Computer Programming with Early Elementary Students with Down Syndrome

STEM in Early Education

Science, technology, engineering, and mathematics are frequently interwoven throughout state standards and national programs (Carr et al., 2012; Israel, Wherfel, Pearson, Shehab, & Tapia, 2015; Nadelson et al., 2013). Researchers concluded early elementary grades was the time to introduce basic STEM curriculum and skills for young students (e.g., Lott, Wallin, Roghaar, & Price, 2013; Nadelson et al., 2013). Students in elementary school tend to have more flexible schedules for teachers to incorporate STEM-related activities into their daily routines (Nadelson et al., 2013). Young students also tend to be motivated, engaged in lessons, and have a drive to explore (Nadelson et al., 2013).

Current state and national standards (e.g., National Council of Teachers of Mathematics [NCTM], 2000; Next Generation Science Standards [NGSS], 2013; U.S. Department of Education, 2016) provide guidance for establishing a foundation for STEM in the early elementary grades. Mathematics is considered a cornerstone of education at this level (Clements & Sarama, 2008) and focuses on number sense, algebra, measurement and data collection, and geometry (NCTM, 2000). The NGSS (2013) provide teachers with direction on how to embed science and engineering instruction into the general education curriculum (e.g., mathematics) to offer a well-rounded learning space for students. Many of the current state standards already have engineering concepts built in (Carr et al., 2012).

Lottero-Perdue, Lovelidge, and Bowling (2010) discussed the need for implementation of STEM standards, specifically engineering, in inclusive elementary classrooms. The authors suggested the five steps of the engineering process (i.e., ask, imagine, plan, create, and improve) provide students with a means to test their own ideas and extend their learning. Even the concept of failing is encouraged by the authors, as it gives the students the opportunity to test their work and spurs thinking of what needs to be done differently to make a project succeed.

Computer Programming and Robotics

Computer programming and computer science are skill areas that are critical for many careers (Kalelioğlu, 2015). Preparation for these careers needs to begin early in students’ schooling, which is a time to promote engaging activities and develop necessary skills (Bers, Flannery, Kazakoff, & Sullivan, 2014; Kalelioğlu, 2015; Kazakoff & Bers, 2012). Kalelioğlu (2015) researched the effects of computer programming using the web-driven application “code.org” on elementary students’ reflective thinking skills. Students in the treatment group accessed modules to further their programming skills. The researchers hypothesized students interacting with the online modules would develop stronger reflective skills given the tasks they had to complete and problems to solve. Although participants did not show significant differences in their reflective skills, the students in the treatment group felt “code.org” was an engaging site, assisting students in strengthening mathematical skills and understanding.

Kazakoff and Bers (2012) studied sequencing skills of kindergarten students’ before and after implementation of a programming and robotics curriculum. Computer programming requires users to logically sequence code, so that programs are developed to run without interference. The robotics curriculum was introduced to the participants because it offered a concrete and tangible way to see the effects of specific code they developed. Students in the treatment group received the programming intervention and showed significant positive changes in mean sequencing scores. The authors also found kindergarten students could develop programming skills, given the right tools and instructions.

Sullivan, Kazakoff, and Bers (2013) continued this line of research on computer programming in studying prekindergarten students’ engagement in a robotics curriculum. The curriculum focused on engineering and mathematics, which included engineering design, programming, and robotics. Students were able to design and build robots to complete tasks (e.g., pushing, moving, and sensing) with adult guidance but also found free exploration and peer assistance of value.

Tangible coding combines the use of computer-based software and physical manipulatives to teach young students how to code. Kwon, Kim, Shim, and Lee (2012) researched learning gains of first-grade students in a treatment group using tangible coding compared to a control group using only a computer-based coding program (i.e., Scratch). Participants in the treatment group made fewer errors in coding and achieved higher levels of programming tasks. As the tasks became more difficult, participants in both groups had trouble successfully completing programming requirements without direct teacher/instructor help.

Computer programming and robotics offer young students an engaging medium to explore areas of STEM and learn important skills that can be built on in other academic content areas (e.g., Bers et al., 2014; Sullivan, Kazakoff, & Bers, 2013). Students in prekindergarten and kindergarten have demonstrated the ability to program when given structured lessons, scaffolding, and guidance from adults (e.g., Bers et al., 2014; Sullivan et al., 2013). Unfortunately, few empirical research studies focused on computer programming have been conducted with students with disabilities (e.g., deaf or communication disabilities; Adams & Cook, 2013; P. Miller, 2009). It is important to research the extent to which students with IDs can comprehend and develop computer programming skills.

