Behavioral Vaccine or Social Suppressant: The Good Behavior Game, a Scoping Review

Arksey and O’Malley ( 2005 ) recommend that scoping review searches of the literature base should achieve both in-depth and broad results, rather than being guided by a highly focused research question (RQ). As such, we initially identified two broad RQs targeting the specific behavior of interest (RQ1) and the capture of unplanned mentions of this behavior within texts (RQ2). Owing to this breadth, the processes of scoping reviews are “not linear but iterative” (Arksey & O'Malley, 2005 , p. 22). Arksey and O’Malley expressly state their methodology “requir[es] researchers to engage with each stage in a reflexive way and, where necessary, repeat steps to ensure that the literature is covered in a comprehensive way” (p. 22). When our research team noticed social interactions frequently included within operational definitions, we added a third RQ to analyze this phenomenon and adjusted our final codebook to systematically explore this pattern. 1. How many studies explicitly target social interactions as a dependent variable? How many studies explicitly target social interactions as a dependent variable? 2. How many and to what extent do articles on the GBG mention unplanned anecdotal insights related to social interactions? How many and to what extent do articles on the GBG mention unplanned anecdotal insights related to social interactions? 3. How do studies include social interactions within operational definitions of target behaviors? a. Specifically, are increases or decreases in social interactions targeted? b. Are the dependent variables (DVs) that included social interactions intended to increase or decrease? c. On the basis of reported results, did the DVs that included social interactions increase or decrease as intended? How do studies include social interactions within operational definitions of target behaviors? a. Specifically, are increases or decreases in social interactions targeted? Specifically, are increases or decreases in social interactions targeted? b. Are the dependent variables (DVs) that included social interactions intended to increase or decrease? Are the dependent variables (DVs) that included social interactions intended to increase or decrease? c. On the basis of reported results, did the DVs that included social interactions increase or decrease as intended? On the basis of reported results, did the DVs that included social interactions increase or decrease as intended? We conducted searching and screening procedures in February 2024 using Academic Search Ultimate through EBSCO Host. We selected the following databases: Academic Search Ultimate, APA PsycINFO, Education Source, ERIC, and the Journal of Applied Behavior Analysis through Wiley. We searched each database individually (not as a collection) and screened the results for duplicates using Zotero (Zotero, 2024 ). Our final search terms included: “Good Behavior Game” OR GBG OR “Caught Being Good Game” OR CBGG ensuring broad coverage. The initial searches yielded a total of 1006 articles across databases. We then limited results to peer reviewed academic journals in English resulting in 886 articles across databases (Academic Search Ultimate: n = 383; Education Source: n = 135; ERIC USU: n = 215; ERIC: n = 96; PsycINFO: n = 216; Jaba : n = 65). We conducted a first round of deduplication in Zotero, resulting in 527 articles. We then imported these articles ( n = 527) into Rayyan, a systematic literature review software (Ouzzani et al., 2016 ), where we removed an additional nine duplicate articles ( n = 518). From there, we removed author indexes, leaving a total of 499 studies for title and abstract screening. We did not limit the searches by year or geographical location. The screening process included: (1) title and abstract screening, (2) double screening and conflict resolution, (3) full text screening, and (4) full text conflict resolution. Figure 1 displays the PRISMA ScR flowchart. PRISMA Flow Chart To be included in the review, articles had to meet the following inclusion criteria: (1) Be a peer-reviewed, empirical study (training materials, intervention materials, opinion pieces, literature reviews were not included); (2) Be single subject or group experimental design; and (3) The title or abstract must mention the Good Behavior Game, GBG, Caught Being Good Game, or CBGG. We included all locations, date ranges, and settings in the initial screening. We excluded articles if the title or abstract stated an inter-dependent group contingency and not GBG specifically. Although common in GBG literature base, CW-Fit and Step-It-Up interventions are referred to as group contingencies separate from the GBG, so we excluded these variations. The first author trained researchers on screening articles using the above inclusion and exclusion criteria. A team of three researchers completed the first stage of article screening (title and abstract screening, n = 499), by selecting “yes,” or “no” within the Rayyan program (Ouzzani et al., 2016 ). If two researchers had a disagreement (e.g., one selected “yes” and the other selected “no”), researchers met as a team to review and screen the article to resolve the conflict. The first and third author double screened all included studies ( N = 499) and the fourth author triple screened a random 49% ( n = 254) of the sample. Throughout the title and abstract screening there were 28 conflicts, resulting in inter-rater reliability (IRR) of 95% and 142 articles moving to the full text