Adherence to language of instruction in Spanish-English dual language early elementary classrooms

ABSTRACT

This study examined teachers’ adherence to the language of instruction in dual language education (DLE) programs among teachers in 19 kindergarten through 3rd grade Spanish-English two-way DLE immersion classrooms (n = 5, 90/10 model; n = 14, 50/50 model). Teachers were observed toward the beginning and end of the school year to determine adherence rates of teacher talk by role, teacher talk by purpose, and teacher redirections of students not speaking the target language. Adherence for teacher talk was high across the school year but was lower for assistant teachers compared to lead teachers. Adherence for instructional talk and behavior management talk was also high, though rates were lower for behavior management talk in the fall for the 90/10 model. Redirections of students not speaking the target language were highest in the fall but negligible in the spring during English whole group instruction in the 50/50 model, and redirections were low but increased over time for teachers providing Spanish instruction. Redirection rates were especially low during small group or individual work time. Implications for future research and considerations for supporting teachers’ adherence to the language of instruction are discussed, including accounting for practices (e.g., translanguaging) reflecting updated views about language separation.

Educators and researchers have identified dual language education (DLE) programming as a promising approach for supporting the educational outcomes of students who are English Learners (ELs).¹ As described in the National Academies of Sciences, Engineering and Medicine’s (2017) report, there are two types of DLE programs: one-way dual language programs and two-way dual language programs. Both types of programs promote goals related to bilingualism, biliteracy, and biculturalism. Whereas one-way dual language programs primarily serve students from one home language group (i.e., EL students learning English and their home language or English proficient students learning in English and another language), two-way dual language programs jointly serve students who are ELs and students who are proficient in English. Two-way DLE programs are distinguished by the language of instruction specifications or goals, whereby programs often follow a 50/50 model (i.e., 50% instruction in the partner language and 50% instruction in English) or a 90/10 model (i.e., 90% instruction in the partner language and 10% instruction in English) that transitions gradually to a 50/50 model over the years. Importantly, use of the partner language at least 50% of the time has been defined as a central tenet among DLE programs (Lindholm-Leary, 2001), as students need opportunities to produce language to build their language skills (Saunders & O’Brien, 2006).

Thus, in considering DLE as a large-scale intervention, adherence to the language of instruction is a key aspect of delivering DLE programs with fidelity. Hulleman et al. (2013) define intervention fidelity as the extent to which a program is implemented as designed, especially regarding the intervention’s core components. This is distinct from implementation fidelity, which centers on “the contextual factors that support the implementation of the intervention’s core components” (Hulleman et al., p. 68). Intervention fidelity is also called adherence, compliance, integrity, faithful completion, or completeness, and differs from other dimensions of implementation, including dosage, quality, differentiation, monitoring, attendance, reach, responsiveness, and adaptation (Berkel et al., 2011). For DLE programs, the heart of fidelity lies in the need for congruence between a DLE program’s theoretical model for dual language instruction and actual classroom practice (Mora et al., 2001), including teachers’ actual language use in the classroom. Indeed, the Center for Applied Linguistics’ (CAL) Guiding Principles for Dual Language Education (Howard et al., 2018, p. 57) Instruction strand Principle 1A specifies the standard that “The program model and corresponding curriculum are implemented with fidelity.” Numerous researchers also point out the importance of fidelity in DLE programming (Li et al., 2016; Mora et al., 2001). Yet, there have been few empirical studies of intervention fidelity in DLE programming, despite the expansion of DLE programs across the U.S. (American Councils Research Center, 2021). Thus, for this study, we prioritize examining intervention fidelity in Spanish-English two-way DLE programming, focusing on adherence to the language of instruction, consistent with CAL Instruction strand Principle 1A.

Fidelity in DLE programs

The DLE fidelity studies that have been conducted have focused on the teachers’ language of instruction or instructional practices. Studies examining the language of instruction have focused on the extent to which teachers actually use the target language of instruction (i.e., English or the partner language, as indicated by the language model of instruction). Torres-Guzmán et al. (2005) collected self-report data from pre-k through 8th grade DLE teachers at 32 schools and showed that only two schools reportedly provided at least 50% of instruction in the partner language despite many of them offering the 50/50 model. Additional qualitative data suggested that either a substantial proportion of the schools was in essence offering second language enrichment programming rather than a true DLE program, or teachers misunderstood the nature of DLE programming (e.g., thinking that a 90/10 model is supposed to have 90% of instruction in English, and 10% in the partner language). Recently, Li et al.’s (2016) study in two-way 90/10 DLE programs found that only 57% of teachers used the partner language (i.e., 90% Spanish or 90% Russian) in 100% of the observations, while 40% of teachers used the partner language about 90% of the time. A case study of two DLE co-teachers showed they did not always adhere to their designated language but still successfully leveraged their own and their students’ language expertise, including using specific pedagogical strategies such as translanguaging to model strategic, purposeful, and dynamic practices common to bilingual communication and communities (Pontier & Gort, 2016).

