Abstract
In the current study, we investigated the effects of an instructional package (i.e., response prompting, sentence frames) on sentence writing for three middle school participants (ages 12–13) with moderate to severe disabilities. We employed a multiple probe across behaviors design to evaluate the efficacy of the intervention package and also used intermittent probes to assess generalization of sentence writing skills to journal writing activities. Our data indicated that the package was effective and produced variable levels of maintenance and generalized responding for all three participants.
In recent decades, research in the area of instruction for learners with moderate to severe disabilities (MSD) has been marked by an increased emphasis on academic content. Researchers have demonstrated that individuals with MSD can make progress across a range of content areas including reading (Browder, Wakeman, Spooner, Ahlgrim-Delzell, & Algozzine, 2006), mathematics (Browder, Spooner, Ahlgrim-Delzell, Harris, & Wakeman, 2008), science (Spooner, Knight, Browder, Jimenez, & DiBiase, 2011), and social studies (LeBlanc, Miguel, Cummings, Goldsmith, & Carr, 2003) and have galvanized the research corps to conduct investigations into new content areas and to ultimately expand expectations for this unique population of learners (e.g., Heinrich, Knight, Collins, & Spriggs, 2016; Root, Browder, Saunders, & Lo, 2016). The majority of investigations have addressed skills in the area of literacy and have demonstrated the efficacy of strategies for increasing participants’ skill repertoires in reading (Hudson, Browder, & Wood, 2013). Despite the critical importance of reading for all learners, literacy instruction also must reflect learners’ use of text in communicating to others (Graham & Harris, 2016). Unfortunately, the extant research on literacy instruction for students with MSD reflects an unbalanced emphasis on reading skills and the need for the increased investigation of strategies for teaching writing.
Written expression plays a valuable role in lives of all students as it provides a means for demonstrating the acquisition of new skills across content areas. In the absence of a sufficient writing repertoire, students must speak responses to instructional directives to receive teacher feedback (e.g., correction, praise). This presents a challenge for many students with MSD who may not use speech to communicate and are thus precluded from access to complete instructional units (i.e., stimulus presentation, response, feedback). In addition, a teacher’s sole reliance on students’ vocal responding may not be efficient, especially in large group arrangements (e.g., general education classrooms), where a teacher’s attention is not continuously available and students are expected to respond using permanent products that can be reviewed at a later time.
Students with poor writing skills also are disadvantaged outside of the classroom setting. Writing is essential to the development and maintenance of social relationships. Recent data suggest that individuals spend an increasing amount of time using electronic media to engage in social interactions (Wartella, Rideout, Montague, Beaudoin-Ryan, & Lauricella, 2016). Students without the skills to produce simple text will be less likely to engage in these electronic interactions and may find themselves increasingly isolated by their restricted access to the digital world. Furthermore, written expression is required for competence across a range of postsecondary settings including higher education, employment, and the community (Marchand-Martella, Martella, Modderman, Petersen, & Pan, 2013; Perin, 2013).
Fortunately, a few research teams have demonstrated an interest in improving the writing skills of students with MSD. A majority of the earlier investigations focused on improving the spelling skills of these students. Three studies involved the use of computer-based match to sample procedures, whereas students learned to select letters to match a written model displayed on a computer screen. Following participants’ responses, computer software delivered differential feedback for correct responses and errors (Stromer, Mackay, Howell, McVay, & Flusser, 1996; Sugasawara & Yamamoto, 2007; Tanji, Takahashi, & Noro, 2013). Similarly, two research teams taught students to spell words in the presence of a model and then gradually faded the model using backward chaining procedures (Purrazzella & Mechling, 2013; Stewart, Hayashi, & Saunders, 2010). Erbas, Turan, Ozen, and Halle (2006) used the “cover-write method,” in which students practiced spelling a word in the presence of the model, covered the model, and then attempted to spell the word in the absence of the model. Finally, McDonnell, Thorson, Allen, and Mathot-Buckner (2000) taught spelling within partner learning activities. Students with and without disabilities were presented with differentiated word lists and then took turns practicing spelling the words and providing feedback. Collectively, these studies demonstrated the efficacy of instruction involving repeated practice and systematic fading of models for teaching spelling to students with MSD. Though spelling is critical in the development of reading skills (Mercer, Mercer, & Pullen, 2011) and may be viewed as a prerequisite to generative written expression, it represents only a kernel of the knowledge required to be an effective writer.
