The Online Effects of Processing Instruction on the Acquisition of the English Passive Structure

Abstract

This study investigates the immediate and delayed effects of processing instruction (PI), which is an input-based pedagogical intervention, and its key component, structured input (SI), which aims to foster making correct form-meaning connections, on the acquisition of the English passive structure. Thirty-four ESL learners, who had not received any formal instruction on the target structure, were randomly assigned to either a PI group (n = 12), SI group (n = 12), or Control group (n = 10). Both the PI and SI groups received 1 hour of computer-based instruction, while the control group did not receive any instruction. A self-paced reading test was used to measure the accuracy of response and response time in selecting the correct pictures. The test was administered before instruction, immediately after instruction, and 3 weeks later. The results indicated that both the PI and SI groups improved significantly in accuracy on both posttests, while the control group did not. No significant gains, however, were found in response time for any of the groups on any of the posttests, although the PI group was comparatively faster than the SI group on the immediate posttest.

1. Introduction

An important question within the field of second language acquisition (SLA) is how L2 learners process language for acquisition in real time. In other words, how do learners make form-meaning connections during online sentence comprehension, and what hinders them from making accurate form-meaning connections? According to VanPatten’s (1996, 2007) Input Processing Theory, learners’ accurate processing of input is hindered by their reliance on lexical-semantic as well as word order cues during comprehension. VanPatten (1996, 2007) argues that for learners to process input accurately, they should be pushed away from their less-than-optimal input processing strategies in the course of comprehension. This could be effectively achieved through processing instruction (PI), a pedagogical intervention based on the Input Processing Theory, which aims to assist L2 learners in making correct form-meaning connections during processing input for comprehension.

Today, there is sufficient evidence suggesting that PI can assist learners to process morphosyntactic forms accurately (see Lee, 2015 for a review). However, the bulk of this evidence comes from offline measures, which may not tap into implicit knowledge and do not provide a picture of moment-by-moment processing of language in real time (Benati, 2022a, 2022b; Henry, 2022). In offline measurements, learners’ comprehension of the sentence is measured once the whole sentence has been processed for meaning, without capturing the time required to process it (i.e., reaction time). Online tests, such as the self-paced reading test, on the other hand, have the advantage of providing us with an account of both accuracy in comprehending the sentence and the time required to process it, which reflects the cognitive effort involved in comprehending the sentence. In other words, learners’ reaction time to the stimuli reveals the level of cognitive effort involved in processing the input. Since the stimuli in these tests are simple and brief and focus on comprehension, learners would be less likely to rely on their explicit knowledge (Benati, 2022a; Suzuki, 2017). However, if they did, this would be reflected in their reaction times. As highlighted by Jegerski (2014) and Benati (2022a), a longer or shorter reaction time reveals the level of difficulty in processing input for comprehension. While longer reaction times indicate more effort involved in processing input, shorter reaction times suggest faster and less effortful processing. As implicit knowledge is generally characterised by being fast and less effortful (Ellis, 2005), learners’ shorter reaction times in processing input on the posttest as compared to the pretest can capture the development of their implicit knowledge of the language (Benati, 2022a; Jegerski, 2014). Thus, more recently, research on PI has witnessed a shift of focus from using offline measures to using online measures such as self-paced reading tests and eye-tracking tasks. However, unlike offline studies, studies using online measures have mostly focused on only the immediate effects of instruction. The present study, thus, investigates both short- and long-term effects of PI by using a self-paced reading test.

2. Background

2.1. Input Processing Theory

This study is theoretically grounded in VanPatten’s Input Processing Theory (VanPatten, 1996, 2007). The Input Processing Theory is basically about what psycholinguistic strategies L2 learners employ to comprehend input (language that learners receive) and derive intake (a subset of input processed for learning). The model consists of two main principles with some corollaries to account for how L2 learners process input and derive intake. These principles, as VanPatten (2015) argues, are mainly universal and independent of learners’ L1. The first principle, the Primacy of Meaning Principle, is based on the limited capacity model of attention, which states that the quantity of information that humans can perceive and take in at any given time is constrained by their limited attentional resources. As a result, humans acquire mechanisms (or perhaps they are born with them) which enable them to be selective in attending to the information surrounding them (Lee & VanPatten, 2003). In the case of SLA, Lee and VanPatten (2003) argue that learners cannot attend to both form and meaning simultaneously, and thus, in the course of comprehension, it is the meaning that L2 learners primarily attend to in the input. This suggests that L2 learners tend to process lexical forms before grammatical forms. Thus, they would tend to use lexical information rather than morphological information to comprehend the overall meaning of a sentence.

The Input Processing Theory, however, is not limited to the primacy of meaning. The second principle of this theory, known as the First Noun Principle, concerns word order. According to VanPatten (1996, 2004, 2007), the noncanonical word order poses a challenge to L2 learners in that they misinterpret the role of the first noun or pronoun they encounter in the sentence. According to the First Noun Principle, learners tend to assign the role of subject or agent to the first noun or pronoun they encounter in a sentence, making them misinterpret the OVS word order as SVO. Thus, an English passive sentence such as Jane was Kissed by Tom would be interpreted as if it were Jane who kissed Tom. Such misinterpretations would lead to deriving inaccurate intake from input and would delay acquisition (Lee & VanPatten, 2003).

