A Simple Technique for Determining Velopharyngeal Status during Speech Production

Abstract

Clinical evaluation of velopharyngeal function relies heavily on auditory perceptual judgments that can be supported by instrumental examination of the velopharyngeal valve. Many of the current instrumental techniques are difficult to interpret, expensive, and/or unavailable to clinicians. Proposed in this report is a minimally invasive and inexpensive approach to evaluating velopharyngeal function that has been used successfully in our laboratory for a number of potentially difficult-to-test clients. The technique is an aeromechanical approach that involves the sensing of nasal ram pressure (N-RamP), a local pressure sensed at the anterior nares, using a two-pronged nasal cannula. By monitoring the N-RamP signal, it is possible to determine the status of the velopharyngeal port (open or closed) during speech production. Four case examples are presented to support its clinical value.

Introduction

Velopharyngeal closure is critical for production of vowels and oral consonants and thus, has a profound impact on speech intelligibility. Several clinical populations, for example, children with a history of a cleft palate or individuals with dysarthria, have impairments in velopharyngeal function that contribute to problems with speech production. Clinical evaluation of velopharyngeal function often relies heavily on the clinician’s auditory perceptual judgment. For example, a clinician might judge a client’s speech to be hypernasal or hyponasal and assign a severity rating to it (e.g., mildly hypernasal). However, auditory perceptual judgments are known to be unreliable¹ and are not necessarily well-correlated to velopharyngeal status^2,3,4. Therefore, auditory perceptual judgments of velopharyngeal function are often supplemented with instrumental evaluation. Instrumental approaches to understanding velopharyngeal function have proven to have both clinical and research value. Unfortunately, some of these are intrusive, difficult to interpret, expensive, and/or unavailable to clinicians. Proposed in this report is a simple, minimally invasive, and inexpensive approach to evaluating velopharyngeal function that has been successfully used for a number of potentially difficult-to-test clients.

Instrumental Approaches to Evaluating Velopharyngeal Function during Speech Production

There are many instrumental approaches to evaluating velopharyngeal function during speech production. The focus here is on those that are commonly used in clinical settings by speech-language pathologists and other clinical professionals, and include selected acoustical, visual imaging, and aeromechanical approaches.

An acoustical approach to velopharyngeal evaluation is attractive because the data are generally easy to record, requiring only a microphone and a recording device. Analysis of the speech signal often involves spectrographic analysis with an emphasis on selected acoustic features; for example, formant bandwidth (coupling of the velopharyngeal-nasal pathway to the oral pathway is associated with wider bandwidths) and formant frequencies (coupling of the velopharyngeal-nasal pathway to the oral pathway usually introduces a low-frequency “nasal formant” into the spectrum). Nevertheless, this type of analysis can be difficult to carry out and interpret without substantial experience and expertise.

Another more widely used acoustical technique for evaluating velopharyngeal function is nasometry. Nasometry requires the use of two microphones, one positioned in front of the lips and one positioned in front of the anterior nares. A cumbersome, horizontally oriented plate is positioned between the microphones to isolate the sound energy from the two sources. Nasometry allows for the calculation of nasalance, which is the quotient of nasal sound pressure level to nasal + oral sound pressure level. Although nasalance values have been found to be strongly associated with perceptual judgments of nasality (at least when specified cut-off scores are used), they have been found to be weakly correlated with estimated velopharyngeal orifice area⁵. The weak correlation between nasalance values and velopharyngeal orifice size is likely due, at least in part, to the effects of variables such as turbulent nasal air flow and degree of mouth opening on nasalance^{5, 6}.

Visual imaging is another common approach to evaluating velopharyngeal function in clinical practice. One type of imaging is videonasendoscopy, which involves passing a flexible endoscope posteriorly through the nasal pathways and viewing the velopharynx from above. Another type is videofluoroscopy, which involves the use of x-ray to view oral, nasal, pharyngeal, and laryngeal structures from both sagittal and coronal perspectives (though not simultaneously). These two imaging techniques permit direct observation of the velum and pharyngeal walls as they move in relation to one another during speaking and swallowing. A major motivation for using a visual imaging approach is to examine contributions of the velum and lateral and posterior pharyngeal wall to velopharyngeal closure, information that is often critical in planning medical management, including surgical and prosthetic interventions. Visual imaging can be expensive, invasive, and poses safety risks and is, therefore, applied selectively.

