causal and semantic relatedness effects on l2 text processing and memory: evidence from self-paced...
TRANSCRIPT
Causal and Semantic Relatedness Effects on L2 Text Processing and Memory:
Evidence From Self-paced Reading and Recall
Shingo Nahatame Kyoei University, Japan
Visit my website!
Introduction Literature Review This Study Method Results & Discussion Conclusion
Introduction
Research Interest: Discourse comprehension in L2
written text consisting of multiple sentences
cf. word recognition, syntactic processing
This study: examines L2 discourse processing and memory
focuses on the connectivity between sentences
Literature Review
Sentence connectivity ・Causal Relatedness (CR) cause-and-effect relations between events
described in the sentences
・Semantic Relatedness (SR) similarity of meaning of words included in
the sentences.
CR and SR play an important role in the comprehension process and production of text
(e.g., Graesser et al., 1994; Kintsch & van Dijk, 1978)
Causal Relatedness High in casual relatedness Mary could not find anything to read in the library. She
went to the bookstore to get new books.
L1 readers process faster and recall better the text high in CR than the text low in CR.
(Barres et al., 2015; Keenan et al., 1984; Myers et al., 1987)
Low in casual relatedness Mary went to the library to look for something to read. She
went to the bookstore to get new books.
Semantic Relatedness High in semantic relatedness
Mary could not find anything to read in the library. She went to the bookstore to get new books.
L1 readers process faster and recall better the text high in SR than the text low in SR.
(Todaro et al., 2010; Wolfe et al., 2005)
Low in semantic relatedness Mary wanted to look for recipes for her dinner party. She went to the bookstore to get new books.
Semantic Relatedness
Text processing is more strongly influenced by CR than SR: The effect of SR is only significant when CR is lower (Wolfe et al., 2005)
High in semantic relatedness Mary could not find anything to read in the library. She went to the bookstore to get new books.
Low in semantic relatedness Mary wanted to look for recipes for her dinner party. She went to the bookstore to get new books.
Semantic Relatedness SR is assessed by latent semantic analysis (LSA) a computational approach for estimating semantic
similarity between the meanings of words using by statistical computations applied to a large corpus of texts (Landauer et al., 2007)
based on direct and indirect relations among words
(i.e., words that tend to be used in similar contexts.)
Context 1: Birds fly in the sky. Context 2: Birds have two wings. = “fly” and “wing” are related via “birds” SR by LSA approximately ranges from 0 to 1.
L2 Studies on CR
・Causal Relatedness (CR) L2 readers recall better sentence pairs high in
CR than those low in CR (Horiba, 1996)
L2 readers process faster sentence pairs high in CR than those low in CR (Shimizu, 2009)
L2 readers judge sentence pairs high in CR to be more coherent than those low in CR
(Nahatame, in press) L2 readers are sensitive to CR of paired
sentences.
L2 Studies on SR
・Semantic Relatedness (SR) SR between sentences computed by LSA is one
of the indications of L2 text readability (Crossley et al., 2012; Crossley et al., 2007; Crossley & McNamara,
2008) L2 readers judge sentence pairs high in SR to
be more coherent than those low in SR (Nahatame, in press)
Few empirical studies have examined the effect of SR on text processing and memory.
This Study
This Study This study aimed to examine co-influence of CR and SR on L2 discourse processing (measured by self-paced reading task) and memory (measured by recall task).
RQ1: Do CR and SR between sentences affect L2 readers’ text processing?
RQ2: Do CR and SR between sentences affect L2 readers’ text memory?
This Study This study aimed to examine co-influence of CR and SR on L2 discourse processing (measured by self-paced reading task) and memory (measured by recall task).
CR will have a strong impact on both L2 text processing and memory.
SR might have an impact on L2 text memory and processing.