Students With ID and STEM

Students with ID have an IQ of less than 70 and significantly impaired adaptive skills (American Psychiatric Association, 2013). Adaptive skills include problem-solving, self-determination, and flexibility (Cote et al., 2010). Goharpey, Crewther, and Crewther (2013) found students with ID develop problem-solving skills at a slower rate than their same-age peers and may benefit from utilizing auditory or visuospatial skills. B. Miller, Doughty, and Krockover (2015) found secondary school students with ID could be taught problem-solving strategies embedded in scientific inquiry and could generalize the skills to novel tasks. Science, technology, engineering, and technology instruction (and computer programming in particular) may offer students with ID an opportunity to learn problem-solving skills that can be applied to social and academic situations (B. Miller, Doughty, & Krockover, 2015).

Although many teachers do not feel young students (i.e., prekindergarten through second grade) with and without disabilities are ready for STEM instruction (Faulkner et al., 2013; Goodnough et al., 2014), prominent researchers have studied EBPs to support student learning in this content area. Browder, Spooner, Ahlgrim-Delzell, Harris, and Wakeman (2008) and Spooner, Knight, Browder, Jimenez, and DiBiase (2011) found EBPs for students with ID in mathematics and science including systematic, explicit instruction, life skills in context (known as in vivo), and opportunities to respond. Spooner and colleagues (2011) found time delay to teach discrete skills and task analytic instruction to teach chained skills emerged as EBPs for students with ID and other severe disabilities. Doabler and Fien (2013) discussed the use of explicit mathematics instruction for students with disabilities and others with difficulties in the subject area. The authors suggested explicit instruction encompassed direct teaching, scaffolding, student practice, and consistent feedback. Other researchers working with young students (prekindergarten through second grade) concluded explicit instruction was necessary and important for student comprehension in programming skills (Fessakis, Gouli, & Mavroudi, 2013; Harlow & Leak, 2014; Sullivan et al., 2013).

Currently, there is a gap in the literature focused on students with IDs learning computer programming skills, especially at a young age and grade level. Students with disabilities must be given the same opportunities as their typically developing peers (Every Student Succeeds Act [ESSA], 2015; Individuals with Disabilities Education Act, 2004). While researchers are continuing to study programming instruction in the early elementary grades (e.g., Bers et al., 2014; Fessakis et al., 2013; Kazakoff & Bers, 2012; Sullivan & Bers, 2013, 2016; Sullivan et al., 2013), research with students with ID is not currently evident in the literature (Sullivan & Heffernan, 2016).

Purpose and Research Questions

The purpose of this study was to explore the feasibility of teaching computer programming skills through explicit instruction to students with IDs, specifically those with Down syndrome. The research questions for this study were:

To what extent does explicit instruction on computer programming in a 1:1 setting increase percentage of correct responses to explicit instructions by first- or second-grade students with Down syndrome?

Participants

First- and second-grade elementary students with Down syndrome (N = 4) were recruited to participate in this study. Selection criteria of participants for the study included (a) diagnosis of Down syndrome, (b) an ID, (c) currently in first or second grade, (d) no prior experience with coding applications, and (e) could identify (choose and label) a square. Initially, four students met criteria for this study, but one was unable to participate due to personal reasons. Three students completed the study requirements (see Table 1). The Woodcock-Johnson III: Quantitative Reasoning subtests were used to assess participant mathematical knowledge and ensure participants were similar in mathematics ability. The lead researcher administered each subtest.

OPEN IN VIEWER

Table 1.

Demographics.

					Woodcock-Johnson III: Quantitative Concepts
Participant	Gender	Grade Level	School	Diagnosis	Grade Level	Age Level
Participant 1	Female	1	Home school	Down syndrome	<K.0	4.1
Participant 2	Female	2	Home school	Down syndrome	K.1	5.2
Participant 3	Male	1	Charter school	Down syndrome	<K.0	3.4

Setting

The setting for this study was a large university in the southeastern United States. All students received the intervention in a one-on-one format (i.e., researcher and participant) in a research room on the university campus. On the floor of the room was a square (measured at 100 cm × 100 cm) created using masking tape. The square was a visual cue for participants of the path the robot was to follow. Students sat next to the taped figure while the researcher sat in front of them to provide directions for the study.

