Motivation plays a critical role in college students’ academic success, particularly for studying and essay writing, which require active problem-solving and sustained use of metacognitive strategies (Britt et al., 2017; Rouet, 2006; Rouet et al., 2017; Snow, 2002). Theories of self-regulated learning emphasize the integration of cognitive and motivational factors in guiding students’ learning processes (Boekaerts, 1991, 1992; Efklides and Schwartz, 2024; Pintrich, 2000; Zimmerman, 1989). Expectancy-value theory (EVT) further delineates motivation by identifying two core components: competence beliefs (i.e., perceived self-efficacy in one’s abilities), which correspond to expectancy, and the perceived importance or enjoyment of tasks, which define task value (Eccles et al., 1983; Eccles and Wigfield, 2002; Wigfield, 1994; Wigfield and Eccles, 1992, 2002).

Translanguaging theory contextualizes how motivation fosters reading among multilingual learners, enabling them to navigate monolingual educational systems that often overlook their linguistic strengths (García and Wei, 2014; Gu et al., 2024). For example, Liu et al. (2024) examined how self-efficacy and engagement, two motivational constructs that align with EVT, influenced the relationship between home language and English reading among Chinese elementary school students. They found that self-efficacy in the home language significantly contributed to English reading comprehension through cross-linguistic transfer, highlighting the importance of leveraging home language literacy skills. Additionally, intrinsic motivation, which is linked to enjoyment and interest in reading, positively influenced comprehension in both the home language and English.

Previous research on multilingual college students suggests that translanguaging aligns with key motivational constructs in EVT. For example, Kamhi-Stein (2003) found that multilingual college students in the U.S. who viewed their home language as an asset were more likely to use cross-linguistic strategies, such as mental translation, to navigate complex academic texts. This positive perception reflects confidence in drawing on home language skills to support reading comprehension in English, aligning with competence beliefs in EVT. Additionally, students who recognized their cultural and linguistic knowledge as valuable for academic success found learning more engaging and meaningful, reinforcing the construct of task value (Eccles and Wigfield, 2002; Kamhi-Stein, 2003). This aligns with evidence that college students studying English as a foreign language make better progress when they demonstrate a growth mindset about their ability to improve (Calafato and Simmonds, 2023). Growth mindset is closely related to self-efficacy, as confidence in one’s ability to improve fosters resilience and persistence (Blackwell et al., 2007; Dweck, 2009; Rhew et al., 2018). Similarly, Rutgers et al. (2024) found that learners’ self-perceptions as multilinguals, including their linguistic, emotional, and evaluative connections to multiple languages, serve as a foundation for fostering resilience and creativity when engaging with complex texts in English.

Directed Self-Placement (DSP) is tied to concepts of motivation in academic success because it offers a student-centered approach to course placement by emphasizing self-assessment and agency (Royer and Gilles, 1998). Unlike standardized placement methods, DSP can incorporate motivation by using reflective questionnaires that prompt students to evaluate their competence, prior experiences, and readiness for specific courses (Ferris et al., 2017; Royer and Gilles, 1998). DSP can also validate the diverse linguistic and educational backgrounds of multilingual learners by encouraging them to reflect on their use of cross-linguistic strategies or home-language literacy skills to navigate academic texts (Johnson and Vander Bie, 2024; Ferris et al., 2017).

However, DSP may also introduce risks, particularly if students misjudge their capabilities, which can lead to potential misplacement (Ferris et al., 2017; Crusan, 2011). Incorporating performance-based reading and writing tasks alongside self-rating surveys may help address this issue for some groups of college students (Aull, 2021). For example, Aull (2021) found that students in 100-and 200-college level courses adjusted their self-assessments after completing a performance-based task, which required them to read an article on fake news and write a 250–300-word summary of its key points. Conversely, students placed in preparatory courses showed no significant changes in their self-ratings. Aull suggests that this may be because these students’ self-perceptions were shaped more by general messages about their abilities than by the task itself.

