The Form Preparation Paradigm

A form preparation task, also called the implicit priming paradigm, has been frequently used to investigate the phonological preparation unit in spoken word production in adults’ native language (J.-Y. Chen et al., Reference Chen, Chen and Dell2002; T.-M. Chen & Chen, Reference Chen and Chen2013; Cholin, Schiller & Levelt, Reference Cholin, Schiller and Levelt2004; Kureta et al., Reference Kureta, Fushimi and Tatsumi2006; Meyer, Reference Meyer1990, Reference Meyer1991; O’Seaghdha et al., Reference O’Seaghdha, Chen and Chen2010). The task involves an associate learning session and a naming session. In the associate learning session, participants memorize some prompt-response word pairs (e.g., night-day, wet-dew, and bread-dough). After the participants indicate that they have memorized all the pairs, a naming session is conducted immediately. In the naming session, lists of prompts are presented. For each list, a set of prompt words appears in a random order and the participants are required to say the corresponding response word as quickly and accurately as possible (e.g., when the word night is presented, participants need to say the word day). The critical manipulation is the context of the list: in the heterogeneous context, the response words do not share any phonological units systematically in common (e.g., three response words are day, sea, pie); in the homogeneous context, the response words share the same ingredient (e.g., the initial phoneme is always /d/ for day, dew and dough). If the shared ingredient is the phonological preparation unit in the target language, the foreknowledge of the ingredient allows the participants to prepare their production in advance, thus facilitating their naming latency in the homogeneous list.

Research using the form preparation task suggests that phonological preparation units differ across languages (Chen et al., Reference Chen, Chen and Dell2002; Kureta et al., Reference Kureta, Fushimi and Tatsumi2006; Meyer, Reference Meyer1990, Reference Meyer1991; O’Seaghdha et al., Reference O’Seaghdha, Chen and Chen2010). Native speakers of alphabetic languages such as Dutch (Meyer, Reference Meyer1990, Reference Meyer1991) and English (O’Seaghdha et al., Reference O’Seaghdha, Chen and Chen2010) select phonemic segment as the preparation unit in spoken word production. Native Mandarin speakers benefit from the same initial syllable but not from the same initial phoneme (Chen et al., Reference Chen, Chen and Dell2002; O’Seaghdha et al., Reference O’Seaghdha, Chen and Chen2010), suggesting that the primary preparation unit is the syllable instead of the phoneme in Mandarin. Native Japanese speakers benefit from the same initial CV (consonant-vowel) mora but not the same phonemic segment (Kureta et al., Reference Kureta, Fushimi and Tatsumi2006), suggesting that the CV mora may be the preparation unit in Japanese.

A limitation of the traditional associative word naming task is that the orthographic information is explicitly represented in the learning session and naming session. Li, Wang, and Idsardi (Reference Li, Wang and Idsardi2015) showed that orthographic cues might encourage speakers to choose the preparation unit that is more compatible with the orthography used in the task. For native Mandarin speakers, the onset was selected as the preparation unit when the materials were presented in Pinyin, a phonetic system that uses the Latin alphabet to transcribe the pronunciation of Chinese characters. However, when the stimuli were presented in Chinese characters, a morphosyllabic writing system, the participants did not show the same pattern of onset facilitation. Kureta, Fushimi, Sakuma, and Tatsumi (Reference Kureta, Fushimi, Sakuma and Tatsumi2015) showed a similar pattern with Japanese speakers who demonstrated significant phoneme facilitation when participants learned materials in romaji (i.e., a phonetic system to write Japanese using the Latin alphabet). However, this phoneme preparation effect was not shown when materials were presented in an auditory task. The above results suggested that presenting visual materials in an alphabetic writing system may encourage speakers to prepare spoken production in smaller units (i.e., the phoneme). Alario, Perre, Castel, and Ziegler (Reference Alario, Perre, Castel and Ziegler2007) suggested that the associative learning session may encourage speakers to use orthographic codes to facilitate memorization and hence speed up the following naming session. As a result, similar to Alario et al. (Reference Alario, Perre, Castel and Ziegler2007), we decided to adopt a simple picture naming task with the form preparation paradigm in order to avoid the influence of orthographic cues. Chen and Chen (Reference Chen and Chen2013) adopted a simple picture naming task with the form preparation paradigm to investigate the role of onset and syllable segment in phonological preparation in Mandarin Chinese. Participants were required to name a list of pictures whose names may share the same onset, the same initial syllable segment or nothing systematically in common. A significant facilitation was only shown in the same syllable segment condition, suggesting that syllable segment instead of onset is the phonological preparation unit in speaking Mandarin Chinese even when no orthographic cues were provided.