Studies examining instructional practices in DLE programs have evaluated the quality of teachers’ practices as being consistent with the Sheltered Instruction Observation Protocol (SIOP; Echevarria et al., 2013), an evidence-based instructional model to address the academic needs of ELs consisting of eight interrelated components: lesson preparation, building background, comprehensible input, strategies, interaction, practice/application, lesson delivery, and review and assessment. Although SIOP is not without its critics (Crawford & Reyes, 2015), Echevarria et al.’s (2011) study investigated teachers’ fidelity to the SIOP lessons created by the research team in a sample of middle school classrooms, finding that teachers rated as high implementers appeared to have students who made the greatest academic gains. In addition, Li et al. (2016) reported an average score of three on a four-point scale on observer ratings of quality instructional reflecting SIOP practices.

The current study

While all dimensions of intervention fidelity are important to increase the likelihood that DLE programming will lead to desired student outcomes, for the purposes of this paper, we focus on one dimension of intervention fidelity: adherence to the target language of instruction. Consistent with Kerper (1986) and the CAL principles, we argue that adhering to the designated language of instruction, regardless of the language of instruction model, is critical to promoting students’ bilingualism, biliteracy, and academic success in English and in the partner language. Adherence to the language of instruction is related to another issue in DLE programs, namely the degree of separation between the two languages used in the classroom. Indeed, CAL Instruction strand Principle 1B, states that “Instruction incorporates appropriate separation of languages to promote high levels of language acquisition” (p. 58). We note, however, as does the CAL Principles document, that the degree to which languages should be separate in DLE classrooms is somewhat controversial, with some recent views discouraging the strict separation of languages to allow for use of strategies, such as translanguaging and language bridging, that leverage teachable moments and natural bilingual communication patterns (García et al., 2017; Palmer et al., 2014; Thomas & Collier, 2012; Urow & Beeman, 2012). Within the lens of fidelity, CAL Instruction strand Principle 1B can be interpreted as aligned with the concepts of adaptation (Berkel et al., 2011) and flexibility within fidelity (Carroll et al., 2007), which moves away somewhat from strict interpretations of fidelity, particularly in some real-world bilingual settings where language mixing is common. Regardless of one’s position on language separation, the first step is understanding how much each language is actually being used in DLE classrooms. As noted above, there is little research examining intervention fidelity in DLE programs, and past studies show relatively low fidelity to DLE instructional models – below what might be considered a reasonable degree of flexibility that still provides enough instruction in each language to achieve DLE program goals. Thus, we aimed to address this gap by focusing on adherence to the language of instruction in a descriptive study of Spanish-English two-way immersion DLE programming offered in two elementary schools.

Therefore, we examined three aspects of the language of instruction adherence that can inform professional development efforts for DLE teachers and efforts to promote intervention fidelity for DLE programming. Specifically, we were interested in adherence to the language of instruction for teachers’ talk. We examined this in two ways. First, we focused on adherence of teacher talk by teacher type, examining whether adherence to the target language of instruction is different for the lead teacher compared to the assistant teacher. Previous studies examining fidelity in DLE programming either focused on lead teachers only or did not specify if assistant teachers were included in the study. Given research showing variations in teacher and assistant teacher talk in English and Spanish in non-DLE settings (Durán et al., 2021; Early et al., 2014; Franco et al., 2019; Sawyer et al., 2016), this variable is worth exploring in DLE settings. Second, we were interested in adherence to different types of teacher talk, such as talk for instructional purposes and talk for behavior management. Again, none of the DLE studies reviewed above considered this dimension of teacher talk in their analysis of adherence. Subsequently, we wished to explore variations in adherence by talk type given that this dimension has been examined in studies of non-DLE early childhood classrooms explicitly designed to support EL children’s home language (Early et al., 2014).

Third, we wished to examine adherence to the language of instruction by noting whether teachers prompted or redirected students when they were using the non-target language. Previous research has documented that students often will respond in either English or Spanish during teacher instruction in Spanish (Arnold, 2017); however, it is not unusual for DLE students in two-way immersion programs to speak in English with peers when instruction is in the partner language (Arnold, 2017; Ballinger & Lyster, 2011; Bucknam & Hood, 2020; Li et al., 2016; Potowski, 2004). For example, Li et al. (2016) noted that about 28% of their observations had students using the partner language 100% of the time, though students spoke to the teacher in the partner language 90–99% of the time in 67% of the observations. Further, student language use varies by instructional setting. Specifically, Carranza (1995) found that students in 50/50 two-way immersion classrooms were more likely to speak in Spanish in whole-group instruction yet used English more when off-task. In addition, Li et al. (2016) observed that students used the partner language with peers during group work 100% of the time in only 20% of the observations, whereas students used the partner language 80–99% of the time in 43% of the observations. As noted by Arnold (2017), teachers set the expectations for language use in the classroom, meaning that if teachers always expect students to speak Spanish and consistently enforce this standard, students are more likely to speak in Spanish. In contrast, for teachers who had more relaxed expectations for student language use, students spoke either Spanish or English. Thus, we considered teacher prompts for students to speak in the target language relevant to adherence, which we examined in whole-group, teacher led instruction, as well as small-group instruction and independent work time. Again, the first step in a research program on language usage in DLE settings is descriptive work such as the current study, to find out the extent to which teachers are prompting students to use the target language.

