Is Acquiring Knowledge of Verb Subcategorization in English Easier? A Partial Replication of Jiang (2007) September 11, 2016 PacSLRF 2016 Chuo University, Tokyo, Japan 1 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 2 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 3 

• Purpose • To investigate integrated knowledge of adult L2 learners of English • How? • Using self-paced reading task • Findings • plural -s: ☓ , verb subcategorization: ̋ Brief overview of Jiang (2007) 4 

• Purpose • To investigate integrated knowledge of adult L2 learners of English • How? • Using self-paced reading task • Findings • plural -s: ☓ , verb subcategorization: ̋ Brief overview of Jiang (2007) 5 

Yu TAMURA Graduate School, Nagoya University Japan Society for the Promotion of Science 6 

• Integrated knowledge • Used spontaneously both in comprehension and production • Unconsciously activated • With minimal cognitive resource • With no or less attention to accuracy • Integrated knowledge <-> automatized performance Brief overview of Jiang (2007) 7 

• Why is automatized performance important? • It’s the ultimate goal of second language acquisition/instruction • SLA is the process of knowledge integration Brief overview of Jiang (2007) 8 

• Selective integration • The difference between child’s L1 acquisition and adult’s L2 acquisition • Some structures are more likely to be fossilized or less likely to be integrated • ESL learner’s knowledge of inflectional morphology never reaches at the level of native speakers • No matter what process it might be, integration of linguistic knowledge has to be selective • Ease of integration depends on linguistic structures Brief overview of Jiang (2007) 9 

• Purpose • To investigate integrated knowledge of adult L2 learners of English • How? • Using self-paced reading task • Findings • plural -s: ☓ , verb subcategorization: ̋ Brief overview of Jiang (2007) 10 

• Self-paced reading task • Required to read as fast as possible • Focus on meaning • Native speakers take longer time to read when they encounter grammatical errors. • Even without instruction • Even when the errors do not prevent comprehension • The delay is the evidence of possessing integrated knowledge Brief overview of Jiang (2007) 11 

• Self-paced reading task • Explicit knowledge cannot work as monitor during the task • Whether or not the learners have integrated knowledge can be measured as whether there is a delay in reading Brief overview of Jiang (2007) 12 

• Participants • Chinese ESL learners (N = 26) • Native speakers of English (N = 26) • Materials • plural morphemes : 32 items • verb subcategorization: 32 items • SVO + NP (10 items) • The mayor promised to offer/*keep the returning advisor a better position soon. • SVO + to infinitives (12 items) • The teacher wanted/*insisted the students to start all over again. • SVO + PP (6 items) • Her parents later married/*found her to a millionaire in Thailand. • SVO + adj (2 items) • Everyone considered/*believed the girl innocent after they had heard the story. Brief overview of Jiang (2007) 13 

• Participants • Chinese ESL learners (N = 26) • Native speakers of English (N = 26) • Materials • plural morphemes : 32 items • verb subcategorization: 32 items • SVO + NP (10 items) • The mayor promised to offer/*keep the returning advisor a better position soon. • SVO + to infinitives (12 items) • The teacher wanted/*insisted the students to start all over again. • SVO + PP (6 items) • Her parents later married/*found her to a millionaire in Thailand. • SVO + adj (2 items) • Everyone considered/*believed the girl innocent after they had heard the story. Brief overview of Jiang (2007) 14 

• Results • NS • significant RT differences in both structures • NNS • significant RT differences in only verb subcategorization Brief overview of Jiang (2007) 15 

• Discussion • Compatible to the results of Jiang (2004) • Why? • L1 influence • Teachability Hypothesis (Pienemann, 1989) • Weak Interface Hypothesis (R. Ellis, 1997) • Starting age (DeKeyser, 2000) • Frequency (N. Ellis, 2002) • However, none of the above factors can fully explain the results Brief overview of Jiang (2007) 16 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 17 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 18 

• Is the delay really the evidence of using integrated knowledge of verb subcategorization? • Ungrammatical version of the test items seemed not to be as much plausible as grammatical versions • ex. An attempt was made to persuade/*give the school board to change the policy. Problems with Jiang (2007) 19 