2.2. Processing Instruction

PI is a pedagogical intervention that is based on the principles of the Input Processing Theory. It is an input-based instructional attempt that aims to assist learners in making correct form-meaning connections and computing sentence structures accurately during online comprehension through circumventing their nonoptimal input processing strategies. PI, as VanPatten (2018, p. 3) states, has the following features:

• “It is predicated on what learners do (and don’t do) during input processing (i.e., sentence-level processing strategies);
• It is input-oriented (learners don’t produce targeted structures during the treatment);
• Input is manipulated in particular ways to alter processing strategies and increase better intake for acquisition (structured input);
• It includes explicit information for the learner on both grammatical structure and processing problems (although the research has shown this is not necessary for its success);
• It follows certain guidelines for the creation of structured input activities.”

As stated in the fourth feature of PI, the provision of explicit information on the grammatical form and processing problems is not essential for PI to be successful. Indeed, a number of studies have now revealed that structured input (SI) is the main causative factor accounting for the effectiveness of PI (e.g., Benati, 2004a, 2004b; Benati & Lee, 2015; Farley, 2004a; VanPatten & Oikkenon, 1996; Wong, 2004). SI is meaningful input that is manipulated in a way that would require the learners to process the grammatical forms accurately in order to comprehend the sentence correctly (i.e., make correct form-meaning connections). Lee and VanPatten (2003, p. 154) provide the following guidelines for constructing SI activities:

• “Present one thing at a time.
• Keep meaning in focus.
• Move from sentences to connected discourse.
• Use both oral and written input.
• Have the learners do something with the input.
• Keep the learners’ processing strategies in mind.”

SI activities are generally composed of two types: referential and affective activities (Lee & VanPatten, 2003). Referential activities refer to those SI activities that require a factual response, which can be judged as correct or incorrect. In other words, in referential SI activities, there is only one correct response, which can be figured out only if the input is processed accurately for meaning. Example 1 below provides a sample referential SI activity for the English passive structure:

Example 1. 

Sample referential SI activity.

The horse was kicked by the donkey.

(a)

The donkey is hurt.

(b)

The horse is hurt.

John is loved by Emily.

Who loves?

(a)

Emily

(b)

John

In these examples, for instance, participants are required to read the sentences and then select the correct option based on their comprehension of the sentence. Upon selection of the option, they receive feedback only on the correctness of their answer (e.g., correct/incorrect or Yes/No).

Affective activities, on the other hand, require a response that is based on the learners’ real-life experiences or opinions. As such, there is no correct or incorrect response in these activities, and responses may vary from learner to learner. Example 2 below provides a sample affective SI activity for the English passive structure. According to Lee and VanPatten (2003), SI activities should start with a set of referential activities and then continue with some affective activities.

Example 2. 

Sample affective SI activity.

Read the sentences and determine whether they are true or false about yourself.

I am helped by my friends when I have a problem.	True	False
I am admired by my family.	True	False

Learners receiving PI would also receive the same type of instruction (i.e., SI activities). However, before the provision of SI activities, they would also receive explicit information about the target structure and the processing problem. Therefore, in practice, the only difference between SI and PI lies in the provision of explicit information in the latter.

2.3. The Processing and Instruction Perspectives of PI

PI can be viewed from two perspectives: processing (P) and instruction (I). From the processing perspective, PI is a pedagogical intervention that provides learners with opportunities to process input correctly through circumventing their nonoptimal input processing strategies (VanPatten, 2015). From the instructional perspective, there are two main components to push the learners away from their default processing strategies: (a) explicit information about the grammatical structure and how the processing problem may negatively affect the processing of input, and (b) SI activities which aim to ensure learners process input accurately. SI activities are indeed the key component of PI in that they manipulate input in a way that learners would make accurate form-meaning connections and parse sentence structures correctly. As noted earlier, there is now ample evidence suggesting that SI is the main causative factor accounting for the effects of PI and that the provision of explicit information does not add any further benefits (e.g., Benati, 2004a, 2004b; Benati & Lee, 2015; Farley, 2004a; VanPatten & Oikkenon, 1996; Wong, 2004). As outlined by Wong (2004, p. 35), SI practice “push[es] learners to abandon their inefficient processing strategies for more optimal ones”.

2.4. Offline Studies on the Passive Structure

As noted earlier, the acquisition of the passive structure is affected by the First Noun Principle. Therefore, a number of studies have selected this structure as the target of instruction in PI/SI research. These studies aimed at finding out whether PI/SI would push the learners away from their reliance on the First Noun Principle in processing the passive structure. In other words, they tried to find out whether learners receiving instruction in the form of PI/SI would parse the OVS sentence structure accurately, without being affected by the First Noun Principle, and whether PI/SI would be more beneficial than other types of instruction in this regard.

