The self-correction of speech errors (McCormick, O’Neill & Siskin)

From LearnLab
Revision as of 15:47, 11 January 2008 by Dawn E. McCormick (talk | contribs) (Findings)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

The self-correction of speech errors

PIs D. E. McCormick, M. C. O’Neill & C. B. Siskin
Graduate Student Researchers Year 2 Jessica Hogan and Mary Lou Vercellotti
Study Start Date September 2005
Study End Date August 2007
LearnLab English as a Second Language (ESL)
Number of Participants Spring 2006 (term 2064) - 94; Summer 2006 (term 2067) - 119; Fall 2006 (term 2071) - 128; Spring 2007 (term 2074) - 94; Summer 2007 (term 2077) - 95; Fall 2007 (term 2081) - 89
Number of Participant Hours (based on 5 hours per participant per term) Spring 2006 (term 2064) - 470; Summer 2006 (term 2067) - 595; Fall 2006 (term 2071) - 640; Spring 2007 (term 2074) - 470; Summer 2007 (term 2077) - 475; Fall 2007 (term 2081) - 445
Data in the Data Shop After editing to ensure anonymity, 2064 data (spring term 2006), 2067 data (summer 2006), and 2071 (fall term 2006) were made available to the Data Shop. Edited data of 2074 and 2077 are waiting to be exported to the Data Shop. Additional data from 2081 is in the process of being edited for anonymity.


The purposes of the second year of this project are: 1) to continue to examine and describe the extent of self-correction among English-as-a-second-language (ESL) students of varying proficiency levels during recorded speaking activities; and 2) determine if the self-correction process has an immediate impact on spoken production. Despite the myriad studies on corrective feedback in second language acquisition and second language pedagogy research literature, few studies address the issue of the learner as the source of feedback. In other words, the student is providing self error correction support. This study, in contrast, will focus on the learner as a primary source of corrective feedback. Based on classroom anecdotal evidence, we believe learners are an effective source of corrective feedback if provided with opportunity, time, and support.

Background and Significance

When we searched the ESL pedagogy literature on self-correction, we found few mentions of student self-correction as we are defining it. One related area that was mentioned was student self-assessment; in other words, students evaluating an aspect of their own second language performance. Genesee and Upshur (1996) suggest that self-assessment provides students with some of the same benefits for which we were looking, including an opportunity to share responsibility with teachers, an increase in individual feedback and learner involvement, and an expansion of the basis of evaluation by the teacher. Lazaraton (2001) mentions “one of the most recent trends in oral skills pedagogy is the emphasis on having students analyze and evaluate the language that they or others produce” (p. 108). She continues to argue that this analysis and evaluation helps students become more “metalinguistically aware” of the language they need to interact in the target language. Although we were not interested in student self-assessment at this time, we believed that the issues raised by Genessee and Upshur (1996) and Lazaraton (2001) related well to student self-correction in that both self-assessment and self-correction involve students in a “meta” activity and increase student involvement in general.

In a discussion about providing pronunciation feedback, Goodwin (2001) mentions self-correction, first in reference to students’ correcting an error after the error is cued by the teacher, and later in reference to students’ monitoring a recording of their speech. Goodwin (2001) states that: "A second way to encourage self-monitoring is to record student speech, in either audio or video format. This is particularly effective if the learners’ first task is to transcribe their speech (not phonetically, just regular orthography). Working with their transcript while listening to their tape, learners can monitor for a specific feature" (p. 130). Our project, then, utilizes Goodwin’s idea, although in a broader context.

Within the research literature, studies by Golonka (2000) and Shehadeh (2001) provide examples of research that involved aspects of student self-correction. In her dissertation, Golonka (2000) investigated variables that predicted oral proficiency gains in 22 adult learners of Russian who studied abroad. One of the categories of variables was “metalinguistic variables,” which included various self-correction and sentence repairs. Within the three groups, nullgainers (those who did not achieve a measurable gain in proficiency after their study abroad experience), gainers (those who progressed from Intermediate High to Advanced on the ACTFL OPI scale after their study abroad experience), and high gainers (those who progressed from Intermediate High to Advanced High on the ACTFL OPI scale after their study abroad experience), differed in regard to self-correction.