• L2 learners tend to use meaning-driven processing mechanism if the task does not require them to use syntactic processing (e.g., Lim and Christianson, 2013) • The RT differences obtained in Jiang (2007) might be due to breakdown of processing meaning Problems with Jiang (2007) 20 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 21 

• Purpose of the Present Study • To investigate the effect of comprehensibility of the test items used in Jiang (2007) The Present Study 22 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 23 

• Japanese undergraduate and graduate students (N = 32) Table 1. Demographic Information of the Participants Participants n M SD Min Max Age 31 24.77 5.35 20 40 TOEIC 32 824.22 113.12 550 990 Note. One participant did not report their age. 24 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 25 

• On the basis of the test items used in Jiang (2004, 2007) • Slightly modified some difficult vocabularies on the basis of JACET 8000 • millionaire -> rich; unwise ->ridiculous etc. • They had to teach the employees Chinese before sending them to China (Grammatical) • *They had to train the employees Chinese before sending them to China (Ungrammatical) • 64 test items (G: 32, UG:32) in total • Half of the items was followed by yes-no comprehension questions Materials and Procedures 26 

• How to identify the target regions? • Jiang (2007) • The teacher wanted the student to start all over again. • *The teacher insisted the student to start all over again. • “reading times for ‘start’ were compared” (p.13) • Shouldn’t it be “to”? • However, in some other items, two words after the target verb should be the target region • We all called him captain at the time. • *We all needed him captain at the time. • They had done little to make their children happy and successful in life. • * They had done little to provide their children happy and successful in life. Materials and Procedures 27 1 2 3 4 significant RT differences were reported in 3 and 4 How could these items be treated equally? 1 2 3 4 1 2 3 4 

• How to identify the target regions? • In Jiang (2007) • It seems the target regions were different across the test items, although the comparison is minimum within each pair • In this study • Target regions were set to be where the ungrammaticality first arises • *The teacher insisted the student to start all over again. • *We all needed him captain at the time. • * They had done little to provide their children happy and successful in life. Materials and Procedures 28 1 2 3 1 2 3 1 2 3 

____ ___ ________ _____ ___ _____ _____ ___ __ ____ _____ The ___ ________ _____ ___ _____ _____ ___ __ ____ _____ ___ teacher ________ _____ ___ _____ _____ ___ __ ____ _____ ____ ___ wanted _____ ___ _____ _____ ___ __ ____ _____ ____ ___ ______ _____ ___ _____ _____ ___ __ ____ _____ ____ ___ ______ _____ ___ _____ _____ ___ __ ____ _____ ____ ___ ______ _____ ___ _____ _____ ___ __ ____ _____ ____ ___ ______ _____ ___ _____ _____ ___ __ ____ _____ ____ ___ ______ _____ ___ _____ _____ ___ __ ____ _____ ____ ___ ______ _____ ___ _____ _____ ___ __ again. _____ ____ ___ ______ _____ ___ _____ _____ ___ __ ____ ࣍΁ Materials and Procedures Computer-based self-paced reading task • Moving window version • Word-by-word manner 29 

• Two counterbalanced forms (A and B) and two sessions • A1, B1: 16 sentences (G:8, UG:8) + 28 fillers • A2, B2: 16 sentences (G:8, UG:8) + 28 fillers • The order of the items was randomized • The order of the two sessions was counterbalanced Materials and Procedures 30 

• Comprehensibility questionnaire • Instructions were all written in Japanese • Five-point Likert scale • 1: ҙຯ͕·ͬͨ͘Θ͔Βͳ͍ (I don’t get the meaning of the sentence at all) — 5: ҙຯ͕ͱͯ΋Α͘Θ͔Δ (I get the meaning of the sentence very well) • The participants answered the questionnaire after they completed the self-paced reading task • The participants did not see the same items which they saw in the self-paced reading task Materials and Procedures 31 

