Investigating the role of dynamic assessment on speaking accuracy and fluency of pre-intermediate EFL learners

Abstract The current study set out to investigate the effect of dynamic assessment (DA) on the speaking accuracy and fluency of pre-intermediate English as a foreign language (EFL) learners. In contrast to static assessment, DA is conceptualized as an interactive approach towards assessment which combines teaching and testing in a unitary instructional intervention. To address the objectives of the present study, initially Preliminary English Test (PET) was given to 93 learners and 62 learners whose scores fell within ± one standard deviation from the mean were selected. The 62 selected learners were divided into an experimental and a control group. Prior to the administration of the treatment, both groups were given a speaking pretest and their fluency and accuracy scores were calculated. Following that, the dynamic assessment group received treatment for grammar and vocabulary in line with an established framework in the literature in which a stepwise mediation from the most implicit to the most explicit feedback shapes the foundation of DA. As for the control group, the participants received the same content in terms of vocabulary and grammar but there was no step-wise mediation. At the end of the treatment, both groups were given a speaking posttest and the scores were used to address the research questions. Moreover, five of the participants in the experimental group were interviewed to seek their perceptions towards the effectiveness of dynamic assessment for their speaking accuracy and fluency. The results of ANCOVA indicated that dynamic assessment significantly impacted the speaking accuracy of the participants; however, DA did not significantly improve the participants’ speaking fluency. The results of interviews also revealed that overall the participants had positive perceptions towards the effectiveness of dynamic assessment for improving their speaking accuracy.


PUBLIC INTEREST STATEMENT
Speaking accurately and fluently is very important in learning a language. Moreover, assessment has also been an important factor in language learning. The present study aimed to investigate the effect of dynamic assessment on the speaking accuracy and fluency of pre-intermediate EFL learners. To do so, a step-wise mediation process was adopted in which the instructor just introduces hints and the learner could finally come up with the correct from. The findings revealed that dynamic assessment significantly impacted the speaking accuracy of the participants but not speaking fluency. Moreover, some of the participants were also interviewed. The results of interviews showed that the participants had positive perceptions towards the effectiveness of dynamic assessment for improving their speaking accuracy. The findings are significant because results call for teachers' and materials developers' attention and consideration to tailored mediation to help students move toward self-regulation. experimental group were interviewed to seek their perceptions towards the effectiveness of dynamic assessment for their speaking accuracy and fluency. The results of ANCOVA indicated that dynamic assessment significantly impacted the speaking accuracy of the participants; however, DA did not significantly improve the participants' speaking fluency. The results of interviews also revealed that overall the participants had positive perceptions towards the effectiveness of dynamic assessment for improving their speaking accuracy.

