1. Developing a practical model of vocabulary acquisition

Paul Meara first noted the lack of an overall theory of vocabulary acquisition in Reference Meara1983, and this is still true today. Of course, there have been numerous theories which cover limited aspects of vocabulary learning. For example, the Revised Hierarchical Model (Kroll & Stewart, Reference Kroll and Stewart1994) posited that the psycholinguistic pathway to L2 meaning is initially through L1 translation equivalents. Ellis (Reference Ellis2002) reviews how frequency partially drives language acquisition, including individual words and formulaic language. Brown and Payne (1994, cited in Hatch & Brown, Reference Hatch and Brown1995) proposed a five-step model of vocabulary learning, although it only dealt with form and meaning. In fact, the very few theories/models available have tried to explain how basic form-meaning connections are created (sometimes in relation to L1 lexicon entries), but there are still none that explain how the many different components of lexical mastery are developed. This is partly because vocabulary knowledge remains an extremely complex construct, which resists any single explanation. This consists of (ideally) very large numbers of both individual words with their inflections and derivatives, and of formulaic sequences. Every lexical item has its own characteristics which may make it relatively easier or more difficult for any particular learner to acquire, with L1 being a major factor (Laufer, Reference Laufer, Schmitt and McCarthy1997). It is probably impossible to devise any explanation of acquisition which can reliably describe how every intrinsically different lexical item is acquired by learners of various L1s.

The best-known and most widely used framework is Nation's (Reference Nation2013) division of vocabulary knowledge into nine components of ‘word knowledge’ (e.g. spelling, word parts, meaning, grammatical functions, collocation). The framework has been extremely useful in describing the totality of what learners need to know, but says nothing about any hierarchical ordering (i.e. which components are typically learned before others, or should be taught before others). This limits its pedagogical value, because it is unclear how the various components relate to each other, and how to best prioritize the various components when teaching.

A limited number of studies have measured knowledge of multiple components from the framework for various research purposes (e.g. Schmitt, Reference Schmitt1998; Webb, Reference Webb2005; Chen & Truscott, Reference Chen and Truscott2010), and generally found that the various measured components were interrelated, but that some appeared to be acquired before others. The most comprehensive of these multicomponent studies was carried out by González-Fernández and Schmitt (Reference González-Fernández and Schmitt2019). They found that all of the components they measured (form-meaning, derivatives, collocation, polysemy) were known to a recognition level before any of those aspects were known to a recall level. This suggests the key descriptor of vocabulary knowledge may not be the word knowledge components themselves (as implied by Nation's framework), but rather recognition vs recall mastery of those components. They also found knowledge of the components conformed to an implicational scale, i.e. that some components were learned before others.

If their acquisition order is shown to be generalizable, this would give teachers and testers a blueprint for how (at least a portion of) vocabulary knowledge is acquired. But the research first needs extending to other word knowledge components and other L1s. This leads to my first two research suggestions:

In an ideal world, it would be best to measure all of the word knowledge components (both receptively and productively) in the same test battery. However, this seems virtually impossible in practical terms, as González-Fernández and Schmitt's battery of only four components took between 2.5 and 3.5 hours for most learners. Their study could be usefully replicated (Porte, Reference Porte and Porte2012), but it should also be extended by measuring word knowledge components they did not measure (spelling, pronunciation, word parts, associations, grammatical functions, constraints of use). By also including one or two of the components they did measure as a comparison, it should be possible to relate any new study's results to the existing implicational scale. For example, González-Fernández and Schmitt found that recognition of the form-meaning link was typically mastered before recognition of derivatives. If a new study found that recognition of correct spelling was mastered before recognition of the form-meaning link, it could be inferred that recognition of correct spelling is also typically mastered before recognition of derivatives. The comparison would be further enhanced if González-Fernández and Schmitt's target words were used in the new studies. Another approach would be to measure all components, but with different target words, carefully controlling the target items to be similar across components. Although this has the limitation of not measuring the same words across the different components, the value of more components being measured concurrently may be worth the trade-off. It also eliminates any possibility of test contamination (where exposure to a target word on one test in the battery may give hints to answering a subsequent test).

