International Journal of Pediatric Otorhinolaryngology (2008) 72, 885—889

www.elsevier.com/locate/ijporl

The effect of task type on fundamentalfrequency in children

Susan Baker *, Barbara Weinrich, Megan Bevington,Katharine Schroth, Emily Schroeder

Miami University, Speech Pathology and Audiology, 2 Bachelor Hall, Oxford, OH 45056, USA

Received 1 November 2007; received in revised form 21 February 2008; accepted 26 February 2008Available online 8 April 2008

KEYWORDSVoice disorders;Fundamentalfrequency;Acoustic analysis

Summary

Objective: The purpose of this study was to examine the influence of task type on thefundamental frequency (F0) produced by young children. Fundamental frequency is aparameter which describes the rate of vocal fold vibration. The influence of task typeon F0 values is important for health professionals designing and implementingassessment protocols for children with voice disorders.Methods: Forty-eight healthy children between the ages of 5.0 and 7.11 years wereevaluated in this study. Each child completed four tasks used to elicit a voice samplefor subsequent analysis of F0. The tasks included: (a) sustaining the vowel/a/(Vowel),(b) sustaining the vowel embedded in a word at the end of a phrase (Phrase), (c)repeating a sentence (Sentence), and (d) counting from 1 to 10 (Counting). Each childwas evaluated by one of two clinicians who had an equal number of years of trainingand were the same age and gender.Results: A repeated measures analysis of variance (ANOVA) was used to examine theinfluence of task type on F0 values. The results revealed a significant difference in F0between the four elicitation tasks ( p = .002). Pair wise comparisons revealed thatCounting elicited higher F0 values compared to Phrase ( p = .018) and Sentence tasks( p = .001). There were no significant interaction effects for task by age, gender, orclinician who evaluated the child ( p > .05).Conclusions: The results of this study revealed that task type does significantlyinfluence F0 values in young children. This finding is clinically important as it indicatesthat the same task should be used to monitor changes in the voice over time inrelationship to surgical or behavioral interventions.# 2008 Elsevier Ireland Ltd. All rights reserved.

* Corresponding author. Tel.: +1 513 529 2553;fax: +1 513 529 2502.

E-mail address: bakerse1@muohio.edu (S. Baker).

0165-5876/$ — see front matter # 2008 Elsevier Ireland Ltd. All rigdoi:10.1016/j.ijporl.2008.02.019

1. Introduction

Acoustic analysis of the voice is one importantcomponent of a comprehensive evaluation of a

hts reserved.

886 S. Baker et al.

Table 1 Age and gender distribution for enrolledparticipants

Males Females Total

Age (years)5.0—5.11 7 8 156.0—6.11 9 6 157.0—7.11 9 9 18

Total 25 23 48

pediatric voice disorder. Acoustic measures, such asfundamental frequency (F0), which is a measure ofthe rate of vocal fold vibration, provide informationabout the tone-generating capacity of the larynx [1].Numerous reports of normative data across a wideage range are available for comparison of F0 valuesobtained during a voice evaluation. Lesions on thevocal fold tissue, such as nodules, polyps, and cysts,tend to result in theproduction of lowerF0 values dueto the increasedmass of the vocal folds [2]. Obtainingacoustic measures is a non-invasive means of estab-lishing baseline vocal function prior to surgical orbehavioral intervention for a voice disorder. Acousticmeasures can also be used to monitor progressthroughout a behavioral treatment program.

Previous research suggests that acoustic mea-sures obtained in a voice assessment may be influ-enced by how they are elicited [3—7]. In a specificexamination of the influence of task type on F0,Zraick et al. [7] completed an analysis of theresponse of adult females (n = 12), adult males(n = 12), and children (n = 12) ranging from 5 to10 years of age, to various elicitation tasks (e.g.,counting, reading, sustaining a vowel). The resultsof this study indicated that there was variability inF0 across tasks within all of the subject groups(males, females, and children). However, the F0

variability was only significant for the adult females.It is unclear from this study why the variability wasgreater in the female subjects, however, after closeinspection of the data, the authors suggested that F0

appears to be influenced by length of utterances,speaking style, gender, and age of the subjects.