An aeromechanical approach is also used frequently to evaluate velopharyngeal function during speech production. Perhaps the most powerful aeromechanical method involves obtaining simultaneous measures of nasal airflow, oral air pressure, and nasal air pressure. By entering these measures into an equation developed by Warren and DuBois⁷, it is possible to estimate the size of the velopharyngeal port during running speech production. Knowing the size of the velopharyngeal port allows the clinician to determine if medical management is necessary or if behavioral management alone may be appropriate.

Perhaps the simplest aeromechanical technique currently in use involves the sensing of nasal airflow. A small mask is attached to a pneumotachometer and placed over the outer nose. The pneumotachometer is connected to a sensitive differential pressure transducer and the signal is calibrated so that mass airflow exiting the nose can be measured. Nasal airflow is relatively well correlated to velopharyngeal port size, but only at small sizes⁸; the correlation decreases as velopharyngeal size increases. Nevertheless, nasal airflow provides an excellent means of determining if the velopharynx is open or closed during speech production. For example, if nasal airflow is detected during oral sound production, it can be inferred that the speaker has a velopharyngeal leak.

The Nasal Ram Pressure (N-RamP) Technique

The motivation for creating a less cumbersome aeromechanical technique came from an interest in documenting velopharyngeal development during sound production in infants. By the 1990’s, Thomas Hixon and Jeannette Hoit had conducted investigations of velopharyngeal function in healthy children as young as 3 years of age⁹ and in healthy adults as old as 97 years⁶. For these earlier investigations, they had used the nasal airflow technique (described above) to determine if the velopharynx was open or closed during speech production. However, when they attempted to place a mask over an infant’s nose, the infant moved his head and arms, tried to push the mask away, and cried most of the time. Hixon and Hoit explored other means of detecting nasal airflow in infants, including thermal imaging and nasal thermistors, but these presented their own challenges. Finally, they tested a technique involving the use of nasal ram pressure and found it to be successful with infants.

The nasal ram pressure (N-RamP) technique requires the use of a two-pronged nasal cannula, such as that shown in Figure 1. This is the same type of nasal cannula that is used to deliver oxygen to sick infants. The two probes that insert into the anterior nares are sometimes cut short so that the distal ends do not touch the nasal membranes. To detect pressure change, the other end of the nasal cannula is connected to a sensitive differential pressure transducer (with a +/− 2 cmH₂O diaphragm, such as is often used to detect airflow during speech production). As shown in Figure 2, the probes of the cannula are inserted into the anterior nares and the tubing is taped to the cheeks. In addition, a small microphone is taped to the forehead. The resultant air pressure signal and acoustic signal are each amplified, displayed on an oscilloscope for on-line monitoring, and recorded on a multichannel digital recorder for later analysis. The infant is also videotaped at close range (with the audio signal from the microphone routed into the video camera). The purpose of the video is to have a record of mouth status and body movements (discussed below), as well as documentation of the communication situation.

Figure 1.

A nasal cannula.

Figure 2.

An infant seated in a high chair with a microphone taped to her forehead and a nasal cannula in place.

In this context, the nasal air pressure signal reflects a ram pressure; that is, the local average pressure in the probe tube. This is akin to the ram pressure that is used to measure aircraft speed. This local pressure is directly related to the local airflow within the tube, but it is not necessarily related to the mass airflow from the nose (because air escapes around the probe tubes). For this reason, Hixon and Hoit decided that it was not useful to attempt to calibrate the pressure signal, as it would not provide meaningful information. What it did provide, however, was critical information that could be used to determine if the velopharynx was open or closed. Specifically, the signal could be interpreted as follows: (a) negative pressure = inspiration through the nasal airway, (b) positive pressure = expiration through the nasal airway, and (c) pressure at zero (atmospheric) = no airflow (breath holding) or a closed velopharynx during inspiration or expiration. Thus, during vocalization (which almost always occurs during expiration), positive pressure indicates an open velopharynx and pressure at zero indicates a closed velopharynx. It is also important to determine if the mouth is open or closed. If the mouth is closed, nasal pressure is necessarily positive during vocalization (except in the case of instantaneous mouth closure for /p/ and /b/). Thus, the vocalizations that are of greatest interest are those produced with the mouth open.

