Nasalance during use of pharyngeal and glottal place of production

Purpose: This study obtained nasalance scores during use of compensatory articulation (CA) and compared nasalance between groups with and without hypernasality and with and without CA. Methods: Speech samples were obtained from 43 individuals with and without velopharyngeal dysfunction during repetition of 20 phrases originating 860 audio recordings and their respective nasometric values. After excluding 143 recordings due to low quality, the remaining 717 samples were rated by three speech language pathologists (SLPs), independently, for presence or absence of hypernasality and CA. Nasalance scores for the 553 samples rated with 100% agreement among the SLPs were grouped according to the auditory-perceptual ratings: Group 1 (G1) included samples without hypernasality and without CA (n=191); Group2 (G2) included samples with hypernasality and without CA (n=288); Group 3 (G3) included samples with hypernasality and with pharyngeal fricative (n=33); Group 4 (G4) included samples with hypernasality and with glottal stop (n=41). Results: Analysis of variance (ANOVA) revealed significant difference nasalance scores which were significantly higher for G2, G3, and G4 (p<0.0001) when compared to G1. The use of pharyngeal fricative (G3), particularly during /f/ (p=0.0018) and /s/ (p=0.0017) productions resulted in nasalance scores significantly higher than scores found for G2. Conclusion: Significantly higher nasalance values where identified during use of pharyngeal fricative.


INTRODUCTION
Cleft lip and palate (CLP) is one of the congenital anomalies with an impact in the structures and function of the velopharyngeal mechanism.Primary palatoplasty does not always warrant adequate velopharyngeal function for speech for individuals who present with the CLP, resulting in velopharyngeal dysfunction (VPD) and related speech disorders.Besides hypernasality, audible nasal air emission and weak oral pressure consonants, speakers with CLP or VPD may use atypical place of production, resulting in a group of consonant production errors (CPE) described in the literature as compensatory articulation (1,2) .A group of authors (1) classified CPE related to CLP and VPD as pre-uvular and post-uvular place of production errors.Among the post-uvular errors, pharyngeal fricatives (PF) and glottal stops (GS) have been described as the most common atypical productions associated to CLP and VPD (3) .The use of CA can compromise intelligibility of speech, having an impact in socio-educational development and vocational placement of the speaker.Individuals with history of CLP and VPD may also experience bullying, due to the speech disorders involving CPE and hypernasality.
Many cleft palate teams around the world follow parameters established for evaluation and treatment of CLP and other craniofacial anomalies as proposed by ACPA (4) .Instrumental assessment of velopharyngeal function is recommended for adequate management of VPD and may involve the use of nasometry to corroborate findings of speech nasality.While nasometric assessment provide a quantifiable measure of speech nasality and has been extensively studied and described in the literature (5)(6)(7) , the recommendation for not conducting nasometric assessment during use of CA is based on clinical impressions.
During our clinical practice, we observe that the use of postuvular place of production leads to impaired velopharyngeal movements during nasoendoscopic and videofluoroscopic assessments.Making decision regarding the management of VPD at the presence of CA is difficult and some cleft palate teams delay surgical procedures for correcting velopharyngeal insufficiency until after speech therapy is conducted to resolve CPE involving use of post-uvular CA.Since nasometry has clinical applications during evaluation and treatment of speech disorders related to VPD, we questioned the possibility of using nasalance scores to corroborate auditory-perceptual findings of acoustic events happening at atypical places in the vocal tract.As reported in the literature (8) , the use of GS surrenders velopharyngeal function once the valving of airflow to generate air pressure for plosion is obtained before the flow reaches the level of the velopharyngeal mechanism.Nasometric scores have been shown to vary according to several aspects including language, age and gender of the speaker, as well as the phonetic context and the length of the speech stimuli (9)(10)(11)(12)(13) .One article (14) also reported that, besides nasal acoustic energy related to voiced resonance, nasometric assessment can also reflect acoustic events related to unvoiced aerodynamic turbulence occurring within the vocal tract.
This study assumes that, besides corroborating auditoryperceptual findings of voiced resonance related to speech nasality, nasometric assessment can also provide information related to other acoustic events occurring within the vocal tract.The hypothesis to be tested is whether the production of frication and plosion at a post uvular place of production may significantly increase or reduce nasalance scores.Finding whether the use of CA can affect nasalance scores, however, will not bring direct useful information for management of VPD, since it is known that the use of pharyngeal place of frication (such as in PF) or laryngeal place of plosion (such as in GS) may minimize (or completely surrender) velopharyngeal activity.Further information about the impact of CA on nasalance scores, however, may provide clinician with a tool for measuring outcome of behavioral management of CPE.This study compares nasalance scores between speakers with and without hypernasality and use of PF and GS.