Thus, the main research questions for the study were 1) to what degree do Spanish-English dual language education teachers show adherence to the target language of instruction, and how does adherence vary as a function of type of teacher (lead vs. assistant) and type of talk (instructional vs. behavior management), and 2) how does adherence manifest in teacher prompts for students to speak in the target language when students are using the non-target language? For both questions, we examined whether rates of adherence or redirections changed from the fall to the spring of the academic year. Based on past literature and the length of time our settings had offered DLE programming, we expected high adherence for lead teacher and instructional talk. Without past research to inform our expectations, we anticipated a moderate amount of adherence for assistant teacher talk and behavior management talk, based on our experience in DLE classrooms. Similarly, we expected frequent redirections for students to talk in the target language would occur in English instruction contexts relative to Spanish instruction contexts.

Method

Participants

Data were drawn from a larger, ongoing study examining early elementary school students’ experiences in Spanish-English two-way immersion DLE programs. Data were analyzed for 19 kindergarten through 3rd-grade teachers (Kindergarten: n = 6, 32%, 1st grade: n = 7, 37%, 2nd grade: n = 2, 11%, and 3rd grade: n = 4, 21%) from two schools in two different districts in the Southeastern United States. One school is located in a university town, where approximately 7% of the population is Hispanic/Latino, and the median household income was $77,000 (U.S. Census, 2020); in contrast, the other school is located in a town in a rural community, where approximately 53% of the population is Hispanic/Latino, and the median household income was $40,440. Teachers in five classrooms used the 90/10 instructional model, where students were primarily taught in Spanish by one teacher. The remaining 14 teachers used the 50/50 instructional model, where students were to receive half of their instruction in Spanish and half in English. Of these 14 teachers, four teachers taught half of their instruction in Spanish and half in English. The other 10 teachers were partnered in teaching pairs, where one teacher taught in English only, and the other teacher taught in Spanish only. In these classrooms, two groups of students switched from one teacher to another mid-day, each day, for instruction in the other language.

At the beginning of the school year, lead teachers were 36 years old on average (SD = 8.48), and most were female (n = 17, 89%). They described their race/ethnicity as follows: White (n = 8, 42%), Hispanic/Latino (n = 6, 32%), multiethnic (n = 4, 21%), and other (n = 1, 5%). The sample was relatively split regarding the number of years teachers had lived in the United States, with 8 (42%) living in the U.S. less than five years and 8 (42%) being born in the U.S. from U.S.-born parents. The remaining teachers (n = 3, 16%) either had lived in the U.S. for 6–10 years, 11–20 years, or were born in the U.S. from immigrant parents. Most teachers reported being fluent in English (n = 17, 89%) and Spanish (n = 14, 74%), with 2 (11%) also reporting they were fluent in an additional language. For educational background, most teachers reported either having a Bachelor’s degree (n = 13, 63%) or a Master’s degree (n = 6, 32%). For educational experience, on average, teachers reported being a teacher for 10.82 years (SD = 9.50), being at their current school or program for 4.55 years (SD = 6.27), teaching in a DLE program for 5.17 years (SD = 5.38), and teaching at their current grade/age level at the time of the study for 5.61 years (SD = 5.86).

Measures

Adherence to target language of instruction

An adapted version of the Language Use Inventory (LUI; LaForett et al., 2014) was used to measure teachers’ adherence to the target language of instruction. The LUI quantifies the amount of English and Spanish used in classrooms, as well as the purposes for each language. The LUI has been used successfully in three different program evaluation studies of classrooms designed to serve young Spanish-speaking children, where Spanish was expected to be used by teachers. Findings from these studies, which involved conducting LUI observations in close to 60 classrooms, have been disseminated via reports to state agencies and have been used to inform policy and practice for young ELs (Early et al., 2014, 2015, 2016). These initial uses of the LUI did not calculate inter-rater reliability, given the very small number of data collectors conducting observations. We used the four original LUI questions which ask the observer to rate the amount of English and Spanish used: 1) by the lead teacher when talking to students; 2) by the assistant teacher when talking to the students; 3) by teachers for instruction; and 4) by teachers for managing students’ behavior. These four questions were scored using a 5-point rating scale (1 = All English; 2 = Mostly English; 3 = An Equal Amount of English and Spanish; 4 = Mostly Spanish; and 5 = All Spanish). For the adapted LUI, we added two questions to rate the degree that the teacher redirected students who were not using the designated language of instruction during teacher-led instruction and during independent/group work time. These questions were also scored on a 5-point scale for the frequency of redirections (1 = Never; 2 = A Few Times; 3 = About Half the Time; 4 = Most of the Time; and 5 = All of the Time) and were scored as Not Applicable (NA) if students were always speaking in the target language or if the instruction setting did not occur.

Observers scored these questions over a series of five-minute observation windows. For LUI items reflecting teachers’ talk (i.e., lead teacher talk, assistant teacher talk, talk for instructional purposes, talk for behavior management purposes), adherence is measured by the percent of observation cycles in which the talk is in the language of instruction. This is calculated by counting the number of cycles where “most” or “all” of the talk was in the language of instruction, divided by the number of valid cycles for the observation. For example, in a classroom where Spanish was the target language of instruction, if lead teacher talk were at 95%, this indicates that data collectors observed the lead teacher talking in Spanish most or all of the time for 95% of valid cycles in the observation. For LUI items reflecting whether teachers redirected students who were not using the target language of instruction (i.e., student talk during whole-group activities, student talk during small-group activities or independent work), adherence is measured by the percent of observation cycles that teachers redirected students when they were not using the language of instruction. This is calculated by counting the number of cycles where teachers redirected students either most or all of the time, divided by the number of valid cycles for the observation in which students were not speaking the target language. For example, in a classroom where Spanish was the target language, if teacher redirections were 60%, this indicates that data collectors observed teachers redirecting students most or all of the time for 60% of valid cycles in the observation when students were not adhering to the language of instruction.