In an attempt to investigate the effects of PI on the acquisition of the English passive structure, Uludag and VanPatten (2012) compared PI to Dictogloss (DG), which, unlike PI, is an output-based instruction. The study measured accuracy in both interpretation and production. The results indicated that while both PI and DG groups improved in their interpretation of the passive structure, the PI group outperformed the DG group. Regarding production, the study found that both PI and DG were equally effective in improving the learners’ accuracy. What made the study more interesting was the PI group’s improvement in production, which was equal to that of the DG group, although the production tests were biased for the DG group, as this group received output practice during the instruction, while the PI group did not. This double interpretation/production effect of PI/SI has also been confirmed by subsequent studies (e.g., Benati, 2015, 2016; Benati & Batziou, 2019).

Benati and Batziou (2019), for instance, investigated whether SI provided on the English passive causatives would be more beneficial if combined with output-based instruction. Learners were assigned to three groups (SI, output-based instruction, and SI + output-based instruction) and were tested on their interpretation and production of the target structure before, immediately after, 3 weeks after, and 24 weeks after instruction. The results of the study revealed that SI was effective in improving learners’ interpretation and production of the target structure on all posttests, irrespective of the presence of output-based instruction. The results also indicated that learners receiving SI and SI + output-based instruction both outperformed those receiving only output-based instruction.

Taking a different focus, Benati (2015) investigated whether providing learners with more SI activities on the Japanese passive structure between the immediate and delayed posttests would be more beneficial. Instruction effects were measured in terms of accuracy in interpretation and production immediately and 4 weeks after instruction. The study found that while SI alone was effective in improving learners’ accuracy in interpreting and producing the target structure on both posttests, additional SI practice provided between the posttests reinforced the effects of instruction, leading to further improvement on the delayed posttest.

In another study, Benati (2016) compared the effects of PI to traditional instruction (TI), which involved mechanical output production. The study aimed to find out whether PI provided on the Japanese passive structure would be more effective than TI as measured by learners’ interpretation and production of this structure immediately and 2 weeks after instruction. The results of the study replicated those found in Uludag and VanPatten (2012). While learners receiving PI and TI both improved in producing the target form, only the PI group gained significant improvement in the interpretation task. The results of the delayed posttest run after 2 weeks showed no change in this pattern.

Overall, the findings coming from these studies show that PI/SI is not only effective in making L2 learners better processors of input, but it is also helpful in developing their underlying L2 system in a way that it is available for speech production in less controlled conditions. These findings reveal that PI/SI can successfully push the learners away from their reliance on the First Noun Principle when processing input for meaning, which in turn affects not only their comprehension of the input but also their production of the linguistic feature.

2.5. Online Studies on PI/SI

The bulk of evidence in support of PI/SI comes from studies using offline measures (e.g., Benati, 2001, 2005, 2015, 2016; Benati & Lee, 2010; Cadierno, 1995; Cheng, 2004; Farley, 2004a, 2004b; Marsden, 2006; Uludag & VanPatten, 2012; VanPatten & Fernández, 2004; VanPatten & Cadierno, 1993; VanPatten & Wong, 2004; Wong, 2010, 2015). While these measures can provide some evidence on the effects of instruction, they run short of providing any account of learners’ processing of language in real time. Online measures, on the other hand, have the advantage of providing a finer-grain analysis of learners’ moment-by-moment sentence parsing and processing of language in real time (Benati, 2022a; Henry, 2022). These measures may also provide a picture of the development of learners’ implicit knowledge, which could be taken as evidence of authentic L2 development (Benati, 2022a; Benati & Chan, 2023; Henry, 2022; Leeser, 2014). Thus, recently, a number of researchers have started to explore the effects of PI/SI by using online measures such as self-paced reading tests (Benati, 2022a; Chiuchiù & Benati, 2020; Lee & Doherty, 2019; Lee et al., 2020; Malovrh et al., 2020) or eye-tracking strategies (Benati, 2020, 2021, 2022a; Issa & Morgan-Short, 2019; Ito & Wong, 2019; Wong & Ito, 2018). Since the focus of this study is on measuring the effects of instruction through the use of self-paced reading tests, only studies employing the same measure are reviewed here.

In an attempt to investigate whether and the extent to which PI would lead to more nativelike input processing behaviour, Lee and Doherty (2019) compared the nonnative speakers’ processing of Spanish active and passive sentences with that of native speakers before and after receiving PI. This investigation was a rather novel attempt in that it not only examined the effects of PI on the accurate processing of input, but it also tested whether learners receiving PI would indeed process input similar to the way native speakers did. Both accuracy and response time were measured in a paired picture matching task. Eye-tracking measures were also used to capture the participants’ processing behaviour of the target forms. The study found that while, before instruction, the nonnative speakers were less accurate and slower in processing passive sentences compared to the native speakers, after instruction, they reached the native speaker level in both accuracy and response time. In terms of input processing behaviour, the results, however, showed that the nonnative speakers did not reach the native speaker level, although they became more nativelike after instruction.