For multilingual learners, understanding how their perceptions of competence interact with specific task demands in English-dominant contexts is essential for accurately assessing their academic level (Baker et al., 2024). Although Aull (2021) did not specifically examine multilingual learners, the author found that students requiring additional support rated their abilities similarly before and after a writing task. This suggests some students may not need a proficiency test to adequately self-assess their general skill level. However, motivation may have influenced how they engaged with different aspects of the task. For instance, students who anticipated needing a preparatory course may have felt more confident and engaged when identifying key ideas in the article, but less so when summarizing main points or structuring sentences with appropriate syntax and cohesion. Extending this to multilingual learners, analyzing how their perceptions of ability and engagement relate to specific task demands at the item level (e.g., performance on individual reading questions or rubric criteria on an essay) could provide more targeted insights for instructional support as they navigate English-only academic contexts.

Building on this idea, Baker et al. (2024) required multilingual college students to rate their confidence on specific test items within their DSP framework. However, they did not examine whether students’ self-assessments on test items aligned with their actual performance, leaving open the question of how students with different motivation levels perform on item-level tasks. The current work employs Explanatory Item Response Models (EIRMs) to explore this gap in the literature, by linking students’ broader self-assessments of motivation to their actual performance on specific test items (De Boeck and Wilson, 2004; Maris and Bechger, 2009). Unlike aggregate scoring approaches, EIRMs analyze how motivation interacts with different linguistic and cognitive demands within a task, offering a more nuanced understanding of self-assessment accuracy across various task characteristics, and for diverse student groups. This approach could reveal patterns that placement assessments often overlook, providing deeper insight into the role of motivation in multilingual learners’ academic performance.

The reading vocabulary assessment comprised 20 items selected from the reading vocabulary subtest of the GMRT-4 (Level 6; MacGinitie et al., 2002). The GMRT-4 reading vocabulary assessment evaluates both the depth and breadth of vocabulary knowledge. Specifically, it assesses participants’ comprehension of individual words within specific contexts, indicating depth, while also encompassing a wide variety of vocabulary words across multiple items, offering insight into breadth.

This assessment was designed to evaluate receptive (passive) vocabulary knowledge, meaning it measured the participants’ ability to recognize and understand words in context rather than produce them actively. Participants silently read short phrases or sentences presented on a computer screen, each containing a single underlined word, followed by five answer options. Their task was to select the option that best matched the meaning of the underlined word in context, which included both higher-and lower-frequency words. A higher-frequency word is one that appears more commonly in daily conversation (e.g., “They want to inspire him”; Responses: express concern for, motivate, attempt to persuade, frighten, save; Answer: motivate). In contrast, a lower-frequency word is less commonly used in everyday speech and may be more typical of academic or specialized language (e.g., “The scientist attempted to extrapolate the data”; Responses: confirm, estimate beyond, analyze, refute, compare; Answer: estimate beyond). The reliability of the items for the current study sample was excellent (Cronbach’s alpha coefficient = 0.92).

We determined word frequency score for each of the 20 underlined words using the Zipf scale from the SUBTLEX-US word frequency database (van Heuven et al., 2014). Zipf scale values range from 1 to 7 (1 = lowest frequency, 7 = highest frequency). The Zipf scores for reading vocabulary items ranged from 1.59 to 3.83. Item 4 had no assigned score in the database and was therefore excluded from analyses. From the 19 items, a score of 2.63 represented the median frequency score across all items. Thus, items with a score of 2.63 and above were considered higher frequency and items below 2.63 were considered the lower frequency items¹.

Reading motivation was assessed with two 5-point Likert-scale items, measuring the two critical dimensions of (a) competence beliefs and (b) task value that are central to EVT (Eccles and Wigfield, 2002; Wigfield and Eccles, 2002). The item assessing competence beliefs prompted students to self-assess confidence in their reading ability with the statement “I am confident in my reading ability,” rated on a scale from 1 (not like me at all) to 5 (very much like me). The item measuring task value for reading asked students to rate reading enjoyment with the statement “I enjoy reading,” also on a scale from 1 (not like me at all) to 5 (very much like me).

