2.1. Information processing: From input to output

When it comes to studying vocabulary learning/acquisition, examining the role of linguistic input and output looms large. Input refers to pieces of language to which the learner is exposed. It is based on such linguistic evidence that the language learner forms linguistic hypotheses. From a comprehension point of view, for ease of access, input is simplified at the lexical, phonological, and syntactic levels in order to facilitate the processing of linguistic features (^{Hatch, 1983}). There are strong arguments in the literature for the role of input in L2 acquisition in general and vocabulary learning/ acquisition in particular and most studies argue that L2 development depends on the quality of the input to which the learner is exposed (^{Ellis & Collins, 2009}). Based on the findings of the related studies, the linguistic features in the input cannot sufficiently explain learning/acquisition by themselves, however (^{Swain, 1995}). There is also another side to input which includes learner-based variables as noticing, processing, storing, and production that is output.

The other side of the coin is output or produced stretches of language by which the learners can move from lexical to syntactic processing. In addition, it provides the language learners with the opportunity to test their hypotheses with regard to how syntactic forms function in L2 (^{Swain, 1985}). Also, output can help the learners notice the gap in their interlanguage consequently it facilitates learning and makes the learners sensitive to upcoming patterns of the target language (^{Swain, 1995}). Furthermore, output has been demonstrated to increase frequency and accuracy of grammatical structures (^{Toth, 2006}; ^{Izumi & Bigelow, 2000}).

2.2. Input, noticing, input-enhancement and vocabulary learning

In the domain of SLA, ^{Lee and VanPatten (2003}, p. 26) put forward a metaphor about language input: “input is to language acquisition what gas is to a car… an engine needs gas to run; without gas, the car would not move an inch… likewise, input in language learning is what gets the ‘engine’ of acquisition going…without it, acquisition simply doesn't happen.”

Second Language Acquisition (SLA) has attracted a considerable interest believing that holding learners' attention to the formal features of second language input is beneficial and essential for second language development. ^{Sharwood-Smith (1993}) stated that input is the ‘potentially processible language data which are made available, by chance or by design, to the language learner’ (p.167). Furthermore, ^{Gass and Mackey (2015}) defined language input as the language data a learner is exposed to. In order that second language acquisition occurs, the learners should necessarily attend the L2 input. Since late 1970's, input has been under investigation in second language acquisition (SLA) research. It is believed that input is a necessary and sufficient condition, while output is a facilitative, but not a necessary condition for L2 acquisition. In 1991, Sharwood-Smith suggested another term, which is ‘input enhancement’, to discuss and highlight the role of grammar in second language acquisition. Additionally, the term “input enhancement” was suggested to demonstrate a learning strategy through which second language learners are exposed to the input for its certain features so that they become perceptually salient and noticeable. Reinders and ^{Ellis (2009}) defined input enhancement as input process in which the target structure forms are highly frequent. Generally, input enhancement can be closely related to be one of the focus on form approaches in second language acquisition. At first, input enhancement was considered to be like consciousness-raising and a type of focus on form technique. ^{Nassaji and Fotos (2011}) believe that input enhancement is used not only to raise learners' awareness of target forms but also to improve learners’ acquisition. If the learners are exposed to enhanced input repeatedly, the target vocabulary items are most likely to be noticed. As literature indicates, one type of input enhancement which is common in studies is textual enhancement (TE). According to ^{Wong (2003}), there are two features of textual enhancement including: Firstly, second language learners read texts to achieve propositional content. Secondly, learners’ attention can be drawn through such typographical cues as bolding, underlying and italicizing. Additionally, evidence shows that typographical cues can also be beneficial in learning and perception of new information and forms.

As far as noticing is concerned, ^{Schmidt (2001}) put forward that “people learn about the things that they attend to and do not learn much about the things they do not attend to” (p.30). ^{Nassaji and Fotos (2011}) stated that textual enhancement is believed to be an “implicit and unobtrusive way” and keep learners’ attention in targeted forms. Learner’s attention is directed to the meaning of the text, and incidentally on form-meaning relationships (^{Ellis, 2008}). In a contrary manner, ^{Lee (2007}) believed that target forms which are textually enhanced may distract learners’ attention from meaning, thereby having a harmful effect on comprehension. ^{Godfroid, Boers, and Housen (2013}) suggested that even though the perception of external stimuli is considered to be an essential part of L2 learning, the main functions of noticing, attention and awareness in second language learning are controversial and not that clear. In the same manner, Nassaji and Fotos (2011) proposed that it is possible to notice the target forms without taking the meaning of the form into account.

Noticing, which can be done through input enhancement, is described as “the process of the learner picking out specific features of the target language input which she or he hears or reads, and paying conscious attention to them so that they can be fed into the learning process” (^{Cullen, 2012}, p. 260).

At first, ^{Schmidt (2001}) believed that conscious attention is an essential part of second language acquisition meaning that conscious attention is a necessary process through which second language acquisition occurs. However, he has changed the view in such a way that noticing and conscious awareness are separated. He proposed that noticing is limited to “awareness at a very low level of abstraction” (Schmidt, 2001, p. 5). Therefore, in order to put Schmidt's and ^{Tomlin and Villa's (1994}) views in the same line, Robinson (as cited in ^{Song, 2007}) defined noticing as "detection plus rehearsal in short-term memory, prior to encoding in long-term memory" (p. 296). ^{Schmidt (1994}) has also suggested that if a target feature in the input is noticed, it might become intake. In other words, the input becomes intake if learners notice and attend to input.

