1. Introduction
An understanding of the nature of L2 lexical knowledge and how it develops has important theoretical and educational implications. Because real language use, such as fluent reading, requires efficient and accurate access to word form and contextually appropriate meaning, we need measures of learning that can assess these aspects of word knowledge. However, L2 vocabulary learning studies tend to measure offline explicit word knowledge (Elgort, Reference Elgort2018) that may not be accessible in online language use. Measuring whether a learner can access learned words accurately and fluently, under time pressure, creates a more precise picture of word knowledge at a given point in the acquisition process. In developing and validating theoretical and computational models of L2 word learning and processing, researchers need both offline measures (such as explicit recognition and recall) and online measures (such as online access to formal and semantic representations, under conditions that significantly restrict access to task strategies). Using tests, tasks and procedures that evaluate both types of knowledge can help researchers identify optimal learning and instructional approaches that can be used in formal and informal learning contexts.
In this article, I propose replications of two word learning studies that investigated quality of lexical knowledge for recently learned L2 (English) words. Both studies were motivated by theoretical frameworks and models of visual word recognition and processing; both used an experimental task from cognitive psychology, form-priming, alongside traditional tests of word knowledge, to measure lexical quality of known and recently learned vocabulary. These studies represent a growing trend of interdisciplinary research investigating issues in applied linguistics by combining theoretical frameworks and research methods from education, second language acquisition (SLA) and cognitive psychology (Mackey & Gass, Reference Mackey and Gass2015). In a recent introduction to a special issue of Language Learning, Rebuschat, Meurers, and McEnery (Reference Rebuschat, Meurers and McEnery2017) argue that the study of ‘a complex phenomenon like language acquisition can significantly benefit from insights, tools, and methods from many disciplines’ (p. 7).
1.1 Background
How do we know if a word is learned? We can ask the learner to indicate whether she knows the word and how well she knows it (Paribakht & Wesche, Reference Paribakht and Wesche1993), or test whether the learner is able to correctly identify its form or meaning in a multiple-choice task (Nation & Beglar, Reference Nation and Beglar2007; Laufer & Aviad–Levitzky, Reference Laufer and Aviad-Levitzky2017), or whether she is able to explain what the word means in her own words (Elgort & Warren, Reference Elgort and Warren2014; Elgort, Candry, Boutorwick, Eyckmans, & Brysbaert, Reference Elgort, Candry, Boutorwick, Eyckmans and Brysbaert2018). These approaches can provide an indication of whether some information about the target word has been encoded and stored in the memory. However, in completing these offline tests, learners are able to resort to a whole range of memory and task strategies. In real language use, such as reading a book or watching a movie, the use of such strategies would require a diversion of available cognitive resources from interpreting the meaning of the text to the processing of individual lexical items possibly causing confusion and misinterpretation (Grabe, Reference Grabe2009; Perfetti & Stafura, Reference Perfetti and Stafura2014; Castles, Rastle, & Nation, Reference Castles, Rastle and Nation2018).
In normal first language (L1) reading, for example, the time needed for the initial lexical access and word identification is somewhere between 140 and 250 milliseconds after the word onset, as shown in event-related brain potentials (ERP) and eye-movement word processing studies (Hauk, Davis, Ford, Pulvermüller, & Marslen-Wilson, Reference Hauk, Davis, Ford, Pulvermüller and Marslen-Wilson2006; Grainger & Holcomb, Reference Grainger and Holcomb2009; Rayner, Reference Rayner2009). Moreover, an initial fast parafoveal preview prior to the first fixation on the word already extracts orthographic information about that word (Veldre & Andrews, Reference Veldre and Andrews2016), and this preview affects the programming mechanism that determines where the fixation will fall, how long readers will fixate on the word or the likelihood of word skipping (Rayner, Reference Rayner1998; Drieghe, Rayner, & Pollatsek, Reference Drieghe, Rayner and Pollatsek2005; Reichle, Pollatsek, & Rayner, Reference Reichle, Pollatsek and Rayner2012). Thus, much of the early lexical processing happens extremely fast in a more or less ballistic, resource-light manner, of which offline tests of word knowledge cannot tell us much. Reading for meaning requires accurate word identification under time pressure that relies on precise orthographic representations of individual words, so that a lower frequency word, petty, is not confused with its higher frequency orthographic neighbour, pretty (Perfetti, Reference Perfetti2007). Therefore, research that investigates approaches to vocabulary teaching and learning needs to evaluate not only whether the learner can explicitly associate form and meaning but also whether she is able to access this knowledge online, accurately and fluently.