The role of rhyme in phonological preparation

Few studies have employed the form preparation paradigm to examine the role of later components of a word (e.g., rhyme or the second syllable of a disyllabic word) in preparing spoken word production. Meyer (Reference Meyer1991) showed that fore-knowledge of the later components of a word did not benefit native Dutch speakers in a form preparation task, presumably because the processing of phonological units is sequential. However, Li et al. (Reference Li, Wang and Idsardi2015) found that native Mandarin speakers showed significant rhyme interference effect with Pinyin cues in an associative naming task with form preparation paradigm. A potential explanation of this interference effect is that lexical competition occurred between the response item and its rhyme neighbors. The contrast between these two studies may be related to the different linguistic features between Dutch and Mandarin Chinese, the simple syllable structure of Chinese may make it easier to activate the lexical information of rhyme neighbors. Therefore, the second goal of the present study was to examine the role of the later component (i.e., rhyme) in phonological preparation of spoken English word production.

It is likely that rhyme would not lead to any phonological preparation effect in English, just as with what has been shown in Dutch. However, it is also possible that a rhyme interference effect would be shown as a result of competition due to different onsets. According to the parallel-then-sequential model proposed by O’Seaghdha and colleagues (see O’Seaghdha & Marin, Reference O’Seaghdha and Marin2000 for a detailed review), all the segments of the target words are activated in a parallel fashion, but are assembled sequentially. As a result, when a list of spoken words share the same rhyme, the shared rhyme may reactivate the previous trial when participants need to produce the target of the current trial, so that the different onsets compete with each other when speakers assign or assemble the phonological units sequentially. Using a continuous repetition procedure without orthographic input, O’Seaghdha and Marin (Reference O’Seaghdha and Marin2000, Exp. 4) revealed that repeatedly producing word pairs that shared the same rhyme (e.g., glory-story-glory-story-. . .) led to an interference effect and the effect was modulated by frequency (i.e., larger interference for high frequency words). Since all the segments were retrieved in a parallel fashion, the rhyme information sent feedback and activated other rhyme neighbors (e.g., glory) when the assignment of segments of the current trial (e.g., story) was initiated. The different onsets (i.e., gl- and st-) then led to competition and thus the interference effect. Although Sevald and Dell (Reference Sevald and Dell1994) showed a facilitative effect when participants kept producing end-related words using the same continuous repetition procedure, O’Seaghdha and Marin (Reference O’Seaghdha and Marin2000) explained that the difference might be due to the fact that Sevald and Dell used less naturalistic materials and their participants were more extensively practiced, so that segmental retrieval and assignment required less time than normal activation of the phonological units.

Repeating rhyme neighbors in an English reading task may also lead to an interference effect due to lexical competition as proposed by Lukatela and Turvey (Reference Lukatela and Turvey1996). In that study, native English speakers named a target word (e.g., nose) slower when its prime shared the target's rhyme (e.g., hose) compared to the neutral condition when the prime did not share any phonological information with the target (e.g., dish). Lukatela and Turvey explained this effect as a result of the prime inhibiting the pre-activation of the target at the lexical level. Li et al. (Reference Li, Wang and Idsardi2015) also used the lexical competition hypothesis to explain their rhyme inhibition effect in native Chinese speakers using the associative word naming task. In their study, Pinyin served as orthographic cues to encourage speakers to attend to the sub-syllabic units, including the rhyme unit, given that it is explicitly represented. Producing a Pinyin trial pre-inhibited its rhyme neighbors, including the next trial, thus interference occurred.