2.3. Input-based activities and vocabulary learning

Task-based language teaching (TBLT) has been known to include production-based tasks. ^{Swan (2005}) put TBLT under question due to this belief that it focuses on ‘pushed output’ and learners cannot grasp opportunities to acquire ‘new linguistic material’. On the contrary, ^{Ellis (2009}) suggests that, tasks can also be ‘input-based’. He believes that these kinds of tasks can be used to give the learners a chance of learning the ‘new linguistic material’ that Swan considers it to be necessary. ^{Shintani (2012}) argues that:

An input-based task aims to promote interlanguage development by directing learners’ attention to second language (L2) input through listening or reading without requiring them to produce the L2. However, L2 production is not prohibited in an input-based task; learners may elect to respond to the input they receive by engaging in language production (p. 254).

The main purpose of the task enriched with input is to provide the learners with samples of target forms and items but does not require the learners to indicate that they have successfully processed the input.

Input-Based Reviewing is a process through which the learners try to review and retrieve the target words via such instruments as word cards in which a target word is on one side and on the other side either its translation in learners’ mother tongue, or its equivalent in second language or a picture is appeared. ^{Takač (2008}) states that “strategies of retrieval play a very important role in learning: every recall of a previously learnt word strengthens the link between knowledge and retrieval cue” (p. 75). As mentioned earlier, ^{Nation (2001}) argues that retrieval can take two different forms, namely receptive and productive. The former deals with the cue which is written form then meaning should be retrieved. The latter deals with the cue which is meaning then the word form should be retrieved. Teachers should make learners more aware of the importance of retrieving for their vocabulary learning and encourage them to integrate this repetition technique into their learning activities (Nation, 2001). Retrieval plays an important role in the strategy of using word cards for vocabulary learning and it makes the word cards more favorable for learners compared to other strategies such as notebooks or lists of vocabulary items (^{Schmitt & Schmitt, 1995}; ^{Waring, 2004}). Because the target words and their meanings are put on different sides of word cards, retrievals with them will be easier for learners compared with word lists where L2 words and their meanings are presented at the same time. The strategy of using word cards for vocabulary learning will be examined in the following section.

The retrieval process is an indispensible part of flashcard learning. So, learners using flashcards should be encouraged to retrieve the meaning of the target word from memory, which leads to a more permanent learning (^{Barcroft, 2007}; ^{McNamara & Healy, 1995}; ^{Nation, 2001}). In addition to the retrieval process, the order of the flashcards is another factor which affects the learning process. According to ^{Baddeley’s (1990}) primacy and recency effects, the items at the beginning and the end of a list are memorized better than the words in the middle. Taking into account this finding, and also the fact that learners have the freedom to change the order of words if they study with flashcards, learners should put difficult words near the beginning, so these words can get more attention.

2.4. Output-based activities and vocabulary learning

There is now a growing consensus on the role that output plays in the language acquisition (^{Russell, 2014}). Simply stated, output is the language a learner produces. Output occurs when a learner discusses and writes within a group of learners who then give immediate feedback for the purpose of solving a problem (^{Swain, 2000}). The findings of Swain's (1985) seminal paper about output countered Krashen's input hypothesis (1980, 1985). The major tenet of her hypothesis (Output Hypothesis) was that the provision of comprehensible input alone is not enough for acquisition to take place and that the production of output forces learners to process language more deeply, attending to both meaning and linguistic form simultaneously (Swain, 1985, 1993, 1995, ¹⁹⁹⁸). As Swain (1985) argues, producing the target language may serve as “the trigger that forces the learner to pay attention to the means of expression needed in order to successfully convey his or her own intended meaning” (p. 249).

The Output Hypothesis can be discussed on the ground of some psycholinguistic and cognitive models of processing as well. According to ^{Anderson's (1983}) ACT model, for example, practicing to produce the language is likely to free up the cognitive resources needed for attending to other aspects of language via the process of automatization. Two important concepts of ^{McLaughlin’s (1987}) Information Processing Model are 1) controlled / automatic processing, and 2) restructuring. Briefly stated, the most important variable facilitating these two mental processes is production practice, which provides the impetus for the shift from controlled to automatic processing and the occurrence of restructuring. Later on, McLaughlin and Heredia explained the functions of practice as follows:

(Production) Practice can have two very different effects. It can lead to improvement in performance as sub-skills become automated, but it is also possible for increased practice to lead to restructuring and attendant decrements in performance as learners reorganize their internal representational framework. It seems that the effect of practice does not accrue directly or automatically to a skilled action, but rather cumulates as learners develop more efficient procedures (^{McLaughlin & Heredia, 1996}, p, 218).

According to ^{Swain (1995}, 2005), one important function of output, among others, is helping learners notice the gap between their linguistic resources and the target language system. To help students reach this level of noticing, several out-based activities can be employed. Output-Based Reviewing is one of them in which learners are asked to generate the word forms. Following productive skills, the concept of the word is changed when they make their own expressions, and grasp other features of the word. Furthermore, learners are required to use those words in such activities as paraphrasing the sentences containing the unknown words. ^{Nation (2001}) suggests that in generative phase, getting the learners to reconstruct what was in the text rather than repeat it, learners can retell of the written input in different ways. Another exemplar of output-based reviewing is that learners review the target forms and are required to tell stories with pictures. Nation (2001) introduced a generation activity, paraphrasing, which can be used in new contexts across the four skills. According to Swain (2005) depth of processing refers to the degree of analysis and elaboration carried out on input through activities like paraphrasing rather than mere repetition, ''with greater depth being associated with longer term and stronger memory traces'' (Swain, 2005, p. 475).