Several researchers have investigated instructional strategies for teaching these students to generate narratives for different purposes. In one of the earliest investigations, Collins, Branson, Hall, and Rankin (2001) used a peer-administered system of least prompts (SLP) procedure to teach high school students to write letters. During intervention, peers used a hierarchy of prompts to evoke the inclusion of four elements during the letter writing task (i.e., heading, greeting, content body, closing). All participants generated letters that included the targeted components but only partially maintained skills at 2 and 4 weeks after instruction. More recently, Pennington, Delano, and Scott (2014) taught students to include six components within resume cover letters written to prospective employers. They also employed the SLP procedure but added a graphing component, in which the students graphed the number of components used during each session. In this study, all the participants met criterion and maintained performance during maintenance probes.
In a series of studies, Pennington and colleagues (Pennington, Ault, Schuster, & Sanders, 2011; Pennington, Collins, Stenhoff, Turner, & Gunselman, 2014; Pennington, Stenhoff, Gibson, & Ballou, 2012) used simultaneous prompting to teach students with autism spectrum disorder (ASD) to construct simple stories from words arranged within computer-based arrays. During intervention sessions, the researchers used gestural prompts to guide students through the construction of simple stories while the software provided auditory feedback as students wrote. Data indicated that all nine participants acquired the targeted skills, demonstrated maintenance, and that some responded across untrained topographies (i.e., speaking, handwriting). Most recently, Pennington and Koehler (2017), taught students with MSD to generate stories using modeling, story templates, and self-graphing. First, the authors presented a short video vignette on a computer tablet and a story template containing sentence starters followed by underlined blank spaces. The teacher asked the student to identify story elements depicted in the video, completed the template, and then read the story with the student. Subsequently, the teacher asked the student to write their own story and then upon completion, graphed the number of story elements used in their narrative. All students increased their use of story elements and maintained improved performance during maintenance probes.
Interestingly, there exists a gap in the literature in the development of writing skills that fall between the spelling and the generation of simple narratives. For example, few research teams directly have addressed the instruction of sentence writing skills. Sentences represent smaller syntactic units that serve as the building blocks of more complex narratives. These building blocks can range in complexity and be viewed as their own “miniature compositions” (Saddler, 2012, p. 6). Only three research teams have targeted sentence writing for students with MSD. In the first investigation, Yamamoto and Miya (1999) taught students to select words from an array to construct five-word sentences about a picture. The computer software provided digital feedback (i.e., beep for errors, fanfare for correct) contingent on their responses. During intervention, the computer trained students to construct sentences in the presence of three pictures. Following intervention, the students produced sentences in response to 24 untrained stimuli. Similarly, Basil and Reyes (2003) used a multimedia software package to teach students to select words to construct a sentence. The software provided students with opportunities to “freely” select words to construct sentences and then provided feedback in the form of a digital reading of the sentence and an animated depiction of sentence content. The researchers reported improvements in sentence production skills across all participants and gains in spelling and phonological synthesis. Despite their optimistic findings, both research teams used experimental designs that precluded their ability to demonstrate a functional relation between their intervention package and treatment outcomes.
To date, there exists a single study demonstrating a functional relation between intervention and the acquisition of sentence writing skills by students with MSD. Pennington, Flick, and Smith-Wehr (2016) used response prompting procedures and sentence frames to teach students to write sentences in response to three types of instructional demands. First, the researchers taught students to write sentences to request access to a preferred stimulus (i.e., I want a________). Second, they presented a pictured stimulus and asked the students to write about what they saw (i.e., I see a_______). Finally, the teacher presented a pictured stimulus and asked the students to describe the animal in the picture (i.e., The_____is ______). For two of the students with motor impairments, the teacher used SLP to teach the selection of words from an array presented on a tablet. For the third student with an existing handwriting and spelling repertoire, the teacher used constant time delay (CTD) to teach the student to generate handwritten sentences. The researchers used a multiple probe design across writing tasks, and data suggested that the intervention was effective for all three participants. Furthermore, all three participants maintained responding at levels above those during baseline conditions and demonstrated generalized responding to untrained stimuli.
