structures 0,28
Complex prepositions 0,07 Phrasal Verbs 0,04 Bi trinominals 0,07 Idioms 0,28
Speech act formulae 0,05 Complex prepositions 0,04 Complex conjunctions 0,07 Complex conjunctions 0,28 Complex conjunctions 0,04 Speech act formulae 0,07 Linking adverbials 0,28 Attitudinal formulae 0,04 Textual sentence stems 0,04 Textual sentence stems 0,28
Idioms 0,03 Commonplaces 0,04 Attitudinal formulae 0,28
Speech act formulae 0,03 Sentence structures 0,02 Proverbs 0,28
Sentence structures 0,01 Idioms 0,02 Variable idioms 0,14
Similes 0,02 Speech act formulae 0,14
Complex prepositions 0,02 Commonplaces 0,14
Figure 4.8 Chunks ratio per type per proficiency level
The table above resembles figure 4.8 in the way that the more difficult chunk types are only used at higher proficiency levels. Students start out with using the easier, low level chunks at the lower levels (level 1 and 2 mainly) and use quite a lot of those at levels 2, 3 and 4. At higher proficiency levels (levels 4 and 5) students start using use more difficult chunk types, even though low(er) level chunks are also still widely used.
Chunks level 1
Figure 4.9 Chunk type ratio in level 1 writings (chunk selected from most frequent to least frequent)
Chunks level 2
Figure 4.10 Chunk type ratio for level 2 writings (chunk selected from most frequent to least frequent)
When looking at the level 2 writings chart, it becomes clear that particles and collocations are still most frequently used. However, students have also started using more complement structures and slot fillers, compared to level 1 writings. Overall, the graph shows that a wider variety of different chunks is used at level 2, compared to level 1. The progression from level 1 to level 2 involves several chunk types. Apart from still using 3 types of lexical
combinations (collocations, prepositional chunks, compounds), also phrasal verbs are being used, such as “go to”, “lay on, “pick up” ”or “pay for”. In addition, some bi- and trinomials are being used, such as “nice and sweet” and “nightlife or seasons”. These chunk types are still short; 2-3 word combinations.
were already used at the first level as well, but at level 2 a wider variety of these chunk types is present and in addition, they are being used more often. Also, use of slot filler chunk has gone up, but all students use the same type of slot filler: X years old”. Furthermore, students have started producing a few complex prepositions such as „‟apart from‟‟, linking adverbials such as „‟in other words‟‟ and even speech act formulae such as „‟to cut a long story short‟‟. It is thus clear that a wider variety of chunks is being used. Also, chunks are a bit more difficult and a bit longer than chunks in level 1 writings.
Chunks level 3
Figure 4.11 Chunk types ratio in level 3 writings (chunk selected from most frequent to least frequent)
As the graph shows quite a variety of chunks is being used in level 3 writings. Some students start using new chunks such as sentence structures, idioms, complex conjunctions, textual sentence stems and attitudinal formulae. These new chunks are not frequently used, which shows in the graph, but some students at least have started experimenting with more elaborate chunks. Particles, collocations and complement structures are still most frequent in the
all still short, common chunks and include much of the same examples as in the level 1 and level 2 writings. Students show that they are no well accustomed to using these chunk types in their writings.
As to the new chunk types, they are of a higher standard because they are longer and often more difficult. These chunks include sentence structures such as “if you finish something, clean it up”, variable idioms such as “Go Horsey go!”, preferred sentences such as
“something like that”, “I think, I‟m not sure”, “I think that”, “because of that”, “that‟s about it”, complex conjunctions and linking adverbials such as “well”, “occasionally”, “then”, “as soon as”. Also, entire phrases like “It was just like I expected”, “it's actually pretty simple”, “I would like to do something with”, “I‟m not sure anymore”, “I don`t think I will get it, “I like X very much”, “it`s just so much fun”, “I‟m terribly sorry”, and “when I get the chance” are all being used in the level 3 texts, in addition to the more basic ones such as the lexical
Chunks level 4
Figure 4.12 Chunk types ratio in level 4 writings (chunk selected from most frequent to least frequent)
In level 4 writings, almost all chunk types used, except for type 23, the slogans (e.g. make love not war). Many of the newly used chunks are scarce in the writings, but some students were able to produce them. Lexical collocations, particles, complement structures and slot fillers are still being used most often. In addition to the new chunks as produced in level 3 writings, chunk types 11, 20, 21 and 22 are new in level 4 writings. These chunks include similes, such as “as blue as the sea” attitudinal formulae such as „‟to be honest‟‟, proverbs such as „‟When in…” and commonplaces such as „‟it‟s a big world‟‟. All these chunk types are high level chunks, scarce in everyday language and therefore difficult to acquire.