1.1.1 Prime lexicality effect
In an attempt to evaluate quality of lexical knowledge for recently learned words, Elgort (Reference Elgort2007, Reference Elgort2011) proposed using the form-priming paradigm developed by Forster and Vereŝ (Reference Forster and Vereŝ1998). I will now briefly outline the rationale behind this approach. Word recognition entails interactive activation and competition and selection processes that involve not only the word itself but other words related to it in form and meaning (McClelland & Rumelhart, Reference McClelland and Rumelhart1981; Forster & Davis, Reference Forster and Davis1984; Grainger & Jacobs, Reference Grainger and Jacobs1996; Coltheart, Rastle, Perry, Langdon, & Ziegler, Reference Coltheart, Rastle, Perry, Langdon and Ziegler2001). At the level of sub-lexical representations (e.g., individual letter representations), a letter-overlap facilitates processing. For example, the word, FUNCTION, should be processed faster if it is preceded by the letter sequence, bunction, that differs from it by one letter. However, at the level of lexical (word) representations, orthographic neighbours (words that differ by a single letter) make processing slower and more difficult due to lateral inhibition between word-level representations (Perry, Lupker, & Davis, Reference Perry, Lupker and Davis2008). So, the word, FUNCTION, will be recognised slower when it is preceded by an orthographic word neighbour, junction. The assumptions of the models of word recognition have been tested using the lexical decision task, in which participants decide whether a sequence of letters is a word. The lexical decision task is often conducted under time pressure, with instructions emphasising both speed and accuracy of responses. In order to investigate how the processing of one word (the target) is affected by certain features of another (the prime), primed lexical decision tasks may be used in which the target is preceded by either a related or an unrelated prime, and response time to the target in the related and unrelated conditions are compared. For example, when a word target is preceded by an orthographically related nonce word prime (e.g., bunction–FUNCTION), lexical decisions about the target (FUNCTION) are faster compared to when the target is preceded by an orthographically (and semantically) unrelated prime (e.g., bathroom–FUNCTION), due to prelexical, letter-level facilitation. If prime and target are both words (e.g., junction–FUNCTION), facilitation is no longer observed and, in some experiments, responses to the target in the related condition may be slower (and/or less accurate) compared to the unrelated condition (Davis & Lupker, Reference Davis and Lupker2006). This is known as the prime lexicality effect (PLE) (Forster & Vereŝ, Reference Forster and Vereŝ1998) because the processing of the target varies as a consequence of the lexical status of the prime (i.e., whether or not it is a word).
Elgort (Reference Elgort2007, Reference Elgort2011) proposed that the PLE could be used as a measure of word learning precisely because it hinges on the lexical status of the prime and thus can test whether a recently learned vocabulary item is processed as a word, without the need to use offline tests of explicit word knowledge. In other words, this approach provides an online measure for evaluating quality of lexical representations established as a result of some learning or instructional treatment.
