The Evolution of Human Speech: The Role of Enhanced ... Evolution of Human Speech: The Role of Enhanced Breathing Control ANN M. MACLARNON* AND GWEN P. HEWITT School of Life Sciences,

The Evolution of Human Speech: The Role of EnhancedBreathing Control

ANN M. MACLARNON* AND GWEN P. HEWITTSchool of Life Sciences, Roehampton Institute London, London, SW15 3SN UK

KEY WORDS vertebral canal; spinal cord; primate vocalizations;intercostal and abdominal muscles

ABSTRACT Many cognitive and physical features must have undergonechange for the evolution of fully modern human language. One neglectedaspect is the evolution of increased breathing control.

Evidence presented herein shows that modern humans and Neanderthalshave an expanded thoracic vertebral canal compared with australopithecinesand Homo ergaster, who had canals of the same relative size as extantnonhuman primates. Based on previously published analyses, these resultsdemonstrate that there was an increase in thoracic innervation duringhuman evolution. Possible explanations for this increase include posturalcontrol for bipedalism, increased difficulty of parturition, respiration forendurance running, an aquatic phase, and choking avoidance. These can allbe ruled out, either because of their evolutionary timing, or because they areinsufficiently demanding neurologically. The remaining possible functionalcause is increased control of breathing for speech.

The main muscles involved in the fine control of human speech breathingare the intercostals and a set of abdominal muscles which are all thoracicallyinnervated. Modifications to quiet breathing are essential for modern humanspeech, enabling the production of long phrases on single expirations punctu-ated with quick inspirations at meaningful linguistic breaks. Other linguisti-cally important features affected by variation in subglottal air pressureinclude emphasis of particular sound units, and control of pitch and intona-tion. Subtle, complex muscle movements, integrated with cognitive factors,are involved. The vocalizations of nonhuman primates involve markedly lessrespiratory control.

Without sophisticated breath control, early hominids would only have beencapable of short, unmodulated utterances, like those of extant nonhumanprimates. Fine respiratory control, a necessary component for fully modernlanguage, evolved sometime between 1.6 Mya and 100,000 ya. Am J PhysAnthropol 109:341–363, 1999. r 1999 Wiley-Liss, Inc.

There has been considerable recent de-bate over the timing and factors involved inthe evolution of human language, includingits physical production, i.e., speech. Manytypes of evidence have been called into play,including brain size and form (e.g., Falk,1980; Holloway, 1983; Tobias, 1987; Aielloand Dunbar, 1993), comparative languagedevelopment (e.g., Savage-Rumbaugh et al.,

1993), linguistic analysis (e.g., Bickerton,1990; Pinker and Bloom, 1990; Wilkins andWakefield, 1995), archaeological evidence of

Grant sponsor: Leakey Trust; Grant sponsor: Boise Fund.*Correspondence to: Dr. Ann MacLarnon, School of Life Sci-

ences, Roehampton Institute London, West Hill, London SW153SN, UK.

Received 1 June 1998; accepted 2 April 1999.

AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 109:341–363 (1999)

r 1999 WILEY-LISS, INC.

tool production and cultural developments(e.g., Noble and Davidson, 1991; Schick andToth, 1993), and skeletal features such asbasicranial flexion (e.g., Lieberman and Cre-lin, 1971; Laitman, 1985), hyoid structure(e.g., Arensburg et al., 1990), and the size ofthe hypoglossal canal (Kay et al., 1998).

Some researchers support the theory thathuman language may have begun to evolvevery early in human evolution, in the austra-lopithecines (Holloway, 1983), whereas oth-ers support a somewhat later start in earlyHomo or Homo erectus (Falk, 1980; Aielloand Dunbar, 1993). However, even if someaspects of human language did appear aslong ago as 2–4 million years ago, it mayinitially have been a very limited ability.Full human language may then have evolvedgradually over a long period of time, al-though other researchers believe it largelyemerged very recently in a sudden burst ofdevelopment at the time of the culturalexplosion of the Upper Palaeolithic, 40,000years or so ago (e.g., Isaac, 1976; Noble andDavidson, 1991). There are also differencesof opinion about the forerunners of fullhuman language. It may have evolved fromthe vocalizations of our primate ancestors(e.g., Steklis and Raleigh, 1979; Steklis,1985; Bradshaw, 1988; Dunbar, 1993; Hauseret al., 1993; Fitch and Hauser, 1995), or itmay have had a more independent origin,such as gestural communication (e.g., Hewes,1973, 1992; Calvin, 1992; Armstrong et al.,1994). Some form of simple protolanguage,without the syntactical complexities of fullhuman language, may have developed first(Bickerton, 1990). However, whatever theevolutionary route and timing of the evolu-tion of modern human language, a range ofcognitive and physical features must haveundergone change from the nonhuman pri-mate condition, and these are unlikely all tohave evolved at the same time. And, al-though some changes may have enhancedlanguage abilities to a much greater extentthan others, individual component changescould very well have provided adaptationaladvantage.

One apparently necessary feature for theproduction of fully modern human speechthat has more or less been ignored in thelanguage debate is the fine control of breath-

ing, and hence of the subglottic air pressurethat fuels sound production and effects someof its intricate variations. On the basis ofwork on the thoracic vertebral canal of thefossil hominid, KNM-WT 15000, it was pre-viously suggested that this fine control wasnot present in Homo ergaster (or early Homoerectus) (MacLarnon, 1993; Walker, 1993).Here, variation in the size of the thoracicvertebral canal is investigated throughoutthe hominid fossil record, and the sugges-tion that the evolution of fully modern hu-man speech required the development offiner breathing control is explored.

METHODSThoracic vertebral canal data

Measurements. The dimensions of thethoracic vertebral canal analyzed here arebased on two measurements of the canaltaken on each thoracic vertebra of the sam-ple specimens. These measurements are:the narrowest height (dorso-ventral diam-eter) of the vertebral canal through eachvertebra (vertebral measurement no. 10;Martin, 1928), and the narrowest width(transverse diameter) of the vertebral canalthrough each vertebra (vertebral measure-ment no. 11; Martin, 1928). From these twomeasurements, the narrowest cross-sec-tional area of the vertebral canal througheach vertebra was calculated using the el-lipse formula (area 5 P/4 3 height 3 width).This formula produces a reasonable approxi-mation of the actual cross-sectional area ofthe vertebral canal (MacLarnon, 1987),though removal of the constant P/4 wouldmake no difference to any of the conclusionsreached here.

The thoracic vertebral canal has fairlyconstant dimensions along its total length inmodern primates (MacLarnon, 1987, 1993),but for ease of comparison the smallestmeasurements from the region for each speci-men were used. These are the minimumdimensions of the canal between the cervicaland lumbar enlargements, where the spinalcord and its bony encasement swell to pro-vide innervation for the fore- and hindlimbs.The minimum cross-sectional area is thesmallest area through an individual verte-bra. As the minimum canal height and widthdo not necessarily occur in the same verte-

342 A.M. MACLARNON AND G.P. HEWITT

bra, the minimum cross-sectional area maybe larger than would be calculated by com-bining the two individual linear measure-ments in the ellipse formula.

Sample. Measurements of the dimensionsof the vertebral canal through the thoracicvertebrae of adult specimens of extant pri-mate species collected for two previous stud-ies (MacLarnon, 1987, 1993) are used here.The original sample (MacLarnon, 1987) com-prised 44 specimens from 37 nonhumanspecies (5 apes, 10 Old World monkeys, 9New World monkeys, 6 lorises, and 7 le-murs), and this sample was used to calculatebaseline scaling relationships (n 5 40, treat-ing males and females of the same speciesseparately). In the second study (Mac-Larnon, 1993), additional ape specimenswere measured giving a total of 2 gibbons, 2orangutans, 4 chimpanzees, and 4 gorillas.Seven adult humans from the Spitalfields’collection at the Natural History Museum,London, were measured (canal height andwidth), and published data for humans weretaken from Haworth and Keillor (1962) (ca-nal width) and Hinck et al. (1966) (canalwidth). Species and sex averages are pre-sented in Table 1.

Measurements of the vertebral canal werealso taken for all available thoracic verte-brae for a series of fossil hominids: Australo-pithecus afarensis (AL 288-1, cast), A. africa-nus (Sts 14, Stw 431), Homo ergaster (orearly H. erectus) (KNM-WT 15000; Mac-Larnon, 1993), Neanderthals (La Chapelle,Shanidar 2 and Shanidar 3 (Trinkaus, per-sonal communication); Kebara 2), and earlymodern Homo sapiens (Skhul 4). This fossilhominid sample comprises most of the knownfossil hominids with sufficiently completevertebral columns for the necessary mea-surements to be taken. For some of thesespecimens, the thoracic vertebral column iscomplete enough to be sure that the mini-mum heights and widths measured on theavailable thoracic vertebrae and calculatedminimum cross-sectional areas are good esti-mates of the actual minima for the region.For other specimens, the minimum measure-ments taken could be a little larger than theactual minima (see Table 2).

Body weights. Body weights of the actualspecimens were used in analyses for some ofthe extant primate species. Remaining bodyweights for these species were taken from alarge collection based on numerous pub-lished and unpublished sources, using datafrom wild-caught specimens wherever pos-sible (Martin and MacLarnon, unpublisheddata) (see Table 1).

Body weight estimates for the appropriatesex of the sample fossil hominid species weretaken from published sources, as indicatedin Table 2. Adult body weight estimateswere used for the juvenile specimenKNM-WT 15000. Analyses based on modernhuman skeletons of known age indicate thatthoracic vertebral canal dimensions in Homosapiens reach adult size by 10 years, orabout halfway through the growth period,and that they only increase by about 20%between first fusion at 2 years and adult-hood (MacLarnon, 1993). The thoracic canaldimensions of KNM-WT 15000, which isestimated to have been at an equivalentstage of development to an 11-year-old hu-man (Smith, 1993), can therefore be treatedas adult, assuming that the growth patternof the vertebral canal in Homo ergaster (orearly Homo erectus) was similar to that ofmodern humans (MacLarnon, 1993).

Statistics. Best-fit lines were calculatedfor bivariate plots as principal major axes.Although debate continues over the mostappropriate line-fitting technique for describ-ing allometric relationships (e.g., Martinand Barbour, 1988; Pagel and Harvey, 1988;Aiello, 1992), where correlation coefficientsare high there is very little difference be-tween the results of the major methods(Aiello, 1992).