• Analysis • Outliers removed (4.5%): • Responses below 200ms • Responses above the Mean RT+3SD of each participant in each condition • t1 = where the ungrammaticality first arises • t2 and t3 = for spill-over effects Materials and Procedures 32 

• Analysis • A series of Generalized linear mixed-effects models (GLMM) • Response variables: Raw RT • Explanatory variables: • grammaticality (condition): 2 levels • comprehensibility: centered around grand mean • word length: centered around grand mean • Gamma distribution and identity link function Materials and Procedures 33 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 34 

Results Table 2. Mean RTs (ms) and SDs (in parentheses) in each condition N = 32 35 t1 t2 ̓3 G 557 (144) 522 (112) 511 (110) UG 546 (128) 555 (135) 534 (112) t 0.69 1.77 1.16 p 0.50 0.09 0.25 Correlation 0.78 0.64 0.49 d -0.08 0.27 0.21 d (paired) -0.12 0.32 0.21 

200 250 300 350 400 450 500 550 600 ｔ1 t2 ｔ3 G UG 36 GLMM including only the main effect of condition found significant RT differences 

Results Table 3. The Results of Paired sample t-tests of the comprehensibility questionnaire N = 32 G UG M 4.12 3.80 SD 0.45 0.56 t1 t = 4.43, p < .001 t2 t = 2.20, p = .04 G UG 0 1 2 3 4 5 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● Figure 2. Box plot of the results of comprehensibility questionnaire. Red points indicate each participant’s mean score and blue points indicate mean scores in each condition. 37 Cohen’s d for item analysis d = 0.54 [0.04, 1.08] 

Results Table 4. The Results of GLMM (region t1) 38 Random effects Fixed effects By Subject By Items Parameters Estimate SE t p SD SD Intercept 584.71 15.29 38.24 <.001 80.17 55.85 Condition -7.81 13.92 -0.56 .57 51.70 50.93 comprehensibility 1.28 14.11 0.09 .92 46.22 — word length 23.17 5.92 3.91 <.001 — — Interaction 39.08 19.81 1.97 .048 — — Note. Number of observations: 999, N = 32, K =32 

Interaction between condition and comprehensibility Region t1 39 condition*c.comp in t1 c.comp rt 500 520 540 560 580 600 620 640 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG Note. grey and pink areas show 95%CI 

Results Table 5. The Results of GLMM (region t2) 40 Random effects Fixed effects By Subject By Items Parameters Estimate SE t p SD SD Intercept 559.69 15.74 35.55 <.001 68.41 44.90 Condition 43.392 14.81 2.93 <.01 63.53 52.71 comprehensibility 2.61 12.99 0.20 .840 48.78 12.32 word length 25.57 5.54 4.62 <.001 — — Interaction 35.56 14.12 2.52 .011 — — Note. Number of observations: 993, N = 32, K =32 

Interaction between condition and comprehensibility Region t2 41 condition*c.comp in t2 c.comp rt 500 520 540 560 580 600 620 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG Note. grey and pink areas show 95%CI 

Results Table 6. The Results of GLMM (region t3) 42 Random effects Fixed effects By Subject By Items Parameters Estimat e SE t p SD SD Intercept 522.25 14.49 36.04 <.001 59.12 36.84 Condition 20.91 12.71 1.65 .10 50.17 42.90 comprehensibility -15.27 12.49 -1.22 .221 41.49 15.71 word length 25.594 4.28 5.98 <.001 — — Interaction -29.19 16.00 -1.82 .068 — — Note. Number of observations: 998, N = 32, K =32 

Interaction between condition and comprehensibility Region t3 43 condition*c.comp in t3 c.comp rt 480 500 520 540 560 580 600 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG Note. grey and pink areas show 95%CI 

Overview • Brief overview of Jiang (2007) • Problems with Jiang (2007) • The present study • Participants • Materials and procedures • Results • Discussion 44 