3.1. Participants

One hundred and fifty intermediate English learners (males and females) at Rezvan and Khan-e-Zaban Institutes in Ardabil were the final participants of the current study. They were selected based on a simple random sampling. The participants were randomly assigned into five groups, namely Noticing through Input enhancement (n=30), Input Enhancement plus Input-Based Retrieving (n=30), Input Enhancement plus Output-Based Reviewing (n=30), Input Enhancement Coupled with Input-Based and Output-Based Reviewing (n=30), and a pilot group (n=30). The participants' age ranged from 16 to 28.

3.2. Materials and Instruments

3.2.1. Preliminary English Test (PET)

In order to select a homogenous set of participants a general proficiency test, Preliminary English Test (PET), was used. It included five sections, namely reading (including 3 parts and 30 items), writing (including 2 parts and 5 items), use of English (including 4 parts and 42 items), listening (including 4 parts and 30 items), and speaking (including 4 parts). Due to time constraints and subjectivity in scoring, the speaking and writing sections were excluded.

3.2.2. Academic Word List

Another instrument was the Academic Word List (AWL) (^{Coxhead, 2000}). The AWL is based on frequency and range data from a corpus of 3.5 million words from academic texts. It includes 570 word families- defined by Coxhead (2000) as “a stem plus all closely related affixed forms” (p. 218). It was employed both in the treatment sessions as the target words and assessment sessions as the pre- and posttests items.

3.2.3. Focus on Vocabulary 2 Mastering the Academic Word List

It is a research-based vocabulary textbook developed by ^{Schmitt, Schmitt, and Mann (2011}) that gives high intermediate to advanced students the advantage they need to succeed in academic environments and master academic words. It has featured 504 out of 570 word families ^{Coxhead (2000}) had identified as frequently used words in academic texts across a wide range of topics. Eight units of this book have been chosen through which 56 academic words have been contextualized, 7 words in each unit. These passages were used as an instrument in which the target words were encountered.

3.2.4. Vocabulary Levels Test (VLT)

It was originally designed by ^{Nation (1990}) and has been wildly used in a number of vocabulary research studies (e.g., ^{Cobb, 1997}; ^{Laufer & Nation, 1999}; ^{Laufer & Paribakht, 1998}; ^{Schmitt & Meara, 1997}). Later it was modified by ^{Schmitt, Schmitt, and Clapham (2001}) and was developed in two versions, where the validation evidence is also presented. Each section of the revised VLT consists of 30 items in a multiple matching format. Items are clustered together in 10 groups for this, so that learners are presented in each cluster with six words in a column on the left and the corresponding meaning senses of three of these in another column on the right. Learners are asked to match each meaning sense in the right-hand column which one single word from the left-hand column. Within each level, there is a fixed ratio of word classes to represent the distribution of English word classes. This ratio is 3 (noun) : 2 (verb) : 1 (adjective) in the latest revised versions (Schmitt et al., 2001). Word classes are not mixed within any one cluster. For the 2001 version of the VLT, two parallel test versions are available that have been established to be relatively equivalent. Both versions will be used both as an assessment instrument and in the validation phase as an established test. The VLT includes four frequency-based levels (the 2,000 word level, 3,000 word level, 5,000 word level, 10,000 word level). It also includes the AWL.

As ^{Brown (1990}) argued reliability strategies for norm-referenced tests, e.g., test-retest and equivalent-from strategies, all depend, in one way or another, on the magnitude of the standard deviation. On the other hand, a relatively high standard deviation is one result of developing a test that spreads students out along a continuum of abilities which is a feature of NRT tests. However, the same strategies may be inappropriate for CRTs because they are not developed for the purpose of producing variance among scores.

Of the three categories applied specifically to the estimation of CRT consistency, threshold loss agreement, squared-error loss agreement or domain score dependability, the first one has been used to find the reliability or dependability of the VLT. To do so, two common threshold loss agreement statistics, i.e., agreement coefficient (ρ _o ) and kappa coefficient (k) have been employed. These two coefficients measure the consistency of mastery-non-mastery classifications across two test administrations. The agreement coefficient is simply the proportion of examinees consistently classified (as masters or as non-masters) on both administrations and kappa reflects the proportion of consistent classifications observed beyond that expected by chance. However, as ^{Brown (2005}) argues, administering a test twice is cumbersome and hard on everyone involved. So several approaches (e.g., ^{Huynh, 1976}; ^{Subkoviak, 1976}; ^{Peng & Subkoviak, 1980}) have been worked out to estimate threshold agreement from one administration. Since these methods required access to computer facilities and appropriate software and assume a somewhat advanced background in test theory, they were also difficult for practitioners to implement. More recently, ^{Subkoviak (1988}) presented practical approaches for approximating both the agreement and kappa coefficients. To reach either of these coefficients from a single test administration, two values are required. The first is the cut-point score converted to a standard score. The second value is one of the NRT internal-consistency reliability estimates. Once the standardized cut-point score and an internal-consistency reliability are in hand, it is just a matter of checking the appropriate tables developed by Subkoviak (1988, pp. 49-50).