In the above described study, all children wereevaluated by the same individual. The speech sam-ples required for the elicitation of F0 in youngchildren (e.g., sustaining vowels, sentence repeti-tion) required amodel from the evaluating clinician.Therefore, individual difference amongst cliniciansis another variable that should be considered in taskelicitations. Despite giving similar verbal instruc-tions, clinicians may vary in their frequency varia-bility while speaking and in their own modal F0.These factors can influence the model provided forthe child during an evaluation.

The purpose of the current study was to examinethe effect of task type on measures of F0 in youngchildren. The specific age range that was targetedwas 5.0—7.11 years. This age range was selected asit represents a developmental age that children areexpected to be able to participate fully in a com-prehensive acoustic protocol (i.e., be able to com-plete all required speech samples). This studyfurther expanded the previous work in the area ofexamining the influence of tasks effects, but with alarger number of children which also allowed for

meaningful gender comparisons. A secondary pur-pose of this study was to examine the effect of theindividual differences between two cliniciansregarding the models provided for the tasks.

2. Methods

2.1. Participants

Forty-eight children between 5.0 and 7.11 years ofage were recruited from an elementary school inCincinnati, Ohio for this study. Attempts were madeto recruit equally across the age range. A descrip-tion of the age and gender of the participants isdisplayed in Table 1. All children were enrolled inage- and grade-appropriate settings and werenative speakers of English. Children were excludedif they were serviced with an Individualized Educa-tion Plan, had unrepaired velopharyngeal incompe-tence, or were perceived by two certified speech-language pathologists (Baker and Weinrich) to havea voice disorder or abnormal oral-peripheral struc-tures. This study was approved by the Miami Uni-versity Institutional Review Board and parent/guardian consent for each child participating inthe study was obtained prior to data collection.

Prior to collecting the experimental sample, eachchild enrolled in the study was required to pass abilateral pure-tone hearing screening. A portablepure-tone audiometer (Maico Model MA39) was usedto screen each child in a quiet room at the school.The child was instructed to raise his or her handwhenever a tone was audible. Pure tones werepresented at 25 dBHL at 1000, 2000, and 4000 Hz.

2.2. Procedure

All experimental procedures were conducted in twoquiet rooms in the elementary school building. Aheadset microphone (MicroMic Series, C420 PP) wasworn by each participant and connected to a digitalaudio tape recorder (TASCAM Sony DAPI) to recordvoice samples. A 2-in. mouth-to-microphonedistance was maintained at all times. All samples

The effect of task type on fundamental frequency in children 887

Fig. 1 Mean fundamental frequency by task type.

were obtained by one of two clinicians and a scriptwas used to elicit all samples in order to standardizethe type and amount of instruction. Children wereassigned to be evaluated by one of the clinicians inthe order in which they were dismissed from theirclassroom (i.e., they were placed with the clinicianthat was available when the child arrived at thetesting location). Both clinicians were 24-year oldfemales, who were completing their 2nd/final yearof graduate training.

Fundamental frequency (F0) measurements wereobtained from each child utilizing four experimentaltasks described below. The name of the each of thetasks was written on one side of an index card. Thechild was presented with a deck of the four cards(task names face down) and asked to select one ofthe cards. The child would then complete that taskand be asked to select another card until the deckwas completed. The clinicians provided one modelof each task for the participant. The child wasallowed one practice production to provide clarity.All children completed three trials of each of thefour experimental tasks. The average of the threetrials was used for data analysis. The child wasencouraged to complete all tasks at a comfortablepitch and loudness level.

2.2.1. VowelThe first task required the participant to sustain avowel, without the context of a phrase. For thistask, each child was instructed to inhale normallyand sustain the vowel/a/for 5 s.

2.2.2. PhraseThe second task required the participant to sustainthe target vowel/a/, which was embedded within aphrase [8]. The phrase that used was, ‘‘I need amop.’’ The children were instructed to hold out thevowel/a/in ‘‘mop’’ for 5 s and they were instructednot to say the/p/at the end of the word.

2.2.3. SentenceThe third task required the child to repeat a sen-tence after it was produced by the clinician. Thesentence was, ‘‘Bob wants a ball.’’

2.2.4. CountingThe fourth task required the child to inhale comfor-tably and count from 1 to 10. The child was encour-aged to say the numbers with a normal rate (not toofast or slow).

2.3. Data measurement

The recorded samples/acoustic signals from the DATrecorder were digitized and analyzed utilizing the

Multi-Dimensional Voice Analysis Program (KayPen-tax, Model 5105). The middle one-second of thesustained vowel samples was used for the analysis.Measures were taken from the onset and offset ofvoicing for the speaking/connected speech samples(Sentence/Counting).