RESULTS

Figures 1–3, parts a, are log-log plots ofeach of the three minimum thoracic canaldimensions vs. body weight for the wholeextant primate sample, plus the fossilhominids. Figures 1–3, parts b, are enlargedversions of the upper right portions of theseplots. The lines drawn in describe the rela-tionships between each of the canal dimen-sions and body weight for the extant pri-mates less humans. The statistics for the

343ROLE OF BREATHING CONTROL IN EVOLUTION OF SPEECH

TABLE 1. Minimum thoracic canal dimensions and body weights for extant primate species

Species Sex

Bodyweight

(g)

n for canalmeasure-

ments

Minimum thoracic vertebral canal measurements (ranges)

Height (mm) Width (mm) csa (mm2)

Original sampleNycticebus coucang m 725 1 2.70 3.92 8.38Perodicticus potto f 935 1 3.72 4.77 14.10Arctocebus calab-

arensis f 212.4 1 2.20 3.36 6.39Euoticus elegantulus m 291 1 2.28 2.65 5.09Galago alleni f 262 1 2.31 2.79 5.35Otolemur crassicau-

datus m 1,165 1 2.91 3.79 9.11Otolemur crassicau-

datus u 1,204 1 3.03 3.26 7.99Microcebus murinus u 66.5 1 1.58 2.10 2.89Lemur catta m 2,350 1 4.41 4.89 17.44Eulemur fulvus u 2,201 1 4.19 4.72 15.53Varecia variegatus u 3,100 1 4.91 5.29 20.82Daubentonia madagas-

cariensis u 2,800 1 3.76 5.11 15.77Indri indri f 6,250 1 6.42 6.86 38.58Propithecus diadema u 6,500 1 5.70 6.00 27.80Callithrix pencillata f 287 1 2.40 2.76 5.35Saguinus oedipus m 325 1 2.32 3.05 5.56Cebus apella u 2,741 1 4.18 5.46 18.26Saimiri sciureus m 805 1 2.91 4.15 9.94Aotus trivirgatus f 1,200 1 3.19 3.91 10.04Cacajao rubicundus m 3,750 1 4.25 6.27 21.43Lagothrix lagotricha m 8,160 1 4.37 6.49 23.68Ateles paniscus u 8,804 1 5.51 6.98 31.63Alouatta seniculus f 5,807 1 4.51 5.92 21.48Colobus satanas m 10,683 1 5.83 6.97 33.01Trachypithecus

obscurus f 6,360 1 5.75 6.69 33.75Cercopithecus aethiops m 4,878 1 5.63 7.85 35.41Cercopithecus mitis f 4,280 3 5.13 (5.17–5.39) 6.61 28.82Cercopithecus mitis m 7,374 3 5.22 (5.00–5.40) 7.44 32.59Papio anubis m 21,920 1 7.43 12.59 76.04Papio cynocephalus m 21,728 1 7.90 11.79 76.32Cercocebus atys f 7,800 1 5.90 7.82 38.47Macaca fascicularis f 3,614 1 4.76 6.63 26.24Macaca fuscata f 9,100 1 5.94 8.15 38.40Hylobates concolor f 5,749 1 5.40 7.13 30.24Hylobates hoolock u 6,700 1 6.34 6.32 32.42Miopithecus talapoin m 1,460 1 4.07 5.46 18.00Pongo pygmaeus m 73,388 1 9.33 13.28 103.68Pan troglodytes f 34,135 1 10.46 12.12 108.99Gorilla gorilla f 82,500 1 15.39 13.69 167.17Gorilla gorilla m 152,600 1 18.15 15.26 229.37

Complete ape sampleHylobates concolor f 5,749 1 5.40 7.13 30.24Hylobates hoolock u 6,700 1 6.34 6.32 32.42Pongo pygmaeus m 73,388 2 10.72 (10.46–12.10) 11.99 (10.70–13.28) 107.49 (108.99–111.29)Pan troglodytes f 34,135 2 9.93 (9.40–10.46) 10.96 (9.80–12.12) 92.60 (76.20–108.99)Pan troglodytes m 34,135 2 9.05 (8.80–9.30) 12.30 (12.00–12.60) 88.45 (87.78–89.11)Gorilla gorilla f 82,500 2 13.35 (11.30–15.39) 13.35 (13.00–13.69) 141.27 (115.37–167.17)Gorilla gorilla m 152,599 2 15.48 (12.80–18.15) 15.73 (15.26–16.20) 196.62 (163.87–229.37)

Human dataSpitalfields (MacLar-

non, 1993) f & m 60,000 7 14.4 (12.3–16.5) 15.9 (13.9–17.9) 207.3 (159.4–263.6)Americans (Haworth

and Keillor, 1962) f & m 60,000 100 17.60White Americans

(Hinck et al., 1966) f & m 60,000 121 17.4 (13.9–20.8)

csa, cross-sectional area; u, unknown.


best-fit lines describing the scaling of theminimum thoracic canal dimensions to bodyweight are presented in Table 3. In eachcase, these were calculated for three samples:the original sample of extant primates lesshumans; the original sample of extant pri-mates less hominoids; and the completesample of apes, including more recentlycollected data (see Table 1).

The samples of nonhominoids and apeswere used to test whether these two groupsform separate grades for the scaling of any ofthe three canal dimensions. In all threecases, the scaling statistics were very simi-lar for the two groups (see Table 3). Inaddition, the equations for lines of the aver-age slope values for the two groups, passingthrough their combined mean values for xand y, were calculated, for each of the threecanal dimensions. Residuals for each speciesfrom these lines were calculated. In all threecases, the results of t-tests to compare theresiduals of the two groups of species showedthat they were insignificantly different (ca-nal height, t 5 0.012; canal width, t 5 0.859;canal cross-sectional area (csa), t 5 0.469;df 5 39, p . 0.05 in all cases). These resultsshow that nonhominoids and apes do notform separate grades for the scaling of any ofthe minimum thoracic canal dimensions.

From the statistics presented in Table 3and from Figures 1–3, parts a, it can be seenthat the minimum thoracic canal dimen-sions are all highly correlated with bodyweight across the nonhuman primates (rvalues 5 0.97–0.98). In Figures 1–3, parts b,the relative positions of the fossil hominidscan be seen more clearly.

In Figure 1, most of the fossil hominidsfall within the narrow range of deviation ofextant primates from the best-fit line. How-ever, the earlier hominids, the australopithe-cines and Homo ergaster, mostly fall at thelower end of this range. Early modern hu-mans and modern humans fall right at theupper edge of the modern nonhuman rangeof deviation from the line, as does Australo-pithecus afarensis, although in this last case,the thoracic canal height plotted is not defi-nitely the minimum for the thoracic region(see Table 2). However, overall there is noclear evidence of any difference between therelative minimum thoracic canal heights ofany of the hominids and the nonhumanprimates, nor of major change in the relativecanal height during human evolution.

In Figure 2, the australopithecines andHomo ergaster all fall close to or below thebest-fit line, within the lower end of therange of deviation of the nonhuman pri-

TABLE 2. Minimum thoracic canal dimensions and body weight estimates used for fossil hominid sample

Specimens Species

Minimum dimensions of the thoracicvertebral canal (vertebra no.) Body weight

estimates(kg)

References forbody weightestimates

Height(mm)

Width(mm)

Cross-sectionalarea (mm2)

Al 288-1 (cast) Australopithecusafarensis

12.3 (T8)2 10.2 (T6 and T8)2 98.5 (T8)2 29.3 (female) McHenry (1992)

Sts 14 Australopithecusafricanus

9.4 (T9) 9.6 (T9) 70.9 (T9) 30.2 (female) McHenry (1992)

Stw 431 Australopithecusafricanus

10.7 (T10) 11.5 (T10) 96.6 (T10) 40.8 (male) McHenry (1992)

KNM-WT 15000 Homo ergaster 10.5 (T1) 11.9 (T10) 123.4 (T10) 681 (thisspecimen)

Ruff and Walker(1993)

Kebara 2 Homo neander-thalensis

12.8 (T6) 16.1 (T3) 211.2 (T6) 76 Ruff et al. (1997)

La Chappelle Homo neander-thalensis

13.5 (T1) 17.3 (T8) 253.0 (T6) 76 Ruff et al. (1997)

Shanidar 2 Homo neander-thalensis

15.3 (T3) 76 Ruff et al. (1997)

Shanidar 3 Homo neander-thalensis

12.5 (T6) 17.0 (T7 and T10) 173.6 (T7) 76 Ruff et al. (1997)

Skhul 4 Homo sapiens 15.8 (T11)2 18.0 (T11)2 223.4 (T11)2 66.6 Ruff et al. (1997)1 KNM-WT 15000 is juvenile, estimated to be at an equivalent stage of development to an 11-year-old modern human (Smith, 1993).Assuming that the growth pattern of the vertebral canal in Homo ergaster was similar to that of modern humans, it would havereached adult dimensions by this developmental stage (MacLarnon, 1993). Therefore, the adult body weight estimate for KNM-WT15000 by Ruff and Walker (1993) is used to compare relative canal dimensions.2 Specimens not complete enough to be certain that measurements are of thoracic minima.


Figs. 1–3. Log-log plots of minimum thoracic canaldimensions (1—height; 2—width; 3—cross-sectionalarea) vs. body weight for extant primates and fossilhominids. The best-fit lines drawn in are the principalmajor axes for the original extant sample less humans(see Table 1). Parts b are enlarged versions of the upper

right hand portions of parts a. For ape species, the rangeof measurements for each sex is shown in parts a, and allspecimens in parts b. For extant humans, sample meansand the maximum ranges of two standard deviations oneither side of these are shown.


mates. However, most of the Neanderthal,the early modern human (Skhul 4), and the3 contemporary modern human samplemeans all fall above the extant nonhumanprimate range. In general, however, therecent hominids have larger relative mini-mum thoracic canal widths than the otherhominids and nonhuman primates. There isjust some slight overlap of the Neanderthaland chimpanzee ranges of deviation fromthe best-fit line.

Figure 3 shows a very similar picture tothe plots in Figure 2.All of the australopithe-cines and Homo ergaster have relative mini-mum thoracic canal cross-sectional areaswithin the lower end of the range of extantnonhuman primates. The Neanderthals andearly and contemporary modern humans allhave larger relative thoracic canal cross-sectional areas than all the nonhuman pri-mates. This last result clearly depends onthe body weight estimates for the fossils.The body weight estimates for the Neander-thals and early modern human would haveto be as high as 109 and 118 kg, respectively,for their relative thoracic canal cross-sec-tional areas to be equal to that of the nonhu-man primate with the largest relative canalarea (a chimpanzee specimen). These wouldbe extremely high estimates, and it there-fore seems clear that the two fossil speciesdo indeed have relatively larger thoraciccanals than extant nonhuman primates.

DISCUSSION

The results presented here demonstratethat the thoracic vertebral canal of earlyfossil hominids was of similar relative size tothat of extant nonhuman primates, andsubstantially smaller than that of modern

humans. This has been shown for the twoaustralopithecine species for which the rel-evant material is available, and for Homoergaster. However, the Neanderthals had athoracic vertebral canal of similar relativedimensions to modern humans, includingearly modern humans from Israel.

Vertebral canal dimensions, particularlyof the middle and upper vertebral column,are quite well-correlated with those of thespinal cord contained within the canal inmodern primates, and therefore canal dimen-sions can be interpreted in terms of the sizeof the spinal cord (MacLarnon, 1995). Evi-dence from the relative sizes of the majorspinal cord tissues of extant primates indi-cates that it is only the grey matter of themodern human thoracic spinal cord that hasexpanded beyond the typical relative size forthe order (MacLarnon, 1993). This demon-strates that it is local innervation in thethoracic region that has increased in hu-mans, as the bulk of white matter contain-ing nerve fibers passing through the tho-racic region, to and from the hindlimbs, is ofexpected relative dimensions for a primate(MacLarnon, 1993). Measurements of thespinal nerves themselves were not collected,but presumably these are also larger in thethoracic region of humans, matching theincrease in the volume of nerve cell bodies inthe grey matter, and so further accountingfor the increase in the girth of the vertebralcanal in that area. This evidence indicatesthat thoracic innervation in earlier fossilhominids, australopithecines and Homo er-gaster, was similar to that of extant nonhu-man primates, but that Neanderthals andearly modern humans had expanded tho-

TABLE 3. Statistics for log-log scaling equations relating minimum thoracic canal dimensions to body weight (g)

y variable Sample n Slope95% confidencelimits on slope Intercept r

Minimum thoracic canal height (mm) Extant primates less humans 40 0.30 0.27–0.32 20.38 0.97Extant primates less hominoids 34 0.27 0.25–0.30 20.30 0.97Apes 7 0.30 0.23–0.36 20.36 0.98

Minimum thoracic canal width (mm) Extant primates less humans 40 0.28 0.25–0.39 20.22 0.97Extant primates less hominoids 34 0.28 0.25–0.32 20.23 0.95Apes 7 0.26 0.19–0.33 20.15 0.97

Minimum thoracic canal cross-sec-tional area (mm2)

Extant primates less humans 40 0.58 0.54–0.62 20.70 0.98

Extant primates less hominoids 34 0.56 0.51–0.61 20.63 0.97Apes 7 0.56 0.49–0.64 20.63 0.99


racic innervation similar to that in extanthumans.

