- ] -

ATTRIBUTION OF AGE FROM PERCEIVED SPEECH

Leo W.A. van Herpt. and Ton Hoebe

l • INTRODUCTION

When we hear someone's voice we form an impression of his appearance . How we trust these judgments shows in our confusion when we find that ' the man ' s voice' belongs to a woman or when a calm, deep, sonorous sound is produced by a five feet tall man with a baby face. But in general these judgements are reliable and rather accurate. Much is still unknown about the cues which are used in this attribution process, but since they have perceptual importance speech must have certain acoustic characteristics which are responsible for eliciting those perceptual processes. Over forty years ago already Sanford (1942) wondered whether this connection is of a sort thatcan only be sensed by judges or whether it could ever be analyzed and stated in a qualitative manner. To find out which acoustic characteris­tics are related s ystematically to the subjective impressions more research

. on the iden t ification of relevant paraineters on both levels must be done. On the perceptual level the long-term voice features that result from ana­tomical speaker-characterizing differences are the most reliable. So primarily the biologically detennined voice-indices of age, sex and phy­sique will be used in our investigations to define criteria of speaker-char­acteristics that can be used for automatic speaker recognition and for the eval uation of coded speech. We started with a series of experiments concerning factors that iRfluence the perception of aging in the human voice. The first study, the one tha t i s reported on i n the present paper, i s a imed at me thodol ogical aspects for subsequent research. The experiment consisted of an evaluative rating by twenty-nine listeners based on two minutes oral reading by fifty Dutch speakers aged from eight­een to eighty-seven. The aim was to explore J ) . factors that influence the accuracy of age estimation on the basis of vocal cues alone and 2) the po­tency of the ten bipolar semantic scales and twelve speech therapeutic terms to differentiate between younger and older speakers. The results will show that listeners are ab le to estimate the speakers' age with reasonable accuracy although several interfering and interacting vari­ables have been identified, but first we'll give a short survey of some aspects considered relevant in literature.

2. LITERATURE

It is not a new asser tion that vocal functioning reflects the on-going phy­sical changes with age and that these changing vocal parameters can be per­ceptually observed. Publications

, in the thirties already showed that listeners are able to esti­

mate the age of speakers solely by listening. Pear (1931) investigated with

I

- 2 -

a s s istence of the BBC the problem of jud g in g the personal ity of n ine speak­_ers of d ifferent ages, sex and interests by their radio voice . On the whole the (4000) l i s teners estimated the ages with fair s uccess. Stern (1932) tried to relate changes of the a ging voi ce as perceived by l isteners to the phy s iological change s of the larynx caused by age . In a study of Herzog (1933) too, a ge.was judged more accurately than one woul d expect by chance. Allport and Cantril (1934) conducted a series of experiments to determine to what extent the natural voice is ·a val i d ind icator of various features of personal i t y . Speakers' ages were judged in two experiments . In the first the voices TNere matched by SO listeners to the ages given; in the·second the ages of three s peakers were merely est imated by the 190 judges. Both experi­ments y iel ded pos itive results . In the last one there was a tendency to cen­ter the ages around the median, corroborating Pear's di scovet:_y of a central tendency in judgment s of age from voice . The general observation that age � stimat ion can be made from voice stimul i has an explanation among other things in physiological changes caused by alteration s of the ·anatomic elements and by changes in the endocrine func­tions , e spec ially of sex hormone production .

The vocal mutation o f puberty (12-15 years) occurring in the course of en­docrinological development is closely bound up with the extraordinary growth r ate of the larynx. Accordin g to Damstee (1973) the vocal cords o f boys can increase in lenght by 1.5 times. Beg inning shortly after the age of 20 the hya line laryngeal cartilages be­come progressively oss ified (Luchsinger and Arnold, 1965). According to Kressner and to Luchs inger (cited by Endres et al . , 1971, p . 1848)'the so-called "Waldeyer's tons ilar ring" - i.e . the palatine tons i l s , the gray columns and the lymph gland s - shows a strong invol ution a fter the 35th to 40th year.' This reduction in size results in an enl argement of the supragl ottic tract. 'In the female the aging of the voice is closely correlated with the meno­paus e around a ge SO.' (Luchsinger and Arnol d, 1965, p. 135). Vocal involu­tion from male climacterium occurs at a later age . The elasticity in the tissues decreases and the mucosal epithel i um tend s to atrophy. Elderly people show atrophy of the internal vocali s muscle by which their g l ottis 'often has a bowed appearance' (Luchsinger and Arnold, 1965, p . 136), 'this means that, to achieve phonation, greater e ffort has to be ex­erted to bring the vocal cords together , and a rather harsh voice is often the result . 1 (Laver and Trudg i l l , 1979 , p . 10) In old age fatty tis sue - so-called lipoid substances - arise in the upper l arynx and around the muscul ar fibrils of the vocal cords (Inhofer, 1913; Ferreri, 1959). In extreme old age, changes in voice derive from a complex of endocrinal proce s ses . 'The mucal flui d s upply often becomes disturb ed ( • . . ) , tissues become les s elas tic and cartilages become calcified and ossi f�2d' (Laver and Trudgill, 1979, p. 10).

All these factors and more can contr ibute to the qual ity of phonation and to changes in quality of the aging voice . (O'ther e_xplanations, e . g. those con­sidering the characteristic audition. changes or socio-emotional factor s, as

- 3 -

proposed by Henton (1982) are not taken into account in thi s report. ) But thes e phenomena did not l ead to a �reat amount of p ercep ti on studies. On the contrary, almos t all res earch on the relati on b etween age and s peech has been concerned with the acoustically measurable aspects of phonation. In particular fundamental frequency (Fo) got much attenti on. With i ncreas ing.age b oth mal es and femal es show - at l east u�til adulthood -a trend of decreas e i n Fo. F rom about the age of sixty the average Fo of men tends to shift upwards , p os s ib l y due to femi nization. Viril ization in women after 5 0, due to an increas e in the tes tosterone-es trogen ratio after the menopaus e, would l ed us to expect a downward shift in Fo, but.here contra­dictory results are ob ta ined. Charlip (196 8) does not find systematic alter­ations a s sociated with age, whereas McGl one and Hol lien (1963 ) f ind a s mall i ncrease i n Fa, as i s also reported by S toicheff (1981) for non-smoki ng wom­en of ol d age. For more conclusive s tatements more comparative data mus t b e gathered with control of variables s uch as subj"ects in p re- or p os tmenopause and of s mok­ers and non-s mokers. It i s not a s i mp l e thi ng to as certai n without l ongitudinal s tudies how the female voice develop s natural l y during the menopaus e. Dams tee (1973, p . 93) s tates: there are s o many grandmothers with l ow and creaky voices that one may wel l wonder if there are s ti l l mature wome·n l eft who have never b een treated with androgenic preparations. (The i nf luence of tes tos teron with whi ch a great many women are treated is very individual .)

Other suppos ed clues of the aging voice which have been analysed acous tical­ly are the s tabil ity (or variabil ity) of pitch a nd intensity, and the rate of speaking and phonation. For an extended review s ee Hel f rich (1979).