To address our primary research questions, we conducted Explanatory Item Response Modeling (EIRM) using the lme4 package in R Studio (Bates et al., 2015). In line with the assumptions of the Item Response Theory (IRT) framework, where the intercept typically represents item easiness, we opted not to include an intercept in our models (De Boeck et al., 2011). By excluding the intercept, we aimed to directly estimate item parameters and simplify the interpretation of model coefficients.

Three models were estimated to explore person-and item-level effects on students’ probability of correctly answering reading vocabulary items (coded as 0 = incorrect, 1 = correct). The first model (i.e., Model 1) estimated person-level factors at level two, including year in college (γ₀₁), monolingual English speaker status with multilingual learner status as the reference group (γ₀₂), high reading competence beliefs (γ₀₃), and high reading enjoyment (γ₀₄) with lower reading competence and enjoyment as the reference (i.e., score of 3 and below). Additionally, this model included item-level parameters at level 1 (γ₁₀). The specification for Model 1 is shown below: L n p i j / 1 − p i j = γ 01 Year in College j + γ 02 Monolingual English j + γ 03 High Reading Competence j + γ 04 High Reading Enjoyment j + γ 10 Item Parameters i j + u 0 j + ∈ i j

In the subsequent two models, we investigated combined person-by-item effects alongside the main effects, which encompassed either reading competence or enjoyment, monolingual English speaker status, and lower-frequency words. Due to the inherent limitations in estimating both item parameters and traditional interaction effects across item and person level effects within a single model (De Boeck et al., 2011), we employed subset differential item functioning (DIF) modeling. Subset DIF requires the creation of new covariates representing the combined effect between dichotomous item and person-level variables (see De Boeck et al., 2011).

In this study, item frequency and motivation were measured on continuous and ordinal scales, respectively. Therefore, we created dichotomous categories for word frequency, reading competence, and reading enjoyment. Items with a word frequency below the median of 2.63 were categorized as lower frequency. As in Model 1, reading competence and enjoyment were classified into high (ratings of 4 and 5) versus relatively lower scores (3 and below). The equation for Model 2 below illustrates the three new dummy-coded covariates with person-by-item effects derived from reading competence, multilingual learner status, with lower frequency words (γ₁₁ – γ₁₃)². A similar third model (Model 3) included person-by-item effects with reading enjoyment and multilingual learner status with lower frequency words. L n p i j / 1 − p i j = γ 01 Year in College j + γ 02 Monolingual English j + γ 03 High Reading Competence j + γ 04 High Reading Enjoyment j + γ 10 Item Parameters i j + γ 11 Monolingual English × Lower Reading Competence × Lower Word Frequency i j + γ 12 Multilingual Learner × High Reading Competence × Lower Word Frequency i j + γ 13 Multilingual Learner × Lower Reading Competence × Lower Word Frequency i j + u 0 j + ∈ i j

Despite the inherent limitations of dichotomizing continuous data and data with multiple levels, we chose this method to gain deeper insights into the combined effects of multilingual learner status, motivation, and word frequency, over and above item-level parameters. Moreover, dichotomization of self-reported reading competence and enjoyment addressed the heavily skewed distribution of self-reported ratings, predominantly ranging from 3 to 5, with only a few students rating themselves a 1 or 2 (as explained later in the preliminary findings).

If the person-by-item effects showed significant results indicating a meaningful relationship between different levels of motivation on multilingual learner status and word frequency, we conducted a planned follow-up analysis. This analysis aimed to further explore whether English exposure among multilingual learners varied between groups characterized by differing levels of reading competence and enjoyment. We examined descriptive statistics, including the mean, standard deviation, and distribution of English exposure rates among multilingual learners across high and lower reading competence beliefs and enjoyment. To assess potential differences in mean English exposure across these groups, we used a Wilcoxon rank-sum test, a non-parametric test selected due to the relatively small sample size of multilingual learners (N = 39), which were divided into lower and high motivation groups.