While cross-sectional studies can be informative, longitudinal studies are usually better at describing acquisition. This would require re-administering the test battery again after a semester or year, to see how the various components were enhanced and to what degree. If this proved difficult to arrange for groups of students, such a longitudinal study might be well suited to a case study approach, if several very cooperative learners could be found.

González-Fernández (Reference González-Fernández2018) found the same results with Spanish learners of English (cognate language) and Chinese learners (noncognate language). But it would also be useful to extend the research to other L1s, as previous research (e.g. Otwinowska & Szewczyk, Reference Otwinowska and Szewczyk2017) shows that L1 is an important factor in L2 vocabulary acquisition.

Eventually it should be possible to model which word knowledge components were typically learned before others (for most words and most learners), or alternatively, whether some components are inherently idiosyncratic regarding their learning burden (i.e. whether they are learned earlier or later depends mainly on the particular word or learner). It should also become clear whether recognition mastery always precedes recall mastery for all components, or just for the four that González-Fernández and Schmitt measured. The results would inform pedagogy in showing where most teaching effort needs to be applied: in moving vocabulary knowledge from Ø → receptive, or from receptive → productive. It would also inform testing, because if an implicational scale could be identified for all/most components, it would only be necessary to test a few components to infer what other components were known.

2. Understanding how vocabulary knowledge develops from receptive to productive mastery

Point 1 is about understanding the various aspects of vocabulary knowledge (e.g. spelling, meaning, word parts, collocations) and how they relate to each other. This point focuses more on acquisition, especially how to move learner knowledge to the more advanced productive level. There is plenty of research, along with teacher experience, to show that receptive mastery of a lexical item (ability to understand it when listening or reading) is generally stronger than productive mastery (ability to produce it in one's own speech or writing). Virtually all research which includes both receptive and productive measures show higher receptive scores (e.g. Melka, Reference Melka, Schmitt and McCarthy1997; Laufer & Goldstein, Reference Laufer and Goldstein2004; Webb, Reference Webb2005). (See Point 5 below for more discussion on receptive/productive and recognition/recall measurements.) But research generally just reports the receptive and productive scores, without considering the relationship between the two. In simple terms with a continuum-based illustration, the relationship might be exemplified like this (although note that a ‘states’-based conceptualization is also possible (e.g. Meara, Reference Meara, Schmitt and McCarthy1997).

Many might think that the intervals (i.e. learning burden) is about equal between Ø → R and R → P for most words, as illustrated in Figure 1. Others might think the major learning is in learning the word in the first instance to receptive mastery, and after that productive mastery follows on without too much trouble (Figure 2).

Figure 1.

Figure 2.

But I think the research indicates that the opposite is true: that learning most words to receptive mastery is relatively easy; it is enhancing that knowledge to productive mastery which is the real challenge (Figure 3).

Figure 3.

It is not difficult to understand why this might be so, as I explain in Schmitt (Reference Schmitt2014). In reading, to comprehend a word, it might be enough to be able to recognize the spelling of a word and remember its meaning. All or most of the other word knowledge components are already in the text (e.g. its collocation and derivative form), which may or may not be utilized to aid comprehension. But when writing, one must know and produce all of the various components independently without prompts. The same holds true for listening/speaking.

I think it is safe to assume that most learners want to be able to employ their vocabulary in their speaking and writing. So for me, an under-researched area of particular interest is how to push learners’ knowledge from receptive mastery to the point where they can independently use lexical items fluently and appropriately in their own output. This leads to my next research suggestion:

Note that this task is not about determining which vocabulary activities are best to begin the learning process, but which are best to enhance partially known vocabulary. These questions will probably require a pretest-treatment-posttest design, with target lexical items being tested both receptively and productively. The targets will need to be known receptively (but not productively) at the beginning of the study. Because it is difficult to know how well individual lexical items are known, a range of items will probably have to be given in the pretest, and only those known receptively selected for inclusion in the study.

