The Illusion of Control: A Critical Look at Confounding Variables in ABA Research

Dr. Scott Frasard is an autistic autism advocate who is a published author and an outspoken critic of operant conditioning approaches to change natural autistic behaviors to meet neuronormative social expectations.

---

Introduction

Applied Behavior Analysis (ABA) emphasizes methodological precision, particularly through its focus on controlling confounding variables to establish internal validity. This emphasis is especially pronounced in single-subject experimental designs, which prioritize clarity in functional relations through tightly managed conditions. Cooper et al.'s (2020) treatment of this topic reflects and reinforces this tradition. Their framework categorizes confounds into four procedural domains: subject, setting, measurement, and independent variable. Each category is accompanied by technical strategies aimed at eliminating or minimizing extraneous influences that might obscure causal interpretations. This structure provides a procedural roadmap for isolating effects and is widely adopted as a pedagogical tool in training new behavior analysts.

Yet, as this article explores, such procedural sophistication can mask a deeper conceptual narrowness that limits the field's interpretive range. The commitment to control often rests on a metaphor that equates scientific rigor with dominance over complexity, as if truth is something to be conquered rather than dialogically engaged. This metaphor does more than shape experimental design; it informs the ethical stance of the researcher, narrows the epistemological lens through which phenomena are understood, and predetermines what is considered legitimate data, relevant behavior, and acceptable forms of evidence. It creates an intellectual environment in which variability is viewed with suspicion and deviation is framed as error, rather than as potentially meaningful expressions of difference.

This critique challenges the field’s assumption that confounding variables can be managed through procedural clarity alone. It argues instead that confounds often emerge from the very paradigms that structure ABA research and shape its methodological instincts. These paradigms include reductive operational definitions that prioritize measurability over meaning, the systematic exclusion of participant assent as a variable of concern, the persistence of an epistemic monoculture that elevates certain forms of knowledge while marginalizing others, and the widespread absence of reflexive engagement with power, identity, and positionality. Each of these elements introduces distortions that cannot be resolved through technical fixes alone. In doing so, this article calls for a reimagining of what it means to produce valid, ethical, and socially relevant knowledge in behavior analysis to one that foregrounds complexity, centers lived experience, and invites methodological pluralism.

The Illusion of Control

The concept of “control” is central to behavioral science, often invoked as the gold standard of scientific credibility. Within ABA, controlling variables is framed as both a methodological imperative and a moral good, often conflating technical discipline with ethical righteousness. This framing functions not merely as a guideline for design but as a marker of identity within the field, delineating behavior analysis from what it regards as less scientific or more speculative disciplines. By emphasizing control as both proof of rigor and proof of legitimacy, the field imbues its methodologies with a sense of authority that resists critical scrutiny. Yet this metaphor of control, while historically foundational, may be deeply misleading, not only methodologically, but also ethically and epistemologically.

Human behavior does not unfold in tidy, isolated sequences. It is influenced by histories, relationships, power dynamics, trauma, and context. These are not peripheral distractions but core constituents of behavior, shaping both the form and function of how individuals act and respond. When researchers aim to control for all such influences within experimental conditions, they often produce environments that are so tightly managed they no longer resemble the settings in which behavior naturally occurs. This artificiality reduces external validity, even as it bolsters internal control. The resulting data may appear stable and predictable, but only because the experimental design has filtered out the very messiness that gives behavior its meaning and significance.

This illusion of control creates a paradox. By attempting to isolate variables to prove functional relations, ABA researchers risk constructing artificial scenarios where findings are internally valid but stripped of ecological depth. The environments in which these findings are produced often lack the relational, cultural, and contextual elements that shape behavior in lived settings. As a result, interventions validated in these artificial contexts may falter when applied in the complexity of real life. Such methodological purism, while it may satisfy procedural standards, becomes a liability when it produces confidence that is not warranted by the limited scope of its inquiry. In these moments, certainty becomes not a mark of truth, but a symptom of conceptual reductionism.

