Articulation rate, pauses, and disfluencies in professional fighters: potential speech biomarkers for repetitive head injury

Abstract

Objective:

As part of a larger study dedicated to identifying speech and language biomarkers of neurological decline associated with repetitive head injury (RHI) in professional boxers and mixed martial artists (MMAs), we examined articulation rate, pausing, and disfluency in passages read aloud by participants in the Professional Athletes Brain Health Study.

Setting:

A large outpatient medical center specializing in neurological care.

Participants, Design, and Main Measures:

Passages read aloud by 60 boxers, 40 MMAs, and 55 controls were acoustically analyzed to determine articulation rate (the number of syllables produced per second), number and duration of pauses, and number and duration of disfluencies in this observational study.

Results:

Both boxers and MMAs differed from controls in articulation rate, producing syllables at a slower rate than controls by nearly half a syllable per second on average. Boxers produced significantly more pauses and disfluencies in passages read aloud than MMAs and controls.

Conclusions:

Slower articulation rate in both boxers and MMA fighters compared to individuals with no history of RHI and the increased occurrence of pauses and disfluencies in the speech of boxers suggest changes in speech motor behavior that may relate to RHI. These speech characteristics can be measured in everyday speaking conditions and by automatic recognition systems, so they have the potential to serve as effective, non-invasive clinical indicators for RHI-associated neurologic decline.

Introduction

Professional fighters exposed to repetitive head injury (RHI) are at risk for degenerative neurologic decline such as chronic traumatic encephalopathy (CTE). Definitive diagnosis of CTE can be made only at autopsy,¹ but research criteria for the clinical manifestation of CTE, traumatic encephalopathy syndrome (TES), has recently been introduced.² Although speech impairment is not a core feature of TES, motor signs (e.g., Parkinsonism, dysarthria, ataxia, etc.) may support the TES diagnosis. As part of a larger study dedicated to identifying speech and language indicators of TES in professional boxers and mixed martial artists (MMAs), the present study examines speech motor behaviors in the form of articulation rate, pausing, and disfluency in passages read aloud by participants in the Professional Athletes Brain Health Study (PABHS), an observational, longitudinal study of athletes exposed to RHI, including boxers and mixed martial artists, that seeks biomarkers of associated decline in cognitive and neurological functions.³

Boxing has long been associated with dysarthria, a group of neurogenic speech disorders associated with motor disturbances,^4–7 but few detailed reports of speech deficits exist. Berisha et al.⁸ examined recordings of speech produced over many years by boxer Muhammad Ali, who was diagnosed with Parkinson’s disease in 1984. They found, for example, that Ali’s speech rate declined by 26% from 1968 to 1981, from over four syllables per second to three syllables per second. In another account of impaired speech in boxers, McMicken et al. (2011) described an amateur boxer with ataxia who exhibited slow and slurred “inebriated” speech.⁹ Because speech production is a complex motor behavior requiring precisely timed, rapid, and accurate movements coordinated by several neurologic systems,⁹ speech changes may serve as the earliest signs of neurologic disease.¹⁰ Two recent reviews discuss the utility of speech biomarkers such as speech rate, pause time, and pause duration for detection of neurologic disturbance including Parkinson’s disease, amyotrophic lateral sclerosis, and RHI.^11–12 For example, Tauro et al. reported that acoustic machine learning techniques differentiated speakers with and without a history of RHI with 85% accuracy.¹³ In this study, we compared fighters exposed to RHI with controls who had no history of RHI exposure on several characteristics of connected speech that have been associated with various types of dysarthria and that can be reliably measured in speech samples collected in clinical conditions. Understanding speech changes associated with RHI will help develop behavioral indicators for the early identification of RHI-related neurologic decline to improve patient outcomes and to track disease progression.