Following ''the well-established contrasting group method'' proposed by ^{Brown (2005}, pp. 236-7) to reach the cut-points, converting these points to standardized cut-point scores according to the formula developed by Brown (2005, p. 203), estimating the internal-consistency reliability for each test through calculating Cronbach alpha coefficient (a), and checking the appropriate tables mentioned above, the agreement coefficient (ρ _o ) and kappa coefficient (k) for the reliability of each test have been approximated as follows:

3.3. Procedure

The initial participants of the present study were 212 male and female students from intermediate classes at the above-mentioned institutes in Ardabil. They sat for the proficiency test (PET) based on which 171 learners were identified as intermediate. Finally, 150 learners whose scores fell within one standard deviation above and below the sample mean were randomly selected as the final participants of the study. Thirty out of the 150 learners were randomly selected and assigned into a pilot group who sat for the VKS consisting of 150 vocabulary items (extracted from the AWL, 60 words in Sublist 8, 60 words in Sublist 9, and 30 words in Sublist 10, respectively) in order to discover the unknown words. Given the results of the VKS, the words the participants recognized through the first and the second categories came to be 71 words out of which 56 words were randomly selected as the target words. Regarding the sample size of lexical items, it is self-evident that the more samples you obtain from your participants, the more valid and reliable your results should be. However, this measurement ideal is always in contrast with the practical reality that the amount of time to administer a study is always limited (Schmitt, 2010) and in practice, a researcher is usually able to measure only a small proportion of lexical items in focus. Furthermore, there is no set answer to the tricky question of how large (or how small) a target sample is acceptable. However, bearing in mind Schmitt's (2010) guidance that the researcher needs to be able to argue that the sample which is measured gives reliable and meaningful information about the vocabulary being discussed and consulting Nation (2017, personal communication), the researchers have come up with the final decision in this regard.

The remaining 120 participants were randomly assigned into four groups each including thirty learners who had already been proved to be homogeneous by meeting the criteria. The four treatment conditions included noticing through input enhancement, input enhancement plus input-based reviewing, input enhancement plus output-based reviewing, and input enhancement plus input-based and output-based reviewing. Each group sat for the pretest the items of which were adopted from the VLT developed by the researchers. The items for the four groups were composed of the words coming from Sublists 8, 9, and 10.

A point must be clarified here that by noticing through input enhancement the first condition is meant i.e., noticing, by input enhancement plus input-based reviewing the second one is meant i.e., noticing combined with retrieval, by input enhancement plus output-based reviewing the third one is meant i.e., noticing plus generation, and in the last group input enhancement is coupled with both input-based and output-based reviewing. What follows is a brief account of what was performed in each group:

Group one : Noticing through Input enhancement: Each session the learners in this group were given a reading passage including at least seven out of 56 target words. Seven of those words were highlighted to enhance input and call learners’ attention. So participants of this group took part in eight treatment sessions.

Group two : Input enhancement plus input-based reviewing: Learners in this group followed the same procedure as the first group did, that is noticing the target words. What is more in this group is that every session the learners reviewed and retrieved seven words introduced in the previous session by using researcher-made word cards out of those words. The idea of using word cards originates from the fact that retrieval or reviewing “… can involve an informal schedule for returning to previously studied items on word cards and the recycling of old material” (^{Nation, 2013}, p. 302). According to Nation (2013), in order to provide students with a chance of reviewing learnt vocabulary, word cards can be applied in different forms including a foreign language word on one side and either its first language translation, or its second language definition or a picture on the other side of the card. From the three procedures to use flash cards the first one, a foreign language word on one side its first language translation on the other side, was used in this study. It is imperative to mention that the input-based reviewing of seven words presented in the first session was done in the second session and the treatment of the second seven words in the third session and so on. Consequently, the participants of this group took part in nine treatment sessions i.e., the treatment of the last seven target words was done in the ninth session.

Group three : Input enhancement plus Output-based reviewing: Learners in this group followed the same procedures as the first group did; that is, noticing the target words through input enhancement. Furthermore, they were required to use those words in such a generative activity as paraphrasing the sentences containing the unknown words. As mentioned above, this approach originates from ^{Nation (2013}) who introduced paraphrasing or using words in new contexts across the four skills as a kind of generation activity. It goes without saying that, like the second group, the output-based reviewing of seven words presented in the first session was done in the second session and the treatment of the second seven words was done in the third session and so on. So the participants of this group, as referred to before, took part in nine treatment sessions i.e., the treatment of the last seven target words was done in the ninth session.

Group four: Input enhancement coupled with input-based and output-based reviewing: This condition was the sum of the three conditions provided for the first three groups, that is, they had noticing through input enhancement; retrieval through using researcher-made word cards; and generation through rewriting the sentences containing the unknown words. Needless to say, this group, too, took part in nine treatment sessions like the second and third groups.

The immediate posttest including the target words presented in the class was administered one week after the last treatment session. Vocabulary learning is incremental in nature, and only research designs considering this element can truly describe it (Schmitt, 2010). One way to meet this requirement is simply adding one or more delayed posttests. So, the participants sat for a delayed posttest two weeks after the immediate one. To exclude any interim immediate posttest exposures and their subsequent practice effect on the delayed posttest, the practical question of the length of the delay arises. There is no clear-cut answer to this question in the literature and as Schmitt (2010) argued ''there is no standard period of delay, and that any delay beyond the immediate posttest is better than nothing'' (p. 156). Following Schmitt (2010), the delay between the posttests in this study was two weeks.