screening stage. The first and second author double screened all articles at this level of screening ( n = 142; IRR = 90%) with 71 articles included for full text coding. We included texts for coding if social interactions by participants, as we defined them below, occurred anywhere in the text. We replicated the definition of social interactions from Walton, Preston, Hoffmann et al. ( 2026 ), that was originally adapted from Groves and Austin ( 2019 ): Positive peer interactions included verbalizations or statements aimed at encouraging peers, congratulating each other on performance (e.g., “That work is great,” “Well done!”), and requests or offers of assistance to one another. Positive physical gestures, which occurred when students were oriented to a peer and used motions such as thumbs up, waves, or high fives. Negative peer interactions included verbalizations or gestures that threatened, provoked, or demeaned a peer (e.g., name calling, laughing at peer mistakes, threatening to hit a peer), or interfering with a peer’s work (e.g., taking a peer’s book or materials). In accordance with Speltz et al. (1982), an interaction was considered directed toward a peer when the student used the peer’s name, their body was oriented toward the peer, or they had physical contact with the peer. (Walton, Preston, Hoffmann et al., 2026 , p. 6) The first and second author developed the codebook in Qualtrics with coders completing a survey for each article. Our team identified coding for group designs and single case designs on the same codebook was ineffective. We therefore split articles into single case ( n = 38) and group studies ( n = 33) with single case articles being coded for this paper. Group designs will be presented in a forthcoming paper. All single case studies ( n = 38) were double coded by the first and second author, with IRR of 93% agreement (range, 80% to 100%). At this stage, we identified the pattern of social interaction behaviors being included in larger operational definitions (meaning how behaviors targeted for change were described, they should be clear, concise, unambiguous, and objective; Cooper et al., 2020 ) and were repeatedly targeted for reduction. To understand this pattern more, we conducted additional coding of combined behavior classes (meaning the overall group, e.g., disruption), subclasses (meaning smaller behavior groups, e.g., verbal disruption and physical disruption folded into the main grouping of disruption), social topographies (examples of the behaviors under each class, e.g., hitting, talking to peers), and graphed variables for all articles identified as targeting behavior that included a social interaction. We then thematically coded behavior classes. Of these articles ( n = 27), 48.1% ( n = 13) were randomly selected for IRR with agreement of 92.5% (range, 72% to 100%). We exported all data from Qualtrics to Excel for analysis. We summarized all articles by the following features: experimental study design, location, participant demographics, school demographics, study purpose, intervention characteristics, dependent variable characteristics, and other reported outcomes. Single-subject designs were coded across the following designs: AB, ABAB reversal, variation of the ABAB reversal, alternating treatment design, multiple baselines across participants, multiple baselines across settings, variation of a multiple baseline (such as an embedded reversal), multi-element design, multiple probe design, or other. Our team collected information regarding the country of implementation as well as geographic region. Geographic region included information such as rural, urban, suburban, or not reported to identify reporting standards within the literature base. We coded for multiple participant characteristics, including total number of participants reported broken down into number of students, number of teachers, and number of schools. Researchers coded for participant age range and/or grade range and classroom type, which included options such as general education, special education resource room, special education self-contained/small group/Pupil Referral Units (PRU; British special education schools for US equivalent of Emotional and Behavioral disability [EBD]), special education additional services such as applied behavior analysis (ABA) clinic, camp, ABA lab, residential/hospital, or other school setting such as cafeteria or hallway. Finally, researchers coded for any reported disability classification including ASD/autism, ADHD/ADD, specific learning disability (SLD), intellectual disability (ID), EBD, oppositional defiance disorder (ODD), behavioral disorder, and/or speech language impairment (SLI). When a disability classification was encountered but not listed, the codebook was adjusted to account for this. We charted these data using frequency tables, binary frequency tables, and we assessed how disability reporting varied across setting and grade level using Excel filtering. The research team coded a variety of school-level variables. These included school type, such as elementary/middle/high/post-secondary, residential, charter, or pre-school. Additional coded variables included the percentage of students receiving free and reduced-price lunch (FRPL), percentage of students receiving Individualized Education Plan (IEP) services, Title 1 status, and the percentage of students reported as English second language (ESL) students. We synthesized data across binary frequency tables representing percentage of articles that did or did not report each