Bilingualism as a Purported Risk Factor
for Stuttering: A Close Look at a Seminal
Study (Travis et al., 1937)
Does bilingualism elevate the risk of stuttering?
Whenever this question is discussed in the litera-
ture, one particular study (Travis et al., 1937) is
almost invariably cited. Is this citation pattern warranted?
Quite a few studies, discussed below, have pointed out short-
comings in Travis et al.’s methods. Yet, the citation pattern
has persisted. Travis et al. reported a stuttering prevalence of
1.8% for monolingual school children and 2.8% for bilingual
ones, a difference they found to be statistically significant.
Many sources citing Travis et al. quote these figures. Some
(e.g., Szmalec, 2013, among many others) cite them as evi-
dence for a causal link between bilingualism and stuttering.
Others note, neutrally, that these figures have been consid-
ered by other authors to constitute evidence for such a link
(e.g., Choo & Smith, 2020; Proctor et al., 2008; Schneider,
2015; Yairi & Ambrose, 2013) and/or point out methodo-
logical flaws in the study (e.g., Bloodstein & Bernstein
Ratner, 2008; Van Borsel, 2011; Van Borsel et al., 2001).
No matter what position one takes in this matter, the
reported clear difference in stuttering prevalence among
monolingual versus bilingual children is striking and note-
worthy, particularly given its practical implications for
However, a difference should only be noteworthy if
it is supported by the data. Here, we ask whether it was in
fact the case that 1.8% of the monolingual group, versus 2.8%
of the bilingual group, in Travis et al. (1937) stuttered.
then ask to what extent small discrepancies in the data affect
the statistical significance of the results. To preview our
results: We find that the raw counts and prevalence rates
reported in Travis et al. are internally inconsistent. Among
other problems, it appears that the 1.8% versus 2.8% dif-
ference is due to several stuttering children being excluded
from the prevalence estimate for the monolingual group.
In addition, we find that the statistical significance of the
results rests on the classification of very small numbers of
participants—far smaller than the uncertainties in the reported
Travis et al. (1937): Methods and Criticisms
Travis et al. (1937) surveyed nine schools in East
Chicago, Indiana, primarily focusing on stuttering, but
also including “foreign accent,”“letter substitution,”“oral
inactivity,”and “aphasia.”The children ranged in age from
4 to 17 years. The reported average age was 8.54 years, and
the reported sample size was 4,827. Each child was given a
questionnaire asking what languages were spoken in the
child’s home and what proportion of the child’s“home
speech”and “school speech”was in English. The question-
naire did not include questions about the length of time
children had been in a multilingual environment. Travis
et al. do not report the distribution of ages across grades
but note that the grades spanned kindergarten through
sixth grade. They further state that the sample included
As noted by others (e.g., Byrd et al., 2016), whether these children
actually stuttered is not clear, particularly as Travis et al. (1937) did
not specify the diagnostic criteria they used. Several facts about the
reported numbers suggest that the stuttering diagnosis was likely
unwarranted in many cases. Nevertheless, given that Travis et al.
considered these children to be diagnosed with stuttering, they will be
referred to as such in this article.
Department of Linguistics, University of California at Berkeley
Correspondence to Susanne Gahl: firstname.lastname@example.org
Editor-in-Chief: Bharath Chandrasekaran
Editor: Julie D. Anderson
Received June 26, 2020
Revision received August 3, 2020
Accepted August 11, 2020
Disclosure: The author has declared that no competing interests existed at the time
Journal of Speech, Language, and Hearing Research •1–5•Copyright © 2020 The Author
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
a group (the “Opportunity group”) they describe as a “group
for mentally retarded children.”Age 17 is unusually old for
6th grade. The practice of placing English Language
Learners in lower grades is one possible reason for this.
Another possibility is that the oldest children surveyed were
in the “Opportunity group.”
All children were asked to read “[g]raded readable
selections containing the most frequently used sounds”
(Travis et al., 1937, p. 187) and were interviewed by the
third author of the study, Jayne Shover. Children who did
not speak English were asked to read passages translated
into their native languages and were interviewed by inter-
preters. Travis et al. themselves pointed out that low lan-
guage proficiency would result in disfluent speech in just
about anyone and that differences in socioeconomic status
could be responsible for differences between the groups.