Design

A changing criterion single case design was used in this study. This design utilizes a baseline phase followed by the treatment phase with several levels. Each level is contingent on a set of criterion. Once criterion has been met, the participant moves to the next level (Gast & Ledford, 2010). Participants began with five sessions in baseline to establish a stable level and trend. Upon completion of the baseline phase, participants entered the treatment phase, which consisted of four levels. The first level required participants to correctly follow at least three of the four steps (i.e., identifying and sequencing programming symbols) in two consecutive sessions before moving to the second level. The second, third, and fourth levels consisted of seven steps each. The participants only moved to the next level when they had correctly completed all steps of the previous level with a maximum of one error over two consecutive sessions.

Instruments

Coding blocks

The main intervention used physical blocks (see Figure 1) to represent code for the students (Interim CSTA K–12 Computer Science Standards, 2016). Students with ID develop knowledge with more ease using concrete objects (i.e., physical blocks) rather than abstract concepts (e.g., iPad application for coding; Jimenez, Browder, & Courtade, 2008; Witzel, Mercer, & Miller, 2003). Each block had a picture or color representing a block of code (i.e., arrow forward, arrow left turn, green for go, and red for stop) and the word it represented in the code (i.e., go, forward, turn left, and stop). The blocks were fitted, so that participants could connect them together.

OPEN IN VIEWER

Figure 1.

Coding blocks.

Blockly application

Many coding programs and tools present code to elementary students using digital blocks that drag and drop to create a program (Kalelioğlu, 2015). In this study, the application “Blockly” was used (see Figure 2), which was developed by Wonder Workshop specifically for Dash and Dot robots (see description below). This application allows students to merge blocks of code to tell the robot to follow a variety of instructions (e.g., movement, sound, and repeat). Students were not assessed on use of Blockly application. Application was only used to make the robot program run.

OPEN IN VIEWER

Figure 2.

Blockly application.

Dash robot

Dash is a robot created by Wonder Workshop (see Figure 3). The robot was designed to engage students through voice, sound, and accessories (e.g., attachments for Legos, smartphone mount). Using applications (Blockly, Wonder, Go, Path), students can control the robot through touch screen devices (e.g., tablets or smartphones).

OPEN IN VIEWER

Figure 3.

Dash robot.

Dependent Measure

The dependent measure in this study was a checklist of steps to be completed by the participant. Each step detailed the direction participants were to complete (e.g., Step 1: Point to coding block “Go”). If a step was completed correctly, a point was awarded and the researcher continued on to the next step. If a step was completed incorrectly during the intervention, the researcher modeled for the student what should have been done and then repeated the step. A point was not awarded if any portion of the step was done incorrectly. There were 25 total points possible over all four treatment levels.

A total count of directions followed correctly was used for calculating percentage of correct responses. Total points for a section were divided by total possible points and multiplied by 100 to calculate the percentage. In the changing criterion model, the participants build upon previous knowledge obtained in prior intervention levels. Total scores resulted from points earned at the current level the participant was in plus the cumulative total points from the levels before. Final percentages were calculated from taking this total score, dividing by points possible overall (25) and multiplying by 100. Table 2 details the individual intervention levels, scoring within those levels, and final percentages for a change between intervention levels.

OPEN IN VIEWER

Table 2.

Dependent Measure Phase Changes.

Phase	Possible Points	Cumulative Possible Points	Points needed for level change	Cumulative points needed	Calculation for Percentage	Percentage Criterion Change
Baseline	25	25	N/A	N/A	0 to 25 25 × 100	0–100
Intervention:
Level 1	4	4	3–4	3–4	3 or 4 25 × 100	12–16
Level 2	7	11	6–7	10–11	10 or 11 25 × 100	40–44
Level 3	7	18	6–7	17–18	17 or 18 25 × 100	68–72
Level 4	7	25	6–7	24–25	24 or 25 25 × 100	96–100

Independent Variable

The independent variable was explicit instruction in coding the robot Dash to move in a square. Explicit instruction is an EBP used by teachers that has had positive results in students with IDs learning and acquiring new skills (Browder et al., 2008; Doabler & Fien, 2013; Spooner et al., 2011). Explicit instruction is defined by Doabler and Fien (2013) to consist of direct instruction, scaffolding, practice, and feedback. Participants in this study worked directly with the researcher to learn the computer coding materials (i.e., coding blocks) and the purpose of developing a code (i.e., make the robot move).

