a Carnegie Mellon University, Department of Psychology, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA b Teachers College Columbia, Department of Human Development, 525W 120th Street, New York, NY 10027, USA c Carnegie Mellon University, Department of Statistics, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA d South Fayette Township School District, 3680 Old Oakdale Road, McDonald, PA 15057, USA

a r t i c l e i n f o

Article history: Received 14 August 2014 Received in revised form 17 April 2016 Accepted 23 April 2016 Available online 31 May 2016

Keywords: Off-task behavior Attention Time on-task Instructional design

* Corresponding author. Carnegie Mellon University PIER, 336E Baker Hall, 5000 Forbes Avenue, Pittsburg

E-mail address: [email protected] (K.E.

a b s t r a c t

This paper reports results from a large-scale observational study investigating attention allocation during instructional activities in elementary school students (kindergarten through fourth-grade). In Study 1, 22 classrooms participated while a more diverse sample of 30 classrooms participated in Study 2. This work investigated temporal patterns in children's attention allocation by collecting observational data on children's on- and off-task behaviors at three different time points (i.e., beginning, middle, and end of the school year) [Study 1]. We also investigated whether patterns of attention allocation changed as a function of student characteristics (gender, grade-level, SES), teachers' instructional design choices (instructional format and duration of an instructional activity), and school type (private, parochial, public charter schools) [Studies 1 & 2]. Children's patterns of attention allocation fluctuated over the course of the school year. Female students were found to be more on-task compared to male students. On-task behavior tended to decline as the instructional duration increased. The lowest rates of on-task behavior were observed while children were engaged in whole-group instructional formats. An effect of school type was found with higher proportions of on-task behavior observed in parochial schools. However, the effect of grade-level was equivocal across studies. These findings can begin to form a foundation for the development of research-based guidelines for instructional design aimed to support engagement among students in elementary classrooms.

Loss of instructional time due to off-task behavior is a well- established problem in educational settings, recognized by re- searchers (e.g., Carroll, 1963; Karweit & Slavin, 1981; Lee, Kelly, & Nyre, 1999) and practitioners (e.g., Lemov, 2010) for over a hun- dred years (cf. Currie, 1884 as cited in Berliner, 1990). Off-task behavior has been documented to negatively impact academic achievement, although the magnitude of this impact is unstable across studies (for reviews see Frederick & Walberg, 1980; Goodman, 1990). Off-task behavior is an indicator that students' attention is not focused on the instructional activity. The link be- tween the quality of attention and task performance has also been documented in the cognitive psychology literature (e.g., Choudhury & Gorman, 2000; Dixon & Salley, 2007; DeMarie-Dreblow & Miller, 1988). Despite considerable prior research on off-task behavior,

, Department of Psychology & h, PA 15213, USA. Godwin).

designing effective, easy to implement, and scalable interventions to reduce off-task behavior has been challenging. Roberts (2002) suggests that many existing interventions may be unsuccessful because they do not take into sufficient account the conditions that lead to off-task behavior. The goal of the present study is to eluci- date some of the factors involved in off-task behavior in elementary school settings.

1. Off-task behavior in elementary school students

There is a variety of reasons why loss of instructional time oc- curs in schools; these reasons include but are not limited to: weather (e.g., snow days), sudden onset interruptions (e.g., an- nouncements over the loudspeakers), and special events. However, student inattentiveness (i.e., off-task behavior during instructional time) has been found to be the biggest factor that accounts for loss of instructional time (Karweit & Slavin, 1981). Prior research esti- mates that elementary school students spend between 10% and 50%

mailto:[email protected]

http://crossmark.crossref.org/dialog/?doi=10.1016/j.learninstruc.2016.04.003&domain=pdf

www.sciencedirect.com/science/journal/09594752

http://www.elsevier.com/locate/learninstruc

http://dx.doi.org/10.1016/j.learninstruc.2016.04.003

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 129

of their time off-task in regular education classrooms (e.g., Fisher et al., 1980; Karweit & Slavin, 1981; Lee et al., 1999; Lloyd & Loper, 1986). Inattention or off-task behavior is a serious chal- lenge educators face. In fact, off-task behavior has been identified as one of the most common reasons for student referrals (Roberts, 2001). While eliminating all off-task behavior is not a realistic expectation, reducing rates of off-task behavior is an important goal given the challenges that off-task behavior causes for classroom management as well as the potential implications of off-task behavior on academic achievement.

A large number of prior studies have examined off-task behavior in elementary school students; however, the generalizability of prior work is limited due to its relatively narrow scope. For example, some studies observed a few classrooms within a single grade level (e.g., Lahaderne, 1968; Samuels & Turnure, 1974). Other studies involved a large number of classrooms (e.g., 18 to 25 classrooms), but only observed a small subset of students within each classroom (e.g., Fisher et al., 1980; Karweit & Slavin, 1981). Indeed, the wide range in estimates of off-task behavior reported in the literature may be partially attributed to the relatively small sample sizes utilized in prior research, as small samples are more susceptible to the influence of extreme data points. In order to establish a more comprehensive understanding of children's on and off-task behaviors during early and middle childhood, research examining children's patterns of attention allocation on a larger scale and across multiple grade-levels is clearly needed.

The present work makes a contribution to the field due to its size and scope: this work includes a large sample size both in terms of the number of classrooms which were recruited (e.g., Study 1: 22 classrooms, Study 2: 30 classrooms) as well as the number of children within each classroom who were observed (i.e., all stu- dents in attendance were included in the study). In contrast to some prior work which tended to focus on one particular grade or a small range of grade-levels, this work conducts observations across a broad range of grade-levels in elementary school (i.e., kinder- garten through fourth-grade).

A second contribution of this work pertains to the detailed coding scheme that was employed to provide a more nuanced examination of the sources of off-task behavior common in elementary school settings. Previous work examining off-task behavior in classrooms tended to treat off-task behavior as a uni- tary construct (e.g., Carnine, 1976; Frederick, Walberg, & Rasher, 1979; Karweit & Slavin, 1981). Consequently, the sources of chil- dren's off-task behavior remain underspecified. In the present study we delineate common types of off-task behavior including off-task peer interactions, self-distraction, and off-task behaviors directed towards aspects of the classroom environment. Identifying common types of off-task behavior in elementary school settings is critical as interventions targeting inattention will be successful only to the extent that they adequately address the source of children's off-task behavior. The types of off-task behavior measured in the present work were based on the results of a teacher survey. Thirty elementary school teachers were asked to rate the frequency of students' off-task behaviors on a scale from 1 to 4, where 1 in- dicates that the behavior occurs rarely and 4 indicates that the behavior is very frequent. Peers (M ¼ 3.21, SD ¼ 0.62) and self- distractions (M ¼ 2.62, SD ¼ 0.90) were identified by teachers as frequent sources of off-task behavior. Additionally, walking around the classroom (or being out of one's seat) was identified as a frequent off-task behavior by 14 of the teachers (M ¼ 2.50, SD ¼ 0.85). Off-task behavior relating to the environment was identified as another common source of distraction (M ¼ 1.83, SD ¼ 0.85). Studying off-task behaviors associated with the class- room environment is of particular interest because of the hypoth- esized link between off-task behavior and visual design features of

elementary school classrooms (e.g., Fisher, Godwin, & Seltmen, 2014; Godwin & Fisher, 2011). Primary classrooms often contain large amounts of stimulating sensory displays intended to increase children's motivation and engagement (Barrett, Zhang, Moffat, & Kobbacy, 2012; Tarr, 2004; Thompson & Raisor, 2013). However, there is no empirical evidence demonstrating that this design choice increases motivation and engagement. By contrast, large amounts of stimulating displays in classrooms have been described as “visual bombardment” (Bullard, 2010, p. 110) and a “cacophony of imagery” (Tarr, 2004, p. 1). Barrett and colleagues (Barrett et al., 2012) recently reported that, contrary to their initial hypothesis, high amounts of color (i.e., the degree and manner in which color was utilized in the classroom walls, furniture, and displays) was negatively associated with elementary school children's achieve- ment scores (although note that a follow-up study by Barrett, Davies, Zhang, and Barrett (2015) reported that very low amounts of color are also negatively associated with achievement, suggest- ing that there may be a level of visual stimulation that is optimal for classroom settings). Furthermore, there is recent experimental evidence supporting the notion that highly decorated learning environments may actually promote off-task behavior in young children and thereby decrease learning (Fisher et al., 2014; Godwin & Fisher, 2011).

Despite a large number of studies documenting rates of off-task behavior in elementary school students, there has been limited research examining the factors associated with off-task behavior. The present work aims to address this gap in the literature by conducting an exploratory study which examines four main research questions: (1) Do patterns of attention allocation change over the course of the school year? (2) Are student characteristics (e.g., gender, grade-level, and SES) related to children's attention allocation patterns? (3) Are instructional design strategies (e.g., instructional format and duration of instructional activity) related to children's tendency to engage in on and off-task behavior? (4) Is school-type related to children's attention allocation patterns? Below we briefly discuss how each of these factors may be related to patterns of attention allocation in elementary school students.

There has been limited investigation of the variability in chil- dren's patterns of attention allocation as a function of time (Martin et al., 2015). The idea that students' attentional capacity fluctuates over the course of the school day is a common belief in education circles. Indeed, teachers' report that they modify their instruction in response to fluctuations in students' levels of attention by avoiding challenging instructional activities following lunch or at the end of the school day (for discussion see Muyskens & Ysseldyke, 1998; Ammons, Booker, & Killmon, 1995). Observational research examining the effect of time of day on school children's classroom behaviors has found that inappropriate behaviors are more frequent in the afternoon compared to the morning (Muyskens & Ysseldyke, 1998). Similar findings have been obtained with chil- dren who have attention deficit disorders (e.g., Antrop, Roeyers, & De Baecke, 2005; Zagar & Bowers, 1983). Furthermore, studies us- ing performance-based measures of attention (e.g., paper and pencil visual search tasks in which participants are asked to locate and cross out a target object from a group of distractors) have found that performance on tests of attention is highest in the mid- morning and declines mid-day, although there is some variability in the observed attention patterns for preschool children and stu- dents in primary grades (e.g., Janvier & Testu, 2007). Although levels of attention have been found to oscillate over the course of the school day, it is currently an open question if and how patterns of attention allocation change across the school year. Specifically, the proportion of on-task behavior as well as the prevalence of different types of off-task behavior may fluctuate as children become more familiar with their teacher, school rules, peers, and

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143130

their classroom environment. For example, self-distractions may be more common in the beginning of the school year, but as time progresses children may become better acquainted with their classmates, leading to more off-task behavior directed towards peers in the middle and end of the school year. Another possible outcome is that children may habituate to their classroom visual environment. Therefore, off-task behavior directed at the visual features of the classroom may decrease from the beginning of the school year compared to the end of the school year.

Student characteristics also likely influence children's patterns of attention allocation. For example, prior work suggests that males exhibit more off-task behavior compared to females (Marks, 2000; Matthews, Ponitz, & Morrison, 2009). Consequently, it is of interest to investigate whether gender differences emerge in a large sample of elementary school children and to investigate whether the specific types of off-task behavior children engage in vary as a function of gender. Grade-level is another factor that may contribute to children's tendency to engage in off-task behavior as prior research has documented that the ability to engage in selec- tive sustained attention improves with age (e.g., Bartgis, Thomas, Lefler, & Hartung, 2008; for review see; Fisher & Kloos, 2016; White, 1970). Additionally, it is possible that specific types of off- task behavior may be more prevalent in younger grade-levels (e.g., self-distraction) while other types of distraction may be more pervasive across grade-levels (e.g. peers). Furthermore, rates of on and off-task behavior may vary as a function of SES. Prior work has found that executive function skills are typically weaker in children from a lower socioeconomic background (e.g., Wiebe et al., 2011); consequently, these children may have greater difficulty inhibiting distractions and thus be more likely to engage in off-task behaviors.