Besides the regularly used lexical combinations chunks and complement structures also slot fillers, preferred sentences, and linking adverbials are used quite often. New linking
of saying things include “I don't think we're going to”, “but I still like X better”, “I`m very good at it”, “what I like about X”. Students have started using more attitudinal formulae, speech act formulae and variable idioms such as “to be honest”, “as a matter of fact”, and “don‟t like doing”
Even though the more easy, low level chunks are still produced most frequently, students do start using more extended and longer chunks, such as idioms and preferred ways of saying things, linking adverbials, conjunctions, and attitudinal and speech act formulae. Whereas in level 3 writings, students merely experimented with using one or two of these chunk types, they are now more embedded in their writings.
Chunks level 5
Figure 4.13 Chunk types ratio in level writings (chunk selected from most frequent to least frequent)
“slang”, they show more attitude in their writings by using phrases or words like “gotta hurry”, “gotta go”, “whatever”. At this point, longer chunks are used at a higher level. For example, preferred sentence stem chunks, preferred sentences and idioms were also present at previous levels but students use new, more elaborate ones in level 5 writings. These chunks include “I'd like to work for X”, “when I am all grown up”, “she's such an amazing talent”, “they're just awesome”, “now I hear that we have to write two pages of…”, “I don‟t know what I‟m planning to do later on”, “I‟ll just take one step a time”, “they managed to escape” , and “I‟ll handle a situation”.Also, textual connectors such as “one day”, “ then”, “so”,
“once”, “for example” are still widely used. Students do not use new chunk types compared to level 4 but they do use more advanced chunks such as speech act formulae (“to cut a long story short”), attitudinal formulae (“honestly”), proverbs and proverb fragments (“When I was in France”) or commonplaces (“that‟s really important”).
The graph indicates that the distinction between high frequent low-level chunks and less frequent high-level chunks is not that salient anymore. Particles and collocations are still used most frequently, but the use of other chunk types does no longer show a big difference in frequency of use as was the case in, for example, level 4.
In summary, analysis on the distribution of chunks types per level showed that short, easy every day chunks are used a lower proficiency levels. These chunks are overall short (2-3) words and common in everyday speech. Low-level chunks are constantly used throughout the writings, but at higher proficiency levels, students also use longer, more difficult chunks which are not that common in everyday language.
4.3 Individual Differences: Development of Individual Learners
The development of six individual learners will now be presented. Of these six learners, three started out at level 1 and three at level 2. These beginning levels were chosen so that
Level 1 Collocations, prepositional chunks, compounds, phrasal verbs, particles, bi- and trinomials, complement structures, preferred sentence stems, slot fillers
Level 2 Collocations, prepositional chunks, compounds, phrasal verbs, particles, bi- and trinomials, complement structures, preferred sentence stems, slot fillers, variable idioms, complex prepositions, adverbials, speech act formulae
Level 3 Collocations, prepositional chunks, compounds, phrasal verbs, particles, bi- and trinomials, complement structures, preferred sentence stems, slot fillers, variable idioms, complex
prepositions, adverbials, speech act formulae, sentence structures, idioms, textual sentence stems Level 4 Collocations, prepositional chunks, compounds, phrasal verbs, particles, bi- and trinomials,
complement structures, preferred sentence stems, slot fillers, variable idioms, complex prepositions, adverbials, speech act formulae, sentence structures, idioms, textual sentence stems, similes, complex conjunctions, attitudinal formulae, proverbs, commonplaces Level 5 Collocations, prepositional chunks, compounds, phrasal verbs, particles, bi- and trinomials,
complement structures, preferred sentence stems, slot fillers, variable idioms, complex prepositions, adverbials, speech act formulae, sentence structures, idioms, textual sentence stems, similes, complex conjunctions, attitudinal formulae, proverbs, commonplaces Figure 4.14 Chunk types per level
Individual learners starting at level 1
Figure 4.13 Chunks acquisition ratio per level student 1.1
When comparing the graph of student 1.1 with the general chunks acquisition pattern as displayed in figure 4.8 and figure 4.14, it shows that there are some differences. This learner does not produce a lot of chunks to begin with and the chunks that are present are mostly produced at levels 2 and 3. When looking at the chunk types that were produced, the graph shows that mainly low(er) level chunks are used such as complement structures, particles, compounds, phrasal verbs and lexical combinations. Quite remarkable is the frequent use of bi- and trinomials at level 4. Low-level chunk types occurring in level 2 and 3 writings is in line the general pattern. Yet, only one chunk type (complement structures) is used at level 1 which is not in line with the general pattern. Also, it would be expected that a wider variety of chunks is produced at levels 2 and especially 3. Many chunk types are not produced at these levels whereas they are present at levels 2, 3 and 4 in the general pattern.