Notwiths tanding the ove·remphas i s that is pl aced i n research on a few objec­tivel y measurab l e acous tical aspects of the voice, undefined perceptual fea­tures of voice quality and voice dynamics a re suppos ed to be the most imp or­tant clues of age. Neverthel es s very l ittle res earch is done about these perceptual f eatures . Except descriptions of the 'b reaking voice' of puberty, work on this s ubject is virtual ly l imited to the des cription of older p eo­p l e's voices .

According to Laver and Trudg i l l (1979, p. 10) 'vocal i ndications of puberty often include whi s p ery voice' and Luch s i nger and Arnold (1965 , p . 132) add that 'In addition to the l owering of the average speaki ng p i tch, the voice is f requently husky during mutation, or it may s ound weak'. The voice of old age is generally describ ed i n negative terms : monotonous, shril l, thi n, sharp , hol low , weak, hoars e, rough, breathy, tremulous trem­b l ing. It i s not clear to wh ich extent these characteristic attributes are based on stereotypes . S tereotypes , defined as a col lection of b el iefs about a group, may or may not be a val id description of real i ty, but the label s are hardly ever acou stical l y defined. And if efforts have been made to de­fine them the resul ts are not unambiguous. According to McGlone and Hol l i en (1963, p. 164) Bach, Lederer and Dinol t should have described the 'senil e' voice as 'monotonous , flat, and occas ionally shril l '. (Th i s ci tation could

I

- 4 -

not be found in the original publication.) McGlone and.Hollien (1963) tried to verify this description but come to the conclusion that 'pitch variabil­ity of speech apparently changes little with advancing age' (p. 169) . In an experiment of Ptacek and Sander (1966) listeners indicated 'that the older speakers tended to use lower pitches and less pitch variability' (p. 276). Mysak (1959) on the contrary finds that the measure of pitch variability 'reflected a general trend toward greater variability as a function of age' (p. 52). Most authors agree that the voice of old age has a smaller vocal range. But also there is a shift of the vocal range toward the lower or higher frequen­cies. Lowering of pitch range can happen through vocal mutation after the menopause and/ or' because the vocal cords become thicker, e.g. through the· occurrence of lipoid substances. A shift to higher frequencies� on the other hand, can be caused by atrophic changes within the vocal cords. But in the male also feminization after the climacterium can result in utilization of falsetto-voice. Other voice clues of old age are voca1 loudness and speakirtg rate. Diminishing vitality is demonstrated in the reduction of intensity which is perceived as a reduction in vocal loudness. And in an :investigation of Ptacek and Sander (1966) listener-judges report rate of reading as an obvious clue for differentiating aged from younger persons. Ptacek and Sander measured in the same study words per minute and found significant rate differences for both impromptu speaking and oral reading between aged and younger subjects. This in contrast to Mysak (1959, p. 52) who did not find this difference for impromptu speaking, although he did measure a reduction of time measures (words per minute; phonation/time ratio) as a function of age for reading ..

Concluding we can say that on the whole listeners are able to estimate with fair success the age of a speaker on the basis of vocal cues alone. In the impression of the listener this is done directly from the voice in all its complexity, because to the listener voice is an inextricable pattern of pitch, loudness, timbre and rhythm. About what happens in the aging larynx quite a lot is known physiologically.

·From the acoustic effects resulting from the anatomic alterations, in parti­cular pitch attributes have been analysed and, the importance accounted to fundamental frequency is indicated in quite a lot of acoustic studies (see Appendix A). But changes in voice quality and voice dynamic features are likely to carry much more information about, among other things, age. Research in this area however, has been neglected; empirical data on the perceptual aspects of the relation between age and speech are very scarce. (See Hartman and Danhauer, 1976; Ptacek and Sander, 1966; Ryan and Burk, 1974; Shipp and Hollien, 1969). Even more lacking, virtually non-existent, are studies on the relationship between objectively measured acoustic stimuli and the perceptually communi­cated percepts of the particular individual characteristics i.e. age. From the foregoing it is clear that much more research on the identification of relevant parameters on both levels must be done before their relationship can be studied systematically.

I

- 5 -

3. EXPERIMENT

The experiment consists of an evaluative rating by 29 listeners based on 2 minutes oral reading text by 50 Dutch speakers from 18 - 87 years-of-·age. On basis of voice alone speakers' ages are judged. The speakers are divided into subgroups, according to their ratings on 12 speech therapeutic terms and furthermore they are rated on 10 bipolar semantic scales.

3. J 'Rating form

The rating form exists of two parts: l) a set of semantic differential scales and 2) twelve terms from the speech therapist1s terminology The age of the speakers is assess,ed twice; before (ARl) and after (AR2) the ratings on the scales and terms are given.

1) The scales Twelve bipolar seven-point scales were selected from a battery of 35 used by Fagel, van Herpt and Boves (1983) for global perceptual ratings of speak­ers. This subset was chosen because literature suggested that the pole-terms might pertain to age. In table l the English equivalents of the Dutch scale terms and their factorloadings as resulting in the factor-solution in the study of Fagel et al. is given (If less than 10% of the variance of the scale concerned is explained by a factor, the value of that factorloading is omit­ted.) The last two columns show the communalities (h2) i.e. the percentage ' of the total variance of � variable explained by the 5 factors and, as a measure of reliability the mean correlations between eight rater groups for each scale (ra).

Table J. Selected bipolar scales (For an explanation see, text).

' English equivalents of F-1 F-2 F-3 F-4 F-5 h2 re Dutch scale terms

SOI spiritless-vivacious 0.80 0.84 0.91 S02 expressive-expressionless -0.80 -0.32 0.82 0.93 S03 colourless-sonorous 0.70 0.45 0.78 0.90 S04 tense-relaxed o'.49 0.33 -0.36 0.59 0.91 SOS polished-slovenly -0.75 0.74 0.93 S06 distinct-indistinct -o 35 -0.59 -0.39 0 .63 0.91 s 11 husky-not husky I .

0.66 0.57 0 .94 S07 dul,1-clear 0.44 0.63 0.67 0.91 S08 powerful-weak -0.37 -0.33 -0.63 0.73 0.86 809 steady-unsteady -0.36 -0.45 -0.49 0.66 0.93 S12 high- low for a (wo)man �0.67 0.46 0.92 SIO quick-slow 0.69 0.49 0.88

- 6 -

With this chosen set of rating scales we hope to be able to describe 'age characteristics' on the same dimensions as found by Fagel et al. The scales SOI up to S04 are clear representatives of the dimension 'Voice Dynamics' (F-l); so.5, S06 and S09 of I Articulation Quality I (F-2) which has to do with intelligibility and preciseness of articulati�n. S03, S07, SOS, 809 and S 11 ·of 'Voice Quality' (F-3) which is strongly associated with perceived clarity as well as with subjective strength. These three factors explain in the study of Fagel et al. 56% of the total variance. The authors also identified a Pitch-factor (F-4) and a Tempo-fac­tor (F-5) which together explain only 6% of the total variance. Nevertheless we incorporated Sl2 and 810 in our battery because both pitch and tempo seem to pertain to the perception of age.

In a pilot investigation Sll 'husky - not husky' and S12 'high - low for a (wo)man' did not differentiate and were eliminated from the battery. 'Husky­not husky' did not function, probably because the speakers did not differ enough from each other with respect to this parameter. 'High - low' is a no­toriously difficult scale, maybe because of the variability of the fundamen­tal frequency or because it is difficult to assess pitch independently from loudness or because the judges have to use different reference points for men and women.