One speech behavior that can be reliably measured is speech rate, the rapidity with which articulatory speech gestures are produced. Slowed speech rate reflects neurologic disturbance even when speech intelligibility is not noticeably reduced^14,15 and has been associated with a variety of neuropathologies, including multiple sclerosis, Parkinson’s disease, stroke, and traumatic brain injury (TBI).^16–18 In this study, we assessed articulation rate, the number of syllables produced per second during a passage reading task. Because calculation of articulation rate does not include pauses or disfluencies, it is considered to be primarily a measure of speech motor function.¹⁹ Healthy adults produce a little more than five syllables per second while reading aloud.²⁰

We also examined the frequency and duration of silent pauses in a passage read aloud. Pauses, or temporary breaks in the forward flow of speech, are associated with separating clauses or sentences, taking a breath, or utterance-planning hesitations.^21–22 Typical adults align their pauses with syntactic boundaries and rarely pause at locations not related to syntax.^23,24 Increased frequency of pauses, longer pause duration, and insertion of pauses in syntactically-inappropriate locations are known to occur in a variety of neurogenic disorders including TBI,^18,25 Parkinson’s disease,^26,27 and ataxia.²⁸

In addition, we documented the frequency and duration of speech disfluencies in fighters and healthy controls. Disfluency rates for healthy speakers in reading are low, about 0.005 events per syllable.²⁹ Disfluencies produced by typical speakers consist of interjections (insertion of extra words such as “um” and “you know”), phrase revisions (e.g., “the rainbow was, the rainbow is”), and phrase repetitions (e.g., “a division of, a division of”).³¹ Neurogenic disfluency (or stuttering), an acquired speech disorder characterized by sound- and syllable-repetitions, prolongations of speech sounds, and silent blocks, has been reported in numerous cases of TBI.^31–34 Although the pathology of neurogenic disfluency is not well understood, it has been associated with lesions of the cerebral hemispheres, subcortical white matter, basal ganglia, thalamus, cerebellum, and brainstem, all of which can be affected in fighters with RHI.³⁴ Neurogenic stuttering is characterized by more frequent interruptions in the forward flow of speech and by differences in the nature of disfluent behavior. Disfluencies associated with neurogenic stuttering include sound- and syllable repetitions, prolongations of speech sounds, and silent, tense blocks between words.³⁰

In this study, we sought to determine if professional fighters differed in articulation rate, number and duration of pauses, and number and duration of disfluencies from age-matched controls with no reported history of brain injury. We hypothesized that slower articulation and increased number of and duration of pauses and disfluencies would occur in fighters because of their increased probability of RHI exposure. We also explored the effect of type of fighting, boxing vs. mixed martial arts, on articulation rate, pausing, or disfluency, postulating that boxers would exhibit more speech deficits than MMA fighters because boxing leads to more repetitive head impacts. Elucidating differences in speech characteristics between fighters and healthy controls may lead to the development of biomarkers for neurologic decline associated with RHI that are non-invasive and easy to obtain in clinical settings.

Methods

Participants

Participants in the study were drawn from the PABHS, which was approved by the Cleveland Clinic Institutional Review Board (IRB). All participants gave informed consent in providing recorded speech samples. Participants were de-identified, and information about visit number, age, number of professional fights, primary language, country of birth, race, and years of education were supplied for each fighter.

To avoid the potentially confounding effects of gender or second language acquisition on speech production, 100 men whose primary language was American English (60 boxers and 40 MMA fighters) were selected for this project. All fighters were at least 18 years of age and had completed at least 10 years of education. Control participants were 55 men who spoke American English as their primary language. Of the 55 control participants, 28 were drawn from the PABHS. These participants had no self-reported history of neurological disorders, head trauma, military service, or participation at a high school level or higher in sports involving head impacts, and they had completed at least 10 years of education. An additional 27 controls were drawn from a previous project on tongue strength and speech rate.³⁷ That study was approved by the IRB of the University of New Mexico, and all participants provided informed consent. These 27 participants reported no history of speech, language, or neurologic deficits and had completed at least a high school education. Summary demographic information can be found in Table 1. Mean age and age range was similar for the fighters (mean = 39.3 years, SD = 10.5, range = 24 to 69) and controls (mean = 41.6 years, SD = 15.1, range = 20 to 78). Thirty-one percent of fighters and 29% of controls were Black, and 67% of controls and 52% of fighters were white.