A number of authors have pointed out flaws in Travis
et al. (1937)’s methods. For example, the criteria for stutter-
ing and bilingualism were unstated (Byrd et al., 2016); the
diagnosis was based on a single assessment (Shenker, 2004;
Van Borsel, 2011; Van Borsel et al., 2001); and, for languages
other than English, the diagnosis was not made by a trained
clinician (Byrd et al., 2016; Van Borsel, 2011). It has also
been pointed out that some of the prevalence estimates would
seem to undermine the claimed causal connection between
bilingualism and stuttering: Bloodstein and Bernstein Ratner
(2008) credit Kathryn Kohnert for pointing out that the
prevalence for trilinguals was lower (2.4%) than for bilin-
guals (2.8%) and higher (7.4%) for children who were mono-
lingual in a language other than English. Along similar lines,
Schneider (2015) points out that the prevalence for the Afri-
can American children (given as 3.76%) exceeds that of the
bilingual group; one might take this as an additional indica-
tion that the diagnoses were inaccurate, in light of the fact
that there is no evidence to suggest that the prevalence of
stuttering is higher in African American children (Proctor
et al., 2008). One might add that Shover interviewed around
145 children a day (assuming a five-day school week over
six and a half weeks, the stated duration of the study), a
pace that precludes careful procedures. In sum, the accu-
racy of the stuttering diagnosis is questionable at best.
Travis et al. (1937)’s Results
Figure 1 depicts Travis et al. (1937)’s table showing
the proportion of children who stuttered in each subgroup,
along with annotations by the present author. The oft-cited
prevalence rates of 1.8% for monolingual English-speakers
(“English Speaking Only”) and 2.8% for children who spoke
English and one other language (“Bilingual”)appearinthe
two leftmost cells in the final row of the table. The annota-
tions in Figure 1 illustrate the method used here for recon-
structing the absolute counts of children in each subgroup,
by combining information from the main text with the per-
centages in the table. For a complete guide to the calcula-
tions, the reader is referred to Items 1–7 in Section 3 of
Supplemental Material S1.
Some of the counts can be reconstructed easily. For
example, Travis et al. (1937) state that there were 10 boys
who did not speak any English and that the stuttering preva-
lence in that group was 10% (cf. Items a and b in Figure 1).
It follows that one boy in the group stuttered. One may find
percentages for such small samples misleading, but the
calculation is unassailably correct. The prevalence of 2.80%
among the bilingual children (cf. Items c and d in Figure 1
and Item 1 in Supplemental Material S1) is also consistent
with other information in the text: Travis et al. mention
that there were 2,322 bilingual children. Indeed, 2.80%
of 2,322 is 65.016, and 65 out of 2,322 is 2.799, which
rounds to 2.80%, matching the reported results.
Upon closer inspection, inconsistencies emerge. Here,
we primarily focus on the monolingual English-speaking
and bilingual groups. Notes on additional contradictions
can be found in Supplemental Material S1.
First, there is a discrepancy between the total re-
ported sample size (4,827) and the sum of the subgroups,
which is 4,837 (cf. Item 4 in Supplemental Material S1). A
clerical error might be a reasonable explanation, as 4,827
and 4,837 differ in a single numerical digit.
A second discrepancy reveals a major problem, how-
ever. It concerns the children classified as stuttering: The
reported counts of English-Only (43), Bilingual (65), Trilin-
gual (2), and Foreign-Only (2) sum to 112 (cf. Item 5 in
Supplemental Material S1). But Travis et al. (1937) state
that a total of 126 children stuttered. The figure 126 has
additional support from the table, in that it matches the
reported overall prevalence of 2.61% (assuming a sample
size of 4,827). Short of assuming a complex set of typos
in the table and in the text, the only way to reconcile
the information is to conclude that 14 of the 126 children
who stuttered were not included in any of the reported
Were these 14 children monolingual or bilingual? For
a complete step-by-step guide to answering this question,
the reader is referred to Items 6 and 7 in Supplemental
Material S1. Here, we summarize the argument briefly:
The key to the discrepancy can be found in the numbers
of boys versus girls who participated in the study. From
Travis et al.’s (1937) text, it is possible to reconstruct the
number of boys (1,213) and girls (1,205), yielding a total
of 2,418 monolingual English-speaking children—19 more
than the reported total of 2,399 for that group (cf. Item 7 in
Supplemental Material S1). This suggests that 19 children
were excluded from the count of the monolingual English
group: 14 who stuttered (as per Item 5 in Supplemental
Material S1) and five who did not. If this is correct, then
these 19 children were included in the sex-specific prevalence
rates and in the count of children who stuttered, but excluded
from the calculation of the stuttering prevalence among
the English-only group.