Next, we examined whether the number of students with only 1 year of college (Freshmen) differed among multilingual learners with varying levels of reading competence and enjoyment. Given the small sample size of multilingual learners and the limited number of students in years 2–4 of college (see Table 1), we compared the number of Freshmen to a combined group of students with more than 1 year of college. This analysis utilized Fisher’s exact test, comparing frequencies in a two-by-two table of multilingual learners with high and lower motivation, categorized by whether they were Freshmen or had more college experience.

The parameter estimates from Model 2 are shown in Table 6. A significant number of the item-level parameter estimates were negative (ps < 0.001), indicating that students were generally less likely to answer items correctly. Years in college (γ01) was significantly related to item-level performance (OR = 1.24, 95% CI = 1.02–1.52, p = 0.038), suggesting that students with more years in college had a higher probability of answering items correctly. However, reading competence beliefs (γ04) did not significantly relate to item-level performance beyond students’ years in college (OR = 1.75, 95% CI = 0.99–3.10, p = 0.08). Similarly, neither monolingual English speaker status (γ02) nor reading enjoyment (γ03) were significantly related to students’ probability of correctly answering the items (ps > 0.05).

Figure 1 illustrates the nature of the combined effect between multilingual status, reading competence, and word frequency. Compared to monolingual English speakers with higher reading competence, multilingual learner students with lower competence exhibited significantly lower probability of success on lower frequency words compared to higher frequency words (OR = 0.17, 95% CI = 0.06–0.54, p = 0.003). This effect was not observed for monolingual English speakers with lower reading competence or multilingual learners with high reported reading competence.

This follow-up analysis examined potential differences in the duration of English exposure and number of years in college between two groups of multilingual learners, categorized by higher and lower reading competence (M[SD] age = 21.7[5.29] and 19.4[1.80], respectively; see Supplementary Table S5 for demographic details). Both groups exhibited similar rates of English exposure, with multilingual learners who reported higher reading competence having an average exposure rate of 0.71 (SD = 0.25; Min - Max = 0.1–0.97), and those with lower competence showing an identical average exposure rate of 0.71 (SD = 0.22; Min - Max = 0.21–0.95). A subsequent Wilcoxon rank-sum analysis confirmed that there were no significant differences in English exposure rates between these groups (p = 0.764; see Supplementary Table S6 for the complete Wilcoxon rank-sum results). Figure 2 illustrates this with a cloud plot, showing that despite the non-normal distribution of English exposure across the two groups, both groups display similar distribution trends with a negative skew.

Next, we used Fisher’s Exact Test to examine the association between group membership (lower motivation vs. higher motivation) and year in college (year 1 vs. years 2–4) among multilingual learners. The test yielded a p-value of 0.301, indicating no statistically significant association between group membership and having more years in college. However, there was an observed trend with fewer Freshman students in the higher motivation group (12 out of 28, 43%) compared to their less motivated peers (7 out of 11, 64%).

Existing DSP models often cite Bandura’s self-efficacy theory, which focuses on competence beliefs, or students’ self-efficacy in their ability to complete a task successfully (Bandura, 1989). While self-efficacy is a critical factor in motivation, it does not account for task value, or the extent to which students perceive a task as meaningful, useful, or engaging. Task value is particularly important in literacy assessments because students are more likely to engage deeply with reading and writing tasks when they see them as relevant to their academic goals (Eccles and Wigfield, 2023, 2024). However, DSP assessments do not typically include task value explicitly, limiting the ability to capture how all aspects of motivation vary across literacy settings.

