Introduction
Technological innovations in computer software and advancements in corpus linguistics have made it easier to analyse the vocabulary in a text. Corpus-driven analysis of multi-million-word corpora has enabled researchers to create word frequency lists and these have been used together with software to profile the distribution of words at different frequency levels within texts. This analysis of vocabulary provides an indication of the difficulty level of a text, the potential for vocabulary learning through reading or listening to the text, as well as the vocabulary knowledge required to understand different types of spoken and written discourse. This area of research, known as lexical coverage or lexical profiling, tends to fall into a number of different categories. The following themes highlight some of the major areas of lexical coverage research.
A. How does lexical coverage affect comprehension?
B. How many words do you need to know to understand written or spoken discourse?
C. To what extent do different text types provide opportunities for incidental vocabulary learning through reading and listening?
D. Which tools and methods can be used to investigate lexical coverage?
Studies in Theme A are experimental in nature and findings are typically based on the percentage of words that learners know in reading or listening passages and their comprehension scores for that passage. This area investigates the amount of lexical coverage needed to reach ‘adequate’ discourse comprehension. It is important to note that ‘adequate comprehension’ is not a clearly defined construct and what is considered adequate in one study might be quite different in another. This line of research indicates that 95% to 98% coverage is needed to establish adequate comprehension of different types of discourse. For reading comprehension, minimal (95%) and optimal (98%) thresholds have been suggested (Laufer & Ravenhorst-Kalovski, 2010Footnote *) while for listening comprehension, 95% lexical coverage is recommended for good comprehension (e.g. van Zeeland & Schmitt, 2013*). These findings indicate that a reasonable comprehension level may be achieved at a density of lexical coverage above 90%, suggesting that learners need to know a very high percentage of words in a text in order to understand it.
Building on the findings in Theme A, the second area of lexical coverage research (Theme B) uses word frequency lists and a representative corpus of a domain to examine the number of words necessary to reach the coverage figures associated with adequate comprehension. For example, 3,000 word families plus proper nouns (e.g. ‘Harry’, ‘London’) and interjections (e.g. ‘oh’, ‘hey’), are sufficient to reach 95% coverage of movies (Webb & Rodgers, 2009a*), TV programs (Webb & Rodgers, 2009b*), and conversation (Nation, 2006*). Findings from this line of research give an indication of the coverage of high-, mid-, and low-frequency words, as well as academic vocabulary in different types of discourse. This is particularly useful in distinguishing the relative values of words in different types of discourse. For example, West's (1953)* General Service List (GSL) was developed to represent the most useful 2,000 words in the English language and provides around 70–90% coverage of discourse with the amount of coverage dependent on the type of text. Coxhead's (2000)* Academic Word List (AWL) covers around 10% of academic written texts and around 4% of academic spoken discourse (Dang & Webb, 2014*).
Studies in Theme C examine the number of encounters with words to determine the potential for incidental vocabulary learning through reading, listening, or viewing the text. These studies draw on the lexical profiling research in Theme B by investigating the potential for learning words incidentally that are deemed to be unknown due to their word frequency level. For example, Webb and Rodgers (2009b*) looked at repetition of vocabulary that was less frequent than the 3,000 word level to determine the potential to incidentally learn lower frequency words through watching television programs. Cobb (2007)* examined the potential for readers to learn the most frequent 3,000 word families through reading different types of text. Studies investigating incidental vocabulary learning have shown that the more often words are encountered, the more likely they are to be learned (e.g. Horst, Cobb, & Meara, Reference Horst, Cobb and Meara1998; Rott, Reference Rott1999; Waring & Takaki, Reference Waring and Takaki2003; Webb, Reference Webb2007). By examining the extent to which text types such as graded readers, listening passages, and classroom-based teacher talk create the opportunity for incidental vocabulary acquisition, we can better understand where vocabulary learning is more likely to occur.