1.1.2 PLE under masked conditions
The approach to evaluating effectiveness of learning interventions proposed in Elgort (Reference Elgort2011) has since been tested in L1 and L2 word learning studies conducted by Qiao and Forster (Qiao, Forster, & Witzel, Reference Qiao, Forster and Witzel2009; Qiao & Forster, Reference Qiao and Forster2013; Qiao & Forster, Reference Qiao and Forster2017). In L1, the PLE has been obtained under non-masked and masked priming conditions, with both adults (Davis & Lupker, Reference Davis and Lupker2006; Andrews & Hersch, Reference Andrews and Hersch2010; Andrews & Lo, Reference Andrews and Lo2012; Qiao & Forster, Reference Qiao and Forster2013; Nakayama, Sears, Hino, & Lupker, Reference Nakayama, Sears, Hino and Lupker2014) and children (Tamura, Castles, & Nation, Reference Tamura, Castles and Nation2017). However, in an L2 word learning study that used masked form-priming as a measure of lexicalisation, Qiao and Forster (Reference Qiao and Forster2017) failed to observe the PLE with Chinese-English bilinguals either for the recently learned L2 vocabulary or for known L2 words. This result is in line with another recent L2 word processing study with Japanese-English bilinguals by Nakayama and Lupker (Reference Nakayama and Lupker2018) who also failed to observe the PLE under masked priming conditions. After considering a number of alternative explanations, Qiao and Forster (Reference Qiao and Forster2017) conjectured that ‘the L2 lexicon is not represented in the same type of memory system as the L1 lexicon, rather it is represented in a system where competitive effects are not observed’ (p. 147). This conjecture has profound theoretical implications for L2 vocabulary acquisition, teaching and learning. For this reason, replications of Qiao and Forster (Reference Qiao and Forster2017) would be an important step in validating this finding and its interpretation.
To sum up, in the PLE, we may have a litmus test of lexicalisation that can be applied in vocabulary learning studies to evaluate quality of lexical representations established for recently learned words. Elgort (Reference Elgort2011) has shown that the PLE can be observed in the L2, under prime awareness conditions, and that form-priming can therefore be used as a measure of lexicalisation in L2 vocabulary learning and processing studies. However, so far, the PLE has not been observed with L2 participants under masked priming conditions.
2. Replication study 1: Elgort (Reference Elgort2011, Experiment 1)
2.1 The original study
Elgort's (Reference Elgort2011) study investigated whether deliberate word learning using a paired-associate learning approach can result in the kind of knowledge that learners need in real language use (such as reading). Although the efficiency of paired-associate learning is not normally contested, the effectiveness of this treatment is less widely accepted. Theories of SLA that prioritise the role of comprehensible input, such as the Monitor theory (Krashen, Reference Krashen1982, Reference Krashen1985; Krashen & Terrell, Reference Krashen and Terrell1983) argue that deliberate learning results in the kind of knowledge that is only useful in monitoring performance, under conditions that afford enough time to do so, but does not result in the kind of knowledge that is needed for online language use. The Acquisition-Learning Hypothesis of the Monitor theory posits a dissociation between the knowledge that is gained implicitly through language use (such as reading or listening) and the knowledge that is gained explicitly through deliberate study, i.e., deliberate learning ‘does not become acquisition’ (Krashen, Reference Krashen1982, p. 83). Elgort's (Reference Elgort2011) motivations for testing the assumptions of the Acquisition-Learning Hypothesis for L2 vocabulary learning were as follows: (1) the Acquisition-Learning Hypothesis has been generalised to the learning of vocabulary, beyond the original context of learning grammar (Krashen & Terrell, Reference Krashen and Terrell1983; Krashen, Reference Krashen1989; Dupuy & Krashen, Reference Dupuy and Krashen1993); (2) the hypothesised dissociation between deliberate learning and acquisition has significant implications for L2 learning and teaching practice and research, and (3) the view that deliberate word learning has limited (or no) value promotes negative attitudes among language teachers towards this method of vocabulary learning.
To test the assumptions of the Acquisition-Learning Hypothesis, Elgort (Reference Elgort2011) used an unambiguously decontextualized deliberate learning treatment using vocabulary flashcards. Participants – 48 university students of advanced L2 proficiency from a range of L1 backgrounds – learned 48 vocabulary items (English) using paper-based within-L2 flashcards for one week (five–six learning sessions), following a recommended repetition and retrieval training regime. Each card had the target pseudoword printed on one side and its L2 definition and an example of use on the other. The pseudowords were created by changing a single letter in real English words.