There is a third discrepancy, described in Item 3 of
Supplemental Material S1. Briefly, there is reason to believe
that the stuttering prevalence for the trilingual children was
determined after excluding three fluent children, increasing
the proportion of children who stuttered in that group.
2Journal of Speech, Language, and Hearing Research •1–5
Travis et al. (1937) do not mention excluding any
participants. It is conceivable that the 19 excluded cases
were in the above-mentioned “Opportunity group.”We
cannot be sure of this, as the number of children in that
group, their language backgrounds, and fluency status are
In any case, it appears that the prevalence for the
monolingual group was determined after excluding 14 stut-
tering and 5 fluent children from the monolingual English-
speaking group, lowering the proportion of children who
stuttered in that group. There is also reason to think that
the proportion of children who stuttered in the trilingual
group was increased by excluding several fluent cases be-
fore calculating the prevalence. Excluding monolingual
stuttering children and trilingual fluent ones shifts the prev-
alence estimates so as to increase the impression of a causal
link between multilingualism and stuttering.
The Robustness of Travis et al. (1937)’s
Excluding 19 (or 24, if we consider the trilingual group)
out of almost 5,000 children may have seemed innocuous to
Travis et al. (1937). Would the findings have been statistically
significant if the 19 missing cases had been included in the
monolingual English group? The answer is no. When the
19 excluded cases are included, the difference (which is now
between 2.36%, i.e., 57 out of 2,418; vs. 2.80%, i.e., 65 out
of 2,322) does not reach significance by a Chi-square test
without Yates’correction, the test likely used by Travis and
(1) = 0.649, p= .42.
We next ask whether the statistical significance of
the results was robust to small changes in stuttering diag-
nosis or language background. To answer this question, we
simulated the process of re-assigning one child at a time
from one fluency group to the other, tracking the associ-
ated pvalues. The full results are shown in Table S1 in
Supplemental Material S1. It will be observed that moving
a single child in the total group of 2,399 monolingual chil-
dren from “fluent”to “stuttering”shifts the pvalue from
.021 to .027. Reassigning two children within the monolin-
gual group and two within the bilingual group suffices to
push the pvalue past the .05 level of significance. Thus,
changes in group sizes far smaller than the discrepancies
we noted in the previous section affect the statistical signifi-
cance of the results.
Travis et al. (1937)’s Influence
Given the age of Travis et al. (1937)’s study and its
many acknowledged flaws, one might expect its influence
to be limited at this point. That is not the case. Literature
reviews have repeatedly pointed out the scarcity of studies
comparing stuttering prevalence among monolingual ver-
sus bilingual children (or adults) and the inconclusiveness
of the extant literature. Pooling all combinations of lan-
guages together, reviews by Van Borsel (2011), Choo and
Smith (2020), and Werle et al. (2019) mention a total of four
group studies since 1937 reporting higher stuttering preva-
lence in bilingual or bidialectal versus monolingual children
(Firozjaei, 2013; Howell et al., 2009; Ralston, 1981; Stern,
1948), along with five others reporting no significant effect
of bilingualism per se (Aron, 1962; Au-Yeung et al., 2000;
McLeod et al., 2014; Mohamadi et al., 2008; Ralston, 1976).
Some of these studies have elicited disagreement (see, e.g.,
Packman et al., 2009’s response to Howell et al., 2009).
The low number of relevant reports and the seemingly com-
pelling results in Travis et al. (1937) may partly explain why
the study has been cited so frequently and why it continues
to be cited.
