Conceptualization of lexical complexity
In the language-learning literature, lexical complexity (LC) has been referred to using a variety of terms, such as lexical diversity (Jarvis, 2002; Malvern & Richards, 1997; McKee, Malvern & Richards., 2000), lexical richness (Daller, Van Hout & Treffers-Daller, 2003; Horst & Collins, 2006; Laufer, 1994; Laufer & Nation, 1995; Ovtcharov, Cobb & Halter, 2006; Read, 2000; Tidball & Treffers-Daller, 2008; Vermeer, 2000, 2004) and lexical proficiency (Harley & King, 1989). The wide variety of terminology used, coupled with the fact that, as Butlé and Housen (2012) point out, there is no commonly accepted definition for this concept, can lead to confusion.
As Bulté and Housen (2012) note, many studies do not give a clear definition of the terms, which makes ensuring uniformity in the interpretation of the concepts rather difficult. These observations emphasize the importance of clearly defining the concept of lexical complexity in order to enable a more correct and precise interpretation and comparison of results and conclusions across a number of studies. OK
Bulté and Housen (2012) present the concept of linguistic complexity as a vast system that involves many aspects of language. The subcategory of linguistic complexity that is of interest here is lexical complexity. The authors define this concept as “the degree of elaboration, the size, breadth, width, or richness of the learner’s L2 system or ‘repertoire’, that is, to the number, range, variety or diversity of different structures and items that he knows or uses . . . ” (p. 25). The definition refers to the complexity of the lexical knowledge the learner has acquired to date rather than the complexity of the language itself. The authors further divide LC into three observable and measurable constructs, which can be used to assess LC: lexical density, lexical diversity and lexical sophistication.
Lexical density can be expressed by two measures: the ratio between the number of lexical words (nouns, adjectives, verbs and adverbs) and the number of function words (pronouns, prepositions, conjunctions, articles and auxiliary verbs), and the ratio between the number of lexical words and the total number of words (Bulté & Housen, 2012). A high ratio indicates a lexically dense text; the higher the ratio, the more lexical words are contained in the text, in comparison either to function words or to the total number of words in the text.
Lexical diversity, as defined by Bulté, Housen, Pierrard and Van Daele (2008), refers to the extent of the learner’s lexical knowledge, or the number of different words he or she knows and uses. It can be expressed as the number of different words or types a learner uses, or it can be expressed as a ratio, such as the Type-Token Ratio (TTR), which is calculated by dividing the number of different words by the total number of words in the text (Bulté et al.). Read (2000) also uses the term lexical variation to refer to this concept.
Bulté et al. (2008) define lexical sophistication as “the perception of a L2 user’s lexical proficiency formed by, among other things, his use of semantically more specific and/or pragmatically more appropriate different words from among a set of related words” (Bulté et al., p. 279). Lexical sophistication can also be considered as the proportion of less-frequent words a learner uses in a text. Frequency lists, which divide words of a given text into lists according to their frequency of occurrence in the language (for example, the 1000 most-frequent word families, the 1001-2000 most-frequent word families, etc.), are used to determine the proportion of less-frequent words used by the learner. Vermeer (2004) establishes a relationship between the degree of difficulty of a word and its order of acquisition. She emphasizes the fact that adults and children learn language differently, and states that in young learners, the frequency of a word can be used to define its degree of difficulty. A greater proportion of less-frequent words therefore indicates a more lexically sophisticated text.
The three aspects of lexical complexity (density, diversity and sophistication) combine to paint a complete picture of the complexity of a learner’s lexical knowledge. The development of a learner’s
LC can be observed through variation, over a period of time, in the measures used to represent these three aspects.
Measures of lexical complexity
Many different measures have been used to assess lexical complexity. As Daller et al. (2003) point out, the consensus among researchers appears to be that there is a need for investigating lexical knowledge; however, how to investigate this lexical knowledge remains an area of dispute. Indeed, there is no commonly accepted measure of lexical complexity, and the validity of some measures is contested by other researchers. The most frequently used measure of lexical complexity appears to be the Type-Token Ratio (Vermeer, 2000). This measure is calculated by dividing the number of types (different words) by the number of tokens (total number of words) in a given text. However, the validity of this measure has been criticized by many researchers (such as Daller et al., 2003; Jarvis, 2002; Laufer & Nation, 1995; Malvern & Richards, 1997; McKee et al., 2000; Serrano Serrano, 2011; Vermeer, 2000), mainly because of its dependence on text length.