Dynamic assessment
Assessment is an important component in all educational settings in general and the EFL context in particular. As Beaumont et al. (2011) maintain assessment has long been regarded as an important factor which contributes to both teaching and learning. Many researchers in English Language Teaching (ELT) contexts (e.g. Antón, 2003Antón, , 2009Barzegar & Azarizad, 2014;Poehner, 2008;Yildirim, 2008) believe that assessment has an important role in stimulating the learning process and can help learners improve their learning. One type of assessment is dynamic assessment. DA is an alternative assessment which is process-oriented and increases learners' responsibilities for their own learning (Crick & Yu, 2008). In this type of assessment, the interaction between teacher and learner has an important role and this interaction captures the distance between the existing knowledge and the potential knowledge learners are capable of acquiring (Regan, 1998). DA is conceptualized as an interactive approach that combines teaching and testing in a unitary instructional intervention with the purpose of enhancing learning via giving suitable kinds of mediation in the form of hints and prompts. Overall, DA builds upon the interaction between a language practitioner and the L2 learners. During this dynamic interaction, the practitioner acts as an intervener in order to give the learners who are the examinees the necessary scaffolding to perform the tasks. As Elliot (2003, p. 16) noted, the "theoretical forefather" of DA is Vygotsky whose notion of Zone of Proximal Development (ZPD) is a key construct in DA. "In DA, teachers act as promoters and provide immediate feedback to whole procedures of a task" (Xiaoxiao & Yan, 2010, p. 25). Xiaoxiao and Yan (2010) further maintain that DA is a meditational and planned teaching in which the focus is on learners' future development, not the outcome of the past development.
According to Murphy (2008) and Poehner (2008), dynamic assessment originates in the Vygotsky's construct of the ZPD put forth within the sociocultural theory. In practical terms, ZPD is defined as the difference between what an individual can accomplish without any assistance and what he/she can achieve by being aided by others, both in assessment and in classroom learning situations (Kozulin et al., 2003). The zone in ZPD is characterized as the gap between what a person is able to do by himself/herself and what he/she can do through collaboration with a more expert one (Daniels, 2001). Moreover, in the view of Chaiklin (2003), ZPD refers to interaction between a more competent individual and a less knowledgeable one on a task, with the collaboration leading to the progress of the less competent one. The very notion of intervention or mediation involved in assessment significantly makes a distinction between dynamic assessments from the so-called traditional or static assessment. As discussed by Haywood and Lidz (2007), mediation is the method used by good teachers and parents, resulting in the promotion of high levels of mental functioning among their children and learners. In fact, the mediation is a reflection of Vygotsky's ideas regarding the instruction within the zone of proximal development, serving as a guide for the assessor in making instructional decisions. This is made possible through the analysis of the student, the text, and the type as well as the extent of mediation the individual should be provided with (Vygotsky, as cited in Mardani & Tavakoli, 2011).
It is possible to implement the mediation process through multiple methods. As mentioned by Lantolf and Poehner (2004), in the literature two primary approaches can be identified to DA, namely, Interactionist and Interventionist. Both consist of three stages: pre-test, mediation, and then post-test. These two approaches differ in terms of how the mediation happens during the course of assessment. In the same vein, Poehner (2008), maintains that Interactionist DA originates in Vygotsky's ZPD where assistance is given amid the interaction between the assessor and the examinee given the learners ZPD. It puts emphasis on the growth of person as a learner or a group of learners (Lantolf & Poehner, 2004). As for interventionist approaches, the tasks and materials are procured with the aim of pinpointing the examinee's problems faced by the learners during the assessment. In this approach, mediation takes the form of hints, prompts, and leading implicit/explicit questions. The mediator proceeds based on the hints, moving from hint to hint and the learner picks the answer he/she thinks is the exact one (Lantolf & Poehner, 2011). There are two subcategories in the Interventionist DA: 1-the "sandwich" and 2-the "cake" formats (Sternberg & Grigorenko, 2002). In the sandwich format, the learner receives instruction at once between the pre-test and the post-test, while in the cake format, the instruction unfolds in graded layers following each test item if necessary. The main difference between these two formats is that instruction and assessment are distinct in the sandwich format while combined in the cake format DA (Wang, 2010).

Dynamic assessment and speaking performance
As a plausible justification for the investigation of the use of DA for L2 speaking, it might be argued that speaking is usually considered as one of the most important skills by L2 learners who may consider L2 speaking to be equal to success in acquiring the language (Richards, 2008). However, the ability to speak in L2 is a very complicated and multi-dimensional process for pre-intermediate EFL learners who do not have adequate opportunity to produce output (Bygate, 2009). Preintermediate or not very proficient EFL learners are not provided with the opportunity to revise and edit the output as speaking is usually transient, unpredictable and occurs in real time (Bailey, 2006). When it comes to assessment, all language skills and components are of importance and speaking is no exception. Due to its importance, speaking has been subject to numerous studies recently (e.g. Marashi & Dolatdoost, 2016;Wahyurianto, 2018;Yufrizal, 2018). As Marashi and Dolatdoost (2016) maintain, speaking is a significant skill in second language learning since the ability to communicate in a foreign language lies at the heart of foreign language learning. The two main components of speaking are fluency and accuracy. Fluency is "the ability to keep going when speaking spontaneously with all available linguistic resources and regardless of grammatical mistakes" (Gower et al., 2005, p. 100). Ellis and Barkhuizen (2005, p. 139) consider fluency as "the production of language in real time without undue pausing or hesitation." Due to the importance of speaking fluency, it has been investigated by many researchers (e. g., Syamdianita et al., 2018;Thomson, 2018;Wahyurianto, 2018;Yufrizal, 2018). Accuracy is "the extent to which the language produced in performing a task conforms to target language norms" (Ellis, 2003, p. 339). Ellis and Barkhuizen (2005, p. 139) define accuracy as "how well the target language is produced in relation to the rule system of the target language". Similar to fluency, accuracy has also been the subject of many studies (e.g., Navidinia et al., 2018;Pourdana & Bahram, 2017;Toni et al., 2017).
Quite relevant to the title of the present study, Ebadi and Asakereh (2017) investigated the effect of DA on the development of speaking skill. Microgenetic and thematic analysis were employed to probe any possible changes in the participants' cognition development in terms of speaking performance. The findings showed that DA had a significant impact on the development in the participants' cognition and movement toward further self-regulation. Moreover, the findings of the thematic analyses the participants' satisfaction with DA. In another study by Ebrahimi (2015), it was revealed that implementing DA to enhance oral proficiency, complexity and accuracy were outstandingly improved but fluency was not affected by mediation through DA. Although DA has gained a lot of interest and attention in all educational contexts in general and in applied linguistics in particular; the amount of research about the effects of DA on foreign language acquisition processes is not sufficient yet. The present study adds a new dimension to DA research on foreign language education as well as the idea of assessment in the Iranian EFL context by investigating students' speaking accuracy and fluency in the light of DA. Moreover, there is a dearth of studies in the literature seeking to address accuracy and fluency simultaneously.