Following the findings of offline studies showing that the main causative factor accounting for the effects of PI is SI, Chiuchiù and Benati (2020) compared the effects of SI to those of textual enhancement on the acquisition of the Italian subjunctive of doubt by using a self-paced reading test. Both SI and textual enhancement are input-based instructional interventions, and the study aimed to find out whether one type of instruction would be more beneficial than the other one in terms of improving L2 learners’ reading time of the verb forms. Reading time was measured to determine whether learners exposed to any of the instructional treatments would show sensitivity to violations of the target structure. The results indicated that only learners receiving SI demonstrated higher sensitivity to violations after instruction. This finding provides some evidence in support of the argument that attention alone is not sufficient for acquisition to take place (Tomlin & Villa, 1994), suggesting that without effective processing of the input and the formation of form-meaning connections, learners do not fully abandon their non-optimal input processing behaviour.

Lee et al. (2020) investigated the effects of PI on the acquisition of Spanish passive structure by running a self-paced reading test. The study tried to find answers to the question of how L2 instruction would make a difference in L2 acquisition. More specifically, the researchers aimed to find out whether instruction delivered in the form of PI would facilitate L2 acquisition through influencing learners’ cognitive processing behaviours. Learners were assigned to either a PI group, receiving instruction on the target structure, or a control group, engaging in some learning activities irrelevant to the target structure (e.g., reviewing the definite or indefinite articles). Accuracy, response time and reading time were measured in a picture selection task. Regarding accuracy, the study found that while the PI group showed improvement on both immediate and delayed posttests, the control group did not achieve any gains on any of the posttests. Response time data, however, indicated only short-term gains for the PI group, as the learners’ response time increased significantly on the delayed posttest run after 8 weeks. Finally, the study found that PI effectively influenced learners’ online processing of input, as reflected in their shorter reading time of the verb forms on both posttests, suggesting that PI was effective in influencing learners’ cognitive processing behaviours used to process the verb forms.

In a follow-up study of Lee et al. (2020), Malovrh et al. (2020) compared the short- and long-term effects of PI and guided-inductive instruction (GI) on the acquisition of the Spanish passive structure through running a self-paced reading test. GI is a type of instruction that focuses on fostering a greater depth of processing of the grammatical rules through engaging learners in creating explicit knowledge of the language. In the study, GI was operationalised as PI with the added task of asking the learners to write down rules regarding the target structure after receiving SI practice. The results of the immediate posttest indicated that while both groups improved in accuracy, learners receiving GI outperformed those receiving PI. Differences between the groups, however, were not durable as both groups’ scores equally decreased on the delayed posttest run after 8 weeks. Regarding response time, the results revealed a faster response time for both groups on both posttests. These findings indicate that involving learners in deeper processing of form during PI treatment may offer further benefits. However, these benefits are only short-term as learners receiving PI both with and without activities leading to deeper processing of form ultimately perform the same in the long run.

Following a different line of inquiry, Benati (2022a) compared the effects of SI to output-based instruction. The study examined what effects SI and meaning-oriented output-based instruction would have on the acquisition of the English passive structure. More specifically, the study aimed to find out whether the superiority of SI over output-based instruction, as evidenced in offline studies, would also manifest in real-time sentence comprehension. Accuracy, response time and reading time were measured through running a self-paced reading test. The results revealed that learners receiving SI improved on all measures after instruction, while output-based instruction did not result in improvement on any of the measures.

The studies reviewed collectively indicate that PI/SI is highly effective in changing L2 learners’ input processing behaviour. These studies reveal that learners receiving PI/SI not only improve in accuracy but also become faster and more efficient processors of input in real time, which could be taken as some evidence suggesting that PI/SI fosters L2 development through influencing learners’ cognitive processing behaviours. This new line of inquiry, however, is still relatively limited, and more studies within the same framework should be carried out (Benati, 2022a). More studies, for example, are required to measure the long-term effects of PI/SI by using online measures such as self-paced reading tests. Currently, only two studies (Lee et al., 2020; Malovrh et al., 2020) have addressed this question, and the results have been mixed. The differing findings of these two studies are despite the fact that the instructional treatment, the participants, the target structure, and the tests in both Lee et al. (2020) and Malovrh et al. (2020) were identical. This further highlights the need for additional research on measuring the delayed effects of PI/SI. The present study, thus, investigates both the short- and long-term effects of PI/SI through running a self-paced reading test. The inclusion of a delayed posttest would provide evidence on whether the effects of instruction on accuracy and mental operations in processing input would be sustainable over time. Additionally, the study involves participants who were native speakers of Persian. To date, no PI studies have addressed a similar group of participants. To meet the aims of the study, the following research questions are formulated:

Q1: What are the short- and long-term effects of PI on learners’ interpretation of English passive constructions as measured by accuracy of response and response time in online tasks?

Q2: What are the short- and long-term effects of SI on learners’ interpretation of English passive constructions as measured by accuracy of response and response time in online tasks?

Q3: Is there a significant difference between the effects of PI and SI as measured by accuracy of response and response time?