One of the main problems in assessing lexical complexity is the influence of text length. On the one hand, a longer text appears to give learners more opportunity to use unique words, making the longer text more lexically complex; on the other hand, a longer text also means greater opportunity for repetition, as the pool of words to choose from becomes more restricted as the text lengthens. TTR is a measure that is directly influenced by text length: as the text lengthens, TTR decreases, giving the text the appearance of being less lexically rich (Malvern & Richards, 2002). The comparison between texts of different lengths through TTR is therefore rendered invalid.
In order to overcome the problem of TTR’s dependence on text length, mathematical transformations have been developed as an alternative. The Guiraud Index is one of those. It is calculated by dividing the number of types by the square root of the number of tokens (Daller et al., 2003). Although it appears to be a more valid measure than TTR (Daller et al., 2003; Vermeer, 2000), the Guiraud Index is similarly criticized for its dependence on text length. Vermeer (2000) claims that as the text lengthens, the measure will first increase, then reach a maximum and begin to decrease. The fact that the measure only seems valid for texts of a certain length prompted Daller et al. (2003) to question its validity for more advanced learners.
Collentine (2004) did not use the traditional measure of TTR or one of its variations to assess LC, but used the number of unique occurrences per 1000 words of text. Similar to TTR in that it provides insight as to the number of different words a learner is able to produce, this scaled measure made it possible to compare texts of different lengths, as the comparison was made on equivalent proportions of text. The method assessed the proportion of unique occurrences according to seven lexical categories (adjectives, adverbs, conjunctions, nouns, prepositions, pronouns, and verbs) and was adopted in order to discriminate between the four primary parts of speech (nouns, verbs, adverbs and adjectives), which constitute the learner’s core lexical base, and more complex parts of speech, which are used to mark discursive coherence. A high number of unique occurrences, or a high ratio in the case of TTR, indicates a lexically diverse text.
In 1995, Laufer and Nation proposed a new measure to assess lexical richness: lexical frequency profiling (LFP). This measure categorizes words of a given text into separate lists according to their frequency of occurrence in the language; thus, the proportion of less-frequent (and therefore more complex) words illustrates the lexical complexity of the text. They set out to prove the reliability and validity of LFP, and their analyses led them to conclude that it was a reliable and valid measure of lexical richness. However, in 2005, Meara put forth a critical analysis of Laufer and Nation’s (1995) use of LFP and questioned the validity and sensitivity of their results. He used a set of simulations (Monte Carlo analysis) to examine the validity of LFP in estimating productive vocabulary size. He concluded that while LFP may allow researchers to distinguish between groups of learners whose vocabulary sizes are very different, it is not as sensitive with groups that have small differences in vocabulary size. He also reported that the sensitivity of LFP seemed to decline with larger vocabulary sizes. This critique was addressed by Laufer (2005), who attempted to clarify the matter by explaining that LFP is a measure of lexical use in writing, rather than a measure of vocabulary size. She maintained that the reason LFP did not distinguish between groups showing only slight differences in vocabulary size was that those learners do not use vocabulary differently. She argued that all areas of lexical competence do not develop the same way, and that an increase of 500 words in vocabulary size did not necessarily lead to a change in vocabulary use.