Aims and research questions
Language as the main tool of communication is not spoken in a vacuum. Knowing a language is not just knowing the grammatical rules and conventions but also knowing when to say what and to whom, that is knowledge of how the system is put to use and action in performing different social actions. Accuracy and fluency are the two fundamental factors which verify the success of English language users from non-proficient ones. This can imply that English language teachers and material developers need to take students' language needs into account and avoid one-sizefits-all approach to language teaching and learning. However, a review of the previous studies indicates that there is inadequate research exploring the effect of dynamic assessment on the speaking performance of EFL learners in general and speaking accuracy and fluency in particular which is the focus of the present study. In fact, a considerable body of research has been carried out on the applicability of dynamic assessment on different language skills. However, insufficient research has addressed the role of DA on the speaking accuracy and fluency of Iranian EFL learners. Therefore, this study is an attempt to shed more light on the effectiveness DA on EFL learners' speaking fluency and accuracy. To address the purposes of the study, speaking accuracy and fluency were considered as covariates and the following research questions were formulated:

Participants
The initial participants of the current study were 93 pre-intermediate EFL learners studying English Translation in Karaj Islamic Azad University, Alborz, Iran. The sampling procedure was convenience sampling in which the researchers recruited the participants to whom they had more access (Ary et al., 2018). The initial participants took the English proficiency test and those whose test scores fell within the limit of ±1SD were selected as the legitimate participants of the study. The participants were all female learners as the researcher did not have access to male participants. Their age range was within 18 to 30 years and Persian was their mother tongue. The participants of the study were the college students of four intact classes who were junior students of English majors. They were of pre-intermediate proficiency level as revealed by the results of PET (Table 1). In the first step of the study, participants of the study were selected in a way that ensured their homogeneity in terms of English language proficiency. To this end, initially 93 EFL learners took the English proficiency test and then those whose test scores fell within the limit of ±1SD were selected as the legitimate participants of the study. Table 1 shows the English proficiency scores of the initial participants and the English proficiency scores of the learners whose scores were within the range of ±1SD (homogenized learners). Table 1, the initial participants were 93 English learners with a mean score of 60.12 (SD = 8.38) in the general English test. After deselecting those whose scores were beyond ±1SD, 62 learners remained with the mean score of 59.98 (SD = 4.42). These learners served as the actual participants of the study.

Preliminary English Test (PET)
First of all, Preliminary English Test (PET) was administered to 93 learners to select homogeneous learners in terms of overall language proficiency. This test consists of 35 reading and 7 writing questions (90 min), 25 listening questions (30 min) and 4 speaking questions (10 min). According to the booklet of PET, approximately 340 h of English language instruction are needed before taking the test. It is a pre-intermediate level test of English language proficiency. (PET shows you are able to use your English language skills for work, study and travel The test used in the present study was taken from Quintana (2015). Regarding the validity and reliability of PET, "Cambridge ESOL examinations are designed around four essential qualities: validity, reliability, impact and practicality" (University of Cambridge ESOL Examinations, 2009, p. 2).

Speaking pre-test and post-test
Since the dependent variable of the present study is EFL learners' speaking fluency and accuracy, the researcher administered PET's speaking test at the beginning and end of the study as both pretest and post-test. Both speaking tests were administered by two examiners. They interviewed two participants at a time (based on the PET test book instructions) which took 10-12 min per interview. The speaking posttest was the same as the test for pre-test and the procedure was exactly the same as the pre-test. However, the tasks chosen from the test bank of PET speaking section were different. The posttest was administered as the final exam of the course.

Speaking fluency rating scale
In the current study, speaking fluency was measured in line with Ellis (1990as cited in Ellis & Barkhuizen, 2005 in terms of the number of syllables produced per min on a task. To do so, the number of syllables produced is counted and divided by the number of min took to produce the spoken output.

Speaking accuracy rating scale
In the present study, speaking accuracy was measured as the number of error-free clauses divided by the total number of independent clauses, sub-clausal units and subordinate clauses multiplied by 100 (Foster & Skehan, 1996 as cited in Ellis & Barkhuizen, 2005). As Ellis and Barkhuizen (2005) maintain, this measure of speaking accuracy is the most frequently used measure in the literature.