Most studies have used multiple-choice items as their receptive format, but these are probably not the best choice, because guessing will almost certainly inflate the scores to some unknowable extent (Gyllstad, Vilkaite, & Schmitt, Reference Gyllstad, Vilkaite and Schmitt2015). I suggest using meaning recall formats for the receptive tests and form recall formats for the productive tests (see Point 5). This virtually eliminates guessing, and also better matches the lexical knowledge necessary for the receptive and productive skills (reading/listening; writing/speaking) (Schmitt, Reference Schmitt2010, Reference Schmitt2014). Using L1 translations is a good way to operationalize these recall measures if one is testing a homogenous L1 group. With heterogeneous groups, it will be necessary to use alternative meaning-based prompts, such as L2 definitions or higher-frequency L2 synonyms.

Research suggests that it takes some time to build up to productive mastery, which prompts Read (Reference Read2000, p. 154) to pose the very interesting question: ‘Is there a certain minimum amount of word knowledge that is required before productive use is possible?’ Answers to receptive-productive questions like this will probably require longer-term longitudinal studies. Just how long is probably a research question in itself, but studies with too short a treatment period (I speculate less than six months) will likely show little change from receptive to productive knowledge. (There is also the issue of appropriacy of use, which I do not have space to address in this piece.)

Any one iteration of a learning activity will be unlikely to result in truly productive knowledge, so research should also enquire how many repetitions it takes to move knowledge to the productive level. The most effective methodology might entail a combination of activities, and so this should also be explored. Research suggests learners must practice vocabulary productively to reach productive mastery (e.g. Laufer, Reference Laufer2005), so the activities investigated will almost certainly require learners to produce output, rather than just practice receptively. It will only be possible to study limited numbers of activities in any one study, so I foresee this strand requiring a number of studies, and only by eventually synthesizing them will we get the bigger picture of the kind of vocabulary activities we should promote after initial (receptive) learning is in place.

3. Getting lexical teaching/learning principles into vocabulary and language textbooks

There has been enough research on vocabulary to suggest a number of principles of good vocabulary instruction. Hunt and Beglar (Reference Hunt and Beglar1998) proposed seven, including:

• • diagnose which of the 3,000 most common words learners need to study

• • provide opportunities for elaborating word knowledge

• • provide opportunities for developing fluency with known vocabulary

• • experiment with guessing from context

In Schmitt (Reference Schmitt2008), I added another six points, including:

• • Learners need large vocabularies to successfully use a second language, and so high vocabulary targets need to be set and pursued.

• • Vocabulary learning is a complex and gradual process, and different approaches may be appropriate at different points along the incremental learning process.

• • Once this initial meaning-form link is established, it is crucial to consolidate it with repeated exposures.

• • It is also important to begin enhancing knowledge of different aspects of word knowledge. Some of these may be usefully learned explicitly (e.g. knowledge of derivative forms), but the more ‘contextualized’ word knowledge aspects (e.g. collocation) are probably best learned by being exposed to the lexical item numerous times in many different contexts.

While principles like these are sound, and would surely lead to better vocabulary pedagogy, the problem is that most teachers do not have the time, expertise, or resources to consistently put them into practice. Take the idea of recycling (repeated exposures). For example, if a teacher taught ten new words a class, it might be possible for her to conscientiously recycle all those words for a while. But eventually, it will become unmanageable to recycle 100–200+ words in a systematic manner.

This task needs to be done by someone with the time and resources to carefully consider (1) which vocabulary is most beneficial for learners (largely high-frequency vocabulary); (2) which activities most effectively teach these lexical items (some activities may be better suited to certain lexical items than others); and (3) how to systematically build recycling and enhancement into a course longer-term. Syllabus designers, and particularly textbook writers (who typically take a year or more to write their books), are best positioned to organize this thoughtful development of vocabulary.