Procedures

Teacher recruitment took place in the fall of the 2018–2019 academic year, following university IRB and district research office approvals. Teachers completed a brief background survey to provide information on demographics and their training and experience as educators. Data collectors who were bilingual in English and Spanish were trained to use the LUI by the first author, and percent agreement within one during training ranged from 95% to 100%. Data collectors subsequently observed each teacher twice in the fall (late September – October) and twice in the spring (April), resulting in a total of 38 observations. The research team worked with the school administrators to identify 60-minute windows where English or Spanish was specified as the target language of instruction, and typically occurred during language arts/literacy, math, or science instruction. Data collectors were assigned to conduct observations during English instruction (50/50 model) or Spanish instruction (50/50 model and 90/10 model). For each visit, teachers were observed for 60 minute over 12, 5-minute LUI cycles. Percent agreement within one among data collectors for 20% of these observations ranged from 95% to 100% in the fall and from 97% to 100% in the spring. Our ability to field a team of multiple data collectors and establish strong inter-rater reliability during both training and data collection enhances our confidence in our study’s LUI data, relative to limitations noted in initial studies using the LUI.

Data analysis plan

We defined the threshold for the minimum number of valid cycles as 10 cycles out of the 12 possible. Adherence ratings were based on the percent of cycles where talk was “most” or “always” in the target language (Research Question 1) or where teachers redirected students to speak in the target language most or all of the time (Research Question 2), over at least 10 cycles. For our research questions, we computed the mean adherence percentages for each LUI variable separately for observations conducted in the fall and spring. We calculated the difference in mean adherence percentage from fall to spring to examine change over time. We disaggregated the data by the language of instruction model and the target language, resulting in three groups: teachers in the 50/50 model providing instruction in English, teachers in the 50/50 model providing instruction in Spanish, and teachers in the 90/10 model providing instruction in Spanish. Because this is a descriptive study of just a few teachers and classrooms, we deemed it inappropriate to conduct inferential statistical tests when comparing percentages observed across settings, teachers, and time.

Results

Research question 1: Adherence for teachers’ talk in the target language

Table 1 presents adherence for teachers’ talk in the target language of instruction for the amount of English and Spanish used: 1) by the lead teacher when talking to students; 2) by the assistant teacher when talking to students; 3) by teachers for instruction; and 4) by teachers for managing students’ behavior. We show the average percent adherence to the language of instruction broken down by observations in three groups of teachers for when the language of instruction was English (50/50 Model), Spanish (50/50 Model), and Spanish (90/10 Model). Average percent adherence is shown separately for fall and spring observations.

Table 1. Target language adherence for teacher talk across fall and spring observations.

	English Instruction (50/50 Model)				Spanish Instruction (50/50 Model)				Spanish Instruction (90/10 Model)
	Fall		Spring		Fall		Spring		Fall		Spring
Type of Talk	n	%	n	%	N	%	n	%	n	%	n	%
Lead Teacher	14	98.8%	14	100.0%	14	97.6%	14	97.6%	10	90.8%	10	100.0%
Assistant Teacher	14	85.7%	12	91.7%	14	80.9%	9	84.6%	10	74.1%	9	84.3%
Instructional	14	99.4%	14	97.6%	14	92.2%	14	91.0%	10	91.8%	10	100.0%
Behavior Management	14	98.5%	14	95.0%	14	94.5%	14	96.9%	10	78.3%	10	92.8%

Note: Percentage refers to the percent of observation cycles where teacher talk was in the language of instruction most or all of the time.

Lead teacher talk in the target language ranged from 90.8% to 98.8% in the fall, with teachers delivering instruction in English showing the greatest adherence and teachers providing Spanish instruction in the 90/10 model showing the least adherence. In the spring, adherence ranged from 97.6% to 100.0%, with teachers providing Spanish instruction in the 90/10 model and teachers providing English instruction in the 50/50 model showing 100.0% adherence. There was little variation from fall to spring for lead teachers in the 50/50 model regardless of whether instruction was in Spanish or English (i.e., change in percent adherence was less than 2.0% from fall to spring). However, for lead teachers providing Spanish instruction in the 90/10 model, adherence increased from 90.8% in the fall to 100.0% in the Spring, an increase of 9.2%.

There was more variability in adherence for assistant teacher talk. In the fall, average percent adherence ranged from 74.1% to 85.7%, with assistant teachers providing English instruction in the 50/50 model showing the greatest adherence and assistant teachers providing Spanish instruction in the 90/10 model showing the least adherence. In the spring, adherence ranged from 84.3% to 91.7%, with assistant teachers providing English instruction continuing to show the greatest adherence. In the spring, teachers providing Spanish instruction in the 50/50 model and teachers in the 90/10 model showed lower but comparable adherence rates. Rates of adherence increased from fall to spring for assistant teachers as follows: 3.7% increase for 50/50 English instruction, 6.0% increase for 50/50 Spanish instruction, and 10.2% increase for the 90/10 model. Notably, there were fewer valid observations of assistant teachers in the spring than the fall, due to assistant teachers either not being present or not speaking during the minimum number of cycles needed to meet the threshold for calculating a percent adherence score. This was especially true for Spanish instruction in the 50/50 model.