Edwards and Collins (2011) postulated that the underlying mathematical model, which was based on Zipf’s law and used in the simulations conducted by Meara (2005), allowed the analysis to be done directly, without recourse to simulations. The results of their analysis suggested that LFP was not particularly sensitive with respect to estimating individual productive vocabulary sizes, and confirmed Meara’s (2005) suggestion that LFP’s ability to distinguish between groups decreased as the size of the vocabulary increased. In 2013, Edwards and Collins delved further into the analysis by exploring a new model of vocabulary learning that was actually an adapted version of the previously proposed model. The motivation for their analysis was the fact that it was assumed, with the earlier model (the “naïve” model), that L2 learners acquired words according to their frequency of occurrence in the language, which was not the case. They attempted to develop a more valid model that took into account the presence of less common words at different points throughout the acquisition process and the fact that words may be acquired through repeated exposure. The model took into consideration the lower-frequency words that are learned before certain higher-frequency ones and therefore better reflected real-world L2 vocabulary acquisition. They concluded that this model provided a modified view of the relationship between LFP and students’ productive vocabulary size, and gave lower estimates of vocabulary size than the naïve model. Although LFP has been used by certain researchers to estimate vocabulary size it will only be used in the present study for the initial purpose for which it was developed in 1995 by Laufer and Nation, i.e., to evaluate the lexical complexity of a text by categorizing words according to their frequency of occurrence.
In 2006, Horst and Collins (2006) assessed the development of IE students’ LC. They initially used Vocabprofile to create LFP as the measure of LC for their study, but this analysis did not produce conclusive results. They then turned to four additional measures (reliance on French, reliance on cognates, morphological variety and number of K1 families) to supplement their initial results. The students participating in the study were IE students who lived in a French environment and had little or no exposure to the English language outside of class. They used French to fill the lexical gaps in their oral productions, thus pointing to the pertinence of studying how their reliance on French changed over time. The researchers looked at the amount of French used by the students, as well as the nature of the French words used. Further to that analysis, they emphasized the importance of using more than one measure to assess LC.
The studies presented above show that a variety of measures are used to assess the development of lexical complexity. Although there is no commonly accepted measure of LC, an element many researchers (Bulté and Housen, 2012; Daller et al., 2003; Horst and Collins, 2006) appear to agree on is the importance of using a number of measures that combine to paint a complete picture of the complexity of a learner’s lexical knowledge. The development of a learner’s LC can be observed through a variation, over a period of time, in the measures used to represent LC. For the purpose of the present study, measures related to the three constructs proposed by Bulté and Housen will be used: lexical density (through the proportion of lexical words versus total words), lexical diversity (through the number of unique occurrences) and lexical sophistication (through LFP, or the proportion of less-frequent words used). Reliance on French words, a measure suggested by Horst and Collins (2006), which appears to be particularly pertinent in the context of IE programs, will be used as an additional measure to supplement the results provided by the three first measures.
Intensive English programs
In the field of second language acquisition, research demonstrates that repeated exposure and practice are favourable to learning new items or structures (Serrano, 2012). Traditional L2 programs, which offer a few hours of instruction every week, have not proven to be particularly effective in teaching a second language, and have been shown to leave students with limited L2 abilities (Collins & White, 2011). Stern (1985) suggests that this drip-feed approach is less effective, for an equivalent number of hours of instruction, than a more concentrated model in which instruction is administered in larger blocks on a daily basis. He states that a compact or intensive course, if planned correctly, “has undoubtedly considerable potential to remedy weak and straggling language programs” (p. 24). More concentrated forms of L2 instruction, such as intensive programs, have therefore been developed to offer an alternative to traditional programs (Serrano & Munoz, 2007). Research in intensive programs has shown that concentrating the hours of instruction can allow students to make substantial L2 gains in a short amount of time (Collins et al., 1999; Collins and White, 2011; Germain, Lightbown, Netten & Spada, 2004; White & Turner, 2005) without any detrimental effects to their first language (L1) (Lightbown & Spada, 1991; 1997; Spada & Lightbown, 1989).
In Quebec, the main goal of IE programs is to help students become functionally bilingual in order to be able to face typical everyday situations in their L2. This is done by presenting students with authentic learning situations that are meaningful and that relate to the students’ real-life needs and interests. Age-appropriate topics are presented through authentic learning situations to prepare the students for real-life communication (SPEAQ, 2012). A communicative approach is used in all IE classrooms, and although all four skills (speaking, listening, reading and writing) are utilized, the main focus is on listening and speaking (Germain et al., 2004).