The motivation measures in the current study aligned more closely with the original Expectancy-Value Theory (EVT) framework, as they broadly assessed competence beliefs and task value without situating them in specific literacy contexts (Eccles et al., 1983). However, motivation should be assessed in specific scenarios that allow students to reflect on different literacy experiences across languages, as well as immediately after completing a literacy task in English. More recent developments in EVT, specifically Situated Expectancy-Value Theory (SEVT), emphasize that motivation is shaped by more specific academic and social contexts (Eccles and Wigfield, 2023, 2024). This means that competence beliefs and task value fluctuate depending on the literacy setting and task demands. For multilingual learners, motivation may differ based on whether they are reading and writing in English or their home language across different literacy contexts in day-to-day life. Incorporating these variations into DSP assessments would allow for a more accurate understanding of how multilingual students engage with academic literacy tasks.

To better assess competence beliefs and task value across different literacy settings, DSP assessment frameworks could consider including measures that capture motivation in both broader language and literacy contexts and task-specific situations. Students could reflect on their competence beliefs across different language and literacy scenarios by responding to statements such as, “I am confident understanding novels in English” or “I am confident reading academic texts in my home language.” Similarly, task value could be assessed with items like, “I enjoy reading newspapers in English” or “I enjoy reading newspapers in my home language.” These measures capture students’ general engagement with literacy across languages but do not fully account for how motivation fluctuates in response to specific academic tasks.

To align motivation measures with specific task demands, students could self-rate their competence beliefs and task value on a Likert scale after completing a performance-based measure during the DSP process. For example, after summarizing a reading, students could evaluate their confidence in the task by responding to statements such as, “How confident were you in identifying the main idea of the article?” or “How confident were you in summarizing the article in your own words?.” Task value could also be further refined by asking students not only whether they found the task engaging but also whether they believed that certain aspects of the task were a necessary part of the overall task goal, as indicated by statements such as, “Understanding the main points of the article was an important part of successfully completing the summary.”

In the current study, the lack of significant findings for task value in relation to vocabulary performance may reflect the broad nature of the motivation questions, which did not distinguish between different aspects of task value or account for how motivation shifts in response to specific tasks. By refining how motivation is measured across different literacy experiences and immediately after performance-based tasks, DSP assessments may better capture the relationship between different aspects of motivation and literacy assessment performance.

Multilingual students employ their full linguistic repertoires in academic settings, using their home languages to aid comprehension and meaning making. Rather than viewing languages as separate systems, multilingual learners blend their linguistic resources to engage with academic tasks. For example, Canagarajah (2011) argues that multilingual students fluidly integrate their languages in academic writing, challenging traditional monolingual standards that dominate assessment practices. Similarly, García and Wei (2014) emphasize that translanguaging is not simply code-switching but an active process of constructing meaning across languages. However, DSP frameworks have not systematically incorporated these perspectives into self-assessments. While some placement models acknowledge multilingual backgrounds, self-assessments primarily focus on English proficiency and self-perceived academic skills rather than the cognitive and linguistic strategies that students use across languages (e.g., Baker et al., 2024; Ferris et al., 2017; Horton, 2022; Johnson and Vander Bie, 2024). Expanding DSP assessments to include translanguaging strategies could provide a more accurate and equitable understanding of multilingual students’ literacy engagement.

Recent studies have developed self-assessment questionnaires that measure translanguaging strategies in academic literacy. For example, Öztürk and Çubukçu (2022) used a questionnaire to examine how college students incorporate translanguaging into reading tasks, finding that students frequently used their home languages for comprehension, paraphrasing, and critical thinking. Similarly, Ali (2024) employed survey-based measures with college students to analyze how translanguaging supports academic literacy, showing that students used it to improve reading, writing, and note-taking. Champlin (2016) also used self-assessments to explore translanguaging in bilingual instruction, emphasizing the need for structured support and training to maximize its benefits. A common finding across these studies is that multilingual students consistently use translanguaging as an academic resource.