Unfortunately, most textbooks lack any obvious systematic approach to vocabulary. Many (most?) textbooks are built around some kind of reading passage, and the vocabulary highlighted largely depends on the topic of that passage. This vocabulary is seldom repeated later as the topics for the next chapters change. This can lead to quite haphazard vocabulary selection (Schmitt & Schmitt, Reference Schmitt and Schmitt2014). For example, Hsu (Reference Hsu2009) examined 20 international General English textbooks, ranging from low intermediate to advanced levels. She found little uniformity between the level of the textbook and the vocabulary required both within and across textbook series. For example, the advanced Reading for real required 4,000–4,500 word families to reach 95% coverage, while the low intermediate Reading for success 2 required 7,000–7,500 families. Another ramification of unsystematic vocabulary selection is a lack of recycling. In one example of this, Matsuoka and Hirsh (Reference Matsuoka and Hirsh2010) analysed the vocabulary from the best-selling New Headway Upper Intermediate English textbook and found that of the 1,005 beyond-high-frequency word families, 66.4% occurred only once, and only 12.1% occurred five times or more. Finally, the vocabulary activities in textbooks are often quite limited in what they teach. Brown (Reference Brown2010) found that the nine General English textbooks he analysed focused mainly on meaning and form, with some attention to grammatical function and spoken form, but did little to enhance knowledge of other types of word knowledge like collocation, derivative form, or constraints on use. I think it would be useful to find out why vocabulary teaching principles are generally not making it into textbooks:

I am not aware of any research which looks at how publishers and textbooks writers go about selecting the vocabulary for their books, and how they develop the activities for that vocabulary. Therefore, a logical first step is to interview them, to determine how much they take account of vocabulary research in their textbook development. Some obvious lines of questioning include the following:

• • Are textbook writers aware of the vocabulary research in the first place?

• • Is vocabulary development an essential component as they conceptualize their books?

• • If so, how do they attempt to operationalize it in their textbooks?

• • What are the constraints which keep them from more fully applying the principles in their books?

• • If they do not believe it is important to highlight vocabulary, is this because they do not believe the end consumer wants it? Or is vocabulary simply too difficult to organize over time? Or are there other reasons?

• • What guidance/guidelines, if any, do writers get from publishers concerning vocabulary content? What guidelines would they like to get that would make a more principled and systematic approach to vocabulary inclusion in textbooks possible?

In addition to this publisher-based research, it would be very useful to explore teachers’ beliefs, attitudes, and usage of textbooks, to consider (1) how these might affect publishers’ attitudes, and (2) how the presently available vocabulary textbook activities are being utilized.

Most schools and teachers rely heavily on textbooks for language content, so I believe it is essential that these books incorporate sound vocabulary principles in a way that the average teacher would never have the time to do properly, i.e. textbooks are the essential conduit for research to influence practice. It is time to find out why this influence has been so meagre to date, and to think of ways of redressing the problem.

4. Exploring extramural language exposure and how it can best facilitate vocabulary acquisition

While many learners around the world struggle to learn 2,000 English words after schooling of hundreds, or even a thousand, hours (Laufer, Reference Laufer2000), children in other countries come into school knowing substantial amounts of English vocabulary. This is especially true in northern Europe. What makes these countries different? It probably has something to do with the social attitude that English is useful and worth knowing (e.g. de Wilde & Eyckmans, Reference De Wilde and Eyckmans2017). But the more important factor is likely to do with the exposure to English these children enjoy. Recent research has shown that children in some countries are exposed to English for several hours per week (e.g. Lindgren & Muñoz, Reference Lindgren and Muñoz2013; Jensen Reference Jensen2016; de Wilde & Eyckmans, Reference De Wilde and Eyckmans2017). This has led to young learners having impressive English vocabulary sizes for their age. For example, in Belgium, learners scored a mean of 66.20 out of 108 on the Peabody Picture Vocabulary Test, despite not having had any formal English instruction when they took the test (de Wilde & Eyckmans, Reference De Wilde and Eyckmans2017). In Iceland, Lefever (Reference Lefever2010) found that most children before the start of formal education could already understand basic spoken English, many could participate in simple conversations in English, and most were in the first stages of understanding written English. Clearly, these learners are acquiring considerable vocabulary and language outside the classroom, and some must come from media exposure (Kuppens, Reference Kuppens2010). There is even some evidence that out-of-class exposure has a larger effect than length of instruction (Peters, Reference Peters2018).