2) Speech therapeutic terms The second part of the raeing-form gathers information in the categories coded in table 2�

Table 2. Judgment-categories

Category Code Qualifications I t--����������-�

Resonance TO l T02 T03 T04

· Phonat ion T05 T06 T07 T08 T09

Respiration TIO TI l Tl2 T13 T14

Good resonance Bad resonance Predominantly middle register Predominantly chestregister

Good phonation Bad phonation Predominantly hypokinetic Predominantly hyperkinetic Dyskinetic Good breath support Bad breath support Crams too many words in one exhalation Superficial breathing Both Tl2 and Tl3

- 7 -

Judgments are given by the raters in the form of a true/false qualification . • All raters, student13 in speech therapy in their third year, were familiar with. these terms, nevertheless an instruc tion of the following type was given: Concerning respira tion. Someone continues speaking on one exhalation when the las t word (or words) he or she utters sound strained. We.speak of shal­low or superficial breathing when a speaker inhales too often and a t 'wrong' moments in a sentence.

3.2 Speakers

Fifty Du tch male and female speakers from 18 - 87 years-of-age were distrib­u ted over 3 age ca tegories as follows:

18 - 39 years 7 female 9 m�le 40 -. 59 II 9 II 8 If

60 - 8 7 II J J II 6 11 A lower boundary of 17 years was chosen because a t that age every speaker has gone through vocal mutation.

3.3 Speech ma terial

The speech ma terial consisted of an identical text of about 2 minu tes dura­tion read aloud by the speakers. It would have been preferable to use fragments from free conversation be­cause, as mentioned in 2.0, contradic tory resul ts are ob tained conce�ning the relation between oral reading ra te and speaker's age. Bu t in this case an oral reading tex t was chosen because tt was considered more important to control for be tween-speaker differences in lexicon and syntax, as this would give information about the age of the speakers. To enable the listeners to judge brea th-control a long compound sentence was introduced into the text. Before reading the required tex t the readers had ample opportuni ty to famil­iarize themselves with this text.

3.4 Raters

The rating experiment was carried out with a group of 29 female s tudents (aged 19 - 31 years) from the Training Course of Speech Therapists in Ams terdam. A sample of 30 ra ters was planned since that would lead to an effec tive re­liability of 0.90 or higher for all bipolar scales according to a procedure described in Fagel e t al. (1983). All raters were female because i t is known that men and women tend to use somewhat different yardsticks; women seem to use the extremes of the scales in a semantic differential more readily than men. This is especially impor tant for scales like 'spiritless - vivacious' and 'dull - clear ' on which female speakers are generally rated closer to the positive extreme, and 'powerful - weak' and 'steady - unsteady' on which

I

- 8 -

male speakers are mostly r ated closer to the pos i tive extreme of the scale (Boves , Fagel and van Herp t , 1982). S t udents in their third year o f train ing had been selec ted because knowledge of terminology of sp eech sc ience was required for the second half of the e­valuation. From an unpublished analysis of the data of the exper iment by Fagel et al . (1983) is known that the rat ings on the 10 b ipo�ar seven-po int scales are not sign i f icantly influenced by this sort o f t ra in ing o f the judges. (The present experiment has.been replicated with $peech Therapy stu­dents in their first year of t ra in ing, to check this f ind ing . Data st ill have to be proc essed . )

3.5 The rat ing procedure

The f if ty fr agments are presented in one session wi th a tea-break between the vo ices of men and those of women . The voices in both parts are present ed in random order with respect to age and preceded by three t r ial-samples to fam­iliarize the list eners with the text,_ allo tted. t ime , scales and scor ing cate­gor ies. The l ist eners were g iven two minut es t o judge each speech fragment.

·The complet e session required almost 2. 5 hours.

3.6 Reliability

3.6.1 The scales

The reliab ility of a scale can be expressed in a number o f ways . For r easons of compar ison we calculated the same measure as Fagel et al . (198�) , the so­called effec t ive reliab ility a ( Rosenthal , 1973) o f each scale.

a = nf/(l+(n- l ) f ) ) ,

in which r is the mean correlation between r aters and n is the number o f raters.

Table 3. Reliab i l i ty o f rat ing scales

Nr Scales Cl

SO! sp i r i tless-v ivacious 0;964 S02 expressive-expressionless 0.969 S03 colourless-sonorous 0.952 S0 4 t ense-relaxed 0.924 SOS polished-slovenly 0.957 S06 dist inct-indis t inct 0. 959 S07 dull-clear 0.965 SOB powerful-weak 0.975 S09 ste ady-unsteady 0.965 SlO qui ck-slow 0.972

I

-- 9 -

As can be seen, ex is dependent on the. number of raters involved. To achieve a high reliability we used 29 rate•rs which resulted in an a > 0. 92 for all scales (see table 3). This fits in with Fagel et al. who report that 'for all scales an effective reliability of 0.90 or higher resulted when at least 25 rate.rs were involved� except for the scales 'loud--soft' and 'quick-slow' that require about 30 raters to reach an effective reliability of 0.90.' (p.322.)

3.6.2 The speech therapeutic terms

The reliability formula used for the scales is not suitable for answers of the falseitrue type. In order to get some insight into the measure of agree­ment between the listeners in their judgments on the therapeutic terms chi­f;qnare i::ests were used to ascertain if the distributions within the catego-T.i.es are. based on char1ce. The distribution of the observations in the categories is shown in table 4 in percentages. The probability of the distributions being attributable to chance is < 0.001 in every case, so that the conclusion may be d rm ... '11 that the ascription of the qualifications is not arbitrary.

Table 4. Distribbt ion of observations in percentages over categories. (Meaning of codes in table 2)

[cat�ory-=_codes N observations Percentages I I Resonance 1223

I I T01 (Good) 546 44.6 T02 (Bad) 677 55.4

T03 (Middle) 357 56.6

I I T04 (Chest) 274 43.4

Phonation 1375 !

I '

T05 (Good) 564 41.0 I '1'06 (Bad) 811 59.0

I T07 (Hypo) 167 20.6 I T08 (Hyper) 471 57.9

f

T09 (Dis) 175 2 l.5 I Respiration l379

TIO (Good) 525 38. 1 i TI 1 (Bad) 854 61. 9 I Tl2 (Shallow) 258 30.2 I

T13 (Superf.) 394 46.1 Tl4 (!2+13) 203 23.7

----

- 10 -

4. RESULTS

4.1 Rating I versus rating 2

Age ratings are given twice. The first age rating (ARI) is an assessment made by the listeners after hearing the speakers during a few seconds. The second assessment is made after finishing the scoring on scales and thera­peutic terms of each speaker, i.e. after hearing the speakers during two minutes. The two sets of ratings are highly correlated (r = 0.98) but although the difference between the means of ARI and AR2 is less than one year (see table 5), comparison of ARI and AR2 of the same individuals by means of a depend� ent t-test showed the differences to be systematic. In almost all cases the second rating is higher than the first one. This holds for all ages, for female as well as for male speakers, and just as well for overrating the age as for underrating. This result suggests that the parameters relevant for age assessment can be 'extracted' from speech samples of only a few seconds.

