Short- and Long-Term Effects of Rote Rehearsal on ESL Learners' Processing of L2 Collocations

Worldwide there is thought to be around 750 million people who speak English as a foreign language (Crystal, 2003, p. 69). For these speakers the difference between make a picture and take a picture may seem arbitrary. However, use of the former is likely to influence how their second language (L2) performance is perceived (Boers, Eyckmans, Kappel, Stengers, Benson, Benson, Yamashita Wolter age: M = 27; SD = 5.17) L1-Polish participants living in the United Kingdom were recruited for the study. They were all advanced learners of English, as demonstrated by the results of Schmitt, Schmitt, and Clapham's (2001) Vocabulary Levels Test (see Table 1). The study included three sessions: a pretest, a treatment with an immediate posttest, and a delayed posttest. Eighteen participants were tested individually via an acceptability judgment task (Yamashita e.g., warm welcome) and 40 verb-noun (VN; e.g., hold a raffle) combinations. They were divided into two counterbalanced lists in such a way that each of them contained 20 frequent (10 AN and 10 VN) and 20 infrequent (10 AN and 10 VN) combinations (see online supporting information). All collocations were carefully matched, such that they all had similar length and high mutual information scores indicating strongly connected combinations (for a discussion of mutual information, see Manning p = .001; r = .78; RR: z = −2.98; p = .003; r = .70) and infrequent AN collocations (RR plus: z = −2.11; p = .04; r = .51; RR: z = −2.72; p = .006; r = .64). As far as the VN collocations are concerned, only the RR condition reached significance on the infrequent items (z = −2.44; p = .02; r = .59). For comparisons between the pretest and the delayed posttest data, only the RR plus condition revealed significant differences on the infrequent AN (z = −2.77; p = .006; r = .68) and infrequent VN collocations (z = −2.07; p = .04; r = .49). However, these results did not differ from the control group where RTs also decreased, which means that no claims about long-term effects of the treatment can be made. RR plus N = 18 RR N = 18 NSs N = 14 Second, given that both RR plus and RR led to shorter RTs on the AN collocations, their effectiveness was compared. No significant differences between the two conditions were found: (frequent AN: z = −1.07; p = .27; r = .25; infrequent AN: z = −0.92; p = .36; r = .22). Moreover, Mann-Whitney tests were run to compare learners and NSs. The pretest results showed that NSs were significantly faster than learners in both the RR plus condition (frequent AN: U = 62; z = −2.43; p = .02; r = .43; infrequent AN: U = 39; z = −3.18; p = .001; r = .56; frequent VN: U = 58; z = −2.58; p = .01; r = .46; infrequent VN: U = 35; z = −3.46; p = .001; r = .61) and the RR condition (frequent AN: U = 58; z = −2.58; p = .01; r = .46; infrequent AN: U = 54; z = −2.74; p = .006; r = .48; frequent VN: U = 60; z = −2.51; p = .01; r = .44; infrequent VN: U = 37; z = −3.26; p = .001; r = .58). Immediately after the treatment, these differences disappeared in both the RR plus and RR conditions. The only exception was the infrequent VN collocations in the RR condition where NSs were still significantly faster than learners (U = 72; z = −2.05; p = .04; r = .36). The delayed posttest revealed that the knowledge of the VN collocations was not retained—learners were again significantly slower than NSs in both the RR plus condition (frequent VN: U = 54; z = −2.74; p = .006; r = .48; infrequent VN: U = 66; z = −2.28; p = .02; r = .40) and the RR condition (frequent VN: U = 45; z = −3.08; p = .002; r = .53; infrequent VN: U = 67; z = −2.24; p = .03; r = .40). Third, Wilcoxon tests were conducted to compare participants' RTs on the frequent and infrequent collocations. During all the sessions, the former were processed significantly more quickly than the latter in both the RR plus (pretest VN: z = −3.57; p = .001; r = .08; pretest AN: z = −3.37; p < .001; r = .87; delayed VN: z = −2.03; p = .04; r = .48) and RR (pretest VN: z = −2.86; p = .004; r = .07; pretest AN: z = −3.59; p < .001; r = .85; delayed VN: z = −3.38; p = .001; r = .80; delayed AN: z = −3.29; p = .001; r = .80) conditions. The only exception was the delayed posttest data on the AN collocations in the RR plus condition, which failed to reach significance (z = −1.85; p = .06; r = .44). As Table 3 demonstrates, error levels on the pretest were relatively low and comparable for learners and NSs. Thus, there was little room for improvement resulting from the treatment. Despite apparent gains on the infrequent collocations, no claims about learners' improved accuracy can be made, because the experimental groups did not differ significantly from the control group that also improved: RR plus (AN: U = 66; z = −0.43; p = .67; r = .08; VN: U = 49.5; z = −1.53; p = .13; r = .03) and RR (AN: U = 55.5; z = −1.33; p = .18; r = .26; VN: U = 65.5; z = −0.4; p = .69; r = .08). RR plus N = 18 RR N = 18 Control N = 8 NS N = 14 When the experimental conditions were compared, no significant differences were found. The only exception to this was in the immediate posttest data, where the infrequent VN collocations elicited significantly fewer errors in the RR condition compared to the RR plus condition (z = −2.12; p = .03; r = .51). However, this improvement was short term, as the delayed posttest data did not reveal this difference (z = −0.91; p = .37; r = .21). Frequency of collocations was found to influence the error rate. The pretest data revealed that in both conditions learners' errors for the infrequent collocations were significantly higher than for the frequent collocations: RR plus (AN: z = −2.88; p = .004; r = .70; VN: z = −3.22; p = .001; r = .76) and RR (AN: z = −3.25; p = .001; r = .77; VN: z = −3.07; p = .002; r = .72). When the same comparisons were made after the treatment, the delayed posttest data revealed that these differences were no longer found: RR plus (AN: z = −1.6; p = .11; r = .38; VN: z = −0.82; p = .41; r = .19) and RR (AN: z = −0.45; p = .66; r = .11; VN: z = −1.63; p = .10; r = .38). This study examined ESL learners' processing of AN and VN collocations following two different instructional treatments: RR and RR plus. Both treatments led to more effective processing of collocations, as demonstrated by faster RTs for the frequent and infrequent AN collocations. The fact that VN collocations did not demonstrate an overall benefit from the treatments indicates differences in L2 learners' processing of different types of collocational patterns. This is likely to have been caused by L1–L2 congruency effects (30 out of 40 VN vs. 1 out of 40 AN were incongruent collocations). Previous research (Wolter Yamahita otherwise attrition will be evident. Furthermore, similar gains in both the RR and RR plus groups seem to indicate that the repetition of L1 and L2 sentences containing the target collocations was the most important factor in facilitating learners' processing of them. It is likely that the use of both languages (L1-Polish and L2-English) contributed to the benefits of the treatment. Importantly, the lack of additional gains in the RR plus group shows that when there are repetitions and presentation of the L1 and L2, there is no further benefit of underlining. This is an interesting finding that betters our understanding of input enhancement, which might be effective only in combination with other types of treatment (Szudarski, 2012). Finally, the study demonstrates the importance of collocational frequency. We found significant differences between learners' processing of the frequent and infrequent collocations, with the former resulting in shorter RTs and higher accuracy rates. This adds evidence to a growing body of research (e.g., Siyanova-Chanturia et al., 2011) showing that L2 learners' processing of formulaic language is sensitive to frequency effects. Moreover, the accuracy analysis revealed that the L2 learners had achieved native-like levels for the frequent collocations. This shows that such combinations do not pose difficulty for ESL learners at a receptive level of mastery, which accords with Laufer and Waldman's (2011) assertion that “the real problem of collocations in L2 does not lie in recognition but in learning to use them properly” (p. 652). In sum, the study provides empirical evidence for the short-term benefits of rote rehearsal in L2 learners' processing of collocations. Repetition was found to facilitate ESL learners' results, with the addition of input enhancement (underlining) not resulting in more gains. The study also revealed that learners' gains were not retained after 6 weeks, which emphasizes the importance of consolidating the newly acquired L2 collocational knowledge. Lastly, the frequency and type of collocations were confirmed as two important factors that influence L2 learners' lexical development. Paweł Szudarski is a lecturer at Kazimierz Wielki University, Bydgoszcz, Poland. Recently he completed his PhD studies at the University of Nottingham. Earlier he worked for the Center for Applied Linguistics in Washington, DC. His research interests include multiword units and phraseology, L2 vocabulary acquisition and testing, English language learners, and corpus linguistics. Kathy Conklin is an associate professor in psycholinguistics at the University of Nottingham. She is interested in multiword units, automatic word activation in an L1 vs. L2, bilingual/monolingual cognitive control, and implicit L2 learning. To investigate L1/L2 language representation and processing, she makes use of behavioral measures, eye-tracking, and EEG. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.