Table 1:

Demographic information with mean (SD) and range.

	Boxers (n = 60)	MMA Fighters (n = 40)	Healthy Controls (n = 55)
Age in Years	43.35	33.13	41.16
	(10.77)	(6.39)	(15.14)
	24 – 69	24 – 47	20 – 78
Years of Education	12.82	13.55
	(1.62)	(1.85)
	10 – 18	11 – 18
Number of Fights	26.95	15.78
	(20.24)	(18.00)
	0 – 85	0 – 78
Race:
Asian	2	0	0
Black	20	11	16
American	1	0	0
Indian/Alaska Native
Pacific Islander	1	0
White	24	28	37
Multiple Races	4	0	0
Other	7	1	2
Unknown	1	0	0

Stimuli

Each participant read aloud the initial paragraph of the Rainbow Passage,³⁸ a standard passage used in dysarthria research and clinical batteries to assess motor speech skills and speech intelligibility. Scripted reading allows for the assessment of connected, relatively natural speech production with little added variance from the cognitive/linguistic differences associated with spontaneous speech tasks. Fighters and PABHS controls were recorded in an examination room at the Cleveland Clinic Lou Ruvo Center for Brain Health in Las Vegas, NV, with varying microphones and recorders. These recordings were provided to the Speech Acoustics and Intelligibility Laboratory at the University of New Mexico under an agreement with the PABHS. The remaining controls were recorded in a sound-treated room in the Department of Speech and Hearing Sciences at the University of New Mexico using a Marantz PMD 670 digital recorder through a Shure SM 10a head mounted microphone positioned about 1 cm from the corner of the speaker’s mouth. Audio recordings for all participants were saved as .wav files for analysis.

Procedures

Each recorded passage was transcribed by the second author using CHILDES CHAT/CLAN software.³⁹ The transcriptions were opened in a text grid in the acoustic analysis package Praat.⁴⁰ In a two-step analysis process, trained student researchers used waveform and spectrogram displays to mark each syllable, pause, and disfluency using standard written definitions,⁴¹ and these preliminary markings were reviewed and adjusted as needed by the first and second authors. A custom Praat script was used to record the duration of each syllable, pause, and disfluency, and duration values were then extracted into a spreadsheet for analysis. (It should be noted that articulation rate and number of pauses and disfluencies can be determined in clinical settings without the need for scripts or marking of individual syllables.)

Acoustic measures

Articulation rate - the number of syllables produced per second - was calculated by adding the duration of all fluently produced syllables and dividing the total time it took in seconds to produce those syllables. It can be difficult to separate speech motor behaviors from cognitive and linguistic abilities in passage reading tasks. However, articulation rate is thought to be primarily a measure of speech motor function because pauses and disfluent speech productions are not included in the calculation.

Pauses were defined as silences in the recorded samples that lasted for at least 100 ms and were not associated with articulatory behaviors such as voice onset time or consonant closure intervals.⁴² Syntactically-appropriate (SA) pauses occur at major or minor clause boundaries and are often accompanied by a breath.²⁴ The 25 SA pause locations were marked for transcribers in a written copy of the passage. Syntactically-inappropriate (SI) pauses occur at non-grammatical locations in a sentence, such as within a prepositional phrase or between a noun and a verb. SI pauses are relatively unusual in healthy speakers but are known to occur in individuals with dysarthria²⁴ and may affect listener comprehension of dysarthric speech.⁴³

Disfluencies were defined as interruptions in the forward flow of speech. Two types of disfluencies were measured in this study - more typical (MT) and less typical (LT).³¹ MT disfluencies are observed in the speech of typical speakers who do not have neurogenic disorders or developmental stuttering. They include interjections and revisions. LT disfluencies, sometimes called “stuttering-like” disfluencies, are associated with developmental and neurogenic stuttering but occur infrequently in typical talkers. LT disfluencies include part-word repetitions and prolongations. LT disfluencies can also include silent pauses, but these were counted as SA or SI pauses in this study because it is difficult to determine whether silences represent disfluent interruptions or purposeful hesitation. Obvious reading errors, such as word substitutions or added words, were not included in either disfluency category or in the calculation of articulation rate.