2.4. Data analysis

A repeated measures analysis of variance (ANOVA),with task as the within-subject factor and age,gender, and clinician who provided the model asbetween-subject factors, was completed. Pair wisecomparisons were completed as needed using pairedt-tests.

3. Results

All children passed the hearing screening as outlinedin the Section 2. All children raised their hand inresponse to all tones.

The mean F0 values for each of the elicitationtasks were as follows: (1) Vowel (M = 240.46,S.E. = 2.96, Range = 182.87—294.07); (2) Phrase(M = 236.55, S.E. = 2.58, Range = 205.88—299.93);(3) Sentence (M = 235.66, S.E. = 2.97, Range = 172.96—294.99); and (4) Counting (M = 246.51,S.E. = 2.76, Range = 212.48—300.83). The mean F0

for all four experimental tasks are displayed inFig. 1. Fundamental frequency values by age, gen-der, and clinician who elicited the task are displayedin Figs. 2—4, respectively. The analysis revealed asignificant difference amongst the four task types,F(3, 41) = 6.12, p = .002.

Pair wise comparisons completed between eachof the task conditions revealed that the meanCounting F0 was significantly greater than thePhrase F0, t(41) = 3.17, p = .018. Counting F0 wasalso significantly greater than the Sentence F0,t(41) = 4.05, p = .001. Significance values for thepair wise comparisons were Bonferroni adjusted.Both the first and second author (SB and BW) who

888 S. Baker et al.

Fig. 3 Mean fundamental frequency by gender.

Fig. 2 Mean fundamental frequency by age.

Fig. 4 Mean fundamental frequency by clinician.

are certified speech-language pathologist, withexperience in perceptual assessment of vocal qual-ity, listened to the counting tasks compared to thesentences and phrase tasks. These authors per-ceived approximately 50% of the counting tasks tobe higher in pitch.

The age � task, gender � task, and clinician �task interactions were not significant indicating thatnone of these factors had an influence on F0 varia-bility across the tasks. Additionally, there were nomain effects for age, gender, or clinician. Significantdifferences in mean F0 values between the agegroups and between male and females were notexpected in theage range represented in this sample.

4. Discussion

The results of this study revealed that task type didhave a significant influence on the F0 produced by

young children. Counting appeared to elicit a higherF0 overall. This task was chosen as it represents anautomatic speech task that relies on very littleinstruction from the person evaluating the child.It would be interesting to know if the higher fre-quency produced during this task was due to theabsence of a clinician model for the task (which inthis study, the clinicians had a lower F0 due to age)or due to the manner in which children in this agerange may recite numbers. Children in this agerange may recite numbers with increased pitchvariability as to imitate counting exercises fromfamiliar music or television programs. Also, impor-tantly, while the clinician models for the Vowel,Phrase, and Sentence tasks were at a lower F0 thanthe participants’ habitual frequency, the F0 valuesproduced by the children in the study for all taskswere comparable to the normative data for this agerange (e.g., [9,10]).

It is important to observe the actual differencebetween the highest and lowest mean F0 producedin this study (235 and 246 Hz) across the 4 tasks. Thisrange represents a one semitone difference (46 to47 ST) and the difference between approximatelyless than one note on a musical scale (A3 to B3).When viewed in this context, the authors feel thatthere are two important conclusions to make fromthis data. First, regardless of the task elicited fromthe child, the F0 produced should be a value thataccurately represents their F0 and can be comparedwith confidence to published normative data. This isimportant in the context of evaluating children whomay not understand or comply with all tasks pre-sented in an assessment protocol, in which casesimpler tasks, such as counting, may be easier toelicit. However, the significance of the difference inF0 produced from various elicitation tasks is impor-tant when these measures are being used to docu-ment clinical progress or outcomes for patients,caregivers, and 3rd party payers. Small increasesor decreases in these measures may be used torepresent change due to an intervention. It is impor-tant for clinicians to realize that completing differ-ent tasks at baseline and post-treatment/post-surgery may influence F0 values.