In summary, limited research has been conducted to investigate the role of rhyme in phonological preparation in English. Although different results have been shown in the continuous repetition task, there was a consensus in the previous literature that language production entails a sequential segmental assignment process (Meyer, Reference Meyer1991; O’Seaghdha & Marin, Reference O’Seaghdha and Marin2000; Sevald & Dell, Reference Sevald and Dell1994). If parallel activation occurs prior to the sequential segmental assignment, a rhyme interference effect may occur. If the phonological activation is strictly sequential, then it is more likely that shared rhyme would not show any form preparation effect in English, just like the case of Dutch. In addition, lexical competition between rhyme neighbors may also contribute to the rhyme interference effect.

The influence of L1 on L2 phonological processing

Very limited research has investigated the phonological preparation unit in the L2 when speakers’ L1 and L2 have different phonological preparation units. As far as we know, the only study that has investigated this issue was Verdonschot, Nakayama, Zhang, Tamaoka, and Schiller (Reference Verdonschot, Nakayama, Zhang, Tamaoka and Schiller2013). In a masked primed visual word naming task, Chinese ESLs (native Chinese speakers who speak English as a Second Language) with high English proficiency showed significant onset facilitation in English. Namely, when the visual prime (e.g., bark) shared the same phoneme onset with the visual target (e.g., BENCH), Chinese ESLs showed a significantly faster response when naming the visual word BENCH compared with a control condition (e.g., when the prime is dark, a word which does not share any phonological unit with the target word). These results suggest that, like native English speakers, highly proficient Chinese ESLs assemble phonological information phoneme by phoneme. However, a limitation of this study is that, in a visual word-naming task, orthographic processing is involved in phonological retrieval and encoding. It remains unclear what the phonological preparation unit would be in the L2 when orthographic information is not present, such as in a conceptually driven production task (e.g., picture naming), and the present study aims to fill this gap.

Although limited research has investigated the influence of the L1 on the phonological preparation unit in the L2, previous literature has shown rich evidence on cross-language transfer of phonological awareness (Chen, Xu, Nguyen, Hong & Wang, Reference Chen, Xu, Nguyen, Hong and Wang2010; Cisero & Royer, Reference Cisero and Royer1995; Dickinson, McCabe Clark-Chiarelli & Wolf, Reference Dickinson, McCabe, Clark-Chiarelli and Wolf2004; Durgunoğlu, Nagy & Hancin-Bhatt, Reference Durgunoğlu, Nagy and Hancin-Bhatt1993). Phonological awareness entails the ability to detect and manipulate phonological units within a word, such as phonemes (e.g., /s/ in the word skip /skɪp/), onsets (e.g., /sk/ in skip), rhymes (e.g., /ɪp/ in skip), or whole syllables (e.g., /skɪp/) (Gillon, 2004). Turning back to the literature about cross-language transfer of phonological awareness, it has been shown that first-grade Spanish-speaking children who showed better phonological awareness in Spanish also performed better in English tasks that tapped into phonological awareness (Durgunoğlu et al., Reference Durgunoğlu, Nagy and Hancin-Bhatt1993). Furthermore, Wang, Perfetti, and Liu (Reference Wang, Perfetti and Liu2005) showed that cross-language transfer of phonological awareness occurs even between languages that have different writing systems. For 8-year-old Chinese–English bilingual children, their onset and rhyme awareness in Chinese were both correlated with their English onset and rhyme skills. In summary, previous literature suggests that the sensitivity and ability to manipulate certain phonological units (e.g., onset, rhyme and phoneme) in the L1 could be transferred to the L2. The present study investigates whether the transfer of phonological awareness also applies to phonological preparation in spoken word production. If the native Chinese speakers transfer their phonological preparation unit in Chinese to their spoken word production in English, they may select syllable as their phonological preparation unit in English. Conversely, if the native Chinese speakers do not transfer their phonological preparation unit in L1 to L2, they may perform like native English speakers and use phoneme as the preparation unit.