For adherence to different types of talk, teacher talk for instructional purposes showed adherence rates ranging from 91.8% to 99.4% in the fall, with teachers delivering instruction in English showing the greatest adherence, and both groups of teachers providing Spanish instruction showing adherence rates comparable to one another (92.2% for 50/50 and 91.8% for 90/10). In the spring, adherence ranged from 91.0% to 100.0%, with teachers providing Spanish instruction in the 90/10 model showing the greatest adherence and teachers providing Spanish instruction in the 50/50 model showing the least adherence. From fall to spring, adherence increased by 8.2% for teachers providing Spanish instruction in the 90/10 model. Adherence in both groups of teachers in the 50/50 model decreased by less than 2.0%.

Adherence for behavior management talk varied, ranging from 78.3% to 98.5% in the fall, with teachers providing instruction in English showing the greatest adherence and teachers providing instruction in Spanish in the 90/10 model showing the least adherence. In the spring, adherence ranged from 92.8% to 96.9%, with both groups of teachers in the 50/50 model showing comparably strong adherence, and teachers providing instruction in Spanish in the 90/10 model again showing the least adherence. Indeed, teachers in the 90/10 model showed an increase in adherence from fall to spring by 14.5%. Teachers providing instruction in Spanish in the 50/50 model also showed an increase in adherence (2.4%), whereas teachers providing instruction in English showed a decrease in adherence (3.5%).

Research question 2: Teachers’ redirection of students to talk in the target language

Table 2 presents adherence for the average percent of time teachers redirected students to talk in the target language during two instructional contexts: 1) whole-group activities and 2) small-group activities or independent work. We present the results broken down by three groups of teachers (i.e., English, 50/50 Model; Spanish, 50/50 Model; and Spanish, 90/10 Model) for observations conducted in the fall and observations conducted in the spring. Results are reported for observations where student talk in the non-target language was observed; observations where students always talked in the target language were not included in the analysis, as there were no opportunities for teachers to provide redirections; thus, the sample sizes for the observations are lower than the analyses of the teacher talk variables.

Table 2. Teacher redirections of students not speaking the target language across fall and spring observations.

	English Instruction (50/50 Model)				Spanish Instruction (50/50 Model)				Spanish Instruction (90/10 Model)
	Fall		Spring		Fall		Spring		Fall		Spring
Instructional Setting	n	%	n	%	N	%	n	%	n	%	n	%
Whole group instruction	6	49.2%	2	0.0%	14	17.1%	14	26.2%	10	2.6%	9	27.3%
Small group instruction/ independent work	9	0.0%	2	0.0%	14	7.1%	11	0.8%	10	0.0%	10	5.8%

Note: Percentage refers to the percent of observation cycles where teachers redirected students most or all of the time.

For whole-group activities, the extent that teachers redirected students to talk in the target language ranged from 2.6% to 49.2% in the fall, with teachers providing instruction in Spanish in the 90/10 model giving the fewest redirections and teachers providing instruction in English in the 50/50 model giving the most redirections. This pattern was reversed in the spring. Teachers providing instruction in English gave no redirections, and teachers providing instruction in Spanish in the 90/10 model gave redirections in 27.3% of instances when students were not speaking in the target language. From fall to spring, redirections decreased by 49.2% for teachers providing instruction in English, and increased by 24.7% for teachers providing instruction in Spanish in the 90/10 models. For teachers providing instruction in Spanish in the 50/50 model, redirections increased by 9.1% from fall to spring, going from 17.1% to 26.2%. Overall, teachers more frequently redirected students to use English compared to Spanish in the fall only, and teachers’ frequency of redirections to use Spanish increased over time.

During small-group instruction or independent work time, redirections ranged from 0.0% to 7.1% in the fall, with redirections occurring only by teachers providing instruction in Spanish in the 50/50 model. In the spring, redirections ranged from 0.0% to 5.8%. Again, teachers providing instruction in English did not give any redirections; teachers providing instruction in Spanish gave the most redirections, although redirections occurred in fewer than 10% of instances. For teachers providing instruction in Spanish, redirections decreased by 6.3% for teachers in the 50/50 model and increased by 5.8% for teachers in the 90/10 model. Overall, frequent redirections for students to use the target language during small group or independent work time occurred during Spanish instruction only.

Discussion

High-quality dual language programming is characterized by teachers adhering to the program’s specified proportions of target language use (e.g., CAL Instruction strand Principle IA), which is an important component of implementing DLE programs with fidelity. The present study examined adherence to the target language among teachers in Spanish-English DLE elementary schools, focusing on teacher talk and teacher redirections of students who are not speaking in the target language. Our data show that teachers’ adherence to the target language was high, especially for lead teachers. This was generally true regardless of whether teachers provided English instruction in a 50/50 model, Spanish instruction in a 50/50 model, or primarily Spanish in a 90/10 model. Yet, adherence was somewhat lower for assistant teachers across all program models in both the fall and spring, with somewhat lower rates generally during Spanish instruction. Adherence was also high for both instructional and behavior management talk types across all program models over the school year, except for somewhat lower adherence for behavior management talk at the beginning of the year for teachers in the 90/10 model.

