Task is defined in various ways. Nunan (2004) offers his own definition: [A] pedagogic task is a piece of classroom work that involves the learner in comprehending, manipulating, producing or interacting in the target language while their attention is focused on mobilizing their grammatical knowledge in order to express meaning, and in which the intention is to convey meaning rather than to manipulate form. The task should also have a sense of completeness, being able to stand alone as a communicative act in its own right with a beginning, middle and an end (p. 4). Skehan (1996a) states that task is an activity in which: meaning is prior to form; there is source sort of relationship to the real world; tasks are problem-oriented; the outcome is the index for assessment of task performance. Ellis (2003) pointed out that Skehan’s definition was made from a researcher’s point of view rather than from the learners’ point of view. Nunan (1989) regarded the communicative task as an example of meaning-based activities, which involved learners in using the target language to serve different communicative purposes. So it is clear that Nunan’s (1989) definition of task stems largely from communicative language teaching. Willis points out that “tasks are activities in which learners use the target language for a communicative purpose (goal) to achieve an outcome” (Willis, 1996a: p. 23) . In other words, tasks are the activities in which learners use whatever target language resources they have in order to solve a problem, do a puzzle, play a game or share and compare experiences (ibid.). Willis’s (1996a) definition of tasks is adopted in this study.

A lot of researchers were interested in task-based language teaching and learning approach in the world.

Robertson, Margaret (2014) explained the congruence between TBLT and expansive learning theory and the benefits of doing so. Huang, Jiuhan (2010) outlined the 10 principles of instructed language learning formulated by Ellis and showed how using Willis’ Task-Based Learning Framework in grammar instruction for adults responses to many of the 10 principles.

In China, there were some researchers interested in the task-based language teaching and learning approach. Zhaochun Sun (2015) reported a tentative study on the task-based teaching of writing to English majors in Chinese settings and Sun found the findings of this research reveal that the application of TBL framework to English writing classrooms in Chinese EFL settings is effective in improving English majors’ writing competence and performance. Ji (2014) reported effects of a task-based approach to public speaking instruction and concluded that task-based instruction supported learner-initiated learning opportunities outside the classroom, thus encouraging greater learner improvements.

Over the past decade or so, various frameworks for implementing TBLT in the classroom have been proposed (Ellis, 2003; Skehan, 1996a, 1996b; Willis, 1996a) . Although these frameworks vary a bit in their terms, they all share three principal phases named pre-task, while task and post task. This study adopted Willis’s (1996a) framework, which is showed in Table 1, for implementing TBLT in experimental group.

According to Skehan (1996a) , “3Ps” approach is regarded as one of the comparative approach towards TBLT. It gains people’s attention when they discuss the implementation of TBLT. The present study attempts to find out the effect of TBLT in comparison with that of 3Ps on learners’ oral performance. Thus, this study used the framework of 3Ps approach (Baker & Westrup, 2003) , illustrated in Table 2, in control group.

According to the related studies, TBLT and 3Ps differ mainly in two ways.

First of all, there is different order of form and meaning in each framework. Willis (1996b) argued that in 3Ps cycle, focus on form comes first and focus on meaning comes later. On the contrary, in the TBLT framework, the order is reversed. What’s more, the views on two pedagogues’ goals namely accuracy and fluency, are different. Willis concluded that “3Ps is based on the assumption that out of accuracy comes fluency. This methodology entails a good deal of language control on the part of the teacher in the interest of accuracy” (1996b:45). However, Willis (1996b) explained that TBLT is based on the assumption that out of fluency comes accuracy. The implementation of this methodology is through a deep-end strategy whereby learners do their best to achieve some kind of resolution to a communicative problem through the effective use of their existing language resources.

Most of the relevant empirical studies into tasks lie in the three pedagogic goals for task-based approaches: accuracy, complexity and fluency. It is Skehan’s (1996a, 1996b, 1998) three-aspect distinction between fluency, accuracy and complexity of linguistic performance that form the basis for measuring learners’ language production in task-based related research. Skehan (1998) listed three areas as follows: a) meaning which is reflected in fluency; b) form-control, reflected in accuracy; c) form-restructuring which is related to the complexity of the language. Skehan concluded that three pedagogicgoals “are also effective indices for measuring performance on a particular task and, in this way, provide a method for assessing the worth of different tasks” (1998: 270).

Task performance provides clues to discover the function of tasks in promoting language acquisition as well as the challenge a task imposes on learners. It can be observed both qualitatively and quantitatively. Qualitative observation concerns general tendencies of linguistic performance to which a task may be related. The quantitative observation mainly concerns the commonly observed aspects of linguistic performance, such as accuracy, fluency, complexity and appropriacy. Besides the qualitative nature of task related performances, many researchers are concerned with task-related performance in the form of linguistic complexity, accuracy and fluency. Skehan (1996a, 1998) argued that these measures are considered theoretically related to the development of

specific aspect of language ability oracquisition process, so they are readily used to describe the status of acquisition.