3. Materials and Methods

3.1. Participants

The participants of the present study were native speakers of Persian learning English as a second language. Following the ad for participants, 48 learners volunteered to take part in the study. First, learners who had previously received explicit instruction on the target structure were excluded from the pool of participants. This was verified by examining the textbooks used in the curriculum of the English language program that the learners had followed throughout their years of instruction. Following this filtering process, only participants who attended all phases of the study (pretest, instruction, immediate posttest, and delayed posttest) and scored below 50% at the pretest were included in the final analysis of the data. The final pool of the participants included 34 adult male and female learners of English within the age range of 18–22 years old (mean age = 19.3; SD = 1.09), who were randomly assigned to either a PI group (n = 12), SI group (n = 12), or a control group (n = 10). The participants were all at the pre-intermediate proficiency level, as indicated by the results of the Oxford Placement Test (Allan, 2004). The study received ethics approval from the Human Research Ethics Team of Macquarie University (Project ID: 13277). Before the start of the study, the participants were informed that their participation in the study was voluntary, and they could withdraw from the study at any stage with no penalty. All the participants agreed to take part in the study by signing an informed consent form.

3.2. Target Structure

The English passive form was chosen as the target structure in this study for two main reasons. First, the acquisition of the passive structure is affected by the First Noun Principle. As noted earlier, L2 learners tend to assign the role of subject or agent to the first noun or pronoun they encounter in a sentence, making them misinterpret the OVS word order as SVO. Second, the difference in how the passive voice is formed in English and Persian poses a further challenge to native Persian speakers (Li & Roshan, 2019). Unlike English, Persian is a head-last language, which is written from right to left, and commonly follows the OSV word order in passive sentences. Thus, a passive sentence such as “Sara was chased by Peter” would be formed as follows in Persian:

.شد	دنبال	پیتر	توسط	سارا
Was (became)	chased	Peter	by	Sara.
Sara by Peter chased was (became).
“Sara was chased by Peter.”

3.3. Instructional Treatment

Both SI and PI groups received 1 hour of computer-based instruction delivered via the DMDX Software, Version 6.4 (Forster & Forster, 2003). The control group, however, did not receive any instruction and only took the pretest and posttests.1 At the outset of the instruction phase, the learners in the treatment groups were provided with instructions on how to work with DMDX and were given four practice items prior to the start of the delivery of instruction.

The instructional materials were comprised of referential and affective SI activities (72 tokens; see the Supplementary Materials), developed for both SI and PI groups, and were designed following the guidelines provided by Lee and VanPatten (2003). The PI group, however, also received explicit information about the processing problem prior to receiving SI practice. More specifically, the participants in the PI group were informed, in their L1, about the potential incorrect interpretation of the first noun or pronoun as the agent (doer) of the action rather than the receiver of the action in passive sentences. Explicit information for the PI group was limited to only the processing problem, as previous research has consistently revealed that explicit information about the target structure does not offer any additional benefits (e.g., Benati, 2004a, 2004b; Benati & Lee, 2015; Farley, 2004a; VanPatten & Oikkenon, 1996; Wong, 2004). During the instructional treatments, no explicit feedback was provided to any of the groups, and the participants received feedback only on the correctness of their performance (e.g., correct or incorrect).

Attempts were made to use familiar vocabulary throughout the instructional materials. However, to further ensure that the learners would not have any difficulty understanding the vocabulary, they were provided with a list of words used during the treatment along with their definitions in their L1, 1 week before the start of the study. The learners were asked to review the words in case they did not know their meanings. They were then tested on their knowledge of the vocabulary right before the start of the study to make sure that none of the vocabulary items would be unknown to them. In rare cases when they did not know the meaning of any of the words, they were given 5 minutes to review those words again.

3.4. Tests and Scoring Procedures

A self-paced reading test, designed via DMDX, was used to measure the online effects of instruction (test items provided in Supplementary Materials). Two versions of the test were developed and administered in ABA format via the computer at three intervals (pretest, immediate posttest, and delayed posttest). The participants received the first version of the test (A) at the pretest, the second version (B) at the immediate posttest, and the first version again (A) at the delayed posttest. The tests were the same in terms of the verbs used but different in terms of agent/patient roles. Each version of the test included 36 items (12 passive sentences, 12 active sentences used as fillers, and 12 passive causative sentences, which were not analysed in this study). All sentences included singular nouns in agent/patient roles and were animate/concrete.

The self-paced reading test required the participant to read each sentence presented word-by-word at their own pace and then select the matching picture accordingly. Before taking the test, the participants were given thorough instructions on how to proceed through the test. They were also given five practice items prior to starting the test. Each word of the sentence would appear with a single press of the space bar on the keyboard and would disappear with a subsequent press of the same key, which would in turn present the next word. Once the last word of the sentence was displayed, a further press of the space bar would show two pictures simultaneously, one of which would be the correct matching picture. The participants then would have to press either the left shift key, labelled as “A” on the keyboard, or the right shift key, labelled as “B” on the keyboard, to select the matching picture. So, a sentence such as “David is massaged by Sara” would appear on the screen word-by-word in the following fashion (Figure 1) after pressing the space bar:

The participants were asked to go through the sentences as fast as they could (but not too fast to make errors) and select the matching picture as soon as they figured out the correct one. Both accuracy and response time were recorded by the DMDX Software. Each correct response was given 1 mark, while incorrect responses were marked as 0. Therefore, the accuracy score could range from 0 to 12. Response time was also calculated in milliseconds only for correct responses and was determined by measuring the amount of time the learners took to select the correct picture (i.e., the amount of time between the presentation of the pictures and the selection of the matching picture).