A third possibility is that Chinese ESLs may initially use syllable as the phonological preparation unit but change it to phoneme after some practice. Jared and Kroll (Reference Jared and Kroll2001) suggest that recent language experience influences bilingual speakers’ performance on spoken word production, although their work was not focused on the phonological preparation unit. In their study, English–French and French–English bilinguals were asked to name written English words with or without French enemies (i.e., an orthographically-driven naming task). A French enemy of an English word is a French word (e.g., FAIT) that shares the same word-body with the English words (e.g., BAIT) but is pronounced differently (e.g., -AIT is pronounced with a short /e/ in French). Participants’ naming latencies were influenced by whether they had read French words recently: the French–English bilinguals who were more dominant in French showed a significantly stronger inhibition in naming English words with French enemies after a French reading session compared to their performance prior to the reading session. In the present study, we employed a block design in the form preparation task and the participants were asked to repeat the response items. Based on Jared and Kroll's work, it is possible that Chinese ESLs would shift their mental orientation to phoneme in the later blocks after some recent practice with English L2.

The Present Study

The present study consists of two picture naming experiments. In the first experiment, a group of native Chinese speakers were asked to complete a simple picture naming task in Chinese. The goals of Experiment 1 were 1) to replicate the results of Chen and Chen (Reference Chen and Chen2013), which showed that the phoneme onset was not the preparation unit in Chinese, 2) to investigate the role of rhyme in speaking Chinese when no orthographic cues are provided, and 3) to compare with Experiment 2 which was conducted in English on native English speakers to address the cross-linguistic differences. We predicted that no onset effect would be presented in Experiment 1, as Chen and Chen (Reference Chen and Chen2013) suggested. With regards to the rhyme effect, one prediction is that no rhyme effect would be presented, since previous literature has suggested that later components may not contribute to phonological preparation (Meyer, Reference Meyer1991). Another possibility is that a rhyme interference effect could be shown, considering that native Chinese speakers have shown rhyme inhibition when the items were presented in Pinyin in Li et al. (Reference Li, Wang and Idsardi2015). In the picture naming task, although there were no Pinyin cues available to encourage participants to attend to the sub-syllabic units, there were no character cues to encourage them to attend to the syllabic unit either. Therefore, there is still a possibility that repetition of rhyming words in the present study may lead to an interference effect, as a result of lexical competition between the response item and its rhyme neighbors as suggested by Li et al. (Reference Li, Wang and Idsardi2015).

In Experiment 2, an English simple picture naming task using the form preparation paradigm was administered among Chinese ESLs and native English speakers. As mentioned earlier, there are advantages to conducting a simple picture naming task instead of the traditional associative naming task: first, the absence of explicit orthographic information can help to avoid the influence of orthographic cues; second, the absence of associative learning session can help reduce the possibility that participants adopt orthographic strategies to facilitate memorization. The critical conditions in both Experiments 1 and 2 were whether the target names (i.e., the name of pictures) shared the same onset, same rhyme, or nothing systematically in common.

The purpose of the Experiment 2 was to address two questions. Firstly, we wanted to compare Chinese ESL speakers and native English speakers on whether they select onset phoneme as the preparation unit. Secondly, we wanted to investigate the role of the rhyme in preparing language production in English among the Chinese ESL and native English speakers. Given that phoneme onset has been shown not to be the preparation unit in Chinese in previous research (e.g., T.-M. Chen & Chen, Reference Chen and Chen2013), and if the preparation unit in the L1 affects the selection of preparation unit in the L2, then the Chinese ESLs would not benefit from the fore-knowledge of the onset in the English picture naming task, in contrast to native English speakers, where onset facilitation is expected. If the preparation unit only depends on the linguistic features of a target language but not on the speaker's L1 experience, both groups would show significant onset facilitation. If the preparation unit could be influenced by recent language experience, the Chinese ESLs may only show onset facilitation in later blocks whereas native English speakers would show it in all blocks.

Concerning our second goal about the rhyme effect, if phonological preparation in English is similar to preparation in Dutch, native English speakers would not show any significant rhyme effect, in accordance with Meyer (Reference Meyer1991). An alternative hypothesis is that an interference effect would be shown due to onset competition (see O’Seaghdha & Marin, Reference O’Seaghdha and Marin2000), perhaps with additional contribution from lexical competition. For Chinese ESLs, if the Experiment 1 on Chinese L1 does not show a rhyme effect, then it is likely that the ESLs would not show such an effect in English either due to language transfer. They may also show a rhyme interference effect due to a joint contribution from phonological and lexical competition.