Figure 4.14 Chunks acquisition ratio per level student 1.2
The student is experimenting with new chunk types as proficiency level goes up but many types have not quite settled yet because they disappear again in level 5 writings.
Figure 4.15 Chunks acquisition ratio per level student 1.3
The third student starting out at level 1 does not show a wide variety of chunk use at any proficiency level. Given the fact that in the general pattern already 7 different chunk types were present in level 1 writings, chunk use of this student is not in line with the general pattern at all. Some high level chunks are acquired at level 3 and 4 but the really complex ones (e.g. speech act formulae, sentence structures, idioms, textual sentence stems, similes, complex conjunctions, attitudinal formulae, proverbs, commonplaces) are never being used, not even at level 4.
Individual learners starting at level 2
Figure 4.16 Chunks acquisition per level student 2.1
Figure 4.17 Chunks acquisition per level student 2.2 ratio
The second student starting out at level 2 hardly produces any chunks at level 3; only some preferred sentences stems are produced. Several low level chunks are produced at level 2, and quite a variety of chunks is used at level 4 and 5, including high level chunks (conjunctions, prepositions, commonplaces and sentence stems). The chunks at levels 2, 4 and 5 are
Figure 4.18 Chunks acquisition per level student 2.3
In this section the results will be discussed and linked to my research questions and the theories as discussed in the background section. I was interested to find out whether there is a pattern for learning chunks, related to input, use, proficiency and chunks schematicity. Also, it was explored whether individual learners follow this general pattern when acquiring chunks and if not, how they deviate from it.
5.1 Summary of the Results
When looking at the proficiency charts in the results section it became clear that in the end, all students progress. Due to language input and language use all students achieve a higher level of proficiency over the course of the three years in which their writings were analyzed. As a result, it can be concluded that use and (spoken) input proves to be important. Even though the students only encounter English several hours a week, they have regular input which allows them to improve and become more advanced. Results also showed that proficiency level must go up before the more difficult chunks can be produced.
Also, it was found that the group of students overall make huge progress between level 1 and level 2 and 3 when it comes to chunks use. Most students seem to have far more difficulty with the step between levels 3 and 4 and 4 and 5. This could be due to the fact that the step between level 1 and level 2 is relatively small compared to the step between level 4 and level 5. So, in between these levels, students manage to pick up quite a lot. When they want to move up to level 4 or 5 students not only still learn to use new chunks but it also seems to be about finetuning or gaining more variety in types of chunks that were already acquired. In any case, it requires a sufficient amount of input, use and language contact to make such a step; the chunk types present at level 4 and 5 are way more elaborate and advanced than those at level 1 or 2.
Furthermore, in the result section it became clear that students show a great deal of variability in their writings. Variability shows that development does not go in a straight line, as fits in with Dynamic Systems Theory. Variability is actually seen as a sign of learning (Verspoor et al, 2011) Even though this group shows a general acquisition, it is an overall pattern for all 17 learners. In-depth analysis of six individual learners has indicated that even though the
5.2 Overall Chunks Acquisition Pattern
As was just reviewed, there is a pattern in which chunks are acquired. Level 1 writings contain more basic, common chunks whereas more advanced writings show use of more advanced chunks. This is line with the assumption that proficiency influences chunk development; proficiency needs to expand before more difficult, longer chunks are used. Because of their difficulty and schematicity but also because of their scarceness, it takes a certain level to successful recognize, incorporate and use those chunks.