As discussed in the Section 2, the two clinicianswho elicited the samples had similar training andwere of the same gender and age. Based on theanalysis, any variability that may have existedbetween these clinicians in terms of presentationof the tasks or the F0 they used to provide modelsdid not have a significant effect on the outcome. Inthis study, any variability of presentation wouldhave been unlikely due to the use of a script toelicit the task, as well as control for the number ofmodels provided and practice productions allowed

The effect of task type on fundamental frequency in children 889

for each child. It can be observed in Fig. 4 thatClinician 1 did elicit higher F0 values overall com-pared to Clinician 2. This is likely due to differencesin the habitual pitch between the clinicians. It isimportant to note that the overall difference in F0

between the two groups of children assessed by theclinicians was not significantly different.

Sources of variability in F0 are important forresearch of this type and potentially future exam-inations of normative data in children. Futureresearch in this area should include assessment ofthe clinician model on the child’s F0. A comparisonof the use of a male and female model would beinteresting with this age group as a male F0 modelwould be much lower than a child’s. In regards toage, it would also be interesting to evaluate theinfluence of elicitation tasks on preschool-age chil-dren who may have greater variability in F0 due totheir ability to understand instructions for the tasks.It has also been suggested that F0 values vary whenanalyzed with different acoustic analysis softwarepackages [11,12]. This source of variability shouldbe considered when obtaining measures for norma-tive data. This protocol should also be implementedin children with voice disorders. Variability in F0

across tasks in this clinical population may be influ-enced as well by limitations in range of phonatoryfrequencies.

Future research should also evaluate the F0 mea-sured from samples specifically elicited by an exam-iner to samples obtained from spontaneous speech.An examination of this correlation is important inunderstanding the reliability of elicited samples as atrue representation of the child’s typical behavior.

5. Conclusions

This study provides evidence that the manner inwhich tasks are elicited does significantly influenceF0 values in young children. This finding is clinicallyimportant as it indicates that the same task shouldbe used to monitor changes in the voice over time inrelationship to surgical or behavioral interventions.Manner of elicitation should be considered whencompleting normative studies that examine acousticmeasures in children and potentially when using

Available online at www

these measures to characterize disordered laryn-geal function.

Acknowledgements

We would like to thank Ms. Allison Schaser and Ms.Ashley Schuler for their assistance with data mea-surement. We would also like to thank Mr. MichaelHughes for his assistancewith the statistical analysis.

Conflict of interest

None of the authors have any financial or personalconflicts of interest with the study presented in thismanuscript.

References

[1] R.H. Colton, J.K. Casper, Understanding Voice Problems: APhysiological Perspective for Diagnosis and Treatment, thirded., Williams and Wilkins, Baltimore, 1990.

[2] J. Stemple, L. Glaze, B. Klaben, Clinical Vvoice Pathology:Theory and Management, Singular Publishing, Canada, 2000.

[3] M. Andrews, A. Summers, Voice Treatment for Children andAdolescents, Singular Thomson Learning, San Diego, 2001.

[4] A. Reich, J. Mason, R. Frederickson, R. Schlauch, Factorsinfluencing fundamental frequency range estimates in chil-dren, J. Speech Hear. Disord. 54 (1989) 429—438.

[5] R. Zraick, W. Marshall, L. Smith-Olinde, J. Montague, Theeffect of task on determination of habitual loudness, J.Voice 18 (2004) 176—182.

[6] R. Zraick, J. Nelson, J. Montague, P. Monoson, The effect oftask on determination of maximum phonaltional frequencyrange, J. Voice 14 (2000) 154—160.

[7] R. Zraick, S. Skaggs, J. Montague, The effect of task ondetermination of habitual pitch, J. Voice 14 (2000) 484—489.

[8] J. Harden, N. Looney, Duration of sustained phonation inkindergarten children, Int. J. Pediatr. Otorhinolaryngol. 7(1984) 11—19.

[9] L. Glaze, D. Bless, P. Milenkovic, R. Susser, Acoustic char-acteristics of children’s voice, J. Voice 2 (1988) 312—319.

[10] Wilson, D.K., Voice Problems of Children, third ed., Williams& Wilkins, Baltimore, 1987.

[11] S. Bielamowicz, J. Kreiman, B. Gerratt, M. Dauer, G. Berke,Comparison of voice analysis systems for perturbation mea-surement, J. Speech Hear. Res. 39 (1996) 126—134.

[12] R. Roark, Frequency and voice: perspectives in timedomains, J. Voice 20 (2006) 325—354.

.sciencedirect.com