One of the major advantages of the N-RamP technique over nasal airflow is that most head or body movements do not affect the N-RamP signal. This is because the sensing probes are free to move within the nasal vestibules and are not sealed airtight into the anterior nares. Thus, they are part of an open system and are not sensitive to the types of artifacts that often prevail in closed aeromechanical systems (such as a nasal mask coupled to the face). The only major exception that we have encountered is when a child bounces vertically in his chair or when a baby is bounced on a parent’s knee. Despite its relative immunity to movement artifacts, the N-RamP signal is very sensitive to local pressure changes that are associated with velopharyngeal opening and closing.^a

The N-RamP technique was first applied to the study of velopharyngeal function during vocalization in a longitudinal study of six infants (from age 2 to 6 months) conducted by Thom, Hoit, Hixon, and Smith¹⁰. All six infants tolerated the nasal cannula and the vast majority of vocalizations recorded from them were nondistress (i.e., “happy”) vocalizations. The results of this study indicated that the frequency of velopharyngeal closure generally increases with age, but that consistent closure for oral sound production is not yet achieved by 6 months. These findings challenged previously published acoustic and perceptual data that suggested that velopharyngeal closure occurs by approximately 4 months of age^{11, 12}.

Case Examples

We have had the opportunity to use the N-RamP technique to evaluate velopharyngeal function in a variety of individuals. Here, we offer descriptions of four case examples that illustrate the ease and value of the N-RamP technique: a healthy infant, a school-age child with a speech sound disorder, a teenager with an orofacial syndrome, and an adult with amyotrophic lateral sclerosis.

Healthy Infant

Z was a healthy male infant who was seen in our laboratory once a month beginning at age 6.5 months and continuing until age 19 months. Z was a typically developing infant, as determined by data collected at each visit using the Brigance Diagnostic Inventory of Early Development (gross and fine motor subtests)¹³, the MacArthur Communicative Development Inventory¹⁴, and the Vocal Development Checklist¹⁵. During his first visit, Z was producing primarily vowel-like productions. He proceeded through the expected phases of babble, protowords, and real words. At 20 months of age he had an expressive vocabulary of roughly 100 words, was producing 2- to 3-word utterances spontaneously, and could imitate utterances cued by the investigator. His utterances were analyzed by breath group (between inspirations) and were classified as being open throughout the breath group, closed throughout the breath group, or closed during part of the breath group.

Shown in Figure 3 is the percentage of nondistress mouth-open utterances that Z produced at each monthly visit wherein his velopharynx was either closed throughout (N-RamP=0) or closed during part of the breath group (N-RamP=0 for part and N-RamP=positive for part). His nondistress utterances included raspberries, syllable utterances (e.g., vowel productions of long or short duration, consonant-vowel combinations produced in isolation or in multiples and reduplicated and variegated babbling), protowords, and words.

Figure 3.

Longitudinal data for Z shown as percentage of breath groups containing nondistress utterances produced with the velopharynx either closed throughout (closed) or closed during part (mixed).

Results show that Z obtained velopharyngeal closure for nearly all vowel-like and babble utterances containing oral consonants at 8 months of age (91% were closed throughout or closed during part). This percentage decreased at 9, 10 and 11 months of age (range 50 to 75%) but was consistently over 85% for all subsequent months. During the period from 9 to 11 months, Z’s consonant inventory increased from two to eight consonants being produced at least twice during a recording session. It is unclear if this increase relates to the lower percentage of utterances being produced with the velopharynx closed. It is clear, however, that the N-RamP technique can be successfully applied to the study of velopharyngeal development in very young children.