Level 1 writings only contain every day, frequently used chunks such as lexical combinations (compounds, particles, collocations, phrasal verbs and bi- and trinomials) complement
structures, slot fillers and preferred sentence stems. Sentences at this level are really short and students show little creativity; students mainly repeat words and (types of) phrases they are familiar with. The low-level chunks used at this level are short; they are 2-3 words long and very common in everyday speech. Students mainly produce constructions mainly use constructions like “at home”, “at school”, “High School”, “I want to”, “really nice”, “going to”; all high frequent words in speech and therefore frequently encountered. Because they are so short and common, it is rather easy for students to pick them up, store and use them. As mentioned in the results section, students have made a considerable jump from level 1 to 2; level 2 writings were already more fluid and contained a wider variety of chunks than level 1 writings. Students who produce level 2 writings have had more input and they have had more opportunity to use the language, which leads to the emergence and use of new and slightly more complex chunks by some of the students. Several students used complex
prepositions, speech act formula or linking adverbials. These chunks are slightly less common and overall a bit longer than chunks at level 1 which explains why they were not used in level 1 writings. The chunks that are used at level 1, mostly compounds, collocations and particles, are still used in level 2 writings but more frequently and in more variety. One of the examples was the use of particles such as “afraid of” or “at home”, which were used at level 1. At level 2, these were still used but also longer versions of these chunks were produced, such as “I am afraid of”, or “at my grandma‟s home” This shows that students rely on previous linguistic experience and build on this, as proposed by Hopper (1998).
shows in their use of chunks. At level 3, a variety of new and more complex chunks is used. This fits in with the assumption that these types of chunks are acquired when becoming more proficient. Language is beginning to look more authentic, due to higher proficiency and use of higher level chunks such as textual connectors. Also, students have had more input and
language contact which results in noticing less commonly used chunks such as conjunctions and idioms. As discussed, it takes a certain amount of input to recognize and process these chunks (Bybee, 2008; Verspoor et al, 2010). Beside recognizing them, some students also use them. Idioms (it‟s raining cats and dogs) and speech act formulae (to be honest) for example, are not used in great numbers but some students do produce them. Still, the proficiency level is too basic for the writings to resemble “real” English.
In level 4 writings, students use a wide variety of chunks; both low level and high level. Overall, the high level chunks tend to be advanced and more complex than those at level 4. These longer and more complex chunks make sentence more fluent, authentic and easier to read. Almost all the chunk types are now being used, except for type 23 (the slogan). So, in addition to the low-level chunk types, now almost all high-level chunk types are being produced as well, as was expected given the relation between proficiency and chunk
production. Students do indeed use scarce, not very common, difficult chunk types now, even though some types are not produced very often. Still, at this level, students are capable of using a wide variety of chunks. Because their proficiency is getting better (e.g. sentences become longer and more complex), students use chunks like “then” or “later”. Also, students are more able to express themselves or express opinions, which shows in the use of attitudinal formulae such as “as a matter of fact” or “to be honest” and preferred sentences such as ”what I like about…” or “I don‟t feel like…”. At this point, students are thus able to recognize longer and more difficult chunks and produce a variety of those types themselves from time to time. However, low-level chunks remain most frequently used, simply because they are most common still.
Chunks are now used more effortlessly, with a more natural feel. Students produce more high-level chunks than at high-level 5, meaning that these chunk types are more embedded in their language system. Due to their high proficiency level and amount of input they will not mainly recognize and produce low-level chunks, they will more and more also start to recognize and use high-level chunks. Thus, at this point, several factors now seem to come together;
frequent input, frequent use, reliance on previous linguistic experience and proficiency which ties in nicely with DST given the interconnectedness of variables and their influence on language learning (e.g. Verspoor et al, 2011). The results section showed that the substantial difference in use between low and high level chunks is getting smaller. This difference will probably never go away, since low-level chunks are simply more embedded in every day speech, but it goes to show that students feel more comfortable using high-level chunk types as well. They are more capable of balancing out their use of chunks because besides noticing and storing low-level chunks they are now able to also more frequently recognize and store high-level chunks.
To sum up, per level new chunk categories emerge; at higher levels longer and more elaborate types of chunks are being used in addition to the frequent, more common low level chunks. Generally, students use a wider variety of chunks when becoming more proficient.
Proficiency, related to use enables students to learn to recognize and use not only low-level, short and common chunks but gradually also the scarce, more difficult ones.