Instructional design choices (i.e., average duration of an instructional activity and instructional format) may also be related to off-task behavior in elementary school children. For example, the duration of an instructional activity may influence children's ability to attend to the ongoing instruction. Specifically, children may be better able to maintain a state of focused attention when instruc- tional activities are shorter in duration, this may be particularly true for younger children who are still developing the ability to efficiently regulate their attention. This possibility is consistent with studies suggesting that in laboratory settings the duration of focused attention increases gradually with development, from approximately 4-min in 2- and 3-year-old children to over 9-min in 5- and 6-year-old children (for review see Fisher & Kloos, 2016). Surprisingly, this issue has not been investigated systematically in genuine education settings. Consequently, there is a dearth of evidence-based guidelines that teachers can utilize to inform their instructional design choices. Teachers typically have considerable autonomy when determining how instructional time is allotted (Rettig & Canady, 2013). At the same time, better insight into what the optimal durations of instructional activity might be for main- taining high rates of on-task behavior in elementary schools would be valuable information for educators.

It is also possible that some types of instructional format (e.g., whole-group instruction, small-group instruction, etc.) are more likely to be associated with higher rates of off-task behavior than other instructional formats. There is evidence indicating that small- group instruction (when groups are formed on the basis of student ability) is more effective than whole-group instruction with regards to student achievement (for reviews see Lou et al., 1996; Kulik, 1992). However, the size of this effect is relatively small and vari- ability in effect sizes across individual studies is substantial. For example, Lou et al. (1996) reported that the effect size of small- group versus whole-group instruction ranged from �1.96 to 1.52, with an average effect size of 0.17. Thus, some researchers have argued that the large variability in effect sizes severely limits the

degree to which it can be concluded that small-group instruction is superior to whole-group instruction (Prais, 1998, 1999). The pos- sibility that the type of instructional format is related not only to achievement but also to off-task behavior has to our knowledge been largely unexplored (see Goodman, 1990); however, some prior research has documented higher rates of student engagement during teacher led activities (BTES as cited in Goodman,1990; Good & Beckerman, 1978; Ponitz & Rimm-Kaufman, 2011). The present work provides a nuanced examination of the relationship between off-task behavior and specific instructional formats (e.g., individual work, small-group or partner work, whole-group instruction at desks, whole-group instruction while sitting on the carpet) in elementary school students. Obtaining evidence to evaluate this possibility is important because it can empower teachers to choose instructional formats that are likely to optimize children's attention allocation.

Lastly, rates of on and off-task behavior as well as the types of off-task behavior that children engage in may vary as a function of school type (e.g., public schools, private schools, parochial schools). For example, schools may have different group cultures or norms regarding student behavior. These shared expectations may influ- ence children's patterns of attention allocation.

2. The present study

Within this paper, we report the results of two studies. Study 1 investigates the temporal patterns in children's attention allocation [Research Question 1]. Additionally, Study 1 examines whether student characteristics and specific instructional design choices are associated with patterns of attention allocation in elementary school children, both in terms of the overall amount of on and off- task behavior but also the form which off-task behavior takes [Research Questions 2 and 3 respectively]. Data for Study 1 contains a relatively homogeneous set of schools, as all participating schools were part of the same public charter school organization. Study 2 investigates whether the results obtained in Study 1 regarding the role of student characteristics and instructional design strategies on children's selective sustained attention can be generalized to a more diverse sample of schools. In Study 2, we collected data from a more heterogeneous sample of schools that varied in terms of the socioeconomic status of the student population [Research Question 3] as well as school type (i.e., public charter schools, private schools, and parochial schools) [Research Question 4].

2.1. Study 1: Temporal patterns of on- and off-task behavior across the school year

In order to address Research Question 1, Study 1 examined temporal patterns in children's attention allocation by collecting observational data on children's on- and off-task behavior over the course of the school year. To this end, we conducted observations at three different time points: the beginning, middle, and end of the school year. For each time point the proportion of on- and off-task behavior was modeled in order to determine if children's attention patterns fluctuated or remained stable throughout the school year. Additionally, we examined the possibility that the prevalence of certain types of off-task behavior may change over time. We also examined whether children's patterns of attention allocation changed as a function of student characteristics [Research Question 2] or based on instructional design strategies [Research Question 3].

2.2. Method

2.2.1. Participants Twenty-two classrooms participated in Study 1. Participating

3 Mind wandering, also referred to as “Stimulus independent thought” (Killingsworth & Gilbert, 2010, p. 932) and daydreaming (see Smallwood, Fishman, & Schooler, 2007), can be considered another form of off-task behavior. However, in the present study mind wandering was not included as a category of off-task behavior due to methodological concerns. This particular form of off-task behavior is not readily observable. Instead, mind wandering is typically assessed using thought sampling (Smallwood et al., 2007). However, it is unclear whether

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 131

classrooms were selected from 5 charter schools located in or near a Northeastern medium-sized city in the United States of America. Five grade-levels were recruited: kindergarten through fourth- grade.1 The distribution across the five grade-levels was as follows: 5 kindergarten classrooms, 4 first-grade classrooms, 5 second- grade classrooms, 2 third-grade classrooms, and 6 fourth-grade classrooms. The average class size was 21 students (10 males, 11 females). However, due to absences the average number of children observed in a single observation session was 18.9 children. The number of children observed per session ranged from 15 to 22 children.

2.2.2. Design and procedure The observation sessions were staggered across three time pe-

riods (Time 1: October 2011eDecember 2011, Time 2: February 2012eApril 2012, Time 3: May 2012eJune 2012). In order to minimize measurement error and obtain more stable estimates of on- and off-task behavior for each classroom, two observation sessions were conducted during each time period for a total of six observations per classroom. However, due to scheduling con- straints, in four of the 22 classrooms only five observation sessions were conducted. Thus, a total of 1282 observation sessions were conducted in Study 1. The average delay between observation sessions within a single time period was 3.6 days (the delay ranged from 1 to 14 calendar days). The average delay across time periods was 90.5 days. Each observation session lasted approximately one hour. The average number of observations per session was 346.13 and the average number of observations per child within a session was 19.27. For the purposes of the analyses reported below, chil- dren in each session were treated as a different set of students because student identifiers could not be collected. As a result, it was not possible to link observations across the six sessions. Therefore, a total of 2402 student-session pairs were observed. A student- session pair refers to a specific student observed by a coder within a specific session. However, treating the children within each session as a different set of students artificially inflates sta- tistical power. In order to mitigate this concern, a more conserva- tive alpha level was used in the analyses reported below. Specifically, the alpha level was adjusted to 0.0083 (the commonly accepted alpha level of 0.05 was divided by 6, the total number of observations, in order to more closely approximate the true size of the sample).

2.2.3. Coding on- and off-task behaviors All coders were trained in the Baker-Rodrigo Observation Method

Protocol (BROMP) for coding behavioral data in field settings (Ocumpaugh, Baker, & Rodrigo, 2012). All coders received extensive training consisting of coding videotapes and live observation ses- sions. Inter-rater reliability was established prior to the study proper. Kappa values ranged from 0.79 to 0.84. This level of reli- ability is in line with past classroom research coding off-task behavior, and exceeded the 0.75 threshold to which Fleiss (1981) refers to as “excellent” in field settings.

Field coding was conducted using the HART app for Android handheld computers (Baker et al., 2012), which enforces the BROMP protocol. Children were observed using a round-robin coding strategy, in order to reduce the tendency of observers to

1 Due to the nature of the IRB approval for this study, no identifying information, including date of birth, was collected. Consequently, we are unable to provide the mean age per grade-level. However, according to the National Center for Education Statistics (2001), the average age at school entry for U.S. kindergarten children is 5.5 years.

2 Two observation sessions were excluded from analysis due to disruptions that occurred during the observation session.

attend to more salient instances of off-task behavior. In the round- robin coding strategy, each child present in the classroom was observed individually in a prescribed order determined at the beginning of each session. Each time a child was observed, the observation lasted for up to 20 s. The first unambiguous behavior observed during the 20 s period was recorded. Quick glances were considered ambiguous behaviors, and coders were instructed to wait for an unambiguous behavior to occur (i.e., a behavior that was sustained long enough for the coder to identify and code the behavior based on the coding scheme described below). If a behavior was noted before 20 s elapsed, the coder proceeded to the next child, and a new 20 s observation period began. This process was repeated for the duration of the session; in this way each child was observed multiple times throughout the observation session. Coders observed the children using peripheral vision or side- glances in order to avoid looking directly at the student being observed. This technique makes it less apparent to the child that s(he) is being observed. This procedure has successfully and reliably captured students' behavior in prior work which assessed behavior and affect in middle and high school students (e.g., Baker, 2007; Baker, D'Mello, Rodrigo, & Graesser, 2010; Ocumpaugh et al., 2012).

Coders first classified children's behavior as on- or off-task using primarily the direction of the child's gaze. If the child was looking at the teacher (or classroom assistant), the instructional activity, and/ or the relevant instructional materials, they were categorized as being on-task. If the child was looking elsewhere, they were cate- gorized as being off-task. Contextual clues (i.e., teacher in- structions) were also taken into consideration when distinguishing between on- and off-task behaviors. For example, if a child was instructed to discuss an idea with a partner, coders would classify conversing with another peer as on-task unless the coders could clearly discern that the conversation was unrelated to the task.

If the child was classified as off-task, the type of off-task behavior was recorded. Five3 mutually exclusive categories of off- task behavior were logged: (1) Self-distraction, (2) Peer distraction, (3) Environmental distraction, (4) Walking, or (5) Other. Self- distraction entailed engagement with something on the child's own body, such as an article of clothing or an appendage, as well as episodes in which the child would close their eyes. Peer distraction was defined as interacting with or looking at another student when not directed to do so. Environmental distractions include interacting with or looking at any object in the classroom that was not related to the task at hand. Walking was operationalized as a student physically walking around the classroom when it was not consid- ered appropriate for the task. Other distractions included student behavior that was off-task but did not clearly align with the five aforementioned categories. A sixth category Unknown was also included to capture rare instances in which it was unknown

young children possess the metacognitive capabilities to self-report the occurrence or frequency with which they experience mind wandering. Relying on self-report may also be particularly problematic since mind wandering can occur without awareness (see Schooler et al., 2011; Smallwood et al., 2007). Additionally, young children may be particularly susceptible to demand characteristics which may diminish the accuracy of thought sampling procedures. A thorough discussion of mind wandering is beyond the scope of the present paper; however, interested readers can refer to Killingsworth and Gilbert (2010), Smallwood et al. (2007) or Schooler et al. (2011) for discussion of recent findings in the mind wandering literature.

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143132

whether the child was on- or off-task, and it was impossible or inappropriate for the observer to relocate in order to obtain a better view of the child. Unknown was also used when students left the classroom for various reasons (e.g., to use the restroom). Since the category Unknown is not informative in terms of children's patterns of attention allocation it was excluded from the analyses. The category Unknown accounted for 5% of the total observations coded.

As discussed previously, instructional format was included as a variable in order to examine whether certain instructional formats elicit differential amounts of off-task behavior. Six different instructional formats were coded: (1) individual work (when stu- dents are working on an activity individually, for instance when the teacher directs the students to complete a worksheet or other ac- tivity by themselves or without any help from classmates), (2) small-group work (when students are divided into smaller groups and each group works on an activity independently of the other groups; this category includes small-group work, partner work, and centers), (3) whole-group instruction at desks (when students are sitting at their desk or table while the teacher instructs the whole class in an activity), (4) whole-group instruction while sitting on the carpet (when students are sitting on the carpet or floor while the teacher instructs the whole class in an activity), (5) dancing, and (6) testing. It is important to note that observations were scheduled during instructional time; however, on rare occasions testing and dancing were observed resulting in a limited quantity of data gathered during these formats. As a result, testing and dancing were excluded from the analyses.

2.3. Results study 1

Overall, the percentage of off-task behavior averaged across all observations (29.30%) was within the range of previously reported estimates of off-task behavior in elementary school students (i.e., 10%e50%). In the present study, the three most common types of off-task behavior observed included: Peer distractions, Environment distractions, and Self-distractions. These three sources of distraction accounted for 85% of children's off-task behaviors. The percentage of children's on and off-task behaviors are reported in Table 1.