Findings • Comprehensibility of the test items • Overall, the grammatical items were rated more comprehensible than the ungrammatical ones • However, some of the grammatical items were rated worse than their ungrammatical counterparts (see Appendix) • Those items were not acquired yet? • The effects of grammaticality and comprehensibility on RT • Possible interaction between grammaticality and comprehensibility • Grammaticality and comprehensibility might not be in a linear relationship Discussion 45 

Findings • Comprehensibility of the test items • Overall, the grammatical items were rated more comprehensible than the ungrammatical ones • However, some of the grammatical ones were rated worse than their ungrammatical counterparts (see Appendix) • Those items were not acquired yet? • The effects of grammaticality and comprehensibility on RT • Possible interaction between grammaticality and comprehensibility • Grammaticality and comprehensibility might not be in a linear relationship Discussion 46 

• Possible interaction between grammaticality and comprehensibility • In region t2 • The more comprehensible, the larger the effect of grammaticality • Learners’ sensitivity to the errors were found only if the sentences were comprehensible • No strong main effect of comprehensibility to the delay of RT • In region t1 and t3 • The less comprehensible, the larger the effect of grammaticality Discussion 47 condition*c.comp in t2 c.comp rt 500 520 540 560 580 600 620 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG condition*c.comp in t3 c.comp rt 480 500 520 540 560 580 600 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG 

Findings • Comprehensibility of the test items • Overall, the grammatical items were rated more comprehensible than the ungrammatical ones • However, some of the grammatical ones were rated worse than their ungrammatical counterparts (see Appendix) • Those items were not acquired yet? • The effects of grammaticality and comprehensibility on RT • Possible interaction between grammaticality and comprehensibility • Grammaticality and comprehensibility might not be in a linear relationship Discussion 48 

• Grammaticality and comprehensibility might not be in a linear relationship • The effect of grammaticality was influenced by the comprehensibility of the test items • L2 learners use both meaning driven and syntactic-driven processing dynamically during self-paced reading • RT differences observed in the study might not be all due to the fact that L2 learners automatized the knowledge of verb-subcategorization Discussion 49 

• The test items used in Jiang (2007) need a careful revision to examine the knowledge of verb-subcategorization • Syntactic position of the target regions should be controlled across the sentences • Ideally, the types of constructions (e.g., SVO + to V, SVO + PP, etc.) should also be controlled • Selective integration? • Number agreement -> less effect of ungrammaticality to the meaning • Subcategorization -> more effect of ungrammaticality to the meaning • These two types of grammatical knowledge should not be directly compared • GLMM would be preferable • to take into account word length • to take into account participants’ and items’ variance • to see the interaction between meaning and syntactic processing Discussion 50 

• Limitations • The participants in Jiang (2007)’s study were more proficient • Determination of the target regions might be different than the original study Discussion 51 

Jiang, N. (2007). Selective integration of linguistic knowledge in adult second language learning. Language Learning, 57, 1–33. doi: 10.1111/j.1467-9922.2007.00397.x Lim, J. H., & Christianson, K. (2013). Integrating meaning and structure in L1–L2 and L2–L1 translations. Second Language Research, 29, 233–256. doi:10.1177/0267658312462019 References 52 

Is Acquiring Knowledge of Verb Subcategorization in English Easier? A Partial Replication of Jiang (2007) contact info Yu Tamura Graduate School, Nagoya University [email protected] http://www.tamurayu.wordpress.com/ 200 250 300 350 400 450 500 550 600 ｔ1 t2 ｔ3 G UG • The test items and the analyses should be revised • The effect of grammaticality was influenced by comprehensibility of the items 53 condition*c.comp in t1 c.comp rt 500 520 540 560 580 600 620 640 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG condition*c.comp in t2 c.comp rt 500 520 540 560 580 600 620 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG condition*c.comp in t3 c.comp rt 480 500 520 540 560 580 600 −1.0 −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 condition G UG 

54 comprehensibility condition k t1 t2 t3 G>UG G 9 569 526 504 UG 9 546 557 544 UG>G G 23 528 510 523 UG 23 548 548 508 All G 32 557 522 510 UG 32 546 555 534 Table 7. Mean RTs (ms) across three types of items in each condition 

55