The ethical implications are equally stark. When participants, particularly those who are disabled, racialized, or otherwise marginalized, are treated as systems to be manipulated rather than as people to be engaged with, the research loses any claim to neutrality. It becomes not a dispassionate inquiry, but an enactment of power that uses the language of science to justify domination. In such contexts, methodological control is not simply a technical act, but rather a political and moral stance. It determines whose agency matters, whose behavior is pathologized, and whose experiences are rendered invisible. Control, in this light, functions not as a neutral design choice but as an ethical position that must be examined with the same scrutiny as any other variable in the research process.

A justice-oriented science must ask: Who benefits from the current framing of control, and who is silenced or marginalized by it? The language of control often disguises deeper questions about power, agency, and legitimacy in research. When the emphasis is placed on eliminating variability rather than understanding it, science risks erasing the very humanity it seeks to study. The goal of research should not be to exert dominance over human complexity, but to engage it with intellectual humility, ethical responsibility, and relational depth. Reimagining “control” as responsiveness rather than domination opens the door to methodologies that are not only empirically rigorous, but also socially accountable and inclusive of the voices most often excluded from scientific narratives.

The Narrow Framing of Confounds

Cooper et al. (2020) categorize confounding variables into four types: subject, setting, measurement, and independent variable. This classification offers clarity and procedural utility, particularly for novice researchers learning to identify and manage threats to internal validity. However, it also reveals an underlying reductionist logic in which confounding is treated as a discrete, technical challenge to be neutralized rather than as an emergent property of the complex interplay between variables, researchers, and contexts. This framing assumes that confounds exist independently of the researcher's worldview and design choices, when in fact they often arise precisely from unexamined assumptions, systemic influences, and the artificial constraints imposed by tightly controlled environments.

What is absent in this framework are systemic, interactive, and emergent confounds arising from intersections of social location, historical marginalization, or institutional norms. These are not ancillary factors, but central influences on how behavior is expressed, perceived, and recorded. For example, the behavior of an autistic child in a clinical setting may be influenced not only by the independent variable being tested, but also by prior experiences of coercion, ongoing anxiety about being observed, the racialized or gendered dynamics of the therapy context, or the child’s lack of communicative agency. Such influences are not merely external noise; they shape the conditions under which the behavior is elicited and interpreted. Their exclusion from formal confound categories reflects a broader failure to recognize how research settings are socially and politically constructed.

Confounds do not always present themselves as clearly delineated threats. Sometimes they are embedded in the very structure of the study itself, shaped by the implicit values, assumptions, and exclusions baked into the research design. When research environments assume neutrality but reproduce exclusionary dynamics, such as who is observed, who defines the behavior, and whose voices are absent, they become confounded not by external interference but by their foundational framing. These entanglements are not incidental or avoidable anomalies; they are constitutive of how knowledge is produced, whose experiences are legible, and what truths are allowed to surface. In this way, methodological design is never neutral; it is always situated, and often complicit in the very biases it claims to transcend.

In many applied settings, the idea of controlling for a variable assumes it is possible to isolate it from its broader context. Yet behavior rarely arises in isolation. Subject and setting variables, for instance, often interact in ways that shape how individuals respond to interventions. The influence of setting is not simply environmental; it includes social cues, power dynamics, prior experiences, and even expectations of what the space represents. A sterile room with clinical observers may elicit behavior that differs significantly from what would emerge in a familiar, affirming environment surrounded by trusted individuals. Additionally, the presence of evaluators can invoke performance anxiety, masking or altering the behavior of interest. Such interactions are not adequately addressed in the four-category model, which privileges experimental neatness over contextual realism.

These are not “noise” to be eliminated; they are part of the signal. When ABA research fails to acknowledge these layers, it does not merely overlook potential sources of bias. It upholds a worldview in which unpredictability and nuance are framed as threats rather than as essential elements of understanding behavior. Treating these dimensions as irrelevant or inconvenient undermines the potential for research to reflect lived reality. A more expansive understanding of confounds would treat their presence not as evidence of failure, but as an invitation to expand the boundaries of interpretation, incorporate multiple perspectives, and refine the relevance of the questions being asked.

