Teachers’ instructional talk in a partly scripted language intervention targeting young second-language learners: developments over time

ABSTRACT

Despite huge investments in interventions designed to support oral language skills in early childhood and beyond, many of the interventions fail to identify impacts on children’s language learning. Programmes may have limited impact because they do not sufficiently succeed in supporting teachers’ instructional talk, and thus, more efficiently promote children’s language learning. The present study examined the extent to which 15 teachers in Norway implementing a language intervention programme designed to enhance students’ second-language learning in first and second grade demonstrated changes in their instructional talk over the 8-week programme. The programme consisted of scripted parts (labelling pictures of targeted words, repeated exposures), as well as soft scripted parts (word relations and definitions) and minimally scripted parts (narratives and explanations that extended the here-and-now). Teachers received professional development that qualified them to implement the programme. Analysis of modifications in teachers’ instructional talk was based on audio-recorded small-group interactions, comparing characteristics of teacher talk at the beginning and end of the 8-week programme. Results revealed that teachers’ instructional talk developed to include more word definitions and extended discourse, talk categories aligned with the less scripted parts of the intervention. Conversely, teacher talk during the scripted parts of the programme did not change.

KEYWORDS:

• instructional talk
• scripted components
• oral language learning
• interventions

From a social-interactionist theoretical perspective, language learning is embedded within and results from the complexity of oral interactions that children participate in during and outside school (Grøver et al. 2019; Halliday 1993; Hoff 2006; Ninio and Snow 1996). A basic quality of language intervention programmes within the social-interactionist approach is the teacher–student language-promoting interactions the programmes support (Dickinson and Porche 2011; Justice, Jiang, and Strasser 2018; Michener, Proctor, and Silverman 2018). In modern multilingual societies, identifying qualities of language intervention programmes that support teacher capacity to engage second-language (L2) learners in language-promoting interaction is crucial.

Features of interactions that support language learning have been adapted into intervention programmes by encouraging conversations using strategies such as open-ended questions (Biemiller and Boote 2006; Wasik, Bond, and Hindman 2006; Zucker et al. 2010), reinforcement and repetitions (Carlo et al. 2004; Roberts and Neal 2004), and expansions of children’s utterances and the use of sophisticated vocabulary words (Dickinson and Porche 2011). Wasik and Hindman (2011) found that an intervention programme based on teacher professional development improved the quality of teachers’ instructional talk (e.g. modelling language or providing children opportunities to use language), and that variation in the quality of teacher talk was associated with children’s language gains. Although most of these intervention programmes have examined diverse samples in prekindergarten and kindergarten, the strategies used to promote language learning are relevant in relation to L2 children who have not yet acquired the level of proficiency needed to understand the language of instruction in school.

Studies examining the efficacy of language intervention programmes on children’s oral language skills have reached inconsistent findings. Meta-analyses by Elleman et al. (2009), Marulis and Neuman (2010), and Rogde et al. (2019) revealed positive impacts of language interventions on some oral language measures but not on others. Other language intervention studies did not find effects on child outcomes despite huge investments (e.g. Yoshikawa et al. 2015). Furthermore, several intervention studies that comprise both monolingual and L2 learners report that interventions are equally effective for both groups (August et al. 2009; Carlo et al. 2004; Proctor et al. 2011), while other studies have demonstrated greater gains for L2 learners (Neuman, Newman, and Dwyer 2011; Snow, Lawrence, and White 2009).

Language learning and scripted intervention components

Key decisions in the construction of language intervention programmes for young learners concern the detailing of teacher manuals. Some intervention programmes offer detailed and manualized guidelines for how teachers should interact with students during activities; that is, the programmes are strongly scripted, while other intervention programmes are less detailed when it comes to how teachers’ instructions are supported. We use the term scripting in this article to refer to prescribed, read-aloud components, which are highly structured with distinct steps and instructions to complete. Scripted versions usually contain step-by-step instruction for teachers to follow from A to B (e.g. reading out each instruction directly from the manual), and do not open up for flexible use and extensions of the targeted activities and instructions. An example is the work by Rogde, Melby-Lervåg, and Lervåg (2016), who provided teachers with detailed, word-by-word lesson scripts to follow throughout the intervention. Other intervention programmes demonstrate a softer type of scripting, for example, by providing some intervention components with specified instructions for teachers to follow (i.e. scripted word definitions), while also inviting teacher adaptations that are not specifically outlined regarding what teachers should emphasise (e.g. Neugebauer et al. 2017). Even less guidelines for instructional talk is offered in intervention programmes that mostly provide teachers with examples of activities to select from within the targeted domain (Wasik and Hindman 2020). The teacher support in these latter programmes may include information on general language-enhancing strategies that teachers receive as part of professional development during workshops before the intervention, and that teachers are encouraged to use in a flexible way throughout the implementation of the intervention. For some intervention programmes, a mixture of scripted and less scripted components is included (Kim et al. 2017).