Experiment 1: Form preparation in Chinese

This experiment used a simple picture naming task with the form preparation paradigm where the critical variable for manipulation was Context, that is, whether the name of the pictures in a list shared the same onset, rhyme or neither (heterogeneous). The same items were used in both the same onset and same rhyme conditions with three lists in the same onset, same rhyme and the heterogeneous conditions, respectively. The heterogeneous condition served as a control condition (See Appendix A for all stimuli for each list). We expected to replicate the results found in previous literature (Chen & Chen, Reference Chen and Chen2013) where participants failed to show a facilitative effect in the same onset condition. We also expected to see a null effect in the same rhyme condition considering that speakers may retrieve and encode the syllable as an integral unit (Chen et al., Reference Chen, Chen and Dell2002; Chen & Chen, Reference Chen and Chen2013; O’Seaghdha et al., Reference O’Seaghdha, Chen and Chen2010), and that there were no orthographic cues (e.g., Pinyin) to encourage speakers to attend to the rhyme unit. An alternative prediction is that a rhyme interference effect might be shown, similar to the finding with Pinyin as the stimuli in Li et al. (Reference Li, Wang and Idsardi2015).

Results and discussion

Response time (RT) analyses were based on correct trials only (approximately 1% of the trials were incorrect and were deleted). Approximately 2% of the RT data was removed because the response failed to trigger the voice key or because hesitation or disfluency errors occurred. RTs that fell above or below 2.5 standard deviations away from the mean RT were also removed. In total, 4.8% of the RT data was removed, while the remaining RT data was log-transformed to improve normality. Analyses were carried out in R, an open source programming environment for statistical computing (R Development Core Team, 2008) with the lme4 package (Bates, Maechler, Bolker & Walker, 2013) and lmerTest package (Kuznetsova, Brockhoff & Christensen, 2013) for linear mixed effects modeling (LLM, GLMM). We reported analyses of the onset session and rhyme session separately in order to make the model more concise.

For the onset session, Context, OR (i.e., the sequence of onset and rhyme session), Block and Sequence (coded as Seq for brevity in R) were entered in the model as fixed effects whereas Participant and Item were entered as random slopes with Context serving as the intercept. The original mixed effect model entered in R for analysis of RT was “LogRT ~ Block * Seq * OR * Context+ (1 + Context|Participant) + (1 + Context|Item)”. The interaction terms were removed one at a time until we found the model with the best goodness-of-fit using the Chi-square test. Onset was not included as a fixed effect because it had already been embedded as an element in each item (i.e., each item had its own onset) and Item had been included as a random slope. Therefore, we decided to collapse onsets and to investigate the overall Context effect.

Likewise, for the rhyme session, Context, Block, OR and Sequence were entered in the model as fixed effects whereas Participant and Item were entered as random slopes with Context serving as the random intercept in the models. An analysis of accuracy rate in both sessions followed a similar procedure with the generalized linear (binomial) mixed-effects regression, but failed to show significant results. Also, the interaction terms were removed one at a time until we found the model with the best goodness-of-fit. See the upper part of Table 1 for summaries of the descriptive data of the onset session, and the bottom part for the rhyme session.

Table 1. Descriptive data of participants’ performance in Experiment 1 with mean reaction time (M), error rates (E%), standard errors (SE), and preparation effects

Note. * p < .05; **p< .01

The onset session

The main effect of Context did not reach significance (F (1, 14.7) = 2.0908, p = .16902) and also did not show a significant interaction with any other variables (ps> .10). As a result, the insignificant interaction terms have been removed. Participants also showed a significant Block main effect (F (2, 3294.9) = 9.7335, p < .001), indicating a possible practice effect that occurred over the course of the experiment considering that the RTs were faster in later blocks (Block 1: 649ms; Block 2: 632ms; Block 3: 632ms). See Table 2 for the full results.