The Absence of Construct Validity

Another critical omission in Cooper et al.'s (2020) treatment is construct validity: the question of whether the variables being manipulated and measured truly capture the constructs of interest. This dimension of validity addresses whether operationalized behaviors genuinely represent the psychological, emotional, or social concepts they purport to measure. In its commitment to observable behavior, ABA often narrows its focus to what can be counted or graphed, bypassing deeper questions about the relevance or authenticity of these measurements. This omission limits the field’s ability to claim that its findings correspond meaningfully to real-world outcomes or lived experiences.

Consider the operationalization of “on-task behavior” as a student’s eye gaze toward a worksheet. While this metric is convenient for data collection, it offers a superficial indicator that may have little to do with actual engagement, comprehension, or interest. It also risks misinterpreting passive behaviors as signs of attention, when in reality the student may be disengaged, confused, or masking discomfort. Similarly, “problem behavior” is often defined in ways that pathologize actions rooted in communication, sensory processing, or boundary setting, especially for neurodivergent individuals. These definitions typically center the observer’s interpretation rather than the participant’s intent or need. “Compliance” is frequently framed as an ideal outcome, regardless of the ethical conditions under which it is achieved or the potential harms of ignoring context, autonomy, and consent. These examples reflect a broader tendency in ABA to substitute measurability for meaning.

In such cases, the quest for internal validity undermines the very meaning and purpose of the research. If what is being measured does not align with what truly matters in lived experience, then the results, however replicable or graphically compelling, are epistemically hollow and ethically questionable. This disconnect becomes especially problematic in applied contexts, where interventions based on flawed constructs can misguide practice and inadvertently harm those they aim to help. It also reveals a tension between technical accuracy and human relevance. Construct validity is not a luxury or an advanced layer of analysis; it is a foundational requirement for research that seeks not only to describe behavior but to inform interventions that respect, empower, and reflect the dignity of those whose lives are most affected by its conclusions.

When operational definitions are constructed without meaningful engagement from those being studied, especially disabled individuals, the resulting measures often reflect institutional priorities or practitioner convenience rather than the lived realities and values of participants. This disconnect is not a minor oversight; it can fundamentally reshape the focus and interpretation of entire studies. The consequence is not just a conceptual mismatch, but an ethical distortion that compromises both the validity and integrity of the research. Behavior becomes shaped, measured, and judged through external frameworks that participants may neither understand, agree with, nor have had any role in defining, resulting in findings that may ultimately serve systems rather than those most impacted by them.

This problem is exacerbated in settings where constructs are defined solely by their observable correlates, without consideration for the underlying experiences or meanings driving those behaviors. A child who avoids eye contact may be categorized as “noncompliant,” even if the behavior is an expression of overwhelm, an attempt to self-regulate, or a culturally normative form of communication. Such labels not only misrepresent the behavior but also risk imposing interventions that suppress rather than support the individual's needs. The absence of construct validity in these cases becomes a site of both methodological failure, because the data do not reflect what they claim to measure, and epistemic injustice, because the perspectives and realities of the person being studied are systematically erased from the research process.

A more rigorous and ethical approach would treat construct definition as a collaborative process rooted in dialogue, transparency, and respect. Engaging participants, families, and communities in defining what constitutes meaningful change would not only increase construct validity but also embed research in a framework of shared dignity and mutual accountability. It would help ensure that the constructs being studied reflect not just external interpretations but the lived experiences and values of those most affected. This approach shifts the central question from “Can we measure it?” to “Does what we’re measuring truly matter to those we are claiming to represent?”

Epistemic Confounds

Most critiques of confounding variables focus on threats to internal validity. Yet confounding also operates at the epistemic level, shaping what is visible, knowable, and valued within a research tradition. In the context of ABA, which is grounded in radical behaviorism, the preference for observable behavior and measurable outcomes is not simply a methodological choice but instead reflects a broader worldview about what kinds of knowledge are legitimate. This foundational assumption acts as an epistemic filter, excluding forms of understanding that are introspective, relational, narrative, or contextually embedded. These excluded ways of knowing are often the very ones that would center the voices, needs, and experiences of participants who are most impacted by behaviorist intervention. As such, epistemic confounding becomes a silent force that structures the boundaries of what counts as data and whose perspectives are allowed to shape the scientific narrative.