Although studies have revealed that teachers can implement a programme with high levels of adherence when it comes to delivering the number of lessons and targeted components in the intervention as intended, high implementation fidelity does not necessarily reflect quality instruction (Harn, Parisi, and Stoolmiller 2013; Justice et al. 2008). Less focus may be placed on the dynamic interactions that occur between the teacher and the students, and on the information provided to students (Bowne, Yoshikawa, and Snow 2016).

We identified three studies that shed light on the impact of intervention scriptedness on teachers’ instructional talk. First, Kim et al. (2017) examined the effectiveness of two conditions of the programme READS, an original version with scripted components and an adapted version in which teachers were offered guidance on how to modify specific parts of the programme. Findings showed that teachers in both conditions implemented the core elements with high levels of fidelity, while teachers in the adapted condition also afforded extensions and modifications in line with the programme’s aim. Students participating in the adapted version performed better on reading comprehension than their peers in the original condition, and were also provided with more exposure to core components of the programme (e.g. longer duration or extensions of activities).

Second, a recent study compared the quality of instructional talk by examining scripted components and teachers’ extensions of these components in relation to students’ language growth (Neugebauer et al. 2017). Quality of instructional talk was examined on two levels: scripted instructions of definitions and word relations, and extensions of these definitions and word relations (which were not scripted but in line with the programme’s aim). Findings revealed that adapted instructional extensions were most beneficial in enhancing students’ language growth.

Third, a large-scale Danish study compared three intervention conditions that differed in their level of scripting;: small-group and large-group conditions with softly scripted lesson plans linked to language and literacy activities, and a small-group condition that differed from the two others by being only minimally scripted. The latter intervention condition provided the same scope and sequence of instructions as for the other two conditions, but without a scripted plan to follow (Bleses et al. 2018). The impact was larger in the minimally scripted condition compared to the other two conditions. Furthermore, the minimally scripted version contributed to gains in children’s vocabulary, with larger gains achieved by second-language learners compared to their monolingual peers.

Interestingly, Dickinson (2011) suggested that interventions may not have the desired impact because they fail to change the ways teachers talk with children, and thus, the teachers’ capacity to support children’s language learning. Relevant in that regard is Smolkowski and Gunn’s (2012) work that showed teachers’ ways of providing opportunities for classroom learning were stable over time. Teachers’ instructional talk and ways of interacting with students reflect their cultural backgrounds and beliefs, and several studies have demonstrated that such talk may not easily be altered (Dickinson and Caswell 2007; Dickinson and Porche 2011; Hindman and Wasik 2012; Neuman and Cunningham 2009).

However, studies of whether teachers who receive professional development and participate in language intervention programmes are able to modify their instructional talk over time as they deliver the programme to their students, may not have distinguished sufficiently between developments in instructional talk that are associated with more scripted parts of the intervention programme and developments in instructional talk that result from, and reflect, less scripted intervention components. Examining teacher instructional talk aligned with components that differ in degree of programme scriptedness may help to identify qualities of teachers’ language-promoting interaction with students, and thus, also provide a basis for future intervention programmes designed to support student language learning.

The present study

The aim of the present study was to examine characteristics of the instructional talk of teachers who participated in a language intervention programme addressing young L2 students in first and second grade, and to assess how the teachers’ instructional talk developed from the early to the later parts of the intervention programme. The teachers participated in an intervention programme that consisted of scripted and less scripted components. Various terms have been used in the literature to describe degrees of scripting, either of an entire intervention programme or of specific components of a programme, such as overly and highly scripted (Parks and Bridges-Rhoads 2012), soft-scripted (Bleses et al. 2018), or largely unscripted (Neugebauer et al. 2017). Given the diversity of terms, and reflecting the characteristics of the intervention programme examined, we distinguished between three levels of programme component scriptedness: scripted components, softly scripted components, and minimally scripted components.