A three-level hierarchical logistic regression was performed with observations nested within students, nested within class- rooms using SAS PROC GLIMMIX. Random intercepts for students and classroom were included in the model. Separate models were run for on-task behavior and the three most common types of off- task behavior (peer distractions, environmental distractions, and self- distractions). For all models the test of the covariance parameters were significant indicating that both random intercepts are needed (all ps < 0.0001). All off-task models are conditional on being off- task. Fixed effects included Time of Year (beginning, middle, and end of the school year), Gender (Males, Females), Grade-level (Kindergarten, First, Second, Third, and Fourth grades), and Instructional Format (individual work, small-group, whole-group instruction at desks, whole-group instruction while sitting on the carpet). Using this approach our results reflect the effect of each independent variable controlling for the others and the correlations induced by the hierarchy.

Table 1 Percentages of on- and off-task behaviors in Study 1.

On-task behavior

70.70%

Sources of off-ta

Peer distractions Environmental distractions Self d

44.12% 24.74% 15.91

2.3.1. Question 1: Do patterns of attention allocation change over the course of the school year?

2.3.1.1. Time of year and on-task behavior 2.3.1.1.1. On-task behavior. Controlling for gender, grade-level,

and instructional format, a significant effect of time of year on rates of on-task behavior was found (F(2, 2341) ¼ 26.16, p < 0.0001). In general, on-task behavior was found to decline at the end of the school year with children engaging in significantly less on-task behavior at the end of the school year than at the beginning of the school year (t(2399) ¼ 3.13, p ¼ 0.0018, Odds ratio [OR] ¼ 1.13, 95% multiplicity-adjusted confidence interval [CI] ¼ [1.03, 1.23]) or the middle of the school year (t(2431) ¼ 7.17, p < 0.0001, OR ¼ 1.31, 95% CI ¼ [1.20,1.44]). Thus, children were 1.13 times more likely to engage in on-task behavior at the beginning of the school year compared to the end of the school year and 1.31 times more likely to be on-task in the middle of the year compared to the end of the school year. However, higher rates of on-task behavior were found in the middle of the school year compared to the beginning of the school year (t(2216) ¼ �4.32, p < 0.0001, OR ¼ 1.17, 95% CI ¼ [1.07, 1.27]).

2.3.1.2. Time of year and the sources of off-task behavior. Next, we examined temporal changes in the three most common types of off-task behavior: peer distractions, environmental distractions, and self-distractions. A significant effect of Time was found for all three types of off-task behavior. However, fluctuations across time were not uniform. The results for each type of off-task behavior are described below.

2.3.1.2.1. Peer distractions. Controlling for gender, grade-level, and instructional format, peer off-task behaviors, relative to all other types of off-task behavior, were found to vary across time (F (2, 2117) ¼ 7.20, p ¼ 0.0008). Children exhibited a significantly higher rate of peer off-task behavior, relative to all other types of off-task behavior, at the middle of the school year compared to the beginning of the school year (t(2073) ¼ �3.43, p ¼ 0.0006, OR ¼ 1.21, 95% CI ¼ [1.06, 1.38]) and the end of the school year (t(2151) ¼ 3.12, p ¼ 0.0019, OR ¼ 1.20, 95% CI ¼ [1.04, 1.38]). Spe- cifically, children were 1.21 (OR) times more likely to engage in peer distractions in the middle of the year compared to the beginning of the year and 1.20 (OR) times more likely to engage in peer dis- tractions in the middle of the year compared to the end of the school year. There was no significant difference in the frequency of peer off-task behavior, relative to all other off-task behaviors, be- tween the beginning and end of the school year (t(2133) ¼ �0.15, p ¼ 0.88, OR ¼ 0.99, 95% CI ¼ [0.86, 1.14]).

2.3.1.2.2. Environmental distractions. Controlling for gender, grade-level, and instructional format, a significant effect of time of year was found on rates of off-task behavior directed towards the environment relative to all other off-task behaviors (F(2, 2331) ¼ 8.08, p ¼ 0.0003). In general, off-task behavior directed toward the environment increased over time, as a fraction of all off- task behaviors, with children exhibiting a significantly higher rate of environment based off-task behaviors (relative to all other types of off-task behavior) at the end of the school year compared to the beginning of the school year (t(2430) ¼ �3.69, p ¼ 0.0002,

Off-task behavior

29.30%

sk behavior

istractions Walking off-task Other distractions

% 3.07% 12.15%

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 133

OR ¼ 1.33, 95% CI ¼ [1.10,1.59]). Children were 1.33 (OR) times more likely to engage in environmental distractions at the end of the school year compared to the beginning of the school year. Similarly children were 1.27 (OR) times more likely to engage in off-task behavior directed toward the environment in the middle of the school year compared to the beginning of the school year (t(2384) ¼ �3.22, p ¼ 0.0013, OR ¼ 1.27, 95% CI ¼ [1.06,1.52]). There was no significant difference in rates of environmental distractions (relative to all other off-task behaviors) between the middle and end of the school year (t(2195) ¼ �0.58, p ¼ 0.56).

2.3.1.2.3. Self-distractions. Controlling for gender, grade-level, and instructional format, a significant effect of time of year was found on rates of self-distraction relative to all other types of off- task behaviors (F(2, 2337) ¼ 32.35, p < 0.0001). The results for the proportion of Self-distractions across time points was similar to the observed pattern of results for Environmental distractions. The frequency of self-distractions relative to all other types of off-task behavior increased significantly across all three time points (all ps < 0.0013). For instance, self-distractions were 1.86 (OR) times more likely at the end of the school year compared to the beginning of the school year (t(2400) ¼ �7.98, p < 0.0001, OR ¼ 1.86, 95% CI ¼ [1.54, 2.24]).

2.3.2. Question 2: Are student characteristics related to children's attention allocation patterns?

In order to evaluate the putative relationship between student characteristics and children's attention allocation, the fixed effects for gender and grade-level were examined. The results are reported below for on-task behavior as well as the three most common types of off-task behavior (peer distractions, environmental distractions, and self-distractions).

2.3.2.1. Effect of gender 2.3.2.1.1. Gender and on-task behavior. Controlling for grade-

level, instructional format, and time of year there was a signifi- cant effect of gender (F(1, 2264) ¼ 58.46, p < 0.0001). Females were more likely to engage in on-task behavior than males, and this difference was statistically significant (t(2264) ¼ 7.65, p < 0.0001, OR ¼ 1.26, 95% CI ¼ [1.19, 1.34]). Female students engaged in 1.26 (OR) times more on-task behaviors than males. For example, at the end of the school year, in an average classroom, in the whole desk format, the average fourth-grade girl has 1.98 on task behaviors for every 1 off-task behavior, while the average boy has 1.57 on task behaviors for every 1 off-task behavior.

2.3.2.1.2. Gender and peer-distractions. Controlling for grade- level, instructional format, and time of year there was a signifi- cant effect of gender on the rates of peer-distraction (F(1, 2048) ¼ 21.37, p < 0.0001). Relative to all other off-task behaviors, females were more likely to engage in peer off-task behaviors compared to males, and this difference was statistically significant (t(2048) ¼ 4.62, p < 0.0001, OR ¼ 1.24, 95% CI ¼ [1.13, 1.36]). Thus when a female is off-task, her off-task behavior is 1.24 times more likely to be an off-task behavior directed towards her peers than for a male in the same grade-level, classroom, instructional format, and time of year.

2.3.2.1.3. Gender and environmental distractions. Controlling for grade-level, instructional format, and time of year there was a significant effect of gender on rates of off-task behavior directed toward the environment (F(1, 2369) ¼ 10.75, p ¼ 0.001). Relative to all other off-task behaviors, males were 1.22 [1.08, 1.38] (OR and 95% CI) times more likely to engage in environmental dis- tractions compared to females, and this difference was statistically significant (t(2369) ¼ 3.28, p ¼ 0.001).

2.3.2.1.4. Gender and self-distractions. Contrary to the results for peer distractions and environmental distractions there was no

significant effect of gender on the rate of self-distractions relative to all other off-task behaviors after controlling for grade-level, instructional format, and time of year (F(1, 2249) ¼ 0.20, p ¼ 0.65).

2.3.2.2. Effect of grade-level. The analyses for grade-level revealed no significant effect of

grade-level for on-task behavior (F (4, 17) ¼ 1.32, p ¼ 0.30) and no significant effect of grade-level for each type of off-task behavior (peer distraction p ¼ 0.16, environmental distraction p ¼ 0.41, self- distraction p ¼ 0.68) after controlling for gender, instructional format, and time of year.

2.3.3. Question 3: Are instructional design choices related to children's attention allocation?

In order to evaluate the role of instructional design strategies on children's tendency to engage in on and off-task behaviors, the fixed effects for instructional format were examined. The results are reported below.

2.3.3.1. Instructional format and on-task behavior. Controlling for gender, grade-level, and time of year there was a significant effect of instructional format on the rates of on-task behavior (F(3, 4803) ¼ 8.35, p < 0.0001). Recall that significance is based on the adjusted alpha level of 0.0083 in which the commonly accepted alpha level of 0.05 was divided by 6, the total number of observa- tions. The p-values are also adjusted for multiple comparisons us- ing the Tukey-Kramer method. In general, on-task behavior was most likely to occur during activities that took place in a small- group format (small-group vs. whole-group carpet: t(4803) ¼ 4.97, adj. p < 0.0001; small-group vs. whole-group desk: t(4803) ¼ 2.97, adj. p ¼ 0.016 [marginally significant based on the more conservative alpha level]; however, the contrast between individual and small-group instruction was not statistically signif- icant adj. p ¼ 0.03). On-task behavior was 1.23 [1.11, 1.37] (OR and 95% CI) times more likely in the small-group format compared to whole-group instruction on the carpet and 1.12 [1.02, 1.24] (OR and 95% CI) times more likely compared to whole-group instruction at desks. There was no significant difference in the rates of on-task behavior across the remaining instructional formats: individual, whole-group at desks, and whole-group on the carpet (all adj. ps � 0.08).

2.3.3.2. Instructional format and peer distractions. Controlling for gender, grade-level, and time of year there was a significant effect of instructional format on rates of peer off-task behavior (F(3, 3914) ¼ 70.87, p < 0.0001). Adjusted for multiple comparisons (using the Tukey-Kramer method) relative to all other off-task be- haviors, peer distractions were most likely to occur during activities that took place in an individual instructional format or small-group format compared to whole-group instruction at desks or on the carpet (all adj. ps < 0.0001). There was no significant difference in the frequency of peer distractions between individual and small- group formats (t(3914) ¼ 1.60, adj. p ¼ 0.38); nor was there a sig- nificant difference in the frequency of peer distractions (relative to other off-task behaviors) between the two whole-group instruc- tional formats: whole-group instruction on the carpet and whole- group instruction at desks (t(3914) ¼ 1.84, adj. p ¼ 0.26). Peer off- task behaviors were approximately twice as likely during individ- ual or small-group instruction compared to whole-group instruc- tion with odds ratios ranging from 1.70 (small-group vs. whole- group instruction on the carpet) to 2.16 (individual vs. whole- group instruction at desks).