The paucity of research on teaching sentence writing to students with MSD warrants further investigation into this critical area of instruction. In addition, the three studies described above primarily involved sentence construction tasks in which students have words available from which to choose. To date, researchers have not demonstrated the effectiveness of intervention on the generation of sentences by students with MSD. These generation tasks are likely more difficult as they require students to use their working memory to recall words and arrange them in a particular order prior to and during writing. Therefore, the purpose of the current investigation was to evaluate the efficacy of instruction on the sentence writing skills of students with MSD. We addressed the following research questions:
Method
Participants
Three males (ages 12–13) with developmental disabilities participated in the study. The participants attended a self-contained special education classroom for students with multiple disabilities within a public middle school. Dino was a 13-year-old Black male with ASD (Severely Autistic range; Childhood Autism Rating Scale [CARS], Schopler, Reichler, DeVellis, & Daly, 1980). Though Dino’s intellectual functioning fell in the borderline disability range (i.e., Full Scale Intelligence Quotient [FSIQ-71]; Wechsler Intelligence Scale for Children–Fourth Edition [WISC-4]; Wechsler, 2003), his most recent formal language assessment indicated a severe delay in communication (i.e., 57; Comprehensive Assessment of Spoken Language [CASL]; Carrow-Woolfolk, 1999) and performance in the “extreme lower” range in written expression (i.e., 61; Kaufman Test of Educational Achievement–Second Edition [KTEA-2]; Kaufman & Kaufman, 2004). Carmen was a 12-year-old White male with Down syndrome and a moderate intellectual disability (i.e., FSIQ-46; Standford–Binet Intelligence Scale–Fourth Edition; Thorndike, Hagen, & Sattler, 1986). Carmen’s most recent formal language assessment indicated a severe delay in communication (i.e., Listening-40, Oral expression-40; Oral and Written Language Scales [OWLS]; Carrow-Woolfolk, 1996) and well below average performance in reading and written expression (i.e., Reading subtest-38, Written expression subtest-19; Woodcock Johnson Tests of Acheivement–Third Edition; Woodcock, McGrew, & Mather, 2001). Allan was a 13-year-old White male with ASD (i.e., 24; Autism Diagnostic Observation Schedule [ADOS], Lord, Rutter, DiLavore, & Risi, 2001) and a moderate intellectual disability (i.e., FSIQ-46; WISC-4). His most recent formal language assessment indicated a severe delay in communication (40-Listening, 40-Oral expression; OWLS) and below average performance in written expression (i.e., Writing score-5; Basic School Skills Inventory–Third Edition [BSSI]; Hammill, Leigh, Pearson, & Maddox, 1998). Though formal reading assessments were not available, all three students were identified by their teacher as primarily sight word readers. Prior to intervention and when asked to write, Dino could generate multiple words but not syntactically correct sentences (e.g., teacher arm hammer out). Carmen and Allan produced lists of single words. All participants had previously produced handwritten text and demonstrated the ability to copy from a written a model.
A special education teacher with an undergraduate degree in moderate to severe disability and elementary education conducted all the sessions. The teacher had 4 years of teaching experience and was currently enrolled in a master’s degree program in ASD and applied behavior analysis.
Settings and Materials
The teacher conducted all sessions in the special education classroom, in a 1:1 instructional arrangement, and at a small table (1.2 × 2.43 m) located in the back of the classroom. During each session, the participant sat next to the teacher, while eight other students received instruction from paraprofessionals.
Instructional stimuli included six colored photographs of animals and 36 sentences written on index cards (7.62 × 12.7 cm). Though animals were not a part of the participants’ current curriculum, we selected to use them because of their familiarity to all the participants. Each color photograph depicted only a positioned animal against a white background. For each picture, we constructed six sentences for use during instruction. We selected three simple sentence formats to teach, (i.e., The [subject] is [adjective], The [subject] [verb], and The [subject] feels [adverb]) and created two different exemplar sentences for each format to potentially facilitate generalization to untrained contexts. In addition, we presented the participant with lined paper and a pencil during instruction and a lined notebook and a pencil during generalization tasks.