In September 2014, 65 first-year non-English-majored graduates in 2 classes from Yangtze University participated in this study. Among the 65 subjects, 33 were females and 32 males, average age 22 with Chinese as the main language. Their majors were chemistry, agriculture, plant protection, finance, biological technology. All 65 students who had passed CET4were divided into Experimental group (EG = 33) and Control group (CG = 32) randomly. Their education background, family background, age, and other factors were same, that was to say, their overall learning and cognitive abilities were almost equal. Both EG and CG were taught by the same teacher. Before the experiment, the independent samples t-tests were used to explore whether non-English-ma- jored graduates in EG or CG generated significant differences in oral English performance (Table 3). The survey found that before the experiment, EG and CG showed no significant difference in oral English performance from Accuracy (Task One (t = .321, P = .825), Task Two (t = .417, P = .543)) to Complexity (Task One (t = 1.352, P = .214), Task Two (t = 1.423, P = .143) (Table 3).

The study was carried out over a period of one semester (September, 2014 to January, 2015). At the beginning of the first semester in the first academic year for non-English-majored graduates’ English course, the TBLT methodology in the Experimental group (EG) and 3Ps methodology in the Control group (CG) were implemented by the same teacher who was male, with 15 years’ teaching experience. In each group, the teacher and students used the same textbook, that was, Listening and Speaking (published by Foreign Language Teaching and Research Press). Every week, non-English-majored graduates in EG were taught for two hours with the TBLT methodology and non-English-majored graduates in CG were taught for two hours with 3Ps methodology. The experiment lasted for 18 weeks.

The instruments used in the study were measures (accuracy, fluency and complexity), two tasks and interviews.

Measures in this study were three: accuracy, fluency and complexity.

According to Skehan (1996a) , accuracy was the “freedom from error”. It was measured in terms of the number of error per 100 words. Mehnert (1998) pointed out that this measure was more sensitive than error-free clauses with learners of relatively low proficiency.

As Skehan notes, “fluency concerns the learners’ capacity to mobilize an inter-language system to communicate meanings in real time” (1996a: 46). According to Ellis (2003) , the fluency can be measured in terms of the number of pauses(i.e. a break of 2.0 second or longer either within a turn or between turns); the number of repetitions (i.e. immediate and verbatim repetition of a word or phrase); reformulations (phrases or clauses that were repeated with some modification either to syntax, morphology, or word order); replacements (i.e. lexical items that are immediately substituted for another) and number of words per minute(the ratio of total number of words to the time that the particular candidate’s spend on the utterance).

Skehan suggested that “Complexity, and its attendant process, restructuring, relates to the stage and elaboration of the underlying inter-language system” (Skehan, 1996a: p. 46) . Complexity is often measured in terms of c-unit (Foster & Skehan, 1996) and t-unit (Foster et al., 2000) . But in this study, the author attempted to apply the indicator of words per AS-unit (the Analysis of Speech Unit) (Foster et al., 2000) as the measure of complexity. Foster et al. (2000) pointed out that, in counting the As-unit, some words should be excluded since the inclusion of these data could distort the perception of the nature of the performance.

According to Foster et al. (2000) , in this study, when we counted the As-unit, we should exclude one-word minor utterances in the following situation: “Okay, Uhuh, Right, Umm”; when the repetition, replacement or reformulation occurs, the final version was counted as As-unit.

Two tasks used in the study were a problem-solving task and a narrative task, both are required to carry out in English. The choice of these two tasks was based on an analysis of tasks commonly used in current English language textbooks (Listening and Speaking published by Foreign Language Teaching and Research Press, 2010). The first task is a role-play. It belongs to a problem-solving task suggested by Willis’s (1996a) task type and similar to the personal task used in the research by Foster and Skehan (1996) . The second task is a narrative task. It is similar to the task used in the research by Foster and Skehan (1996), Skehan and Foster (1999) and Mehnert (1998) .

Task One: A Tour to Beijing (Unit 2 taken from Listening and Speaking)

Yang is a local tour guide in Beijing. While accompanying her guests to tour round the city of Beijing, she is introducing the top attractions to them. The day’s arrangements are as follows. Refer to the information below and have a discussion about the arrangements. One of you is Yang, and the others play the role of the tourists.

Task Two: Describe your own apartment to your partner (Unit 3 taken from Listening and Speaking)

Imagine you have your own apartment (Table 4). Describe it to your partner, who will draw a picture of it with precise ratios of length, width and height.

Interviews: All 33 non-English-majored graduates in EG were interviewed about the TBLT methodology.

Before the research experiment (September 2014) and after the experiment (February 2015) two tasks were conducted to compare changes between the two groups of non-English-majored graduates in the TBLT methodology and the 3Ps methodology. In this study, social scientific software SPSS 17.0 was used for statistical analysis to the collected data. In order to find out whether TBLT methodology were superior to 3Psmethodology, comparison of Means was adopted to compare two groups of non-English-majored graduate students’ average scores of their pretest and after the experiment on the basis of two samples. And the independent sample T-test was adopted to exanimate if there were significances between EG and CG before the experiment and after the experiment.

Independent samples t-tests were used to find out whether any of the instruction conditions (TBLT, 3Ps) generated significant results

As shown in Table 5, we could know the results from independent samples t-tests on the instruction conditions for two tasks between EG and CG. On Task One, only Repetition and the words per AS-unit, which are respectively measures of fluency and complexity, attained significance at the .015 level and at the .01 level, respectively. As for Task Two, Number of errors per 100 words, Replacements, Number of words per minute and Words per AS-unit, these four variables of measurement attained significance at the .05 level.The descriptive statistics on the measures of accuracy, fluency and complexity was showed in Tables 6-8 respectively.

M stands for Mean; S stands for Standard Deviation.

M stands for Mean; S stands for Standard Deviation.