Table 2. Results of the ANOVA approach to linear mixed-effect model analysis of RT of the onset session in Experiment 1

Note. ***: p < .001

The rhyme session

A significant rhyme interference effect was shown (F (1, 11.7) = 7.7152, p = .017) and as in the onset session, Context did not show significant interaction with any other variables (ps> .10). There was a significant Block × Seq × Context three-way interaction (F (2, 3251.5) = 3.6229, p = .02681), as well as a rhyme interference effect that was shown in all Block × Seq combinations with different effect sizes. Similar to the onset session, a significant Block main effect was shown (F (2, 3251.6) = 15.219, p < .001) since the RTs were faster in later blocks (Block 1: 647ms; Block 2: 633ms; Block 3:627ms). See Table 3 for the full results.

Table 3. Results of the ANOVA approach to linear mixed-effect model analysis of RT of the rhyme session in Experiment 1

Note. *: p < .05; ***: p < .001.

Consistent with previous literature (Chen & Chen, Reference Chen and Chen2013), native Chinese speakers did not benefit from the fore-knowledge of the onsets of words during picture naming, suggesting that the phoneme onset is not their phonological preparation unit. A significant rhyme interference effect was shown. It is possible that lexical competition could be the reason for this interference effect, this interference effect is in line with the lexical competition hypothesis proposed in Lukatela and Turvey (Reference Lukatela and Turvey1996) and the finding in the Pinyin condition in Li et al. (Reference Li, Wang and Idsardi2015) that used the form preparation paradigm. In the present study, naming a target picture (e.g., /pəŋ2/, meaning awning) may pre-inhibit its lexicalized rhyme neighbors. In particular, the item of the next trial (e.g., /təŋ2/, meaning pain) was pre-inhibited. Thus, it took longer time for the participants to name the next trial (i.e., /təŋ2/). Note that Lukatela and Turvey (Reference Lukatela and Turvey1996) employed reading materials that started with orthographic input instead of a conceptually driven production task. Li et al. (Reference Li, Wang and Idsardi2015) employed an associative naming task that also involved orthographic processing. Therefore, further investigation is needed to examine whether lexical competition or pre-inhibition explanation is applicable to the interference effect shown in the picture naming task in the present study.

Another interpretation of the rhyme interference effect is phonological competition as a result of different onsets, according to the parallel-then-sequential model. For example, the shared rhyme may activate /pəŋ2/ when participants needed to produce /təŋ2/, so that /t/ and /p/ competed when speakers assigned or assembled the phonological units sequentially. Given that no onset effect was shown in Experiment 1, it is likely that a larger unit than onset (e.g., syllable) was retrieved as an integral unit in language production. Therefore, we suggest that it is more likely that lexical competition rather than phonological competition led to the rhyme interference effect. At least, the lexical competition could be a stronger effect than the phonological competition. Importantly, the rhyme interference effect was shown in our study in a simple picture naming task without orthographic cues, suggesting that this effect is not an artifact but an intrinsic property in Chinese word production. However, given the different results shown and different methodology used in the previous studies about the effect of rhyme repetition in language production (Lukatela & Turvey, Reference Lukatela and Turvey1996; O’Seaghdha & Marin, Reference O’Seaghdha and Marin2000; Sevald & Dell, Reference Sevald and Dell1994), future research is called for to conduct in-depth investigation about the mechanism underlying rhyme effect across different languages.

Experiment 2: Form preparation in English

Experiment 2 aimed to investigate the phonological preparation unit in English among Chinese ESLs and native English speakers. If the preparation unit in the L1 affects the selection of the preparation unit in the L2, Chinese ESLs would not show an onset facilitation effect while native English speakers would. If the preparation unit only depends on the linguistic features of the target language but not on speakers’ L1 experience, both groups should show significant onset facilitation. If the preparation unit is influenced by recent language experience, Chinese ESLs may show onset facilitation only in later blocks. For native English speakers, if the rhyme unit functions similarly to Dutch, they may not show any rhyme effect. However, it is also possible that lexical competition between rhyme neighbors and phonological competition between the different onsets of rhyme neighbors may jointly lead to a rhyme interference effect. For Chinese ESL speakers, they may show a rhyme interference effect if their L1 has an influence.