English Vocabulary in Use
"English Vocabulary in Use" authored by Stuart (2004) is a family of self-study and classroom texts for vocabulary development. The books follow the format of the English Grammar in Use titles with presentation of new vocabulary on the left-hand pages and practice exercises on the facing righthand pages. This book contains one hundred units covering vocabulary on different topics (e.g. sightseeing, hotel, airport, crime, family members, etc.). It helps learners expand their vocabulary

Semi-structured interviews
A set of semi-structured interview questions (See Appendix A) were prepared and interviews were conducted to examine students' perceptions of the effectiveness of dynamic assessment in terms of speaking fluency and accuracy. To report the results of the interviews, content analysis approach recommended by Auerback and Silverstein (2003) was adopted by following six stages including: • Getting familiar with data, • Coming up with initial codes, • Looking for themes among codes, • Reviewing the themes, • Defining and labeling the themes, • And producing the final report To address validity of the interview questions, the researcher appealed to expert opinion. Thus, after developing the first set of questions, two PhD. holders in TEFL with 20 years of teaching experience reviewed the questions and their comments were addressed. To assure consistency between the codings of the two raters (the authors), the number of agreements and disagreements between the authors' codings were calculated using Holsti's (1969) coefficient of reliability. Holsti's coefficient of reliability indicates the number of agreements per total number of coding decisions. The value was 0.90, which indicates excellent agreement. This number revealed that the inter-raters' coding result was satisfactory. Only five learners in the experimental group were interviewed to explore their perceptions in terms of the effectiveness of DA. There were two reasons behind the participation of five learners. First of all, with interviewing five learners the data reached saturation as the raters noticed the emergence of recurrent themes. As Mason (2010) notes, saturation refers to the point where the data collection process no longer produces any relevant or new data. In a similar vein, as Charmaz (2006, p. 113) maintains, interview data are saturated "when gathering fresh data no longer sparks new theoretical insights, nor reveals new properties of your core theoretical categories". Moreover, the qualitative data in the current study were not the salient data type and were used as peripheral data to substantiate the quantitative data. Therefore, only five learners were interviewed in this study. Note should be taken that in studies based on a grounded theory approach, a higher number of participants is needed to create "categories from the data and then to analyze relationships between categories" (Charmaz, 1990(Charmaz, , p. 1162; however, since the design of the present study was not a grounded theory approach and the qualitative data had a confirmatory function, the researcher assumed that five participants was sufficient. Moreover, as mentioned earlier, with five participants the collected qualitative data reached saturation which is a condition based on which the number of participants for a qualitative research project is determined.