School-Age Child with Speech Sound Disorder

B, a 6-year old boy, was referred for an evaluation of his velopharyngeal function. According to his school speech-language pathologist, B was “struggling to produce certain high-pressure consonants,” which generated a concern that he might have a velopharyngeal leak. He was diagnosed with a moderate-severe speech sound disorder, mild motor programming deficit, and severe tongue thrust. Upon oral examination a slightly asymmetric elevation of the uvula was noted. An audiometric screening indicated that his hearing sensitivity was within normal limits (20 dB HL ¹⁶), and a language screening indicated that his receptive and expressive language were within normal limits (CELF-4 Screening ¹⁷).

Velopharyngeal function during speech production was evaluated using the N-RamP technique. The speech tasks included imitation of nonsense words imbedded in a carrier phrase (e.g., “Say /isi/ again”), imitation of high-pressure consonant loaded sentences (e.g. “The puppy was playing with a rope”), picture naming using the Iowa Pressure Articulation Test (IPAT¹⁸), and conversational speech.

The protocol was designed to determine the percentage of high-pressure consonants produced with a closed velopharynx. However, due to his tongue thrust and speech sound disorder, many of the targeted high-pressure consonants were substituted with a /θ/ or another off-target consonant. Because of this, it was not possible to calculate quantitative data as originally intended. Nevertheless, it was possible to make certain qualitative observations. For example, we found that when B attempted to produce a high-pressure consonant in the context of an upcoming nasal consonant, he often generated a large nasal pressure pulse during the pressure consonant. This is illustrated in Figure 4 during the production of /k/ (which he substituted for /g/) in “again.” The large positive pressure spike associated with the /k/ indicates that his velopharynx was open momentarily. Figure 5 also shows a large positive pressure spike during the production of a fricative consonant that preceded a nasal consonant when producing the name “Sean.” Apparently, B opened the velopharynx inappropriately early in anticipation of the upcoming nasal consonant. It is important to note that this pattern was not consistent; there were many instances in which B produced similar utterances with appropriate velopharyngeal closure.

Figure 4.

Glottal waveform and nasal pressure (P_nas) signal for the target utterance “again” produced by S. Note that the transcription in the figure is based on S’s actual production and that it does not match the target “again.” A positive P_nas peak can be seen near the release for the stop consonant /k/.

Figure 5.

Glottal waveform and nasal pressure (P_nas) signal for the target utterance “Sean” produced by S. Note that the transcription in the figure is based on the S’s actual production and that it does not match the target utterance. A positive P_nas peak can be seen during production of the fricative.

We concluded that B did exhibit a velopharyngeal leak during what should have been oral sound productions, but that the velopharyngeal leak was inconsistent and not attributable to a structural or peripheral neural impairment. The inappropriate velopharyngeal opening was concerning, although it appeared to be inconsistent and had less impact on his speech output than the pervasive tongue thrust. A report of the findings was sent to his school speech-language pathologist so that treatment could be tailored to include strategies such as biofeedback to address the unusual velopharyngeal pattern.

Young Adult with Goldenhar Syndrome

S, a 21-year old male with hemifacial microsomia (Goldenhar syndrome), referred himself to the University of Arizona Clinic for Adult Communication Disorders for an evaluation 12 months after he had undergone a Le Fort I osteotomy. This is a surgical procedure wherein the maxillary bone and hard palate are separated from the skull and repositioned; the procedure is designed to move the maxillary bone forward for better alignment with the mandible. S stated his primary complaint as follows: “I have had VPI since the surgery. It was worse right after the surgery but it improved after about 2–3 months. People would ask me to repeat what I said. It has continued to get better but I still don’t sound like I did before the surgery.” His parents gave a similar account.

A full clinical evaluation was completed that included an oral mechanism examination, intelligibility testing, nasometry (Nasometer; Kay Elemetrics, Lincoln Park, NJ), an aeromechanical assessment (PERCI-SARS; Chapel Hill, NC), as well as data collection to assess velopharyngeal status using the N-RamP technique described in this paper. The speech protocol consisted of three productions each of plosive consonants (p, b, t, d, k, g) and fricative consonants (s, z, ʃ, ʒ) in syllable initial, final, and medial positions combined with the vowels /i/ and /ɑ/. In addition, sentence reading (GOS.SP.ASS sentences ¹⁹), passage reading (Rainbow Passage ²⁰), and conversational speech samples were recorded.