To capture translanguaging practices effectively, DSP assessments could incorporate two distinct but complementary approaches. The first focuses on students’ broader use of translanguaging in daily literacy activities, asking them to reflect on how they engage with reading and writing across languages in both academic and non-academic contexts. For example, students could respond to statements such as “When you read for school, do you mentally translate difficult words into your home language?” or “How often do you switch between English and your home language when reading news articles, texting, or writing notes?” These questions provide insight into how students integrate multiple languages as part of their regular literacy habits, not just in formal academic settings.

The second approach situates translanguaging within specific DSP tasks, allowing students to reflect on their language use immediately after completing a reading or writing assignment. After summarizing an article, students could indicate whether they relied on their home language for comprehension by answering questions such as, “Did you mentally translate any parts of the article into your home language while reading?” or “While writing your summary, did you think in your home language before translating your ideas into English?” By assessing translanguaging both broadly and in direct relation to performance-based measures, DSP frameworks can gain a more comprehensive understanding of how multilingual students navigate academic literacy in English while drawing on their full linguistic resources.

Traditional DSP frameworks incorporate a range of assessment methods, including broad self-assessments, holistic writing samples, and institution-specific placement models (e.g., Baker et al., 2024; Ferris et al., 2017). While some DSP surveys include post-task self-assessments (e.g., Ferris et al., 2017), others rely on more general measures of self-confidence that do not capture how motivation interacts with specific literacy tasks (Toth and Aull, 2014). Aull (2021) has emphasized the need for more refined and task-specific DSP assessments, particularly those that account for the nuanced ways multilingual students engage with academic literacy.

EIRMs provide a more detailed understanding of how motivation, linguistic background, and literacy performance interact at the item level (Wilson et al., 2008). Traditional assessments often produce broad placement decisions, but EIRMs allow for a more precise analysis of how different learner profiles engage with specific dimensions of a literacy task. For example, Magliano et al. (2007) emphasized that multiple-choice reading assessments require various skills, including both text matching and inferencing. Similarly, rather than treating reading proficiency as a single construct, EIRMs distinguish between these subskills to identify how learner characteristics, such as self-efficacy levels or translanguaging tendencies, relate to specific aspects of reading comprehension and task performance.

Findings from the current study highlight the value of this approach. The EIRM analysis suggested that multilingual students with lower reading self-efficacy performed significantly worse on lower-frequency vocabulary items than both their monolingual peers and multilingual students. However, our findings also suggested that some multilingual students may perform just as well as their monolingual English speaking peers if they endorse high reading competence beliefs. While Aull (2021) and Ferris et al. (2017) took a step toward more accurate self-assessment by having students rate their competence immediately after a literacy task, our results suggest that incorporating data-driven methods such as EIRM analyses can provide a more precise understanding of how learner characteristics relate to different components of a literacy task. For example, a student may report feeling less confident in their reading ability on a vocabulary assessment, which aligns with their overall score, but may not recognize that their difficulties are specifically in decoding lower-frequency vocabulary. Similarly, another student may rate themselves as having high competence on a reading comprehension task but could still require support in areas such as inferencing. Rather than relying solely on self-reported general skill levels, these methods help anticipate how different students engage with specific literacy demands, allowing DSP frameworks to more effectively identify and support multilingual learners in targeted areas.

Expanding EIRMs beyond vocabulary to writing placement could further enhance DSP frameworks. Traditional writing assessments often rely on binary placement decisions, such as assigning students to remedial or standard courses based on a single aggregate cut-off score, but EIRMs can evaluate performance across multiple dimensions, including organization, coherence, and grammar. Uto (2021) applied a multidimensional Item Response Theory (IRT) approach to rubric-based performance assessments and demonstrated that writing ability consists of distinct yet interrelated competencies rather than a single unified skill. While IRT provides a more detailed understanding of how students perform across these dimensions, EIRM analysis builds on this by examining how learner characteristics, such as multilingual background or motivation levels, influence performance on specific aspects of writing. Incorporating EIRMs into DSP would not only allow for a more precise assessment of writing proficiency but also help identify how multilingual learners and students with varying levels of motivation engage with different writing demands.