Procedure
Initially, PET was given to 93 learners and based on means and standard deviation, 62 learners were selected based on the PET inclusion criteria and were then randomly divided into an experimental and a control group. As it is mentioned in the PET booklet, this test requires two examiners to administer the speaking part of the test. One examiner acts as both assessor and interlocutor and manages the flow of interaction by asking question and setting up the tasks, while the other just acts as an assessors and does not participate in the conversation. The scoring procedure of PET used in this study was exactly in accordance with PET scoring guidelines provided by Cambridge. Throughout the test, students were assessed on all four skills required by PET. The final score of each participant receives is an aggregate of the marks from each of the separate skills. In the next phase of the study and prior to administering the treatment, both groups were given a speaking pretest. In the current study the researcher used just the speaking part of the PET exam as a pretest (See Appendix B). The speaking performance of both groups were recorded and then transcribed. Moreover, based on the fluency and accuracy measure criteria mentioned earlier, the fluency and accuracy scores for both groups were calculated. Following that, the treatment began. To do so, the experimental group was exposed to eight successive sessions of dynamic assessment and the control group received non-dynamic assessment. It is worth mentioning that the treatment was introduced to the context by just one teacher. In other words, one teacher performed the role of mediator in experimental group and the very same teacher ran the conventional classroom in control group. Unlike the experimental group which received step-wise mediation from the most implicit type of feedback to the most explicit type, in the control group no step-wise mediation was used. In the control group, if the learners produced erroneous sentences or used vocabulary items wrongly, their errors were written on the board. Then, after asking the learner to provide the correct once, the teacher would provide the correct answer along with the pertinent explanation. The dynamic assessment group, on the other hand, received treatment in the following steps in line with Poehner (2008). Though it must be noted at the very outset of the study, according to Poehner (2008): "Intending to mediate development in the L2 classroom entails being open to providing any form of mediation learners require without concern for standardization of the procedure or adherence to a set repertoire of mediating techniques. Recall that the hierarchy of mediating moves developed by Aljaafreh and Lantolf (1994) emerged from their analysis of tutorlearner interactions and does not represent an exhaustive inventory. (Indeed, it is difficult to imagine what a "complete" list of mediation would look like!) While one may certainly enter an interaction with a plan that includes forms of mediation that might be offered, interaction in the ZPD requires that this plan be altered and perhaps even abandoned at any moment." (p. 104) Therefore, it is noteworthy that based on Poehner's (2008) comment, since instructional contexts and cognitive phenomena are to a large extent unpredictable, coming up with a set of rigorously well-established steps for dynamic assessment seems rather unfeasible. Therefore, the whole procedure was situation and learner specific. Moreover, as Poehner (2008), states in order to carry out dynamic assessment a rough list of steps as a foundation which is based on the principles of ZDP can just give the researcher a starting point.
Considering the above-mentioned points, the following steps were taken to mediate and intervene in a dynamic assessment manner to provide an assessment on vocabulary and grammar: The texts available in the "English vocabulary in use" book were distributed to the participants. In this book the vocabulary items under instructions appear on the left pages. Therefore, first the left pages were handed out. The participants were given 10 min to study this part. Next, the exercise sections which are on the right-hand pages were given to the learners. Since there were several exercises following each unit, the exercises were given to the participants, one at a time. The teacher used instruction check questions to assure that the learners were on the right track. Ample time was given to the participants to complete each exercise. The intervention in line with Poehner (2008) consistent with the principles of dynamic assessment unfolded in the forthcoming stages after the learners were done with the exercises. Poehner (2008, p. 104) mentions that, "no adherence to a set repertoire of mediating techniques" is required for dynamic assessment and "indeed, it is difficult to imagine what a 'complete' list of mediation would look like!" (p. 104). Thus, the following steps were used as the basis of dynamic assessment procedures for the current study: The mediator asked one of the learners to read the answer to the first item in the exercise section. When the answer was provided by student, the teacher asked the learner why s/he thinks the answer was right. If the learner could not think of a reason, the researcher tried to provide her/ him with some intervening steps, e.g., asking her to look at the meaning of the vocabulary in the left-hand page and asking the learner to look at the context in which the vocabulary had been used. If the learner was not able to justify the reason again, the teacher provided more contexts in which the vocabulary had been used. In case the answer was wrong up to this level, the learner had the opportunity to choose the right answer and provide a reason behind her choice. If the learner was still unable to think of a reason and provide the correct answer, the teacher would provide more hints until the learner was able to choose the correct answer and provide a reason. The above-mentioned procedure was followed for each item of the exercises. When the learners were done with the exercises, the researcher moved on to the next stage which was the production of the vocabulary they had learnt. They were asked to use each of the words under instruction in a sentence. The teacher asked one of the learners to read her first sentence. If the vocabulary had been used wrongly, the teacher started the mediation and intervention process in the same manner provided for the exercise section.
As for grammar, if the learner's production was not correct and/or appropriate, the mediator used the mediation guidelines by (Ebrahimi, 2015, p, 113) as follows: (1) "Teacher pauses (2) Teacher repeats the whole phrase questioningly As it can be inferred, step 1 is the most implicit while step 8 is the most explicit mediation which may be provided. As it may have been noticed, the dynamic assessment group received treatment in the steps specified above in line with Poehner (2008).
As for the control group, the participants received the same content in terms of vocabulary and grammar. However, there was no step-wise mediation. In his group, the researcher provided the learners with the conventional type of feedback in which the errors were spotted and the learners corrected them and then checked with the teacher. At the end of the treatment, both groups were given a speaking posttest and their speaking performances were recorded and transcribed and then scored drawing on the fluency and accuracy scoring criteria mentioned earlier. The related scores were used to address the research questions. Furthermore, to ensure the reliability of rating scores, the researcher asked a colleague who was a Ph. D holder in applied linguistics to score the data and they did the scoring collaboratively. Correlation coefficient between the two raters for the pretest was 72% and for the posttest it was 78% which are quite satisfactory. Ultimately, five participants from the experimental group were interviewed to explore their perceptions towards the effectiveness of dynamic assessment in terms of their speaking fluency and accuracy. The whole period of treatment was eight sessions which lasted for only 3 weeks as every week three sessions were held.
It should also be noted that the researcher adopted an interventionist, sandwich design dynamic assessment in the present study. The DA used in the current study was interventionist as during the assessment mediation took the form of hints, prompts, and leading implicit/explicit questions. It was a sandwich format as the learners received instruction between the pre-test and the posttest. To analyze the data, the researcher used an analysis of covariance (ANCOVA). ANCOVA is used to test the main effect of categorical variables on continuous dependent variables, controlling for the effects of selected other continuous variables, which co-vary with the dependent variable which is the covariate (Pallant, 2010). In the present study, the pretests of speaking accuracy and speaking fluency were considered as the covariates. As Pallant (2010) maintains, to run ANCOVA there is no need to assure that the performances of the groups on a certain variable are statistically insignificant prior to the treatment as the pretest scores are considered as covariate and ANCOVA calculates the mean differences on the posttest while taking into account any possible differences between the scores of the groups on pretest.