Three types of extracurricular exposure (sometimes referred to as extramural exposure) usually found to be important include: watching English-language television (with subtitles or captions) (e.g. Kuppens, Reference Kuppens2010; Lindgren & Muñoz, Reference Lindgren and Muñoz2013), playing computer/internet English games (e.g. Kuppens, Reference Kuppens2010; Sylvén & Sundqvist, Reference Sylvén and Sundqvist2012; Jensen, Reference Jensen2016), and particularly for older learners, consuming English-language reading material, whether on paper or on the internet (e.g. González-Fernández & Schmitt, Reference González-Fernández and Schmitt2015; Garnier & Schmitt, Reference Garnier and Schmitt2016; Macis & Schmitt, Reference Macis and Schmitt2017). However, these studies do not typically indicate the relative importance of these various kinds of extramural exposure, nor examine in much detail the precise nature of the exposure (e.g. the prominent English features in the computer games being used). Having finer-grained detail about the nature of extramural exposure, and studying how this directly leads to L2 acquisition, would allow more concrete suggestions about how to promote the most effective use of extramural exposure in a range of contexts, and how to best integrate it with classroom instruction.

Most research into extramural exposure has used surveys or interviews to determine the category (e.g. watching TV or playing computer games in the L2) and the extent of exposure, and then matched the answers with learner scores on vocabulary and other language tests. As shown above, this methodology has demonstrated that extramural exposure of various types is related to better L2 proficiency. What is less clear is how maximize the benefits of this exposure for vocabulary acquisition. Research on several approaches could be useful. The first involves marrying supplementary explicit instruction to the extramural exposure. Two small studies (Miller & Hegelheimer, Reference Miller and Hegelheimer2006; Ranalli, Reference Ranalli2008) have shown that supplementary materials (e.g. word lists of the vocabulary in SIM games and vocabulary exercises) lead to better vocabulary learning when playing the games. Potential research studies could match extramural exposure with a variety of supplementary materials provided in language classrooms to determine which types of material are the most effective in promoting vocabulary growth. Some obvious candidates include: lists of words from the extramural exposure, a variety of explicit exercises focusing on those words, and strategy training exercises to help learners manage and learn new words (e.g. exercises which train learners to concurrently focus on both L2 audio and L1 subtitles when watching subtitled media (Kuppens, Reference Kuppens2010)). This approach could be informative about which supplementary materials are most effective with particular types of extramural exposure, or whether some combination is best. It would be useful to include three conditions: one in which learning gains from the extramural exposure alone were measured, one where the learning from just the explicit materials was measured, and a combined condition. This would illustrate how much learning accrues from the interaction of exposure and materials, and how much simply comes from the explicit materials themselves.

Another approach would involve using multiple vocabulary tests to determine what vocabulary knowledge comes from different types of extramural exposure. By using both recognition/receptive and recall/productive tests, it should be possible to better describe the degree of mastery that extramural exposure typically leads to.