The study was part of a randomised controlled 8-week trial targeting oral language learning in young L2 learners in the early elementary school years. Children who received the intervention showed statistical significant improvements in oral language skills compared to children who were enrolled in the waiting-list control group, with an overall effect size of 0.35 immediately following the intervention (Heller, Lervåg, and Grøver 2019). The intervention programme included scripted components (various types of labelling with detailed guidelines for implementation), more softly scripted components (e.g. invitations to define a word in ways that were not prescribed or detailed in the guidelines to the teachers), and minimally scripted components (e.g. teacher invitations to children to extend the here-and-now, reflecting specific children’s interest and experience). We expected that teachers may use more of the less scripted components as they got to know the students in addition to students developing their second-language skills, even within the short time span of eight weeks. If this is the case, the design of intervention studies should take into consideration that teacher talk may develop during an intervention. We examined how the components of the intervention programme, varying in degrees of scriptedness, appeared in the teachers’ interactions with students, and whether we could identify a development in the teachers’ uptake of the various intervention components. More specifically, we asked the following two research questions:

RQ1. What characterized teachers’ instructional talk in the early and late parts of the intervention?

RQ2. Did teachers modify their instructional talk during the intervention?

Methods

Participants

The participants were 15 (13 female) teachers and 65 (37 female) students. All teachers worked as resource teachers in the children’s schools, but were not their classroom teachers, and therefore, did not know the students before the intervention study. Resource teachers were mainly responsible for second-language support and adaptive education in the schools. In Norway, a four-year college or university degree is required to qualify as a teacher. The participating teachers were experienced, with 11.79 years (SD = 10.57) of mean teaching experience in elementary schools. A total of 16 teachers participated in the randomised, controlled study (Heller, Lervåg, and Grøver 2019), but one teacher did not complete the requested audio recordings on which the present study was based, and thus, was not included.

Participating students were all L2 learners receiving the intervention. To be included in the study, the children scored in the lower range (1.5 SD below the mean or lower) on a standardised language test in Norwegian, Norsk Som Læringsspråk (Norwegian as a Language for Learning, NSL; Frøyen et al. 2011). The students had a mean age of 77.05 months (SD = 6.730 months). Although the children spoke various first languages, Urdu, Arabic, Somali, Kurdish, Turkish, and Tamil accounted for 66.1% (for a more detailed description of the total sample, see Heller, Lervåg, and Grøver 2019). The 15 groups had a mean of 4.40 students (SD = 0.63).

Ethical approval was obtained from the Norwegian Social Science Data services. The schools’ principals, the resource teachers and the parents of the participating children offered informed written consent.

Intervention

The main part of the programme had clearly scripted guidelines for implementation (the labelling of the visual material), while other programme components were presented as examples of language-supporting instructional talk (softly scripted), and others represented encouragement to extend talk beyond the here-and-now (minimally scripted). The lesson design aimed at enhancing oral language learning and affording students with a baseline of words for accessing the curriculum. As all students had low levels of L2 proficiency, words that were crucial for comprehending and learning curricular topics across subjects were chosen (i.e. acquire everyday content words already known by their monolingual peers). Explicit explanations of word meaning supplemented with visual material targeting semantic categories, such as ‘home’ or ‘hospital,’ and words useful for talking about them, such as for ‘home,’ apartment, elevator, and hall, and for ‘hospital,’ wound, patient, and ambulance, were the main content in each lesson. This usage of semantic categories aligns with studies documenting the efficacy of teaching word knowledge through conceptual categories, such as interventions targeting taxonomic usage (sorting words in hierarchical order; Neuman, Newman, and Dwyer 2011), and those examining the effects of taxonomic and thematic groups (Hadley et al. 2019; Pinkham, Kaefer, and Neuman 2014). Children received 64 30-min lessons over 8 consecutive weeks.

The professional development the teachers received comprised a one-day workshop (7 h) in which the rationale and the aim for the intervention was presented. The introduction of the programme’s components was combined with practice in small groups supervised by school psychologists. In the latter, activities that were part of the scripted components of the intervention were first demonstrated, followed by the teachers practicing the skills with supervision. These activities were then discussed in plenary to clarify and provide further examples if needed. Teachers were encouraged to draw students’ attention to word relations and definitions in a less prescribed form (the softly scripted components of the programme), as well as to invite children to tell about events in and outside the classroom (the minimally scripted components). Finally, teachers were told to adapt their instructional talk based on students’ progress throughout the intervention, and various examples were illustrated and discussed (i.e. how to extend their language use and level of abstraction). Support beyond this workshop was provided by telephone or email.