The main research question of the study investigated whether deliberate paired-associate learning would trigger vocabulary acquisition, conceptualised as lexicalisation, i.e., the establishment of lexical representations of the recently learned L2 vocabulary items. Lexicalisation is characterised by precise formal-lexical representations (knowledge of form) and integrated lexical semantic representations (knowledge of meaning) that become available in an online manner, under the conditions that limit opportunities to deploy conscious strategies. In Experiment 1, lexicalisation was operationalised as the PLE, i.e., lexicalised recently learned items should not facilitate the processing of orthographically related word targets in a form-priming lexical decision task due to lateral inhibition between orthographic-neighbour words. Conversely, if flashcard learning does not result in lexicalisation, the studied items will not be processed as words and will speed up the processing of related word targets, as a result of the letter-level facilitation in the absence of lexical-level competition. Other measures of acquisition in Elgort's study (not considered in the present article) included masked repetition priming, semantic priming, and a measure of fluency of access to lexical representations – a coefficient of variation (Segalowitz & Segalowitz, Reference Segalowitz and Segalowitz1993).
In addition, because no previous research had tested whether the PLE can be observed for known L2 words (in other words, whether lexical representations of known L2 words are accessed and processed in the same way as those of known L1 words), Elgort (Reference Elgort2011) included an additional set of items in the form-priming experiment; in this set, known L2 words were used as related form-primes. This is because the PLE could not have been assumed in the L2 since, even in the first language, the PLE is more likely to be observed for good (but not poor) spellers (Andrews & Hersch, Reference Andrews and Hersch2010; Andrews & Lo, Reference Andrews and Lo2012).
2.1.1 The form-priming experiment: design and results
In the form-priming experiment, participants made lexical decisions on known L2 word targets preceded by related or unrelated form-primes. Both primes and targets were presented for 522 milliseconds (i.e., about half a second). Two sets of stimuli were used in this experiment: the original set of primes and targets from Forster and Vereŝ (Reference Forster and Vereŝ1998, Experiment 1), in which related primes were real English words, and a new set of primes and targets, in which the deliberately learned pseudowords were used as related primes.
The results showed the presence of the PLE in both stimuli sets: significant priming was observed when target words were preceded by orthographically related nonwords (letter sequences created using the same approach as the pseudowords but not included in the learning treatment), but not when they were preceded by either real word primes or by the deliberately learned pseudoword primes. This result provided the first confirmation that the PLE could be observed in the L2 (under prime awareness conditions) and, crucially, that deliberately learned L2 words could be lexicalised (acquired) and processed in a manner similar to known L2 words.
2.2 Approaches to replication
2.2.1 An approximate replication
There are a number of reasons why this study warrants replication. Although priming had been used in bilingual studies since the 1980s, Elgort (Reference Elgort2011) was the first study to use form-priming – an experimental paradigm from cognitive psychology – to evaluate theoretical SLA claims about the relationship between learning and acquisition of L2 vocabulary. Because this study continues to be cited by vocabulary researchers, presumably because its findings are seen as relevant in the field, approximate replications are needed to test whether the findings reported in the original study are reliable (Marsden, Morgan-Short, Thompson, & Abugaber, Reference Marsden, Morgan-Short, Thompson and Abugaber2018) and can be obtained outside of the original context (Porte, Reference Porte2012).Footnote 1 Cumming (Reference Cumming2014) argues that ‘[a] study that keeps some features of the original and varies others can give a converging perspective, ideally both increasing confidence in the original finding and starting to explore variables that influence it’ (p. 10).
It is worth pointing out that an internal follow-up study by Elgort and Piasecki (Reference Elgort and Piasecki2014) extended the findings of Elgort (Reference Elgort2011); they found that using L1 (rather than L2) definitions in the flashcard treatment with a group of L1 German speakers had also resulted in the lexicalisation of the studied L2 vocabulary items, as indicated by the presence of the PLE. Elgort and Piasecki (Reference Elgort and Piasecki2014) made a number of other changes to the original study design: they used 32 original 7- and 8-letter items and 16 new 6-letter items (that replaced the original 9-letter ones);Footnote 2 they also used a group of L2 university students whose L1 was German (whereas the participants of the original study were from a range of L1 backgrounds). Importantly, Elgort and Piasecki (Reference Elgort and Piasecki2014) observed the PLE for known L2 words with a new group of L2 participants using the stimuli from the original study, thus increasing the confirmatory power of the finding that known L2 words are processed visually in a way similar to known L1 words, under experimental conditions that do not restrict prime visibility. This finding suggests that, for speakers of alphabetic languages (such as German), native-like quality of orthographic representations in L2 English is achievable.