Gaming is an important type of extramural exposure for many learners, and it would be useful to better understand what kinds of games are most beneficial. Jensen (Reference Jensen2016) is a model of how finer-grained analyses can prove informative. She classified the games into the following categories: games with both oral and written English input, with oral but no written English input, with written but no oral English input, with oral English input and written Danish input, and Danish oral input and written English input. She found that gaming with ‘both spoken and written English input’ was significantly related to scores on the Peabody Picture Vocabulary Test, and to a much lesser extent, gaming with ‘only written English input’. Likewise, different games might be better for boys vs girls. Sylvén and Sundqvist (Reference Sylvén and Sundqvist2012) found that not only did boys engage in gaming more than girls, but they tended to play different types of games; they preferred first-person shooter or multiplayer games, while girls tended towards single-player simulation games. This makes a difference because Sylvén and Sundqvist consider multiplayer games (particularly Massively Multiplayer Online Role-playing games, such as EverQuest 2 and World of Warcraft) ‘highly beneficial for L2 acquisition because they provide learners with opportunities for engagement with rich target language input [and output] as well as for scaffolded interaction’ (Reference Sylvén and Sundqvist2012, p. 315).

Further finer-grained research into games could explore whether different types of game promote different types of vocabulary (e.g. Do shooter games highlight ‘action’ verbs? Do multi-player games promote the vocabulary of commanding or negotiating?). An analysis of the vocabulary in games could show what words are being presented, e.g. the percentage of high-/mid-/low-frequency words through a Lextutor analysis (www.lextutor.ca), vocabulary that realizes language functions through referral to references such as Nattinger and DeCarrico (Reference Nattinger and DeCarrico1992), and specialist vocabulary. A lexical analysis could also show whether words are being repeated enough to make incidental learning viable (Cobb Reference Cobb2007), although this would also require the additional step of determining the number of repetitions required for this kind of input, as it might differ from reading, on which most incidental learning research has been carried out. Other research directions include exploring the role of gaming visual input in vocabulary learning (e.g. in what ways does the visual input support the acquisition of vocabulary?), and the nature of gaming interaction (e.g. what kinds of interaction are most facilitative for vocabulary learning?).

5. Developing more informative measures of vocabulary knowledge

Any research into vocabulary acquisition is only as good as the tests used to measure the learning, and most vocabulary tests are not validated to any great degree (Schmitt, Nation, & Kremmel, Reference Schmitt, Nation and Kremmel2019). Most studies have used only a single measure, so in Points 1 and 2, I suggested the value of using test batteries, including measures of multiple types of word knowledge, and at receptive and productive masteries (see also Webb, Reference Webb2005). Furthermore, most studies to date have measured only some aspect of the form-meaning link, often with a multiple-choice format (i.e. at the ‘recognition’ level of mastery), so I have also argued that it is better to measure form-meaning knowledge at the levels of meaning recall and form recall (Point 2).

Until now, I have used the terms receptive and productive quite loosely when speaking about testing, as have most commentators. What we really want in vocabulary measurement is the ability to infer what learners can do with the target words. (Nobody interprets test scores as simply words that learners can answer on a vocabulary test!) Receptive knowledge entails knowing a lexical item well enough to extract communicative value from speech or writing. Productive knowledge involves knowing a lexical item well enough to produce it when it is needed to encode communicative content in speech or writing. That is, receptive/productive knowledge of vocabulary is usage-based, and should presumably be measured with skill-based instruments. However, it is hardly ever measured this way.

This is because it is very difficult to measure vocabulary knowledge in context. All skills require more than just vocabulary knowledge. A target word interacts with the other words in a text in semantic, grammatical, morphological, and phraseological ways, and thus it is difficult to measure understanding or production of a single lexical item in context without the context becoming part of what is being measured. This is why vocabulary is often measured in isolation, and usually at just the form-meaning link level. Typical tests involve measuring target words with either a multiple-choice or matching test format in which learners recognize the correct form or meaning, or with the meaning given and the L2 word form needing to be recalled (form recall). Alternatively, the form can be given, with the meaning needing to be recalled (meaning recall).

