Word Frequency Dominance and L2 Word Recognition
September 12, 2016Vocab@Tokyo
Meiji Gakuin University, Tokyo, Japan1
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
2
Yu TAMURA (Nagoya University)Mitsuhiro MORITA(Hiroshima University)Yoshito NISHIMURA (Nagoya University)
3
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
4
• Morphology• Inflectional morphology
• -ed, -ing, 3rd-person -s, plural -s, -er
• Derivational morphology• prefix
• pre- (e.g., precondition), dis- (e.g., disagree)
• suffix• -able (e.g., wearable), -ish (e.g., boyish)
Introduction5
Morphological Processing
• Morphology• Inflectional morphology
• -ed, -ing, 3rd-person -s, plural -s, -er
• Derivational morphology• prefix
• pre- (e.g., precondition), dis- (e.g., disagree)
• suffix• -able (e.g., wearable), -ish (e.g., boyish)
Introduction6
Morphological Processing
• Morphology• Inflectional morphology
• -ed, -ing, 3rd-person -s, plural -s, -er
• Derivational morphology• prefix
• pre- (e.g., precondition), dis- (e.g., disagree)
• suffix• -able (e.g., wearable), -ish (e.g., boyish)
Introduction7
Morphological Processing
• Recognition process• Visual word recognition
• How morphology is processed in reading• Auditory word recognition
• How morphology is processed in listening
Introduction8
Morphological Processing
• Recognition process• Visual word recognition
• How morphology is processed in reading• Auditory word recognition
• How morphology is processed in listening
Introduction9
Morphological Processing
Findings of This Study• No evidence of direct access to the inflected
(plural) forms -> Morphological decomposition
10Introduction
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
11
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
12
• The more frequent, the faster• Three positions of the morphological processing
mechanism• Full-form storage model (e.g., Sereno &
Jongman, 1997)• Obligatory decomposition (e.g., Taft, 2004)• Dual-route model (e.g., Baayen, Dijkstra, &
Schreuder, 1997)
Background13
Frequency Effects
• The more frequent, the faster processing • Three positions of the morphological processing
mechanism• Full-form storage model (e.g., Sereno &
Jongman, 1997)• Obligatory decomposition (e.g., Taft, 2004)• Dual-route model (e.g., Baayen, Dijkstra, &
Schreuder, 1997)
Background14
Frequency Effects
• Full-form storage model (e.g., Sereno & Jongman, 1997)• Base forms and inflected forms
• stored separately• show frequency effects
Background15
Frequency Effects
rule rules
rule rules
• The more frequent, the faster processing • Three positions of the morphological processing
mechanism• Full-form storage model (e.g.,Sereno &
Jongman, 1997)• Obligatory decomposition (e.g., Taft, 2004)• Dual-route model (e.g., Baayen, Dijkstra, &
Schreuder, 1997)
Background16
Frequency Effects
• Obligatory decomposition (e.g., Taft, 2004)• Inflected forms
• are always decomposed• do not show frequency effects
Background17
Frequency Effects
rule rules
rule rules
• The more frequent, the faster processing • Three positions of the morphological processing
mechanism• Full-form storage model (e.g., Sereno &
Jongman, 1997)• Obligatory decomposition (e.g., Taft, 2004)• Dual-route model (e.g., Baayen, Dijkstra, &
Schreuder, 1997)
Background18
Frequency Effects
• Dual-route model (e.g., Baayen, Dijkstra, & Schreuder, 1997)• Frequently occurred inflected forms
• are processed as a whole• show frequency effects
Background19
Frequency Effects
kid kids
kid kids
rule rules
rule rules
High frequent inflected formsLow frequent inflected forms
faster
• Frequency difference between base forms and inflected forms• Singular-dominant nouns
• Singular (base) forms > plural (inflected) forms• e.g., ball, box
• Plural-dominant nouns• Plural (inflected) forms > singular forms (base)• e.g., kids, tears
Background20
Frequency Dominance
• Baayen et al. (1997)• Dutch• No Reaction Time (RT) difference between
• Plural dominant plurals and plural dominant singulars
• Highly frequent inflected forms would not be decomposed but processed as a whole
• Support dual-route model• New et al. (2004)
• French and English• Support Baayen et al. (1997)
Background21
Frequency Dominance
• Morita (2007)• Investigated whether the frequency of the
inflected words would affect the processing of the base forms
• Cumulative frequency (sg + pl) predicts the lexical decision time for native speakers of English
• -> dual-route or decomposition• Surface frequency (sg only) predicts the lexical
decision time for Japanese L2 learners of English• -> full-form strage?