2.3.3.3. Instructional format and environmental distractions. Controlling for gender, grade-level, and time of year there was a

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143134

significant effect of instructional format on rates of environmental distractions (F(3, 3914) ¼ 28.58, p < 0.0001). Adjusted for multiple comparisons (using the Tukey-Kramer method) relative to all other off-task behaviors environmental distractions were least likely to occur during activities that took place in an individual instructional format or small-group format; all adj. ps < 0.0001. There was no significant difference in the rates of environment based off-task behavior, relative to all other off-task behaviors, across these two instructional formats (t(3914) ¼ �1.50, adj. p ¼ 0.44, small- group:individual OR ¼ 1.19, 95% CI ¼ [0.88, 1.58]). Relative to all other off-task behaviors, environmental distractions were most likely to occur during whole-group instruction at desks and whole- group instruction on the carpet. Specifically, children were 2.07 [1.60, 2.68] (OR and 95% CI) times more likely to exhibit environ- mental distractions during whole-group instruction on the carpet than during individual instruction and environmental distractions were 1.75 [1.36, 2.26] (OR and 95% CI) times more likely during whole-group instruction on the carpet than during small-group work. Similarly, students were 1.94 [1.51, 2.51] (OR and 95% CI) times more likely to engage in environmental distractions during whole-group instruction at desks compared to individual instruc- tion and 1.64 [1.29, 2.09] (OR and 95% CI) times more likely to engage in environmental distractions during whole-group instruc- tion at desks compared to small-group work. There was no signif- icant difference in the rates of environment based off-task behavior across the two whole-group formats (t(3914) ¼ 0.74, adj. p ¼ 0.88, desk:carpet OR ¼ 0.94, CI ¼ [0.75, 1.18]).

2.3.3.4. Instructional format and self-distractions. Controlling for gender, grade-level, and time of year there was a significant effect of instructional format on rates of self-distraction (F(3, 3914) ¼ 139.14, p < 0.0001). Adjusted for multiple comparisons (using the Tukey-Kramer method) relative to all other types of off- task behavior, self-distractions were highest during whole-group instruction on the carpet compared to all other instructional for- mats (all adj. ps < 0.0001). For example, self-distractions were 5.24 [3.95, 6.90] (OR and 95% CI) times more likely during whole-group instruction on the carpet than during small-group work. The sec- ond highest rates of self-distraction, relative to all other types of off- task behavior, occurred during whole-group instruction at stu- dents' desks compared to small group and individual instructional formats (both adj. ps < 0.0001). There was no significant difference in the frequency of self-distractions, relative to all other off-task behaviors, between small-group and individual instructional for- mats (t(3914) ¼ �0.90, adj. p ¼ 0.81, small-group:individual OR ¼ 1.13, 95% CI ¼ [0.80, 1.58]).

2.4. Discussion study 1

Several novel findings have emerged from Study 1. First, the results from Study 1 indicate that children's patterns of attention allocation are not stable across the school year. Variations in the proportion of both on-task and off-task behavior were observed as a function of the time of year (i.e., the beginning, middle, or end of the school year). The findings suggest that on-task behavior declines by the end of the school year. The frequency with which children engaged in different sources of off-task behavior was found to fluctuate over time. Notably, peer off-task behavior was found to increase during the middle of the school year while self- distractions and environmental distractions increased at the end of the school year. The latter finding is notable because it contradicts the possibility that over the course of the school year children habituate to their environment and engage in less off-task behavior related to the environment e Study 1 suggests the opposite to be the case. Recall that in the present work we

observed children twice at each time point in order to minimize aberrations and obtain more stable estimates of children's behavior. However, future research should aim to replicate these findings using a greater density of observations at each time point.

Second, children's patterns of attention allocation also varied as a function of certain student characteristics (i.e., gender). For example, in line with prior research, females tended to engage in more on-task behavior than males. Interestingly, there were also differences in the types of off-task behaviors that children tended to engage in as a function of gender with males being more likely to engage in off-task behavior directed toward the environment (relative to other off-task behaviors) and females engaging in more peer off-task behaviors (relative to other off-task behaviors). In contrast to the results for gender, there was no significant effect of grade-level on children's on or off-task behaviors.

Third, the results from study 1 also suggest that instructional design choices are related to children's attention allocation, as rates of on and off-task behavior varied as a function of the instructional format (e.g., small-group, individual, etc.). On-task behavior was most likely to occur during small-group instruction. Different instructional formats also appear to elicit different types of off-task behavior. For instance, peer distractions were most common in in- dividual or small-group instructional formats, whereas environ- mental distractions were least likely to occur during these two instructional formats but were more likely to occur during whole- group instruction.

Taken together the results from Study 1 indicate that children's patterns of attention allocation are influenced by external factors such as time of year and instructional design strategies; however, there is also some evidence to suggest that certain student char- acteristics such as gender influence the rates of children's on-task behavior and to some extent the form in which off-task behavior may take.

2.5. Study 2: Assessing the generalizability of the relationship between student characteristics, instructional design strategies, and attention allocation

Study 1 contained a relatively homogeneous set of schools, as all 5 of the participating schools were part of the same public charter school organization. Thus, it is an empirical question as to whether the results obtained in Study 1 would replicate with a more diverse sample. The goal of Study 2 was therefore to examine whether the findings obtained in Study 1 could be generalized to other schools and student populations. Consequently, in Study 2 a more hetero- geneous sample of schools was recruited. The schools varied in terms of the socioeconomic status (SES) of the student population as well as school type (i.e., public charter schools, private schools, and parochial schools). Specifically, Study 2 investigated how pat- terns of attention allocation change as a function of student char- acteristics (gender, grade-level, SES) [Research Question 2], instructional design strategies (i.e., instructional format and average duration of an instructional activity) [Research Question 3], and school type (i.e., public charter schools, private schools, and parochial schools) [Research Question 4].

2.6. Method

2.6.1. Participants Thirty classrooms participated in Study 2. The classrooms were

recruited from 9 schools which included: 4 charter schools, 3 pri- vate schools, and 2 parochial schools. Data were also collected based on the percentage of students belonging to low-income families. The distribution of schools in each SES quartile was as

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 135

follows: 5 schools had 0e25% low-income students; 0 schools had 25e50% low-income students; 3 schools had 50e75% low-income students; 1 school had 75% and above low-income students.

Students were recruited from five grade-levels: kindergarten through fourth-grade. The distribution across the five grade-levels was as follows: 7 kindergarten classrooms, 7 first-grade classrooms, 7 second-grade classrooms, 3 third-grade classrooms, and 6 fourth- grade classrooms. Information regarding the class size and gender distribution was provided by 20 out of 30 participating teachers. The average class size was 18.71 students (9.25 males, 9.95 fe- males). Due to absences, the average number of children observed in a single observation session was 18.58 children. The number of children observed per session ranged from 14 to 23.

2.6.2. Design and procedure Each classroom was observed two times during the school year,

resulting in a total of 60 observation sessions. The observation sessions occurred between October 2012 and December 2012. The average delay between observation sessions within a single time period was 3.8 calendar days (the delay ranged from 1 to 10 days). Each observation session lasted approximately one hour. The average number of observations per session was 263.93 and the average number of observations per child within a session was 16.17. A total of 1113 student-session pairs were observed. Recall that a student-session pair refers to a specific student observed by a coder within a specific session. As mentioned in Study 1, treating the children within each session as a different set of students artificially inflates statistical power. As a result, a more conservative alpha level was employed; specifically the alpha level was adjusted to 0.025 (the standard alpha level of 0.05 was divided by 2, the number of observations, in order to more closely approximate the true sample size).

Coding of on-task behavior was identical to that in Study 1. Coding of off-task behavior was modified in one important way: Based on reports from coders, we split the environmental distrac- tions category into two separate categories to distinguish between distinct types of off-task interactions with the classroom environ- ment. Specifically, in Study 2 environmental distractions were defined more narrowly as looking at or interacting with elements of the classroom visual design (e.g., charts, posters, etc.); inappro- priate use of objects that were a part of the assigned task (e.g., playing with a pen instead of using it for its intended purpose) were coded as off-task behavior related to school supplies (whereas in Study 1 such instances were coded as environmental distractions). The decision to code supplies separately from environmental dis- tractions was based on the idea that interventions that target the classroom visual environment may be more feasible than in- terventions that address the inappropriate use of school supplies. Presumably the classroom environment can be streamlined (i.e., extraneous posters and charts can be removed) while school sup- plies are by definition tools that students need to complete their assignment and are therefore less amenable to intervention. Furthermore, prior laboratory research has indicated that the classroom visual environment (charts, posters, educational

Table 2 Percentages of on- and off-task behaviors in Study 2.

On-task behavior

73.58%

Sources of off-ta

Peer distractions Environmental distractions Supplies distraction

49.07% 12.10% 16.86%

displays, etc.) can be a source of distraction for young children (Fisher et al., 2014). Thus, it was of interest to determine whether this finding would generalize from the laboratory to real-world classrooms. Therefore, a total of six categories of off-task behavior were coded: (1) Self-distraction, (2) Peer distraction, (3) Environ- mental distraction, (4) Supplies, (5), Walking, and (6) Other. Similar to Study 1, the category Unknown was utilized when coders could not establish whether a child was on- or off-task. Since the category Unknown was not informative in terms of children's patterns of attention allocation it was excluded from the analyses. The category Unknown accounted for 3% of the total observations.

In addition to gender and grade-level, SES was also included as a student characteristic. For each school, SES was based on the per- centage of students from low-income families, obtained from the Pennsylvania Department of Education for the 2012e2013 aca- demic year. As such, lower values represent schools in a more affluent community.

Predictor variables pertaining to instructional design strategies included the instructional format and the average duration of an instructional activity. The same four instructional formats were included: (1) individual work, (2) small-group or partner work, (3) whole-group instruction at desks, (4) whole-group instruction while sitting on the carpet. The average duration of an instruc- tional activity (sec) was operationalized as the total duration of an observation session divided by the number of activities (defined as the number of transitions between instructional activities plus one). Duration of an instructional activity was included in order to investigate whether children were better able to maintain a state of focused attention when instruction consisted of small blocks of activities versus instructional activities that occurred over a longer duration. Transitions were noted every time the teacher paused instruction to change from one activity to another (e.g., tran- sitioning from working on a math problem to listening to a short story). In many cases, transitions coincided with a change in instructional format (e.g., switching from whole-group instruction to small-group instruction); however this was not always the case as transitions could occur without a change in instructional format (e.g., with children rotating from one small-group activity to another). Transitions were frequently marked by the teacher asking the children to get out new instructional materials (e.g., “Please get out your math binders”) or requesting that students change locations (e.g., “Please put your notebooks away and come to the carpet”).

2.7. Results study 2

Similar to Study 1, and to previous reports in the literature (Karweit & Slavin, 1981; Lee et al., 1999), 26.42% of children's observed behaviors were coded as off-task. In line with the patterns of attention allocation observed in Study 1, the most common types of off-task behavior were: peer distractions (49.07%), self-distrac- tions (11.56%), environmental distractions (12.10%) and supplies (16.86%); see Table 2 for the percentages of children's on and off- task behaviors.

Off-task behavior

26.42%

sk behavior

s Self distractions Walking off-task Other distractions

11.56% 3.17% 7.24%

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143136

In order to investigate changes in the students' patterns of attention allocation based on student characteristics, instructional design strategies (i.e., format and average duration of an instruc- tional activity), and school type, a three-level hierarchical logistic regression was performed with observations nested within stu- dents, nested within classrooms using SAS PROC GLIMMIX. Random intercepts for students and classroom were included in the model. Separate models were run for on-task behavior and the four most common types of off-task behavior (peer distractions, envi- ronmental distractions, self-distractions, and supplies). The tests of the covariance parameters were significant in all but one model (self-distractions) indicating that both random intercepts are needed (all other ps < 0.0001). All off-task models are conditional on being off-task. Fixed effects included: Gender, Grade-level, SES, Instructional Format, Instructional Duration, and School Type (private, charter, parochial).

2.7.1. Question 2: Are student characteristics related to children's attention allocation patterns? 2.7.1.1. Effect of gender

2.7.1.1.1. Gender and on-task behavior. Consistent with the findings of Study 1, there was a significant effect of gender after controlling for SES, grade-level, instructional format, duration, and school type (F(1, 1062) ¼ 23.96, p < 0.0001). Females had signifi- cantly higher rates of on-task behavior than males (t(1062) ¼ 4.89, p < 0.0001). The corresponding odds ratio is 1.26 with 95% CI [1.15, 1.37], indicating a 26% higher ratio of on-task behavior for females compared to males. As an example, for an average private school classroom of third graders at the beginning (i.e., duration ¼ 0) of whole desk instruction, an average female of average SES has 2.07 on task behaviors for every one off-task behavior, while an average male has only 1.65 on task behaviors for every one off-task behavior.