Procedures

Data collection. Teachers audiotaped their instruction and interaction with students during three lessons in the second (time point 1; t1) and seventh weeks of the intervention (time point 2; t2). We selected for transcription and analysis the first out of the three lessons in each respective category, as this was the lesson when teachers introduced for the first time the new semantic category to be worked with (‘home’ in week 2 and ‘hospital’ in week 7). The approximate first 20 min of each lesson were selected for transcription (t1; M = 18.13 min, SD = 3.45; t2; M = 17.34 min, SD = 1.96). This main part of each lesson consisted of the same framework in which the targeted category and its related words were introduced for the first time through labelling (scripted) accompanied by conversations extending the here-and-now children’s interest and experience (minimally scripted), in addition to talk about word relations and definitions (softly scripted).

The audio recordings were transcribed following the format of Codes for the Human Analysis of Transcripts (CHAT) from the Child Language Data Exchange System (CHILDES) (MacWhinney 2000). Utterances were identified by intonation, pauses, and conversational turns that indicated a break in the flow of speech. Transcriptions were limited to utterances in Norwegian. We were not able to identify with sufficient reliability individual student utterances in the audio recordings, and therefore, the transcripts did not distinguish between individual student speakers.

Coding scheme development and measures. The applied coding scheme builds on recent studies in which features of teachers’ instructional talk were examined. We sought to incorporate talk categories that captured a continuum of language skills, ranging from basic word level skills at one end to extended discourse in the other, describing the variety of linguistic input students may be exposed to in classrooms.

Talk categories developed to identify how intervention components varying in scriptedness appeared in the classroom

Talk categories aligned with scripted components of the intervention. To identify instructional talk that reflected the scripted parts of the intervention, we developed two categories that were derived and adapted from Bowne, Yoshikawa, and Snow’s (2016) conceptual information category (concrete information about words): simple labelling and elaborated labelling.

Simple labelling. This category was used for labelling pictures, either isolated by basic naming of the targeted word (sykepleier, ‘nurse’) or through basic sentences (dette er en sykepleier, ‘this is a nurse’).

Elaborated labelling. This category included information about the meaning of a word, as provided by describing illustrations, acting out the word, facts, examples, and information of what was not characteristic of the targeted word (e.g. dette er noe barn er redd for (sprøyte), ‘this is something children are afraid of (a syringe)’).

A code was given for each utterance that offered either simple or elaborated labelling. The coding also applied to utterances that repeated, reinforced, or reformulated a previous utterance coded as labelling.

Talk categories aligned with the softly scripted components of the intervention. To address instructional talk reflecting the softly scripted parts of the intervention, we developed two categories, building on definition and word relation categories from studies by Bowne, Yoshikawa, and Snow (2016), who distinguished between definitions and examples of word usage, and Neugebauer et al. (2017), who also examined definitions and word relations. These categories, conceptual definitions and usage-based examples of definitions, differed from the labelling categories previously described in the completeness of information provided in the teacher guidelines.

Conceptual definitions. This category encompassed information about word meaning provided through developmentally relevant definitions, such as sykepleier det er hun som passer på de syke, gir dem mat og medisiner/ (‘a nurse is someone who takes care of people that are sick, gives them food and medicine’). Teachers were not offered prescribed guidelines for conceptual definitions, and therefore, provided students with the teachers’ own adapted versions. When teachers spontaneously offered synonyms for targeted words, they were also included in the conceptual definitions category, such as in the following definition of a doctor: en lege er det samme som en doktor (‘a doctor is the same as a physician’).

Usage-based examples of definitions. We applied this category to identify context-related demonstrations of word use such as in the following examples: når man reiser til noen land så må man ta en reisevaksine (sprøyte) (‘sometimes when you travel to another country you need to get a vaccine’) or man tar også røntgenbilde av tennene hos tannlegen (‘we can also take an x-ray of the teeth at the dentist’). This category was distinguished from the former category by less formal descriptions to exemplify word usage.

In a similar vein as for the talk categories aligned with the scripted components of the intervention, each utterance identified as either a conceptual or usage-based example of definitions was counted. This also applied to utterances that repeated, reinforced, or reformulated an utterance identified as a definition.