3.4. Data Analysis

Since the study enjoys a delayed posttest and two dependent variables, to measure within-group differences across different measurement times (from pretest to the immediate posttest and from the immediate posttest to the delayed posttest) four one-way repeated measures ANOVAs were run to analyze the data obtained from the VLT. Moreover to measure the between-group differences, three one-way ANOVAs followed by subsequent post-hoc comparisons were run for the three time intervals among the four groups for the data obtained from the VLT.

4.1. Results

Within-group analyses

First of all, the normality of the distribution of scores was assessed to decide on using either parametric statistics or their non-parametric equivalents. So, the data obtained was assessed through a series of Kolmogorov-Smirnov tests confirming that the scores obtained are normally distributed. Since each subject was measured on the continuous scale on three occasions, one-way repeated measures ANOVA was conducted to detect the possible significant differences in the participants' performance through the three time intervals.

Table 2 shows the descriptive statistics for the scores obtained from VLT for the first group over the pretest, the immediate posttest, and the delayed posttest. It is revealed that the first group's scores increased from the pretest to the immediate and delayed posttests but there was a slight decline from the immediate posttest to the delayed one.

Table 3 shows the results of the one-way repeated measures ANOVA for the VLT obtained from the first group over the pretest, the immediate posttest, and the delayed posttest. The results revealed that there is a significant difference in the performance of the participants in the first group over the three time intervals (Wilks' Lambda = .10 with a probability value of .000, F (2, 28) = 1.361, p < .0005, multivariate partial eta squared (η²) = .990). To determine the exact location of the difference, the pairwise comparisons are also required (Table 4)

Based on the results of descriptive statistics (Table 2) coupled with the post-hoc Pairwise Comparisons for the first group (Table 4), it was revealed that the group made gains significantly from the pretest (M = 8, SD = 2.913) to the immediate posttest (M = 12.63, SD = 2.965) (p = .000) and the delayed posttest (M = 12.53, SD = 2.874) (p = .000) but not from the immediate posttest to the delayed posttest (p = .249).

To analyze the data to answer the second research question, that is, the possible effect of the second condition, i.e., input-enhancement combined with input-based reviewing, the same procedures followed for the first research question have been followed the findings of which are presented in the tables 5 through 7.

Based on table 5, it is revealed that the second group's scores increased from the pretest to the immediate and delayed posttests but there was a drop from the immediate posttest to the delayed one.

Based on the results of the one-way repeated measures ANOVA for the VLT obtained from the second group over the pretest, the immediate posttest, and the delayed posttest (Table 6), it was revealed that the difference in the performance of the participants in the second group over the three time intervals was significant (Wilks' Lambda = .004 with a probability value of .000, F (2, 28) = 3.436, p < .0005, multivariate partial eta squared (η²) = .996).

Merging the results of the descriptive statistics (Table 5) with the post-hoc Pairwise Comparisons for the second group (Table 7), it was revealed that the group made gains significantly from the pretest (M = 7.63, SD = 3.068) to the immediate posttest (M = 21.83, SD = 3.291) (p = .000) and the delayed posttest (M = 19.50, SD = 2.418) (p = .000). It was also revealed that, unlike the first group, the second group's decline from the immediate posttest to the delayed one is significant (p = .000).

The results obtained from the third group across three time intervals replicate those for the second group- that is, the third group's scores soared from the pretest to the immediate and delayed posttests but there was a decrease from the immediate posttest to the delayed one (Table 8).

To ensure that these changes are statistically significant, the results of the multivariate tests are displayed in Table 9:

The within-subjects results (Table 9) revealed that the difference in the performance of the participants in the third group over the three time intervals was significant (Wilks' Lambda = .005 with a probability value of .000, F (2, 28) = 2.715, p < .0005, multivariate partial eta squared (η²) = .995). The results yielded from the pairwise comparisons (Table 10) coupled with the descriptive statistics mentioned above (Table 8) replicated those of the second group, that is, that the third group made significant improvement from the pretest (M = 7.40, SD = 2.283) to the immediate posttest (M = 29.93, SD = 4.218) (p = .000) and the delayed posttest (M = 27.80, SD = 3.718) (p = .000). It was also observed that the third group's fall from the immediate posttest to the delayed one is significant (p = .000).

The fourth research question aimed to find the possible effect of the last condition introduced for the first time in this work, that is, the impact of input-enhancement combined with input-based and output-based reviewing on vocabulary learning and retention. To do so, the same procedures conducted for the first three groups were followed for the last group.

Like the second and third groups, the fourth group, too, gained improvement from the pretest to the immediate and the delayed posttests (Table 11). Based on the results of the multivariate tests (Table 12), the difference between the three time intervals is significant (Wilks' Lambda = .005 with a probability value of .000, F (2, 28) = 2.590, p < .0005, multivariate partial eta squared (η²) = .995).

To find the exact location of the differences between the pretest and the immediate and delayed posttests, the results of the pairwise comparisons are displayed in Table 13 according to which it can be asserted that all three time intervals are significantly different from one another. Furthermore, taking the results of the previous findings into account (Tables 11 and 12), it can be concluded that the scores soared significantly from the pretest (M = 7.67, SD = 2.657) to the immediate posttest (M 41.23, SD = 5.425) (p = .000) and the delayed posttest (M = 36.50, SD = 4.841) (p = .000). The difference between the immediate posttest and the delayed one is also significant (p = .000). The forth research question is also partially rejected, that is, the impact of input-enhancement combined with input-based and output-based reviewing on vocabulary learning is statistically significant but its effect on vocabulary retention is under question.