Measurement of Dependent Variables
During probe and intervention conditions, we collected data on the percentage of correct sentences written. A sentence was scored as correct if it matched the structure of the targeted sentence type and accurately described the picture. Sentences could be scored as accurate if they included novel descriptors or ones presented as models during instruction. We ignored participants’ omissions of capitalization and punctuation.
During generalization probes, we scored the number of overall sentences used during unstructured writing opportunities. A sentence was scored as correct if it contained a minimum of a subject followed by a verb and was syntactically correct.
Reliability
A second observer, either the first or third author, collected dependent variable and procedural fidelity data across probe and intervention conditions. We collected data for 50% of probe sessions and 40% of intervention sessions for Dino, 45% of probe and 50% of intervention sessions for Carmen, and 50% of probe and 48% of intervention sessions for Allan. We also collected data for 54%, 50%, and 36% of generalization probes for Dino, Carmen, and Allan, respectively.
We used the point-by-point method to calculate interobserver agreement (IOA) across raters. The number of agreements was divided by the total number of agreements plus disagreements and multiplied by 100%. We determined the mean agreement across participants to be 99.1% across probe conditions (i.e., 100%, 98%, 100%) and 99.3% across intervention conditions (i.e., 100%, 97.5%, 100%). We also used the point-by-point method to calculate IOA across the number of sentences used during journal writing (i.e., generalization) probes. Two coders, the first and third authors, scored randomly selected writing samples. Following the calculation of IOA, the coders met to reconcile the scores and agreed upon final score for all disagreements. Prior to reconciliation, IOA was 93% for Dino, 82% for Carmen, and 100% for Allan.
We collected procedural fidelity data on the teacher’s implementation of 46 steps during probe and 26 steps during intervention sessions. We calculated procedural fidelity by dividing the number of observed steps by the number of planned steps and then multiplying by 100%. Procedural fidelity was 99% across probe and 99% across intervention sessions.
Experimental Design
We used a multiple probe (Horner & Baer, 1978) across sentence types to evaluate the efficacy of the intervention package on the participants’ acquisition of sentence writing responses. Prior to intervention, we conducted at least three full probe sessions across all three sentence types. We then introduced intervention to the first sentence type (i.e., The [subject] is [adjective]) and continued until the participant met criterion levels of responding (i.e., 100% accuracy across three consecutive days). We then conducted three full probe sessions across all sentence types and subsequently introduced intervention on the next sentence type (i.e., The [subject] [verb]). Again, once the participant met criterion on the second sentence type, we conducted three full probe sessions and introduced intervention on the final sentence type (i.e., The [subject] feels [adverb]). We selected this design because of its capacity to demonstrate a functional relation with nonreversible behaviors and because it limits participants’ exposure to baseline conditions, whereas they would be asked to perform difficult writing tasks in the absence of instructional support. In addition, we collected intermittent data on students’ writing responses during unstructured journal writing activities to assess generalized responding.
Procedures
Teacher training
We used behavioral skills training (BST) procedures to instruct probe and intervention procedures. Prior to baseline and training sessions, the first author sent a list of procedures to the classroom teacher (second author) via email. Subsequently, the two met to review procedures and then engaged in didactic training. First, the researcher modeled the procedures and then directed the teacher to perform them until she completed each step with 100% accuracy.
Probe conditions
During probe sessions, the teacher assessed participants’ performance across three sets of writing tasks. The teacher presented three probe trials per writing task for total of nine trials per session. We only presented three trials per stimuli to reduce the amount of writing during probe conditions and to limit potentially inhibitive effects of increase response effort during baseline performance. At the onset of each session, the teacher presented paper and a pencil to the participant, followed by vocal signal for the probe session to begin (e.g., “Time to work on writing”). During each trial, the teacher delivered an attention cue (e.g., “Look at me”), presented a picture of an animal, delivered the directive associated with each sentence type (i.e., “Write a sentence telling me what the ________looks like,” “Write a sentence telling me what the ___________ is doing,” “Write a sentence telling me how the _______feels”), and stated, “Remember, if you need help spelling you may ask.” The teacher waited 15 s for the participant to start writing. If the student did not start writing, stopped writing for 15 s, or indicated that they were finished (e.g., vocal response, pushed paper away), the teacher discontinued the trial and delivered nonspecific praise for on task behavior.