variable. Researchers coded the reported aims of each study including: comparison of game variations, comparison of the GBG intervention with another intervention, outcomes of the GBG on secondary non-targeted behaviors, behavior reduction, behavior increase, both behavior increase and decrease, component analysis of different intervention features. We numerically analyzed these data within a table to identify patterns of aims. Researchers coded a variety of intervention characteristics. We reported if the intervention used was the GBG or the CBGG. Included game rules were coded as providing the full list, only examples, or failed to report rules entirely. We also reported if the studies provided rationale for the selection of the rules, and if the rules of the intervention were aligned with the targeted DVs. We analyzed these data for frequency of responses, characteristic overlap, and study aim overlap. For coding of the DV characteristics, we first identified if the included studies targeted a social interaction as either (a) its own response class, meaning social interaction behaviors were their own DV; or (b) if a social interaction was included as part of a combined DV or combined response class. All DVs with inclusion of social interactions were coded for purpose; options included: targeted for increase, decrease, both, or general observable effects, and researchers reported the provided definition. Of the studies that included social interactions as part of a larger dependent variable, we coded the name of the combined class, all subclasses, and all social topographies; these were then thematically grouped. Finally, our team coded if studies included a psychometric tool that included a form of student social outcome or if any anecdotal reporting of student social interactions was reported within articles. We summarized our analysis and synthesis by charting data across frequency tables in Excel, heat maps, binary frequency tables displaying percentage of occurrence across articles to account for varied sample sizes across educational settings. Table 1 presents our data on publication outlets. The majority of articles included in this review were published in the Journal of Applied Behavior Analysis (39.5%; n =15). Followed by the Journal of Positive Behavior Interventions (10.5%; n = 4). Twelve other journals were included with the majority publishing either one or two articles. These distribution data indicate the GBG studies that included social interactions have a wide audience and reach, with many journals focusing on research in schools and behavior intervention support. The year of publication for each article is represented in Fig. 2 and these range from 1969 to 2022. Publication of articles on the GBG that included social interactions significantly increased after 2011, with 31.6% ( n = 12) being published between 2020 and 2022, indicating an increasing trend of social interactions being included within publications in some form. Frequency of Articles Published Across Decades The articles included in this review were largely published from within the USA (81.6%; n = 31). The UK represented 10.5% ( n = 4) of the articles, and other European countries represented 7.9% ( n = 3) of the articles. Figure 3 shows most of the research in this area is focused within the USA. Additionally, geographic regions were coded in this review, and the majority of articles (52.6%; n = 20) did not report this information. Pie Charts of County of Implementation (left) and Geographic Location (right) This review included single case research designs. The most common research designs included variations of ABAB (50%; n = 19) and the multiple baseline design across settings and participants (28.9%; n = 11). There were additional variations reported with embedded reversals (return to baseline conditions) consisting of 10.5% ( n = 4) of the articles included. Figure 4 displays the percent each design type was present across articles. Percentage of Design Type Presence Across All Articles Each article’s research aims were analyzed using a binary frequency table of presence or absence due to the unit (aim) not being mutually exclusive. This allowed for articles with multiple aims to be reflected in the data. These data are represented in Fig. 5 and indicate decreasing problem behavior alone (47.4%; n = 18) as the most common occurring research aim. Fewer articles explored increasing desired behavior while decreasing problem behavior (15.8%; n = 6). The second most common purpose within included articles was “other purpose” (23.7%; n = 9) which included paper-specific research questions (e.g., feasibility [Donaldson et al., 2018 ]) or teacher effects [Joslyn & Kronfli, 2021 ; Rubow et al., 2018 ]) Percentage of Study Aim Presence Across All Articles We coded articles for GBG variation and how rules were reported within each article. Figure 6 displays data on study variation. The most common variation investigated was the GBG (71%; n = 27), followed by a combination of the GBG and CBGG (26.3%; n = 10), and then CBGG alone (2.6%; n = 1). We identified the articles that investigated the GBG, with the inclusion of social interactions, the majority aimed to decrease behavior (93%; n = 25). Game Variation and Purpose Across Articles Regarding the reporting of game rules, 57.9% ( n = 22) provided the full list of rules, 10.5% ( n = 4) provided only narrative examples, and 31.6% ( n = 12) provided neither rules or examples within the text. Game rules across educational settings are displayed in Fig. 7 . Frequency