Overall, the rates of target language adherence observed in our study are higher than those reported in previous research (Li et al., 2016; Torres-Guzmán et al., 2005); however, those studies were conducted with a large number of schools which could have resulted in more variability in those samples. On the one hand, such high rates of adherence are encouraging for ensuring that students are experiencing the hypothesized threshold of language exposure needed to gain bilingual skills. Evidence of high adherence to the target language can help alleviate some concerns raised about the dilution of language exposure in DLE programs. For example, some have expressed concern that there are subtle ways that the inequitable dominance, value, and prioritization of English over Spanish appears in 50/50 DLE programs, such as allowing certain patterns of code-switching, using English in more important instructional settings, and using Spanish for behavior management and lesser contexts (DePalmer, 2010; Palmer, 2009). Practices in DLE programs that elevate the value of and provide ample exposure to languages other than English are needed to ensure that DLE programs realize their equity-aligned goals to support ELs, given that such programs were originally designed to lift up EL students’ strengths and promote their success. This point is especially pertinent given concerns about inequitable access to DLE programs, where increased enrollment of non-ELs in DLE programs becomes an enrichment for advantaged students at the expense of ELs (Najarro, 2023; Valdez et al., 2016).

Still, it is worthwhile to consider whether the overall high adherence rates in teacher instructional talk in our sample may indicate that, while teachers’ talk is responsive to CAL Instructional strand Principle 1A, perhaps teachers are being too strict in their separation of instructional languages as cautioned in CAL Instructional strand Principle 1B and others in the field (Palmer et al., 2016; Thomas & Collier, 2012). In such a scenario, our teachers may have missed opportunities for using translanguaging, language bridging, and other strategies to deliver instruction flexibly but still with fidelity. Another possible explanation might be related to the bilingual skills within our teacher sample; not all our teachers were bilingual in English and Spanish, and this would clearly impede teachers’ capacity to leverage flexible bilingual pedagogical strategies and would result in higher adherence to the target language of instruction.

Nonetheless, given that past research has primarily focused on lead teachers, our study adds to the literature by contributing new knowledge related to assistant teachers in DLE settings. Future research on target language adherence should continue to include the role of the assistant teacher, including gathering information to understand whether assistant teachers have the needed fluency in the target language to meaningfully engage with students in the target language. Talk by assistant teachers who have limited fluency in the target language could have implications for the quality and amount of language exposure students receive. It will also be important to examine the relative proportion of lead teacher talk vs. assistant teacher talk to fully understand the amount of language exposure in the target vs. non-target language. Relatedly, lead and assistant teachers may employ a specific division of roles between them regarding teaching and supporting students, particularly during small group and independent work time, to ensure that each student’s specific learning needs are met; these arrangements between lead and assistant teachers, and the specific students with whom they are working, could influence real-time decisions about adherence to the target language (Pontier & Gort, 2016).

Further, while we generally found high adherence for teacher talk both for instructional talk and behavior management talk, more research is needed. For instance, lower adherence for behavior management talk in Spanish was observed in the fall in 90/10 (kindergarten and 1st grade) classrooms, which suggests that student age, acclimation to the new classroom environment, and the particular immersion context may play a role in teachers’ use of more English talk to address student behavior. Future research could also explore the extent to which assistant teachers are engaged in instructional vs. behavior management activities. While teacher use of student home language for behavior management purposes has been documented in non-DLE settings (Figueras-Daniel & Li, 2021; Franco et al., 2019; Sawyer et al., 2016; Wallace Jacoby & Lesaux, 2014), less is known about whether this is the case in DLE programs at the early elementary school level. To our knowledge, ours is the first study to examine lack of adherence to the target language for behavior management talk, and documents that adherence is lower in some Spanish instruction contexts – indicating that teachers may be more likely to use English to support behavior management for students who are stronger in English. Thus, this finding replicates, in principle, those from non-DLE settings showing that teachers are more likely to use students’ home language for behavior management purposes (e.g., Spanish). Future research should continue to explore such possibilities in DLE programs.

Across program models, teachers infrequently redirected students to speak in the target language, indicating low fidelity for holding students accountable for talk in the target language. This was especially true when students were in small groups or working independently. The highest rates of redirections occurred during the fall in 50/50 program models, particularly when the target language was English. However, frequent redirections to speak in Spanish during whole group time increased from fall to spring. Again, while there has been limited examination of this topic in past studies from an adherence perspective, our results are consistent with previous findings indicating that it is not unusual for students to talk in English during independent work time during Spanish instructional time (Arnold, 2017; Ballinger & Lyster, 2011; Bucknam & Hood, 2020; Li et al., 2016; Potowski, 2004). It may be more difficult for teachers to monitor the talk of all students during these activities, particularly if teachers are leading other activities (e.g., small groups) or are not in close physical proximity to students. To better understand the factors that influence whether teachers prompt students to speak in the target language, future research should examine the extent to which prompting changes over the course of the school year. One possibility is that teachers may be more diligent in prompting at the start of the school year, as our data suggest, at least in Spanish instructional contexts. In addition, student adherence to the target language as they gain fluency over the school year may result in fewer opportunities for teachers to redirect students. Further, teachers might be less strict with target language redirections for younger students. While the present study did not examine the role of peers, future research should consider the extent to which talk in the non-target language is peer-driven, particularly in peer-learning contexts that have less direct supervision by teachers.