In the following paragraphs calculations concerning ARl are left out because the differences between the first rating and the second are small and system­atic.

Table 5, Mean age ratings (l1.Rl and AR2) and mean true age (AGE) of male and female spe.akers of each speaker-age group.

Female Male Total Age-group

N AR1 AR2 AGE N ARl AR2 AGE N ARI AR2 AGE

18-J9 7 26. l 26. 3 25.] 9 34.0 34.1 26.2 16 30.6 30.7 25.8 40-59 9 43.6 4f�. 2 50. 1 I 8 44. 1 44.7 48.8 17 43.9 44.4 49.5 60-87 11 63.4 64.4 77. 6 6 64.8 65.8 74.3 17 63.9 64.9 76.4

Total 27 47.2 47.8 54.8 23 45.6 46. l 46.6 50 46.4 47.0 51.1

4.2 Age ratings

First, the results show that there is a close relation between rated and real ages. The correlation between AR2 and AGE is 0.82. Nevertheless in a t-test with paired observations of all speakers the dif­ference between AGE and AR2 is significant (p < 0.001). Closer analysis revealed that the rating errors made are a function of age of the speakers. The rating errors expressed as the mean of the absolute differences between AR2 and AGE for the subsequent age-groups are 6.3, 10.0 and 12.7 years respectively. (p < 0.001). Also, as can be seen from table 5, there is a central tendency in the

I

f I l

- 1 1 -

ratings: older people are rated too young, younger people too old. However this error is not signifiaant for female speakers in the.youngest age group (18 - 31 years) which is exactly the same age group as the one we recruited the female listeners from. The corr-elation AR2 - AGE for this group is 0.97. This result suggests an interaction of the variables age and sex between listeners and speakers. The picture is further complicated by a speaker-sex-effect. In all age groups the mean rated age of men is.higher than that of women.

4.3 Age and therapeutic terms

With the aid of t-tests in the case of two judgment-categories .and with an one-way analysis of variance followed by Tukey's Honestly Significant Dif­ference Test in the case of three therapeutic subcategories we analyzed the different speech therapeutic qualifications on differences between rated and true ages. (See table 6)

Table 6. Mean age of subjects categorized according to speech therapeutic terms. (Meaning of the codes in table 2.) All differences are significant (p < 0.01) except between sub­groups indicated with o .

� Categories main �roups Categories Sub roups

1-J'..2 l AGE AR2 AGE

I RES. TOI (Good) 42.86 45.98 T03 (Middle) 45.46 50. 12 0

T02 (Bad) 49.25 53.63 T04 (Chest) 49.48 so. 140 I

I I I

Diff. -6.39 I

-1.6s Diff. -4.02 -0.02 I

T05 (Good ) l+O. 41 I (Hypo) PHON. 42.66 1T07 44.57 46.82

T06 (Bad) 5 l. 09 55.60 T08 (Hyper) 52.900 57.860 Diff. -10.68 -12.94 T09 (Dys) 52.400 57.730

RESP. TIO (Good) 42.44 45.99 Tl2 (Shallow) 50.220 55.420 Tl I (Bad) 49.42 53.03 Tl3 (Super f.) 48.4Jo 50.24

Diff. -6.98 I

-7.04 T 1 l+ (12+1 3) 50 .48o 55.540

For all main categories the 'good' speakers are significantly younger than the 'bad' speakers both on rated and on true age. In other words, listeners consider speakers to be younger when their resonance (TOl/02), respiration (TIO/II) and/or especially their phonation (TOS/06) ·is good. Their assump­tions seem to be justified since the real·AGE of the 'good' speakers is in reality also significantly lower than that of the bad speakers (Table 6). Speakers who use the chestregister (T04) are rated four years older than those who speak in the middle register (T03). In this case, however, there

I

- 12 -

SCALE NRS ,----�---,.--�----�----�...,.�--�---.--�---,�-------r-------,

6.

s. 2.

9. 4.

8. 3.

7. I •

10.

6.

s. 2. 9.

4. 8.

3. 7. I .

JO .

6. s.

2. 9. 4. 8.

3. 7.

1. 1 o.

0

Fig. 1 .

TOI (good)

T05 (good)

TIO (good)

2 3 lf SCALE VALUES

MAIN RESONANCE GROUPS

T02 (bad)

MAIN PHONATION GROUPS

T06 (bad)

MAIN RESPIRATION GROUPS

TI 1 (bad)

.5 6

Mean scores on ten bipolar scales of 'good' and 'bad' speakers on three therapeutic concepts. All scales are oriented with the positive poles to the left.

7

- 13 -

is no difference in real age between both groups. This attribution error suggests that rating a higher pitched voice younger than a low-pitched voice is a stereotype. The group with bad phonation is subdivided in three groups (T07, T08 and T09). There is no difference in real or rated age between the hyper- and dyskinetic grol\pS, but both are estimated about eight years .older than the hypokinetic group. The difference is in reality even more., namely I I years, so these concepts are apparently used in the attribution of age. Bet�een the three subgroups of speakers with bad breath-support (TI2� Tll and T14) there is no significant difference in mean age ratings although the mean real age of the 'superficial' breath group is five years ybunger. In this aspects vocal age seems not to coincide with chronological age

4.4 Scales and therapeutic terms

Relating the scales to the therapeutic terms reveals a high correlation be­tween both ratings. For all scales, except 'quick-slow' (SJO), the mean scores are significantly more on the positive pole of the scale for speak­ers rated as 'good' in Resonance or Phonation or Breathing. The profiles (see fig. 1) differentiate clearly between 1 good' and 'bad 1 groups of speak­

ers. The correspondence of the three profiles is caused by the interrela­tion of the categories Resonance, Phonation and Respiration: 12 from 50 speakers were scored as 'good', 24 speakers as 'bad' on all three categories. (For further analyses the true/false qualifications will be scaled.) The 'good' and 'bad' groups, on their turn differentiate significantly both on rated and true age, as shown in table 6. Nevertheless the correlations between the ten bipolar scales and AR2 are, although all significant at less than !%, never higher than 0.44. Those be­tween AGE and the scales are all in the same direction but on the average 0. 05 lower (Table 7)

I S ca_l_E_� s_· __ A_R

_2

, SOI S02 S03 S04 SOS S06 S07 S08 S09 810

0.204 0.193 0.267 0.296 0.210 0.271 0.371 0.248 0.439 0.170

AGE

o. 120 0. 1j2 o. 184 0.295 0.153 0.252 0.300 0.231 0.417 0.121

J

Table 7. Correlations between age and scales. All scales are oriented with the positive pole to the left. p < 0.01 for all correlations.