Possible explanations for increasedinnervation by the thoracic spinal nerves

The thoracic spinal nerves innervate themuscles of the wall of the thorax (intercostaland subcostal muscles, transversus thora-cis), the skin overlying this region, and thelongitudinal muscles of the thoracic verte-bral column (erector spinae, transversospina-lis). They also supply muscles of the anteriorabdominal wall (external and internal ob-lique muscles, transversus abdominis, rec-tus abdominis), but make little contributionto the innervation of the posterior abdomi-nal muscles. Fibers from the first, or firstand second, thoracic spinal nerves pass tothe brachial plexus and contribute to theinnervation of the arms, but the upper limbsare largely innervated by cervical spinalnerves. The diaphragm, which lies betweenthe thoracic and abdominal cavities, is alsoalmost entirely cervically innervated. Thisreflects the fact that it originates more crani-ally and migrates caudally during develop-ment (Romanes, 1984).

The thoracic and abdominal muscles thatare the major targets of innervation fromthe thoracic spinal nerves have three or fourmain functions in humans. They are in-volved in the maintenance of upright pos-ture and control of the trunk during move-ment, they are sometimes involved inexpulsion such as coughing or defecating,and they are involved in the control ofbreathing. An increase in thoracic innerva-tion is therefore presumably associated withone or more of these activities. Flexion,extension, and rotation of the trunk areactively controlled by trunk muscles, includ-ing the abdominal muscles and the longitu-dinal muscles of the spine, many of whichare thoracically innervated. Bipedal locomo-tion necessitates good control of the uprighttrunk, in order to prevent its inefficientflailing about during locomotion, and for theinitiation of forward motion from standstill(Lovejoy, 1988). More particularly, it hasbeen proposed that the origin of bipedalismwas related to the evolution of accuratethrowing (Fifer, 1987), which involves arch-ing of the back and twisting at the waist

followed by rapid release (Fifer, 1987) asimportant components of the well-coordi-nated, sequential movements needed to hita distant target, such as prey (Calvin, 1982,1992). Whether efficient bipedalism evolvedfor this particular reason, or primarily forlocomotion, its development might have re-quired an increase in the neural control oftruncal movements, including thoracic neu-ral control.

However, as the Nariokotome boy demon-strates, Homo ergaster was fully bipedal andhad a flexible waist. There are very fewdifferences between its postcranial skeletonand those of modern humans except that itwas more robust (Ruff and Walker, 1993;Walker, 1993). Hence, any increase in thecontrol of the trunk necessary for fully devel-oped bipedalism would have taken placeprior to the evolution of Homo ergaster, andyet this species had thoracic innervationsimilar to that of modern nonhuman pri-mates, and not the increased innervation ofmodern humans. Therefore, the evolution ofbipedalism does not explain the evolution ofincreased thoracic innervation (Walker, 1993;Walker and Shipman, 1996).

When expulsion for activities such ascoughing, vomiting, defecation, and parturi-tion is forceful, abdominal-muscles are con-tracted to increase intraabdominal pres-sure, and expiratory intercostal musclesmay be contracted against a closed glottis, soincreasing intrathoracic pressure (Romanes,1984). For most of these activities there is noreason to think that their control mighthave increased during human evolution.However, human parturition is among themost difficult of any primate species, whereasin great apes giving birth is relativelystraightforward (Rosenberg, 1992). The hu-man difficulty results from the large size ofthe neonatal head in relation to the size ofthe birth canal through the pelvis, and thecomparative shapes of the head and canal.Calculations based on pelvic size and shape,known adult brain size, and the relationshipbetween neonatal and adult brain size inprimates demonstrate that birth would prob-ably have been no more difficult in australo-pithecines than in apes, though the mecha-nism may have altered (Rosenberg, 1992).Evidence from other early hominid species is


scarce, but it seems probable that the firstspecies in which the neonatal head wouldhave squeezed through the birth canal wasHomo ergaster (or early Homo erectus), inwhich fast brain growth must have contin-ued after birth in order to attain adult brainsize (Begun and Walker, 1993). Greater mus-cular force might therefore have been re-quired for birth in Homo ergaster comparedwith typical primates and earlier hominids.This could have involved more powerfulcontractions from abdominal and expiratoryintercostal muscles, but the primary muscleinvolved is the uterine muscle, which ismainly sympathetically innervated. Be-cause of this, and the fact that there is noevidence of increased innervation of abdomi-nal and expiratory intercostal muscles untilafter Homo ergaster, increasingly difficultbirth is apparently not the functional reasonfor increased thoracic innervation duringhuman evolution.

The one remaining function of the musclesinnervated by the thoracic spinal nerves isthe control of breathing. Three functionalchanges, or suggested functional changes,during human evolution could have re-quired an increase in the neural control ofbreathing. Firstly, as Carrier (1984) pointedout, the evolution of upright posture enabledthe decoupling of the rhythm of breathingfrom a one-to-one relationship with the loco-motor cycle, as the forelimbs no longer trans-mitted strong forces associated with weighttransference and propulsion through thethoracic cage. This, Carrier (1984) went onto say, would have enabled bipeds to runmore efficiently at more variable speedsthan quadrupeds, being capable of matchingbreathing rates to oxygen requirementsacross a wide range of running speeds.Hence, they could have hunted prey to thepoint of exhaustion by chasing them atspeeds inefficient for the less flexible quadru-peds, for whom a particular speed (for eachgait pattern) is optimal.

Once again, however, the timing of thispossible evolutionary change is not corre-lated with the post-ergaster (or early H.erectus) evolution of increased thoracic inner-vation, as early Homo or Homo erectus areproposed as the first endurance runners(Carrier, 1984; Trinkaus, 1984). It also seems

improbable that the ability to vary the rateof rhythmical breathing according to vari-able oxygen requirements would need agross increase in the innervation of themuscles involved, especially as this abilitymust to some extent be present in otherspecies. Nevertheless, the changes high-lighted by Carrier (1984) in the transmis-sion of locomotor forces may have beenimportant preadaptations for the later evolu-tion of greater control of breathing for otherpurposes.

Secondly, if early human evolution wentthrough an aquatic phase as some havesuggested (Hardy, 1960; Morgan, 1972, 1982;Verhaegen, 1995), swimming and diving togather food resources may have been facili-tated by increased breath control (Morgan,1982) enabling, for example, rapid inspira-tions followed by long slow expirations,rather than the usual more evenly dividedbreathing cycle. More detailed investiga-tion, though, suggests that the human abil-ity for greater breath control is not particu-larly well-suited to swimming and diving,and therefore probably did not evolve as anaquatic adaptation (Patrick, 1991). In anycase, the aquatic phase is proposed to havetaken place earlier than any of the knownhuman fossil remains, or at least earlierthan A. afarensis (i.e., .3.7 Mya) (Morgan,1991). Therefore, the evolutionary increasein thoracic innervation identified here, whichtook place more recently than 1.6 Mya, can-not have been an aquatic adaptation.

The third possible functional reason forthe evolution of increased breathing controlis that it was necessary for the evolution ofmodern human speech. This could have beena secondary change necessitated by otherdevelopments. The human larynx is set lowerin the throat than that of other primatespecies (Negus, 1949; Lieberman, 1968). Theextended pharynx above the human larynxis manipulated for the production of variousphonetic components as air is expired. Therepositioning of the vocal apparatus mayincrease the likelihood of choking as theresult of food entering the trachea ratherthan the esophagus (Laitman, 1985). Thispotential danger is combated in all mam-mals by interrupting respiration while swal-lowing (McFarland et al., 1994; McFarland


and Lund, 1995), but the suggested in-creased need for antichoking muscular ac-tion in humans may have required someincrease in breathing control. However, itdoes not seem likely that this would havebeen sufficiently complex and variable torequire gross neural development.

Much more likely, it is aspects of speechitself that required a significant increase inthe innervation of breathing muscles, for, asCampbell (1968) stated, ‘‘the subtlety ofcontrol required of the intercostal musclesduring (human) speech makes demands ofthe same order as those that are made onthe small muscles of the hand.’’ Evidence tosupport this statement is discussed below. Itcomes from two main sources: studies ofhuman speech production itself, and com-parisons of the production of human speechand the vocalizations of nonhuman pri-mates.

Evidence that modifications to quietbreathing are important in human speech

Work by respiratory physiologists hasdemonstrated the importance of modifica-tions to quiet breathing patterns for theproduction of the long, punctuated, andmodulated utterances typical of humanspeech (e.g., Draper et al., 1959; Taylor,1960; Ladefoged, 1968; Mead et al., 1968;Sears and Newsom Davis, 1968; Bouhuys,1974; Campbell, 1968, 1974; Proctor, 1974;Hoit et al., 1988; Hixon and Weismer, 1995).Respiratory muscles control the subglottalair pressure (i.e., below the larynx) thatfuels voice production in the upper respira-tory tract. Most human speech takes placeon expirations alone, interspersed with rapidinspirations, in a pattern very different fromthe more evenly divided breathing cycles ofquiet breathing. Subglottal air pressure isalso maintained and regulated throughoutexpirations of varying lengths. Respiratorymuscles are active in preventing the generalfading of sound volume that would resultfrom the uncontrolled elastic recoil and gravi-tational deflation of the lungs and rib cage.They also enable sound production to con-tinue beyond the point at which the normalminimum lung volume in quiet breathing isreached. Together, these complex aspects ofbreathing control enable adult humans to

speak fluently in long sentences, withoutdisruptive pauses for inspirations, and withthe necessary quick inspiratory pausesplaced at meaningful linguistic boundaries(Lieberman and Lieberman, 1973; Lieber-man, 1984).

Control of subglottal pressure is also im-portant to several other features which varyin human speech, including its intensity orloudness, the emphasis of particular syl-lables or phonemes, and pitch and intona-tion patterns (Ladefoged, 1968; Proctor,1974; Hoit et al., 1990a; Stathopoulos andSapienza, 1993). Changes in the length andshape of higher vocal tract structures, par-ticularly the larynx, also affect these aspectsof speech, and the relative importance ofrespiratory and laryngeal control can varybetween individuals (Ladefoged, 1968;Stathopoulos and Sapienza, 1993). Evi-dence, however, suggests that respiratorycontrol is the more important in modifyingloudness and emphasis, both in speech(Lieberman, 1984; Hoit et al., 1990a; Statho-poulos and Sapienza, 1993) and in singing(Sundberg et al., 1993). In addition, al-though most phoneme production is con-trolled in the upper respiratory tract, forsome languages there is also evidence thatthe production of certain consonants is differ-entiated by differences in subglottal pres-sures (Ladefoged, 1968).

At both the larger scale of breath cycles,and the finer scale of detailed features withinphrases and words, therefore, subglottalpressure must be well-controlled for theproduction of human speech. Even at thelarger scale this is quite complex. The recoilproperties of the lungs and rib cage aredifferent at different volumes, and so theproduction of a particular subglottal pres-sure requires different patterns of muscleactivity according to the lung volume (e.g.,Mead et al., 1968). Added to this is the finercontrol needed for subtle variation in empha-sis, pitch, and intonation, and again, at anygiven lung volume, different respiratorymuscle actions are needed to produce par-ticular changes in air pressure, and differentlinguistic requirements may be superim-posed on each other, e.g., falling pitch at theend of some phrases and emphasis of aparticular syllable (Ladefoged, 1968). Differ-


ent languages make different demands, butall require subtle control of subglottal airpressure to convey their meaning (Lieber-man, 1984).