Before the session began, the participants were told the purpose of the session. Then, the participants were shown each of the coding blocks they would use during the session. The researcher showed the block and explicitly said the name of the block to the participants. The participants were then asked to repeat the name of the block. If a participant could not say the name of the block, they were given the option to point at it as they tried to say the name. The participants were then asked to identify a particular block from a group of two or three (e.g., identify block Go from blocks “Stop” and “Forward”). Steps were repeated if the participants identified a block incorrectly. The final step consisted of the participants creating a particular code after identifying all coding blocks to be used. The researcher showed the participant all the blocks needed for the creation of the code and asked the student to identify the blocks in response to questions posed (e.g., “First we need ‘Go.’ Which block is ‘Go’?”). The researcher and participant read each individual block in the code together as a review before Dash followed the program (through Wonder Workshop application Blockly).

Procedure

Participant 1 (P1)

P1 was in first grade and was home schooled. The baseline phase was completed over five sessions and P1 was unable to complete any portion of the code to make Dash move in a square (see Figure 4). P1 entered intervention beginning with the first level to code Dash to move in a straight line. She was able to successfully complete the treatment phase over three sessions (4%, 12%, and 16% accuracy). In the second level, the researcher explicitly taught P1 to create a code to move Dash in a straight line, turn left, and move in a straight line again (i.e., ½ of square). The first session of this level was positive, with only one incorrect response (40% accuracy). The second session of this level was done without any errors and successful completion of the code (44% accuracy). The first session of the third level was completed with one mistake (68% accuracy) but followed with a second session of two mistakes (64% accuracy). P1 continued the third level over two more sessions, consecutively recording only one mistake and no mistakes (68% and 72% accuracy). In the final (fourth) level, P1 successfully completed the coding operations over three sessions (92%, 96%, and 100% accuracy) to make the robot move in a square and completed the study.

OPEN IN VIEWER

Figure 4.

Participant 1 coding interventions.

Participant 2 (P2)

P2 was in second grade and recently began home schooling. Level and trend were stable during the five baseline sessions. During baseline Sessions 3, 4, and 5, she placed one or two blocks in the correct place to create a code (i.e., first block had to be Go, the second block had to be “Forward”). Placement of these blocks was most likely by chance, although she may have recognized and read Go as being first (see Figure 5).

OPEN IN VIEWER

Figure 5.

Participant 2 coding interventions.

The first level of the intervention required P2 to follow directions to recognize coding blocks (i.e., Go, Stop, and Forward) and place them in the correct order to tell Dash to move in a straight line. P2 reached the benchmark of this level in two consecutive sessions (12% and 16% accuracy) making a maximum of one error in the first session and successfully completing the level without error in the second session. The second level built upon the first, requiring P2 to recognize a new coding block (i.e., turn left) and create a code to tell Dash to move forward, turn left, and move forward again. P2 was able to identify the coding blocks and follow all directions correctly to create the code over two consecutive sessions (44% and 44% accuracy). She was able to identify all blocks needed in developing and creating the code for Dash to move in ¾ of a square during the third level. P2 completed the third treatment level over two sessions (72% and 68% accuracy). She then began the fourth level with only one mistake in the first session (96% accuracy) and completed the final session with a perfect score (100% accuracy), successfully completing the study.

Participant 3 (P3)

P3 was in first grade and attended a charter school developed specifically for the inclusion of students with and without disabilities in the general education classroom. Five baseline sessions were completed by P3 showing a stable level and trend (see Figure 6). P3 began all baseline sessions unsure of what to do with the coding blocks and at some points tried to put the blocks into the robot. The final two sessions of correct responses were due to putting the coding block Go in the right position.

OPEN IN VIEWER

Figure 6.

Participant 3 coding interventions.

P3 began the first level of intervention to code Dash to move in a straight line. He was only able to correctly respond to one step in each of the first two sessions at this level (4% and 4% accuracy). P3 showed an increase in understanding during the third session of this level. He correctly responded to explicit instruction in three of the four instances (12% accuracy). In Session 4, P3 was only able to correctly respond in two of the four instances (8% accuracy). P3 showed gain scores in responses (12% and 16% accuracy) during Sessions 5 and 6 and successful completion of the first level. P3 made two mistakes during the first session of the second level (36% accuracy). He was able to complete the second level in the following two sessions, making one mistake in each (40% and 40% accuracy). During the third level, P3 completed two consecutive sessions with only one mistake (68% and 72% accuracy) to successfully move into the final (fourth) level. P3 then completed the fourth level over two sessions (96% and 96% accuracy), successfully coding Dash to travel in a square.