I

- 14 -

The significance of the scales suggests that acceptable scales were used. The values being low means that the scale values cannot be used for the pre­diction of age without some form of manipulation. Further analysis, e.g. stepwise multiple regression analysis and the delimiting of weighting fac­tors may be necessary. However, the scales can give an insight into the fac­tors that place the speaker groups in perceptual space. E.g.· the differences between two subgroups of bad phonation, hypo- (T07) and hyperkinetic (T08), are significant on scales SOI up to 804, 807 and SOS. Except on 804 the hy­perkinetic group is assessed on all scales more positively (i.e. in the di­rection of the scores for good phonation). In other words a hyperkinetic speaker is perceived as more vivacious, more expressive and more sonorous, as clearer and more powerful but as less relaxed than a hyperkinetic speak­er - but both are in all these respects less so than the speaker with good phonation (T05). In figure 2 the relative differences of these six differ­entiating bipolar scales are represented in scale units. And again, differences in ages between the three groups are in the same direction: AR2 - TOS: 40.41 T08: 44.57 T07: 52.90 years

2

Cl) 3 p:: z

7 i:x4 ....:i <t: u Cl) 8

4

•

;

•

� : : .. :

•

0

AGE - T05: 42.66 T08: 46.82 T07: 57.73 years

r I T05 I T07 ...

' T08 ' ' '• \

\ \ It ' ' '

).

·- - --

2 SCALE UNITS

Relative differences - in scale units - between hypo- (T07) and hyperkinetic (T08) speakers corn- -pared on six bipolar scales with 'good' phonating speakers (TOS). All scales are oriented with the positive poles to.the left. All differences are significant (p < 0.00 1 )

- 1 5 -

5 . 0 CONCLUDING REMARKS

Fir s t o f al l , it mus t be emphas i z e d again that neglect o f the p ercep tual at tribu t e s of - amon g o ther things - age j udgment from vo ice s t imul i , i s unwarran t e d . L i s teners appear to be ab l e to e s t ima te speake r s ' ages fairly we l l f rom vocal p arame t er s . In genera l , speake r s who s core h i gh on yoice qual i ty f e a­tures and on vo ice dynamic featur e s · are s upposed to b e younger than speak­e r s who do wors e . A suppos i t i on wh i ch appears j us t i f ied s ince there is a h i gh corr e l a t i on b e tween chrono l ogical age and i t s e s t ima t ion in thi s s tudy . None the l e s s , the pred i c t ion o f the true age o f one individual remains pre­car ious , b ecause there are s everal interfering and interac.ting var iab l e s . For ins tance , there i s a central t endency i . e . o lder peop l e are rated too young , younger peop le too o l d . But at the s ame t ime there s e ems t o be an interaction b e tween speaker s ' age and l is tener ' s age , i . e . s omeb ody ' s age i s b e s t e s t imated b y s omeone o f the s ame age . S ince the l i s teners i n t h e pres­ent s tudy were all of the young e s t age group , th is age- interac t ion phenom­enon migh t a l s o account for our f ind ing that the s i ze o f the rat ing erro r s are a f unc t ion o f age of the speaker . Moreover there can b e a l i s t ener­sp eaker s ex interac t i on , a l though it is a l s o po s s ib l e that - f o r s o c i o l o gi­cal cause s - on ly women make this rat ing-error ( e s t ima t ing men ' s age too high) . Dif f erences in rat ings may also oc cur as a resul t o f the material used (oral r eading versus impromp tu speaking ) , o f the t ime a l l otted · t o the l i s t en­e r s to g ive the i r rat ings ( s e conds ver sus minu t e s ) o r of the typ e o f l i s t en­er taking part in the exper iment ( tr ained ver sus unt raine d ) . The d i f f eren t i a t i on o f the perceptual rat ings in the p r e s ent s tudy g ive con­f i dence that th is type of de s �r i p t i ve tools are u seful in e s tab l i shing cri­teria for def ining speaker character i s t ic s . The b ip o l ar scal e s d id func t i on we l l but the relat ion t o the therap eutic te rms i s no t qui te c l ear . Th is i s caus e d by the c orre l a t ion b e tween the cat­e gor i e s Re sonance , Phona t ion and Re sp irat ion . Fur ther analys e s on the rela­t ion o f scal e s and terms wi l l be p er formed but are hampered by the fact that the therapeutic terms are mer e ly s cored as True or Fal s e . O ther r a t ing p ro­cedures for this type of termino lo gy are pref erab l e e . g . having the terms rated in sma l l clus ters of be tween 2 and 4 on 7- and 4-po int s c a l e s , as pro­p o s ed by van B e z o o ij en ( 1 9 8 4 ) , or s imp ly by us ing as a s c o re the percen tage of j udges that g ive a pos i t ive rat ing . From the bat tery of b ip o l ar scales the scal e ' h igh-low' was el iminat e d b e­caus e i t d id no t f unct ion properly ( s ee 3 . 1 ) . In further res earch a s o l ut ion for this no tor iously d i f f i cu l t scale mu s t be f ound to enabl e the descrip t ion o f speakers on the p i tch d imens ion o f p ercep tua l space . Given the amount of avai lab l e phy s ica l data concerning f undamental frequency i t i s d e s irab l e tha t sub s equent s tudies a r e conduc ted r e l a t ing acous tic' measures conne c ted with Fo b o th to age and s ex and to the percept i on of age and s ex . As ind icated already i n the l a s t s entence , in subsequent s tudies age-ind i­ces wi l l b e inves t i gated in combinat ion with o ther b i o logically d e termined marke rs in speech . Much i s s t i l l unknown about the exac t nature of the cues wh ich are used in the at t r ibut ion proce s s , but l i s t eners a s s e s s b io l o g ical character i s t ic s such as sex, age and phys ique w i th reas onab l e val i d i ty , r e l iab i l i ty and accuracy from vo cal cues as has b e en s hown , among o thers by

I

- 1 6 -

Las s e t al . ( 1 9 76 , 1 9 7 8 , 1 980) . S ince these cues have percep tual importance they should be us eful in e s tab l ishing cri ter ia f or the evaluat i on of coded speech as wel l as for def ining general s peaker-character i s t ics whi ch can b e used i n a s tepwise process o f automa tic recogn it ion .

ACKNOWLEDGEMENT

We would l ike to thank Hel een Deigh ton-van W i tsen , Wil Fage l and Lou i s Pols for read ing an earl ier vers ion o f thi� ar t icle and for their comment s on it .

REFERENCES

Al lport , G . W . ( 1 9 3 4 ) . Judging per s ona l i ty from vo ice . J . of Soc . P sych . , S , pp . 3 7-5 5 .

Bach , A. C . , Lederer , F . L . & Dino l t , R . ( 1 94 1 ) . S en i l e changes in the l aryn­geal musculature . Arch . Otolaryng . , 34 , 4 7-5 6 .

Bezooij en , R . van ( 1 9 84 ) . The aud i tory descrip t ion of long- term speech charac ter i s t ics : Some prac t ical remarks . Proceed ings 8 , p . 1 - 1 2 . Cathol ic Univers i ty , Nijme gen .

Bove s , L . , Fage l , W . P . F . & Herp t , L . W . A . van ( 1 9 8 2 ) . Opvat tingen van vrou­wen en mannen over de spraak van mannen en vrouwen . De Nieuwe Taal­gids 75 ( 1 ) , 1 -2 3 .

Charl ip ,. W . S . ( 1 96 8 ) . The aging f emale vo ice : s e l ec ted fundamen tal frequen­cy character i s t i c s and l i s teners j udgements . D i s sertat ion , Purdue Univers ity . (Dissertation Ab s tract s , 1 968 , 29 B , 1 1 9 8-B . )

Dams tee , P . H . ( 1 9 7 3 ) . De pathologi sche s tembandfunkti e . S tafleu ' s · Weten­s chapp e l ij ke Ui t g . mij b . v . , Leiden .