To ensure that the participants’ performance in the tests would not be affected by their knowledge of the vocabulary used in the tests, attempts were made to use words that would be familiar to the participants. However, similar to the instruction phase, the participants were also provided with a list of the vocabulary items used in the tests along with their definitions in their L1, 1 week before the beginning of the study. They were asked to review the words in case they did not know their meanings. They were then tested on their knowledge of the words before they took the tests, and if they did not know the meaning of any of the words, they were given 5 min to review those words again.

3.5. Procedure

This study employed a pretest-treatment-posttest-delayed posttest design to measure the online effects of PI/SI. The self-paced reading pretest was run 1 week before the treatment, and the immediate posttest was run immediately after the treatment. Three weeks later, the delayed posttest was administered. Figure 2 presents a schematic illustration of the procedure of data collection in this study.

3.6. Data Analysis

Initially, accuracy scores and response times were calculated for each participant using the R Software, Version 4.4 (R Core Team, 2024). The processed data were then entered into SPSS (Version 30) for further analysis. A mixed-design ANOVA, with time as the within-subject factor and group as the between-subject factor, was used to analyse the data. Pairwise comparisons were also run to examine differences across time within each group as well as between the groups.

In order to minimise the effects of outliers, response times with more than 2.5 standard deviations (SDs) above the mean were replaced with the value of 2.5 SD above the mean per subject in each condition. This affected 3.1% of the whole RT data. Moreover, RT data from one participant in the SI group was excluded from analysis due to missing data points.

4. Results

4.1. Accuracy

Descriptive statistics for accuracy scores are presented in

Table 1. Descriptive statistics for accuracy.

	Group	Mean	Std. Deviation	n
Pretest	PI	2.92	1.92	12
	SI	3.08	2.71	12
	Control	2.40	2.41	10
Immediate posttest	PI	9.42	2.84	12
	SI	9.92	2.50	12
	Control	2	2.49	10
Delayed posttest	PI	9.67	3.05	12
	SI	11	2.55	12
	Control	1.70	1.94	10

.

A 3 (Time: pretest, posttest 1, posttest 2) × 3 (Group: PI, SI, Control) mixed-design ANOVA was conducted to examine the effects of time and group on accuracy scores. Mauchly’s test of sphericity was not significant (p = 0.666), indicating that the assumption of sphericity was met. Therefore, unadjusted F-tests were used for within-subject comparisons. The results of the ANOVA test indicated a significant main effect for Time (F(2, 62) = 38.42, p < 0.001, η² = 0.553), a significant main effect for Group (F(2, 31) = 45.77, p < 0.001, η² = 0.747), and a significant Time × Group interaction (F(4, 62) = 11.67, p < 0.001, η² = 0.429). These findings suggest that the accuracy scores changed over time, the groups performed differently overall, and the pattern of change over time varied across groups. Therefore, a set of pairwise comparisons with post hoc tests was run to check where the differences lay.

Pairwise comparisons indicated that the three groups started at a similar level at the pretest (p = 1.000, d < 0.26). Therefore, any differences between the groups at the posttests would be due to instruction. The results of pairwise comparisons with Bonferroni correction indicated that both PI and SI groups performed significantly better at both immediate and delayed posttests, while the control group did not improve at any of the posttests. More specifically, the PI group demonstrated a significant achievement from the pretest to the immediate posttest (p < 0.001, d_av = 2.73) and the delayed posttest (p < 0.001, d_av = 2.72), with no significant differences between the immediate and delayed posttest results (p = 1.000, d_av = 0.08). Similarly, the SI group achieved significantly higher accuracy scores on both the immediate posttest (p < 0.001, d_av = 2.63) and delayed posttest (p < 0.001, d_av = 3.01), with no significant differences between the immediate and delayed posttest results (p = 0.745, d_av = 0.42). The accuracy scores for the control group, however, did not change at any of the posttests (p = 1.000, d_av < 0.32). Figure 3 illustrates accuracy gains over time for each group.

Group comparisons also revealed no significant differences between PI and SI at any of the posttests (immediate posttests: p = 1.000, d = 0.19; delayed posttest: p = 0.653, d = 0.47), with both groups significantly outperforming the Control group at both the immediate posttest (PI vs. Control: p < 0.001, d = 2.77; SI vs. Control: p < 0.001, d = 3.17) and delayed posttest (PI vs. Control: p < 0.001, d = 3.11; SI vs. Control: p < 0.001, d = 4.09).

Table 2. Group comparisons for accuracy.