Within the first analysis (to determine the linguistic and metalinguistic variables in the pre-immersion OPI impact proficiency development) nullgainers corrected themselves an average of 3.7 times as compared to eight times by gainers during the OPI interviews. In the second analysis (which focuses on the metalinguistic awareness to determine the differences between gainers and high gainers), nullgainers self-corrected an average of 3.7 times, gainers self-corrected an average of 6.2 times, and high gainers an average of 9.8 times, with each measure significant at p = .005. The self-corrections mentioned in Golonka’s 2000 study refer to sentence repair in which the participant made changes in sentences because the participant decided to change the utterance patterns or because something in the utterance necessitated a change (e.g., lacking a vocabulary item). Within all variables studied and in both analyses, self-correction was one of the variables distinguished as significant for discriminating between nullgainers, gainers, and high gainers. The researcher argues for a connection between self-correction and higher L2 gain and that the result “implies that self-correcting behaviors are positive and desirable and should be encouraged from a pedagogical and self-instructional viewpoint” (p. 113). In our case, the focus on self-correction within the recorded speaking activity may serve as tool to develop self-correction skills that can be used in contexts outside of the classroom.

Shehadeh (2001) examined the relationship between self- and other- initiations and modified output. Modified output was defined as “the modifications non-native speakers (NNS make to their output in order to make an initial utterance or part of an utterance more accurate or more comprehensible in response to: (a) other-initiation or (b) self-initiation” (p. 437). Other-initiations occurred when the listener signaled his or her total or partial lack of understanding of the speaker’s “trouble-source” utterance (p. 455). The signal was an explicit verbal utterance, e.g., repeating the problematic item, repeating part of the problematic item with a question word, asking a direct question, or stating something like “ I don’t understand.” Self-initiation occurred when the speaker noticed the listener’s lack of comprehension. In the first case the listener reacted to the trouble-source first, while in the second case the speakers reacted first.

The non-native speaker participants involved 35 adults of various first language backgrounds who were learning English as a second language in the United Kingdom. The native speakers (NS) involved were university teachers and postgraduate students. The three tasks included: a) a picture description task in which a NNSs described a picture for a NS or a NNS to reproduce, b) an opinion exchange task in which the participants (NS/NNS and NNS/NNS pairs) discussed a newspaper article, and c) a group decision-making task in which the NNSs created a constitution for an invented country. Although self- and other-initiations prompted modified output, more modified output was produced from self-initiations across four out of five interactional contexts. Shehadeh (2001) argues that the results indicate the need for more interactional contexts within classroom practice. In addition, the researcher argues that because of the amount of modified output due to self-initiations, “learners need both time and opportunity for self-initiated, self-completed repair of their messages” (p. 451). We believe that the classroom recorded speaking activities in which students are given an opportunity engage self-correction, in the ELI are one way to provide students with the time and opportunity to modify their own output.




Research question

Two research questions for this project have a descriptive purpose:

  • Research Questions 1: What errors are students able to identify within given proficiency levels (low-intermediate, high-intermediate, advanced)? across levels?
  • Research Question 2: Of the errors identified by students, what errors are they able to self-correct within given proficiency levels (low-intermediate, high-intermediate, advanced)? across levels?

An additional research question has been added for the second year:

  • Research Question 3: Can students use a self-correction learning event to immediately improve spoken production accuracy?

Independent variables

  • Primary: Language Proficiency (low-intermediate, high-intermediate, advanced)
  • Secondary: First language (L1), gender, prior language learning experiences, topic, etc.


  • Hypothesis 1: Students will be able to identify errors, though the ability will vary across proficiency levels.
  • Hypothesis 2: Students will be able to self-correct errors, though the ability will vary across proficiency levels.
  • Hypothesis 3: Students can use a self-correction learning event to immediately improve spoken production accuracy, though ability will vary across proficiency levels.

Dependent variables

  • Identification/correction of errors
  • Accuracy
  • Path choice


We have just completed data collection and and working on analyses. However, we would like to share some general observations:

First, all proficiency levels have been able to identify and self-correct errors to some degree.

Second, students are able to identify and self-correct errors in the areas of grammar, vocabulary, and pronunciation, with grammar being the area with the most frequent corrections.

In a sample of data from 16 students, students in all three levels identified and corrected both program target and non-target grammar structures. The majority of errors identified and corrected were related to noun and verb errors, and the majority of errors identified and corrected related to program target grammar structures. In addition, the data revealed that the students tended to be fairly accurate in their corrections, with level 3 at 79% accuracy, level 4 at 82% accuracy, and level 5 at 87.4% accuracy. Below are some examples of correct corrections (original > correction):

Level 3: I play and do anything with my friend > friends

Level 4: it’s nice sport > it’s a nice sport

Level 5: We always shopping food in the open market > we always shop for food in the open market

Third, viewing the self-correction process as a potential learning event space has helped us realize that what constitutes “correction” is a complex process. Data show that students follow these possible path choices:

1. Identify a correct form as correct (considering its orthographic and/or audio forms); therefore, no alternative form provided.
2.Identify a correct form as incorrect (considering its orthographic and/or audio forms).
2.1. provide a correct alternative
2.2. provide a partially correct alternative
2.3. provide an incorrect alternative
2.4. state that the student cannot provide an alternative
3. Identify an incorrect form as incorrect (considering its orthographic and/or audio forms)
3.1. provide a correct alternative
3.2. provide a partially correct alternative
3.3. provide an incorrect alternative
3.4. state that the student cannot provide an alternative
4. Not identify an incorrect form as incorrect; therefore, no alternative form provided.