Method
Method Participants
• 45 undergraduate and graduate students in Japan (26 males and 19 males; ages varied from 19 -26 years old)
• Native speakers of Japanese and learners of
English. • Independent users of English; B1 to B2 levels of
CEFR (confirmed by the test)
Method Materials
• L2 reading proficiency test (based on Eiken; Cronbach’s α = .80) • 20 sets of paired sentences
(adapted from Nahatame, in press; originally from Wolfe et al., 2005)
simple syntactic structures and highly frequent words
four types of the first (prime) sentence and one common second (target) sentence.
Method Materials
1a. Mary could not find anything to read in the library. (CR-High / SR-High)
1b. Mary wanted to look for recipes for her dinner party. (CR-High / SR-Low)
1c. Mary went to the library to look for something to read. (CR-Low / SR-High)
1d. Mary was having a dinner party for her office. (CR-Low / SR-Low)
2. She went to the bookstore to get new books. (Target sentence)
Method Materials
1a. Mary could not find anything to read in the library. (CR-High / SR-High)
1b. Mary wanted to look for recipes for her dinner party. (CR-High / SR-Low)
1c. Mary went to the library to look for something to read. (CR-Low / SR-High)
1d. Mary was having a dinner party for her office. (CR-Low / SR-Low)
2. She went to the bookstore to get new books. (Target sentence)
Method Materials
1a. Mary could not find anything to read in the library. (CR-High / SR-High)
1b. Mary wanted to look for recipes for her dinner party. (CR-High / SR-Low)
1c. Mary went to the library to look for something to read. (CR-Low / SR-High)
1d. Mary was having a dinner party for her office. (CR-Low / SR-Low)
2. She went to the bookstore to get new books. (Target sentence)
Method Materials
1a. Mary could not find anything to read in the library. (CR-High / SR-High)
1b. Mary wanted to look for recipes for her dinner party. (CR-High / SR-Low)
1c. Mary went to the library to look for something to read. (CR-Low / SR-High)
1d. Mary was having a dinner party for her office. (CR-Low / SR-Low)
2. She went to the bookstore to get new books. (Target sentence)
Method Materials
1a. Mary could not find anything to read in the library. (CR-High / SR-High)
1b. Mary wanted to look for recipes for her dinner party. (CR-High / SR-Low)
1c. Mary went to the library to look for something to read. (CR-Low / SR-High)
1d. Mary was having a dinner party for her office. (CR-Low / SR-Low)
2. She went to the bookstore to get new books. (Target sentence)
Method • Causal manipulation was ensured by the preliminary
study in Nahatame (in press) with L2 readers of English. Across all pairs, sentences high in CR were judged to
be more causally related than those low in CR
Method • Semantic manipulation was ensured by employing LSA. Across all sentence pairs, sentences high in SR had
higher LSA values between sentences than those low in SR
Method Procedure
• L2 reading test
• Self-paced reading task Participants were randomly assigned to one of four
presentation lists
conducted on the computer
sentence by sentence reading
Method Procedure
• Self-paced reading task
Was the first batter new boy on the team?
???
Comprehension question
Self-paced reading
(sentence by sentence)
Time
750 ms
until response
until response
until response
until response
She went to the bookstore to get new books.
Mary could not find anything to read in the library.
Ready?
Method Procedure
• Cued recall task given the first sentences of each pair as cues (e.g., Mary could not find anything to read in the library.) wrote down the second sentences of each pair in their
L1 (e.g., She went to the bookstore to get new books.)
Method Scoring and Analysis
<Reading Times> • Reading times were divided by the number of syllables
to control for sentence length. • Reading times exceeding ±2.5 SDs above the mean for
each participant were excluded (3.22%).
Method Scoring and Analysis
<Reading Times> • Linear mixed-effects (LME) modeling (R & lme4 package) Dependent variables: Reading times for the second sentence Random effects: Participants and items Fixed effects (continuous): CR (judgment ratings), SR (LSA values), and
Proficiency (test scores) and their interaction By-subject random slopes for SR, CR, and their interaction, and
by-item random slopes for Proficiency. Covariates: sentence length and word frequency All factors were centered around its mean value. Models were selected based on their AIC. Significance: t ≥ 2
Method Scoring and Analysis
<Recall Protocols> • When a target sentence was reproduced verbatim or
paraphrased, it was given one point (binary scoring of 1 or 0).