The relatively few studies of voice disor-ders resulting from disruption of breathingcontrol also confirm the importance of breathcontrol in fluent human speech. Patientswith cervical cord damage from C4/5 to C7/8and consequent impairment of thoracic andabdominal muscle function suffer from anassociated constellation of speech problems.These include the inability to produce otherthan short phrases, reduced loudness, andreduced contrasts in terms of stress ormonotonal speech, as well as slow and there-fore longer inspirations (Hixon and Putnam,1983; Hoit et al., 1990a).

The subglottal pressure required to pro-duce the specific air flow rate needed for aparticular intensity or pitch of speech soundalso changes with variation in the resistanceof the upper respiratory tract. Changes inthe shape of the larynx, pharynx, and mouthare the main causes of the production ofdifferent phonemes. Vibration of the vocalfolds in the larynx causes the production ofsound from the outflow of air from the lungs.Changes in the shape of the pharynx andmouth filter the frequency profile of thissound, or cause it to be released in unevenbursts. These variations are perceived bythe listener as the different phonemes ofhuman speech. All these changes to theshape of the upper respiratory tract alter itsresistance to air flow. Therefore, the subglot-tal pressure needed to produce the specificair flow rate required for a particular pitchor intensity of speech varies according to theoften rapidly changing sequence of pho-nemes produced. Subtle feedback mecha-nisms between the upper respiratory tractand the respiratory muscles are no doubtinvolved (e.g., Gould and Okamura, 1974).The flexibility and power of these systems isdemonstrated by the ability of patients withinjury to the upper respiratory tract to com-pensate by varying subglottal pressure out-side the normal range in order to produceintelligible speech (e.g., Anthony, 1980).

The neural control of subglottal pressureduring human speech must also involvecognitive factors, and hence feedback be-

tween the brain and spinal cord. At the finerlevel, variation in emphasis, pitch, and into-nation, which involve respiratory control,can alter the meaning conveyed by pho-nemes, words, and phrases (Ladefoged,1968). There is also evidence that at thegrosser level, the volume of air inspired atthe start of a phrase reflects the amountneeded to produce it before making the nextinspiration at a suitable linguistic point(Lieberman and Lieberman, 1973; Lieber-man, 1984; Winkworth et al., 1995). Finally,the muscular movements needed to producea particular rate of air flow also vary accord-ing to body position, general activity levels,and the amount of liquid in the stomach(Bouhuys, 1974; Hixon et al., 1976; Lieber-man, 1991). Taking all these factors to-gether, human speech requires very fast,fine control of subglottal pressure whichresponds to cognitive factors and is inte-grated with control of the upper respiratorytract and other body changes. This makesconsiderable demands on the control of therespiratory muscles, for, as Campbell (1968,p. 137) said, ‘‘Moving air is easy, but control-ling it is difficult.’’

Evidence that thoracic neural control isparticularly important for the control of

breathing during human speech

The main muscles involved in humanbreathing are the diaphragm, internal andexternal intercostals plus some abdominalmuscles, rectus abdominis, and external andinternal oblique muscles. All of thesemuscles, apart from the diaphragm, arethoracically innervated. Other muscles havealso had a respiratory role attributed tothem, such as the scalenes, sternocleidomas-toid, trapezius, pectoralis major and minor,serrati, sacrospinalis, levatores costarum,transversus thoracis, erector spinae, subcla-vius, latissimus dorsi, and quadratus lumbo-rum, but their role in breathing is moreminor, including their role, if any, duringspeech (Campbell, 1968; Romanes, 1984;Sharp and Hyatt, 1986).

During quiet breathing the main musclesof inspiration are normally described as thediaphragm, the interchondral portion of theinternal intercostal muscles, and the poste-rior external intercostal layer. Expiration


involves the action of the interosseous por-tion of the internal intercostals, togetherwith passive recoil of the rib cage and lungs(Campbell, 1968, 1974; Borden and Harris,1984). The role of the abdominal muscles inquiet breathing is less clear, although theyhave been reported to be active throughoutinspiration and expiration (Gould and Oka-mura, 1974). However, this may only be thecase when the body is in an upright standingposition, but not when it is supine (Hoit etal., 1988). In the former circumstances, theycould be acting to pull the contents of theabdomen inwards so that the costal fibers ofthe diaphragm are stretched and its me-chanical efficiency is increased (Hoit et al.,1988). However, whatever the involvementof the abdominals during quiet breathing,both cervically and thoracically innervatedmuscles are involved.

From the evidence presented above, it isclear that the intricacies of human speechplace demands on the neural control ofmuscles that are of a very different orderfrom those of quiet breathing. However,analyzing the detailed muscle actions in-volved is problematic precisely because oftheir variability and subtlety. A further com-plicating factor is that different individualsappear to adopt different and sometimesinefficient muscular strategies to achieveparticular aspects of speech (Ladefoged,1968; Hixon et al., 1976; Hixon and Putnam,1983; Hoit et al., 1990a; Winkworth et al.,1995). In addition, results derived from theuse of different techniques such as monitor-ing the electrical activity of muscle groupswith EMG (Draper et al., 1959; Taylor, 1960;Sears and Newsom Davis, 1968; Campbell,1974), or deformations of the chest wallusing kinematic observations (Hixon et al.,1976; Hixon and Weismer, 1995), or linear-ized magnemometers or respiratory plethys-mography (Proctor, 1974; Hixon and Put-nam, 1983; Hoit et al., 1990b, 1994;Winkworth et al., 1995) have sometimesbeen contradictory. Interpretation has alsovaried. In their now classic work, the Edin-burgh group concluded that the intercostalmuscles play a much more important rolethan abdominal muscles during conversa-tional speech (Draper et al., 1959; Ladefo-ged, 1968). However, Hixon et al. (1976) and

Hixon and Weismer (1995) have criticizedthis view, suggesting a more central role forthe abdominals. For present purposes,though, such differences are not particularlyimportant, as both these sets of muscles arethoracically innervated.

There is general agreement that the inter-costal muscles involved in inspiration forquiet breathing (i.e., the external intercos-tals, interchondral portions of the internalintercostal muscles) are active during therapid inspirations required for speech. Hixonet al. (1976) also emphasized the importanceof the abdominal muscles in this phase ofthe speech breathing cycle. They suggestedthat these muscles are tensed before aninspiration, which raises the diaphragmslightly, stretching its costal fibers and in-creasing its potential contribution to rapidinspiration. However, there is also evidencethat the diaphragm is much less importantduring phonation than it is in quiet breath-ing (Draper et al., 1959; Taylor, 1960;Bouhuys et al., 1966; Campbell, 1974). Proc-tor (1974) even stated that the diaphragm iscompletely relaxed during most speech andsinging. The external (or inspiratory) inter-costal muscles may also continue to be ac-tive during the initial moments of speechexpiration, particularly for soft speech(Draper et al., 1959). According to Draper etal. (1959), this enables the increase in sub-glottal pressure due to relaxation at thestart of a speech sequence to be restrictedappropriately. However, Hixon and Weismer(1995) have strongly criticized this interpre-tation.

Expiration during speech is initially theresult of elastic recoil of the lungs, but atsome point (precisely where is a matter ofdispute) this is insufficient to maintain sub-glottal air pressure at the required level. Atthis stage the internal (expiratory) intercos-tal muscles and abdominal muscles becomeactive, increasing the expiratory drive tomaintain subglottal pressure (Draper et al.,1959; Hixon et al., 1976). This allows speechto be produced at a consistent volumethroughout a phrase on a single breath. Thesame muscles can also increase subglottalpressure briefly for emphasis of a particularsyllable or for the production of particularconsonants (Mead et al., 1974; Borden and


Harris, 1984). The relative contribution ofthe two sets of muscles in regulating expira-tions during speech is still a matter ofdispute, and it is unclear whether the appar-ently contradictory results produced ema-nate from the different experimental condi-tions used or the different measurementtechniques (e.g., Draper et al., 1959; Hoit etal., 1988; Stathopoulos and Sapienza, 1993),or from different strategies adopted by differ-ent individuals.

Understanding of the precise musculardeployment in breathing control during hu-man speech is clearly far from complete.However, it is reasonably well-establishedthat the most important muscles involved,both in inspiration and expiration, are theintercostal muscles and a set of abdominalmuscles, all of which are thoracically inner-vated, and therefore the more detailed differ-ences in interpretation of results are notparticularly important here. There is alsogeneral agreement that complex coordina-tion of the movements of these muscles isrequired to produce the subtlety of controlneeded for the intricacies of human speech.This coordination may include load-compen-sating reflexes within the spinal cord, incor-porating feedback from muscle spindles andtendon organs with which the intercostaland abdominal muscles are richly suppliedcompared to the diaphragm (e.g., Sears andNewsom Davis, 1968; von Euler, 1968; New-som Davis and Sears, 1970; Campbell, 1974;Grassino and Goldman, 1986; Iscoe, 1998).

Central control of speech breathing isserved by neural pathways that are function-ally and anatomically distinct from thoseinvolved in the control of respiration formetabolic purposes (Phillipson et al., 1978).Metabolic respiration is controlled by thepons and medulla of the hindbrain, involv-ing vagal and chemoreceptor reflexes,whereas the control of breathing for phona-tion is more complex, involving forebrainand midbrain pathways (Purves, 1979;Zhang et al., 1994; Davis et al., 1996; Mur-phy et al., 1997). At the level of the spinalcord, several descending tracts and reflexloops involved in the control of breathing areintegrated (Newsom Davis and Sears, 1970;Milic-Emili and Zin, 1986). Reflexes originat-ing from the proprioceptors of the lower

intercostals might also affect phrenic moto-neurons, producing an excitatory effect onthe diaphragm (Decima et al., 1967; vonEuler, 1968). While the picture again is farfrom complete, there is clear evidence ofsignificantly different and increased neuralcontrol of breathing for speech production inhumans compared with that for metabolicrespiration. And, at the level of the spinalcord, these differences occur in the thoracicregion.

Human language compared withnonhuman primate vocalizations

Human language differs from the vocalcommunication of nonhuman primates inmany ways. In broad terms, these can bedivided into cognitive differences and differ-ences in sound production, though clearlythe two aspects must be interrelated formeaningful communication.As shown above,sophisticated control of breathing is veryimportant to human speech, or the physicalproduction of human language, and it issuggested that this control increased duringhuman evolution. An important part of theevidence for this is the difference in theamount of thoracic innervation in extanthumans and nonhuman primates. In addi-tion, if the association made here betweenan evolutionary increase in thoracic innerva-tion and breathing control is correct, thenthe breathing control required for nonhu-man primate vocalizations should be sub-stantially less than that needed for humanspeech. The evidence for this part of thesuggested evolutionary picture is consideredbelow.

Precise comparisons of breathing patternsduring human speech and nonhuman pri-mate sequences of vocalizations are difficultfor a number of reasons. The most importantof these is that most researchers who haveworked on nonhuman primate vocalizationshave not focused on breathing related tovocalization, and hence information in thisarea is often only available incidentally, orhas to be estimated from sonograms selectedand reproduced for other purposes. Whatevidence there is suggests that sequences ofvaried, discrete sounds are commonly pro-duced by nonhuman primates on a series ofboth inspirations and expirations, with only


single sound units expressed on individualair movements (e.g., pant hoot, Pan tro-glodytes (Marler and Hobbett, 1975; Marlerand Tenaza, 1977; Clark and Wrangham,1993); sections of gibbon songs, Hylobatesspp. (Geissmann, 1993; Gittins, 1984; Hai-moff, 1983, 1984); whoop gobble, Cercocebusspp. (Waser, 1982); two-phase or roar grunt,Papio spp. (Andrew, 1963a,b; Byrne, 1981,1982); gecker, Macaca fuscata (Green, 1975);loud low-pitched calls, Callicebus moloch(Robinson, 1979); grunt series, Eulemur ful-vus and Nycticebus coucang (Andrew, 1963a);click-grunt, Otolemur crassicaudatus andPerodicticus potto (Andrew, 1963a)), ratherthan multiple units on extended expira-tions, which is the human pattern. Thisqualitative distinction implies that the inter-relationship between vocalization andbreathing is very different in our species.