Studies under Theme D relate to lexical profiling methodology. These include discussions of issues that should be considered when designing and evaluating studies of lexical coverage. For example, Reynolds (2013*) questions whether the unit of counting in lexical profiling studies should be the ‘word family’. Word families are made up of a headword (e.g. ‘access’), its inflections (‘access’, ‘accesses’, ‘accessing’, ‘accessed’), and derivations (‘accessible’, ‘accessibility’, ‘inaccessible’, ‘inaccessibility’) and are the typical unit of counting in the studies found in this timeline (e.g. Nation 2006*; Webb & Paribakht 2015*). The rationale behind using word families is that if a learner knows the headword, he or she may understand the meaning of the inflections and derivations when they are encountered in context. Currently, there is a debate in lexical profiling studies – whether to use word families or lemmas as a unit of counting. Although there are some proponents of the lemma-based approach (e.g. Gardner & Davies, Reference Gardner and Davies2014), there is little research investigating this issue and an answer for this question remains to be determined. Thus, further research needs to be done in this area (see Nation, 2016 for discussion*). Another issue is the vocabulary size at which proper nouns (e.g. Baltimore, Harry) should be considered to be understood by learners. Research under this theme also discusses tools (e.g. software, frequency lists) that are commonly used to perform the coverage analysis.
In addition to the four themes discussed in this manuscript, we were tempted to include studies of the lexical frequency profiles of learner language use (e.g. Laufer & Nation, Reference Laufer and Nation1995; Meara, Reference Meara2005). Although these studies do have a similar methodology to those of Theme B, their focus on productive knowledge contrasts the research in this timeline which is all focused on receptive vocabulary knowledge. Lexical frequency profiling of learner language measures the extent of lexical sophistication and thus might be better incorporated in timelines that are devoted to lexical richness. Another theme not included in this timeline is word lists (e.g. high-/low-frequency lists, academic and technical word lists). There has been ample research on word frequency lists since the early 1950s that has resulted in the creation of lists such as West's (1953) GSL and Brezina and Gablasova's (Reference Brezina and Gablasova2015) New GSL. Similarly, there are several lists of academic vocabulary such as Coxhead's (2000*) AWL and Gardner and Davies’ (Reference Gardner and Davies2014) Academic Vocabulary List. In addition, there are lists of academic collocations (Ackermann & Chen Reference Ackermann and Chen2013), phrasal expressions (Martinez & Schmitt, Reference Martinez and Schmitt2012), and lexical frames common in academic and spoken English (Gray & Biber, Reference Gray and Biber2013). In this timeline, frequency lists such as the GSL (West 1953*) and Nation's (2012*) British National Corpus (BNC) are included, together with the British National Corpus/Corpus of Contemporary American (COCA) word lists that have been used in analyses within the articles summarized in the timeline. However, there was not sufficient space to include all of the available word lists, and ideally these studies would be detailed in a separate timeline.
Overall, our timeline reveals a relatively steady number of studies published on lexical coverage from the 1980s. Research might have been spurred by the introduction of technological tools that reduce the challenges of corpus-driven studies. The availability of the RANGE program (Heatley, Nation, & Coxhead 2002*) and the development of the BNC word lists (Nation, 2006*) have likely led to an increase in the number of studies conducted in recent years. With a relatively small number of studies investigating the relationship between lexical coverage and comprehension, further research is still warranted. However, there is value in researching each of these themes in more depth (Schmitt et al., 2017*). Moreover, as advances in lexical profiling tools occur there will likely be additional innovation regarding how these lines of research are conducted.
Ulugbek Nurmukhamedov is an Assistant Professor of TESOL in the M.A. TESOL program at Northeastern Illinois University (Chicago, Illinois). His research focuses on second lanaguge (L2) vocabulary instruction and computer-assisted language learning. He has published in these areas in journals including TESOL Journal, ELT Journal, Writing & Pedagogy and International Journal of Lexicography.
Stuart Webb is a Professor of Applied Linguistics in the Faculty of Education at the University of Western Ontario. His research interests include vocabulary studies, extensive reading and listening, and language learning through watching television. His articles have been published in journals such as Applied Linguistics and Language Learning. His latest book (with Paul Nation) is How vocabulary is learned (2017, Oxford University Press).