• 30% of recall protocols were assessed by two raters separately (95% agreement).
Method Scoring and Analysis
<Recall Protocols> • Generalized linear mixed-effects (LME) modeling Dependent variables: Recall scores (binary) the function glmer and the family = binomial argument. Random effects: Participants and items Fixed effects (continuous): CR (judgment ratings), SR (LSA values), and
Proficiency (test scores) and their interaction By-subject random slopes for SR, CR, and their interaction, and
by-item random slopes for Proficiency. Covariates: sentence length and word frequency All factors were centered around its mean value. Models were selected based on their AIC.
Results & Discussion
Results & Discussion Reading Times: A significant main effect of CR, which was qualified by a significant SR × CR interaction
Random effects
Fixed effects By Subject By Items
Parameters Estimate SE t SD SD
Intercept 448.65 26.76 16.77* 107.44 92.26
cSR −56.82 39.00 −1.46 –––––
–––––
cCR −22.31 3.31 −6.74* –––––
–––––
cPro −5.14 4.98 −1.03 ––––– –––––
cSR x cCR −50.38 21.56 −2.34* –––––
–––––
cSR x cPro 3.13 11.26 0.28 –––––
–––––
cCR x cPro 1.32 0.97 1.36 ––––– –––––
cSR x cCR x cPro 4.02 6.32 0.64 ––––– –––––
Note. Model formula: RT ~ cSR * cCR * cPro + (1 | subject) + (1 | item)
Higher CR resulted in the shorter reading times; the effect was stronger when SR was higher.
Results & Discussion
The effect of SR was prominent only when CR was higher.
Results & Discussion
Reading Times (RQ1): <Causal relatedness>
• CR consistently affects L2 readers’ text processing. When the sentences have a clear cause-and-effect
relation, the processing is facilitated. (Shimizu, 2009)
The effect of CR is stronger when SR is higher.
Results & Discussion
Reading Times (RQ1): <Semantic relatedness>
• SR also influences L2 readers’ text processing. When the sentences include words that tend to be
used in similar contexts, the processing is facilitated.
The effect of SR is only clear when CR is higher. = The effect of SR might be secondary to CR on text
processing (Wolfe et al., 2005) .
Results & Discussion
Recall performance: Significant main effects of SR, CR, and Proficiency. There was a significant interaction between SR and CR.
Results & Discussion
Random effects
Fixed effects By Subject By Items
Parameters Estimate SE z SD SD
Intercept 0.58 0.14 4.11* 0.72 0.20
cSR 2.66 0.52 5.09* –––––
–––––
cCR 0.40 0.04 8.49* –––––
–––––
cPro 0.09 0.39 2.23* –––––
–––––
cSR x cCR −0.81 0.30 −2.71* –––––
–––––
cSR x cPro −0.02 0.15 −0.13 ––––– –––––
cCR x cPro 0.02 0.01 1.83 ––––– –––––
cSR x cCR x cPro −0.10 0.09 −1.15 ––––– –––––
Note. Model formula: Recall ~ cSR * cCR * cPro + (1 | Subject) + (1 | item)
Higher CR resulted in the better recall; the effect was prominent when SR was lower.
Results & Discussion
Higher SR resulted in the better recall; the effect was prominent when CR was lower.
Results & Discussion
Recall performance (RQ2): <Causal relatedness>
• CR consistently affects L2 readers’ text memory. When the sentences have a clear cause-and-effect
relation, the memory is enhanced. (Horiba, 1996; Shimizu, 2009)
The effect of CR is stronger when SR is lower.
Results & Discussion
Recall performance (RQ2): <Semantic relatedness>
• SR also consistently influences L2 readers’ text memory. When the sentences include words that tend to be
used in similar contexts, the memory is enhanced.
The effect of SR is stronger when CR is lower. = CR and SR functioned in a complemental manner in
text memory (Wolfe et al., 2005).