Results
As the first step of data analysis, descriptive statistics were calculated. Table 2 presents the descriptive statistics for the pretest and posttest scores of the accuracy and fluency.
In order to find answers to the research questions, ANCOVA was utilized. In this analysis, accuracy and fluency pretests served as the covariates, dynamic assessment was the independent variable and accuracy and fluency posttests were the dependent variables. To run ANCOVA, it is required that data are normally distributed, relationships are linear, slope of regressions are homogeneous, there is no multicollinearity and variances are homogeneous (Pallant, 2010). Therefore, these requisite statistical assumptions were checked and it was revealed that the assumptions (i.e. normality of data, linearity, homogeneity of variances and regression slopes) were not violated.

Addressing the research questions
The first research question set out to investigate the effect of dynamic assessment on the speaking accuracy of the Iranian EFL learners. Table 3 indicates the results of ANCOVA for the accuracy scores  According to the results of ANCOVA on the accuracy post-test scores, it was found that there was a significant difference between the dynamic assessment group and control group in terms of accuracy scores (F = 267.58, p = 0.001 < 0.05). In order to understand which group had performed better in the accuracy measure, marginal mean scores were checked (Table 4). "Estimated Marginal Means provides us with the adjusted means on the dependent variable for each of our groups. 'Adjusted' refers to the fact that the effect of the covariate has been statistically removed" (Pallant, 2010, p. 309).
According to the results in Table 4, as far as marginal means is concerned, dynamic assessment group had better accuracy scores meaning that dynamic assessment positively affected the speaking accuracy of the learners. As it is clear in Table 5, Dynamic assessment group with the score of 26.62 outperformed control group who scored 33.95. It should be noted that lower scores in accuracy measure points to less error or higher accuracy. Thus, the answer to the first research question of the study was that dynamic assessment significantly affects the speaking accuracy of Iranian EFL learners. To calculate the effect size, which indicates the magnitude of the differences between the means, partial eta squared was consulted in line with Tabachnick and Fidell (2007). As noticed in Table 3, the partial eta squared for group row is .819 which is a relatively large effect indicating that the difference between the means of the two groups is largely due to the effect of the independent variable that is dynamic assessment.
The second research question targeted the effect of dynamic assessment on speaking fluency of the Iranian EFL learners. To explore this question, the output of ANCOVA on fluency scores was checked (see Table 3).  According to the results of ANCOVA on the fluency post-test scores, no significant difference was found between the dynamic assessment group and control group in fluency scores (F = 0.92, P = 0.34). As seen in Table 5, the group effect is not significant, which indicates that for the fluency measure, there was no significant difference between the control and experimental groups.

Interview results
To seek out learners' perceptions towards the contributions of dynamic assessment towards fluency and accuracy, five learners in the experimental group were randomly interviewed. The thematic analysis further revealed the followings: Did you find dynamic assessment useful for improving the accuracy of your speaking? Why?
Four of the five interviewees answered that the use of dynamic assessment had helped them improve their speaking accuracy. When they were asked for the reason, they mostly responded that dynamic assessment was a thought-provoking type of assessment in which the teacher helped them step-by-step and did not provide them with the right answer suddenly. One of the interviews said: I really found dynamic assessment useful for my speaking accuracy. This type of assessment helps you to move slowly and the teacher gives you time to correct yourself. So, because you discover your own mistake and you correct it yourself, so you can improve the correctness of your speech.
This finding is consistent with the claimed assumption that dynamic assessment is very likely to improve the examinees' motivation and self-confidence in doing L2 speaking tasks (Poehner, 2005;Yakışık & Çakır, 2017).