Intervention conditions
During intervention conditions, the teacher presented a single directive (e.g., “Write a sentence telling me what the ______looks like”) until the participant met criterion (i.e., 100% accuracy across three consecutive sessions). At the beginning of each session, the teacher stated, “Today we are going to work on writing sentences.” The teacher then presented the rule, “A sentence names a character and tells more (Kame’enui & Simmons, 1990),” asked the participant to repeat the rule and then delivered praise contingent on his response. Next, the teacher implemented a CTD procedure. First, the teacher conducted two sessions of 0-s delay trials. During each session, the teacher conducted five trials. During each trial, she presented a picture of a different animal and a directive for the participant to write. She then randomly selected one of the written models and presented it to the students with the statement, “You can use this to help you.” After the participant wrote the modeled sentence, the teacher read the sentence aloud and delivered vocal praise. Finally, the teacher asked “Did you follow the rule” and then pointed to the subject and predicate while labeling them as “character and more.” The participant was not required to answer the teacher’s question. On the third session, the teacher began 15-s delay trials, and again, she conducted five trials per session. The teacher presented a picture of an animal and a directive and then waited 15 s for the participant to start writing. If the participant did not start writing within 15 s, the teacher presented the written model. If the participant made an error or stopped writing for 15 s, the teacher interrupted the response, erased the incorrect word, and presented the written model. Errors included writing a word out of sequences, omitting a word, writing about a subject that did not correspond to the picture, or using a sentence format that did not correspondent to the directive. The teacher delivered vocal praise for prompted and unprompted correct responses.
Maintenance and generalization probes
Following participants’ meeting of criterion on each sentence type, we conducted maintenance probes using procedures that were identical to those in full probe sessions. To assess whether participants would generalize their sentence writing responses to new contexts, we provided students with the opportunity to write in a journal when they arrived at school (i.e., approximately 2 hr before probes or intervention sessions). We collected generalization data during the first baseline probe condition and following the introduction of intervention on the first sentence type. The teacher presented a wide ruled notebook to the participants and directed them to write about a topic of their choice. She also reminded him that they could ask for spelling assistance and to vocally indicate when they were finished. The students were permitted to respond until they stopped writing for 1 min. At that time, the teacher asked the students whether they were finished. If the students indicated that they were not finished, they were permitted to continue until another minute without writing elapsed. The teacher delivered nonspecific praise for writing attempts.
Results
During baseline sessions, all three participants did not consistently generate any of the targeted sentence structures. Following the introduction of intervention, the participants generated all three types of sentences and maintained levels of responding at levels above those in baseline sessions. Furthermore, the participants wrote more sentences in their journals directly following the introduction of intervention.
Dino
Acquisition
Dino’s performance is depicted in Figure 1. During baseline sessions, Dino did not generate any sentences. Following the introduction of intervention on the first sentence structure, data showed an immediate effect (i.e., change in level) and that he met criterion within six sessions. On the next sentence type, Dino’s data, again, showed an immediate effect and a rapid acceleration to criterion (i.e., within six sessions). On the final sentence type, Dino’s data showed an immediate effect and that he met criterion within four sessions.
Dino’s performance across baseline and intervention sessions and generalization probes.
Maintenance probes
During maintenance probes, Dino maintained responding at nearly 100% accuracy with the exception of Session 28. In this session, he responded correctly to probes on the first sentence type but made errors on the other sentence types.
Generalization probes
Prior to intervention, Dino wrote one to two sentences during journal writing tasks. Immediately following intervention, we observed an immediate change in level as Dino wrote five sentences in his journal. Following 2 days of writing five sentences, we observed a gradual decelerating trend as the number of sentences used in his writing decreased to two and three sentences. One opportunity to write was provided following intervention across all three sentence structures. During this opportunity, he generated five sentences in his journal.
Carmen
Acquisition
Carmen’s performance is depicted in Figure 2. During baseline sessions, Carmen did not generate sentences. Following the introduction of intervention on the first sentence structure, data showed an immediate effect and a steadily accelerating trend toward criterion. He met criterion within seven sessions. On the next sentence type, Carmen’s data showed an effect during Session 2 and variable performance until he met criterion. He met criterion within 13 sessions. On the final sentence type, Carmen generated three sentences during the second session and demonstrated variable but high levels of responding until he met criterion in the 12th session.