IE in Quebec is generally offered in sixth grade, but some school boards choose to offer the program in fifth grade (Germain et al., 2004). While the first IE program was developed on a five-month/five-month model, other models have been created over the years. These are grouped into two categories: massed programs and distributed programs. In the former, the intensive period is condensed into five months, with students devoting half of the school year to the regular curriculum, and the other half to ESL; in the latter, the intensive program is spread out over the ten months of the school year (Collins et al., 1999). Some examples of distributed programs include alternating between 4 days/1 day and 1 day/4 days; 2.5 days/2.5 days; 1 day/1 day; half a day/half a day. Schools and school boards have the freedom to decide on the structure of the intensive program. Of all these models, the massed program appears to be the most widely used (SPEAQ, 2012).
Overall, studies on IE programs in Quebec show that intensive instruction can have a positive impact on learning a L2. In fact, reading and listening comprehension, basic communication skills, along with fluency, confidence, and mean length of runs are some of the areas that appear to benefit from a five-month intensive program (Collins et al., 1999; Lightbown & Spada, 1994; Spada & Lightbown, 1989; White & Turner, 2005). The present study looks at the learning process of students participating in a five-month program, which provides an ideal context in which to investigate the development of LC in a relatively short amount of time.
Intensive English in Quebec
Much of the research involving IE programs in Quebec has focused on the development of global components of students’ language, such as fluency (Spada & Lightbown, 1989), listening and reading comprehension (Collins & White, 2011; Spada & Lightbown, 1989), correctness of language (Lightbown & Spada, 1991) and oral production or performance (Collins & White, 2011; Spada & Lightbown, 1989; White & Turner, 2005). There is now a substantial body of evidence showing that IE programs allow students to make considerable gains in their L2 in a short amount of time, and that show that IE students outperform their regular-program counterparts on a variety of tasks pertaining to all four language skills (Collins & White, 2012).
In terms of vocabulary development, research has shown that word recognition steadily improves over the course of an intensive program, suggesting that students’ vocabulary has grown by the end of the intensive period. Collins and White (2011) suggest that by the end of the program, IE students are familiar with approximately 75% of the 1000 most-frequent words of the language. It has also been observed that IE students tend to be more talkative (i.e., to produce more words) than students following a regular ESL curriculum (Spada & Lightbown, 1989), and that they use a wider range of expressions and a more varied vocabulary when participating in an oral task (White & Turner, 2005).
While little research has specifically dealt with lexical knowledge and development, Spada & Lightbown’s (1989) finding that IE students use a more varied vocabulary than their regular-program counterparts might suggest that IE programs could promote gains in lexical abilities, including the development of LC. This is confirmed in Laufer’s (1994) idea that:
The process of acquiring an additional language, second or foreign, has often been described and discussed in terms of the learner’s progress along the Interlanguage continuum, from a non-existent knowledge of L2 towards native-like competence, without necessarily reaching this ideal stage. If this is the view we take of language acquisition, then lexical acquisition research would have to account for the gradual increase in the learner’s vocabulary size, as the most striking difference between the vocabulary of native speakers and that of language learners is in the number of words they can control, particularly in free production: speech or writing (p. 21).
One of the few studies to investigate lexical development in an IE context was conducted by Horst and Collins (2006), who studied the development of lexical richness in written productions of sixth-grade IE students. They based their research on the hypothesis that the students’ writing would show improved lexical richness over the course of an intensive program. The study involved a total of 230 students, all of whom were 11- or 12-year-old French-speaking Québécois who lived in an environment that provided little exposure to English outside the classroom. The researchers used LFP to measure the lexical richness of students’ texts at four times during the program; after 100, 200, 300 and 400 hours of instruction. At each of the four testing times, students were shown a picture prompt and asked to write about what they thought had happened before, during, and after the event depicted. They were given 20 minutes to complete the task, and were allowed to use French words to fill lexical gaps.
Students’ texts were analysed using Vocabprofile. The software divided the words of the text into four categories: the 1000 most-frequent word families of English (K1), the 1001-2000 most-frequent word families of English (K2), an academic word list (AWL), and an off-list, which included all the words that did not fit into one of the three previous categories. The authors hypothesized that over time, the proportion of words in the K1 band would decrease and the proportion of words in the other three categories would increase, thus showing that the students were using a greater amount of less-frequent (and therefore more complex) words as time went by. However, the proportion of words from the K1 band slightly increased over time, and the proportion of words in the K2 band actually decreased, contrary to hypothesis. These initial results therefore showed that the young beginning learners of English did not use greater amounts of less-frequent words over the course of an intensive program. However, the authors were not prepared to say that there had been no lexical development. They conducted further analyses using other measures and were able to conclude that although Vocabprofile did not suggest any development in LC, the use of four alternate measures (reliance on French words, reliance on cognates, variety of word families within the 1000 level and morphological variety of items within word families) showed substantial improvement in LC over the course of the IE program.