Between-group analyses

To see whether the four conditions have differential effects on vocabulary learning and retention or not, three one-way ANOVAs were run for the three time intervals among the four groups. The first one was conducted for the data obtained from the pretests of the four groups:

Based on the one-way ANOVA run (Table 15), the four groups were taken to be equal in their initial proficiency in the target words, F = .24, p = .86.

To discover the potential immediate outperformance of the groups over one another, the data of the immediate posttest was also analyzed through a one-way ANOVA accompanied by post-hoc comparisons the results of which are presented in tables 16-18:

Based on the results of the one-way ANOVA run (Table 17), the between-group differences in the immediate posttest were significant (F = 265.042, p = .000). The post-hoc test (Table 18) run showed that the differences between the groups were significant. More specifically, taking into account the descriptive statistics (Table 16), it was concluded that the second group outperformed the first group, the third group outscored the first and the second ones and the fourth group outperformed the three previous groups.

The same procedures were followed to find the potential long-term differences in the performance of the four groups.

Based on the results of the one-way ANOVA run (Table 20), the between-group differences in the delayed posttest were also significant (F= 251.11, p = .000). These results coupled with the descriptive statistics (Table 19) and the results obtained from the post-hoc test (Table 21) showed that in the delayed posttest, like the immediate posttest, the second group outperformed the first group, the third group outscored the first and the second ones and the fourth group outperformed the three previous groups.

4.2. Discussion

Regarding the first condition, it was observed that the exposure to input which is textually enhanced through boldfacing does facilitate the learning of vocabulary and has immediate effect on it. In the long term, however, the results are fairly vague. Although the group gained significant improvement from the pretest to the delayed posttest, there was a slight decline in the performance of the group from the immediate posttest to the delayed one. Putting all mentioned findings into account, it can be argued that the first condition, i.e. noticing through input-enhancement, has significant positive effects on vocabulary learning but its effect on vocabulary retention should be dealt with greater care and conservatism. Theoretically speaking, the first part of this finding i.e., positive effects of noticing through input-enhancement on vocabulary learning, confirms the original psycholinguistic underpinnings of attention and/or awareness in SLA in general and second language vocabulary learning in particular. As ^{Leow (2015}) argues one irrefutable key process in all of the psychology-based theoretical underpinnings is the role attention plays in the L2 learning process. Particularly stated, the finding is in line with Schmidt's (1990, 1994, 2001) Noticing Hypothesis through which he argues that only L2 data that are noticed can be converted into intake and noticing is ''the necessary and sufficient condition for the conversion of input to intake'' (Schmidt, 1994, p. 209). The finding is also in line with ^{Sharwood-Smith's (1991}, 1993) ''input enhancement'' by which he meant the guidance teachers provide for promoting second language learners’ self-discovery or conscious awareness of the formal features of the L2. Regarding the empirical side of the coin, this finding is in line with ^{Baleghizadeh, Yazdanjoo, Fallahpour (2018}), ^{Izumi (2002}), ^{Jourdenais et al. (1995}), Leow (2001), ^{Lee and Lee (2012}), and ^{Rassaei (2014}). Regarding the second finding, i.e. the lack of positive effects of noticing through input-enhancement on vocabulary retention, several facts should be referred to. First of all, there are few empirical studies that have used delayed posttests to see if the positive effects of input enhancement actually persist in the longer term (^{Sharwood-Smith & Truscott, 2014}). However, theoretically speaking, the finding can be justified from some perspectives. An item of information can be consolidated and transferred from working memory to long-term memory by rehearsal. The more often we retrieve a particular item of information from long-term memory, the easier it becomes to access it and the less likely it is to be lost. Information that is rarely retrieved may decay and language attrition is an example of such decay (^{Field, 2004}). On the other hand, vocabulary knowledge is more likely to experience such attrition compared to other linguistic aspects (Schmitt, 2010). So, we can argue that the decline from the immediate posttest to the delayed one can be the result of language attrition because between the two posttests there was no language exposure.