Receptive tests are easier to develop because test writers are in control. They can select the target items and embed them in contexts which are not too informative so that the meaning cannot be inferred from the context. But just because a learner can understand a lexical item in one non-defining test context, does that mean they will be able to understand it in a variety of real-world contexts? Then there is the issue of oral/written language. Does the ability to comprehend a lexical item on a written test imply the ability to understand it when listening? Or vice versa? These issues suggest the need to better understand how receptive test formats work:

One approach to this task is to investigate current item formats in terms of the information they give about the ability to understand words when reading or listening. A recent study showed that different existing formats do have explanatory power in predicting receptive skills, in this case, overall vocabulary size in predicting general reading proficiency. Laufer and Aviad-Levitzky (Reference Laufer and Aviad-Levitzky2017) found that both meaning recognition and meaning recall vocabulary measures correlated with reading scores at .91–.92. But it would be even more interesting to discover what the test formats showed about the ability to understand the particular target words when reading. Target words could be placed in existing test formats, and then also embedded in multiple authentic reading and listening contexts. Comprehension questions would then test understanding that was directly reliant on knowledge of the target words. It would also be important to control and limit the surrounding context so answers could be not arrived at through lexical inferencing strategies. The ability of learners to comprehend the target words in a variety of reading/listening texts would be the benchmark of what learners can do with the target words. This benchmark could then be compared to the scores from the various item formats to see how well each format reflects the reading/listening comprehension of the target words. This type of concurrent validation would be very useful in knowing how to interpret the scores from the various test formats. If a quick and simple item format was shown to reliably predict the ability to comprehend a word, this would be a very good result, as large numbers of words could be tested, and practitioners and researchers could be confident whether learners actually knew the words well enough to understand them in real-world language.

However, it may turn out that existing formats are unable to adequately describe comprehension. A second approach would be to develop new formats. Space limitations prohibit discussion of the many possible formats, but good places to look for inspiration are Tools for Researching Vocabulary (Meara & Miralpeix, Reference Meara and Miralpeix2017) and Paul Meara's lognostics website (www.lognostics.com), which illustrate a number of interesting experimental formats. But whether Meara's formats or completely original ones are explored, they will need the same kind of concurrent usage-based validation evidence as discussed above.

Turning to productive tests, things get trickier. Most measures are post-hoc computerized analyses of learner output (e.g. type-token ratios, lexical sophistication, lexical density – see Coh-Metrix (www.cohmetrix.com) and TAALES (www.kristopherkyle.com/taales.html). But these merely describe the lexical items produced, not a learner's complete productive vocabulary. Just because a learner produces one word in speaking or writing (e.g. difficult), it does not mean that they do not know other possibilities, like onerous or hard; they just happened to select difficult in that instance. Any output a learner produces will inevitably contain only a small percentage of the words they are capable of producing. This makes it almost impossible to directly measure the complete range of a learner's productive vocabulary. However, there are ways in which it might be indirectly measured, and I think these are worth pursuing.

One indirect approach which I think holds promise (and is relatively doable) entails potentially establishing the ratio between receptive and productive knowledge. There have been many studies which compared ‘receptive’ and ‘productive’ knowledge, but they have mainly looked at only the form-meaning link (e.g. Fan, Reference Fan2000; Laufer & Goldstein, Reference Laufer and Goldstein2004). My suggestion involves sampling a large, representative number of words, either from frequency lists or from a forthcoming list of the best-known lemmas in English (Schmitt et al., Reference Schmitt, O'Sullivan, Anthony, Dunn and Kremmelforthcoming).Footnote ¹ The researcher would then have learners say/write sentences for each of these words. From my experience, participants often produce sentences which do not really show the meaning/usage of the target word (including one very naughty example ‘I like the word raspy’), so it might be necessary to do the task one-on-one to probe further and ask for new sentences if necessary. Developing a set of criteria for acceptable answers would also be important. This approach would be very time-consuming and would probably have to be carried out as case studies. But the end result should be a good estimate of the learners’ productive vocabulary. This estimate could then be compared to scores from one or more receptive tests. If the productive/receptive ratio was relatively constant across learners, then only a receptive test need be given, and a learner's productive vocabulary size could be calculated using the established ratio. Even if the ratio proves not to be stable across learners, the results of this study would still be a valuable contribution towards understanding the relationship between productive and receptive vocabulary.