Effect Size

Effect size was calculated using τ_novlap, as no significant trend changes occurred in baseline for any of the participants. τ_novlap analyzes overlap between baseline phase and intervention phases taking into account both nonoverlapping and overlapping data (Parker, Vannest, & Davis, 2011). Effect sizes were calculated using an online program (i.e., http://www.singlecaseresearch.org/calculators/tau-u; Vannest, Parker, & Gonen, 2016). Effect sizes for all three participants were large (0.93–1.00; P1 × τ_novlap = 0.95, P2 × τ_novlap = 1.00, P3 × τ_novlap = 1.00). Combined effect size for all three participants was also large (τ_novlap = 0.982).

Interrater Reliability

Interrater reliability was completed by an outside observer using the archived video recordings. An explanation of the project and the scoring rubric was provided for the observer. The researcher coded three video files alongside the interrater to ensure agreement on student performance and to answer any questions. The first author was the primary observer and the interrater was a doctoral candidate with over 10 years’ experience working with elementary students with disabilities using explicit instruction.

Interobserver agreement was calculated using a point-by-point method. The formula for this method is dividing the number of agreements between the two observers by the total agreements plus disagreements and multiplying by 100 to get a percentage. A random number generator was used to determine the videos the interobserver would code (i.e., 40% of video recordings throughout baseline and intervention phases for participants). Interobserver agreement was 94.8%.

Social Validity

Participants and parents were asked to complete a survey to measure their satisfaction with the intervention’s goals, procedures, and outcomes (Wolf, 1978). Participants responded to a Likert-type scale survey questions using emoticons to aid in engagement and understanding, which represented a 3-point Likert-type scale (i.e., sad face (1) = disagree, face without smile or frown (2) = neither agree nor disagree, smiling face (3) = agree). All participants responded choosing smiling faces (i.e., agree) regarding working with the Dash robot, constructing code, and using the robot in the future. Parents of the participants responded to a Likert-type scale survey representing 5 points (1 = strongly disagree, 2 = disagree, 3 = neither nor disagree, 4 = agree, and 5 = strongly agree). Parents indicated that they strongly agreed or agreed (only one instance) to all survey questions regarding the intervention’s goals, procedures, and outcomes.

Limitations and Future Implications

There are several limitations to be noted within this study. The research design could have been strengthened with at least three data points per treatment level (What Works Clearinghouse [WWC]; Kratochwill et al., 2010). The researchers felt only two data points were necessary to move between levels due to immediacy and stability of data in each of the treatment levels. Participants completed only one level per session and could not move on to the next level until the next session. At baseline, students demonstrated an inability to advance through levels without explicit instruction, deeming constraint necessary. Following the baseline and treatment phases, participants were not required to complete a maintenance phase. The researchers’ purpose of this study was to assess whether students with Down syndrome could learn skills related to computer programming when taught explicitly, as there are no studies currently in the literature working with the same population and skills in coding. Future studies should require a minimum of three data points per treatment phase to align with WWC standards. Participants should be given the opportunity to continue constructing code to move the Dash robot in a square regardless of treatment level (i.e., not constrained in changing levels). Finally, participants should be monitored after completion of the treatment phase for maintenance of skills acquired.

Results should be generalized with caution, as there were only three participants. More research is needed on students with and without disabilities and their ability to learn computer programming skills at a young age. Research is also needed to assess whether the skills taught using the explicit coding intervention generalize outside the research setting and to novel problems. It may also be important to fade out prompts through scaffolding methods.

Future studies will need to focus on expanding from teaching basic computer programming skills to higher level thinking. Skills for computer programming include knowledge of technology, ability to recognize errors in code and correct them, and developing code to meet a certain criteria. Future studies should also focus on the ability of students to generalize tasks learned (e.g., coding robot to travel in a square) to develop new code (e.g., coding robot to follow a path). Researchers have documented the importance of teacher guidance in these programming activities for all students, regardless of disability, to support learning and growth and minimize frustration and boredom (e.g., Harlow & Leak, 2014; Kalelioğlu & Gülbahar, 2014; Sullivan et al., 2013). Future research should also include large group studies for students with disabilities, interactions with peers, and ability to complete basic programming tasks.