Regarding the second condition, first, it is asserted that input-enhancement combined with input-based reviewing has immediate significant positive effect on the participants' vocabulary knowledge (here called vocabulary learning). ^{Ellis (2003}) explained that two types of input-based approaches, structured input and consciousness-raising, can be best used in teaching grammar. The first approach, structured input, has been adapted to the teaching of L2 vocabulary through making use of flash cards. Such structured input task occupies an important part in processing instruction (^{Takimoto, 2007}). ^{VanPatten (1996}) argues that processing instruction entails an explanation of the relationship between a given form and the meaning it can convey and ''structured input activities in which learners are given the opportunity to process form in the input in a controlled situation so that better form-meaning connections might happen compared with what might happen in less controlled situations'' (p. 60). So, the substantial gains for the participants after reviewing the words through form-meaning connections and the controlled activity can be attributed to what Ellis (1997) termed ''structured input''. Furthermore, this finding lends support to previous studies (^{De la Fuente, 2002}; ^{De Jong, 2005}; Ellis & He, 1999; ^{Shintani, 2012}), which have shown that input-based instruction benefits vocabulary knowledge. Second, comparing the results of the pretest with the delayed posttest, we observed that in the long-term, too, the participants maintained the positive effect of the treatment. However, they experienced a significant decline from the immediate posttest to the delayed one leading us to cast doubt on the long-term significant positive effect of the second treatment on vocabulary knowledge as well (here called retention). Like the first case, this decline can be justified by the lack of exposure between the immediate and the delayed posttests. Furthermore, we can argue that not only was input-based reviewing unable in compensating this lack of exposure but also the situation got worse, that is, the decline from the immediate posttest to the delayed one was significant. This conclusion is in line with ^{van Zeeland and Schmitt (2013}). They compared the effects of four levels of input frequency in one learning session. Although they found small positive effects of repeated input on the immediate posttest, no significant difference was found on the posttest conducted 2 weeks after learning. Based on their finding, ^{Nakata (2016}) argued that ''although increasing frequency in one learning session enhances learning in the short term, its advantage diminishes over time'' (p. 4). This finding can also be analyzed through the distinction drawn between working memory and long-term memory (Baddeley, 1999, ²⁰⁰⁷; ^{Baddeley & Hitch, 1974}; ^{Gathercole & Alloway, 2008}) that will be discussed at end of this section. According to ^{Nation (2013}), flash cards can be employed as a tool to review the input already provided. However, in order for such a strategy to succeed students should enjoy some skills to benefit its advantages. As ^{Nakata (2008}) claimed about his non-native Japanese participants, it is likely that the subjects of the current study, too, could not take full advantage of benefits offered by cards due to lack of experience.

The findings pertinent to the third condition, i.e. input-enhancement combined with output-based reviewing, replicated those of the second condition discussed above. Specifically speaking, it can be concluded that the input-enhancement combined with output-based reviewing positively affects the participants' vocabulary knowledge (here called vocabulary learning). Furthermore, based on the significant difference between the pretest and the delayed posttest, it can be argued that the treatment has long-term significant positive effect on vocabulary knowledge as well (here called retention). However, there was a significant decline from the immediate posttest to the delayed one. The positive effects of input-enhancement combined with output-based reviewing on vocabulary learning can be discussed with regards to ^{Swain's (1985}) Output Hypothesis through which she argued that learners must be pushed to produce output in the second language in order to develop accuracy. This finding is also consistent with some works studying and confirming the noticing function of output (^{Adams, 2003}; ^{Kwon, 2007}; ^{Horibe, 2003}; ^{Izumi & Bigelow, 2000}). The possible effect of this treatment on vocabulary retention, to the best of our knowledge, has not been the subject of empirical research. The finding reveals that although the treatment has worked regarding the participants' improvement from the pretest to the immediate posttest, its effectiveness is under question when the delayed posttest is concerned. To put it differently, the positive effect of this treatment like the previous one on the long-term retention of vocabulary is rejected and we can claim that output-based reviewing, too, couldn’t save the words from being faded away.

Putting the findings related to the second and third conditions into account, the positive effect of the fourth condition, input-enhancement combined with input-based and output-based reviewing, on vocabulary learning is predictable. The condition had positive effect on vocabulary learning, that is, the group's improvement from the pretest to the immediate and delayed posttests was significant. However, the participants experienced a significant decline from the immediate posttest to the delayed one. Reviewing the theoretical perspectives discussed above, we can claim that the first finding is consistent with Noticing, Input and Output hypotheses. However, the second finding, the decline from the immediate posttest to the delayed one, is intriguing. Since the amount of exposure in this group was more than the previous three groups, it was predicted that the groups' performance would not decline from the immediate posttest to the delayed one. However, the prediction was not confirmed and retention has not been achieved through this condition. In contrast to the comparative studies, the potential efficacy of combining input-based and output-based practices has not received much research attention. Although some studies were conducted to find the effect of such synergy on grammar learning (^{Gass & Torres, 2005}; ^{Kirk, 2013}; ^{Jabbarpoor & Tajeddin, 2013}; ^{Tanaka, 1999}, ²⁰⁰¹), no such study was conducted for vocabulary learning.

What discussed above was concerned with the within-group differences over three testing periods for each condition. To discuss the differential effects of the vocabulary teaching procedures in focus, the between-group differences should be taken into account too.

Based on the date gained in the immediate posttest, the results revealed that 1) the students treated through input enhancement+input-based reviewing outperformed those treated through input enhancement alone, 2) the students treated through the input enhancement+ouput-based reviewing performed better than those treated through the input enhancement alone and those treated through input enhancement+input-based reviewing and finally, 3) the students treated through input enhancement+input- and output-based reviewing outperformed those treated through input enhancement alone, those treated through input enhancement+input-based reviewing and those treated through input enhancement+out-based reviewing.

With respect to the first finding, as ^{Sharwood-Smith (1993}) himself cautioned the enhanced input might or might not be further processed into the language system, that is, L2 knowledge. This caution can be interpreted that while ''enhanced input is noticed and taken in by the learner, such linguistic information may not be further processed due to the kind of evidence to which learners are exposed'' (^{Leow, 2015}, p. 166). Moreover, Leow's (2015) assertion about the grammatical benefits of input enhancement can be employed to justify the superiorly of input enhancement coupled with input-based reviewing over input enhancement without such reviewing in the realm of vocabulary learning. After having a critical review of the literature on grammatical benefits of input enhancement, he argued that a conflation of input enhancement with one or more variables (e.g., instruction, feedback, metalinguistic information, oral discussion, etc.) appears to contribute to better grammatical development. In terms of vocabulary instruction, one of those variables is input-based reviewing operationalized, in this study, through using flash cards. The finding is in line with ^{Kornell (2009}) and ^{Komachali and Khodareza (2012}) which found the positive effect of using flashcard-based instruction on vocabulary learning. The finding also confirmed two recent studies investigating the efficacy of flashcard-based instruction, ^{Lin-Fang (2013}) and ^{Sinaei and Asadi (2014}).