Reliability for acoustic measures

Intra-rater and inter-rater reliability were measured for syllable durations and occurrence and duration of pauses and disfluencies. For intra-rater reliability, 20 randomly selected recordings whose final markings were created by the second author were re-transcribed and re-marked in Praat by her, and five additional recordings whose final markings were created by the first author were reanalyzed by her. Percent agreement for pause and disfluency occurrences was measured by calculating the number of times the original marked events matched the re-measured events. Intra-rater reliability for durations of pauses and disfluencies was obtained by comparing the original durations with the re-measured durations via Pearson product-moment correlations for each of the two authors. Inter-rater reliability for pause and disfluency occurrence was obtained by calculating percent agreement for twenty passages originally measured by the second author with remeasurements performed by the second author. Inter-rater reliability for pause and disfluency durations was performed via Pearson correlations between the second author’s original durations for the twenty passages with durations measured by the first author.

Statistical analyses

Statistical analyses were conducted using IBM SPSS Statistics Version 28.⁴⁴ Normality of data was assessed using Kolmogorov-Smirnov and Shapiro-Wilk tests in conjunction with values of skewness and kurtosis. Homogeneity of variance was assessed with Levene’s test for parametric tests. ANOVA was used to compare the three groups of speakers (controls, boxers, MMAs) for variables that were normally distributed (articulation, number and duration of SA pauses). When Levene’s test indicated non-homogeneity of variance, Games-Howell multiple post-hoc comparisons were conducted. The non-parametric Kruskal-Wallis test and Bonferroni-adjusted pairwise comparisons were used to compare the groups for variables that were not normally distributed (number and duration of SI pauses, LT disfluencies, and MT disfluencies). To provide estimates of practical significance,⁴⁵ effect sizes were calculated using ƞp² for ANOVA and Cohen’s d for Kruskal-Wallis tests and Games-Howell post-hoc comparisons.

Results

Reliability

Intra-rater percent agreement for identification of pauses and disfluencies in the original analysis and re-analysis was high (98.3 to 99.1%), and Pearson product-moment correlations for repeated measurement of duration of syllables, pauses, and disfluencies ranged from r = .94 to .98 (p < .01). Inter-rater reliability measures also indicated sufficient reliability for acoustic measurements. Agreement between the two judges for identification of pauses and disfluencies was 92%, and correlations for syllable, pause, and disfluency duration ranged from r = .90 to .94 (p < .01).

Articulation rate

Univariate ANOVA for group (boxers, MMAs, controls) revealed a significant and medium effect, F(2, 152) = 9.81, p < .01, ƞp² = .11. Mean articulation rate for boxers was 4.59 syls/s (SD = .51), MMAs 4.63 syls/s (SD = .54), and controls 5.04 syls/s (SD = .68). Games-Howell post-hoc comparisons revealed that both groups of fighters had significantly slower articulation rates than controls, and effect sizes were medium (boxers vs. controls mean difference = −.450 syls/s, p < .01, Cohen’s d = 0.75; MMAs vs. controls mean difference = −.4.8 syls/s, p < .01, d = 0.67.). Boxers did not differ from MMAs (mean difference = −.043 syls/s, p = .92). See Figure 1.

Figure 1:

Box and whisker plot of articulation rates for boxers, MMA fighters, and healthy controls showing medians, interquartile ranges, minimum and maximum values, and outliers (>1.5 * IQR) for each group.