Despite the paucity of direct evidence, it ispossible to make a series of predictions ofadditional quantitative differences thatwould be expected if human vocalizationdoes indeed involve greater breath controlthan that of nonhuman primates, and to testthese largely using information that can begleaned from the literature. These predic-tions are that nonhuman primates will haveless extended exhalations during vocaliza-tions than humans, more equal divisions ofbreathing cycles into the inspiratory andexpiratory portions, and a drop in fundamen-tal frequency through single expirations ofmore than a very short duration. In addi-tion, it is expected that nonhuman primateswill be limited in their rate of production ofdiscrete sounds (particularly if this involvesa series of inspirations and expirations), andrestricted in the order in which sounds canbe produced relative to the phase of thebreathing cycle.

Compared with quiet breathing, duringspeech, humans take quicker inspirationswith increased volume and they extend theexhalation phase of the breathing cycle (e.g.,Borden and Harris, 1984). There is no goodevidence that nonhuman primates modifythe inspiratory phase, although occasionallyit has been suggested (Larson et al., 1994;Davis et al., 1996). The duration of speakingbreaths (expirations) in human speech nor-mally ranges between about 2–6 sec (Hoit et

al., 1994; Kien and Kemp, 1994; Mitchell etal., 1996), although it can be more than 12sec (Winkworth et al., 1995). As far as it ispossible to assess from published informa-tion (including estimations from publishedsonograms), the longest calls given by nonhu-man primates on a single breath are muchshorter than this, e.g., individuals of Indriindri and Alouatta palliata can producecalls of up to 5.0 sec and 4.8 sec, respectively(Table 4).

However, a comparison of the maximumdegree to which the duration of a restinginhalation is extended during vocalizationsis probably more important than a compari-son of the absolute duration of calls. Restingbreathing rate scales allometrically withbody weight in mammals (Stahl, 1967) andis absolutely slower in larger species. There-fore, assuming that exhalations make up50% of a breathing cycle, which is a reason-able approximation (see below), the dura-tion of resting exhalations will be longer inlarger-bodied species. Available data indi-cate that the majority of nonhuman primatespecies only extend the duration of exhala-tions to 2–3 times the resting duration,compared with humans who can produce amore than 7-fold increase (see Table 4).Some nonhuman species can extend exhala-tions to at least 4–5 times resting duration,e.g., some Hylobates spp., Indri indri, andAlouatta palliata. These species all havelaryngeal air sacs (though H. klossi may bean exception; Geissmann, 1993), which hu-mans lack. This adaptation could be analternative means of increasing the dura-tion of exhalation during vocalization (Fitchand Hauser, 1995), though one that is lesspowerful and less controllable in terms ofvariability of the rate of air release, incomparison with enhanced thoracic breathcontrol in humans. The evidence indicatesthat humans can increase both the absoluteand relative duration of exhalations duringvocalization substantially more than anynonhuman primate, despite our lack of airsacs (see Table 4).

Mainly as a result of extending exhala-tions, humans are apparently unique amongprimates in the extent to which they candistort an even breathing cycle for vocaliza-tions to a greater extent than any other


TABLE 4. Respiratory data for call duration and call rate: examples from species of a range of body sizes and respiratory adaptations

Species

Mean bodyweight1

(kg)

Mean restingbreathing rate2

(min21)

Duration oflongest call3

(sec21)

Call length asa multipleof resting

exhalation4Call rate5

(sec21)

Call rate as amultiple of resting

breathing rate6

References from whichcall duration and callrates were estimated

Without air sacsGalago senegalensis 0.19 82.4 0.5 1.4 3.0 2.2 Andrew (1963a), Zimmermann et al. (1988)Tarsius spectrum 0.20 82.0 0.7 2.0 6.0 4.3 Haimoff (1986)Otolemur crassicaudatus 1.2 51.1 0.7 0.6 2.3 2.7 Andrew (1963a), Zimmermann (1990)Hylobates pileatus 5.45 34.4 2.1 2.4 2.0 3.5 Haimoff (1984, 1986), personal observationHylobates lar 5.6 34.2 2.8 3.1 2.0 3.6 Haimoff (1984), Raemakers and Raemakers

(1985)With air sacs

Alouatta palliata 6.55 32.9 4.8 5.1 3.0 5.6 Sekulic and Chivers (1986)Hylobates concolor 7.55 31.6 4.5 4.7 3.0 5.7 Deputte (1982), Haimoff (1984), Haimoff et

al. (1987)Indri indri 10.5 29.0 5.0 4.8 3.0 6.2 Haimoff (1986), Thalmann et al. (1993)Pan troglodytes 45.9 19.8 1.6 0.6 4.0 12.5 Marler and Tenaza (1977), Clark and

Wrangham (1993)Gorilla gorilla 117.55 15.7 3.2 1.6 5.4 20.6 Harcourt et al. (1993)

With enhanced breathing controlHomo sapiens 60.00 18.5 12.0 7.2 12–15 Lieberman et al. (1992), Winkworth et al.

(1995)1 Body weight data from Martin and MacLarnon (unpublished data).2 Breathing rates estimated from body weight using the allometric formula, respiration rate 5 53.5W20.26, where W 5 body weight in kg (Stahl, 1967).3 Duration of longest calls given on a single breath calculated from published data, including sonograms. Breathing patterns either described in source or implied by definitions.4 Duration of resting exhalation calculated as 50% of duration of a single breathing cycle (see text).5 Fastest rate of call units given on sequences of inhalations and exhalations (call rate) calculated from published data, including sonograms. Breathing patterns either described in source or impliedby definitions.6 Call rate as a multiple of resting breathing rate calculated by dividing the absolute call rate by the number of resting breathing cycles per second.

primates. In human speech, exhalations typi-cally comprise 85% of each cycle (e.g., basedon data in Borden and Harris, 1984; Mitchellet al., 1996). Published sonograms, and somepublished data, suggest that expirations usu-ally fill between 35–65% of breathing cyclesduring the vocalizations of nonhuman pri-mates, e.g., 55% for the pant hoots of Pantroglodytes (calculated from data in Marlerand Hobbett, 1975), 67% for train grunts inGorilla gorilla (calculated from Harcourtet al., 1993), and 33–57% for the woo-ahsequence of Hylobates agilis (estimated fromsonograms in Gittins, 1984). During humanspeech, therefore, which is only produced onthe expiratory phase, pauses in sound pro-duction for inspiration are minimized, andthe length of expirations can be varied overa considerable range, enabling inspiratorybreaks to be placed at linguistically suitablepoints (Winkworth et al., 1995; Davis et al.,1996; Mitchell et al., 1996).

The fundamental frequency (F0) of soundproduction falls as subglottal pressure falls.Without breath control, expiration essen-tially results from the relaxation of therespiratory muscles, and vocalizations onsuch expirations would be expected to showa decline in F0 through their duration. Inhumans, F0 can be controlled voluntarily, assubglottal pressure can be controlled, and F0only falls in 63% of spontaneous utterances(Lieberman, 1983, as corrected by t’Hart,1986, and cited in Hauser and Fowler 1992).A rise or fall in F0 at the end of a phrase inhuman speech affects meaning; it may indi-cate a question or the end of an utterance, orit may facilitate (or prevent) turn-taking inconversation. Control of F0 changes throughan expiration is an important aspect ofhuman speech and language.

Nonhuman primates commonly show adrop in fundamental frequency through avocalization on one exhalation, e.g., the coosof rhesus macaques, the wrrs of vervets, andthe girneys of rhesus macaques (interpretedfrom evidence in Andrew, 1963a; Hauser,1991; Hauser and Fowler, 1992). However,there are other vocalizations produced onsingle expirations which have a flat or risingF0 (e.g., segments of gibbon songs; Haimoff,1984). In almost all cases, these are of veryshort duration (Hauser and Fowler, 1992),normally no longer than about 1 sec, and

many are much shorter. In a survey of theliterature, very few cases were found oflonger vocalizations with a rising inflection.Usually, these are associated with the pres-ence of air sacs, e.g. a long whoo of up to 4 secgiven by a male Hylobates concolor leucog-enys in response to a female song bout (Fig.4.6 in Deputte, 1982). There are also a veryfew examples where air sacs are not present:e.g., H. lar, one rising note lasting approxi-mately 2.8 sec (Fig. 5 in Raemakers et al.,1984), and H. klossi, a long, flat whoo lasting2.5 sec (Fig. 2 in Tenaza, 1976), thoughneither of these is very extended. It is appar-ently broadly true to say that nonhumanprimates do not produce substantially ex-tended exhalations without a declining F0,except by the use of air sacs, which humansdo not possess. It is possible that nonhumanprimates have no choice about this drop infrequency because they are unable to controltheir rate of breath release to the samedegree as humans. There is also no evidencefrom nonhuman primates of the complex,fast-changing patterns of rising and fallingF0 on a single expiration that characterizehuman speech, and play a central role inconveying meaning, be it through tonal dif-ferentiation of phonemes, or intonationacross a phrase.

As well as fundamental frequency, hu-mans can also modify the volume or ampli-tude of speech to emphasize any syllable orpart of a phrase, providing linguisticallyimportant flexibility. No evidence of suchcontrol has been reported in nonhuman pri-mates.

Human speech is characterized by theproduction of rapid sound sequences whichLieberman et al. (1992) stated are 10 timesfaster than those of any nonhuman primate.This rate of sound sequencing is an essentialelement in rapid information transfer. Com-parison between humans and nonhumanprimates of the rates of production of dis-crete sounds is fraught with difficulties,mostly related to the definition of a ‘‘unit ofsound.’’ Human speech is usually parti-tioned into phonemes or syllables, unitswhich are meaningful in terms of percep-tion, and which are not defined solely bytheir acoustic properties. The units used todescribe nonhuman primate vocalizations


are essentially acoustic, and there is nogeneral agreement about their comparabil-ity between species, or between differentstudies, and in particular about their compa-rability to the units of human speech.

There are two main factors that togetherresult in the rate of sound sequencing dur-ing vocalization: first, the number of soundunits (e.g., notes, syllables, phonemes) pro-duced on a single expiration or inspiration,and second, the breathing rate, i.e., thenumber of breathing cycles per second. Non-human species can produce sequences ofsound units on a single breath using anumber of mechanisms, though most of theseare quite simple repetitions, which do notapproach the highly varied sound sequencesof human speech. Many species producetrills and quavers by simple tongue or lipmovements, e.g., Cercopithecus aethiops(Hauser and Marler, 1992), C. mitis (Hauseret al., 1993), and Macaca fuscata (Green,1975). Some species can produce more com-plex sound sequences on a single breath,e.g., two more distinctive units are producedin the waa-quaver of Hylobates lar, by open-ing and closing the mouth (Haimoff, 1984).More distinctive still, the bitonal scream ofthe siamang involves two sound units on asingle expiration, produced by inflation anddeflation of the laryngeal airsacs (Haimoff,1984). The multimodulated squeals of Sykes’monkey, Cercopithecus albogularis, mightinvolve independent and simultaneous vibra-tions of the vocal folds in the larynx and thevocal lip, using a combination of air flowfrom the lungs and air sacs (Brown andCannito, 1995). The more complex sequencesseem to involve the use of air sacs (whichhumans do not possess), and they are consid-erably less variable and flexible than thesound sequences which humans produce ona single breath by complex manipulations ofthe upper respiratory tract. Nonhuman pri-mates are apparently severely limited in theextent to which they are able to divide asingle expiration into clearly delineatedsound units, something which Provine (1996)also suggested, based on the limitations onchimpanzees’ ability to laugh. They there-fore would not be capable of utilizing veryextended expirations to produce complexsound sequences, as in human speech.