Carmen’s performance across baseline and intervention sessions and generalization probes.
Maintenance probes
During maintenance probes, Carmen showed variable performance. Overall, he maintained responding on average at levels above those in baseline sessions. Carmen generally produced sentences during these probes but often failed to discriminate between the requests and generated sentences using the most recently taught structure.
Generalization probes
Prior to intervention, Carmen generated lists of words in his journal. Following the introduction of intervention, he immediately increased the number of sentences written (i.e., five sentences), but then, his data reflected a gradual decelerating trend to the use of three or four sentences per opportunity to write. Three opportunities to write were provided following intervention across all three sentence structures. During these opportunities, he generated three or four sentences.
Allan
Acquisition
Allan’s performance is depicted in Figure 3. During baseline sessions, Allan did not generate sentences. Following the introduction of intervention, Allan’s data showed an immediate effect and rapidly accelerating trend. He met criterion within seven sessions. On the next sentence type, Allan generated two sentences during the second session and continued to make variable but rapid progress to criterion. He met criterion in 10 sessions. On the final sentence type, Allan generated five sentences during the second session and demonstrated variable but high levels of responding until he met criterion in the 10th session.
Allan’s performance across baseline and intervention sessions and generalization probes.
Maintenance probes
During maintenance probes, Allan maintained responding at levels higher than those during baseline sessions. Similar to Carmen, Allan consistently produced sentences during these probes but often failed to discriminate between the requests and generated sentences using the most recently taught structure.
Generalization probes
Prior to intervention, Allan generated list of words in his journal (e.g., days of the week). Following intervention, Allan demonstrated an immediate increase in the use of sentence in his journals. Following several days of writing five sentences in his journal, his data reflected a gradual deceleration in his generation of sentences. Three opportunities to write were provided following intervention across all three sentence structures. The resulting data reflect that the levels of sentence usage were higher than those during baseline sessions.
Nonparametric Measure of Effect
We used Tau-U (Parker, Vannest, Davis, & Sauber, 2011) to measure nonoverlap between phases and to provide a nonparametric measure of effect. We first measured nonoverlap between baseline and intervention conditions across all three tiers (i.e., sentences types) for all three participants. We then calculated an overall effect across all participants. For Dino, Carmen, and Allan, we calculated the effect to be 1.0, .94, and .90, respectively. The weighted average across all participants was .94, indicating a strong effect.
Discussion
In the current investigation, we evaluated the effects of an instructional package on the sentence generation of three middle school students with MSD. We also sought to determine whether explicit writing instruction would produce changes in students’ use of sentences during unstructured writing activities (i.e., journal writing). Our findings suggest a functional relation between the instructional package and the number of accurate sentences generated during intervention sessions. During maintenance probes, students’ responding was variable but overall reflected levels were higher than those during baseline conditions. On multiple probes, the participants wrote complete sentences but failed to discriminate between teacher directives, thus writing a single sentence type in response to all three directives. In addition to the following instruction, all three participants used more sentences during journal writing activities. These findings are important in that they add to the accumulating evidence suggesting that students with MSD benefit from explicit writing instruction and further the case for targeting the development writing skills within the curriculum of these unique learners.
Overall, our findings corroborate those of Pennington et al. (2016) in demonstrating the utility of response-prompting strategies and predictable writing routines for improving the sentence writing skills of students with MSD. In both the investigations, researchers used response-prompting strategies that have been demonstrated to be effective across many applications but also presented predictable sentence models or frames that may have served to support student learning. Researchers and interventionists have long-used frames (e.g., I want, I see) to support students in learning and generalizing communication responses (Betz, Higbee, Kelley, Sellers, & Pollard, 2011; Ganz & Simpson, 2004; Hernandez, Hanley, Ingvarsson, & Tiger, 2007). Furthermore, researchers in the area of writing instruction for learners with high incidence disabilities have recommended the importance of predictable routines in the design of writing instruction (Graham & Perin, 2007). Additional work is required to understand the role of and refine the use of sentence frames during intervention. In the current study, we taught students to write using one sentence frame at a time and in response to a limited set of stimuli. We found that during maintenance probes, students often erred by using the single frame in response to different writing requests. For example, given the request to write about how an animal might feel, the participant used the frame the (subject) (verb). It may have been more effective to teach students to use multiple frames to write in response to different requests during each intervention session. This enhanced multiple exemplar approach might have improved students’ responding during maintenance probes, resulting in the increased use of sentence types during generalization activities. We also presented a single picture for each referent (i.e., animal) during each instructional session. This may have had an inhibitive effect in the use of frames outside of the training context. That is, the amount variation in instructional stimuli was not sufficient as to facilitate the participants’ transfer of the new writing skill to new antecedent conditions.