5.3 Individual Development and Individual Differences
The investigation of individual learning patterns showed some interesting results. First of all, learners starting out at the same level developed differently from each other. So, they do not only deviate from the general acquisition pattern as discussed in the previous section from time to time, they also deviate from one another in how they do so. The figures displaying the total of chunks per level ratio showed that, overall, students‟ progress from using easier, shorter everyday chunks at lower levels and when receiving more input and becoming more proficient go on with using more difficult chunks. Yet, when looking at individual learners, it became clear that some students use far more chunks than others do, at different levels. Also, some students use a wide(r) variety of chunks than their peers do at certain levels.
As previously discussed, the analysis of the individual learners showed that each learner shows a different learning pattern. Surely, there are overall similarities that, in this case, lead to a general pattern of acquisition, but still each learner learns in their own way. Some learners struggled to attain higher levels of proficiency and producing more elaborate chunks whereas others mainly produced chunks at higher levels. Also, some learners used similar chunk types at each proficiency level; they acquire only a few more difficult chunk types at higher levels but continue on mainly using the basic chunk types. Some students use chunks at a particular level, stop using them at the next level and then use them again at a later stage. The individual analysis of six learners also showed that none of the learners actually uses all types of chunks.
There could be several explanations for these different learning patterns. Some students might be more grammar focused and therefore use less or different types of chunks than their peers do. They might therefore be a bit slower in picking up the more elaborate chunks. Also, some students might need more time or input before a chunk type is correctly established into their language system. Even though frequent use and frequent input are also beneficial for language learning (Ellis, 2002; MacWhinney, 2008; Verspoor et al, 2011) there might very well be differences between learners as to how much input and use is needed for successful settlement of language patterns. Some students pick up a chunk and are able to use it immediately
whereas others need several encounters before the chunk is stored. This might have to do with language aptitude, but also with other factors such as motivation. It is difficult for the students to use all types because many students only produce about 3 writings per level. Since the texts are quite short, it is quite a challenge to use the complete variety of chunks. Furthermore, it has been discussed that some chunks are more scarce than others (such as slogans or similes). Finally, different kind of topics allow different types of chunks so thus might also influence chunks production.
individual learning displays at what points students settle, show progress or show difficulty. With this information, a researcher can opt to find out why this is the case; what factors are at work that make something happen. The individual analyses thus show that variability in second language development can provide insight into the developmental dynamics of second language (L2) learners (Verspoor et al, 2008).
In line with previous research, this study has highlighted the importance of regular input. The OTTO project found that bilingual students learn a lot language implicitly; they pick up language structures because of the large amount of input they receive. Since regular students do not have that many hours of English a week, it might be interesting to investigate how much out of class contact with the English language the students have exactly and see if and how that makes a difference. Also, this thesis has discussed the influence of learning factors other than input and use. Besides these crucial factors, also motivation, language aptitude or contextual factors might influence language learning. Due to time and space limitations, these factors were not explicitly included in the research, although there were sometimes
mentioned. Further research might put more focus on these issues, including an investigation of how they relate to language learning.
The topic of research in this paper was the acquisition of formulaic sequences (by learners of English). Several important factors that would stimulate chunks acquisition were discussed and also the relation between acquisition and proficiency has been explained. In addition, the schematicity and difficulty of different chunk types was discussed, leading to the first
research question: is there a general pattern or order in which chunks are acquired, related to factors as use, input and proficiency? It was found that there is indeed such a pattern; more easy, low lever chunks were acquired before high level chunks. Since proficiency is related to language use and input, it was also found that students mainly use simply, common and short chunks at lower level chunks. When becoming more proficient they start to produce chunks that are longer and more difficult and also more scarce. Since low-level, short chunks are very common in everyday speech, it does not take a large amount of input to be able to produce those whereas scarce and less common chunks do require more input and use, but also higher proficiency.
The results are in line with several studies as discussed in the background, arguing that frequent input and use is vital for successful language learning (e.g. MacWhinney, 2008; Verspoor et al, 2011) and that proficiency is related to use of chunk types: more advance chunks are used at higher proficiency levels (Antoniou, 2008, Verpoor et al, 2010). Learners eventually recognize the patterns that frequently occur together and are able to store them as such; beginning with the easier ones (such as phrasal verbs) but later on they also proved to be able to store entire phrases or sentences. Still, as seen in the individual development of six learners, there are other factors at work besides use and input that influence language.