The next logical step in verifying the main finding of the original study would be an approximate replication of Experiment 1 (Elgort, Reference Elgort2011) with learners whose L1 is not alphabetic, such as Chinese. Because the Chinese writing system is orthographically distant from English, it may not be possible to observe the PLE with Chinese speakers of English, even with known L2 words. This is because visual word recognition and processing in an orthographically distant L2 has a host of challenges (Koda, Reference Koda, Coady and Huckin1997; Muljani, Koda, & Moates, Reference Muljani, Koda and Moates1998; Hamada & Koda, Reference Hamada and Koda2008; Perfetti & Dunlap, Reference Perfetti, Dunlap, Koda and Zehler2008). Developmental differences in young readers from distinct writing systems (English and Chinese) suggest that the brain adapts to ‘the special features of the orthography’ (Cao, Brennan, & Booth, Reference Cao, Brennan and Booth2014). English readers, for example, develop brain regions associated with phonological processing and fine-grained form recognition, while Chinese readers develop regions associated with holistic visual orthographic processing (Cao et al., Reference Cao, Brennan and Booth2014). If orthographic representations of English words are less specified, less precise for Chinese speakers compared to speakers of alphabetic L1s, such as German, it may not be possible to observe the PLE in English with Chinese speakers and other measures of lexicalisation would need to be found for this learner population.
Alternatively, if the quality of L2 lexical representations is affected primarily by learning and instructional methods (and to a lesser degree by the learner's L1), differences in the visual processing of known words may be due to how these words had been learned. For example, the PLE may be observed for known L2 words with Chinese-English bilinguals who learned English in an English as a second language (ESL) but not English as a foreign language (EFL) context. In this case, with the right learning treatment, it should be possible to replicate the PLE for the recently learned words even if the PLE is not observed for ‘known’ L2 words. Such an outcome with Chinese speakers would provide even stronger support to the finding of the original study that deliberate L2 word learning involving repeated form and meaning retrieval (such as flashcards) can result in the acquisition of high-quality lexical knowledge. Researchers conducting replications of Elgort's original study need to keep in mind that, in order to test the Acquisition-Learning Hypothesis, a certain threshold of explicit word knowledge must be ensured for the deliberately learned items before testing their lexicalisation.
2.2.2 A conceptual replication
The robustness of the main finding of Elgort's study – that deliberate word learning can lead to acquisition (operationalised as lexicalisation) – can be tested by conducting a conceptual replication, in which lexicalisation of deliberately learned vocabulary is tested using a different method that, perhaps, affords higher ecological validity, while keeping the learning treatment the same. Because researchers have recently questioned the degree to which response times in lexical decisions represent the actual time-course of lexical processing (Kuperman, Drieghe, Keuleers, & Brysbaert, Reference Kuperman, Drieghe, Keuleers and Brysbaert2013; Dirix, Brysbaert, & Duyck, Reference Dirix, Brysbaert and Duyck2018), using an alternative method to evaluate lexicalisation would test whether or not the original findings might have been an artefact of the form-priming methodology (Porte, Reference Porte2012; Porte & McManus, Reference Porte and McManus2019; personal communications). For example, lexicalisation could be investigated by observing and analysing how vocabulary items are read in meaningful, informative contexts, since known words are recognised and processed faster and with less effort in such contexts (Stanovitch & West, Reference Stanovitch and West1983; Tyler & Wessels, Reference Tyler and Wessels1983).