Table 1 shows the percentage of breath groups produced with the velopharynx closed throughout and closed during part. Note that the reading passage^b and conversational speech samples contained a mixture of oral and nasal consonants; therefore, the number of breath groups that were closed during part were expected to be higher than if only breath groups containing oral consonants were analyzed. This analysis approach was selected because it was judged to be more directly comparable to the analysis approach typically used for nasometry data (for which mean nasalance values are calculated across speech samples containing both oral and nasal sounds). The N-RamP data indicated that S achieved velopharyngeal closure on the majority (67 to 90%) of syllable utterances (CV, VCV, and VC) in which the consonant was a stop. By contrast, he achieved closure on only a minority of the syllable utterances (25 to 37% were closed throughout, and 15 to 34% were closed during part) in which the consonant was a fricative.

Table 1.

Speech tasks produced by S and percentage of utterances identified as being produced with the velopharynx closed throughout and closed during part of the breath group.

Speech Task	Closed Throughout %	Closed During Part %
/b/	84	0
/p/	82	0
/d/	88	0
/t/	90	0
/g/	67	0
/k/	72	0
/s/	25	32
/z/	31	15
/ ʃ /	37	24
Reading (Rainbow passage)	76	14
Conversational speech (across all consonants)	72	17

To compare data from different instrumental techniques used during the same session for S, a correlation was computed to determine level of agreement for the N-RamP data (closed throughout only) and the nasometer data. Three correlations were computed. These included the pooled syllable data (CV, VCV, VC) data (r=−.78), passage reading (r=−.84), and conversational speech sample (r=−.89). In all cases, the correlation was negative because a high mean nasalance score (reflecting the fact that there is considerable acoustic energy emanating from the nose) corresponded to a small percentage of utterances produced with the velopharynx closed and a low mean nasalance corresponded to a high percentage of utterances produced with the velopharynx closed. Although we do not expect a perfect correlation because nasometry has been shown to be susceptible to variations in oral acoustic impedance²², our correlations suggest that there is a reasonable level of agreement between the two techniques. Direct comparison of the data from the PERCI-SARS was not completed because S had a severely deviated nasal septum and results differed when the placement of the nasal flow and nasal pressure transducers were switched. In general, larger velopharyngeal port openings were found for utterances containing fricative consonants compared to those with stop consonants. This is consistent with both the nasometry and N-RamP findings. We concluded that S might benefit from a period of trial behavioral therapy, with an emphasis on achieving velopharyngeal closure in fricative consonant contexts. We recommended that the N-RamP signal be used for visual biofeedback so that S could monitor his velopharyngeal closure on line.

Adult with Amyotrophic Lateral Sclerosis

C, a 57-year old male diagnosed with amyotrophic lateral sclerosis (ALS) came to the clinic for a speech evaluation to determine if he would be a candidate for augmentative and alternative communication (AAC). His speech was slow, effortful, and of abnormally low loudness. Velopharyngeal function was assessed using the N-RamP technique while he produced the nonword and sentence repetition tasks described in the previous two cases.

Results indicated that he closed his velopharynx during only 26% of his high-pressure consonant productions (11.1% were closed throughout the production; 14.8% were closed during part of the production). Table 2 contains the results for each of the target sounds tested.

Table 2.

Target sounds produced by C and percentage of utterances identified as being produced with the velopharynx closed throughout and closed during part of the breath group.