Talk category aligned with the minimally scripted components of the intervention. To identify talk that reflected the minimally scripted components of the intervention, and that invited student contributions extending the here-and-now, we developed the talk category extended discourse which included sequences of talk that were explanations, narratives, or a combination. This category built on the definition of extended discourse provided by Snow, Tabors, and Dickinson (2001, 2): ‘talk that requires participants to develop understandings beyond the here and now and that requires the use of several utterances or turns to build a linguistic structure, such as in explanations, narratives, or pretend’. Usually, this type of talk emerged from the targeted words (e.g. a teacher asking a follow-up question, ‘Has anybody been to a hospital?’ or ‘What does vaccine mean? What do we use it for?’), but could also be introduced by a child who contributed a related story or introduced a word associated with the category. We marked in the transcripts where a sequence of extended discourse started and ended, and counted the number of utterances within the sequence. Given the second-language skills of the participating students, we included all utterances appearing within the segment, and utterances that were simple confirmations (e.g. utterances 6 and 9), negations, and repetitions of the previous utterance (e.g. utterance 5). The following example is an excerpt from a sequence identified as extended discourse, and was introduced by the teacher asking whether everybody had a home. The children followed up on her question by confirming and disagreeing. The teacher then stated that most people have a home:

Example 1

Extended Discourse Emerging from the Category Home

Table

1.	TEA:	De fleste har det.
		(Most people have a home.) (pause)
2.	TEA:	Men det er kanskje noen som ikke har det.
		(But maybe there is someone who does not have one.)
3.	TEA:	Vi skal snakke om hjem, om ting som vi har hjemme.
		(We’re going to talk about home, about things we have in a home.)
4.	TEA:	Noen bor i li (.) små hus, og noen bor i høye hus.
		(Some live in small houses; others live in apartment buildings.)
5.	Child:	Jeg bor i høyhus.
		(I live in an apartment building.)
6.	TEA:	Ja.
		(Yes.)
7.	TEA:	Og hvor bor +/.
		(And where does +/.)
8.	CHI:	Jeg bor også i høye hus.
		(I live in an apartment building, too.)
9.	CHI:	Jeg også.
		(Me too.)
10.	TEA:	Mm, og noen bor i et eget hus.
		(mm, and some live in their own house.)

Tokens and types of talk. Studies using tokens and types as indicators of talk quantity and talk diversity, respectively, have typically reported on large variations from one classroom to another, regarding teacher and student contributions (Grøver Aukrust 2007; Rydland, Grøver, and Lawrence 2014). Thus, we decided to use tokens and types as a control measure of talk quantity and diversity when examining degrees of scripted instructions.

Tokens. Talk quantity was measured by the number of tokens (the sheer number of words), using the FREQ option in Computerized Child Language Analysis (CLAN; MacWhinney 2000).

Types. Similarly, talk diversity was measured by calculating the number of different words (types) in each transcript. To develop a talk index that differed from the tokens index, we created a list of exclusion words against which all transcripts were screened. The list comprised numbers, vocal gestures (‘mm,’ ‘oi’), personal names, and the 150 most common words according to a Norwegian word frequency list. All morphological variants of a word were counted as one word type (e.g. jente-jenter ‘girl’ and ‘girls’). However, when words were derived from the same word root, but differed semantically, they were both included (e.g. mørk-mørket, ‘dark’ and ‘darkness’). All transcripts were examined visually to check for misspellings that might have occurred during transcription. As the transcribed recordings varied somewhat in duration, we developed a density measure for the tokens (number of tokens per minute) and the types (number of types per minute).

Reliability. The first author and a research assistant independently double-coded one transcript and discussed the applicability of the codes. When the two coders disagreed, the final coding criteria were arrived at through discussion between them. The first author and the research assistant then assessed the reliability of the coding scheme based on 20% of the recordings (six recordings in total, pre- and postrecordings). Interrater reliability was computed using Cohen’s kappa, and had a mean average of .84 (ranging between .78 and .93), which indicated substantial agreement (Landis and Koch 1977).

Analytical approach

Identifying characteristics of teachers’ instructional talk and possible modifications of teacher talk throughout the intervention were the aims of this study. To respond to RQ1, we examined the frequency of talk categories in the teachers’ instructional talk, in the early and late parts of the intervention. To respond to RQ2, we used paired sample t tests to estimate and compare modifications in instructional talk at the two time points. Although the research questions addressed teachers, the student groups’ talk developed during the intervention, and we included examination of changes in student groups’ talk as a check. Additionally, measures of tokens and types of talk for teachers and the student groups were included as check points when we examined the scripted and less scripted talk categories. These decisions allowed us to comprehend teachers’ discourse utterances more fully. As lesson duration varied across groups, measures of talk exposure were calculated per minute, and no statistically significant difference was found between the two time points (t(14) = 1.013, p = .328). Complete sets of data on the five categories of teachers instructional talk measures of linguistic input were retrieved for all teachers.