This conceptual replication could be done by recording participants’ eye-movements while reading deliberately learned vocabulary items embedded in sentence or paragraph contexts. This eye-tracking technique offers a relatively natural non-invasive way of studying word processing in real time, without requiring participants to perform any additional tasks beyond reading. The quality of formal-lexical representations of recently learned L2 words can be examined using first-fixation durations (i.e., duration of the first fixation on a word that is indicative of the initial lexical access and word identification) (Rayner, Reference Rayner1998; Pellicer-Sánchez, Reference Pellicer-Sánchez2015; Dirix et al., Reference Dirix, Brysbaert and Duyck2018; Elgort, Brysbaert, Stevens, & Van Assche, Reference Elgort, Brysbaert, Stevens and Van Assche2018). Researchers may also consider using other eye-movement measures to evaluate lexicalisation, such as number of re-fixations and regressions back to the recently learned items; however, in order to test the original claims of the Acquisition-Learning Hypothesis, the chosen eye-movement measures need to reflect visual word processing that does not involve conscious monitoring. Finally, because of high individual differences in eye-movements between readers, the eye-movements on the deliberately learned pseudowords would need to be compared with those on known L2 words (of high- and low-frequency) and non-studied pseudowords. Similar to the PLE measure used in the original study, these eye-movement measures would evaluate learners’ processing of deliberately learned vocabulary items online, under conditions that do not restrict their visibility.
3. Replication study 2: Qiao and Forster (Reference Qiao and Forster2017)
3.1 Background to the study
In a deliberate word learning study with L1 (English) participants, Qiao and Forster (Reference Qiao and Forster2013, Experiment 2) found that deliberate vocabulary learning resulted in lexicalisation, operationalised as the presence of the PLE under masked priming conditions. Qiao and Forster (Reference Qiao and Forster2017) aimed to replicate this finding with L2 participants. Forty-four undergraduate EFL students from a university in China participated in a deliberate word learning study followed by the form-priming experiment. They were native Chinese speakers in their second year of a Business English major who had been studying English for more than ten years. The learning items were 24 pseudowords created by changing one letter in an English word (e.g., baltery – battery), 6–8 letters long. The items were assigned a meaning using a one-sentence definition and a corresponding picture. The participants completed four lab-based learning sessions over two weeks. Each pseudoword was first presented to the participants on its own and then followed by the corresponding picture and definition. Then, the participants completed a multiple-choice picture-word and definition-word matching task. At the end of each learning session, participants completed a primed lexical decision task, in which the pseudowords from the learning session were presented as targets (e.g., battery-BALTERY). English word pairs (e.g., protect-PROJECT) not used in the learning sessions were also included in the lexical decision task, in order to establish whether the PLE could be observed with known L2 words, under masked priming conditions.
After the four learning sessions, participants completed the final lexical decision task under masked priming conditions: the primes were presented in the lower case for 50 milliseconds; they were preceded by a mask presented for 500 milliseconds and followed by the target in the upper case presented for 500 milliseconds. The recently learned pseudowords were used as primes in this experiment and the words they were derived from were used as targets (e.g., baltery – BATTERY). In addition, 24 nonwords (letter sequences created using the same approach as the pseudowords but not included in the learning phrase) were used as related nonword primes (e.g., dresent – PRESENT). Unrelated prime-target pairs were used in the control condition.
In this study, the authors did not observe the PLE with the deliberately learned items: the word targets preceded by the orthographically related studied pseudowords were processed significantly faster (rather than slower) than those in the control condition. Moreover, the PLE was not observed for the L2 words that were considered known. Because the results obtained by Qiao and Forster (Reference Qiao and Forster2013, Experiment 2) with the L1 participants were not replicated with the L2 (Chinese) participants, the authors concluded that the PLE cannot be achieved in the L2. In their discussion, Qiao and Forster (Reference Qiao and Forster2017) reiterated the Episodic L2 Hypothesis articulated previously by Witzel and Forster (Reference Witzel and Forster2012) and Jiang and Forster (Reference Jiang and Forster2001) which conjectures that L2 words may be stored differently than L1 words: in non-lexical, episodic-like memory.