The second interview question was:
Did you find dynamic assessment useful for improving the fluency of your speaking? Why? Why not?
All the five participants unanimously answered that the use of dynamic assessment was not useful for their speaking fluency. When the interviewer asked for the reason, the participants mentioned that the use of assessment created a sort of stop in the course of their talk all the time and perhaps even decreased their fluency of talk at times. One of the interviews commented: Dynamic assessment cannot improve speaking fluency since dynamic assessment does not let you speak fast. Maybe it can help your fluency later but not at the time of the assessment.
The potentially impeding nature of DA to fluency in oral performance has been previously mentioned by Estaji and Farahanynia (2019) who found that L2 participants may feel anxious while processing both oral performance and error correction. Moreover, the participants may be worried about not to be able to understand teacher's scaffolding while doing speaking tasks, thereby reducing their fluency in L2 oral performance.
The third interview question was: Do you prefer the use of dynamic assessment to learn English over the conventional feedback?
Four of the participants mentioned that they preferred dynamic assessment over the conventional type of feedback as dynamic assessment gave them time to correct their own mistakes. One of the participants held that: I like dynamic assessment more than the conventional assessment as this assessment is really fun and it also helps me discover and correct my own mistakes, too.
This exciting and interesting nature of DA may be justified in the light of the fact that DA provides the students with a more comfortable and stress-free learning environment in which the learners know that they will be assisted in case they face with challenging learning tasks. This stress-free nature of DA has been previously recognized for Iranian EFL learners (Mardani & Tavakoli, 2011). Also, the participants in the study carried out by Estaji and Farahanynia (2019) stated that the teacher in DA sessions tried to create a personal relationship with students by using gestures and verbal feedback to students, a situation which led to the emergence of a friendlier learning atmosphere in the class.

Discussion
The present study aimed at investigating the effect of dynamic assessment on the speaking fluency and accuracy of Iranian pre-intermediate EFL learners. The study administered a preand post-test to the participants to discover the change after a period of eight mediated sessions. The mediation procedure presented to the learners followed Poehner (2008). The results of ANCOVA indicated that dynamic assessment significantly impacted the speaking accuracy of the participants; however, DA did not significantly improve the participants' speaking fluency. The results of interviews also indicated that generally the participants had positive perceptions towards the effectiveness of dynamic assessment for improving their speaking accuracy.
As it was shown in the result section of the study, as far as accuracy of speaking was concerned, it was found that there was a significant difference between the dynamic assessment group and control group. This means that in post-test exam, experimental group made use of hint and prompts provided by the teacher in better way in comparison to their counterparts in control group which received no hint and prompts by the mediator. Moreover, statistical analysis of performance of participants in regards to fluency yielded different results. According to the result of ANCOVA on the fluency post-test scores, no significant difference was found between the dynamic assessment group and control group in fluency scores. Accordingly, it was concluded that dynamic assessment does not have any significant effect in comparison with the control procedure on the speaking fluency of the learners. The results of the current study support the experiment done by Ebrahimi (2015) in which the researcher conducted an experimental research on the role of DA on complexity, accuracy and fluency (CAF) of oral production of students. The study yielded the same result. Complexity and accuracy were enhanced while fluency remained unchanged after the intervention. These findings are in line with other research on dynamic assessment which indicated positive effects of DA on language learning (e.g. Birjandi et al., 2013;Haywood & Tzuriel, 2002;Hidri, 2014;Poehner, 2008). The findings of this study are also similar to Haywood and Lidz (2007) account that DA, as an interactive process, consistently and objectively measures the learners' progress in response to cues, strategies, feedback, etc. which are presented to them during the assessment.
The findings of the study are also in line with previous studies (e.g. Antón, 2003Antón, , 2009Barzegar & Azarizad, 2014;Poehner, 2008;Xiaoxiao & Yan, 2010;Yildirim, 2008) concerning the general effectiveness of dynamic assessment. The concept of ZPD is the underlying rationale for dynamic assessment which involves a combined instruction and assessment process. According to Caffrey et al. (2008), dynamic assessment culminates in a more intimate atmosphere and thus more beneficial feedback takes place as a result. The results of the current study support the findings of previous studies by Albeeva (2008), Pishghadam et al. (2011), Sadeghi and Khanahmadi (2011) etc. illustrating the positive effect of DA on ESL/EFL learning. Employing dynamic assessment in EFL classes results in more participation of L2 learners in the learning process and enhances learners' motivation (Hill & Sabet, 2009). Similar to the findings of the current study, Hill and Sabet (2009) carried out a study in which they employed dynamic speaking assessment on Japanese students and found that dynamic assessment proved useful for the speaking skill of the participants. The analysis of the data revealed that, the observed speaking accuracy difference between the control and experimental group which received mediation such as hints, explanations, prompts and leading questions was significant. This finding can be explained on the basis of ZPD. As Bekka (2010) notes ZPD has an important role and is considered as the main foundation of dynamic assessment. Via dynamic assessment and with ZPD in mind, the interaction between teachers/ assessors and learners transpire and the learners' potential to learn more merges. Put it another way, to assess learning potential is to recognize the ZPD and then via interaction assist the learner to acknowledge and accept responsibility for his own learning.
In the present study, it was found that dynamic assessment was useful in enhancing the speaking accuracy of the participants. This finding can be explained via the interview contents. As mentioned by one of the participants, dynamic assessment helped her improve her speaking accuracy as this type of assessment unfolds slowly and provides time for the learner to improve her performance. The findings of the present study may offer numerous pedagogical implications for foreign language policy makers, practitioners, teacher educators, and material developers. Given the beneficial effects of DA, EFL teachers can employ it as an effective vehicle through which different dimensions of language can be taught more effectively. Since pre-intermediate EFL learners may not have adequate opportunity to enhance their speaking inside the class, DA has the potential to help teachers to provide their learners with further speaking activities so that they can improve their own speaking accuracy beyond the walls of the classroom. DA also enables EFL practitioners to diagnose their learners' weaknesses and to provide students with the necessary mediation. As enhancing speaking accuracy of preintermediate EFL learners might be demanding and formidable because of their scanty opportunities of practicing speaking, inclusion of DA in EFL contexts might be considered as an alluring and viable alternative for syllabus designers and material developers. Additionally, the findings of the present study can enhance the awareness of teachers concerning the positive effect dynamic assessment has in improving speaking accuracy. In other words, the outcome of the current study can help teachers properly analyze and acknowledge the problems associated with learners' speaking accuracy and take due assessment steps to deal with them. Consequently, this will contribute to an integration of assessment and instruction and thus enhancing students' more effective learning. DA principles can improve EFL classroom assessment practices and make them more effective by providing the learners with situationspecific assistance. The findings of the present study can also have implications for materials developers. They are encouraged to develop materials which are more in line with ZPD and dynamic assessment practices.