Background22
Frequency Dominance
• How do L2 learners of English process and represent regularly inflected words?
• Hypothesis• If…
• frequent inflected forms < infrequent base forms -> highly frequent inflected forms are processed as
a whole• frequent inflected forms > infrequent base forms
-> inflected words are decomposed• frequent inflected forms > infrequent inflected forms
-> frequency of the base forms matter
Background23
Research Questions
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
24
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
25
• 72 Japanese undergraduate students
Table 1. Descriptive statistics of the TOEIC score
The Present Study26
Participants
N M SD Min MaxTOEIC score 72 575.42 104.19 325 800
1. Frequency list of nouns (both singular and plural forms) from British National Corpus (BNC)
2. 18 words which double or triple in frequency of singular form compared to plural form -> singular-dominant words
The Present Study27
Stimuli
3. 18 words which double or triple in frequency of plural form compared to singular form -> plural dominant words
4. 18 words whose frequency of singular and plural form was almost same. -> control words
The Present Study28
Stimuli
• The cumulative frequency (sg + pl) was controlled among the three groups
Table 2. Mean Frequency and SD in Parentheses
The Present Study29
Stimuli
k singular plural base
sg-domminant 18 69.865(25.849)
21.684(10.931)
91.549(34.342)
pl-dominant 18 22.571(18.661)
69.898(43.345)
92.469(59.779)
control 18 47.064(23.202)
43.893(24.664)
90.958(46.185)
Note. frequency is based on per million
The Present Study30
StimuliTable 3. List of Test Items
singular-dominant plural-dominant control
concept image parent proceeding topic element
film ball pound kid rabbit trend
science target standard tear bone secret
jacket video pupil resident store lesson
box hat individual finding principle firm
colour map detail critic horse step
bar context relation boot rule drug
network station resource participant function sport
college tower skill chemical plant document
• Judge whether the target words were real English words or not
• 54 test items (18*3) presented either in singular or plural form
• Carefully counterbalanced
• The same number of filler items were included
The Present Study31
Lexical Decision Task
• Incorrect responses removed (6.6%)• Outliers (M+3SD and RT below 200ms) removed (1.4%)• Generalized linear mixed-effect model (GLMM)
• Response variable• Raw RT
• Explanatory variable• Presentation (2 levels)
• singular or plural• Frequency dominance (3 levels)
• sg-dominant, pl-dominant, control • Post-hoc multiple comparison
The Present Study32
Analysis
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
33
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
34
35
Reaction Time
Results
k M SD 95%CILL UL
sg-domminant pl 9 838 246 818 858sg 9 765 232 747 783
pl-dominant pl 9 922 324 896 949sg 9 857 288 834 880
control pl 9 824 280 802 846sg 9 719 212 702 735
Table 4. Descriptive Statistics of Reaction Time (ms)
Note. N = 72. CI= Confidence Interval; LL = lower limit; UL = upper limit
Results36
Note. Error bar represents 95%CI
Results37
Note. Error bar represents 95%CI
Significant differences
Results38
Note. Error bar represents 95%CI
Results39
Note. Error bar represents 95%CI
Significant differences
No significant differences
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
40
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
41
• Singular forms judged faster than plural forms irrespective of the frequency dominance
• Singular forms• sg-dominant = control < pl-dominant
• Plural forms• sg-dominant = control < pl-dominant
Discussion42
Summary of the Results
• Singular forms judged faster than plural forms irrespective of the frequency dominance
• Singular forms• sg-dominant = control < pl-dominant
• Plural forms• sg-dominant = control < pl-dominant
Discussion43
Summary of the Results
• Singular forms judged faster than plural forms irrespective of the frequency dominance• Pl-dominant plurals did not show frequency
advantage• L2 learners always decompose plural
inflections
Discussion44