Regarding the second finding, most of the studies investigating the efficacy of output-based activities have mainly focused on grammar and compared its effect with some other grammar instructions the results of which are far from conclusive and mixed enough to urge researchers to do more works to clarify the point. The story is blurrier in terms of vocabulary instruction and the studies on input- and output-based instruction of vocabulary have revealed mixed results. While many studies confirm the positive role of output production in the development of learners’ vocabulary (^{DeKeyser, 1997}; ^{Hashemi & Kassaian, 2011}; ^{Izumi & Bigelow, 2000}; ^{Jalilifar & Amin, 2008}; ^{Kwon, 2007}; ^{Soleimani, Ketabi, & Talebinejad, 2008};), in some other studies, input-based instruction appears to be more effective in L2 vocabulary development (^{Long, Inagaki, & Ortega, 1998}; ^{Shintani, 2011}). Some others, like ^{Rassaei (2012}), concluded that L2 knowledge might develop through both input-based and output-based instruction. This finding of the current study confirms the noticing function of output in vocabulary learning. As ^{Swain (2005}) claims while attempting to produce the target language, learners may notice that they do not know how to say precisely the meaning they wish to convey. Consequently, they might pay more attention to something they need to discover about their L2. Output not only makes learners notice their interlanguage defects, but also activates the internal cognitive process of second language acquisition, so as to promote language acquisition. It also provided the participants with the chance of hypothesis testing and the consolidation of interlanguage knowledge.

With regard to the third finding, we can argue that continuing to investigate the two types of practice in order to determine which should predominate the other may not be the best avenue to pursue, especially since both input and output are widely used in language classrooms. Differently stated, uniquely focusing on one approach or the other may be unfruitful; rather, we should look after the best ways to merge these two approaches in order to exploit the unique benefits of each simultaneously. Research combining input- and output-based instructional approaches has been scarce. There are few studies (e.g., ^{Gass & Torres, 2005}; ^{Hanaoka, 2007}; ^{Izumi, 2002}; ^{Kirk, 2013}; ^{Leeser, 2008}; ^{Tanaka, 1999}, 2001) investigating such synergy based on which we can claim that the combination seems to be favorable on the whole whenever grammar instruction is concerned. However, to the best of our knowledge, no attempt has been made to study the effect of the integration of input- and output-based practices on vocabulary learning. Integrating input- and output-based practices, the participants enjoyed the retrieval advantage of the first one and the noticing function of the latter.

Regarding the between-group differences in the long term, i.e. the differences reported in the delayed posttest, the results replicated those for the immediate posttest. Regarding the superiority of input enhancement combined with input-based reviewing through flash cards over input enhancement without such intervention, it is worth mentioning that vocabulary learning through flash cards can be categorized as a kind of rehearsal. Rehearsal is defined as an activity through which learners encode new information into their long-term memory through overt or silent articulation (Tanaka, 2008). Based on cognitive psychology literature, Tanaka (2008) concludes that ''a successful recall from memory yields superior retention to mere presentation of the target item because the very act of retrieving information from memory strengthens retrieval routes to memory'' (p. 5). According to the phenomenon known as the ''retrieval practice effect'' (^{Baddeley ,1997}; Ellis, 1995; ^{Nation, 2013}), testing one’s memory to recall the L2 word form or its meaning is beneficial to long-term retention.

The second argument drawn from the between-group differences is the superiority of the input enhancement combined with output-based reviewing over the input enhancement combined with input-based reviewing. Although the argument that vocabulary used in production tasks was recalled better than words practiced in non-production tasks has been questioned by some studies (e.g. ^{Dekeyser & Sokalski, 1996}; ^{Horibe, 2003}; ^{Sakai, 2004}), it is supported by a number of empirical studies (^{Ellis & He, 1999}; ^{Hulstijn & Laufer, 2001}; ^{Hulstijn & Trompetter, 1998}; ^{Joe, 1998}; ^{Nobuyoshi & Ellis, 1993}). One rationale underlying the long-term effectiveness of output is that production induces high cognitive involvement, and is more efficient in leaving durable imprints in memory than simple exposure to target words in the input (^{Kwon, 2007}). Furthermore, production provides L2 learners’ with an opportunity to promote noticing on their lexical knowledge, either on their deficiency of vocabulary or on the gap between their receptive and productive vocabulary knowledge.

Vocabulary retention gained through input enhancement coupled with input- and output-based reviewing has been reported to overtake both input enhancement without any of them and with one of them. In this condition, the input condition provides learners with a good chance to understand the meanings of the target words, and such comprehension is transferred to the production of words when they have extra opportunity to produce those words in the output phase. According to ^{Lee and Benati (2009}), general consensus exists that the effects of instruction will fade, wear off, or diminish over time, but that the results of effective instruction should not disappear completely. They employed the term 'durative effects’ to refer to posttesting that occurs beyond the time frame of examining the immediate effects of instruction. The last treatment employed in the current study is one of those effective instructions that had durative effects, preventing vocabulary diminution and resulting in durable learning.