Conclusion
To conclude, both the statistical results and the opinions of the participants corresponded in terms of the effectiveness of the whole assessment procedure. The results of the current study illustrated the effectiveness of implementing DA to learners' speaking accuracy performance. Comparing the findings of this study and other similar studies, it is concluded that the findings of the present study confirm other studies indicating that DA is an effective approach to language learning in general and speaking accuracy in particular. The results of the present study added to the previous theoretical literature concerning the effectiveness of dynamic assessment in improving speaking accuracy at the pre-intermediate level. This research's findings added a new dimension to the current empirical literature of dynamic assessment to second language performance and second language production; previous studies on implementing dynamic assessment paid little attention to the details of aspects of production which they are accuracy and fluency in oral production of learners and also have not paid any attention to the improvement of learners' oral production with regards to aspects of accuracy and fluency by integrating it to DA procedures. Consequently, present findings lay a path between accuracy and fluency and assessing learner's production and establish a firm basis for the integration of details of essential aspects of oral production within teaching and assessment tasks and using them to improve students' performance. This is best shown by the fact that while the learners with different proficiencies performed the same tasks, each of them needs specific and individual scaffolding which catered to his or her needs. So we may conclude that this feature of DA highlights learners need. Thus, DA seems to be more fair and ethical approach to assessment (Shabani, 2016).
The results showed that DA assisted the participants in improving their speaking skills in various ways. The participants developed and retained a positive attitude toward DA. This suggests that DA has promising potential as a classroom practice, especially for pre-intermediate proficient students in a remedial classroom. The current study, like most other studies, was not free from limitations. Firstly, Ellis's (1990) definition of speaking fluency was adopted in the present study as the researchers found calculating fluency scores based on Ellis's (1990) more feasible compared to other definitions of speaking fluency. It should, however; be noted that Ellis's (1990) definition of speaking fluency has been subject to criticism and future researchers are encouraged to take such criticisms into account when dealing with speaking fluency. Moreover, the researcher had access only to female participants. The same study can be done with male participants. Furthermore, the age range of the participants of the study was within 18 to 30. The same study can be carried out with other age ranges to improve the generalizability of the findings. The present study investigated the fluency and accuracy of students' performance; other studies can take into considerations other criteria such as complexity or the trade-off effects of these two criteria simultaneously. [to student A and then again to student B, one or more of these questions] Do you think English will be useful for you in the future?
Tell us about your school.
What do you enjoy doing at the weekends?
What did you do yesterday? Thank you.

[to both students]
In the next part, you are going to talk to each other.

Part 2 (2-3 min)
[to both students] I'm going to describe a situation to you.
At the end of term, your class is going to spend a day at the seaside. Talk together about the things you will do there.

[give students Picture 1A]
Here is a picture with some ideas to help you. [give students a few moments to look at the picture] I'll say that again.
At the end of term, your class is going to spend a day at the seaside. Talk together about the things you will do there.
All right? Talk together.
give students about 2 min to talk together. Do not join in unless they have problems, for example, to make a suggestion, e.g. Some people might like to go for a walk.] [give students about 2 min to talk together. Do not join in unless they have problems, for example, to offer a comment, e.g. I prefer to fly but I know it's not good for the environment.] Thank you. That's the end of the test.