2.7.1.1.2. Gender and peer distractions. As in Study 1, there was a significant effect of gender on rates of peer off-task behavior after controlling for SES, grade-level, instructional format, duration, and school type, (F(1, 865) ¼ 7.41, p ¼ 0.007). Peer distractions were more frequent among females than among males (t(865) ¼ 2.72, p ¼ 0.007). Females were 1.24 [1.06, 1.45] (OR and 95% CI) times more likely to engage in peer off-task behavior (vs. other off-task behaviors) compared to males.

2.7.1.1.3. Gender and environmental distractions. In contrast to Study 1, there was no significant effect of gender on rates of envi- ronmental distractions, as a fraction of all other off-task behaviors, after controlling for SES, grade-level, instructional format, duration, and school type (F(1, 962) ¼ 1.60, p ¼ 0.21).

2.7.1.1.4. Gender and self-distractions. Based on the more con- servative alpha value of 0.025, the effect of gender on self-distrac- tion rates (as a fraction of all off-task behavior) was marginally significant e controlling for SES, grade-level, instructional format, duration, and school type (F(1, 970) ¼ 4.14, p ¼ 0.04). Self-distrac- tions were more frequent among males than among females (t(970) ¼ �2.03, p ¼ 0.04). Males were 1.27 [1.01, 1.59] (OR and 95% CI) times more likely to engage in self-distractions compared to females. Note this pattern of results is not consistent with Study 1 in which no significant gender difference for self-distractions was found.

2.7.1.1.5. Gender and supplies. A significant effect of gender on children's off-task behavior directed toward supplies e controlling for SES, grade level, instructional format, duration, and school type was found (F(1, 990) ¼ 16.71, p < 0.0001). Off-task behavior directed towards supplies was more frequent for males than for females (t(990) ¼ �4.09, p < 0.0001). Specifically, males were 1.54 [1.25, 1.90] (OR and 95% CI) times more likely to engage in off-task behavior oriented toward supplies compared to females.

2.7.1.2. Effect of grade-level 2.7.1.2.1. Grade-level and on-task behavior. In contrast to Study

1, a significant effect of grade-level on children's on-task behavior was found after controlling for gender, SES, instructional format, duration, and school type (F(4, 25) ¼ 3.84, p ¼ 0.01). Adjusting for multiple comparisons (using the Tukey-Kramer method), third graders engaged in significantly less on-task behavior compared to students in fourth grade (t(23) ¼ 3.38, adj. p ¼ 0.02, OR ¼ 0.41, 95% CI ¼ [0.19, 0.83]). Third graders also tended to exhibit lower rates of on-task behavior compared to first graders; however, this differ- ence was only marginally significant based on the more conserva- tive alpha level of 0.025 selected previously (t(25) ¼ 3.04, adj. p ¼ 0.04, OR ¼ 0.45, 95% CI ¼ [0.20, 0.97). All other grade-level comparisons were not statistically significant (all adj. ps � 0.23).

2.7.1.2.2. Grade-level and peer distractions. After controlling for gender, SES, instructional format, duration, and school type, the effect of grade-level on peer distractions, relative to all other off-task behaviors, was not statistically significant (F(4, 21) ¼ 1.78, p ¼ 0.17).

2.7.1.2.3. Grade-level and environmental distractions. After con- trolling for gender, SES, instructional format, duration, and school type, the effect of grade-level on environmental distractions, relative to all other off-task behaviors, was not statistically significant (F(4, 21) ¼ 1.25, p ¼ 0.32).

2.7.1.2.4. Grade-level and self-distractions. Controlling for gender, SES, instructional format, duration, and school type, the effect of grade-level on self-distractions, relative to all other off-task behaviors, was statistically significant (F(4, 1083) ¼ 5.71, p ¼ 0.0002). Adjusting for multiple comparisons (using the Tukey- Kramer method), fourth graders tended to engage in more self- distractions (as a fraction of all off-task behavior) than both kin- dergarteners (t(1334) ¼ 3.71, adj p ¼ 0.002; OR ¼ 2.22, 95% CI ¼ [1.24, 4.00]) and first graders (t(1220) ¼ �4.44, adj p < 0.0001; OR ¼ 2.36, 95% CI ¼ [1.39, 4.00]). All other grade-level comparisons were not statistically significant (all adj. ps � 0.15).

2.7.1.2.5. Grade-level and supplies. After controlling for gender, SES, instructional format, duration, and school type, the effect of grade-level on children's off-task behavior directed toward supplies, relative to all other off-task behaviors, was not statistically signif- icant (F(4, 21) ¼ 0.37, p ¼ 0.83).

2.7.1.3. Effect of SES. No significant effect of SES was found on children's patterns of

attention allocation after controlling for student characteristics, instructional design strategies, and school type (on task p ¼ 0.28, peer distractions p ¼ 0.29, environmental distractions p ¼ 0.70, self- distractions p ¼ 0.83, supplies p ¼ 0.90).

2.7.2. Question 3: Are instructional design strategies related to children's attention allocation? 2.7.2.1. Effect of instructional format.

A graphical overview of the distribution of on-task and off-task behavior across instructional formats is provided in Fig. 1.

2.7.2.1.1. Instructional format and on-task behavior. After con- trolling for gender, grade-level, SES, duration, and school type, a significant effect of instructional format on children's rates of on- task behavior was found (F(3, 2169) ¼ 27.61, p < 0.0001). Compared to all other instructional formats, the highest rates of on- task behavior occurred during individual and small-group in- struction; all adj. ps � 0.0001. The largest OR was 1.62 [1.30, 2.02] for small-group compared to whole-group instruction on the car- pet, and the smallest OR was 1.51 [1.28, 1.77], for individual compared to whole-group instruction at desks. There was no sig- nificant difference in rates of on-task behavior occurring between individual and small-group formats (t(2169) ¼ �0.83, adj. p ¼ 0.84, individual:small-group OR ¼ 1.07, 95% CI ¼ [0.86, 1.33]). On-task

Fig. 1. Depicts the full distribution of on and off-task behavior as it differs across instructional formats for the baseline group in Study 2 (i.e., male, third-grade, average SES, average classroom, private school). Please refer to the text for a discussion regarding the statistical significance of the differences displayed here.

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 137

behavior was least likely to occur during whole-group instruction and there was no significant difference between rates of on-task behavior during whole-group instruction on the carpet and whole-group instruction at desks (t(2169) ¼ �0.04, adj. p ¼ 1.00, desk:carpet OR ¼ 1.00, 95% CI ¼ [0.83, 1.22]).

2.7.2.1.2. Instructional format and peer distractions. After con- trolling for gender, grade-level, SES, duration, and school type, there was a significant effect of instructional format on children's rates of peer distractions, relative to all other types of off-task behavior (F(3, 1015) ¼ 47.09, p < 0.0001). As in Study 1, peer dis- tractions were most frequent during individual instruction and small-group formats, followed by whole-group instruction on the carpet and whole-group instruction at desks. All formats were significantly different from each other (all adj. ps � 0.0001; whole- group instruction on the carpet vs. whole-group at desks was marginally significant, t(1087) ¼ 2.69, adj. p ¼ 0.04) with the exception of individual and small-group formats in which there was no significant difference in the rates of peer distractions, rela- tive to all other off-task behaviors, across these two instructional formats (t(645) ¼ 0.75, adj. p ¼ 0.88). The largest difference was individual vs. whole-group at desks with an OR of 3.11 with 95% CI [2.32, 4.18].

2.7.2.1.3. Instructional format and environmental distractions. After controlling for gender, grade-level, SES, duration, and school type, there was a significant effect of instructional format on chil- dren's rates of environmental distractions, relative to all other types of off-task behavior (F(3, 1582) ¼ 18.69, p < 0.0001). Adjusting for multiple comparisons (using the Tukey-Kramer method) environ- mental distractions (relative to all other off-task behaviors) were most frequent during whole-group instruction on the carpet and whole-group instruction at desks (all adj. ps < 0.0001), and there was no significant difference between these two group instruc- tional formats (adj. p ¼ 0.99). Environmental distractions were almost 3 times more likely to occur in whole-group formats compared to any other instructional format (OR's ranged from 2.72 to 2.82). Furthermore, there was no significant difference between rates of environmental distractions during individual instruction or small-group formats (adj. p ¼ 1.0).

2.7.2.1.4. Instructional format and self-distractions. After con- trolling for gender, grade-level, SES, duration, and school type, there was a significant effect of instructional format on children's rates of self-distractions, relative to all other types of off-task behavior (F(3, 1582) ¼ 45.34, p < 0.0001). Adjusting for multiple comparisons (using the Tukey-Kramer method), self-distractions (relative to all other off-task behaviors) were most frequent during whole-group instruction on the carpet (all adj. ps � 0.007), fol- lowed by whole-group instruction at desks (both adj ps � 0.0001), and least likely to occur during individual and small-group in- struction which were not significantly different from each other (t(1582) ¼ 1.49, adj. p ¼ 0.44). The OR for whole-group instruction on carpet to whole-group instruction at desks is 1.68 [1.11, 2.54], and the ORs for whole-group instruction at desks to individual and small-group are 3.13 [1.95, 5.05] and 4.48 [2.73, 7.35] respectively.

2.7.2.1.5. Instructional format and supplies. After controlling for gender, grade-level, SES, duration, and school type, there was a significant effect of instructional format on children's rates of off- task behavior oriented towards supplies, relative to all other types of off-task behavior (F(3, 1582) ¼ 10.32, p < 0.0001). Adjusting for multiple comparisons (using the Tukey-Kramer method) supplies (relative to all other off-task behaviors) were most frequent during whole-group instruction at desks and during small-group instruc- tion (all adj. ps � 0.002 with the exception of the contrast between the small-group format and individual instruction which was marginally significant based on the more conservative alpha level of 0.025; t(1582) ¼ �2.70, adj. p ¼ 0.04). There was no significant difference in the rates of off-task behavior directed toward supplies (as a fraction of all off-task behaviors) between the small-group format and whole-group instruction at desks (t(1582) ¼ �0.77, adj. p ¼ 0.87). Additionally there was no significant difference be- tween whole-group instruction on the carpet and individual instructional formats (t(1582) ¼ 1.30, adj. p ¼ 0.57). Off-task behavior directed towards supplies was more than twice as likely in small-group (OR ¼ 2.22, 95% CI ¼ [1.25, 3.97]) and in whole- group instruction at the desks (OR ¼ 2.51, 95% CI ¼ [1.50, 4.20]) than in whole-group instruction on the carpet. Similarly, compared to the individual instructional format off-task behavior directed towards supplies was 1.69 [1.02, 2.80] (OR and 95% CI) times more likely in small-group formats and 1.91 [1.30, 2.81] (OR and 95% CI) times more likely in whole-group instruction at the desks.

2.7.2.2. Effect of instructional duration. Recall that the average duration of an instructional activity (sec)

was operationalized as the total duration of an observation session divided by the number of activities (defined as the number of transitions between instructional activities plus one). In many cases, transitions coincided with a change in instructional format; however, transitions could occur without a change in instructional format (e.g., rotating from one small-group activity to another). The duration of activities observed ranged from 6.7 to 39.7 min with a median duration of 12.8 min, 25% of the activity durations were less than 10.6 min and 25% of the activity durations were longer than 17.1 min.

2.7.2.2.1. Duration and on-task behavior. After controlling for gender, grade-level, SES, instructional format, and school type, a significant effect of instructional duration (sec) on children's rates of on-task behavior was found (F(1, 954) ¼ 7.04, p ¼ 0.0081). The slope estimate is �0.0174 (95% CI [�0.0301, �0.0047]) log odds units per additional minute of activity length. For example, comparing an activity of 10 min in length to one of 30 min in length, the estimated OR for on task behavior is 1.42 times higher for the shorter instructional activity (i.e., 10 min) compared to the longer instructional activity (i.e., 30 min). This finding suggests that elementary school children are better able to maintain a state of

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143138

focused attention when instruction consists of relatively short (e.g., 10 min) blocks of instructional activities compared to instructional activities that occur over a longer duration (e.g., 30 min or longer). This finding is consistent with laboratory studies which point to a gradual increase in the duration of focused attention during early childhood (Ruff & Lawson, 1990; Sarid & Breznitz, 1997). See Fig. 2 for a graphical depiction of the effect of instructional duration on on-task behavior across the different instructional formats.

