The Monster Study in Iowa (1939) — When Researchers Cross an Ethical Line

Why Did Scientists Want to Change a Child's Perception of Their Own Speech?

In the first half of the twentieth century, stuttering was explained in different, often conflicting ways. Wendell Johnson of the University of Iowa was developing the idea that the responses of other people could play an important part in the emergence of the disorder: a child hesitates in the ordinary course of speaking, and adults begin treating that hesitation as an error. Under this account, being watched, corrected and subjected to concern might lead the child to monitor their own speech increasingly consciously. This was not a neutral description of an established mechanism, but a research hypothesis. [V01; V02]

In 1939, his student Mary Tudor produced a master's thesis examining the effects of evaluative labelling on speech fluency. She posed four questions about how removing, retaining or introducing the labels “you stutter” or “you speak normally” would affect measured disfluency. The central, controversial question was whether a child previously regarded as a normal speaker could be made less fluent by being given a negative label. The thesis did not explicitly state that its aim was to cause permanent clinical stuttering; that distinction matters if the researchers are to be criticised for what they actually planned. Even the final question, however, reveals the boundary between observing a phenomenon and creating conditions that the researcher expects to be unfavourable to the participant. [V01; V01M]

There is nothing inherently objectionable about studying the influence of language and social expectations. The problem lies in the choice of test. The researchers did not simply observe people's reactions to everyday criticism. They designed the negative label as an intervention and directed it at children who depended on adults within an institution. The question of what causes a speech disorder therefore acquired another unavoidable question: who was entitled, for the sake of a hypothesis, to decide what these children would hear about themselves? [V01; V03; V07]

Twenty-Two Children and Four Different Treatments

Tudor helped assess the speech of children living at the Iowa Soldiers' and Sailors' Orphans' Home in Davenport. In 1939, 256 children were initially screened, and 22 participants were selected from that wider group for the main experiment. Staff at the institution had already labelled ten as children who stuttered; the evaluators treated the other twelve as speakers without that label, although ordinary hesitations occurred in their speech too. The distinction matters: a historical label applied by institutional staff is not equivalent to a modern independent clinical diagnosis. [V01M; V02]

The participants were then divided into four groups. Five children already labelled as stuttering were to have that label removed and were reassured that they spoke normally. Another five retained their previous label. Of the twelve children assessed as normal speakers, six received positive evaluations. The remaining six were given the opposite message: their speech supposedly showed a problem to which they needed to pay constant attention. [V01M; V02]

How Were the 22 Children Divided?

“Stuttering” and “normal speech” here refer to the study's historical participant classifications, not contemporary independent clinical diagnoses. The six children in group IIA received negative labelling during research sessions; the additional instructions to institutional staff were reportedly largely ignored. This table describes the experimental design, not established medical outcomes. Source: Ambrose and Yairi (2002), pp. 191–192, reviewing the Tudor study.

That last group, designated IIA in the study, became the focus of the later controversy. Between the beginning of the year and May 1939, these six children attended approximately eight or nine sessions with the researcher, according to the later analysis. Her instructions required them to monitor each interruption in their speech, interpret it as a sign of stuttering and repeat what they had said correctly. Tudor told staff at the home to continue drawing attention to these supposed faults outside the sessions. She also recorded, however, that staff largely ignored those instructions. The negative messages delivered during research sessions are therefore documented; their continuous reinforcement throughout the day is not. [V01M; V02; V02P]

The number 22 in many accounts creates the impression that all 22 children received the same negative treatment. They did not. The six children in group IIA were specifically exposed to an attempt to attach a new negative label to them. The others received different interventions, some supportive and others serving as comparisons. The study's ethical responsibility should therefore be judged not by dramatically adding up presumed victims, but by examining what each group actually experienced and how its members were protected. [V01M; V02]

When a Child Begins to Believe an Adult Who Says Something Is Wrong

During a research session, a child answers a question, searches for a word, pauses and starts the sentence again. This example illustrates the principle of the intervention; it is not a record of any particular participant's words. Such interruptions occur in ordinary speech, including among people who do not stutter. In the Iowa experiment, however, precisely such a moment could become “evidence” of a supposed speech defect because an adult with research authority interpreted it that way. What mattered was not only what the child said, but how others judged the child's speech. [V01; V02]

Tudor wrote in her thesis that, in her observation, all six children in the negatively labelled group spoke less, gave shorter answers, responded more slowly and became more withdrawn. She also described increased self-consciousness and embarrassment when speaking. These were her assessments, not the results of separate, independent behavioural testing. They are historically important research observations, but not equivalent to a subsequently confirmed independent diagnosis of stuttering. Nor are they sufficient to calculate how many individuals experienced lasting psychological harm. [V01M; V02X]