Results

To respond to the first research question, we present descriptive statistics on the occurrence of simple labelling, elaborated labelling, conceptual definitions, usage-based examples of definitions, and extended discourse produced by teachers (see Table 1). There was substantial variability among the teachers across all measures at both time points. As displayed in Table 1, predominant talk categories in the beginning of the intervention were simple and elaborated labelling, both aligned with the scripted components of the intervention. Talk categories aligned with softly scripted components (definitions and usage-based examples of definitions) and minimally scripted components (extended discourse) appeared much less frequently. In comparison, the less scripted components were often used by the end of the intervention (t2).

Table 1. Teachers’ talk categories per minute.

	Time 1					Time 2
Variable name	N	M	SD	Min	Max	M	SD	Min	Max
Simple labelling	15	4.72	2.48	1.99	11.93	3.60	1.98	1.63	7.92
Elaborated labelling	15	1.10	.65	.23	2.06	.75	.47	.10	1.71
Conceptual definitions	15	.16	.16	.00	.56	.40	.33	.00	1.05
Usage-based examples of definitions	15	.17	.19	.00	.62	.58	.49	.00	1.89
Extended discourse	15	.58	.63	.00	1.75	3.10	1.99	.46	7.41

The second research question sought to identify whether changes in the appearance of teachers’ instructional talk had occurred between the two time points. To assess these aspects, paired t tests were conducted. Statistically significant increases were found for conceptual definitions (t(14) = –.2.709, p = .017), usage-based examples of definitions (t(14) = –2.800, p = .014), and extended discourse (t(14) = –4.649, p = .000). However, neither simple labelling (t(14) = 1.663, p = .119) or elaborated labelling (t(14) = 2.018, p = .063) changed statistically significantly, although the latter approached a statistically significant decrease. Simple labelling and elaborated labelling were both talk categories aligned with the scripted components of the intervention, while the talk categories demonstrating statistically significant increases were aligned with the softly scripted and minimally scripted components.

As students’ language skills developed during the intervention (Heller, Lervåg, and Grøver 2019), we also examined how these talk categories appeared in students’ talk to check whether modifications in teacher talk reflects changes in student talk. Similar frequency patterns of talk categories were found in the students groups at t1 and t2 (see Table 2). Moreover, statistically significant positive correlations were found at both time points between teachers’ and student groups’ use of labelling (t1; r = .72, p = .002; t2; r = .68, p = .005) and extended discourse (t1; r = .90, p = .000; t2; r = .84, p = .000). We found positive statistically significant correlations at t2, but not at t1, between teachers’ and student groups’ use of conceptual definitions (t1; r = –.34, p = .220; t2; r = .55, p = .034) and usage-based examples of definitions (t1; r = .38, p = .161; t2; r = .66, p = .007). Elaborated labelling was positively statistically significantly correlated only at t1 (t1; r = .54, p = .036; t2; r = –.05, p = .880). Furthermore, examinations of changes in children’s talk using paired sampled t tests revealed statistically significant increases in conceptual definitions (t(14) = –2.487, p = .026), usage-based examples of definitions (t(14) = –2.851, p = .013), and extended discourse (t(14) = –4.375, p = .001) in the student groups, in addition to elaborated labelling (t(14) = –4.237, p = .001). In contrast, the use of simple labelling (t(14) = 2.111, p = .053) demonstrated a marginally statistically significant decrease.

Table 2. Student groups’ talk categories per minute.

	Time 1					Time 2
Variable name	N	M	SD	Min	Max	M	SD	Min	Max
Simple labelling	15	6.90	3.00	2.59	13.95	5.39	2.32	1.90	10.18
Elaborated labelling	15	.54	.42	.00	1.33	1.14	.75	.07	2.96
Conceptual definitions	15	.02	.05	.00	.20	.07	.10	.00	.32
Usage-based examples of definitions	15	.16	.19	.00	.76	.66	.63	.00	2.12
Extended discourse	15	.46	.55	.00	1.84	2.46	1.63	.10	5.07

Finally, we used the token density and types density by the teachers and the student groups as a check when examining and interpreting the prevalence and modifications of the scripted-aligned talk categories. Teachers produced, on average, 74.46 tokens per minute (SD = 22.71 tokens), and 6.54 types per minute (SD = 2.17 types) at t1. By the end of the intervention, the mean number of tokens per minute was 72.91 (SD = 13.60), and 8.08 types per minute (SD = 1.53). Students as a group produced, on average, 21.32 tokens per minute (SD = 5.78 tokens), and 2.73 types per minute (SD = 1.00 types) at baseline. By the seventh week, the students produced, on average, 31.32 tokens per minute (SD = 12.70), and 4.63 types per minute (SD = 1.70). Furthermore, teachers provided the same number of tokens at both time points (t(14) = .330, p = .746), while the diversity of words used was statistically significantly higher at t2 (t(14) = –3.624, p = .003). In comparison, students as a group demonstrated higher density of different word types (t(14) = –5.486, p = .000) and talked more (higher token density) by the end of the intervention (t(14) = –4.125, p = .001). Thus, teachers allowed themselves to be less scripted as students as a group developed into more talkative interaction partners.