2.7.2.2.2. Duration and peer distractions. After controlling for gender, grade-level, SES, instructional format, and school type, the effect of instructional duration (sec) on children's rates of peer off- task behavior, as a fraction of all other off-task behaviors, was not statistically significant based on the more conservative alpha level of 0.025 (F(1, 392) ¼ 2.91, p ¼ 0.09).

2.7.2.2.3. Duration and environmental distractions. After con- trolling for gender, grade-level, SES, instructional format, and school type, the effect of instructional duration (sec) on rates of environmental distractions, as a fraction of all other off-task be- haviors, was not significant (F(1, 731) ¼ 2.31, p ¼ 0.13).

2.7.2.2.4. Duration and self-distractions. After controlling for gender, grade-level, SES, instructional format, and school type, the effect of instructional duration (sec) on rates of self-distractions, as a fraction of all other off-task behaviors, was marginally significant based on the more conservative alpha level of 0.025 (F(1, 1582) ¼ 4.89, p ¼ 0.027). However the estimate was negative (�0.033 log odds units per additional minute of activity length) suggesting that as the duration of a lesson increased the rate of self- distractions decreased.

2.7.2.2.5. Duration and supplies. After controlling for gender, grade-level, SES, instructional format, and school type, the effect of instructional duration (sec) on rates of off-task behavior directed towards supplies, as a fraction of all other off-task behaviors, was not significant (F(1, 770) ¼ 0.07, p ¼ 0.79).

Fig. 2. Displays the fraction of on-task behaviors for each instructional format at two durations (10 min and 30 min) for the baseline group in Study 2 (i.e., an average third- grade male with average SES in a private school in an average classroom). Error bars represent the 95% Confidence Intervals. Note that females in the same grade, class- room, SES, and school type show an analogous pattern.

2.7.3. Question 4: Is school type related to children's attention allocation? 2.7.3.1. Effect of school type.

Recall that three different types of schools were recruited for the present study: Parochial schools, Private schools, and Charter Schools (public). The analyses below examine potential differences in the rates of children's on-task behavior as a function of School Type as well as differences in the frequency of four types of off-task behavior: Peer distractions, Environment distractions, self-distrac- tions, and Supplies.

2.7.3.1.1. School type and on-task behavior. After controlling for gender, grade-level, SES, instructional format, and duration, a marginally significant effect of school type on children's rates of on- task behavior was found; F(2, 24) ¼ 3.81, p ¼ 0.04 (based on the more conservative alpha level of 0.025). Parochial schools had higher proportions of on-task behavior than private schools (t(24) ¼ 2.73, p ¼ 0.01). Children attending parochial schools were 2.10 [1.07, 4.14] (OR and 95% CI) times more likely to engage in on- task behavior compared to children attending private schools. For example, for an average third-grade boy of average SES in whole- group instruction at desks with the median activity duration in an average classroom at a parochial school the expected on-task rate is 2.77 on-task behaviors for every 1 off-task behavior, while a corresponding boy from a private school would have 1.32 on-task behaviors per one off-task behavior. Rates of on-task behavior were not significantly different between charter schools and private schools nor was there a significant difference between charter schools and parochial schools (both ps � 0.36).

2.7.3.2. School type and sources of off-task behavior. There was no significant effect of school type on any of the sources of off-task behavior (conditional upon being off-task) after controlling for gender, grade-level, SES, instructional format, and duration (peer distractions p ¼ 0.38, environmental distractions p ¼ 0.87, self-dis- tractions p ¼ 0.67, supplies p ¼ 0.67).

2.8. Discussion study 2

Even in a more heterogeneous sample, gender remained an important student characteristic that was related to children's patterns of attention allocation: Female students engaged in more on-task behavior than male students. Additionally, there were gender differences regarding the types of off-task behavior that each gender tended to engage in. For instance, peer distractions were more common in female students whereas male students were more likely to engage in self-distractions and off-task behavior directed towards supplies.

A significant effect of grade-level on children's on-task behavior was found in Study 2. Third graders engaged in less on-task behavior than first or fourth graders. Additionally, effects of grade-level were obtained for specific types of off-task behavior (fourth graders engaged in more self-distractions compared to first graders and kindergarteners) while no effect of grade was found for peer distractions, environment, or supplies. These findings differ from those obtained in Study 1 in which grade-level was not a significant predictor of children's' attention allocation.

In Study 2 a new student characteristic, SES, was added to the model. However, at least in the present sample, no significant effect of SES was found on children's patterns of attention allocation after controlling for student characteristics, instructional design strate- gies, and school type. One possibility is that SES is confounded with school type. As such the unique contribution of SES may be minimal once school type is taken into account. Consequently, we conducted a follow-up analysis in which school type was dropped from the model of children's on-task behavior. The effect of SES on children's

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 139

on-task behavior controlling for gender, grade-level, instructional format, and duration remained non significant (F(1, 28) ¼ 1.73, p ¼ 0.20).

Similar to Study 1, instructional format was found to be a consistent predictor of children's attention allocation in Study 2. Instructional format was a significant predictor of on-task behavior with the highest rates of on-task behavior occurring during indi- vidual and small-group instruction. Instructional format also influenced how children went off-task. For instance, whole-group instruction elicited higher rates of off-task behavior directed to- ward the environment as well as more self-distractions, while peer distractions were most frequent during individual and small-group formats.

The effect of the duration of an instructional activity on chil- dren's attention allocation revealed that on-task behavior be- comes less frequent as the length of the instructional activity increases. While duration of an activity was a significant predictor of whether or not children would go off-task, in general the duration of an instructional activity did not influence the type of off-task behavior children would engage in e with the exception of self-distractions which were more frequent during shorter instructional activities.

In Study 2, three different types of schools participated: paro- chial schools, private schools, and public charter schools. A signif- icant effect of school type was found for children's rates of on-task behavior with higher rates of on-task behavior evident in parochial compared to private schools. However, in the present sample school type was not an important determinant of the type of off- task behavior that children would engage in as there was no sig- nificant effect of school type on any of the sources of off-task behavior (conditional upon being off-task).

3. General discussion

The present study provides a systematic examination of specific factors that may influence elementary school students' on- and off- task behavior. The following factors were explored: Time of year, student characteristics (gender, grade-level, SES), instructional design strategies (instructional format and duration of instructional activity), and school type (private, parochial, public charter schools). The following findings emerged from the present work (see Table 3 for a summary of key findings).

First, children's pattern of attention allocation is not uniform across Time of year [Research Question 1; Study 1]. Specifically, children's on-task behavior was found to decline by the end of the school year. Additionally, the three most common types of off-task behavior (Peer, Environment, and Self Distractions) were all found to oscillate over the course of the school year with peer off-task behavior increasing in the middle of the year and envi- ronmental and self-distractions both increasing by the end of the school year.

Second, certain student characteristics influenced children's on-task and off-task behaviors [Research Question 2]. Gender was a significant predictor of children's patterns of attention allocation across studies. In both Study 1 and Study 2, our find- ings revealed that female students were on-task significantly more than male students. These findings align with previous research suggesting that in elementary school females are consistently more engaged than males (Marks, 2000). It is possible that in this and prior studies the amount of off-task behavior in females was under-estimated if females tend to engage in off-task behaviors that are less noticeable than off-task behaviors in males. However, our findings do not support this possibility. Specifically, females tended to engage in more peer off-task behaviors than males. Arguably, off-task peer

interactions are a highly noticeable type of off-task behavior; therefore, it appears unlikely that this study under-estimated the rates of off-task behaviors in females. The effects for males were more equivocal. In study 1 males tended to engage in more environment based off-task behavior while in Study 2 males tended to engage in greater rates of self-distractions and off-task behavior directed towards supplies than female students. It is possible that these inconsistencies emerged because in Study 2 we separated off-task behavior directed to the visual environment from off-task behaviors involving supplies, with the latter cate- gory driving the gender differences observed in Study 1.

Variations in the absolute proportion of on-task behavior and off-task behavior across grade-levels were not uniform across the present studies. While results from Study 1 indicate no signifi- cant effect of grade-level on students' patterns of attention allocation, Study 2 found some evidence that grade-level influ- enced children's rates of on and off-task behavior. For example, third graders were least likely to engage in on-task behavior (compared to first and fourth grade students). While higher rates of self-distractions were found among fourth graders (compared to kindergarten and first graders). The inconsistent effect of grade-level across studies may be due in part to the under- sampling of third-grade classrooms in both reported studies. While we attempted to recruit a representative sample of different grade-levels, third grade teachers were considerably less likely to volunteer to participate in the study than teachers in other grade-levels. One possible explanation for third grade teachers' reluctance to volunteer may be due to added pressures that third grade teachers encounter with the onset of standard- ized testing in third grade. Overall, the reported results suggest that in elementary school grade-level is an unstable factor in influencing the prevalence of students' on-task and off-task behaviors.

In the present study socioeconomic status (SES) did not influ- ence children's patterns of attention allocation. One possible explanation for this result is that SES is confounded with school type. However, this possibility was not supported in a follow up analysis in which school type was removed from the analysis and the effect of SES remained non-significant. Additionally, it is possible that any influence from SES may be obscured by under- sampling of very low SES schools. Future research will need to more fully evaluate the influence of SES on attention allocation particularly in these more vulnerable populations.

Third, we also examined the relationship between teachers' instructional design choices and children's on- and off-task behavior [Research Question 3]. Specifically, we investigated two components of instructional design: type of instructional format (e.g., individual work, small-group work, whole-group work, etc.) and the average duration (sec) of an instructional activity. We found that instructional format influenced the overall rate of on-task behavior as well as the form that off-task behavior takes in both a homogeneous sample of charter schools (Study 1) and within a heterogeneous sample of private, paro- chial, and charter schools (Study 2). Thus, the effect of instruc- tional format appears to be generalizable to a wide range of schools and student populations.

Across both Study 1 and Study 2 we observed a consistent association between the type of instructional format and on-task behavior. Specifically, higher rates of on-task behavior were found during small-group work. These results are largely consistent with the findings that small-group instruction is associated with better learning outcomes compared to whole- group instruction (Lou et al., 1996): greater proportion of on- task behavior for small-group instructional activities might mediate the relationship between instructional type and learning

Table 3 An overview of the key findings from Study 1 and Study 2. Consistent effects across both studies are highlighted in blue (with lighter blue highlighting effects that were partially consistent across studies).