One of the additional measures they looked at was reliance on French. When writing their texts, students had been informed that they could use words from their L1 to fill lexical gaps. The number of French words per text decreased considerably from one testing time to the next, and the kinds of French words that the students used also changed; over time, the French words students produced were, in increasing proportion, relatively low-frequency words. This decrease in the use of French suggested that they were able to use a greater number of English words. The number of different word families and the number of types per family were two other measures that were also used to further investigate the development in lexical complexity, and it was noted that over time, students used more word families in their compositions, along with a greater number of types per family.
Development of lexical complexity
As pointed out above, only one study (Horst & Collins, 2006), has analysed the development of LC in IE students, thus highlighting the pertinence of examining other similar contexts in order to obtain a clearer picture of how the development of LC has been assessed.
While much of the research on the development of LC in learners of a second language has been conducted with adults or teenagers, some researchers have also studied the phenomenon in children. This is the case of Bournot-Trites (2007), who studied French immersion students’ development in various aspects of language, including lexical development. The study followed two groups of students and analysed their progress from fifth grade to seventh grade through a composition task that was completed at the beginning of fifth and the end of seventh grade. Three measures were used to assess the students’ lexical development: diversity (number of different verbs divided by the total number of words), sophistication (number of less-frequent verbs divided by the total number of verbs) and the total number of words contained in the text. Results showed greater diversity in fifth grade, but greater sophistication and mean total number of words in seventh grade. These results led the author to suggest that the immersion had a positive impact on students’ lexical development at an advanced stage of learning.
Another study involving young learners was conducted by Lo and Murphy (2010). In that study, the immersion context was a central element, as the researchers attempted to determine whether the language-learning context had an impact on the learners’ vocabulary knowledge and growth. The two learning contexts studied were regular second-language instruction programs and immersion programs in Hong Kong. The students participating in the study were in seventh and ninth grades, and for either grade level, some of the students were following a regular program, while the others were enrolled in an immersion program (the immersion program began in seventh grade; therefore, immersion-program students were either in their first or third year of English immersion).
A free-composition task, which was analyzed through LFP, was used to assess development of LC. The results of the analysis showed that both in seventh and ninth grade, immersion students include a significantly larger proportion of less-frequent words (which appear beyond the first 2000 words on the frequency lists). Hence, the authors concluded that the students in immersion programs used a greater proportion of low-frequency words in their compositions. They also noted that, by as soon as the end of the seventh grade, an advantage was observed on a number of measures for the immersion students, indicating that even after only one year in an immersion program, students were able to show more advanced levels of vocabulary than their regular-program counterparts. Overall, these findings suggest a clear advantage for immersion programs in terms of the development of lexical complexity when measured through written tasks.
The findings presented above have shown that, through the use of LFP, immersion contexts tend to favour the development of LC in children. This differs from research in IE contexts, where research has yet to use LFP to demonstrate significant improvement in LC. The fact that conclusive results have been found in immersion contexts introduces the idea that it may be pertinent to further investigate whether LFP could be used to measure development of LC in an IE context.
The data for this research project comes from an ongoing large-scale research project that examines IE programs and is conducted by Dr. Leif French at the Université du Québec à Chicoutimi. The data collected for the project covers numerous students enrolled in programs based on a variety of intensive models. However, the present study will focus on a sample of those students and will only present data based on tasks directly relevant to the research questions and hypotheses outlined in the introduction.