Target Sound	Closed Throughout (%)	Closed During Part (%)
/p/	40	10
/b/	0	0
/t/	20	0
/d/	0	0
/k/	16	33
/g/	0	0
/v/	0	0
/f/	33	50
/s/	50	16
/z/	0	20
/ θ /	0	50
/ ð /	0	0
/ ʃ /	0	40
/ ʒ /	0	0
/ tʃ /	0	25
/j/	0	0

Although our focus was on the status of the velopharynx during high-pressure consonant productions, the N-RamP waveforms provided us with other valuable information about velopharyngeal behavior. For example, we noted several instances wherein N-RamP was positive (indicating an open velopharynx) during productions of vowels that surrounded the high-pressure consonants, one of which is illustrated in Figure 6. This figure contains two consecutive productions of “Say /isi/ again” in which the velopharynx was closed during the /s/ segment in both; however, the second production (on the right) shows that the velopharynx was open during the /i/ that preceded the /s/ and the /i/ that followed the /s/. This velopharyngeal opening was not perceptible to the ear.

Figure 6.

Glottal waveform and nasal pressure (P_nas) signal for two repetitions of the target utterance “Say /isi/ again” produced by C. The first production (left) of the utterance indicates a closed velopharynx during the “/isi/” (P_nas=0). During the second production of “/isi/” (right), P_nas is positive during the vowels that precede and follow the consonant, indicating an open velopharynx.

Although C produced the majority of the high-pressure consonants with his velopharynx open, we noticed that several of these productions sounded as if they had been produced with the velopharynx closed. To investigate this further, we conducted an informal perceptual study with two speech-language pathology graduate students and two certified speech-language pathologists serving as listeners.

The listeners were presented with the speech samples (auditory only) and were asked to indicate “whether the velopharynx is open or closed” during the target sound productions. The two graduate students identified the velopharynx as closed when it was open on 54% and 55.8% of the productions. The two speech-language pathologists identified the velopharynx as closed when it was open on 16% and 28.4% of the productions. These findings suggest that the ear is not always a good judge of velopharyngeal status and that velopharyngeal function may be impaired before it is detected in clients with ALS.

Conclusions

The N-RamP technique has been used successfully in our laboratory to evaluate clients with a variety of etiologies, as well as to track velopharyngeal development in healthy infants and young children. We see N-RamP as a powerful clinical tool for a number of reasons: (a) it is minimally invasive, comfortable, and does not interfere with speech production; (b) it is inexpensive, accessible, and does not require much training or experience to use; (c) its data collection and interpretation/measurement procedures are relatively quick; (d) it can be used with clients of any age with neural, structural, or functional disorders; (e) it reveals information about velopharyngeal function that is not easily detected in the acoustic signal; and (f) it has potential as a biofeedback tool for behavioral management of velopharyngeal dysfunction.

Biographies

Kate Bunton Ph.D. is CCC-SLP is an Associate Professor in the Department of Speech, Language, and Hearing Sciences at the University of Arizona. Dr. Bunton received her B.S. from the University of Iowa in Speech and Hearing Sciences, her M.S. and Ph.D. are from the University of Wisconsin-Madison in Communicative Disorders. She completed her postdoctoral studies at the University of Arizona as part of the National Center for Neurogenic Communication Disorders. Dr. Bunton’s research relates reductions in speech intelligibility to segmental and suprasegmental production difficulties in persons with dysarthria, including the effects of velopharyngeal dysfunction.

Jeannette D. Hoit, Ph.D. is Professor of the Department of Speech, Language, and Hearing Sciences at the University of Arizona (UA) and a speech-language pathologist. Dr. Hoit received her B.A. in Anthropology from the UCLA, her M.A. in Communicative Disorders from SDSU, her Ph.D. in Speech and Hearing Sciences from UA, and pursued postdoctoral study in the Harvard School of Public Health Respiratory Biology Program. She is a Fellow of the American Speech-Language-Hearing Association and past-Editor of the American Journal of Speech-Language Pathology. Her research focuses on speech physiology, with particular emphasis on normal aging and development, neuromotor speech disorders, and respiratory function and dysfunction, including ventilator-supported speech and speaking-related dyspnea.

Keegan Gallagher, B.A., B.S. is a doctoral student of the Department of Speech, Language, and Hearing Sciences at the University of Arizona (UA). She received her B.A. in Anthropology from Arizona State University, and her B.S. in Speech, Language, and Hearing Sciences from the UA. Her research focuses on speech physiology and development in infants and children, speech impairment in individuals with neurodegenerative diseases, and augmentative and alternative communication.