Endres , W . , Bambach , W . & Flo s s er , G . ( 1 9 7 1 ) . Voice spectrograms as a func­t ion of age , voice d i s guise and vo ice imi tat ion . J . Acous t . Soc . Am. , 4 9 , 1 84 2- 1 848 .

Fagel , W . P . F . , Herp t , L . W . A. van & Boves , L . ( 1 98 3 ) . Analy s i s o f the per­ceptual qual i t i e s of Dutch speaker ' s v� ice and pronunc iat ion . Speech Communicat ion 2 , 3 1 5- 3 2 6 . .

Ferrer i , G . ( 1 9 5 9 ) . Senes cence o f the larynx . I t . Gen . Rev . Oto-rhino l aryng . , 1 , 640-709 .

:Hartman , D . E. & Dannauer , J . L . ( 1 9 7 6 ) . P ercep tual feature s o f speech for males in four p erce ived age decades . J . Acous t

'. · So c . Am. , 5 9 ,

p . 7 1 3- 7 1 5 . Helfrich , H . ( 1 9 7 9 ) . Age markers in speec h . p . 6 3- 1 07 . In : K . R. S cherer &

H . Gile s , S oc ia l markers in Speech . Cambridge Univers i ty Pres s . Henton , C . G . ( 1 9 8 2 ) . Aspec t s o f sex-spec i f ic l anguage d i f f erence a t the

phonat ic and phono logical l evel s . The s i s . Phone t ic s Laboratory , Oxford .

Herzog , H . ( 1 9 33 ) . S timme und P er sonl ichke i t . Z schr . P sycho l og i e , 1 30 , 300-3 7 9 .

Inhofer , R . ( 1 9 1 3 ) . Die Ermudung der S t inune (Phonastenia) . Verlag Kur t Kab i tzsch , Wiirzburg .

- 1 7 -

Las s , N . J . , Hu ghes , K . R . , Bowyer , N . D . , Water s , L . T . & Bourne , V . T . ( 1 97 6) . S p e aker s ex ident i f i c a t ion trom vo iced and f i l tered i s o l a ted vowe l s . J . Acous t . S o c . Am . , 5 9 , 675-6 7 8 .

Las s , N . J . & Dav i s , M . ( 1 9 76) . An inve s t igat ion o f speaker h e i gh t and we i gh t ident if icat ion . J . Acou s t . S o c . Am. , 6 0 , 700- 7 0 3 .

Las s , N . J . B ev�r l y , A . S . N i co s ia , D . K . & S imp s o n , L . A . ( 1 9 7 8.) . An inves ti­gat ion o f s p eaker h e i gh t and we i gh t ident i f icat ion by means of d ir e c t e s t imat i on . J . o f Phone t ic s , 6 , 69- 76 .

Las s , N . J . , Almer ino , C . A . , Jourdan , L . F . & Wa l sh , J . M . ( 1 980) . The e f f e c t o f f i l t ered spe ech on sp eaker race and s ex ident if i c a t ion . J . o f Phone t ics , 8 , 1 0 1 - 1 1 2 .

Las s , N . J . , Brong , G . W . , C i c co l e l l a , S . A . , Walter s , S . C . &. Maxwe l l , E . L . ( 1 980) . An inve s ti ga t ion o f speaker h e i gh t and we i gh t d i s cr iminat ions

by means o f p a ir ed compar i s on j ud gmen t s . J . of Phone t i c s , 8, 205-2 1 2 . Las s , N . J . , Hendr icks , C . A . & I turr iaga , M . A . ( 1 980) . The cons i s tency o f

l i s t ener j ud gmen t s i n s p e aker height and we i ght ident i f i c a t ion. J . o f Phone t i cs , 8 , 439-44 8 .

Laver , J . & Trud g i l l , P . ( 1 97 9 ) . Phon e t i c and l ingu i s t ic markers in speech . 1 -3 2 . In : K . R . S cherer & H . Gil e s , S o c i a l marker s in Speech . Cambrid ge Univer s i ty P r e s s .

Luchs inger , R . & Arno l d , G . E . ( 1 959) . Lehrbuch d er S t imm- und Sprachheil­kund e . Spr inger , Berl in .

Luchs inger , R . & Arno l d , G . E . ( 1 96 5 ) . Vo ice-Speech-Language . C l inical Com­mun i co logy : I t s Phy s i o l o gy and Patho l o gy . Wadswo r th Pub l i sh ing Co . , Inc . Belmont , Ca'l i forni a .

McGlone , R . E . & Hol l ie n , H . ( 1 96 3 ) . Vo cal p i tch charac ter i s t ic s o f a ged women . J . Speech Hearing Re s e ar ch , 6 , 1 64-7 0 .

My s ak , E . D . ( 1 95 9 ) . P i t ch and dura t i on charac t er i s t i c s o f o lder mal e s . J . Speech Hear ing Res earch , 2 , 46-54 .

Pear , T . H . ( 1 9 3 1 ) . Vo ice and Personal i ty . London : Chapman & Hal l . P tacek , P . H . & Sander , E . K . ( 1 9 6 6 ) . Age recogn i t ion f rom vo ice . J . Speech

Hear ing Research 9 : 2 7 3-27 7 . Ro sentha l , R . ( 1 97 3 ) . E s t imat ing e f f e c t ive re l iab i l i t i e s in s tud i e s tha t em­

p l o y j ud ge s ' r a t ing s . J . C l inical P s ycho l . , Vo l . 2 9 , 342-345 . Ryan , W . J . & Burk , K . W . ( 1 9 74) . Perceptual and acous t i c corr e l a t e s of aging

in the speech of mal e s . J. Comm . , 7 , 1 8 1 - 1 92 . S anf ord , F . H . ( 1 942) . Speech and per sona l i t y . P sy cho l og i c a l bu l l e t i n , 39 ,

8 1 1 -845 . Shipp , F . T . & Ho l l i en , H . ( 1 9 69 ) . Percep t ion o f the a g ing ma l e voice . J .

Speech Hearing Res earch , 1 2 , 703- 1 0 . S t ern , H . ( 1 9 32) . Ueber e in i ge Ursachen and F o l gen der Al tersersche inungen

in der S t immgebung . Mschr . Ohrenhe i l kund e , 66 , p . 665 . S t o iche f f , M . L . ( 1 98 1 ) . Speak ing fundamental f r equency charac t er i s t i c s o f

nonsmoking fema l e adul t s . J . Speech Hear ing Res ea r ch , 24 , 437-44 1 .

- 1 8 -

RE f . NR 20 rr.:_--( �r!)---.---.....--....--..--..----.---...---,-�-.-�-.----..,--,..---.-�-�--r--.....-�

28 ------ ( 1 26) l 2 • { I S) 2 1 • • •

9 • ( I S ) 3 5 • ( 5 )

4 ( 2 4 ) 1 7 2 6 l 5

•

- ( 34 1 )

- (467)

FEMALE SUBJECTS

7 30 1 8 23

34 3 3

--- ( 1 00) ( 66)

(37) ---- (27 )

--- ( 1 1 )

5 32

8 2

25 1 9

( J 1 1 ) ---- ( 1 6 )

( 2 ) -------- ( 1 8) --------- (30)

�--�----------------------------� ( 25)

0 I 0 20 30 40 50 6 0 7 0 80 YEARS-OF-AGE

Fig . I . Age group s o f f emal e subj e c t s in s tudies on fundamental frequency .