The students in the study were all enrolled in a sixth-grade IE program. Although various models exist in terms of time distribution for intensive ESL programs, the model applied was among the most frequently used ones (Germain et al., 2004). For the first five months of the school year, the students focused on the regular sixth-grade curriculum. The students then spent the five remaining months learning English. During the five-month English period, the students received approximately 370 hours of instruction based on a communicative approach, which emphasized the development of oral skills. No formal academic content was taught in English, and project-based pedagogy was adopted, focusing on activities that were stimulating for the students and that aimed at helping them develop functional language that would be useful to them in everyday life, were they to be placed in an English-speaking environment.
Participants for this study (n = 56)4 were French-speaking students (11 to 12 years old) from two different elementary schools in the Saguenay region (one class of 28 and one class of 29 students). Initially, 57 students were to participate in the study, but one participant had to be excluded, as he was absent at one of the two testing times. All students had begun to learn English in fourth grade and, prior to the start of the intensive program, had received approximately 120 hours of formal classroom instruction. All had completed the regular sixth-grade academic program in the first part of the school year and were devoting the remainder of the school year to learning English.
A two-part vocabulary test was administered to the students both at the beginning (T1) and at the end (T2) of the program (see Appendix A). In the first part, students were presented with a list of 60 English words, which they were asked to translate into French. In the second part, they were presented with a list of 60 French words, which they were asked to translate into English. All of the items on both lists were common words taken from fifth-grade ESL textbooks. A score out of 60 was computed for both parts of the test, and a composite score out of 120 was also calculated. For the purpose of the present study, the results of the vocabulary test were used to assess students’ initial level of proficiency. A median split was used to divide the students into high- and low-proficiency groups.
This measure was chosen in order to group the students on a measure which was independent from the four other measures of LC, but which was still related to the students’ lexical knowledge.
An oral narration task based on picture cues was completed by the students at T1 and T2. Although picture-cue tasks are generally not used for teaching purposes in communicative contexts, as they do not represent an authentic communication situation, they are used in research to study various aspects of students’ language, and are deemed particularly useful in studying vocabulary development. An important reason for using picture stories in research is the fact that while they provide researchers with the opportunity to control the language elicited, they also give students the freedom to produce language that adequately reflects their speaking abilities (Rossiter, Derwing & Jones, 2008).
Rossiter et al. (2008) also emphasize the importance of choosing a picture story that is clear and void of any ambiguity so that the students focus on the language rather than the meaning of the pictures. The picture story used in this task corresponded well to the criteria suggested by Rossiter et al. More specifically, each picture frame depicted only one action, was detailed enough for readers to understand the context and action, contained no cultural content that might be difficult for students to interpret, and had a clear sequence of events with no flashbacks or flash-forwards.
Students were provided with a 20-minute planning phase, during which they were able to look at the story and take notes as to what was happening in the pictures, and how they planned to relate the story. Although they were not allowed to use their notes for the actual narration, this planning phase enabled them to gain familiarity with the task and, thus, to reduce the cognitive load on language processing during the narration. Foster and Skehan (1996) postulate that planning time gives rise to greater linguistic complexity, as it gives the students the opportunity to attempt more ambitious ideas and to provide greater clarity in the relationship between ideas. The planning phase therefore makes it likely that students were able to use vocabulary that was more contextually appropriate, and that they were able to use the extent of their vocabulary to a greater degree.
The specific picture-cue task used in this study was an adapted version of the book Frog, Where Are You? by Mercer Mayer (1969). The adapted version, created by Leif French at the University of Québec at Chicoutimi, was composed of a series of 16 images and designed specifically to reduce the time needed to tell the Frog Story in a classroom situation while maintaining the integrity of the original storyline (see Appendix B). Thus, the story is that of a young boy who catches a frog and puts it in a jar by his bed. During the night, the frog gets away. In the morning, the young boy is alarmed to find that the frog is no longer in the jar. He looks for it everywhere in his bedroom, but to no avail. He therefore decides to take the search outside with his dog. He goes into the woods, where he encounters a deer, and gets pushed off a low cliff into a brook. He then returns home with his dog, but without the frog. The next day, the young boy goes back outside to look for the frog, and finds it sitting on a log by the river with its family.
The same picture story was presented to the students at T1 and T2. On both occasions, the students were provided with a booklet in which they could take notes and plan each of the three parts of the story. Both the booklet and story were clearly labelled with a time frame (yesterday and this morning, now, tomorrow) in order to provide a basis for indirectly eliciting specific temporal elements (see Appendix B). However, students were not explicitly instructed to use the past, present and future tenses for the different parts of the story.
