The Cross-Script Length Effect: Evidence for Serial Processing in Reading Aloud Kathleen Rastle (Royal Holloway University of London), Linda Bayliss (Royal Holloway University of London), Jelena Havelka (University of Kent), Taeko Wydell (Brunel University) Background Reading aloud latency increases as the number of letters in a stimulus increases, but this increase is larger for nonwords (e.g., pon, plon, plont) than it is for words (e.g., pan, plan, plant). This length by lexicality interaction (Weekes, 1997) is one of the key findings used in support of dual-route theories of reading aloud that incorporate a serial nonlexical procedure (e.g., the DRC model; Coltheart et al., 2001). Nonwords in the DRC model are read primarily by the serially-operating nonlexical route, and that is why they show a particularly strong effect of length. Experiment 1 Japanese Kana comprises Hiragana ( きききき ) and Katakana ( きき きき ) scripts, with a large number of spoken words only ever appearing in one script. These words could be transcribed into the other script and read aloud but they would be orthographically unfamiliar. Thus, a single spoken word can be presented as an orthographic word or as an orthographic nonword in Japanese Kana. Will the length by lexicality interaction be observed under these conditions? Experiment 2 Serbian comprises Cyrillic and Roman alphabets, each of which has a set of unique letters that occur only in that alphabet. The alphabets share seven common letters that are pronounced identically in each, and four ambiguous letters that are pronounced differently in each. Discussion: •The length effect was larger in the Japanese study when stimuli were presented in an orthographically-unfamiliar form than when they were presented in an orthographically-familiar form, and in the Serbian study when stimuli were read as pronounceable nonwords than when they were read as words. •These writing systems are perfectly spelling-to-sound consistent, suggesting that the length by lexicality interaction is not the result of a frequency by consistency interaction. •Targets were very closely matched – phonologically identical in the Japanese study and orthographically identical in the Serbian study – making it unlikely that the effects were due to peripheral serial processes such as eye movements or articulation. •These data constitute the strongest evidence to date for serial nonlexical processing in reading aloud. References: •Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). The DRC model: A model of visual word recognition and reading aloud. Psychological Review, 108, 204-256. •Plaut, D.C., McClelland, J.L., Seidenberg, M.S., & Patterson, K. (1996). Understanding normal and impaired word reading: Computational principles in quasi-regular domains. Psychological Review, 103, 56-115. •Weekes, B.S. (1997). Differential effects of number of letters on word and nonword naming latency. Quarterly Journal of Experimental Psychology, 50A , 439-456. The length by lexicality interaction has also been used to argue against PDP models of reading aloud (e.g., Plaut et al., 1996). In English, however, length is confounded with consistency, with longer stimuli typically being more inconsistent than shorter stimuli. Thus, the length by lexicality effect might also be explained by PDP models as a frequency by consistency interaction, with low- frequency inconsistent exemplars (i.e., long nonwords) being particularly difficult to read aloud. Our research addresses this ambiguity in the interpretation of the length by lexicality interaction by taking advantage of the special properties of the Japanese and Serbian writing systems. Critically, both of these writing systems are perfectly spelling-to-sound consistent, eliminating the confound between Results showed the predicted interaction between length and lexicality, F 1 (2,76)=13.16, p<.01; F 2 (2,53)=3.41, p<.05. The length effect was more than doubled when the same orthographic strings were interpreted as nonwords than when they were interpreted as words. Results showed the predicted interaction between length and lexicality, F 1 (3,54)=31.09, p<.01; F 2 (3,384)=18.55, p<.01. The length effect was 70% larger when the same phonological strings were presented as nonwords than when they were presented as words. Participants: 52 native Serbian speakers (26 in the Roman condition and 26 in the Cyrillic condition). Materials: 60 bivalent targets (30 Roman words; 30 Cyrillic words), that were 4, 5, or 6 letters long. Targets were presented with fillers containing unique Roman or Cyrillic letters to the two groups of participants in order to bias their interpretation. Procedure: Same as Experiment 1. Bivalent words comprise ambiguous and common letters, and have two possible pronunciations depending on the alphabet in which they are read. These words are meaningful in one alphabet but pronounceable nonwords in the other. Will the length effect be larger when bivalent stimuli are treated as nonwords than when they are treated as words? Participants: 21 native Japanese speakers. Materials: 208 Hiragana words and their Katakana transcriptions, 208 Katakana words and their Hiragana transcriptions. Stimuli were 3,4,5 or 6 morae. Procedure: Participants read words aloud as quickly and accurately as possible in Hiragana and Katakana blocks. RTs hand-marked via visual inspection of the speech waveform. Word Length Typical Script Transcribed Script 6 morae 689 ms 789 ms 5 morae 656 ms 745 ms 4 morae 615 ms 671 ms 3 morae 600 ms 638 ms Length Effect 89 ms 151 ms Word length Words Nonwords 6 letters 647 ms 806 ms 5 letters 622 ms 724 ms 4 letters 599 ms 679 ms Length Effect 48 ms 127 ms print Feature Analysis Orthographic input lexicon Phonological output lexicon speech Letter Analysis Phoneme system /p l O n/ GPC translation
The Cross-Script Length Effect: Evidence for Serial Processing in Reading Aloud Kathleen Rastle (Royal Holloway University of London), Linda Bayliss (Royal.