Number and duration of pauses

Figure 2 shows the number of pauses for each group. All participants produced several SA pauses during the reading passage. ANOVA revealed a significant difference in the number of SA pauses among groups, F(2, 152) = 7.51, p < .01 with a medium effect size (ƞp² = .11). Boxers produced an average of 12 pauses (SD = 3.77), MMAs 10.25 pauses (SD = 2.78), and controls 9.91 pauses (SD = 2.35). Games-Howell post-hoc comparisons showed that boxers produced more SA pauses than controls (mean difference = 2.09, p < .01, d = 0.66) and MMAs (mean difference = 1.75, p = .02, d = 0.53). MMAs did not differ significantly from controls in number of SA pauses (mean difference = 0.34; p = .81).

Figure 2:

Box and whisker plot of number of SA and SI pauses for boxers, MMA fighters, and healthy controls showing medians, interquartile ranges, minimum and maximum values, and outliers (>1.5 * IQR) for each group.

The univariate ANOVA for duration of SA pauses also showed a significant difference among the three groups, F(2, 152) = 7.81, p < .01, ƞp² = .09. Mean SA pause duration for boxers was 572 ms (SD = 17 ms), MMAs 465 ms (SD = 14), and controls 493 ms (SD = 12). Boxers produced longer pauses than controls (mean difference = 78 ms, p < .02, d = 0.37 and longer pauses than MMAs (mean difference = 106 ms, p = .01, d = .51). MMAs did not differ from controls (mean difference = 28 ms, p = .49).

Not all speakers produced pauses in atypical locations: 43 of 60 boxers, 22 of 40 MMA fighters, and 30 of 55 controls produced SI pauses. The Kruskal-Wallis test revealed significant differences in number of SI pauses among the groups, H(2) = 9.27, p < .01. Boxers produced a mean of 4.02 SI pauses (SD = 5.50), MMAs 1.48 SI pauses (SD = 2.10), and controls 1.40 SI pauses (SD = 1.82). Boxers had significantly more SI pauses than controls, H(1) = 22.17, p = .04, d = 0.6. and than MMAs, H(1) = 22.17, p = .02, d = 0.63. MMAs did not differ from controls, H(1) = −0.86, p = 1.00. SI pause duration differed across groups, H(2) = 18.86, p < .01. Mean SI pause duration for boxers was 231 ms (SD = 55 ms), for MMAs 224 ms (SD = 24), and controls 360 ms (SD = 148). Controls produced longer SI pauses than both boxers, H(1) = −26.68, p < .01, d = 1.16 and MMAs, H (1) = −25.73, p < .01, d = 1.28.

Number and duration of disfluencies

Figure 3 shows the number of MT and LT disfluencies produced by the three groups. MT disfluencies were produced by 41 boxers, 30 MMAs, and 20 controls. The Kruskal-Wallis test showed a significant difference in number of MT disfluencies among the three groups, H(2) = 19.46, p < .01. Boxers had more MT disfluencies than controls, H(1) = 21.17, p < .01, d = 0.72, as did MMAs, H(1) = 31.11, p < .01, d = 0.69. Boxers and MMAs did not differ from one another, H(1) = 1.07, p = 1.00. There was no significant difference among the groups in duration of MT disfluencies, H(2) = 2.22, p = .33. Average MT disfluency duration was 973 ms for boxers (SD = 656 ms), 746 ms for MMAs (SD = 380), and 861 ms for controls (SD = 339).

Figure 3:

Box and whisker plot of number of MT and LT pauses for boxers, MMA fighters, and healthy controls showing medians, interquartile ranges, minimum and maximum values, and outliers (>1.5 * IQR) for each group.