Procedure for oral narration
Instructions for the task were presented to the students in French by the researcher. They were given 20 minutes to prepare and plan the story. To do so, a planning booklet was given to them, and they were given specific instructions as to how the story had to be constructed. The booklet was divided into three distinct parts, and each one included a space for specific vocabulary the students deemed might be useful to them for the task. The instructions for the task, which were given to the students by the researcher at the beginning of the task, were also contained in the preparation booklet.
At the end of the 20-minute period, students were asked to step into another room, where a researcher was waiting to record their story. Students were asked to narrate the story in English. They were allowed to use French to compensate for lexical gaps, but were not encouraged to do so. They could take as long as they wished to tell the story and were recorded on a computer using the Audacity software program.
Measures of lexical complexity
Several authors (Bulté & Housen, 2012; Daller et al., 2003; Horst & Collins, 2006) stress the importance of using multiple measures to provide a complete picture of the students’ progress over time. In the present study, four different measures were therefore used in conjunction to illustrate students’ development of LC over the course of the five-month intensive program (see Table 1).
The lexical diversity of the narrations produced by the students was gauged by the number of unique occurrences (different words) per 100 words of text produced by the students both at T1 and T2. The diversity provided an indication of whether the students’ productive vocabulary seemed to have grown.
Since the texts produced by the students were of different lengths and were likely to be longer at T2 than at T1, a measure independent of text length had to be calculated. By expressing the total number of unique occurrences per 100 words, and thus matching the texts on a comparable ratio rather than simply using the number of unique occurrences in the text, narrations of various lengths could be analysed and compared.
Lexical sophistication (use of less-frequent words)
Students’ use of less-frequent words was measured through lexical frequency profiling, with the Vocabprofile software program. Vocabprofile analyzes texts and categorizes each of the words (tokens) into one of four lists: the 1000 most-frequent word families in English (K1), the 1001 to 2000 most-frequent word families in English (K2), an academic word list (AWL), and an off-list, which is a list of all the words that did not fall into one of the three previous categories. All of the words contained in the off-list therefore appear beyond K2. Calculating the proportion of words that fall into each of the categories, both at T1 and T2, made it possible to assess whether students’ vocabulary seemed to include a greater amount of less-frequent words and therefore become more complex over time.
The reason for using LFP is that, contrarily to measures that indicate that learners are using a more varied vocabulary, it demonstrates whether the learners’ vocabulary is becoming more complex over time. As Horst and Collins (2006) point out, LFP offers an objective way of determining the degree of difficulty of a word, as it is based on recognized frequency lists, and therefore renders valid comparisons across studies.
Although much of the research on lexical complexity has been conducted with adult learners of various levels, Laufer and Nation (1995) suggest that LFP could be a valid way of measuring young learners’ lexical complexity. This position is further supported by Vermeer (2004), who postulates that word frequency can be used to operationalize the degree of difficulty of words and who used the MLR, which is similar to LFP, to compare the lexical richness of children learning Dutch as a first and as a second language. Her results led her to suggest that LFP could be a valid way of measuring LC with beginners, including children with low proficiency levels. Finally, Lo and Murphy (2010), also used LFP (through Vocabprofile) with young learners, and the conclusive results of their study led them to believe that LFP is indeed a valid measure of children’s lexical richness.
Reliance on French
In their study, Horst and Collins (2006) did not succeed, using only the measures provided by LFP, to determine that IE students’ lexical richness improved over a five-month program. They therefore turned to other measures, such as reliance on French words, to illustrate the progress made by the students.
In the data for the present study, it was observed that many students used French sporadically during the narrations to fill lexical gaps. It was specified, in the instructions for the task, that the students could, if needed, resort to using their L1, but they were instructed to limit their use of French as much as possible. Observing how the use of French changed over time therefore became another indicator of the development of LC.
Procedures for data analyses
The narrative data used for the analysis had previously been transcribed from the audio format into a written format. The resulting text contained no spelling mistakes, which helped assure proper analysis by the software. Grammatical errors, however, had been left intact.