This exclusion has cascading effects. When only that which can be directly observed is considered legitimate, phenomena that are internal, relational, or affective become scientifically invisible. Emotions, intentions, meaning-making, and interpersonal dynamics are either ignored or reframed in terms of observable outputs, stripping them of their depth. This leads to a form of knowledge production that reinforces its own narrow assumptions, mistaking absence of data for absence of importance. Over time, entire domains of human experience are marginalized in favor of what can be coded and counted, narrowing the field's understanding of behavior and diminishing its relevance to lived realities.

Narrative, relational, and phenomenological approaches are not merely alternative methods. They represent fundamentally different epistemologies that prioritize meaning-making, context, and subjectivity as legitimate dimensions of human experience. These methods allow researchers to explore questions that cannot be answered by frequency counts or stimulus-response chains—questions about identity, agency, emotion, and belonging. Their exclusion from ABA's research paradigm is not a neutral omission. It constitutes a form of epistemic violence, where entire domains of lived experience are rendered invisible. This exclusion ensures that certain questions will never be asked, certain truths never considered, and certain experiences never legitimized within the scientific record. The result is a narrowed vision of science that fails to account for the fullness of human complexity.

This narrowing becomes especially dangerous when it informs policy or clinical intervention. When epistemic limitations become embedded in treatment protocols or behavioral goals, they shape how distress, autonomy, and communication are interpreted. If a child’s distress is only legible through the lens of behavior, rather than meaning, then interventions will target surface behavior while neglecting underlying emotional, sensory, or relational needs. This reductionist lens can lead to interventions that suppress rather than support, conform rather than empower. The epistemic frame of ABA research thus not only restricts the scope of valid knowledge but also distorts the moral logic of care by substituting compliance for wellbeing and procedure for presence.

To treat epistemic assumptions as outside the realm of confounding is to ignore the most profound ways in which research can be biased. Confounding does not only occur when variables covary. It also occurs when entire frameworks of meaning, knowledge, or identity are excluded from the research process. When only certain ways of knowing are allowed to shape the design, measurement, and interpretation of data, then the research becomes confounded by its own epistemic limits. Pluralism in epistemology, which includes narrative, embodied, cultural, and relational knowledge systems, is not a threat to scientific rigor. It is its deepening, its expansion, and its ethical repair, a necessary step toward a more inclusive and truthful science.

Researcher Positionality and Reflexivity

Perhaps the most unacknowledged confound in ABA research is the researcher themselves. The notion that experimenters are neutral instruments, capable of executing protocols without bias, is a persistent myth that continues to shape methodological training and practice in the field. This belief not only fails to recognize the subjectivity inherent in all human inquiry but also obscures the ways in which power, belief, identity, and positionality influence every stage of research. From the formulation of questions to the operationalization of constructs, from the choice of methodology to the interpretation of data, the researcher's worldview is not just present but profoundly influential. To ignore this influence is to allow unchecked bias to masquerade as objectivity.

Unlike research traditions that require reflexivity, such as feminist, critical, or ethnographic methodologies, ABA rarely calls on researchers to examine their own assumptions, values, or social positioning. The field has historically emphasized procedural objectivity over personal accountability, reinforcing the notion that a well-designed study can stand apart from the worldview of the person conducting it. Yet who conducts the research, how they define behavior, what populations they study, and what outcomes they privilege are all deeply shaped by the researcher's positional lens. Ignoring this influence does not eliminate bias; it embeds it silently into the research process, allowing power and perspective to operate invisibly under the banner of neutrality.