Results and discussion

The procedure of data cleaning was same as the one used in Experiment 1. For the Chinese ESL group, 4.0% of the RT data was removed in total including incorrect answers (approximately 1%); failing to trigger the voice key, hesitation, or disfluency errors (1.2%); or RTs that were more than 2.5 standard deviations from the mean (1.8%). Similar to the Chinese ESL group, 4.3% of the native English speaking RT data was removed. Approximately 1% of data loss was due to incorrect answers; 1.5% was due to failing to trigger the voice key, hesitations, or disfluency errors; and 1.5% due to the RT being more than 2.5 standard deviations away from the mean. The remaining RTs were log-transformed to improve normality.

The procedure of data analysis was the same as that used in Experiment 1. For each session, the results of each Language Group were reported separately, and Proficiency (i.e., the BNT score, coded as Pro in R for brevity) was entered as a subject-level covariate for the Chinese ESL group. As a result, the original model including this covariate was “LogRT ~ Block * Seq * OR * Context+ (1 + Context|Participant) + (1 + Context*Pro|Item)” for the Chinese ESL group. Similarly to Experiment 1, the insignificant interaction terms were removed to find the model with the best goodness-of-fit. Next, we ran a new model in which the data of the two language groups were included, and Language Group (coded as Lang in R for brevity) was treated as a new fixed effect in the model. Since we did not measure native English speakers’ proficiency with the BNT, the Pro variable was not included in this model. See Table 4 for summaries of the descriptive data of the onset session, and Table 5 for the rhyme session.

Table 4. Descriptive data of participants’ performance in the Onset Session of Experiment 2 with mean reaction time (M), error rates (E%), standard errors (SE), and preparation effects

Note. * p < .05; ** p < .01; *** p < .001

Table 5. Descriptive data of participants’ performance in the Rhyme Session of Experiment 2 with mean reaction time (M), error rates (E%), standard errors (SE), and preparation effects

Note. * p < .05; ** p < .01.

The onset session

For the Chinese ESL group, the main effect of Context failed to reach significance (F (1, 11.1) = 3.297, p = .09648) and Context also did not show a significant interaction with Sequence or OR (ps > .10). However, Context did show a significant interaction with Block (F (2, 3252.2) = 4.154, p = .01578). In Block 1, the difference between the two contexts was 0ms; in Block two, participants showed a 9ms facilitation effect in the homogeneous context; in Block 3, participants showed 20ms facilitation in the same homogeneous context. Post-hoc analysis suggested that the effect of Context was only significant in Block 3 (p = .004), while it was not significant in the 1^st (p = .841) or 2^nd block (p = .286). A significant Block main effect was also found (F (2, 3252.1) = 47.714, p < .001), possibly due to a practice effect given that the RT showed a trend of being faster in the later blocks (Block 1: 678 ms; Block 2: 649 ms; Block 3: 644 ms). See Table 6 for the full results.

Table 6. Results of the ANOVA approach to linear mixed-effect model analysis of RT of the onset session in Chinese ESL group

Note. *: p < .05; ***: p < .001.

For the native English group, the main effect of Context was significant showing an onset facilitation effect (F (1, 12.7) = 18.8163, p < .001). All other main effects failed to reach significance. Context did not show a significant interaction with Sequence or OR (ps > .10). However, Context did show a significant interaction with Block (F (2, 3249.4) = 8.6155, p < .001). Although participants showed facilitation in all blocks, the effect size was larger in later blocks (Block 1: 8ms; Block 2: 16ms; Block 3: 36ms). The facilitation in Block 1 did not reach significance (p = .126), but reached significance in Block 2 (p = .016) and Block 3 (p < .001). See Table 7 for the full results.

Table 7. Results of the ANOVA approach to linear mixed-effect model analysis of RT of the onset session in native English group

Note. *: p < .05; **: p < .01; ***: p < .001.

After the two groups were combined, the critical result is the interaction between Language Group and Context or any three-way or four-way interactions that involve these two variables. Unfortunately, the combined model did not show a significant Lang × Context interaction (F (1, 23.8) = 2.4766, p = .128). Lastly, there were no significant three-way or four-way interactions that involved these two variables. The absence of the interaction may be due to the significant onset facilitation in Block 3 of the Chinese ESL group.