LT disfluencies were produced by 24 boxers, nine MMAs, and 11 controls. The Kruskal-Wallis test revealed significant group differences in LT disfluency frequency, H(2) = 9.18, p < .01. Pairwise comparisons showed that boxers produced more LT disfluencies than controls, H(1) = 18.74, p = .02, d = .60 but not significantly more than MMAs, H(1) = 16.18, p = .08. MMAs did not differ from controls, H(1) = 2.57, p = 1.00. There was a significant difference in LT disfluency duration among the three groups, H(2) = 7.06, p <.01. Boxers produced significantly longer disfluencies than MMAs, H(1) = 11.99, p = .05, d = 0.74. No other pairwise comparisons yielded significant group differences. Mean LT disfluency durations were 716 ms (SD = 617 ms) for boxers, 369 ms (SD = 232) for MMA fighters, and 433 ms (SD = 266) for controls.

Discussion

We examined articulation rate, pauses, and disfluencies in speech produced by professional fighters and healthy control speakers to identify potential speech indicators of neurological decline associated with RHI. Both types of fighters, boxers and MMAs, differed from controls in articulation rate, producing syllables at a slower rate than controls by nearly half a syllable per second on average. Despite overlap among the groups – 42% of fighters’ rates fell within the interquartile range for controls - this difference was robust. Only 4% of fighters had articulation rates that were faster than the 75^th percentile for controls, whereas 54% of fighters had rates slower than the 25^th percentile for control speakers. Reduced speech rate is associated with several types of dysarthria due to neurologic damage,¹⁰ including those caused by damage to one or both cerebral hemispheres (unilateral upper motor neuron and spastic dysarthrias), the basal ganglia control circuit (hypokinetic dysarthria), and the cerebellar control circuit (ataxic dysarthria). The finding of significantly reduced articulation rate is consistent with results of two previous case studies of speech in boxers.^8,9

Interruptions in the forward flow of speech, pauses and disfluencies, were also measured in this study. Two types of speech pauses were documented in this study: syntactically appropriate (SA) pauses in locations such as at the ends of sentences and between clauses, and syntactically inappropriate (SI) pauses in grammatically unexpected locations (e.g., within prepositional phrases, between nouns and verbs). Boxers produced significantly more SA pauses per passage on average than controls or MMAs, and mean SA pause duration for boxers was 79 ms longer than for controls. Boxers also produced two to three more SI pauses on average than MMAs or controls. Overlap among the three groups in number of pauses was observed. However, 48% of boxers produced more SA pauses than the 75^th percentile for controls (11 pauses). Whereas only 55% of MMAs and control participants produced at least one SI pause while reading the passage, 72% of boxers produced at least one, and nine boxers produced 10 or more SI pauses.

Increased pausing, especially at syntactically unnatural locations, has been associated with several forms of dysarthria, including those associated with Parkinson’s disease^24,46,47 and TBI.¹⁸ Hammen and Yorkston,⁴⁷ for example, reported that participants with Parkinson’s disease had twice as many SI pauses than healthy controls. Excessive pausing may reflect poor respiratory support for speech: speakers with dysarthria may need to take more frequent breaths than healthy speakers if they have reduced vital capacity, poor vocal fold closure, and/or velopharyngeal insufficiency. Pause behavior can also be affected by cognitive and linguistic factors, and increased pausing is associated with TBI and dementia.^{18,19,48–50}

The speech of fighters was also characterized by more frequent disfluencies. For MT disfluencies (i.e., interjections, phrase revisions), fighters produced more than one additional disfluency per passage than controls. Only 36% of speakers in the control group produced at least one MT disfluency, whereas 67% of boxers and 75% of MMAs did. Seven boxers and four MMAs produced five or more MT disfluencies while reading the passage aloud. The occurrence of LT disfluencies, the type associated with developmental or neurogenic stuttering, was also elevated in boxers. Only 20% of controls and 23% of MMAs produced at least one LT disfluency, but they were present in 40% of boxers. Seven of the boxers produced three or more LT disfluencies, but none of the MMAs or controls produced more than two.