Factor Behavior Study 1 Study 2

Time of Year

Time of Year:

Beginning (T1) Middle (T2)

End (T3)

On-task Less on-task behavior at end of the year:T3 < T1 and T2; T2 > T1 -

Peer off-task behavior More peer off-task behavior in middle of the year: T2 > T1 and T3 -

Environmental off-task behavior Environmental off-task behavior increased over the year: T3 and T2 > T1 -

Self-distractions Self-distractions increased over the year:T1 < T2 < T3 -

Student Characteristics

Gender

On-task Females more on-task: F > M Females more on-task: F > M

Peer off-task behaviors Females more peer off-task behavior: F > M Females more peer off-task behavior: F > M

Environmental off-task behavior Males more environmental off-task behavior: M > F ns

Self-distractions ns Males more self-distractions: M > F+

Supplies off-task behavior - Males more supplies off-task behavior: M > F

Grade Level On-task ns 3

rd graders less on-task behavior: 3rd < 1st + and 4th

Self-distractions ns 4 th graders more self-distraction:

4th > K and 1st

Instructional Design Strategies

Instructional Format:

Individual (Indv) Small-Group (SG)

Whole-Group Carpet (WG Carpet)

Whole-Group Desk (WG Desk)

On-task On-Task most likely in SG: SG > WG Carpet / WG Desk+ On-Task most likely in SG and Indv: SG / Indv > WG Carpet / WG Desk

Peer off-task behavior Peer off-task most likely in SG and Indv: SG / Indv > WG Carpet / WG Desk Peer off-task most likely in SG and Indv:

SG / Indv > WG Carpet / WG Desk

Environmental off-task behavior Environmental off-task most likely during WG: WG Desks/WG carpet > SG / Indv Environmental off-task most likely during WG:

WG Carpet / WG Desks > SG / Indv

Self-distractions Self-distractions most likely during WG Carpet:

WG Carpet > all other formats WG Desks > SG / Indv

Self-distractions most likely during WG Carpet: WG Carpet > all other formats

WG Desks > SG / Indv

Supplies off-task behavior - Supplies off-task behavior most likely during WG Desk and SG:

WG Desk / SG > WG Carpet WG Desk / SG+ > Indv

Average Duration of Instructional Activity

On-task - On-task behavior declines during longer instructional activitiesNegative association

Self-distractions - Self-distractions declined during longer instructional activitiesNegative association +

School Based Factor

School Type On-task - More on-task behavior in Parochial schools: Parochial > Private

Note. The symbol ‘+’ denotes a marginally significant result

K .E.

G od

w in

et al.

/ Learn

in g an

d In stru

ction 4 4 (2 016

) 12

8 e 14

3 14

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 141

outcomes. Certain instructional formats were also found to elicit different types of off-task behavior. For example, across Study 1 and Study 2, whole-group instruction (e.g., whole-group in- struction on the carpet and at the desks) was associated with more off-task behavior directed toward the environment. There- fore, it is possible that streamlining the classroom visual envi- ronment may help reduce off-task behaviors in this common instructional format. Overall, the finding that certain types of instructional format are associated with more on-task behavior than others indicates that further research is necessary and important given the potential implications for instruction. However, it is worth noting that instructional duration likely varies across different instructional formats; hence, differences attributed to the instructional format may also reflect effects of the duration of an instructional activity.4 Additional research is necessary to further explore this possibility.

In regard to the effect of the average duration of an instructional activity, we found that on-task behavior became less frequent as the length of the instructional activity increased. Thus, children may be better able to maintain a state of focused attention when instruction consists of small blocks of instructional activities versus instructional episodes that occur over a longer duration (cf. Ruff & Lawson, 1990; Sarid & Breznitz, 1997). This finding is in line with prior empirical work with adults in which attention was found to wane over time (for review see Middendorf & Kalish, 1996): at the beginning of a lecture, most adult students were able to attend to the lecture for up to 18 min before a lapse in attention was observed; however, by the end of the lecture period adults' atten- tion span had decreased to 3e4 min segments (Johnstone & Per- cival, 1976 as cited in Middendorf & Kalish, 1996). However, to our knowledge no prior studies have examined the effects of duration of instructional activity on children's on-task behavior. Attention regulation skills are certainly not as developed in elementary school children as in adults. In this light, it is interesting to note that in the present study the median duration of an instructional ac- tivity was 12.8 min, and 25% of the instructional activity durations were longer than 17.1 min. It is possible that instructional duration interacts with other factors (e.g., activity type, difficulty level, topic interest, novelty, etc.), which may collectively contribute to chil- dren's ability to maintain a state of selective sustained attention. Further research is needed to provide educators with evidence- based guidelines on the optimal length of an instructional activity for children at each grade-level; however, the outcomes of this study suggest that long instructional activity durations (i.e., 30 min) may be suboptimal in elementary grades.

Lastly, the relationship between patterns of children's attention allocation and School Type was investigated [Research Question 4]. A marginally significant effect of school type was found, as school type influenced students' rates of on-task behavior. Specifically, students from parochial schools had higher rates of on-task behavior (compared to students attending private schools). How- ever, the type of school (private, parochial, charter) children attend was not found to be a determining factor of the source of children's off-task behavior. It is not clear what aspects of parochial schools explain the observed patterns of increased on-task behavior found in this sample.

4 In the present study, the average duration of an instructional activity is not linked with instructional format. Thus within the current data set it is not possible to determine whether the duration of activities within particular instructional formats vary systematically. As mentioned previously, a change in an instructional activity often coincided with a change in instructional format; however, this was not always the case as an activity could change while the instructional format remained constant (e.g., when children rotate from one center to another the format remains small-group even though the instructional activity changes).

4. Limitations and future directions

The present work reported a number of novel findings; never- theless there are also important limitations that should be raised and addressed in future research. First, although eye gaze is a common measure of visual attention (see Henderson & Ferreira, 2004; Just & Carpenter, 1976 for reviews), it is admittedly not a perfect measure of attention. For example, it is possible for students to appear to be on-task or attending to the teacher while they are actually daydreaming. Conversely students may be looking else- where while still listening to the teacher's instruction. In the latter case one could argue that the student is not off-task as some amount of attention is still being attributed to the learning activity. However, the student described above is by definition in a state of divided attention which has been demonstrated in the prior liter- ature to be less optimal for performance than a state of selective sustained attention (e.g., Craik, Govoni, Naveh-Benjamin, & Anderson, 1996). To address this limitation, in addition to eye gaze future work may benefit from utilizing multiple measures of attention such as activity level or gross motor movements (e.g., Milich, 1984), and body posture (e.g., D'Mello & Grasser, 2010). Finally, it would also be useful to corroborate visual eye gaze measures with performance-based measures of attention (e.g., K- CPT: Conners & Staff, 2001; Track-It: Fisher, Thiessen, Godwin, Kloos, & Dickerson, 2013) and teacher or parent reports (e.g., CBCL: Achenbach & Rescorla, 2001).

Second, children's patterns of attention allocation may vary as a function of the characteristics of the learning task. For example, children's ability to maintain a state of selective sustained attention may be greater in self-directed learning activities compared to learning contexts that are more structured (e.g., Geary, 2011). Similarly, the difficulty level of the instructional task may interact with students' patterns of attention allocation. For instance, attention may decline when the difficulty level of the task exceeds a student's ability level (e.g., Shernoff, Csikszentmihalyi, Schneider, & Shernoff, 2003; Imai, Anderson, Wilkinson, & Yi, 1992). Future research should explore in more depth how these factors modulate students' patterns of attention allocation in classroom learning environments.

Inattention or off-task behavior is a significant problem in educational settings as inattention reduces students' opportunities to learn. Therefore, it is imperative that researchers isolate ante- cedents of off-task behavior in order to identify potential avenues for intervention. The present work provides a detailed exploration of elementary school children's attention allocation patterns and highlights how the time of year, student characteristics (e.g., gender, grade-level, SES), instructional design strategies (e.g., instructional format and duration of an instructional activity), and school-based factors (e.g., school type) contribute to children's on and off-task behaviors in classroom settings. The present findings are a first step in providing empirical evidence to inform in- terventions that aim to better engage students.

Acknowledgements

We thank Megan Miller, Laura Pacilio, and Jessica Meeks for their help collecting data. We also thank the children, parents, and teachers who made this project possible. The work reported here was supported by a Graduate Training Grant awarded to Carnegie Mellon University by the Department of Education (R305B090023) and by the Institute of Education Sciences, U.S. Department of Ed- ucation (R305A110444). The opinions expressed are those of the authors and do not represent views of the Institute or the U.S. Department of Education.

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143142

References

Achenbach, T., & Rescorla, L. (2001). The manual for the ASEBA school-age forms & profiles. Burlington, VT: University of Vermont, Research Center for Children, Youth, and Families.

Ammons, T. L., Booker, J. L., Jr., & Killmon, C. P. (1995). The effects of time of day on student attention and achievement (ERIC Document no. ED 384 592).

Antrop, I., Roeyers, H., & De Baecke, L. (2005). Effects of time of day on classroom behavior in children with ADHD. School Psychology International, 26(1), 29e43.

Baker, R. S. (2007). Modeling and understanding students' off-task behavior in intelligent tutoring systems. Proceedings of ACM CHI 2007: Computer-Human Interaction, 1059e1068.

Baker, R. S. J. D., D'Mello, S. K., Rodrigo, M. M. T., & Graesser, A. C. (2010). Better to be frustrated than bored: the incidence, persistence, and impact of learners' cognitive-affective states during interactions with three different computer- based learning environments. International Journal of Human-Computer Studies, 68(4), 223e241.

Baker, R. S. J. D., Gowda, S. M., Wixon, M., Kalka, J., Wagner, A. Z., Salvi, A., et al. (2012). Towards sensor-free affect detection in cognitive tutor algebra. In Pro- ceedings of the 5th international conference on educational data mining (pp. 126e133). Worcester, MA: International Educational Data Mining Society.

Barrett, P., Davies, F., Zhang, Y., & Barrett, L. (2015). The impact of classroom design on pupils' learning: final results of a holistic, multi-level analysis. Building & Environment, 89, 118e133.

Barrett, P., Zhang, Y., Moffat, J., & Kobbacy, K. (2012). A holistic, multi-level analysis identifying the impact of classroom design on pupils' learning. Building & Environment, 59, 678e689.

Bartgis, J., Thomas, D. G., Lefler, E. K., & Hartung, C. M. (2008). The development of attention and response inhibition in early childhood. Infant and Child Devel- opment, 17, 491e502.

Berliner, D. C. (1990). What's all the fuss about instructional time? The nature of time in schools: Theoretical concepts, practitioner perceptions. New York: College Press.

Bullard, J. (2010). Creating environments for learning: Birth to age eight. Prentice Hall. Carnine, D. W. (1976). Effects of two teacher-presentation rates on off-task behavior,

answering correctly, and participation. Journal of Applied Behavior Analysis, 9, 199e206.

Carroll, J. (1963). A model of school learning. Teachers College Record, 64, 723e733. Choudhury, N., & Gorman, K. S. (2000). The relationship between sustained

attention and cognitive performance in 17-24 month old toddlers. Infant and Child Development, 9, 127e146.

Conners, C. K., & Staff, M. (2001). Conners' Kiddie continuous performance test (K- CPT): Computer program for windows technical guide and software manual. Tor- onto, ON: MHS.

Craik, F. I. M., Govoni, R., Naveh-Benjamin, M., & Anderson, N. D. (1996). The effects of divided attention on encoding and retrieval processes in human memory. Journal of Experimental Psychology, 125(2), 159e180.

DeMarie-Dreblow, D., & Miller, P. H. (1988). The development of children's strate- gies for selective attention: evidence for a transitional period. Child Develop- ment, 59, 1504e1513.

Dixon, W. E., & Salley, B. J. (2007). “Shh! we're tryin' to concentrate”: attention and environmental distracters in novel word learning. The Journal of Genetic Psy- chology, 167(4), 393e414.

D'Mello, S., & Grasser, A. (2010). Mining bodily patterns of affective experience during learning. In A. Merceron, P. Pavlik, & R. Baker (Eds.), Proceedings of the third international conference on educational data mining (pp. 31e40).

Fisher, C. W., Berliner, D. C., Filby, N. N., Marliave, R., Cahen, L. S., & Dishaw, M. M. (1980). Teaching behaviors, academic learning time, and student achievement. In C. Denham, & A. Liberman (Eds.), Time to learn: A review of the beginning teacher evaluation study (pp. 7e22).

Fisher, A. V., Godwin, K. E., & Seltmen, H. (2014). Visual environment, attention allocation, and learning in young children: when too much of a good thing may be bad. Psychological Science, 25(7), 1362e1370.

Fisher, A. V., & Kloos, H. (2016). Development of selective sustained attention: the role of executive functions. In J. A. Griffin, P. McCardle, & L. Freund (Eds.), Ex- ecutive function in preschool-age children: Integrating measurement, neuro- development, and translational research (pp. 215e237). Washington, DC, US: American Psychological Association.