	Groups	Std. Error	Sig.
Pretest	PI vs. Control	1.015	1.000
	SI vs. Control	1.015	1.000
	PI vs. SI	0.968	1.000
Immediate posttest	PI vs. Control	1.125	<0.001 *
	SI vs. Control	1.125	<0.001 *
	PI vs. SI	1.072	1.000
Delayed posttest	PI vs. Control	1.111	<0.001 *
	SI vs. Control	1.111	<0.001 *
	PI vs. SI	1.059	0.653

Notes: * = Significant difference.

presents the results of group comparisons.

4.2. Response Time in Picture Selection

Descriptive statistics for response time in picture selection are presented in

Table 3. Descriptive statistics for response time in picture selection.

	Group	Mean (ms)	Std. Deviation	n
Pretest	PI	3308.87	1344.01	12
	SI	4771.08	2958.97	11
Immediate posttest	PI	2999.88	1168.75	12
	SI	4053.43	923.14	11
Delayed posttest	PI	3255.96	1932.57	12
	SI	3341.85	2184.08	11

. As the Control group did not demonstrate any improvement in accuracy on passive sentences across the posttests, response time was not analysed for this group, as it would not yield meaningful information about changes in processing efficiency in the absence of learning.

A 3 (Time: pretest, posttest 1, posttest 2) × 2 (Group: PI and SI) mixed-design ANOVA was conducted to examine the effects of time and group on response time in picture selection. Mauchly’s test of sphericity was found to be significant (p = 0.042), indicating a violation of the sphericity assumption. Therefore, the Greenhouse-Geisser correction (ε = 0.787) was applied to within-subjects comparisons. Unlike the accuracy scores, the results of the ANOVA test indicated no significant main effect for Time (F(1.76, 36.96) = 0.997, p = 0.370, η² = 0.107) or Group (F(1, 21) = 3.382, p = 0.080, η² = 0.139). Similarly, no significant Time × Group interaction was found (F(1.76, 36.96) = 0.864, p = 0.417, η² = 0.040). These findings suggest that the learners’ response time did not change significantly over time, although there was a decreasing trend, particularly for the SI group, as illustrated in Figure 4.

Pairwise comparisons also confirmed that the change in response time in none of the groups reached statistical significance at any of the posttests (p > 0.05). However, a significant difference was found between the two groups at the immediate posttest (p = 0.027, d = 0.99).

Table 4. Group comparisons for response time in picture selection.

	Groups	Std. Error	Sig.
Pretest	PI vs. SI	944.105	0.136
Immediate posttest	PI vs. SI	422.020	0.027 *
Delayed posttest	PI vs. SI	858.300	0.921

Notes: * = Significant difference.

presents the results of group comparisons.

5. Discussion

This study investigated the online effects of PI and SI on the acquisition of the English passive structure as measured by a self-paced reading test. The research questions asked what effects PI and SI would have on learners’ interpretation of the English passive structure as measured by accuracy of response and response time in the short and long terms. The results of the analysis of accuracy data indicated that both PI and SI were highly effective in improving learners’ interpretation of the passive structure, with no significant differences between their effects. Before instruction, learners misinterpreted the passive sentences as reflected by selecting the wrong picture. This suggests that the learners assigned the role of agent or subject to the first noun they encountered in the sentence, as predicted by the First Noun Principle. After instruction, however, learners in both PI and SI groups improved significantly in selecting the correct picture, which suggests that they accurately interpreted the first noun they encountered in passive sentences as the patient or object, and not the agent or subject, of the sentence. This finding supports the results of previous online and offline studies showing that PI and SI can assist learners to make accurate form-meaning connections and parse sentence structures correctly through pushing them away from their default non-optimal input processing strategies (e.g., Benati, 2015, 2016, 2022a; Lee & Doherty, 2019; Lee et al., 2020; Malovrh et al., 2020; Uludag & VanPatten, 2012). The results also provide further support to the findings of previous offline studies showing that the main causative factor accounting for the effectiveness of PI is the availability of structured input. As noted earlier, evidence coming from offline studies has consistently revealed that explicit information does not offer any further benefits (e.g., Benati, 2004a, 2004b; Benati & Lee, 2015; Farley, 2004a; VanPatten & Oikkenon, 1996; Wong, 2004). The findings of this study reveal that making learners aware of the processing problem does not make PI more effective, either, as both learners with and without receiving information about the processing problem performed equally well on the posttests with no significant differences between them. This finding suggests that SI alone is sufficient in pushing learners away from their nonoptimal input processing strategies, and explicit information does not add any further benefits in terms of improving accuracy. Therefore, in line with offline studies, it could be suggested that the main causative factor accounting for the effects of PI on improving learners’ accuracy is the availability of SI practice, not explicit information.