6. Measuring fluency as part of vocabulary competence

Being able to employ vocabulary in the four skills involves more than just knowledge of lexical items. It is also necessary to be able to listen and speak in real time, or interlocutors will soon tire of the halting, disfluent communication. Likewise, if learners are unable to read at a sufficient rate, the slow word-by-word decoding makes it difficult to understand the constantly developing meaning structure of an extended text. For fluent reading, words need to be recognized quickly, automatically, and accurately (Grabe, Reference Grabe2009). This leads Daller, Milton, and Treffers-Daller (Reference Daller, Milton and Treffers-Daller2007) to propose fluency as one of the three basic components of vocabulary knowledge (in addition to size and depth). However, while there has been considerable research into vocabulary size and some into depth (see Schmitt, Reference Schmitt2014 for an overview), there has been very little into measuring the fluency with which vocabulary can be employed, or how it can be acquired. But if fluency is seen as an essential requirement for using vocabulary, we should begin measuring it in our vocabulary research.

This task could be seen as an extension of Point 5's encouragement to develop and use more informative measures of vocabulary. But how can we measure fluency in vocabulary use? The trick is measuring the fluency of the individual lexical items of interest, rather than just overall reading/listening/speaking/writing speed. I think there are a number of methodologies that might prove useful for this. For reading, eye-tracking can show the fluency in which individual words and phrases are read, in terms of the number and duration of the eye fixations on the target items (e.g. Pellicer-Sánchez, Reference Pellicer-Sánchez2016). Timed lexical decision tasks have also been widely used for the measuring fluency at the individual word level. In writing, keystroke-logging software can show how fluently the target items are typed, in terms of duration and corrections (e.g. Miller, Lindgren, & Sullivan, Reference Miller, Lindgren and Sullivan2008). For listening, measuring comprehension of ideas realized by target vocabulary in real-speed speech may be workable. In speaking, psycholinguistic/technical measurement options are available (e.g. the use of PRAAT (www.fon.hum.uva.nl/praat/) for the analysis of speech), but for the purposes of measuring how well vocabulary is produced orally, using a panel of judges may well be just as good and far more practical.

Once a suitable technique for measuring fluency has been selected for a particular skill, a fluency test could be added to the more typical vocabulary measures. A number of vocabulary learning activities and exercise types could be explored in a pretest-treatment-posttest design with a battery of tests: minimally recognition/recall tests, better yet measures of vocabulary in use (Point 5), and best of all also including a measure of fluency. This would give a much more comprehensive view of the type and amount of learning that accrues from different task types. I would guess that some activities/exercises are good at promoting some aspects of lexical proficiency, but less useful for other aspects. To give you some ideas, the following studies have looked at the effect of vocabulary activities on the development of written fluency: Snellings, van Gelderen, and de Glopper (Reference Snellings, van Gelderen and de Glopper2002), Fukkink, Hulstijn, and Simis (Reference Fukkink, Hulstijn and Simis2005), Elgort (Reference Elgort2011), and Pellicer-Sánchez (Reference Pellicer-Sánchez2015). Note that there has been relatively little fluency research into auditory recognition or spoken lexical production.

Fluency would probably be most useful for charting the incremental improvement of both receptive and productive levels of mastery, i.e. from the slow, halting ability to comprehend/produce lexical items to quick and automatic ability. It would also allow us to reconceptualize the vocabulary learning timelines in Point 2 to something more nuanced, which might be key to understanding the movement from receptive to productive mastery (Figure 4):

Figure 4.