Figure 1. Travis et al. (1937)’s results, annotated by the present author. The table as originally published is depicted in the broken line rectangle.
Travis et al. (1937) table is Copyright © 1937 American Speech-Language-Hearing Association.
Gahl: Contradictory Data in Travis et al. (1937) 3
Another reason for the position of the study in the
literature may be that it meshes well with some proposals
about stuttering etiology. A full discussion of the links be-
tween Travis et al. (1937) and subsequent theories falls out-
side the scope of this commentary. However, as pointed out
in Shenker (2011), the clinical recommendation (no longer
made by ASHA but still very much alive, if anecdotal re-
ports are any indication) to avoid or delay exposure to a
second language in children considered at risk of stutter-
ing is rooted in the idea that multilingualism might cause
excessive demands on the developing system. As Shenker
(2011) also points out, excessive demands on the linguistic
system have in turn been held to cause disfluencies under
the Demands and Capacities Model (see, e.g., Adams, 1990),
a model that has been influential in clinical practice (see,
e.g., Bernstein Ratner, 2000; Packman et al., 2004; and
Shenker, 2004, for discussion).
The Travis et al. (1937) study has had an enormous
influence on the literature about the intersection of bilin-
gualism and stuttering. Gauging its impact on clinical prac-
tice is difficult. What is clear is that the study plays into a
still-prevalent notion that has informed clinical and educa-
tional practice and policy that bilingualism might hinder
children’s development in some way. Travis et al. (p. 189)
comment that the difference between 1.8% and 2.8% “is re-
ally quite small”and that “[i]t is not to be overlooked that
97.20 per cent of the bilinguals do not stutter.”But parents
and clinicians may not actually consider the difference to
be small: 1.8% represents a substantial decrease (of 36%!)
over 2.8%. Parents may be eager to eliminate any avoidable
risk factor, despite the negative consequences of heritage
language loss and the positive consequences of strong heri-
tage language proficiency (see, e.g., Lee & Suarez, 2009;
Oh & Fuligni, 2010; Wong Fillmore, 2000). It is therefore
important to realize that the purported effect of bilingual-
ism on stuttering in the Travis et al. study was not just
small, but apparently unsupported by the data in the first
I thank Yusuf Ali for drawing my attention to Travis et al.
(1937) in the course of working on his UC Berkeley Honors Thesis,
Fillmore for her encouraging comments on an earlier version of
Adams, M. R. (1990). The demands and capacities model I: Theoreti-
cal elaborations. Journal of Fluency Disorders, 15(3), 135–141.
Aron, M. L. (1962). The nature and incidence of stuttering among
a Bantu group of school-going children. Journal of Speech and
Hearing Disorders, 27(2), 116–128. https://doi.org/10.1044/jshd.
Au-Yeung, J., Howell, P., Davis, S., Charles, N., & Sackin, S.
(2000). UCL survey of bilingualism and stuttering. Journal of
Fluency Disorders, 25(3), 246. https://doi.org/10.1016/S0094-
Bernstein Ratner, N. (2000). Performance or capacity, the model
still requires definitions and boundaries it doesn’t have. Journal
of Fluency Disorders, 25(4), 337–346. https://doi.org/10.1016/
Bloodstein, O., & Bernstein Ratner, N. (2008). A handbook of stut-
tering. Thomson Delmar Learning.
Byrd, C. T., Haque, A. N., & Johnson, K. (2016). Speech-language
pathologists’perception of bilingualism as a risk factor for
stuttering. Journal of Communication Disorders, Deaf Studies
& Hearing Aids, 4(2), Article 158. https://doi.org/10.4172/2375-
Choo, A. L., & Smith, S. A. (2020). Bilingual children who stutter:
Convergence, gaps and directions for research. Journal of Flu-
ency Disorders, 63, Article 105741. https://doi.org/10.1016/
Firozjaei, A. K. (2013). Prevalence of stuttering in bilingual and
monolingual primary schools. International Research Journal
of Applied and Basic Sciences, 4(6), 1328–1331.