The second factor contributing to the rateof sound sequencing, i.e., breathing rateduring vocalization, is therefore especiallyimportant, as nonhuman primates are appar-ently limited in the extent to which they canproduce varied sounds on a single air move-ment. As mentioned previously, restingbreathing rate is faster in smaller thanlarger species (Stahl, 1967). For a medium-sized primate of 5–6 kg, a 3–4-fold increaseof resting breathing rate, well within thenormal physiological range, would enable 2or 4 sounds per sec, depending on whethersounds are produced on both inhalation andexhalation, or only one part of the cycle. Asimilar rate of sound production could re-quire a 6–7-fold increase from resting breath-ing rate for humans, and more than an8-fold increase for gorillas. Clearly, the physi-ological demands of producing fast breath-ing rates are greater for larger species. Airsacs may again have an important role here,and larger-bodied species that vocalize onbreathing rates more than 3–4 times restingbreathing rate generally have air sacs, e.g.,Pan spp., Gorilla, larger-bodied Hylobatesspp., a range of larger New and Old Worldmonkey species, and Indri (see Table 4).Humans do not possess air sacs, but theability of our species to produce fast soundsequences on single extended exhalationsprovides an alternative, and apparentlymuch more effective, means of overcomingthe physiological limitations of large bodysize.

The communicatory power of the vocaliza-tions of nonhuman primates is further re-stricted because they normally only producespecific sound units on either exhalations orinhalations, but not both, which restrictstheir order of production. However, the se-quence of note production definitely can beimportant in conveying meaning. For ex-ample, Mitani and Marler (1989) showedthat playbacks of reordered sound sequencesof the great call of Hylobates agilis producedno vocal response from individuals of thespecies. Particular sequences of sounds pro-duced by nonhuman primates can convey avariety of information (e.g., Seyfarth andCheney, 1992), such as the species or indi-vidual status of the caller, the type of ap-proaching predator, mating interest, or terri-


torial claim. However, most authors agreethat nonhuman vocalizations contain noth-ing approaching human syntactical construc-tions, the encoding systems enabling rapidand highly flexible information transfer. Cog-nitive differences between humans and non-human primates are no doubt of great impor-tance here, but so are physiological andmorphological differences. The latter in-clude several features of the upper respira-tory tract such as the much greater ability ofhumans to mould the pharynx and mouthstructures to filter the spectrum of harmon-ics generated by the larynx in order toproduce different phonemes. However, theimportance of the human ability to produceextended, controlled expirations should notbe underestimated. It is not known whethernonhuman primates could be trained to takeon a more human pattern of exhalations, butthere is no evidence from their habitualvocalizations that they would be capable ofdoing so.

Evidence of differences in the muscularcontrol of breathing for human speechand nonhuman primate vocalizations

There are few detailed studies of breathcontrol during vocalizations in nonhumanprimates, apart from the extensive series oflaboratory studies on respiratory muscleactivity and the neural control of vocaliza-tions in Saimiri sciureus (Jurgens, 1982,1988; Jurgens and Kirzinger, 1982, 1984;Jurgens and Richter, 1986; Sutton and Jur-gens, 1988; Jurgens and Schreiver, 1991;Kirzinger and Jurgens, 1994) and the stud-ies on macaques (Macaca nemestrina and M.fascicularis) (West and Larson, 1993; Lar-son et al., 1994). These workers reportedthat vocalizations occurred only on exhala-tions and primarily involved abdominalmuscles rather than intercostal muscles,which were active prior to vocalizations inthe study of macaques by West and Larson(1993) and not consistently active duringvocalizations in the studies of either genus.Jurgens and Schreiver (1991) speculatedthat the intercostal muscles in S. sciureuscould be involved in supporting the thoraxand providing anchorage against which theabdominal muscles could act, rather thanproducing respiratory drive. West and Lar-

son (1993) suggested that in macaques thediaphragm could have a central role duringvocalizations similar to that of the intercos-tals in humans. This evidence suggests thatthe pattern of muscular recruitment duringnonhuman primate vocalizations is very dif-ferent from that involved in producing theair flow for human speech (as describedabove), in which the intercostals have aprimary role, although the actions of theabdominal muscles may be more similar.However, all such comparisons are difficultbecause most of the vocalizations in thenonhuman primate studies were not pro-duced naturally. Rather, they were elicitedby stimulation of electrodes in the periaque-ductal region of the midbrain, and all wereof relatively short duration (100–600 msec).More definitive analysis of the relative mus-cular involvement during nonhuman pri-mate vocalizations and human speech awaitsfurther physiological research.

CONCLUSIONS

Evidence presented here on the size of thethoracic vertebral canal in a range of fossilhominids has been interpreted as showingthat thoracic innervation among australopi-thecines and Homo ergaster (or early Homoerectus) was similar to that of extant nonhu-man primates. Sometime later in humanevolution, the grey matter in this part of thecord expanded, and Neanderthals and earlymodern humans had expanded thoracic in-nervation, like extant humans. It seemsmost probable that this increased innerva-tion evolved to enable enhanced breath con-trol, and the most likely functional reasonfor this was the evolution of human speech,i.e., the physical production of language.Full human language requires extended ex-halations for vocalizations and increasedcontrol of volume, emphasis, and intonationcompared with nonhuman primates andtherefore presumably compared with earlyhominids. Such features require fast, intri-cate, flexible, and integrated neural controlof intercostal and abdominal muscles.

Enhanced breath control, which is a neces-sary feature for fully modern language,therefore was not possible for earlierhominids up until at least 1.6 Mya, the timeof Homo ergaster (or early Homo erectus).


However, this feature was present inhominids of 100,000 mya or less, includingboth Neanderthals and early modern hu-mans. Many evolutionary changes, both cog-nitive and physical, were required for thedevelopment of fully modern language fromour nonhuman ancestors. A range of paleon-tological and archaeological evidence hasbeen interpreted as indicating the probabletime or time range of their evolution (seeTable 5). The present evidence from thethoracic vertebral canal suggests that thelanguage abilities of early hominids were atmost severely limited to short, unmodulatedutterances, lacking rapid sequencing, per-haps something like the protolanguage ofBickerton (1990).

With such a long and important time gapbetween the most recent fossil hominid withan unexpanded thoracic canal and the earli-est fossils with expanded canals, the evi-dence presented here can only make a lim-

ited contribution to the great debate overmodern human origins, i.e., the single origintheory (Stringer and Andrews, 1988) vs.multiregionalism (Wolpoff et al., 1984). How-ever, with the early date of about 1.8 myanow attributed to Homo erectus in the FarEast (Swisher et al., 1994), it seems likelythat the first major migration out of Africawas of hominids lacking the developed breathcontrol necessary for modern human speech,like other African hominids up to at least 1.6mya (KNM-WT 15000). Assuming such anevolutionary change would be highly un-likely to evolve twice, and that sufficientgene flow to carry it thousands of miles isvery improbable, this seems to act againstAsian Homo erectus being the ancestor ofmodern Asians, as the multiregional theoryproposes. Based on the same types of as-sumption, the expanded thoracic canalsshown by Neanderthals and early modernhumans suggest that they had a common

TABLE 5. Summary of suggested dates for evolution of language or contributory features

Date Features Evidence References

3.5 Mya or more recently Increased brain size,increased size of par-ticular brain parts, brainasymmetries

If the initial enlargement ofthe brain was important,language may haveevolved at any time fromabout 2 Mya. However,more detailed featuresthat have been associatedwith earlier language evo-lution require very carefulinterpretation in the lightof recent findings.

Holloway (1983)Falk (1980)Aiello and Dunbar (1993)Walker (1993)Petersen et al. (1988, 1989)

1.6 mya–100,000 years ago Enlargement of vertebralcanal, indicatingincreased control ofbreathing

Absent in australopithe-cines and Homo ergaster,but present in Neander-thals and early modernhumans

MacLarnon (1993)MacLarnon and Hewitt

(1995)Present paper

400,000–300,000 years ago Basicranial flexion present,indicating laryngealdescent and high pharynx

Present in some archaicHomo sapiens (e.g.,Kabwe, Petralona), butnot Neanderthals

Laitman et al. (1992)Lieberman (1984)

.300,000 years ago Hypoglossal canal as largeas modern humans, indi-cating tongue richly sup-plied with motor nervesand complex motor coordi-nation possible

Absent in australopith-icenes, but present inarchaic Homo sapiens(e.g., Kabwe, Swans-combe, and Neanderthals)and early modern humans

Kay et al. (1998)

.100,000 years ago Hyoid bone human-like Kebara Neanderthal Arensburg et al. (1990)

40,000 years ago Cultural development ofLate Stone Age and UpperPalaeolithic, indicatingsymbolic behavior (e.g.,variety of materials used,variety of objects fash-ioned, including nonutili-tarian objects, jewelry,grave goods, cave art)

Modern humans only Isaac (1976)Noble and Davidson (1991)Mithen (1996)


ancestor after the evolution of thoracic ex-pansion, more recently than 1.6 Mya, asgenetic evidence now indicates (Krings etal., 1997). However, evidence from the verte-bral canal does not enable more precisedetermination of how distant or recent thisancestor was. More fossil evidence from this‘‘gap’’ is needed to clarify the more recentevolutionary history of this neglected aspectof the evolution of human language, theevolution of sophisticated neural control ofbreathing required for speech production.

ACKNOWLEDGMENTS

Access to specimens in their care wasgranted by: Theya Molleson and ChrisStringer (Natural History Museum, Lon-don), Yoel Rak (Tel Aviv University), FrancisThackeray (Transvaal Museum, Pretoria),Phillip Tobias (University of the Witwa-tersrand, Johannesberg), and Joe Zias (Rock-efeller Museum, Jerusalem). Erik Trinkauskindly made measurements available of LaChapelle and Shanidar 2 and 3, and AlanWalker measured KNM-WT 15000. TheLeakey Trust and the Boise Fund providedgrants for A.M.M. to travel to Israel andSouth Africa. Many colleagues helped tohone the ideas presented here. We particu-larly thank Alan Walker for his enthusiasticencouragement.

LITERATURE CITED

Aiello L. 1992. Size and shape in human evolution. JHum Evol 22:127–147.

Aiello L, Dunbar RIM. 1993. Neocortex size, group size,and the evolution of language. Curr Anthropol 34:184–193.

Andrew RJ. 1963a. The origin and evolution of the callsand facial expressions of the primates. Behaviour20:1–109.

Andrew RJ. 1963b. Trends apparent in the evolution ofvocalization in the Old World monkeys and apes. In:Napier J, Barnicot NA, editors. Symposium 10, Theprimates. London: Zoological Society of London. p.89–104.

Anthony J. 1980. Aerodynamic and phonetic analysis.In: Edwards M, Watson A, Watson H, editors. Ad-vances in the management of cleft palate. Edinburgh:Churchill Livingstone. p 232–246.

Arensburg B, Schepartz LA, Tiller AM, VandermeerschB, Duday H, Rak Y. 1990. A reappraisal of theanatomical basis for speech in middle palaeolithichominids. Am J Phys Anthropol 83:146.

Armstrong DF, Stokoe WC, Wilcox SE. 1994. Signs of theorigins of syntax. Curr Anthropol 35:349–368.