METHODS
This study was approved by the Human Research Ethics Committee at the research site.Sampling was conducted during a period of three months in which a group of 43 patients, with operated CLP, accepted to participate in the study.To control for some factors that could affect listeners' ratings and nasalance scores recording from participants with dysphonia, nasal obstruction on both nares and nasal snort (turbulent audible nasal air emission) were not included in the study.Patients with syndromes or other conditions that could affect performance or compliance during speech recording were not recorded.A series of 20 phrases was used for audio and nasometric recordings (Table 1).The stimuli are currently used for speech recordings at the research site and the phonetic composition was designed according to the recommendations of Henningsson et al. (1) .To favor identification of CPE, 20 phrases (12 loaded with high pressure sounds, 8 loaded with low pressure sounds, each with recurrence of a single target consonant) were repeated once by the participants.Nasalance scores were obtained using Nasometer model 6400 (Kay PENTAX).An audio-recording was captured simultaneously during nasometric assessment using an AKG -C420 microphone attached to the Nasometer's separator plate.
A total of 860 audio and nasometric samples (Table 2) were recorded (43 participants versus 20 stimuli = 860 phrases).Using Sound Forge 8.0, the audio samples were edited into a material for listeners' ratings of speech production.During editing, 143 samples (17%) presented with external noise and were excluded due to low quality for auditory-perceptual ratings.The remaining 717 recordings were grouped according to the stimuli and presented to three experienced speech language pathologists (SLPs) for ratings of speech nasality and use of PF and GS.The SLPs worked daily in the assessment of speech disorders related to CLP for over five years.They were instructed to listen to each phrase, independently, as many times needed until they were able to indicate the presence or the absence of hypernasality and use of PF or GS.During identification of PF and GS, listeners had to indicate which targets in each stimulus were substituted or co-produced with the use of PF and GS.All ratings were compiled into a Microsoft Excel © table and only the samples rated with 100% agreement among Nasalance and compensatory articulation CoDAS 2014;26(5):395-401 the three SLPs were kept.A total of 164 (23%) samples were rated with less than 100% agreement among the listeners and were excluded.
The remaining 553 samples were distributed into four groups according to the auditory-perceptual ratings: G1 included samples without hypernasality and without CA; G2 included samples with hypernasality and without CA; G3 included samples with hypernasality and with pharyngeal fricative; and G4 included samples with hypernasality and with glottal stop.Nasalance scores for the 553 samples were calculated.Mean nasalance scores and standard deviations were identified for each group in each of the 20 stimuli.Analysis of variance (ANOVA) was used to compare means between the groups, followed by Tukey's test with significance established at p<0.05.

RESULTS
This study initially established a data bank of speech recordings representative of presence and absence of hypernasality, PF and GS.To assure representativeness of the samples, all audio recordings with adequate quality for auditory-perceptual ratings were presented to three listeners who rated the samples independently.Only the samples with 100% agreement between the raters (consensus) were grouped for comparison of nasalance findings.PF or GS were identified with consensus by the SLPs when all targets within the stimuli were produced with the CA and when co-productions were not used.Review and discussion of ratings with less than 100% agreement during independent ratings were not pursued in this study.In the group of samples excluded due to lack of consensus between the raters, 27% involved phrases produced with fricative targets, 24% plosive targets and 18% liquid targets.
Table 5 presents the difference between nasalance means for G2 and G3.Scores identified for /f/ and /s/ targets where significantly higher for G3 (productions with pharyngeal fricatives and hypernasality) when compared to /f/ and /s/ targets in G2 (productions only with hypernasality).For some stimuli, the sample size was too small for inferential statistics (No size).