of Rules Shared Across Studies Researchers coded for a variety of school-level demographic variables to map the extent to which additional contextual information was provided. Table 2 displays these data. The majority of articles did not report socioeconomic status (SES) across all variables. Of the included articles, 71.2% ( n = 27) did not report percentage of students receiving free and reduced-priced lunch ( n = did not report the percentage of students receiving ( n = of articles did not report Title 1 and ( n of studies did not report the percentage of students. Of the included articles, ( n = 13) reported participant disability Each disability was coded using binary frequency The most frequently represented classification reported was ( n = followed by ( n = and were each represented at ( n = disability was present in two studies and was reported for one study Of the articles that included social interactions ( n = occurred within educational settings through high and occurred in special education settings (e.g., for students with special education We across grade level and educational setting in Table 3 . The majority of the articles included in this review occurred within general education settings n = Of these articles, only two & ; & ) explicitly reported student percent ( n = 12) of articles occurred within special education settings. Of these articles, ( n = 4) occurred within special educational settings (e.g., resource group and ( n = occurred within education settings. For a social interaction to be considered a specific the behavior had to be defined as its own response with only social interaction topographies (e.g., defined as or If the operational definition provided a response with social interaction topographies other topographies (e.g., to talking to peers), then we considered the social interaction as a combined Of the included ( n = across both and defined social interaction behaviors as its own study (2.6%; Groves & 2020 ) anecdotal data Of the included articles, ( n = across both and included social interactions as part of combined Table 4 displays the distribution of combined and specific DVs across educational setting & and across study Of articles within general education ( n = ( n = included some form of social interaction within a larger Of the articles conducted within special education and n = ( n = 9) combined some form of social interaction into a larger Of the articles that combined social interaction into a larger DV ( n = ( n = 27) targeted this DV for reduction. Of total articles with combined DVs ( n = 22) combined social interaction topographies (e.g., talking to first into larger subclasses (e.g., verbal disruption), and then into one DV (e.g., that was Articles this data as disruption n = target behaviors ( n = behaviors ( n = behaviors ( n = or combined problem behaviors ( n = 1). The ( n = of articles combined multiple social interaction topographies talking to into a larger behavioral verbal but graphed each in their results Across the articles that combined social interactions within a larger DV for behavior reduction, the frequency distribution of behavior subclasses combined from two to ( = = Across included articles ( n = there were total subclasses ( = 1 to and total social topographies ( = 1 to Across all included articles ( n = there were total subclasses ( = 2 to and total social topographies ( = 1 to Across all articles with combined DVs ( n = 27), of the articles included the of or talking occurred in of the articles, n = n = Figure displays a binary frequency of occurrence across settings. subclasses included in articles were verbal talking to and within of Presence for DV for Figure displays the percentage each social interaction occurred within articles. special education, of all articles ( n = 9) included talking to within a behavior with the to the combined behavior class. general education the most occurring social n = 12) included into a larger of disruption was peers, with peer’s this being the third most behavior is to that the social behavior did not have separate data so be a decrease in the combined DV was to of social or a combination of the of Presence for DV for coding and in-depth analysis of results, we analyzed the specific study outcomes 5 ). Study aims and behavioral outcomes had high of Of all included articles ( n = reported effects of behavior (81.6%; n = behavior increase (10.5%; n = and increasing and decreasing behaviors n = study did not include social interactions within their DVs but was included to inclusion of anecdotal social information & 2020 ). articles targeting social interactions as their own only one included a psychometric of social behaviors & 2016 ). The only used was the & ). Additional anecdotal information regarding social interactions was identified in studies across social results, and analysis identified (a) the to change (e.g., we n = (b) anecdotal of peer interactions (e.g., n = peer (e.g., an opinion of of team of one n = and of classroom ( n = 1). This scoping review was limited to articles published in conducted in countries where English is not the language of publication not have which the of literature and a into the analysis. Behavioral definitions, and intervention across and not represented in this research should the literature base across different to outcome variations across different Additionally, this review was to single case research research of other be to if are limited to different experimental designs. researchers also to at literature to and other for additional