Disfluent speech is associated with several acquired neurologic conditions including Parkinson’s disease^35,51 and TBI.^31–34 The higher levels of LT disfluencies over MT disfluencies in TBI and Parkinson’s disease may reflect motoric disturbances in these disorders.³⁵ MT disfluencies, on the other hand, are thought to be influenced by cognitive and linguistic factors such as word retrieval skills and semantic/syntactic planning ability.¹⁹ The finding of increased MT and LT disfluencies in boxers may reflect both cognitive and/or linguistic and motor issues associated with RHI.

Boxers differed more from the control group than the MMAs did, particularly in the frequency of pauses and disfluencies. Potential explanations may relate to the fight moves considered to be acceptable in the two combat sports. Because hits below the belt are not allowed in boxing, the majority of blows are concentrated on the torso, neck, and head.⁵² MMA rules allow for a wider variety of fighting moves such as punches, kicks, takedowns, and chokeholds.⁵³ Differences in the nature or frequency of head trauma may lead to differences in speech symptomatology in boxers and MMAs. It should be noted, however, that boxers in this study were also older than MMAs (boxer mean = 43.5 years, MMAs mean = 33.1), had experienced more fights than MMAs (boxers mean = 27.0 fights, MMAs mean = 15.8) and had fewer years of education (boxers mean = 12.8 years, MMAs mean = 13.6). Thus, no strong conclusions regarding speech differences due to fighting style can be drawn from this study.

Limitations and future directions

This study represents a first step in documenting speech characteristics of professional fighters exposed to RHI: fighters had slower rates of speech, more pauses, and more disfluencies than healthy controls. In future studies with larger sample sizes, factors such as age, level of education, cognitive ability, race, and RHI exposure should be explored. Part of the control group for this study was drawn from a different project, so it is possible that differences in data collection or participant characteristics could account for some of the variance. Articulation rates, pause frequencies, and disfluency rates for the controls were in line with other studies, however.²⁰

It is difficult to disentangle motor deficits from cognitive-linguistic factors in reading tasks. Although articulation rate is thought to be primarily a measure of speech motor function,¹⁹ speaking speed, pausing and fluency are influenced by cognitive, linguistic, and social factors such as dialect in addition to neuromuscular functions.^14,20 Differences in education level and literacy between fighters and healthy controls may also have influenced the measures obtained from passages read aloud in this study. The Rainbow Passage has a reading grade level of 5.5,⁵⁴ and although all participants completed at least the 10^th grade, their reading ability was not assessed. To minimize the impact of these factors, we plan to examine diadochokinetic rates for repeated speech syllables to capture speech motor behavior as has been done recently for military service members and veterans with TBI⁵⁵ and athletes with sports-related concussion.⁵⁶ In addition, we are examining verbal fluency and discourse characteristics to more fully understand cognitive and linguistic deficits in fighters.

Finally, the measures in this study do not fully describe the motor speech deficits associated with RHI in fighters. We are currently investigating listener perceptions of speech intelligibility and speech subsystem integrity in boxers with the aim of producing profiles of motor speech deficits associated with neurogenic decline. In other future studies, we will relate acoustic and perceptual speech measures with imaging data available through the PABHS to investigate longitudinal course of speech changes with RHI.

Conclusions

In summary, the finding of slower articulation rate in both boxers and MMA fighters compared to individuals with no history of RHI and the increased occurrence of pauses and disfluencies in the speech of boxers suggest changes in speech motor behavior and/or cognitive/linguistic difficulties related to RHI, although additional and more specific investigation is required to determine causal relationships. Acoustic speech measures show great promise in early identification of RHI-related behavioral changes. Tauro and colleagues recently differentiated speakers with and without a history of RHI with 85% accuracy using automated analysis of 13-second voice recordings, including the correct classification of boxers who were just 21 years old.¹³ Articulation rate, pausing, and disfluencies are relatively easy to measure, even in everyday speaking conditions and by automatic recognition systems, so they have the potential for early clinical identification of RHI-associated neurologic decline and to track changes in behavior over time.