Discussion

This study explored patterns of teachers’ instructional talk, and whether modifications of talk occurred in relation to degrees of scripting when participating in an oral language intervention for young L2 learners in the early elementary years. In response to RQ1, we found high usage of labelling in teachers’ instructional talk in the beginning of the intervention, while this feature was less prominent at t2, with definitions and extended discourse emerging. In response to RQ2, we demonstrated that teachers modified their linguistic input during the intervention period, as statistically significant changes in instructional talk were found in softly scripted and minimally scripted instructional talk (definitions, usage-based examples of definitions, and extended discourse). Scripted instructional talk (labelling) showed stability throughout the intervention. These results are important, as some teachers tend to simplify their speech with second-language learners, which does not provide children with language-rich environments needed for learning (Aarts et al. 2016).

Dickinson (2011) documented how difficult it is to change teachers’ practices, and Smolkowski and Gunn (2012) reported in a similar vein on high stability in instruction over time. We argue that the distinction between scripted and less scripted components in interventions is important to address, as it can provide insights into the mechanisms underlying change. Moreover, different levels of scripting offer teachers various directions for enriching the linguistic input in teacher–student exchanges, which can support students’ language learning. Interventions that include less scripted components in addition to prescribed parts provide teachers with the autonomy and flexibility that make it feasible to adapt to developing children’s progress and level of proficiency throughout the intervention. We believe the less scripted components of the present intervention allowed teachers to incorporate more advanced talk in their exchanges with students, while also inviting students to elaborate on their knowledge related to the lesson’s topic. In the literature, adaptations of instructions in interventions have recently received increased attention, emphasising a need to consider the possibility of flexible adaptations as an active ingredient in interventions. Combining different levels of scripted instructions can also give teachers greater agency, which may enhance the quality of the implementation and students’ outcomes (Kim et al. 2017). This perspective concurs with findings from Neugebauer et al. (2017), where adaptations in the form of extensions from the scripted instructions were the most beneficial component in promoting students’ language growth. In addition, when Bleses et al. (2018) compared different language interventions, varying in their degree of scripting, the less scripted version was the most effective compared to the two more scripted versions, with second-language learners having the greatest gains. Affording teachers greater agency by inviting them to adapt the intervention to the specific students they work with may provide teachers with more opportunities to fine-tune their interactions with their students, compared to when they are asked to follow a more scripted intervention manual.

Scripted intervention components can offer an opportunity for teachers to acquire a new skills toolkit that can be incorporated in their instruction over time. Therefore, it may lead to changes in their instructional practices, which can promote children’s language learning (Neugebauer et al. 2017). Furthermore, manuals provide examples of how to initiate language-rich interactions with students. For instance, some teachers may be less comfortable adapting the instruction to students’ needs during the intervention. This can be due to limited competence or experience with strategies promoting language learning. Accordingly, some teachers may profit more from following a set of scripts than others, for example, unexperienced teachers. Scripted guidelines may also be a support in an introductory phase of an intervention for teachers who do not know their students when they start working with them, such as was the case for the sample of teachers in the present study. In this regard, this sample differed from previous studies in which students’ classroom teachers usually were the main provider of the intervention (Rogde, Melby-Lervåg, and Lervåg 2016; Wasik, Bond, and Hindman 2006; Yoshikawa et al. 2015). Working with an unfamiliar group of students adds additional challenges when implementing an intervention; challenges such as building relations with students over a short amount of time, and having limited information about students’ background, knowledge, and level of proficiency. Thus, scripting can be the first step to improve the quality of instruction. Scripted intervention guidelines can be efficient as they require less preparation time and help teachers be consistent in implementing the programme, and thus, may ease evaluations of implementation fidelity.

However, scripting may restrict teachers’ professional autonomy and creativity, and reduce instructional quality if teachers become focused on following scripted guidelines rather than attending to individual students’ needs (Bleses et al. 2018; Parks and Bridges-Rhoads 2012). Taking time to explain when misunderstandings occur and provide additional information are instructional qualities that may be neglected if teachers are too preoccupied with following and getting through a scripted programme. Thus, strongly scripted guidelines can reduce the developmental appropriateness of teachers’ interactions with students. Finally, the optimal balance between scripted and less scripted parts of an intervention may vary depending on educational cultures in specific countries. Scandinavian countries typically value child-oriented learning activities, and manualized programmes are uncommon, emphasising instead that teachers should decide on practices and methods that fit the group they work with. This may be one reason why the less scripted version of a Danish programme was more effective than the scripted ones (Bleses et al. 2018).