A particular difficulty was that Tudor had not gathered comparable, systematic baseline data on these children's personalities, emotions and everyday behaviour before the intervention, nor had she measured those aspects afterwards using an adequate comparative procedure. It was therefore impossible to reliably distinguish changes arising from the experiment from those associated with existing circumstances or from the encounter with the researcher itself. This methodological gap does not erase the children's experiences. It does, however, prevent subjective research notes from being treated automatically as proof of causation. [V02; V02X]

Even so, the imbalance of power is unmistakable. A child in institutional care did not have the ordinary opportunity to tell an adult: “This is your experiment, and I do not want to take part.” To that child, an adult's evaluation could become an explanation of their own abilities. If the child began speaking less as a result, that matters on a human level, even where the research method cannot reliably establish the precise path leading to that outcome. [V01; V03; V07]

What the Measurements Showed — and What They Did Not

The study included audio recordings of speech samples and assessments by five judges before and after the intervention. Tudor reported individual changes, but these did not consistently move in the expected direction. The original analyses lacked adequate statistical tests, and the categories of speech interruptions did not reliably distinguish ordinary hesitation from features associated with stuttering. Fluency ratings among children in the most controversial group varied: some improved, others worsened, while still others showed no clear change. The author also reported noticeable behavioural responses. [V01M; V02]

In 2002, researchers Nicoline Grinager Ambrose and Ehud Yairi reanalysed the original data. They found that the results did not empirically support any of the four experimental questions. In group IIA, the overall assessment of speech fluency did not deteriorate significantly from the beginning to the end; none of the five independent judges described the speech of these six children as stuttering at the end. The study therefore failed to demonstrate that negative labelling caused new clinical stuttering. But a small sample and inadequate measurements cannot establish that the children experienced no emotional distress. [V02; V02P]

The conclusions must go no further than the evidence allows. It has not been shown that the researchers caused chronic stuttering through this experiment. Nor has it been shown that the intervention caused no distress to anyone. The research data are not sufficiently robust to resolve every question about possible emotional and communicative consequences, and in particular they do not support reliable conclusions about each individual's long-term outcomes. [3]

This distinction does not make the experiment less serious. Instead, it prevents its critics from repeating the very error that made the research methodologically problematic: deriving certainty from a preferred interpretation when the measurements do not establish it. A critique of research ethics is stronger when it does not depend on proving something the original documents cannot prove. [V01; V02; V07]

Was the Case Really Hidden for Six Decades?

The experiment is often described as a secret that remained entirely unknown until 2001. That account is too simple. Mary Tudor submitted her master's thesis to the university in 1939; although the study did not become a widely cited conventional journal article, it did not disappear without a trace. Speech-language pathologist Franklin H. Silverman had already written about the “Monster Study” in the Journal of Fluency Disorders in 1988. [V01; V06]

A different turning point came in 2001, when journalistic reporting brought the case out of a limited professional circle and into public debate, reconnecting the study with its former participants. The difference between “there had previously been little public attention” and “nobody knew the document existed” is not trivial. The first describes public reception; the second creates the false impression that the archival record had been entirely erased. [V06; V08]

Following the reports in 2001, the University of Iowa issued a public apology. Contemporary reporting in the university newspaper, The Daily Iowan, records a formal apology on 13 June 2001 and subsequent letters to former participants. The apology matters, but it does not in itself establish a clinical diagnosis or quantify damages. It means the institution had to confront publicly decisions taken under its academic auspices. [V08; V05]

The subsequent controversy also raised questions about the disclosure of former participants' personal information. Ambrose and Yairi warned that linking names, ages and labels in publicly accessible archival material could expose sensitive personal details. They also cautioned that a journalist's interpretation of scientific results should not replace professional assessment. Neither point makes the disputed experiment disappear or denies participants the right to learn what happened to them. When examining a history involving real children, accuracy is owed equally in accounts of the researchers, the reporters and those whose lives are being reported. [V02; V08]

The Court Allowed the Lawsuit, but Did Not Prove Every Consequence

After the public disclosure, former participants and representatives of deceased participants' estates began pursuing claims for compensation. An important legal development was the Iowa Supreme Court's decision in Nixon v. State on 30 September 2005. The court chiefly considered whether the state could prevent the action by invoking sovereign immunity and the age of the events at issue. [6]

The decision allowed the proceedings to continue. The judgment expressly explained that, when considering a motion to dismiss, the facts had to be viewed in the light most favourable to the claimants, and that the ruling should not be understood as a final determination of the substantive merits of their claims. Overlooking this distinction turns a procedural decision into a supposed scientific finding about what happened to each child. [6]