20

28

36 2 2 3 5 1 4 3

24 1 0 ! : 1 3 3 1 i 6 29 27

5

Age in year s i s indicated on the hor i zontal axi s . B e tween brackets the number o f s ubj ects included i n each s tudy . The numbers along the ver t ical ax i s refer to t h e re ferences given in the appendix .

- { 1 4 )

( 1 26)

• ( ! 5 ) • ( 1 5 )

• ( 1 5) --- ( 1 2) --------- (24)

(260) - (27)

- ( 1 6) • ( 8 )

( 1 5 7 )

( 6 )

( 2 4 )

MALE SUBJECTS

�-----------------------------� (39 ) �---------------------------- (4)

---- (5 )

0 I 0 20 3 0 40 50 60 70 80 YEARS-OF-AGE

F i g . II . S imilar to f i g . I - Mal e subj e c t s .

I

::- 1 9 ...,

APPENDIX A

In f igure I and I I a summary i s g iven o f acou s t ic s tud i e s o f vocal p itch character is t ic s o f mal e and f emal e s ubj e c t s o f d if f erent ages . The hori z on­t a l l ine s in both f igur e s repres ent the age ranges o f the subj e c t s u s ed in the s tud ie s which are ind i c a ted by the ref erence number s a l o'ng the ver t i ­c a l axi s . W e d o no t r e p o r t actual data o f Fo measur ement s because the s tud­i e s d if f e r in var ious r e levant var iabl e s . E . g . in s everal c a s e s i t i s not c l ear what k ind of s p eech material i s u s ed ( oral read ing o r impromp tu s p e ak­ing) . Var i ab l e s that are contro l l ed only in a f ew inve s t i ga t ions are ( a­mount o f ) smoking o f c i gare t t e s , o r ( amount o f ) s t re s s . Concerning the sub­j ec t s it i s no t a lways c l ear if the mea sure s are d e termined pre- or p o s t­menopau s e / c l imac ter ium, or i f the women d id o r d id n o t u s e the p i l l , and with young subj ec t s if they are mal e or femal e .

Notwi ths t and ing the f a c t that this i s but a l imi ted summing up o f var i ab l e s invo lved and even though there are many s tud i e s o n the s ubj e c t , i t w i l l b e c l e ar that ther e i s a great need o f l on g i tud inal s tudie s , preferab l y execu­ted cro s s-cu l tural ly

REFERENCES OF APPENDIX A

l . Bach , A . C . , Lederer , F . L . & Dino l t , R . ( 1 9 4 1 ) . Seni l e changes in the l aryngeal mus culature . Arch . Otol aryngo l o gy , 3 4 , 4 7-56 .

2 . Char l ip , W . S . ( 1 968) . The ag ing f ema l e vo ice : s e l e c t ed fundamental f re­quency charac ter i s t ic s and l is t ener j ud gements . Ph . D . Di s s . , Purdue Univer s i ty .

3 . Curry , E . T . ( 1 93 9 ) . An obj e c t ive s t udy o f the p i t ch charac ter i s t i c s o f the ado l e s cent mal e vo ice . Ph . D . D i s s . , S tate Univer s ity o f Iowa . (Shortened ver s ion in : S p e e ch Monographs , 1 94 0 , 7 , 48-62 : The p it ch character i s t i c s of the ado l es c ent mal e v o i c e . )

4 . Duff y , R . J . ( 1 95 8 ) . The vocal p i tch charac t er i s t i c s o f e l even- , thir­t een- , and f i f t een-year-old f emal e speakers . P h . D . Di s s . , Univ . o f Iowa . ( Short ened vers ion in : Language and Spe ech , 1 97 0 , 1 3 , 1 4-24 : Fun­damental f requency character i s t i c s o f adol e scent f ema l e s . ) Note : The unpub l i shed data for the mean fundamenta l frequenci e s o f

the 1 1 , 1 3 and 1 5 y ear-o lds o f Duf f y ' s s tudy o f 1 958 contain a s l i ght c a l cu l at ion error . Tab l e I , p . 1 6 of the a r t i c l e o f Duff y bas ed o n this d i s s er t a t ion contains the c o rr e c t f ig­ure s .

5 . Endres , W . , B amb ach , W . & F l o s ser , G. ( 1 97 1 ) . Voice s p e ctrograms as a funct ion o f age , vo ice d isgui s e and v o i c e imi t a t ion . J . Acous t . Soc . Am. , 4 9 , 1 8 4 2- 1 848 .

6 . Fairbanks , G. ( 1 94 2 ) . An acous t i c a l s tudy o f the p it ch o f infant hunger wai l s . Ch ild Devel opment , 1 3 , 227-2 3 2 .

7 . F i t ch , J . L . & Holbrook , A . ( 1 9 7 0 ) . A mod a l vocal fundament a l f requency o f young adu l t s . Arch . Otolaryngol ogy , 9 2 , 3 7 9-382 .

I

- 2 0 -

8 . G i lbert , H . R . & We i smer , G . G . ( 1 97 4 ) . The e f f ec t s o f smoking on the speaking fundamental frequency of adu l t women . J . P sy chol inguis­t ic Res earch , 3 , 2 25-2 3 1 .

9 . Hammond , J . M . ( 1 9 4 7 ) . An obj e c t ive s tudy o f the p i tch charac t er i s t i c s o f e i ght year- o l d gir l s dur ing o r a l read ing . M . A . The s i s , S tate Unive.r s i ty o f I owa . No t e : Data from this the s i s are used in: Fairbanks , G . , Herber t ,

E . L . & Hammond , J . M� ( 1 9 4 9 ) . An acous t ic s tudy o f vocal p i tch in s even- and e i ght-year-old gir l s . Child Develop­ment , 20 , 7 1 -7 8 . S ee a l s o : Fairbanks , G . ( 1 9 50 ) . An acous t i c comp ar i son o f voc a l p itch in s even- and e i ght-year-old children . Ch ild Development , 2 1 , 1 2 1 - 1 29 .

1 0 . Hanl ey , T . D . ( 1 94 9 ) . An ana l y s i s o f vo cal frequency and durat ion charac­t er is t i c s o f s e l e c t ed s amp l e s o f s peech from general Amer i c an , e a s tern Ame r i c an and s outhern Ameri c an d ia l e c t r egions . Ph . D . D i s s . , Univ er s i ty o f Iowa . Art ic l e based on this d i s s er t a t ion : An analys i s o f vocal frequen­c y and dura t ion character i s t ic s o f s e l e cted d ia l e c t region s . Speech Monographs , 1 95 1 , 1 8 , 7 8-9 3 .

1 1 . Hanl e y , T . D . , Snidecor , J . C . & Ringe l , R . L . ( 1 96 6 ) . S ome acous t i c d if­f erences among languages . Phonet i c a , 1 4 , 9 7- 1 07 .

1 2 . Herb er t , E . L . ( 1 94 2 ) . An obj e c t ive s tudy o f the p i tch charac ter i s t ic s o f s even-year-o l d g i r l s dur ing oral read ing . M . A . The s i s , S tate Univer s ity of I owa . S e e note under 9 .