It is also possible that the deviations from the target language we observed could reflect teachers’ intentional use of pedagogical strategies that aim to facilitate bilingual language acquisition and/or be responsive to individual students’ current levels of language proficiency or specific learning needs; there could also be variations in the extent to which teacher adherence varies in response to interactions between student proficiency, learning needs, and the specific content being taught. For example, translanguaging is a pedagogical strategy where teachers welcome and accept the full range of students’ linguistic repertoires to help them develop content knowledge, specific skills, and language (Vogel & Garcia, 2017). Examples of translanguaging in DLE settings (Garcia et al., 2011; Tabaku, 2023) could be perceived as lack of adherence to the target language. For example, teachers may purposefully coordinate instruction to preview content in one language, view it in the other language, and review it again in the original language; in this example, some of the teacher’s instructional talk would not be in the target language. Another example is accepting and valuing younger students’ responses in the non-target language, with the teacher repeating and responding in the target language; here, the teacher is being strategic in choosing not to redirect the student to speak in the target language. Related to this last example is the pedagogical strategy of language bridging (Urow & Beeman, 2012), where teachers have students engage in activities to compare and contrast the features of the target and non-target language. While our observational tool did not allow for us to document whether intentional translanguaging or language bridging practices were occurring, the pedagogical benefits as well as the positive affirming of students’ skills and identities as bilingual learners suggests that the value of these practices should be accounted for in conceptualizations of acceptable thresholds for target language adherence. Moreover, flexible pedagogies such as translanguaging and language bridging may also be equity-aligned strategies that ensure teachers have the agency and freedom to help ensure that all students get what they individually need in their learning environment. At the same time, it is important to be mindful that too much flexibility can potentially lead to imbalanced language exposure favoring English, would could create inequities for EL students.

Implications for practice

While overall we found high levels of teacher adherence to the language of instruction, past research has documented how some teachers receive mixed messages regarding how to successfully adhere to their DLE program’s expectations (Palmer et al., 2016). Thus, we focus on three ways our findings could have implications for DLE practitioners, focusing specifically on potential implications for improving assistant teachers’ adherence to the target language and also behavior management talk – the two areas from our data where we observed somewhat lower adherence rates.

First, our examination of adherence for teacher talk across teacher role and talk type showed generally high rates, with the exceptions of lower rates for talk by assistant teachers and talk for behavior management in the fall for teachers providing instruction in Spanish in the 90/10 model. To increase assistant teachers’ adherence, and adherence for behavior management talk, DLE programs might consider whether adding more visual supports (e.g., pictures of routines and classroom expectations) might be helpful. Visual supports are frequently recommended to support language-minority students attending contexts where instruction is primarily provided in the students’ second language (August, 2018), and visual supports are commonly included as part of professional development interventions aiming to equip teachers to support dual language learners (Castro et al., 2017; LaForett et al., 2012). For assistant teachers supporting students on academic tasks, they may need to use more visual supports such as pictures and graphic organizers for complex vocabulary and concepts, as has been recommended for elementary-aged learners (August, 2018). For behavior management talk, it is understandable that teachers may break from the target language to redirect students’ behavior to help them learn classroom rules and to address behavior challenges especially when safety is an issue. Here, picture chart schedules, cue cards, and other visual supports are typically recommended as part of social and emotional learning classroom approaches to promote positive student behavior (Jones et al., 2014). Visual supports can also include text in the target language to support communication between teachers and students to convey classroom expectations.

Second, we recognize that schools may have to make compromises when deciding how to staff classrooms. As a result, they may not be able to ensure that assistant teachers have high fluency in the target language of instruction, have the needed fluency in the academic language required for communicating about academic content, and/or have specialized training to teach the academic content. These situations prompt the need for working with assistant teachers to identify whether these or other factors, may be interfering with their use of the target language. As mentioned above, visual supports may be helpful to assistant teachers when supporting students on academic tasks. Also, collaborative lesson planning between the lead and assistant teacher should include opportunities for the lead teachers to model for assistant teachers strategies to engage with students on specific content. Such planning should also aim to identify what supports or resources assistant teachers may need, especially if they are likely to support students during times when the lead teacher is occupied (e.g., when teaching a small group lesson and other students are engaged in independent work). Given the various benefits of professional learning communities (PLCs; Brown et al., 2018), school leadership also might explore whether PLCs could be helpful to learn what professional development supports may help assistant teachers in their efforts to adhere to the target language. Emergent research examining PLCs in DLE settings has shown that it can improve teacher collaboration, deepen understanding of biliteracy strategies, and increase comfort in delivering curriculum content (Sánchez, 2018). These suggestions may also be useful for supporting monolingual teachers, like some of those in our sample, who were not fluent in Spanish, to create opportunities to use translanguaging strategies with students.

Finally, our data suggest there is need for ensuring that students are speaking in the target language during small group instruction and independent work time, absent of strategic use of the non-target language, such as with translanguaging or language bridging practices. As students progress through the grades, their day is more likely to include small-group instruction and independent work time, and it is important to ensure that all students get a high enough threshold of exposure to the target language to promote the development of language fluency. While it may not be feasible for teachers to attend to each student (e.g., delivering small group instruction, providing one-on-one support to a student), it remains important to identify strategies to prompt students to speak in the target language. A possible strategy may be to explore seating arrangements that leverage the skills of students with strong skills in both languages to support students with weaker skills in the target language, particularly to the extent that students may be more likely to socialize during independent work time. Other research shows that students in DLE programs who have strong skills in both English and Spanish are more socially connected with their classmates compared with students who have stronger skills in only one language (Franco et al., 2017); these students may be good candidates for strategic seating arrangements as they may be able to support peers to stick to the target language. Other strategies for leveraging student skills may involve peer modeling and monitoring to promote fellow peers to use the target language, consistent with a recent peer-mediated intervention that showed improvements in peer relationships and language development (Karem & Hobek, 2022).