In 2007, an agreement to settle for a total of $925,000 for six claimants was publicly announced: three former participants and three represented estates. Contemporary reporting noted that the agreement still awaited review by the state's claims board; I have not independently examined the original settlement agreement in this review. The claimants' allegations of long-term consequences were serious and formed part of the legal dispute. Under the publicly reported terms, the state denied legal liability, while the claimants were to withdraw their claims. The settlement was therefore not a judicial determination independently establishing every alleged psychological or medical consequence. [V04; V04B; V03]

We therefore have three distinct layers of history: the research record from 1939, the scholarly review of the evidence in 2002, and the legal proceedings between 2003 and 2007. All concern the same experiment, but they do not answer the same question. The first shows what the researchers did and observed. The second assesses what can be concluded scientifically. The third concerns access to legal remedies and the settlement of claims. It would be misleading to use any one as a substitute for the others. [V01; V02; V03; V04]

Why the Year 1939 Is No Excuse for Failing to Protect Children

The institutional framework of modern research oversight did not yet exist in its later form in 1939. The Belmont Report, an important American statement of ethical principles for research involving human participants, was not published until 1979. It would misrepresent history to judge the 1939 study as though it had been required to complete the exact forms and procedures developed decades later. But the difference in regulations does not settle whether the researchers treated the children with care and fairness. [V07; V09]

The Belmont Report later explicitly set out three principles: respect for persons, beneficence in weighing benefits and harms, and justice in selecting participants. These principles were not legal requirements formulated in 1939. They provide clear language for questions the Iowa case had raised decades earlier: did the child understand the research, who assessed its risks, and why were children living in an institution selected to test negative responses? [11]

Tudor's original account reveals another troubling detail. She gave staff at the home instructions to label the children, but recorded that they largely disregarded those instructions. From a research perspective, this presented a problem of procedural standardisation. From the children's perspective, it raised an even more serious issue: the adults who exercised everyday authority over them had been drawn into a study that was not shown to have been designed around the children's interests and informed consent. [V01M; V02]

A serious assessment does not require the claim that everyone on the research team consciously wanted to cause a permanent disorder. The documents do not reliably support that conclusion. The question is whether, at the planning stage, the researchers should have recognised the danger that a child might come to see as a personal defect something that adults had repeatedly asserted with authority. Responsibility in research is not confined to avoiding malicious intentions; it also includes anticipating foreseeable risks and protecting those who cannot adequately evaluate those risks for themselves. [V02; V07]

A Methodological Failure Can Also Be an Ethical Warning

The Iowa study could be read merely as an unusual historical failure of a hypothesis about the origins of stuttering. But its more important lesson lies in the relationship between testing a theory and protecting people. The researchers intervened in children's perceptions of themselves without gathering sufficiently reliable data to distinguish among the many possible explanations of the consequences. In other words, the risk was concrete, while the scientific benefit was not evidentially assured. [V01; V02]

In 2002, Ambrose and Yairi showed that the original numerical findings supported far more limited interpretations than the popular stories about “artificially created stutterers”. At the same time, they acknowledged the possibility of unpleasant emotional reactions to the research intervention and discussed its historical ethical context. Their position is not a simple certificate of rehabilitation for the experiment; it is a reminder that assessments of scientific effects and moral acceptability must remain separate. [V02; V02X]

Research involving human participants today needs precisely that distinction. Good intentions or an interesting theory do not remove the duty to weigh risks. Equally, a failure to establish statistical proof does not mean that participants' experiences should go unheard. Those experiences can be taken seriously without converting them into diagnoses that no one has reliably established. The worst response would be to replace the researcher's unproven explanation with an equally unproven journalistic certainty. [V02; V07; V08]

Who Gets to Make a Child the Test of a Hypothesis?

The greatest problem with the “Monster Study” is not its name. It is the straightforward description of what was done: six children were told by adults that they supposedly had a speech disorder, even though the researchers had previously classified them as normal speakers. In an institution, that message carried a different weight from a casual remark by another child. It came from people involved in the structures of education, care and scientific authority. [V01M; V03]

An accurate historical critique therefore does not need to prove that the children were made to stutter for life. The researchers set out to discover what would happen if a child was deliberately given a negative label — yet the question of how that label might harm the child was not protected by safeguards as rigorous as the research ambition itself. That remains objectionable even if the final measurements do not support the original hypothesis. [V01; V02; V07]

If this history teaches us anything today, it is not that science should be replaced by moral outrage. It teaches us to demand more of science: reliable evidence, honest acknowledgement of uncertainty, serious treatment of participants and procedures that do not subordinate their rights to the appeal of a hypothesis. The Iowa experiment did not have to be shown to cause chronic stuttering for the protection of children to become a pressing question. That question should have been asked before the first research session in which a child heard that their ordinary speech amounted to a disorder. [V01; V02; V07]