This lack of reflexivity can have far-reaching implications. For example, a behavior analyst operating from a highly compliance-oriented framework may inadvertently interpret resistance as dysfunction rather than as a form of agency, protest, or communicative intent. Acts of refusal or withdrawal may be mischaracterized as treatment failures instead of being recognized as data about unmet needs or misaligned goals. Without critical self-awareness, the researcher’s own values, biases, and lived experiences become invisible variables, shaping the study’s framing, execution, and interpretation without ever being named or examined. This undermines both the ethical integrity and epistemic reliability of the research.

Integrating reflexive practice into ABA would not diminish its scientific value; it would enhance its credibility, transparency, and ethical robustness. By acknowledging their own positionality, researchers make visible the values, assumptions, and interpretive lenses that inevitably shape the research process. Transparency about the researcher’s role and perspective allows for a more honest accounting of the variables influencing the research and invites a richer, more contextually grounded interpretation of data. Reflexivity transforms the researcher from a presumed neutral observer into an accountable participant in the co-construction of knowledge, opening pathways for more inclusive and justice-oriented scientific inquiry.

To move toward more ethically sound and methodologically transparent research, ABA must cultivate a culture of self-interrogation that is active, sustained, and institutionally supported. This involves more than individual introspection; it calls for structured practices that examine how one’s training, identity, institutional affiliations, and cultural norms shape the framing, execution, and interpretation of research. Such inquiry should be built into methodological design, peer review, and training curricula to ensure that reflexivity becomes a norm rather than an exception. In doing so, ABA can begin to close the gap between procedural rigor and epistemic justice, aligning its scientific practices more closely with the values of respect, accountability, and relational integrity.

Assent and Coercion as Methodological Variables

Finally, ethical violations, especially those concerning autonomy and consent, must be recognized not only as moral failings but as sources of methodological distortion. When participants are unable to provide informed, voluntary assent, the conditions under which data are collected are altered in ways that compromise validity. Coercion, fear, or histories of learned compliance can all result in behaviors that conform outwardly to experimental expectations while masking inner distress, resistance, or disengagement. These effects are not anomalies. They are integral to understanding how power operates within research settings. Without accounting for them, data may appear precise but are, in fact, artifacts of compromised agency.

ABA’s traditional focus on compliance as an outcome masks this distortion and reinforces a framework that privileges control over collaboration. Behavior performed under coercion is not equivalent to behavior performed with agency or understanding. When compliance is achieved through control rather than consent, the apparent success of an intervention may reflect submission, fear, or habituated passivity rather than genuine learning or transformation. Such data, while appearing orderly and consistent, is ethically suspect and scientifically weak because it fails to capture the subjective experience and voluntary engagement of the participant.

The exclusion of assent as a variable of interest reveals a larger cultural assumption within ABA: that behavioral outcomes matter more than the relational and contextual conditions under which they occur. This mindset reduces ethical engagement to procedural minimalism, prioritizing technical compliance over genuine consent. In doing so, it fails to interrogate how coercion, institutional authority, or the absence of communicative reciprocity can deeply undermine both the dignity of participants, and the integrity of the data collected. Research conducted without attention to assent is not ethically neutral. It is shaped by the unexamined power dynamics that define what is considered valid participation and whose agency is acknowledged.

If researchers do not ask whether participants freely chose to engage in an intervention, they risk generating conclusions that reflect institutional goals rather than participant wellbeing. This oversight can result in research findings that affirm programmatic efficiency or practitioner success while failing to capture the lived experiences or desires of those receiving the intervention. The issue becomes particularly urgent in research involving disabled or non-speaking individuals, whose consent is often presumed, extracted under duress, or bypassed entirely through proxy decision-making. In such cases, what is interpreted as assent may actually reflect habituation, fear of punishment, or lack of alternatives rather than genuine agreement. Without rigorous attention to these ethical dynamics, the validity of the entire research endeavor is called into question.

[W]hat is interpreted as assent may actually reflect habituation, fear of punishment, or lack of alternatives rather than genuine agreement

Participatory design, assent protocols, and relational ethics should be seen as safeguards not just for participants, but for the integrity and credibility of the science itself. These approaches foster a research environment in which behaviors are interpreted within the context of agency, meaning, and consent, rather than reduced to mechanistic outputs. They shift the emphasis from control over participants to collaboration with them, emphasizing that data derived from authentic, voluntary participation are more valid, generalizable, and ethically sound. In this way, ethical research design and methodological validity become not just compatible but mutually reinforcing, strengthening both the moral and scientific foundations of behavioral inquiry.