Our data also showed teachers were more likely to redirect students to speak the target language during whole group time, as they may have more bandwidth to do so and/or they may use these instructional settings to model expectations for the class. However, there were observed differences in the extent to which teachers redirected students to speak in English compared to Spanish, with more redirections to speak English in the fall but redirections occurring throughout the school year for students to use Spanish. DLE programs should be mindful of such discrepancies, as they could implicitly convey that speaking in English is more highly valued than speaking in Spanish; such messaging is counterproductive to the goals of DLE efforts and may inadvertently reinforce negative messages about bilingualism in society more broadly. Encouragement and prompting from teachers for students to use the target language, particularly for Spanish, is especially needed given strong societal pressure to speak English (Kaveh & Lenz, 2022; Kaveh & Sandoval, 2020). Finally, it will be important for teaching teams within grade levels to work together to identify how strategies to prompt and redirect students to speak in the target language should account for student grade level and the language skills of the students in the classroom as a whole, to ensure that deviations from the target language are strategic.

Limitations

The primary limitations of this study are that it is descriptive, and solely focuses on data gathered over the course of one school year. We also did not collect specific contextual information that could help us better understand our findings. For example, we did not have complete data on the home language backgrounds for all students in the participating classrooms, despite our attempts to obtain this information from our school district partners (e.g., such disaggregation at the classroom level was considered potentially identifying). With such information, we might better understand the extent to which teachers’ adherence (or lack of) reflected efforts to be responsive to the language skills of students in their classrooms (e.g., intentionally engaging in translanguaging practices). In addition, we did not collect demographic background data on assistant teachers, and therefore do not have information about their educational background and experience, or their fluency in the target languages. This information would be useful in considering the extent to which either of these factors – or the combination of them – played a role in assistant teachers’ lower adherence rates. Knowing the specific pedagogical practices used in these DLE programs could have influenced target instruction adherence, but such an examination was outside the scope of the larger study from which these data were drawn. It is also possible that teachers could have changed their language practices in the presence of observers.

We also note that our primary measure, the LUI, is a relatively new instrument. As mentioned earlier, the LUI has been used in three program evaluation studies prior to the present study. While those program evaluation studies did not allow for calculating interrater reliability, the present study allowed us to calculate and demonstrate strong interrater reliability during the training phase and for the data collected, increasing our confidence in the observers’ ratings. Nonetheless, we acknowledge that as a newer instrument, there may be aspects of the LUI that could be improved. For example, future uses of the LUI may consider adding procedures to identify mechanisms that can capture additional information to contextualize the adherence data, such as teacher practices and the content of language interactions. Another consideration in future uses of the LUI may be to separate the ratings of instructional talk, behavior management talk, and redirections by lead and assistant teachers. Finally, while the LUI is intended to focus on teacher talk, it may be helpful to include data points on the extent to which students were talking in the non-target language to better contextualize the data on teacher redirections.

Conclusion

Taken together, our findings have implications for dual language programs’ efforts to create instructional contexts that are consistent with their goal to promote bilingual skill development. As per the CAL Principles (Howard et al., 2018), adherence to the language of instruction – both for teachers’ own talk and for the expectations teachers set for students’ talk – is critical for ensuring that students are exposed to the proportions of each target language needed for them to develop strong bilingual skills. Findings from this study indicate that school leaders and teaching staff should be particularly mindful of three areas that could compromise high-quality DLE programming: talk from the assistant teacher, talk for behavior management purposes, and teachers’ lack of redirecting students to speak in the target language, especially during small group and independent work time. Indeed, implementation science (Hulleman et al., 2013) emphasizes the importance of ensuring high-quality delivery of interventions and programs – which includes establishing that delivery is characterized by a high degree of fidelity – before determinations can be made about the success of such interventions and programs. Thus, school improvement plans should consider how to avoid or address adherence pitfall areas through professional development experiences, supports for delivering instruction, and leveraging the physical structure and social and linguistic composition of the classroom.

Acknowledgements

We acknowledge the contribution of Ashley Beauchamp for her support with the copyediting of the manuscript. We are grateful to the many undergraduate and graduate students, research staff, and school personnel, students, and families who contributed to the success of this project.

Disclosure statement

Dr. LaForett declares having served as a consultant to the Institute of Education Sciences, U.S. Department of Education, and received honoraria for this work. The remaining authors have no conflicts of interest to declare.

Additional information

Funding

This research was supported by the Institute of Education Sciences, U.S. Department of Education, through Grant R305A180063 to the University of North Carolina at Chapel Hill. The opinions expressed are those of the authors and do not represent views of the Institute or the U.S. Department of Education.

Notes

1. We recognize other terms used to describe children who are exposed to more than one language, such as dual language learners or multilingual learners. We are using the term English Learners (ELs) as this is what is currently used by the U.S. Department of Education to describe this population of students per the 2015 Every Student Succeeds Act (2015, Pub. L, 114-95).