The paths taken by the student produce different path effects, basically: a) exiting the path with learning to some degree or 2) exiting the path with mislearning to some degree. Path Choices 3.1 and 3.2 would indicate the transfer of knowledge components in that students have used language knowledge components to identify and correct errors. Also, the student’s path choice reflects sense making in that the students’ error-identification and correction process is an indication of their trying to engage in higher-level thinking.


Annotated bibliography

Presentation to the PSLC Advisory Board, December, 2006


Genesee, F., & Upshur, J. A. (1996). Classroom-based evaluation in second language education. New York: Cambridge University Press.

Golonka, E. M. (2000). Identification of salient linguistic and metalinguistic variables in the prediction of oral proficiency gain at the advanced-level threshold among adult learners of Russian. Unpublished dissertation, Bryn Mawr College, Bryn Mawr, Pennsylvania, USA.

Goodwin, J. (2001). Teaching pronunciation. In M. Celce-Murcia (Ed.), Teaching English as a second or foreign language (3rd ed.) (pp. 117-137). US: Heinle & Heinle.

Lazaraton, A. (2001). Teaching oral skills. In M. Celce-Murcia (Ed.), Teaching English as a second or foreign language (3rd ed.) (pp. 103-115). US: Heinle & Heinle.

Shehadeh, A. (2001). Self- and other-initiated modified output during task-based interaction. TESOL Quarterly, 35 (3), 433-457.

Further Information


McCormick, D. E., O'Neill, M. C., & Smith, D. (in review). Using focused practice and self-correction to develop grammatical ability. Language Learning and Technology.

Peer Reviewed Conference Presentations

April 2007: "Self-correction and fluency in ESL speaking development," Nel de Jong, Dawn E. McCormick, M. Christine O'Neill, Claire Bradin Siskin; part of a colloquium, "CALL contributions to SLA", American Association for Applied Linguistics 2007 Conference in Costa Mesa, California.

May 2006: "Serving three mistresses in CALL: Students, teachers, researchers," Dawn E. McCormick, M. Christine O'Neill, and Claire Bradin Siskin), CALICO Symposium, Honolulu.

Other Presentations

September 2007: Presentation of the software component at the “Multimedia Showcase” sponsored by the Robert Henderson Media Center at the University of Pittsburgh.

September 2006: Presentation of the software component at the “Multimedia Showcase” sponsored by the Robert Henderson Media Center at the University of Pittsburgh.

September 2005: Presentation of the software component at the “Multimedia Showcase” sponsored by the Robert Henderson Media Center at the University of Pittsburgh.

The following presentations by Claire Bradin Siskin are associated with this project in that her work with the software program Revolution, the software program used to design the RSA, has been a major impetus for other projects and the presentations listed below:

November 2007: "Designing Your Own CALL-Based Proficiency Test" (paper, with Elizabeth Wylie-Ernst, Brett Wells, Sarah Williams) ACTFL Conference, San Antonio.

October 2007: "Designing Your Own CALL-Based Proficiency Test," (paper, with Elizabeth Wylie-Ernst, Beatrice DeAngelis, Brett Wells Sarah Williams), PSMLA Conference, King of Prussia.

June 2007: "Revolution for the Next Revolution in Managing MediaI," ALLT Conference, Boston, MA.

May 2007: "Revolution for Low-cost Data Collection in CALL,"CALICO Symposium, San Marcos, TX.

March, 2007: "A Lab Director's Perspective on Assessing Speaking Proficiency," NEALLT Conference, Ithaca, NY.

March 2007: “Misconceptions, myths, and metaphors in CALL research,” TESOL: CALL IS Academic Session.

May 2006: "Revolution Templates for Language Learning" (Courseware Showcase), CALICO Symposium, Honolulu.

May 2006: "Rapid Creation of Internet-based Multimedia Applications Without Web Browser Hassles" (paper, with Devin Asay), CALICO Symposium, Honolulu.

April 2006: "Revolution for Non-Programmers, or Yes, There Is Life After HyperCard!" (paper), NEALLT Conference, Philadelphia.

March 2006: "Comes the Revolution: Templates for Interactive CALL Materials" (Developer's Showcase), TESOL Conference, Tampa.

Plans for future:

- Continue transcription.

- Continue data analysis.


<insert here>