In summary, native English speakers showed a trend of facilitation at the very beginning (i.e., Block 1), and the effect reached significance in Block 2 and Block 3. Chinese ESLs did not show a trend of onset facilitation at the beginning (i.e., Block 1), but began to exhibit a trend of facilitation starting from the second block and reached significant onset facilitation in the third block. The above contrasts suggest that native English speakers benefit from the fore-knowledge of the phoneme very quickly in a naming task. On the other hand, Chinese ESLs do not routinely prepare spoken word production in small units (i.e., onset); however, they become attuned to onsets after practice in producing monosyllabic English words. This result suggests that the selection of phonological preparation unit in the L2 might be related to recent language experience.

The rhyme session

The Chinese ESL group showed a significant Context main effect (F (1, 9.4) = 34.155, p < .001), but it was in the direction of an interference effect from the homogeneous condition. Context also interacted significantly with Sequence and OR (F (1, 13.1) = 12.158, p = .00398 for Context and Sequence; F (1, 13.3) = 10.927, p = .00554 for Context and OR) where they showed significant interference effects for all combinations (ps < .05). In the Context and Sequence interaction, participants showed greater interference when they received the homogeneous lists first (37ms interference) compared to the smaller interference effect when the heterogeneous lists were given first (22ms interference), a difference of 15ms. This interaction may be the result of a practice effect considering that participants tended to react faster in the lists given later. For comparison in the Context and OR interaction, participants showed a larger interference effect when the onset session was given first (34ms interference) compared to when the rhyme session was given first (25ms interference), a difference of 9ms. It is possible that participants who received the onset session first may be more aware of the manipulation of the materials (i.e., homogeneous vs. heterogeneous Context). As a result, by the time they received the rhyme session, it was easier for them to be influenced by the manipulation, thus resulting in a larger Context effect. A Block main effect was also found (F (2, 3241.2) = 57.092, p < .001), possibly due to a practice effect where participants showed faster RTs in later blocks (Block 1: 694ms; Block 2:661ms; Block 3: 659ms). See Table 8 for the full results.

Table 8. Results of the ANOVA approach to linear mixed-effect model analysis of RT of the rhyme session in Chinese ESL group

Note. *: p < .05; **: p < .01; ***: p < .001.

The native English group also showed a significant rhyme interference effect (F (1, 14.67) = 5.2806, p = .03672) and Context showed a significant interaction with Sequence (F (1, 13.72) = 5.5765, p = .03356). The participants showed a significantly larger interference effect when they received the heterogeneous lists first (33ms interference effect) compared to when they received the homogeneous lists first (7ms interference), for a total difference of 26ms. It is possible that this interaction is due to a fatigue effect. There was also a complicated Block × Sequence × Context three-way interaction. See Table 9 for the full results. When the two language groups were entered into the regression model, the interaction between Language Group and Context was not significant (F (1, 23.6) = 2.226, p = .149) since both groups showed a rhyme interference effect.

Table 9. Results of the ANOVA approach to linear mixed-effect model analysis of RT of the rhyme session in native English group

Note. *: p < .05; **: p < .01.

The rhyme interference effect was similar to the effect found in Experiment 1. We suggest that this interference effect can be interpreted in the same way using the lexical and phonological competition hypotheses. For lexical competition, the activation of a trial (e.g., buck) inhibited the pre-activation of its rhyme neighbors, namely the following trials (e.g., suck). For phonological competition, the parallel retrieval of segments made the repetition reactivate the previous trial (e.g., buck) when speakers were producing the current trial (e.g., suck), and then onset competition (e.g., /b/ vs. /s/) occurs due to the sequential assignment of phoneme segments. Unlike Experiment 1 in which no onset effect was shown, both Chinese ESLs and native English speakers showed onset facilitation to some extent in Experiment 2, suggesting that onsets and rhymes may be processed separately in English. Therefore, it is likely that lexical competition and phonological competition have jointly contributed to the rhyme inhibition in Experiment 2. For Chinese ESLs, although Context showed significant interactions with other variables (i.e., Sequence and OR), the difference between the two sequences or the two ORs was just the size of the interference effect. In both sequences (i.e., homogeneous first or heterogeneous first) and ORs (i.e., onset first or rhyme first), Chinese ESLs showed significant rhyme interference. Therefore, we believed that the interference effect was an intrinsic property but not a practice or fatigue effect resulting from other factors (e.g., Sequence).