Fisher, A., Thiessen, E., Godwin, K., Kloos, H., & Dickerson, J. (2013). Assessing se- lective sustained attention: evidence from a new paradigm. Journal of Experi- mental Child Psychology, 114(2), 275e294.

Fleiss, J. L. (1981). Statistical methods for rates and proportions (2nd ed.). New York: John Wiley.

Frederick, W. C., & Walberg, H. J. (1980). Learning as a function of time. Journal of Educational Research, 73, 183e194.

Frederick, W. C., Walberg, H. J., & Rasher, S. P. (1979). Time, teacher comments, and achievement in urban high schools. Journal of Educational Research, 73(2), 63e65.

Geary, K. E. (2011). The impact of choice on child sustained attention in the preschool classroom (Unpublished Thesis). Baton Rouge, Louisiana: Louisiana State University.

Godwin, K., & Fisher, A. (2011). Allocation of attention in classroom environments: consequences for learning. In L. Carlson, C. H€olscher, & T. Shipley (Eds.), Pro- ceedings of the 33rd annual conference of the cognitive science society (pp. 2806e2811). Austin, TX: Cognitive Science Society.

Good, T. L., & Beckerman, T. M. (1978). Time on task: a naturalistic study in sixth- grade classrooms. The Elementary School Journal, 78(3), 192e201.

Goodman, L. (1990). Time and learning in the special education classroom. SUNY Press.

Henderson, J., & Ferreira, F. (2004). The interface of language, vision, and action: Eye movements and the visual world. New York, NY: Taylor & Francis.

Imai, M., Anderson, R. C., Wilkinson, I. A. G., & Yi, H. (1992). Properties of attention during reading lessons. Journal of Educational Psychology, 84(2), 160e173.

Janvier, B., & Testu, F. (2007). Age related differences in daily attention patterns in preschool, kindergarten, first grade & fifth grade pupils. Chronobiology Inter- national, 24(2), 327e343.

Just, M., & Carpenter, P. (1976). Eye fixations and cognitive processes. Cognitive Psychology, 8, 441e480.

Karweit, N., & Slavin, R. E. (1981). Measurement and modeling choices in studies of time and learning. American Educational Research Journal, 18(2), 157e171.

Killingsworth, M. A., & Gilbert, D. T. (2010). A wandering mind is an unhappy mind. Science, 330, 932.

Kulik, J. A. (1992). An analysis of research on ability grouping: Historical and contemporary perspectives. In Research-based decision-making series. Storrs, CT: University of Connecticut. National Research Center on the Gifted and Talented (ERIC Document Reproduction Service No. ED 350 777).

Lahaderne, H. M. (1968). Attitudinal and intellectual correlates of attention: a study of four sixth-grade classrooms. Journal of Educational Psychology, 59(5), 320e324.

Lee, S. W., Kelly, K. E., & Nyre, J. E. (1999). Preliminary report on the relation of students' on-task behavior with completion of school work. Psychological Re- ports, 84, 267e272.

Lemov, D. (2010). Teach like a champion: 49 techniques that put students on a path to college. San Francisco, CA: Jossey-Bass.

Lloyd, J. W., & Loper, A. B. (1986). Measurement and evaluation of task-related learning behaviors: attention to task and metacognition. School Psychology Review, 15(3), 336e345.

Lou, Y., Abrami, P. C., Spence, J. C., Poulsen, C., Chambers, B., & d'Apollonia, S. (1996). Within-class grouping: a meta-analysis. Review of Educational Research, 66(4), 423e458.

Marks, H. (2000). Student engagement in instructional activity: patterns in elementary, middle, and high school years. American Educational Research Journal, 37(1), 153e184.

Martin, A. J., Papworth, B., Ginns, P., Malmbergm, L.-E., Collie, R. J., & Calvo, R. A. (2015). Real-time motivation and engagement during a month at school: every moment of every day for every student matters. Learning and Individual Dif- ferences, 38, 26e35.

Matthews, J. S., Ponitz, C. C., & Morrison, F. J. (2009). Early gender differences in self- regulation and academic achievement. Journal of Educational Psychology, 101(3), 689e704.

Middendorf, J., & Kalish, A. (1996). The ‘change-up’ in lectures. The National Teaching and Learning Forum, 5(2), 1e4.

Milich, R. (1984). Cross-sectional and longitudinal observations of activity level and sustained attention in a normative sample. Journal of Abnormal Child Psychology, 12(2), 261e276.

Muyskens, P., & Ysseldyke, J. E. (1998). Student academic responding time as a function of classroom ecology and time of day. The Journal of Special Education, 31(4), 411e424.

Ocumpaugh, J., Baker, R. S. J. D., & Rodrigo, M. M. T. (2012). Baker-Rodrigo observation method protocol (BROMP) 1.0. Training manual version 1.0. Technical Report. Manila, Philippines: Ateneo Laboratory for the Learning Sciences.

Ponitz, C. C., & Rimm-Kaufman, S. E. (2011). Contexts of reading instruction: im- plications for literacy skills and kindergarteners' behavioral engagement. Early Childhood Research Quarterly, 26, 157e168.

Prais, S. J. (1998). Raising schooling attainments by grouping pupils within each class. National Institute Economic Review, 165, 83e88.

Prais, S. J. (1999). Within-class grouping: a rejoinder (response to article by Philip C. Abrami et al. in this issue p. 105). National Institute Economic Review, 169, 109e110.

Rettig, M. D., & Canady, R. L. (2013). Elementary school scheduling: Enhancing in- struction for student achievement. NY: Routledge.

Roberts, M. (2001). Off-task behavior in the classroom: Applying FBA and CBM. Retrieved from: http://www.nasponline.org/communications/spawareness/Off- Task%20Behavior.pdf.

Roberts, M. (2002). Research in practice: practical approaches to conducting func- tional analyses that all educators can use. The Behavior Analyst Today, 3, 83e87.

Ruff, H. A., & Lawson, K. R. (1990). Development of sustained, focused attention in young children during free play. Developmental Psychology, 26, 85e93.

Samuels, S. J., & Turnure, J. E. (1974). Attention and reading achievement in first- grade boys and girls. Journal of Educational Psychology, 66(1), 29e32.

Sarid, M., & Breznitz, Z. (1997). Developmental aspects of sustained attention among 2- to 6-year-old children. International Journal of Behavioral Develop- ment, 21, 303e312.

Schooler, J. W., Smallwood, J., Christoff, K., Handy, T. C., Reichle, E. D., & Sayette, M. A. (2011). Meta-awareness, perceptual decoupling and the wandering mind. Trends in Cognitive Science, 15(7), 319e326.

Shernoff, D. J., Csikszentmihalyi, M., Schneider, B., & Shernoff, E. S. (2003). Student engagement in high school classrooms from the perspective of flow theory. School Psychology Quarterly, 18(2), 158e176.

Smallwood, J., Fishman, D. J., & Schooler, J. W. (2007). Counting the cost of an absent

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref1

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref2

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref3

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref4

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref5

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref6

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref7

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref8

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref9

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref10

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref11

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref12

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref13

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref14

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref16

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref17

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref18

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref19

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref20

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref22

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref23

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref24

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref25

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref26

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref27

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref28

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref29

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref30

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref31

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref32

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref72

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref33

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref69

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref70

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref34

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref35

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref36

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref37

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref38

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref39

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref40

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref41

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref42

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref43

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref44

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref45

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref47

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref48

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref49

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref50

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref51

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref52

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref53

http://www.nasponline.org/communications/spawareness/Off-Task%20Behavior.pdf

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref55

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref56

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref57

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref58

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref59

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref61

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref62

K.E. Godwin et al. / Learning and Instruction 44 (2016) 128e143 143

mind: mind wandering as an underrecognized influence on educational per- formance. Psychonomic Bulleting & Review, 14(2), 230e236.

Tarr, P. (2004). Consider the walls. Young Children, 59(3), 88e92. Thompson, S. D., & Raisor, J. M. (2013). Meeting the sensory needs of young chil-

dren. Young Children, 68(2), 34e43. White, S. (1970). Some general outlines of the matrix of developmental changes

View publication statsView publication stats

between 5 and 7 years. Bulletin of the Orton Society, 20, 41e57. Wiebe, S. A., Sheffield, T., Nelson, J. M., Clark, C. A. C., Chevalier, N., & Andrews

Espy, K. (2011). The structure of executive function in 3-year-old children. Journal of Experimental Child Psychology, 108(3), 436e452.

Zagar, R., & Bowers, N. D. (1983). The effect of time of day on problem solving and classroom behavior. Psychology in the Schools, 20, 337e345.

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref62

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref63

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref64

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref65

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref66

http://refhub.elsevier.com/S0959-4752(16)30027-5/sref68

https://www.researchgate.net/publication/303691151

Off-task behavior in elementary school children

1. Off-task behavior in elementary school students
2. The present study

2.1. Study 1: Temporal patterns of on- and off-task behavior across the school year
2.2. Method

2.2.1. Participants
2.2.2. Design and procedure
2.2.3. Coding on- and off-task behaviors

2.3. Results study 1

2.3.1. Question 1: Do patterns of attention allocation change over the course of the school year?

2.3.1.1. Time of year and on-task behavior

2.3.1.1.1. On-task behavior

2.3.1.2. Time of year and the sources of off-task behavior

2.3.1.2.1. Peer distractions
2.3.1.2.2. Environmental distractions
2.3.1.2.3. Self-distractions

2.3.2. Question 2: Are student characteristics related to children's attention allocation patterns?

2.3.2.1. Effect of gender

2.3.2.1.1. Gender and on-task behavior
2.3.2.1.2. Gender and peer-distractions
2.3.2.1.3. Gender and environmental distractions
2.3.2.1.4. Gender and self-distractions

2.3.2.2. Effect of grade-level

2.3.3. Question 3: Are instructional design choices related to children's attention allocation?

2.3.3.1. Instructional format and on-task behavior
2.3.3.2. Instructional format and peer distractions
2.3.3.3. Instructional format and environmental distractions
2.3.3.4. Instructional format and self-distractions

2.4. Discussion study 1
2.5. Study 2: Assessing the generalizability of the relationship between student characteristics, instructional design strategie ...
2.6. Method

2.6.1. Participants
2.6.2. Design and procedure

2.7. Results study 2

2.7.1. Question 2: Are student characteristics related to children's attention allocation patterns?

2.7.1.1. Effect of gender

2.7.1.1.1. Gender and on-task behavior
2.7.1.1.2. Gender and peer distractions
2.7.1.1.3. Gender and environmental distractions
2.7.1.1.4. Gender and self-distractions
2.7.1.1.5. Gender and supplies

2.7.1.2. Effect of grade-level

2.7.1.2.1. Grade-level and on-task behavior
2.7.1.2.2. Grade-level and peer distractions
2.7.1.2.3. Grade-level and environmental distractions
2.7.1.2.4. Grade-level and self-distractions
2.7.1.2.5. Grade-level and supplies

2.7.1.3. Effect of SES

2.7.2. Question 3: Are instructional design strategies related to children's attention allocation?

2.7.2.1. Effect of instructional format

2.7.2.1.1. Instructional format and on-task behavior
2.7.2.1.2. Instructional format and peer distractions
2.7.2.1.3. Instructional format and environmental distractions
2.7.2.1.4. Instructional format and self-distractions
2.7.2.1.5. Instructional format and supplies

2.7.2.2. Effect of instructional duration

2.7.2.2.1. Duration and on-task behavior
2.7.2.2.2. Duration and peer distractions
2.7.2.2.3. Duration and environmental distractions
2.7.2.2.4. Duration and self-distractions
2.7.2.2.5. Duration and supplies

2.7.3. Question 4: Is school type related to children's attention allocation?

2.7.3.1. Effect of school type

2.7.3.1.1. School type and on-task behavior

2.7.3.2. School type and sources of off-task behavior

2.8. Discussion study 2

3. General discussion
4. Limitations and future directions
Acknowledgements
References