One of the main aims of this study was to measure the long-term effects of PI and SI. The results revealed that the effects of PI/SI were durable as measured 3 weeks after instruction. This finding aligns with previous offline studies demonstrating the durability of the effects of PI/SI even after 14 weeks (Marsden, 2006) or 8 months (VanPatten & Fernández, 2004). It also corroborates the findings of Lee et al. (2020), which showed that PI had durable effects 8 weeks after instruction. The findings, however, run contrary to those of Malovrh et al. (2020), who found no delayed effects for PI 8 weeks after instruction. As mentioned before, the studies of Lee et al. (2020) and Malovrh et al. (2020) were identical in terms of participants, PI treatment, target structure, and tests. Yet, Malovrh et al. (2020) did not discuss the contrary findings in their study compared to their previous one (i.e., Lee et al., 2020), making any interpretation of the conflicting findings difficult. However, as the bulk of evidence coming from offline studies also points to the durability of the effects of PI/SI instruction, it seems that PI/SI would have long-term effects when tested through online measures as well.

The major focus of this study, however, was to test the effects of PI and SI as measured by response time in selecting the correct picture. Unlike accuracy, the results of the analysis of response time data indicated that none of the instructional treatments resulted in a significant decrease in response time at any of the posttests, indicating no change in the learners’ mental operations in processing meaning. This finding is in contrast to the findings of Lee and Doherty (2019) and Benati (2022a), which showed a significant decrease in response time immediately after instruction. One reason for this conflicting finding could be the rather small number of participants in the present study. According to Stevens (1996) and Pallant (2007), non-significant results may indeed be due to insufficient power, which may arise from small sample sizes. It should, however, be noted that this finding is partially in line with that of Lee et al. (2020), who similarly found no significant decrease in response time on the delayed posttest run 8 weeks after instruction. This may be taken as some evidence suggesting that while PI/SI may have immediate effects on the processing behaviour of the learners, this effect may fade away over time. This suggestion, however, should be taken with caution until further research with larger sample sizes is carried out.

While the lack of difference in response time after instruction may be partly due to the small sample size, it could also be attributed to the participants’ L1. As noted earlier, in addition to the First Noun Principle affecting the acquisition of the passive structure, the way the passive voice is formed in Persian and English poses a further challenge to native speakers of Persian (Li & Roshan, 2019). Therefore, 1 hour of instruction might not have been sufficient for changes to occur in the learners’ mental operations in processing meaning, although it led to improvement in accuracy. That is, although 1 hour of instruction in the form of 72 tokens of SI activities was sufficient to affect the learners’ accuracy in processing the target structure, faster processing of the structure might have required further exposure to SI activities. In fact, both Lee and Doherty (2019) and Benati (2022a), who showed faster response times after instruction, incorporated longer instructional periods compared to this study.

When comparing the groups, the results, however, revealed a significant difference in response time between PI and SI on the immediate posttest, indicating that the participants in the PI group became faster in accurately processing the input compared to those in the SI group. This difference, however, disappeared on the delayed posttest. This finding may suggest that while the provision of explicit information does not offer any additional benefits over SI in accuracy, it may make learners more efficient in processing the input. This advantage, however, was only short-term as eventually participants in both groups reached a similar level of processing efficiency by the delayed posttest. Moreover, the interpretation of this between-group difference must be made with caution. It should be noted that, although not statistically significant, participants in the SI group began with a slower response time at the pretest compared to those in the PI group. This initial disparity may have contributed to the significant difference observed at the immediate posttest. As such, attributing the advantage observed in the PI group solely to the instructional treatment may oversimplify the findings, particularly given the clear decreasing trend in response time observed in the SI group across the testing sessions. Future research employing a larger sample size may shed further light on whether the provision of explicit information may actually offer additional benefits in terms of response time.

6. Conclusions and Directions for Further Research

This study examined the immediate and delayed effects of PI and SI on learners’ interpretation of English passive constructions as measured by accuracy of response and response time in picture selection using a self-paced reading test. The study found that both instructional interventions were highly and equally effective in improving learners’ interpretation of the target structure. The positive effects of instruction were also sustained 3 weeks after instruction, which further supports the beneficial effects of both types of instruction. Given that the acquisition of the passive structure is affected by the First Noun Principle, it could be concluded that both PI and SI can push the learners away from relying on their non-optimal input processing strategies in processing input for comprehension.

The study, however, found no significant decrease in response time for any of the instructional treatments. The lack of effects for treatment in response time could be partly due to the small sample size in the present study. Thus, further research employing a larger sample size should be carried out to determine whether PI/SI would result in significant changes in response time, providing a more comprehensive understanding of the potential effects of instructional treatments on processing efficiency. In addition to response time in picture selection, future research may also investigate whether instruction would result in faster reading times on verb forms and whether any potential effects would persist over time. Measuring reading time would provide a better picture of moment-by-moment processing of language in real time and would provide additional information about learners’ cognitive processing behaviours (Benati, 2022a, 2022b; Henry, 2022).

Finally, future research may investigate whether input modality (i.e., auditory or written) in testing would affect how learners receiving different types of instruction, including PI/SI, process input in terms of accuracy of response, response time, and reading time. There are some arguments and evidence to suggest that processing auditory input may generally be more challenging than processing written input (e.g., Hama & Leow, 2010; Leow, 1995; Park, 2004; Wong, 2001). Yet, surprisingly, little research has addressed the role of input modality in testing learners’ processing of input. This is despite the fact that PI is basically about making learners better processors of input, which further highlights the pertinency of input modality in testing learners’ input processing.