If ABA research is to be both rigorous and just, it must reconceptualize ethics as foundational to methodology rather than as a parallel or secondary concern. This means embedding ethical considerations into the very architecture of research design, from question formulation to measurement and analysis. Moving beyond compliance as a goal, the field must embrace co-authored, consent-driven processes that respect the agency, dignity, and insight of those whose lives the research seeks to impact. Such a shift would not only strengthen the scientific validity of findings but also align the research more closely with principles of justice, mutual accountability, and human relevance.

Conclusion

Controlling for confounding variables is a cornerstone of experimental design, serving as a central tenet in efforts to establish causal relationships. However, the concept presented in Cooper et al. (2020) reflects a narrowly technical understanding of confounds, that prioritizes procedural clarity while overlooking the deeper conceptual, ethical, and relational dimensions of research. This article has argued that true research integrity demands more than internal validity. It requires a broader methodological pluralism, a grounding in ethical principles that center participant agency, and an epistemic humility that acknowledges the limitations of any single worldview. Without these, the control of confounds risks becoming a performance of scientific rigor rather than a meaningful safeguard against bias.

The metaphor of control, though foundational in behavioral science, distorts both the aims and ethics of research when it overshadows complexity, silences participants, or privileges replicability over meaning. In privileging a model of science that seeks to dominate rather than engage with complexity, ABA risks reducing its methodologies to exercises in containment rather than exploration. ABA’s dominant approaches often treat confounds as procedural problems to eliminate, applying technical fixes to what are often deeply rooted conceptual, social, and ethical entanglements. This reductive framing constrains the field’s ability to grapple with the messiness of real-world behavior and limits the relevance of its findings to the lived experiences of those most affected.

To move forward, ABA must broaden its methodological imagination beyond procedural control and toward a more inclusive, reflective scientific paradigm. This involves embracing construct validity not as a technical afterthought but as a central concern that shapes how behavior is defined and measured. It also requires acknowledging epistemic bias as a structural force that influences what kinds of knowledge are seen as credible, and examining researcher positionality to uncover how identity and power shape the framing and interpretation of research. Finally, it demands a recognition of how power and coercion distort data collection, interpretation, and application, especially in contexts involving marginalized or vulnerable populations. Only by integrating these dimensions can the field produce research that is not only procedurally sound but also ethically and socially accountable.

Summary of Issues with ABA's Approach to Confounding Variables:

• Overreliance on internal validity at the expense of construct, ecological, and social validity

• The illusion of control that strips away complexity and masks ethical compromise

• A narrow typology of confounds that ignores systemic, intersectional, and emergent variables

• Operational definitions that distort or oversimplify meaningful constructs

• Exclusion of epistemic diversity, reinforcing behaviorism’s limitations as the sole lens

• Lack of reflexivity, treating researchers as neutral rather than as co-constructors of knowledge

• Absence of assent protocols, ignoring coercion and power dynamics as confounding factors

In reframing control not as domination but as dialogic responsiveness, ABA can evolve toward a more human-centered, ethically grounded science that values complexity and shared meaning-making over rigid proceduralism. This shift entails not only reconsidering how variables are managed but also reimagining the relationship between researcher and participant as one of mutual respect and co-authorship. This transformation is not merely a methodological enhancement. It is a moral imperative grounded in the principles of justice, inclusivity, and the recognition of each participant's full humanity.

Reference

Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.). Pearson.

---

Thank you for taking the time to read this post. If you enjoyed it, please do click LIKE and click SHARE to share it with your network. Be sure to check out my book, "A Reflective Question to Ponder: 1,200+ Questions on Autism to Foster Dialogue" available in paperback and eBook. My newest book, "Autism Advocacy Unleashed: A Socratic Journey to Social Justice" is also available in paperback and eBook.

Thank you!