= When the text is low in CR, participants were more likely to depend on SR to retrieve text information, and vice versa.
Results & Discussion
Conclusion
Causal relatedness between sentences affect L2
text processing and memory. Semantic relatedness,
which is estimated based on contextual-usage
meaning of words, is also a deciding factor in L2
text processing and memory, although for text
processing, its effect is overwhelmed by the effect
of causal relatedness.
Conclusion Take-home message
Limitations • Participants’ proficiency (only intermediate-
level learners were included) • Use of short texts: Investigation using the longer
and whole passage • Statistical analysis: Updating and reexamination
of the models
References Barnes, M. A., Ahmed, Y., Barth, A. & Francis, D. J. (2015). The relation of knowledge-text
integration processes and reading comprehension in 7th- to12th-grade students. Scientific Studies of Reading, 19, 253–272. doi: 110.1080/10888438.2015.1022650
Crossley, S. A., Allen, D., & McNamara, D. S. (2012). Text simplification and comprehensible input: A case for an intuitive approach. Language Teaching Research, 16, 89–108. doi: 10.1177/1362168811423456
Crossley, S. A., Louwerse, M., McCarthy, P. M., & McNamara, D. S. (2007). A linguistic analysis of simplified and authentic texts. The Modern Language Journal, 91, 15–30. doi: 10.1111/j.1540-4781.2007.00507.x
Crossley, S. A., & McNamara, D. S. (2008). Assessing second language reading texts at the intermediate level: An approximate replication of Crossley, Louwerse, McCarthy, and McNamara (2007). Language Teaching, 41, 229–409. doi: 10.1017/S0261444808005077
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative text comprehension. Psychological Review, 101, 371–395.
Horiba, Y. (1996b). The role of elaborations in L2 comprehension: The effect of encoding task on recall. The Modern Language Journal, 80, 151–164. doi: 10.1111/j.1540-4781.1996.tb01155.x
Keenan, J. M., Baillet, S. D., & Brown, P. (1984). The effect of causal cohesion on comprehension and memory. Journal of Verbal Learning and Verbal Behavior, 23, 115–126. doi: 10.1016/S0022-5371(84)90082-3
Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psychological Review, 85, 363–394. doi:10.1037/0033-295X.85.5.363
Landauer, T., McNamara, D. S., Dennis, S., & Kintsch, W. (2007). Handbook of Latent Semantic Analysis. Mahwah, NJ: Erlbaum
References Myers, J. L., Shinjo, M., & Duffy, S. A. (1987). Degree of causal relatedness and memory. Journal of
Memory and Language, 26, 453–465. doi: 10.1016/0749-596X(87)90101-X Nahatame, S. (in press). Standards of coherence in second language reading: Sentence connectivity
and reading proficiency. Reading in a Foreign Language, 29. Shimizu, H. (2009). The effects of causal relatedness on EFL learners’ reading comprehension and
inference generation. ARELE (Annual Review of English Language Education in Japan), 20, 31–40. Todaro, S., Millis, K., & Dandotkar, S. (2010). The impact of semantic and causal relatedness and
reading skill on standards of coherence. Discourse Processes, 47, 421–446. doi: 0.1080/01638530903253825
Wolfe, M. B. W., Magliano, J. P., & Larsen, B. (2005). Causal and semantic relatedness in discourse understanding and representation. Discourse Processes, 39, 165–187. doi: 10.1080/0163853X.2005.9651678
Acknowledgment This research was supported by Grants-in-Aid for Research Activity Start-up (15H06571) from the Japan Society for the Promotion of Science. I am grateful to Dr. Michael B. W. Wolfe for providing the experimental materials for this study.
Shingo Nahatame, Email: [email protected] Website: https://sites.google.com/site/snahatame Twitter ID: @Nahacchi
Causal and Semantic Relatedness Effects on L2 Text Processing and Memory:
<Reading Times>
<Recall Performance>
• CR and SR have an influence on processing.
• SR is secondary to CR.
SR measured by LSA.
• CR and SR affect memory in a complemental manner.