DISCUSSION
Identifying speech samples representative of use of PF and GS was the first task addressed in this study.Representativeness of a sample was established with consensus auditory-perceptual ratings obtained with three SLPs.With the objective of testing the hypothesis that the use of PF or GS would have a significant impact in nasalance scores, only samples for which the judges agreed 100% regarding presence or absence of hypernasality, PF and GS were maintained in the study.Obtaining similar sample size for the four groups of interest, however, was not possible and only 13% of the samples in the overall data bank were rated as representative of use of PF (6%) and use of GS (7%).These findings, however, cannot be interpreted as indicative of frequency of use of PF and GS in the population with CLP since participants' selection was biased to identification of samples representative of the four groups of interest in the study.Authors point to the importance of conducting a careful sampling once the goal is to obtain listeners' perceptual ratings of the recordings (1,15,16) .From the initial pool of 860 recordings, 17% were lost due to poor recording quality, indicating the importance of listening the recordings during or immediately after the data collection.While clinical demands may require a quick process of sampling, special care with the recording procedures as well as with the length and phonetic context of the stimuli are needed to document speech outcome accurately.Sell (16) and Henningsson et al. (1) suggested the importance of using speech stimuli involving recurrence of a single consonant target to identify the use of CPE related to CLP.Balancing the phonetic context of the speech stimuli to maximize identification of both, hypernasality and CPE, however, is a complex task.While ratings of speech nasality may be facilitated with the use of longer stimuli (such as short passages instead of short phrases), the use of shorter stimuli (short phrases, words, syllable repetitions) with recurrence of a single target may favor the identification of CPE and may be applied at an earlier age.
For multiple listeners' ratings, care must be taken with listeners experience and preparation.The judges in this study had extensive experience in evaluating speech disorder related to CLP and have a history of over five year of daily work with management of CLP and VPD in a large craniofacial team.Extensive training was not provided but samples representative of presence and absence of hypernasality, PF and GS were played to the SLPs prior to the task and could be used by the listeners as reference if needed.For this study, all ratings were done independently and after all listeners finished the ratings, only samples for which the listeners agreed 100% (consensus) regarding the presence or the absence of the variables of interest were maintained in the study.The findings of this study support prior literature that indicates that PF is commonly observed for fricative targets while GS were used for both, plosive and fricative targets (3) .PF and GS were not used for liquid targets, as commonly observed (17) .
While it is recommended that cut-off values should be stimuli specific, findings for G1 support prior studies which suggest that a cut-off value of 27% or lower can be used for clinical interpretation of nasalance scores obtained for Brazilian Portuguese language (18,19) .Within the groups with hypernasal speech, mean nasalance for stimuli with high pressure sounds (plosive and fricative) were similar to mean nasalance obtained for stimuli targeting low pressure sounds (liquids).This was an expected finding since the presence of audible turbulent nasal air emission was controlled to avoid increase in the scores due to nasal snort (14,20) .
Looking into the hypothesis tested in this study, we observed a significant difference in nasalance score between the group with hypernasality and the group with hypernasality and use of PF, for the stimuli targeting /f/ and /s/ sounds.The current finding, therefore, does not support Ferreira et al. (21) , who suggested that the use of CA has no impact in nasalance scores.Still this finding must be interpreted with care.The sample sizes for the stimuli were limited once ratings were distributed across the four groups, as indicated in Table 2.This study suggests that, while evaluating a patient that uses backing of fricative targets into post-uvular place of production, one can expect even higher nasalance measures than observed for speakers who are hypernasal but use no PF.Because only samples with PF showed significantly higher nasalance values when compared to samples with hypernasality and use of GS, the presence of frication generated at the pharynx appears to add acoustic features to the speech signal that resonates into the upper vocal tract and can be pick-up by the microphones of the Nasometer.These findings point towards a trend for higher nasalance scores when frication is produced at the pharynx; however, interpretation is limited due to small sample sizes, single repetition of each stimuli, unbalanced vowel distribution and unbalanced number of syllables among the stimuli.Further studies involving larger and better-controlled speech samples are needed for a better understanding of this finding.
Nasometric evaluation involving speech produced during use of CA is not recommended to corroborate speech nasality and velopharyngeal function findings.But when we consider that not all articulatory gestures used by speakers with VPD and history of CLP are captured by human ears (22,23) , higher nasalance findings during use of PF may suggest presence of pharyngeal articulatory gestures and may help corroborating clinical identification of PF.Finally, considering that use of biofeedback of the speech signal with nasometry is a behavioral therapeutic strategy useful with speakers when addressing nasal snort, one can also explore the use of nasometric feedback when addressing CPE involving PF.

CONCLUSION
This study established nasalance scores for speech samples with and without hypernasality and with and without use of pharyngeal fricative and glottal stop.In general, the nasalance values for the samples with hypernasality were higher than the scores for samples without hypernasality.More specifically, however, when pharyngeal fricatives were used in substitution of oral targets /f/ and /s/, the scores where significantly higher than those obtained for speech with hypernasality but without the use of CA.
*AFG was involved in the elaboration and design of the study, collected and prepared data, was involved in data analysis and interpretation and also in the preparation and revision of this manuscript; VCCM MIP-K were involved in the design of the study, in data analysis and interpretation and also in the preparation and revision of this manuscript; JRPL was involved with study design, data analyses and revision of this manuscript; TAG was involved in data collection, in data analysis and interpretation and also in the preparation and revision of this manuscript; JCRD was involved in the study design, collected and prepared data for comparisons and was involved in data analysis and interpretation and also in the preparation and revision of this manuscript.

Table 2
also indicates that 288 samples (52%) were rated with presence of hypernasality and absence of CA, while 74 (13%) samples were rated as representative of hypernasal speech and

Table 2 .
Summary of targets, recorded samples, number of samples not included due to low quality (No Quality), number of samples included and rated by listeners (Included), number of samples excluded due to lack of consensus by listeners (Excluded), sample used in this study (Samples), and number of recordings in each group (G1, G2, G3, G4)

Table 5 .
Difference in nasalance scores between the groups with hypernasality and with CA with p-value indicative of significance when <0.05