Zhao & MacWhinney - Learning the English Article
English Article Usage
Documentation of this study is currently in progress.
Background and Significance
The current project focuses on the development of a cognitive tutoring system for the teaching of English articles – one of the most difficult grammatical forms for second language learners to learn and master. Articles are particularly difficult for learners whose first language (e.g., Chinese and Japanese) does not use articles. There are three factors that make this a difficult target structure: (1) there are dozens of difficult and conflicting rules determining article choice; (2) misuses of the articles usually do not cause miscommunication and therefore learners tend to ignore these errors; and (3) classroom instruction does not provide enough opportunities for learning many of the functions and cues that determine article choice. Cognitive tutoring systems can provide address each of these problems by giving simple illustrations of relevant cues, providing consistent feedback, and sampling across a wide range of genre types and usages.
The research goal of the article tutor project is to promote robust learning and mastery of the English articles among Chinese EFL learners illuminated by principles from: (1) Experimental Psychology: Practices make perfect; Feedback promotes learning; (2) Developmental Psycholinguistics: Language is learned in context; Cue conflicts are crucial for learning; (3) Human-Computer Interaction: rule-based and exemplar-based instruction promotes learning in different ways; and (4) Second Language Acquisition: explicit types of instruction is in general more effective than implicit types of instruction; accurate metalinguistic knowledge representation is important. Synthesizing the above principles, the Cognitive Article Tutor designs exercise with nine genres of texts with rich article usages and provides explicit instruction in the form of explicit feedback.
Explicit versus implicit instruction:
There is a major distinction between explicit and implicit instruction in second language teaching and learning. This distinction is often operationalized in terms of explicit and implicit feedback given to students in the instructional settings. Following Dekeyser (1995), explicit instruction consists of explicit deduction (explicit rule presentation) or explicit induction (instructions to orient learner attention to forms or to induce metalinguistic hypotheses); implicit instruction indicated that no explicit rule statement took place in the treatment and no instructions attending to particular forms or formulating metalinguistic hypothesis were given to learners. Norris & Ortega (2000) did a meta-analysis study and examined the effectiveness of instruction methods in different instructional settings. They concluded that, in general, explicit types of instruction are more effective than implicit types of instruction.
Similar to the general findings of L2 instructional studies, the available intelligent computer assisted language learning studies also suggested that explicit feedback is superior to implicit feedback especially when the learning task involves relatively complex structures whose grammatical rules are not salient in light of the examples. The most effective iCALL feedback is to “to respond to errors by giving a metalinguistic explanation in the form of a rule” (Hanson, p. 49) This general finding gives strength to the potential benefit of using cognitive tutor to teach the English articles, which is a complex and non-salient grammatical category.
Does the Cognitive Article Tutor that provides practice with corresponding explicit feedback increase L2 learners' performance of article usage in written production?
The Cognitive Article Tutor that provides practice with corresponding explicit feedback helps to increase L2 learners' performance of article usage in written production.
The independent variable of the current study is the explicit feedback provided by the Cognitive Article Tutor.
The explicit feedback of each grammatical rule is associated with one type of usage of English articles. Each explicit feedback is composed of three parts: (1) one grammatical rule name, (2) an explanation of the rule, and (3) several examples to further explain the rule. Whenever a learner make a mistake with one article choice, the tutor automatically provides explicit feedback composed of the above three levels of explanation.
For example, one rule of the English articles is named "Non-count Abstract Noun". This is a rule associated with the zero article. The rule explanation describes as follows: "The article should be omitted when referring to a non-count abstract concept, emotion, or principle, even if this noun is modified by a preceding adjective". Following that, several examples are given to further explain the rule: (a) Prudence is the better part of valor, (b) Friction tends to resist gravity, (c) Statistical analysis could clear up the issue, and (d) I strive for clarity in my prose.
The dependence variable of the current study is learners' performance of article usage in written production.
Norris & Ortega (2000) identified four general types of measurements in SLA studies testing the effect of explicit and/or implicit instruction: (a) metalinguistic judgments if the research participant was required to evaluate the appropriacy or grammaticality of L2 target structures as used in item prompts (e.g., grammaticality judgment tasks); (b) Selected response measures required participants to choose the correct response from a range of alternatives, typically either in answer to comprehension questions based on the use of the target L2 form(s) or in order to complete a sample segment of the target language with the appropriate target form(s) (e.g., multiple choice tests providing four options in verbal morphology); (c) constrained constructed response if they required the participant to produce the target form(s) under highly regulated circumstances, where the use of the appropriate form was essential for grammatical accuracy to occur. Constrained constructed response measures required learners to produce L2 segments ranging in length from a single word up to a full sentence, but all such measures were designed with the intent to test L2 ability to use the particular form within a highly controlled linguistic context (e.g., sentence combining with relative pronouns); (d) free constructed response measures were those measures that required participants to produce language with relatively few constraints and with meaningful communication as the goal for L2 production (e.g., oral interviews, written compositions).
Norris & Ortega (2000) suggested that the type of outcome measures used in individual studies likely affects the magnitude of observed instructional effectiveness. Average effect sizes associated with metalinguistic judgments and free constructed response measures were substantially lower than those associated with selected-response or constrained constructed-response measures. Thus, study findings within the research domain may vary by as much as 0.91 standard deviation units depending on the type of outcome measure or measures employed. Therefore Norris & Ortega suggested researchers to triangulate outcome measures in order to overcome the possible bias of particular measurement that is more likely to produce larger effect size.
To guarantee triangulation of outcome measures, the present study makes use of four outcome measures: (1) Untimed Grammaticality judgment test (~10min); (2) Untimed article choice test (~10min); (3) Timed free writing task (15min); (4) Untimed article rule explanation task (~5min).
Untimed grammaticality judgment test (GJT) is a metalinguistic judgment test, which allows us to investigate the explicit knowledge representation of learners’ acquisition of articles. In the untimed GJT, participants are asked to judge whether a sentence is grammatical or not. Both grammatical and ungrammatical sentences examining English articles as well as other grammatical categories (past tense, subjunctive mood, relative clause, third person singular) are included in the GJT as control items. Untimed article choice test is paragraph level cloze test which requires participants to fill in all the articles in the given paragraphs. Timed free writing task is to ask participants to write as much as they can within 15 minutes. The participants are given picture prompts for the writing task. The untimed article rule explanation task is to give participants sentences with correct use of English articles and to ask participants to choose from a pool of four article rule explanations to explain what are the target rules in the given sentences. Except for the free writing task, the other three tasks are graded based on the correct responses that the participants supply. The free writing task is graded with all the noun phrases identified and judged by native speakers for accuracy of usage. Each piece of writing will produce one mean accuracy rate of article usage.
We ran a pilot study among 60 non-English major Chinese learners of English in a Chinese university in Beijing in July 2010. The purpose of the pilot study was to test the effectiveness and accessibility of the instructional materials that are going to be used in the Cognitive Article Tutor. We performed the same pre-test, immediate post-test and delayed post-test with the 60 participants. We used the instruction materials to teach the English article rules to 40 learners. The rest 20 participants received English pronunciation training during the time when the 40 participants were receiving English article instruction. At the end of the article instruction, the 40 participants were asked to fill in a questionnaire about their feedback of the article instruction. Right now we are in the process of doing data analysis and synthesizing questionnaire feedback from the 40 learners who received article instruction so that we can improve the Cognitive Article tutor instructional materials.