3.2 Approaches to replication
3.2.1. Close replications
Because Qiao and Forster's (Reference Qiao and Forster2017) study was conducted with EFL Chinese speakers, a logical next step would be to test their findings with ESL Chinese university students who have lived and studied in an English-speaking country for at least 2–3 years and are likely to have higher L2 proficiency. Keeping the L1 of the learners the same as in the original study, this replication would verify whether the PLE may be observed with participants whose L2 proficiency is higher and who have been exposed to native L2 input. Proficiency has been shown to affect how known and recently learned L2 words are processed. A word learning study by Elgort, Perfetti, Rickles, and Stafura (Reference Elgort, Perfetti, Rickles and Stafura2015), for example, showed that lexical semantic processing of recently learned words by high-proficiency L2 students was more like L1 word processing, but this was not the case for lower-proficiency learners (see also Phillips, Segalowitz, O'Brien, & Yamasaki, Reference Phillips, Segalowitz, O'Brien and Yamasaki2004). Differences in L2 word processing between higher- and lower-proficiency L2 participants have also been reported in ERP (Midgley, Holcomb, & Grainger, Reference Midgley, Holcomb and Grainger2009; Newman, Tremblay, Nichols, Neville, & Ullman, Reference Newman, Tremblay, Nichols, Neville and Ullman2012) and functional magnetic resonance imaging (fMRI) (Abutalebi, Reference Abutalebi2008) studies. Therefore, it is possible that Chinese speakers of English whose L2 proficiency is higher may achieve native-like lexical quality needed to observe the PLE. If this conjecture does not bear out, we can reject the possibility that L2 proficiency is at the core of the failure to obtain the PLE in masked priming with L2 participants.
Another close replication of Qiao and Forster's (Reference Qiao and Forster2017) could be conducted with EFL speakers of an alphabetic language. As detailed in Section 2.2.1, the L1 of the learners (particularly, its orthographic distance from the L2) may affect how known L2 words are processed. It is possible that the absence of the PLE in Qiao and Forster's study is due to a different mechanism of L2-English word processing by Chinese speakers, and not due to how L2 word representations are stored, more generally. This conjecture is in line with findings of an fMRI study into lexical processing of late Chinese-English bilinguals (Cao, Tao, Liu, Perfetti, & Booth, Reference Cao, Tao, Liu, Perfetti and Booth2013). Cao et al. (Reference Cao, Tao, Liu, Perfetti and Booth2013) reported that that neural network of Chinese-English bilinguals partly assimilates English into the Chinese system, especially in the visual stages of word identification, and that the involvement of the native (Chinese) network in L2 (English) processing becomes greater with higher proficiency in English. On the other hand, for EFL speakers whose L1 is alphabetic and who are likely to process L2 English words in the same manner as native English speakers, the PLE may be observed more readily with both known and recently learned vocabulary. This conjecture can be tested using a close replication of Qiao and Forster (Reference Qiao and Forster2017) with L2 speakers of an alphabetic language that encourages a high level of fine-tuning of orthographic representations from an early age, such as Dutch (Marinus, Nation, & de Jong, Reference Marinus, Nation and de Jong2015). An absence of the masked PLE in such a replication would generalise the findings of the original study to learners whose L1 is alphabetic, supporting Qiao and Forster's conjecture ‘that the L2 lexicon … is represented in a system where competitive effects are not observed’ (p. 147).