1 3 . Hol l i en , H . & Jackson , B . ( 1 9 7 3 ) . Normat ive data on the speaking funda­mental frequency charac ter i s t i c s o f young adu l t male s . J . of Pho­net ic s , I , 1 7 7-200 .

1 4 . Ho l l ien , H . & Mal c i k , E . ( 1 96 7 ) . Eval uat ion o f c r o s s-s e c t ional s tud ie s o f ado le s cent voice change in mal e s . Speech Mono graphs , 34 , 80-84 .

1 5 . Hol l i en , H . & P aul , P . ( 1 9 6 9 ) . A s econd evaluat ion o f the speaking fun­d amental f requency charac ter i s t i c s o f p o s t-ad o l e scent g ir l s . Language and Spe ech , 1 2 , 1 1 9- 1 24 .

1 6 . H o l l ien , H . & Sh ipp , F . T . ( 1 9 7 2 ) . Speak ing fundamental frequency and chrono l o gical age in male s . J . Speech Hear . Res . , 1 5 , 1 55- 1 59 .

1 7 . Ho l l ien , H . , Mal c i k , E . & H o l l i en , B . ( 1 96 5 ) . Ado l e s c ent voice changes in southern whi t e mal e s . S p eech Mono graphs , 3 2 , 8 7-90 .

1 8 . Ho l l ie n , H . , To lhur s t , G . C . & McGlone , R . E . ( 1 9 8 2 ) . Speak ing fundamen­t a l f requency as a func t ion o f vocal intens i ty . Unpub l i shed r e­port t o the Dreyfus Found a t ion . ( C i t ed by O . d e P into and H . Hol l ien , J . Phone t ic s , 1 98 2 , 1 0 , p . 369 . )

1 9 . Honj o , I . & Is shiki , N . ( 1 980) . Laryng o s cop i c and vp ice characte r i s t ic s o f aged p e r s ons . Arc h . O t o laryngol o gy , 1 980 , 1 06 , 1 49 - 1 50 .

2 0 . Keat ing , P . & Buhr , R . ( 1 97 8 ) . Fundamental frequency in the speech o f infants and children . J . Acou s t . S o c . Am. , 6 3 , 5 6 7-5 7 1 .

I

- 2 1 -

· 2 1 . C ited in s everal art ic l e s with d i ff erent t i tle s : Kr ike l as , H .

1 . An analys i s of the p i tch character i s t ic s of seven-year-o ld girl s dur ing oral read ing . M . A . The si s , Stanford Univer s i ty , 1 95 9 .

2 . An analys i s o f the fundamental frequency charac·t er is t ics o f 1 4-and 1 6-year-o l d f emale speakers . M . A . The s i s � Stanford Univer­s i ty .

22 . Las sman , F . M . ( 1 94 2 ) . An obj ective s tudy o f the p i tch character i s t ics of e igh t-year-ol d boys during oral read ing . M.A. The s i s , S tate Unive r s ity o f Iowa . Not e : Data from th i s the s i s are used in : Fa irbanks , G . , Wiley ,

J . M . & Las sman , F . M . ( 1 94 9 ) . An acou s t ical s tudy o f the vocal p i tch in s even and e ight-year-old boy s . Chi l d Devel­opment , 20 , 6 3- 6 9 . S e e a l s o : Fairbank s , G . ( 1 950) . An acous t ic compari s on of vocal p itch in s even- and e i ght-year-old chi ldren . Ch i l d Devel opment , 2 1 , 1 2 1 - 1 2 9 .

2 3 . L inke , E . ( 1 9 7 3 ) . A s tudy o f p i tch charac ter i s t ic s o f fema l e vo ices and the ir r e l a t ionship to vocal e f f ect ivene s s . Ph . D . Di s s . , Univers i ty o f Iowa . Ar t i c l e ba s ed on th i s d i s sertat ion and under the same t i t l e i s publi shed in Fo l i a Phoniatr ica , 1 9 5 3 , 25 , 1 7 3- 1 85 .

24 . Maj ewski , W . , Ho l l ie n , H . & Zalewski , J . ( 1 9 7 2 . Speaking fundamental frequency o f Po l i sh adu l t mal e s . Phone t ica , 25 , 1 1 9 - 1 25 .

2 5 . McGlone , R . E . & Ho l l ien , H . ( 1 9 6 3) . Vocal p itch chara cter i s t ic s of aged women . J . Speech Hear . Res . , 6 , 1 64- 1 7 0 .

2 6 . Michel , J . F . , Hol l ien , H . & Moore , P . ( 1 9 6 6 ) . Speaking fundamental fre­quency character i s t ic s o f 1 5- , 1 6- and 1 7 -year-old girl s . Language and Speech , 9 , 4 6- S l .

2 7 . Mys ak , E . D . ( 1 95 9 ) . P i t ch and durat ion charac ter i s t i c s o f older mal e s . J . Speech Hear . Re s . , 2 , 4 6-54 .

2 8 . Oordt , H . W . A . van & Dro s t , H . A . ( 1 9 6 3 ) . Devel opment o f the f requency range of the voice in ch i l dr en . Fol ia Phoniatr ica , 1 5 , 289-29 8 .

2 9 . Philhour , C .W . Jr . ( 1 948) . An exper imental s tudy o f the relat ionship s between p ercept ion of vocal p itch in conne c t ed sp eech and cer ta in me asure s of vocal f re quency . Ph . D . Di s s . , Univers i ty of Iowa . ( C i t ed a . o . in 4 , 2 3 , 24 and 2 5 . )

3 0 . P into , 0 . de & Hol l ien , H . ( 1 9 8 2 ) . Speaking fundamental frequency char­acter i s t i c s of Aus tral ian women : then and now . J . of Phone t i c s , J O , 367-37 5 .

3 1 . Pronovost , W . ( 1 9 4 2) . An exp e r imenta l s tudy o f methods for determining natural and hab i tual p it ch . Speech Mono graph s , 9 , 1 1 1 - 1 2 3 .

3 2 . Saxman , J . H . & Burk , K . W . ( 1 96 7 ) . Speak ing fundamental f requency charac­teristics of midd l e-aged f ema l e s . Folia Phoniatrica, 1 9 , 1 67- 1 7 2 .

3 3 . Snidecor , J . C . ( 1 940) . S tud i e s in the p itch and duration charac ter i s tics o f super ior speaker s . Ph . D . Di s s . , Iowa C i ty . Art ic l e based on th i s d i s ser tat ion . J . Spe ech Hear . Res . , 1 95 1 , 1 6 , 44-52 .

34 . Sto ichef f , M . L . ( 1 98 1 ) . Speaking fundamental · frequency character i s t ic s o f nonsmoking f ema l e a d l t s . J . Speech Hear . Res . , 24 , 437-44 1 .

I

- 2 2 -

35 . Weenink , D . J . M . ( 1 985) . Fundamental frequency of nine-year-old speakers. Unpub l i shed dat a , IFA University of Ams terdam.

36 . Wiley , J . E . ( 1 9 4 2 ) . An obj ect ive s tudy o f pitch character i s tics of seven­year-old boys dur ing oral read ing. M. A . The s i s . S tate Univers ity of Iowa. See no te under 2 2 .