Despite our findings suggesting that changes in students’ increased use of sentences in their journal entries corresponded with the introduction of intervention, it is unclear as to what variables may have contributed to these changes. Only Allan consistently used sentences trained during intervention within journal entries. In fact, they comprised the majority of his entries, though he often included untrained subjects and predicates. Interestingly, Dino and Carmen included few of the sentences trained during intervention. Their entries generally included a combination of complete sentences and fragments. Dino retold segments of a narrative from his favorite cartoon, whereas Carmen wrote about daily events (e.g., eating a hotdog, spending the night with a friend). Dino’s differences in responding across tasks may indicate that he could generate sentences prior to intervention, but they were under control of a limited set of stimuli. Carmen’s differences appeared to be a result of an increase in his use of requests for spelling assistance. During intervention, the teacher reminded Carmen at the beginning of intervention that he could request spelling assistance. He often received a reinforcer (i.e., spelling assistance) during intervention sessions which may have resulted in increased requests for spelling help during journal writing tasks. During journal writing tasks, he asked for spelling assistance often.
It is also important to note that participants’ responses during journal writing activities may also be related to their vocal communication repertoires. Dino and Allan produced restrictive narratives around either a limited set of sentence structures or content. Though all three participants were reported to have severe communication delays, Dino and Allan, both with ASD, often used repeated vocal statements during social interactions.
Finally, we used a dependent variable (i.e., percentage of accurate and complete sentences) that was not highly sensitive to change. For example, our measure was not able to capture a student’s progression from generating a single word about a picture to the generation of a syntactically correct two-word sequence (e.g., dog jumps, big dog). In the future, researchers and practitioners might consider using correct writing sequences (CWS; Videen, Marston, & Deno, 1982) to measure student’s progress in generation of text. A CWS is two adjacent words that are spelled correctly and make syntactic sense. In addition, the first-scored CWS occurs between the blank and the first word in a writing sample and the last occurs between the final word and the blank. The CWS measure, as well as other curriculum-based measures, is well established for use with students with learning disabilities (Romig, Therrien, & Lloyd, 2016) but has not been used in the research literature for students with MSD. Future researchers should consider the use of CWS with students with MSD to capture students’ progress.
Limitations
Several limitations should be considered when interpreting our findings. First, we conducted this investigation across only three participants. We attempted to strengthen our findings by using a design that permitted for both intersubject and intrasubject replication, but further replication will be required to increase the confidence in the efficacy of the test intervention package. Second, we elected to conduct only three baseline probes prior to introducing intervention on the first type of sentence to reduce potential student frustration with the experimental arrangement. This is a violation of a recommendation set forth in Kratochwill et al. (2013), for a minimum of five baseline data points, and limits one’s capacity for interpreting stability prior to intervention. Third, we failed to collect baseline probes at least once in every 8 days for the Carmen and Allan’s third sentence type. Again, this violates the aforementioned recommendations, resulting in a decreased inability to detect cyclical variability. Fourth, we did not collect procedural reliability data during the journal writing probes, and therefore, the resulting data are threatened by potential errors in the teachers’ presentation and feedback during these probes. It is also important to note that the teacher provided fewer opportunities to write in the journals as the students progressed. This may have affected students’ performance on these tasks, and the absence of the probe data may have obscured relevant response patterns.
Conclusion
In light of the ubiquity of written expression in the daily life of most individuals, it is surprising that we do not yet have a well established set of practices for developing this complex repertoire in learners with MSD. Without this technology, these students are at further risk for isolation as they are precluded from a range of interactions that occur between readers and writers. In the current study, we sought to add to the small but emerging body of literature demonstrating that students with MSD are amenable to explicit writing instruction. Our findings indicated that we met these objectives but also highlighted the great expanse of work yet to be done.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.