Begun D, Walker A. 1993. The endocast. In: Walker A,Leakey R, editors. The Nariokotome Homo erectusskeleton. Cambridge, MA: Harvard University Press.p 326–358.

Bickerton D. 1990. Language and species. Chicago:University of Chicago Press.

Borden GJ, Harris KS. 1984. Speech science primer.Baltimore: Williams and Wilkins.

Bouhuys A. 1974. Breathing: physiology, environmentand lung disease. New York: Grune & Stratton.

Bouhuys A, Proctor DF, Mead J. 1966. Kinetic aspects ofsinging. J Appl Physiol 21:483–496.

Bradshaw JL. 1988. The evolution of human lateralasymmetries: new evidence and second thoughts. JHum Evol 17:615–637.

Brown CH, Cannito MP. 1995. Modes of vocal variationin sykes monkey (Cercopithecus albogularis) squeals.J Comp Psychol 109:398–415.

Byrne R. 1981. Distance vocalisations of guinea baboons(Papio papio) in Senegal: an analysis of function.Behaviour 78:283–312.

Byrne R. 1982. Primate vocalizations: structural andfunctional approaches to understanding. Behaviour80:241–255.

Calvin WH. 1982. Did throwing stones shape hominidbrain evolution? Ethol Sociobiol 3:115–124.

Calvin WH. 1992. Evolving mixed-media messages andgrammatical language: secondary uses of neural se-quencing machinery needed for ballistic movements.In: Wind J, Chiarelli B, Bichakjian B, Nocentini A,editors. Language origin: a multidisciplinary ap-proach. Dordrecht: Kluwer Academic Publishers. p163–179.

Campbell EJM. 1968. The respiratory muscles. Ann NYAcad Sci 155:135–140.

Campbell EJM. 1974. Muscular activity in normal andabnormal ventilation. In: Wyke B, editor. Ventilatoryand phonatory control systems. London: Oxford Uni-versity Press. p 3–11.

Carrier D. 1984. The energetic paradox of human run-ning and hominid evolution. Curr Anthropol 25:483–495.

Clark AP, Wrangham RW. 1993. Accoustic analysis ofwild chimpanzee pant hoots: do Kibale forest chimpan-zees have an acoustically distinct arrival pant hoot?Am J Primatol 31:99–109.

Davis PJ, Zhang SP, Winkworth A, Bandler R. 1996.Neural control of vocalization—respiratory and emo-tional influences. J Voice 10:23–38.

Decima EE, von Euler C, Thoden U. 1967. Spinalintercostal-phrenic reflexes. Nature 214:312–313.

Deputte BL. 1982. Duetting in male and female songs ofthe white-cheeked gibbon (Hylobates concolor leucog-enys). In: Snowdon CT, Brown CH, Petersen MR,editors. Primate communication. Cambridge: Cam-bridge University Press. p 67–93.

Draper MH, Ladefoged P, Whitteridge D. 1959. Respira-tory muscles in speech. J Speech Hear Res 2:16–27.

Dunbar RIM. 1993. Coevolution of neocortical size,group size and language in humans. Behav Brain Sci16:681–735.

Falk D. 1980. A reanalysis of the South African australo-pithecine natural endocasts. Am J Phys Anthropol53:525–539.

Fifer FC. 1987. The adoption of bipedalism by thehominids: a new hypothesis. Hum Evol 2:135–147.

Fitch WT, Hauser MD. 1995. Vocal production in nonhu-man primates: acoustics, physiology, and functionalconstraints on ‘‘honest’’ advertisement. Am J Primatol37:191–219.

Geissmann T. 1993. Evolution of communication ingibbons. Ph. D. thesis, University of Zurich.

Gittins SP. 1984. The vocal repertoire and song of theagile gibbon. In: Preuschoft H, Chivers DJ, Brockle-man WY, Creel N, editors. The lesser apes: evolution-ary and behavioural biology. Edinburgh: EdinburghUniversity Press. p 354–375.


Gould WJ, Okamura H. 1974. Interrelationships be-tween voice and laryngeal mucosal reflexes. In: WykeB, editor. Ventilatory and phonatory control systems.London: Oxford University Press. p 347–369.

Grassino AE, Goldman MD. 1986. Respiratory musclecoordination. In: Macklem PT, Mead J, editors. Hand-book of physiology: the respiratory system. Bethesda,MD: American Physiological Society. p 463–480.

Green S. 1975. Variation in vocal pattern with socialsituation on the Japanese monkey (Macaca fuscata): afield study. In: Rosenblum LA, editor. Primate behav-iour: developments in field and laboratory research.New York: Academic Press. p 1–102.

Haimoff EH. 1983. Occurrence of anti-resonance in thesong of the siamang (Hylobates syndactylus). Am JPrimatol 5:249–256.

Haimoff EH. 1984. Acoustic and organizational featuresof gibbon songs. In: Preuschoft H, Chivers DJ, Brockl-eman WY, Creel N, editors. The lesser apes: evolutionand behavioural biology. Edinburgh: Edinburgh Uni-versity Press. p 333–353.

Haimoff EH. 1986. Convergence in the duetting ofmonogamous Old World primates. J Hum Evol 15:51–59.

Haimoff EH, Yang XJ, He SJ, Chen N. 1987. Preliminaryobservations of wild black-crested gibbons (Hylobatesconcolor concolor) in Yunnan Province, People’s Repub-lic of China. Primates 28:319–335.

Harcourt AH, Stewart KJ, Hauser M. 1993. Functions ofwild gorilla close calls. 1. Repertoire, context, andinterspecific comparison. Behaviour 124:89–122.

Hardy A. 1960. Was man more aquatic in the past? NewScientist 7:642–645.

t’Hart J. 1986. Declination has not been defeated—areply to Lieberman et al. J Acoust Soc Am 80:1838–1840.

Hauser MD. 1991. Sources of acoustic variation inrhesus macaque (Macaca mulatta) vocalizations.Ethology 89:29–46.

Hauser MD, Fowler CA. 1992. Fundamental frequencydeclination is not unique to human speech: evidencefrom nonhuman primates. J Acoust Soc Am 91:363–369.

Hauser MD, Marler P. 1992. How do and should studiesof animal communication affect the interpretation of achild’s phonological development? In: Ferguson C,Menn L, Stoel-Gammon C, editors. Phonological devel-opment. Timonium, MD. York Press. p 663–680.

Hauser MD, Evans CS, Marler P. 1993. The role ofarticulation in the production of rhesus monkey, Ma-caca mulatta, vocalizations. Anim Behav 45:423–433.

Haworth JB, Keillor GW. 1962. Use of transparencies inevaluating the width of the spinal canal in infants,children and adults. Radiology 79:109–114.

Hewes GW. 1973. Primate communication and the ges-tural origin of language. Curr Anthropol 14:5–24.

Hewes GW. 1992. Primate communication and the ges-tural origin of language. Curr Anthropol 33:65–84.

Hinck VC, Hopkins CE, Savara BS. 1966. Sagittaldiameter of the cervical spinal canal in children andadults. Radiology 79:97–108.

Hixon TJ, Putnam AHB. 1983. Voice disorders in rela-tion to respiratory kinematics. Semin Speech Lang4:217–231.

Hixon TJ, Weismer G. 1995. Perspectives on the Edin-burgh study of speech breathing. J Speech Hear Res38:42–60.

Hixon TJ, Mead J, Goldman MD. 1976. Dynamics of thechest wall during speech production: function of thethorax, rib cage, diaphragm and abdomen. J SpeechHear Res 19:297–356.

Hoit JD, Plassman BL, Lansing RW, Hixon TJ. 1988.Abdominal muscle activity during speech production.J Appl Physiol 65:2656–2664.

Hoit JD, Banzett RB, Brown R, Loring SH. 1990a.Speech breathing in individuals with cervical spinalcord injury. J Speech Hear Res 33:798–807.

Hoit JD, Hixon TJ, Watson PJ, Morgan WJ. 1990b.Speech breathing in children and adolescents. J SpeechHear Res 33:51–69.

Hoit JD, Shea SA, Banzett RB. 1994. Speech productionduring mechanical ventilation in tracheostomized in-dividuals. J Speech Hear Res 37:53–63.

Holloway RL. 1983. Human palaeontological evidencerelevant to language behaviour. Hum Neurobiol 2:105–114.

Isaac GL. 1976. Stages of cultural elaboration in thePleistocene: possible archeological indicators of thedevelopment of language. Ann NY Acad Sci 280:275–288.

Iscoe S. 1998. Control of abdominal muscles. ProgNeurobiol 56:433–506.

Jurgens U. 1982. Amygdalar vocalization pathways inthe squirrel-monkey. Brain Res 241:189–196.

Jurgens U. 1988. Central control of monkey calls. In:Todt D, Goedeking P, Symmes D, editors. Primatevocal communication. Berlin: Springer-Verlag. p 162–167.

Jurgens U, Kirzinger A. 1982. The effects of lesions inthe cortical face area, supplementary motor cortexand anterior cingulate cortex on vocalization in themonkey. Behav Brain Res 5:105.

Jurgens U, Kirzinger A. 1984. The contribution ofdifferent brain-areas in vocalization—a lesioning studyin the squirrel-monkey. Int J Primatol 5:351.

Jurgens U, Richter K. 1986. Glutamate-induced vocaliza-tion in the squirrel-monkey. Brain Res 373:349–358.

Jurgens U, Schreiver S. 1991. Respiratory muscle activ-ity during vocalization in the squirrel monkey. FoliaPrimatol (Basel) 56:121–132.

Kay RF, Cartmill M, Balow M. 1998. The hypoglossalcanal and the origin of human vocal behaviour. ProcNatl Acad Sci USA 95:5417–5419.

Kien J, KempA. 1994. Is speech temporally segmented—comparison with temporal segmentation in behavior.Brain Lang 46:662–682.

Kirzinger A, Jurgens U. 1994. Role of extralaryngealmuscles in the phonation of subhuman primates. JComp Physiol 175:215–222.

Krings M, Stone A, Schmitz RW, Krainitzki H, Stonek-ing M, Paabo S. 1997. Neanderthal DNA sequencesand the origin of modern humans. Cell 90:19–30.

Ladefoged P. 1968. Linguistic aspects of respiratoryphenomena. Ann NY Acad Sci 155:141–151.

Laitman JT. 1985. Evolution of the hominid upperrespiratory tract: the fossil evidence. In: Tobias PV,editor. Hominid evolution, past, present and future.New York: Alan R. Liss, Inc. p 281–286.

Laitman JT, Reidenberg JS, Gannon PJ. 1992. Fossilskulls and the hominid vocal tract: new approaches tocharting the evolution of human speech. In: Wind J,Chiarelli B, Bichakjian B, Nocentini A, editors. Lan-guage origin: a multidisciplinary approach. Dor-drecht: Kluwer Academic Publishers. p 385–397.

Larson CR, Yajima Y, Ko P. 1994. Modification in activityof medullary respiratory-related neurons for vocalisa-tion and swallowing. J Neurophysiol 71:2294–2304.

Lieberman MR, Lieberman P. 1973. Olson’s ‘‘projectiveverse’’ and the use of breath control as a structuralelement. Lang Style 5:287–298.

Lieberman P. 1968. Primate vocalisations and humanlinguistic ability. J Acoust Soc Am 44:1574–1584.

Lieberman P. 1984. The biology and evolution of lan-guage. Cambridge, MA: Harvard University Press.


Lieberman P. 1985. The physiology of cry and speech inrelation to linguistic behavior. In: Lester BM, Bouky-dis CFZ, editors. Invant crying. New York: Plenum.

Lieberman P. 1991. Uniquely human: the evolution ofspeech, thought, and selfless behavior. Cambridge,MA: Harvard University Press.