An inclusion criterion in the study was that students demonstrated second-language skills in need of support (scored at least 1.5 SD below the mean). It is therefore reasonable to believe that the higher proportion of labelling talk categories in the early exchanges was helpful in providing a joint platform of understanding. Then, as teachers got to know their students better, and the students started talking more, the teachers apparently were more likely to demonstrate talk that was less scripted. To know a word, a learner needs to know the concept behind the label. Therefore, merely hearing the same word multiple times during a lesson is not enough to fully understand the concept, nor does it provide a richer language environment. According to the social-interactionist perspective on language learning that the present study is framed within, quality features of language develop through the dynamic interactions between teacher and students as questions and background experiences are brought into the conversation (Halliday 1993; Ninio and Snow 1996). Teachers’ ability of being responsive to and following up on students’ contributions promote language learning (Hansen 2018). Recent research on quality of linguistic input has shown that the most impactful language practices were those targeting conceptual information, definitions, and word relations (Bowne, Yoshikawa, and Snow 2016; Neugebauer et al. 2017; Silverman et al. 2014). Moreover, the present study indicated that features building on conceptual definitions and usage-based examples of these in combination with extended discourse can help teachers create enriched language environments for second-language learners. Affording students with semantic categories and semantically related words allows them to draw connections between what they already know and relate it to words within the same topic. This aligns with studies demonstrating the effectiveness of teaching children conceptually aligned categories (Hadley et al. 2019; Neuman, Newman, and Dwyer 2011; Pinkham, Kaefer, and Neuman 2014). Repeated exposure of this kind of linguistic input, therefore, may offer an opportunity to go beyond the simple word to learn the concept behind the words. These are all factors crucial in promoting second-language learners’ oral language skills. Using semantic categories can also make it easier for teachers to talk about definitions and include examples that connect with children’s’ own experiences. This may lead to topic-related stories and explanations offering children multiple opportunities to engage in conversations.

Limitations

This study had limitations. First, the small sample size reduced the statistical power of the analysis. Therefore, we were not able to explore whether variations across individual teachers’ instructional talk predicted child outcomes above the effect of receiving the programme that previously has been demonstrated (d = .35 when the intervention group was compared with a control group; see Heller, Lervåg, and Grøver 2019). Second, we were not able to reliably identify individual children’s utterances in the audiotaped observations of instructional talk, and therefore, coding of the children’s utterances was conducted at the group level. We cannot exclude the possibility that the student groups’ participation in labelling, definitions, and extended discourse at both time points in some groups may have resulted from specific students’ contributions, while in other groups reflected a participation structure in which all students joined. Video-taped observations would have allowed a more fine-grained analysis of teacher–student interactions at both time points, as well as the possible development of these interactions over the intervention period.

Conclusion

This study addressed characteristics of teachers’ instructional talk in the early and late parts of an intensive second-language intervention that had demonstrated substantial effects on young second-language learners’ oral language skills. We examined whether potential developments of teachers’ instructional talk over the time span were aligned with intervention components that differed in degree of scriptedness. The findings suggested that developments in teachers’ instructional talk, in particular, were demonstrated within talk categories aligned with the less scripted part of the intervention, while we found no changes in teachers’ labelling talk that was associated with the more scripted parts of the intervention. Promoting oral language skills in students relies on teachers providing language-rich environments where word knowledge, vocabulary, and conceptual knowledge are incorporated in the teachers’ instructional practices. Sparsity of concept-rich interactions in classrooms often results in fewer encounters with unfamiliar words underlying oral comprehension and reading comprehension, which are particularly important for second-language learners to reach proficiency. Very few studies have examined the extent to which teachers change their instructional talk during an intervention. This may have implications for the design of intervention programmes. Future research should address how modifications in teachers’ instructional talk in relation to scripting affect child outcomes. In particular, more emphasis should be placed on fine-grained methodology as the analysis of teacher–student interactions.

Acknowledgements

This research was supported by the Department of Education at the University of Oslo. The authors want to thank all the schools, teachers, and students participating in this project, the school district superintendent in the municipality, and a special thanks to former colleagues in the school psychology service.

Disclosure statement

No potential conflict of interest was reported by the author(s).