Before the data was analyzed, all French words contained in the transcriptions were tagged. This was an important step because it made the French content easier to identify for the analysis and helped avoid any misinterpretation by the software: some words, such as original, exist both in English and in French. Such a word, if used in French in one of the students’ texts, could have been considered an English word by the software had it not been tagged.
In order to go through with the first phase of the analysis, which included measures of lexical diversity and density, several measures had to be calculated for each learner’s text. First, the total number of words7 contained in the text needed to be calculated. In order to do so, the transcriptions had to be altered by removing all pauses and sounds that did not constitute words. The overall number of words was then calculated using a word processor’s statistics function. Then, the total number of English words was calculated. This was also done using a word processor; all the words tagged as French words were removed, and then the total number of words was once more identified using the word processor’s statistics function. It is important to know that the French words that were removed from the text were copied and saved in another file, for an analysis of the use of French. Finally, texts were submitted to Vocabprofile first to identify the lexical density of texts (expressed as a ratio calculated by the software by dividing the number of lexical words by the total number of words), and then to identify the number of unique occurrences, which is the number of different words of the text.
Following this, the measure of lexical density was calculated by dividing the number of unique occurrences by the total number of English words and multiplying by 100. It should be mentioned that the text submitted to the software for this analysis was the version that had been pruned of all French words.
Next, lexical sophistication was assessed through LFP with Vocabprofile. The text submitted for this analysis was void of all pauses, sounds and French words. These items were removed so as not to influence the results of the analysis; all items that fell into the off-list were therefore words that appeared beyond the 2000 most-frequent in the language but not on the academic list. As for proper nouns, Vocabprofile has a function that considers all capitalized words as part of the 1000 most-frequent of the language. This allows the proper nouns to be considered among the easier rather than the more difficult words of the language. The fact that no punctuation, and therefore no capitalization of new sentences, was used in the transcriptions made the use of that function possible. Other words, such as onomatopoeias, were also considered to be among the first 1000 of the language. The results provided by the software included fours measures: the proportion of the 1000 (K1) and the proportion of the 2000 (K2) most-frequent words of the language, the proportion of academic words (AWL), and the proportion of off-list words. All of these measures were expressed as percentages.
To assess reliance on French, the total number of French words (tokens) for each text was calculated. This was done using the file containing the French words for each text, extracted from the data for the first analysis. The number of French tokens was then divided by the overall number of words in the text, and the measure was converted to a percentage to make comparisons between texts easier.
Table des matières
TABLE OF CONTENTS
1. INTRODUCTION AND PROBLEM STATEMENT
2. RESEARCH QUESTIONS AND HYPOTHESES
2.1. Research Questions
3. THEORETICAL FRAMEWORK
3.1 Lexical complexity
3.1.1 Conceptualization of lexical complexity
3.1.2 Measures of lexical complexity
3.2 Intensive English programs
4. LITERATURE REVIEW
4.1 Intensive English in Quebec
4.2 Development of lexical complexity
5. RESEARCH METHOD
5.1 Learning context
5.3.1 Vocabulary test
5.3.2 Oral narration
5.3.3 Procedure for oral narration
5.4 Measures of lexical complexity
5.4.1 Lexical density
5.4.2 Lexical diversity
5.4.3 Lexical sophistication (use of less-frequent words)
5.4.4 Reliance on French
5.5 Procedures for data analyses
5.6 Statistical analysis
6.1 Development of lexical complexity from T1 to T2
6.1.1 Lexical density
6.1.2 Lexical diversity
6.1.3 Lexical sophistication
6.1.4 Reliance on French
6.1.5 Overall observations
6.1.6 New measures of lexical complexity
6.1.7 Overall development of lexical complexity
6.2 Differences in the development of low-proficiency and high-proficiency learners
6.2.1 Results at T1 and T2
6.2.2 Gains in lexical complexity from T1 to T2
6.2.3 Vocabulary test
6.2.4. Overall differences between groups
6.3. Summary of results
7.1 Development of lexical complexity over the course of a five-month IE program
7.2 Differences in the development of low-proficiency and high-proficiency learners
7.3 Pedagogical implications
8.1. Further research