3.2.2 Conceptual replications
A conceptual replication of Qiao and Forster (Reference Qiao and Forster2017) could further test their conjecture that L2 words are not lexicalised and/or not stored in the same manner as L1 words. Qiao and Forster made an assumption that the pseudowords were learned at the end of their treatment because accuracy and response speed on the multiple-choice matching task had improved. However, they did not test participants’ knowledge of the word form (spelling) that is critical in obtaining the PLE (Andrews & Hersch, Reference Andrews and Hersch2010; Andrews & Lo, Reference Andrews and Lo2012; Tamura et al., Reference Tamura, Castles and Nation2017). A key design change in the proposed conceptual replication of Qiao and Forster (Reference Qiao and Forster2017) would therefore be an addition of a dictation posttest (such as that used in Elgort, Reference Elgort2011), in which participants retrieve the written form of the learned pseudoword cued by its definition (or the picture) used in the learning treatment. The results of this posttest of explicit word knowledge will need to be accounted for in the form-priming analysis that measures lexicalisation. A number of approaches could be used: (a) the learning treatment could continue until individual participants are able to produce at least two thirds of the pseudowords accurately; (b) the form-priming data of the participants unable to reach this learning target could be excluded; (c) the pseudowords that had not been retrieved accurately (e.g., below an 80% accuracy threshold) could be excluded from the data analysis; and/or (d) the pseudowords that had not been correctly retrieved in the posttest by a participant could be excluded from the analysis for that participant. This design change would ensure that the presence of the PLE is tested only for the vocabulary items that have been successfully learned (i.e., attained a certain threshold of explicit knowledge). If the original result is replicated and the masked PLE is still not observed after the revised design is implemented, the original conclusion that deliberately learned L2 words are not lexicalised for Chinese EFL speakers in the same way as for native English speakers would be verified.
Similar to the proposed conceptual replication of Elgort (Reference Elgort2011), the findings and theoretical conjecture articulated in Qiao and Forster (Reference Qiao and Forster2017) can be further probed by using an alternative (more ecologically valid) method of recording the learners’ eye-movements while reading. The rationale for this conceptual replication is the same as that articulated for the Elgort's study but the measure of lexicalisation needs to be matched with masked form-priming used by Qiao and Forster that severely restricts prime visibility and awareness. This can be achieving by using the gaze-contingent boundary paradigm (Rayner, Reference Rayner1975), in which readers’ parafoveal preview of the upcoming target vocabulary item is manipulated (the target item is briefly replaced by another) while their eyes are still fixated on the preceding word. When the eyes cross an invisible boundary just before the target item, the correct target item is restored. Readers are typically unaware of this manipulation as the stimulus change occurs during the eye-movement when vision is supressed (the saccade). When the preview of an upcoming word is invalid (e.g., it is replaced with an orthographically illegal nonword), the reader will fixate longer on that word. This is known as the preview benefit effect (Rayner, Reference Rayner1998, Reference Rayner2009; Schotter, Angele, & Rayner, Reference Schotter, Angele and Rayner2012) because availability of a correct preview of a word speeds up its processing. Veldre and Andrews (Reference Veldre and Andrews2015, Reference Veldre and Andrews2016) showed that readers with higher quality of lexical representations experienced a greater preview benefit effect and were more likely to extract lexical information from a parafoveal word. This suggests that the preview benefit effect may be suitable for assessing lexicalisation. This conceptual replication could be conducted with L2 English speakers whose L1 is either Chinese or an alphabetic language, following the logic of the close replications detailed in 3.2.1.
4. Conclusions
In this article, I have proposed a number of replications of two studies – Elgort (Reference Elgort2011) and Qiao and Forster (Reference Qiao and Forster2017) – that touch upon two theoretical frameworks relevant to the SLA of vocabulary: the Acquisition-Learning Hypothesis and the Interactive Activation and Competition Framework of L2 word recognition and processing. The close and approximate replications aim to test the influences of learners’ L1 (logographic vs alphabetic) on their ability to lexicalise deliberately learned and known L2 vocabulary. The conceptual replications aim to verify the two aforementioned theoretical hypotheses addressed in the original studies by selecting a more ecologically valid measure of lexicalisation – eye-tracking during sentence (or paragraph) reading, while keeping the deliberate learning procedure the same as in the original studies. The findings of these replications would extend our understanding of how L2 words are stored, processed and learned, and will have important implications for L2 teaching and learning.
Irina Elgort is Senior Lecturer in Higher Education at Victoria University of Wellington. Irina's research interests include approaches and factors that affect L2 word learning and measures of word knowledge and processing. Her work has been published in Applied Linguistics; Bilingualism: Language and Cognition; Language Learning; Studies in Second Language Acquisition; Language, Cognition and Neuroscience, and other journals.