Howell, P., Davis, S., & Williams, R. (2009). The effects of bilin-
gualism on stuttering during late childhood. Archives of Disease in
Childhood, 94(1), 42–46. https://doi.org/10.1136/adc.2007.134114
Lee, J. S., & Suarez, D. (2009). A synthesis of the roles of heritage
languages in the lives of children of immigrants: What educators
need to know. In T. G. Wiley, J. S. Lee, & R. W. Rumberger
(Eds.), The education of language minority immigrants in the
United States (pp. 136–171). Multilingual Matters.
McLeod, S., Verdon, S., & Kneebone, L. B. (2014). Celebrating young
Indigenous Australian children’s speech and language compe-
tence. Early Childhood Research Quarterly, 29(2), 118–131.
Mohamadi, H., Nilipour, R., & Yadegari, F. (2008). Stuttering prev-
alence among Kurdish-Farsi students: Effects of the two lan-
guages similarities. Iranian Rehabilitation Journal, 6(1), 83–88.
Oh, J. S., & Fuligni, A. J. (2010). The role of heritage language
development in the ethnic identity and family relationships of
adolescents from immigrant backgrounds. Social Develop-
ment, 19(1), 202–220. https://doi.org/10.1111/j.1467-9507.
Packman, A., Onslow, M., & Attanasio, J. (2004). The demands
and capacities model: Implications for evidence-based practice
in the treatment of early stuttering. In A. Bothe (Ed.), Evidence-
based treatment of stuttering (pp. 65–79). Erlbaum.
Packman, A., Onslow, M., Reilly, S., Attanasio, J., & Shenker, R.
(2009). Stuttering and bilingualism. Archives of Disease in Child-
hood, 94(3), 248.
Proctor, A., Yairi, E., Duff, M. C., & Zhang, J. (2008). Prevalence
of stuttering in African American preschoolers. Journal of
Speech, Language, and Hearing Research, 51(6), 1465–1479.
Ralston, L. D. (1976). Stammering: A possible result of language
interference in a bi-dialectal community, Nevis, West Indies.
Paper presented to the Society for Caribbean Linguistics.
Ralston, L. D. (1981). Stammering: A stress index in Caribbean
classrooms. Journal of Fluency Disorders, 6(2), 119–133. https://
4Journal of Speech, Language, and Hearing Research •1–5
Schneider, S. (2015). Bilingualer Erstspracherwerb: Zweisprachig
von Anfang an. UTB.
Shenker, R. C. (2004). Bilingualism in early stuttering: Empirical
issues and clinical implications. In A. Bothe (Ed.), Evidence-
based treatment of stuttering (pp. 81–96). Erlbaum.
Shenker, R. C. (2011). Multilingual children who stutter: Clinical
issues. Journal of Fluency Disorders, 36(3), 186–193. https://
Stern, E. (1948). A preliminary study of bilingualism and stutter-
ing in four Johannesburg schools. South African Journal of
Communication Disorders, 1(1), 15–24.
Szmalec, A. (2013). Bilinguisme et bégaiement chez l’enfant. Enfance,
Travis, L. E., Johnson, W., & Shover, J. (1937). The relation of
bilingualism to stuttering: A survey of the East Chicago, Indiana,
schools. Journal of Speech Disorders, 2(3), 185–189. https://doi.
Van Borsel, J. (2011). Review of research on the relationship be-
tween bilingualism and stuttering. In P. Howell & J. Van Borsel
(Eds.), Multilingual aspects of fluency disorders (pp. 247–270).
Van Borsel, J., Maes, E., & Foulon, S. (2001). Stuttering and bilin-
gualism: A review. Journal of Fluency Disorders, 26(3), 179–205.
Werle, D. R., Byrd, C., & Coalson, G. (2019). Description of mul-
tilingual participants who stutter: An update 2011–2018. Com-
munication Disorders Quarterly. Advance online publication.
Wong Fillmore, L. (2000). Loss of family languages: Should edu-
cators be concerned? Theory Into Practice, 39(4), 203–210. https://
Yairi, E., & Ambrose, N. (2013). Epidemiology of stuttering: 21st
century advances. Journal of Fluency Disorders, 38(2), 66–87.
Gahl: Contradictory Data in Travis et al. (1937) 5