Morphological Processing
• Singular forms judged faster than plural forms irrespective of the frequency dominance
• Singular forms• sg-dominant = control < pl-dominant
• Plural forms• sg-dominant = control < pl-dominant
Discussion45
Summary of the Results
• Singular forms• sg-dominant = control < pl-dominant• Surface frequency advantage was only found
between sg-dominant and pl-dominant
• No clear evidence of the surface frequency effect• Frequency of the inflected forms had no effect on
the RT for the base forms
Discussion46
Morphological Processing
• Singular forms judged faster than plural forms irrespective of the frequency dominance
• Singular forms• sg-dominant = control < pl-dominant
• Plural forms• sg-dominant = control < pl-dominant
Discussion47
Summary of the Results
• Plural forms• sg-dominant = control < pl-dominant• No frequency advantage for pl-dominant plurals
• No evidence of direct access to the plural forms• High frequency inflected words were decomposed• Access latency for inflected forms might be
affected by base form frequency
Discussion48
Morphological Processing
• The experiment only focused on the surface frequency (cumulative frequency was controlled)
• The results were entirely on the basis of lexical decision task
-> priming task etc. might be needed
Discussion49
Limitations
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
50
Overview• Introduction• Background• The Present Study• Results• Discussion• Conclusion
51
• How do L2 learners of English process and represent regularly inflected words?• They decompose the inflected words
irrespective of frequency dominance-> Obligatory decomposition?• No RT difference between control words and
sg-dominant words• There still remains the possibility that L2
learners access abstract lexical entries which include both singular and plural forms
Conclusion52
Word Frequency Dominance and L2 Word Recognition
contact info Yu TamuraNagoya University
http://www.tamurayu.wordpress.com/
53
• Base form frequency seems to matter
• Inflected words always decomposed
• L2 learners access abstract lexical entries (sg + pl forms)
Baayen, R. H., Lieber, R., & Schreuder, R. (1997). The morphological complexity of simplex nouns. Linguistics, 35, 861–877. doi:10.1515/ling.1997.35.5.861
Morita, M. (2007) nihonjin eigo gakusyusya no meishi tansuukei ninshiki niokeru hinndo kouka: hyousou hindo to ruiseki hindo. [Frequency effects on recognition of singular nouns by Japanese learners of English: Surface frequency and cumulative frequency]. Bulletin of the Graduate School of Social & Cultural Systems at Yamagata University, 4, 9–19.
New, B., Brysbaert, M., Segui, J., Ferrand, L., & Rastle, K. (2004). The processing of singular and plural nouns in French and English. Journal of Memory and Language, 51, 568–585.
Sereno, J. A., & Jongman, A. (1997). Processing of English inflectional morphology. Memory & Cognition, 25, 425–437. doi:10.3758/BF03201119
Taft, M. (2004). Morphological decomposition and the reverse base frequency effect. The Quarterly Journal of Experimental Psychology. A, Human Experimental Psychology, 57, 745–765.
References54
55
GLMM
Results
Note. Number of observation = 3581. N = 72; K = 54. Dominance: 1 = control, 2 = pl-dominant, 3 = sg-dominant
Random effectsFixed effects By Subject By Items
Parameters Estimate
SE t p SD SDIntercept 925.32 23.12 40.03 <.001 67.18 52.15Dominance2-1,3 85.87 23.60 3.64 <.001 — —Dominance3-1,2 -27.10 20.92 -1.29 .195 — —Presentation1-2 -70.23 5.57 -12.62 <.001 — —Dom2-1,3:Pres 8.39 14.30 0.59 .557Dom3-1,2:Pres -23.317 12.06 -1.93 .053 — —
56
Post-hoc Multiple Comparison
Results
Dominance Estimate SE z p
control 65.26 9.16 7.12 <.0001
pl-dominant 56.87 10.85 5.24 <.0001
sg-dominant 88.57 8.52 10.39 <.0001
Simple main-effect of presentation (pl vs sg)
57
Post-hoc Multiple Comparison
Results
Presentation comparison Estimate SE z p
pluralctrl - pl -81.68 24.56 -3.33 .003ctrl - sg 15.44 21.65 0.71 .756pl - sg 97.12 30.64 3.17 .004
singularctrl - pl -90.06 24.76 -3.64 <.001ctrl - sg 38.76 21.90 1.77 .179pl - sg 88.57 8.52 10.39 <.001
Simple main-effect of frequency dominance