Jan 21, 2016
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Cross-Script Length Effect: Evidence for Serial Processing in Reading Aloud
Kathleen Rastle (Royal Holloway University of London), Linda Bayliss (Royal Holloway University of London), Jelena Havelka (University of Kent), Taeko Wydell (Brunel University)
Background
Reading aloud latency increases as the number of letters in a stimulus increases, but this increase is larger for nonwords (e.g., pon, plon, plont) than it is for words (e.g., pan, plan, plant). This length by lexicality interaction (Weekes, 1997) is one of the key findings used in support of dual-route theories of reading aloud that incorporate a serial nonlexical procedure (e.g., the DRC model; Coltheart et al., 2001). Nonwords in the DRC model are read primarily by the serially-operating nonlexical route, and that is why they show a particularly strong effect of length.
Experiment 1
Japanese Kana comprises Hiragana ( きんぎょ ) and Katakana ( キンギョ ) scripts, with a large number of spoken words only ever appearing in one script. These words could be transcribed into the other script and read aloud but they would be orthographically unfamiliar. Thus, a single spoken word can be presented as an orthographic word or as an orthographic nonword in Japanese Kana. Will the length by lexicality interaction be observed under these conditions?
Experiment 2
Serbian comprises Cyrillic and Roman alphabets, each of which has a set of unique letters that occur only in that alphabet. The alphabets share seven common letters that are pronounced identically in each, and four ambiguous letters that are pronounced differently in each.
Discussion:
•The length effect was larger in the Japanese study when stimuli were presented in an orthographically-unfamiliar form than when they were presented in an orthographically-familiar form, and in the Serbian study when stimuli were read as pronounceable nonwords than when they were read as words.•These writing systems are perfectly spelling-to-sound consistent, suggesting that the length by lexicality interaction is not the result of a frequency by consistency interaction.•Targets were very closely matched – phonologically identical in the Japanese study and orthographically identical in the Serbian study – making it unlikely that the effects were due to peripheral serial processes such as eye movements or articulation.•These data constitute the strongest evidence to date for serial nonlexical processing in reading aloud.
References:
•Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). The DRC model: A model of visual word recognition and reading aloud. Psychological Review, 108, 204-256.•Plaut, D.C., McClelland, J.L., Seidenberg, M.S., & Patterson, K. (1996). Understanding normal and impaired word reading: Computational principles in quasi-regular domains. Psychological Review, 103, 56-115. •Weekes, B.S. (1997). Differential effects of number of letters on word and nonword naming latency. Quarterly Journal of Experimental Psychology, 50A, 439-456.
The length by lexicality interaction has also been used to argue against PDP models of reading aloud (e.g., Plaut et al., 1996). In English, however, length is confounded with consistency, with longer stimuli typically being more inconsistent than shorter stimuli. Thus, the length by lexicality effect might also be explained by PDP models as a frequency by consistency interaction, with low-frequency inconsistent exemplars (i.e., long nonwords) being particularly difficult to read aloud.
Our research addresses this ambiguity in the interpretation of the length by lexicality interaction by taking advantage of the special properties of the Japanese and Serbian writing systems. Critically, both of these writing systems are perfectly spelling-to-sound consistent, eliminating the confound between length and consistency. If the length by lexicality interaction arises as a result of serial nonlexical processing, then it should persist in both of these languages.
Results showed the predicted interaction between length and lexicality, F1(2,76)=13.16, p<.01; F2(2,53)=3.41, p<.05. The length effect was more than doubled when the same orthographic strings were interpreted as nonwords than when they were interpreted as words.
Results showed the predicted interaction between length and lexicality, F1(3,54)=31.09, p<.01; F2(3,384)=18.55, p<.01. The length effect was 70% larger when the same phonological strings were presented as nonwords than when they were presented as words.
Participants: 52 native Serbian speakers (26 in the Roman condition and 26 in the Cyrillic condition).
Materials: 60 bivalent targets (30 Roman words; 30 Cyrillic words), that were 4, 5, or 6 letters long. Targets were presented with fillers containing unique Roman or Cyrillic letters to the two groups of participants in order to bias their interpretation.
Procedure: Same as Experiment 1.
Bivalent words comprise ambiguous and common letters, and have two possible pronunciations depending on the alphabet in which they are read. These words are meaningful in one alphabet but pronounceable nonwords in the other. Will the length effect be larger when bivalent stimuli are treated as nonwords than when they are treated as words?
Participants: 21 native Japanese speakers.Materials: 208 Hiragana words and their Katakana
transcriptions, 208 Katakana words and their Hiragana transcriptions. Stimuli were 3,4,5 or 6 morae.
Procedure: Participants read words aloud as quickly and accurately as possible in Hiragana and Katakana blocks. RTs hand-marked via visual inspection of the speech waveform.
Word Length Typical Script
Transcribed Script
6 morae 689 ms 789 ms
5 morae 656 ms 745 ms
4 morae 615 ms 671 ms
3 morae 600 ms 638 ms
Length Effect 89 ms 151 ms
Word length Words Nonwords
6 letters 647 ms 806 ms
5 letters 622 ms 724 ms
4 letters 599 ms 679 ms
Length Effect 48 ms 127 ms
Feature Analysis
Orthographic input lexicon
Phonological output lexicon
speech
Letter Analysis
Phoneme system
/p l O n/GPC translation
Related Documents