Lieberman P, Crelin ES. 1971. On the speech of Neander-thal man. Linguistic Enquiry 11:203–222.

Lieberman P, Laitman JT, Reidenberg JS, Gannon PJ.1992. The anatomy, physiology, acoustics and percep-tion of speech—essential elements in analysis of theevolution of human speech. J Hum Evol 23:447–467.

Lovejoy CO. 1988. Evolution of human walking. Sci Am259:118–125.

MacLarnon A. 1987. Size relationships of the spinal cordand associated skeleton in primates. Ph.D. thesis,University of London.

MacLarnon A. 1993. The vertebrate canal. In: Walker A,Leakey R, editors. The Nariokotome Homo erectusskeleton. Cambridge, MA: Harvard University Press.p 359–390.

MacLarnon A. 1995. The distribution of spinal cordtissues and locomotor adaptations in primates. J HumEvol 29:463–482.

MacLarnon A, Hewitt GP. 1995. The hominid vertebralcanal and the evolution of human speech. Am J PhysAnthropol [Suppl] 20:139 [abstract].

Marler P, Hobbett L. 1975. Individuality in long-rangevocalisation of wild chimpanzees. Z Tierpsychol 38:97–109.

Marler P, Tenaza R. 1977. Signalling behaviour of apeswith special reference to vocalizations. In: Sebeok TA,editor. How animals communicate. Bloomington: Indi-ana University Press. p 965–1033.

Martin R. 1928. Lehrbuch der Anthropologie. Jena:Gustav Fischer.

Martin RD, Barbour AD. 1988. Aspects of line-fitting inbivariate allometric analyses. Folia Primatol (Basel)53:424–468.

McFarland DH, Lund JP. 1995. Modification of mastica-tion and respiration during swallowing in the adulthuman. J Neurophysiol 74:1509–1517.

McFarland DH, Lund JP, Gagner M. 1994. Effects ofposture on the coordination of respiration and swallow-ing. J Neurophysiol 72:2431–2437.

McHenry HM. 1992. Body size and proportions in earlyhominids. Am J Phys Anthropol 87:407–431.

Mead J, Bouhuys A, Proctor DF. 1968. Mechanismsgenerating subglottic pressure. Ann NY Acad Sci155:177–182.

Mead J, Hixon TJ, Goldman MD. 1974. The configura-tion of the chest wall during speech. In: Wyke B,editor. Ventilatory and phonatory control systems.London: Oxford University Press. p 58–75.

Milic-Emili J, Zin WA. 1986. Relationships betweenneuromuscular respiratory drive and ventilatory out-put. In: Macklem PT, Mead J, editors. Handbook ofphysiology: the respiratory system. Bethesda, MD:American Physiological Society. p 631–646.

Mitani J, Marler P. 1989. A phonological analysis of malegibbon singing behaviour. Behaviour 109:20–45.

Mitchell HI, Hoit JD, Watson PJ. 1996. Cognitive linguis-tic demands for speech breathing. J Speech Hear Res39:93–104.

Mithen S. 1996. The prehistory of mind: a search for theorigins of art, religion and science. London: Thames &Hudson.

Morgan E. 1972. The descent of woman. London: Souve-nir Press.

Morgan E. 1982. The aquatic ape. London: SouvenirPress.

Morgan E. 1991. Why a new theory is needed. In: RoedeM, Wind J, Patrick JM, Reynolds V, editors. Theaquatic ape: fact or fiction. London: Souvenir Press(E&A), Ltd.

Murphy K, Corfield DR, Guz A, Fink GR, Wise RJS,Harrison J, Adams L. 1997. Cerebral areas associatedwith motor control of speech in humans. J ApplPhysiol 83:1438–1447.

Negus VE. 1949. The comparative anatomy and physiol-ogy of the larynx. London: William Heineman MedicalBooks, Ltd.

Newsom Davis J, Sears TA. 1970. The proprioceptivereflex control of the intercostal muscles during theirvoluntary activation. J Physiol [Lond] 209:711–738.

Noble W, Davidson I. 1991. The evolutionary emergenceof modern human behaviour: language and its archae-ology. Man 26:223–253.

Pagel MD, Harvey PH. 1988. Recent developments inthe analysis of comparative data. Quart Rev Biol63:413–440.

Patrick JM. 1991. Human respiratory adaptations forswimming and diving. In: Roede M, Wind J, PatrickJM, Reynolds V, editors. The aquatic ape: fact orfiction? London: Souvenir Press (E&A) Ltd., p 237–246.

Petersen SE, Fox PT, Posner MI, Mintun M, RaichleME. 1988. Positron emission tomographic studies ofthe cortical anatomy of single-word processing. Na-ture 331:585–589.

Petersen SE, Fox PT, Posner M, Mintun M, Raichle ME.1989. Positron emission tomographic studies of theprocessing of single words. J Cogn Neurosci 1:163–170.

Phillipson EA, McLean PA, Sullivan CE, Zamel M. 1978.Interaction of metabolic and behavioural respiratorycontrol during hypercapnia and speech.Am Rev RespirDis 117:903–909.

Pinker S, Bloom P. 1990. Natural language and naturalselection. Behav Brain Sci 13:707–784.

Proctor DF. 1974. Breathing mechanics during phona-tion and singing. In: Wyke B, editor. Ventilatory andphonatory control systems. London: Oxford Univer-sity Press. p 39–57.

Provine RR. 1996. Laughter. Am Sci 84:38–45.Purves MJ. 1979. What do we breathe for? In: von Euler

C, editor. Central nervous control mechanisms inbreathing. Oxford: Pergamon Press. p 7–12.

Raemaekers JJ, Raemaekers PM. 1985. Field playbackof loud calls to gibbons (Hylobates lar): territorial,sex-specific and species-specific responses. Anim Be-hav 33:481–493.

Raemaekers JJ, Raemaekers PM, Haimoff EH. 1984.Loud calls of the gibbon (Hylobates lar): repertoire,organisation and context. Behaviour 91:146–189.

Robinson JG. 1979. An analysis of the organization ofvocal communication in the titi monkey Callicebusmoloch. Z Tierpsychol 49:381–405.

Romanes GJ. 1984. Cunningham’s manual of practicalanatomy: combined fourteenth edition. London: Ox-ford University Press.

Rosenberg KR. 1992. The evolution of modern child-birth. Yrbk Phys Anthropol 35:89–124.

Ruff CB, Walker A. 1993. Body size and body shape. In:Walker A, Leakey R, editors. The Nariokotome Homoerectus skeleton. Cambridge, MA: Harvard UniversityPress. p 234–265.

Ruff CB, Trinkaus E, Holliday TW. 1997. Body mass andencephalization in Pleistocene Homo. Nature 387:173–176.


Savage-Rumbaugh ES, Murphy J, Sevcik RA, BrakkeKE, Williams SL, Rumbaugh DM. 1993. Languagecomprehension in ape and child. Monogr Soc ResChild Dev 58:5.

Schick KD, Toth N. 1993. Making silent stones speak.London: Weidenfeld & Nicolson.

Sears TA, Newsom Davis J. 1968. The control of respira-tory muscles duringvoluntary breathing. Ann NYAcad Sci 155:183–190.

Sekulic R, Chivers DJ. 1986. The significance of callduration in howler monkeys. Int J Primatol 7:183–190.

Seyfarth RM, Cheney DL. 1992. Meaning and mind inmonkeys. Sci Am 267:78–84.

Sharp JT, Hyatt RE. 1986. Mechanical and electricalproperties of respiratory muscles. In: Macklem PT,Mead J, editors. Handbook of physiology: the respira-tory system. Bethesda, MD: American PhysiologicalSociety. p 389–414.

Smith BH. 1993. The physiological age of KNM-WT15000. In: Walker A, Leakey R, editors. The Narioko-tome Homo erectus skeleton. Cambridge, MA: Har-vard University Press. p 195–200.

Stahl WR. 1967. Scaling of respiratory variables inmammals. J Appl Physiol 22:453–460.

Stathopoulos ET, Sapienza C. 1993. Respiratory andlaryngeal function of women and men during vocalintensity variation. J Speech Hear Res 36:64–75.

Steklis HD. 1985. Primate communication, comparativeneurology and the origin of language re-examined. JHum Evol 14:157–173.

Steklis HD, Raleigh MJ. 1979. Requisites for language:interspecific and evolutionary aspects. In: Steklis HD,Raleigh MJ, editors. Neurobiology of social communi-cation in primates: an evolutionary perspective. NewYork: Academic Press. p 283–314.

Stringer CB, Andrews P. 1988. Genetic and fossil evi-dence for the origin of modern humans. Science 239:1263–1268.

Sundberg J, Titze I, Scherer R. 1993. Phonatory controlin male singing—a study of the effects of subglottalpressure, fundamental-frequency, and mode of phona-tion on the voice source. J Voice 7:15–29.

Sutton D, Jurgens U. 1988. Neural control of vocalisa-tion. In: Steklis HD, Erwin J, editors. Comparativeprimate biology, neurosciences. New York: Alan R.Liss, Inc. p 625–647.

Swisher CC, Curtis GH, Jacob AG, Getty A, Suprijo A,Widiasmoro. 1994. Age of the earliest known hominidsin Java, Indonesia. Science 263:1118–1121.

Taylor A. 1960. The contribution of the intercostalmuscles to the effort of respiration in man. J Physiol[Lond] 151:390–402.

Tenaza R. 1976. Songs, choruses and countersinging of

Kloss’ gibbons (Hylobates klossi) in Siberut Island,Indonesia. Z Tierpsychol 40:37–52.

Thalmann U, Geissmann T, Simone A, Mutschler T.1993. The indris of Anjanaharibe-Sud, NortheasternMadagascar. Int J Primatol 14:357–381.

Tobias PV. 1987. The brain of Homo habilis: a new levelof organisation in cerebral evolution. J Hum Evol16:741–761.

Trinkaus E. 1984. Comment on: Human running andhominid evolution, by D.R. Carrier. Curr Anthropol25:491.

Verhaegen M. 1995. Aquatic ape theory, speech origins,and brain differences with apes and monkeys. MedHypotheses 44:409–413.

von Euler C. 1968. The proprioceptive control of thediaphragm. Ann NY Acad Sci 155:204–205.

Walker A. 1993. Perspectives on the Nariokotome discov-ery. In: Walker A, Leakey R, editors. The NariokotomeHomo erectus skeleton. Cambridge, MA: HarvardUniversity Press. p 411–430.

Walker A, Shipman P. 1996. The wisdom of bones: insearch of human origins. London: Weidenfeld & Nicol-son.

Waser PM. 1982. The evolution of male loud calls amongmangabeys and baboons. In: Snowdon CT, Brown CH,Petersen MR, editors. Primate communication. Cam-bridge: Cambridge University Press. p 117–143.

West R, Larson CR. 1993. Laryngeal and respiratoryactivity during vocalization in macaque monkeys. JVoice 7:54–68.

Wilkins WK, Wakefield J. 1995. Brain evolution andneurolinguistic preconditions. Behav Brain Sci 18:161–182.

Winkworth AL, Davis PJ, Adams RD, Ellis E. 1995.Breathing patterns during spontaneous speech. JSpeech Hear Res 38:124–144.

Wolpoff M, Wu X, Thorne AG. 1984. Modern Homosapiens origins: a general theory of hominid evolutioninvolving the fossil evidence from East Asia. In: SmithFH, Spencer F, editors. The origin of modern humans:a world survey of the fossil evidence. New York: AlanLiss. p 411–483.

Zhang SP, Davis PJ, Bandler R, Carrive P. 1994. Brainstem integration of respiration and vocalisation: roleof the midbrain periaqueductal gray. J Neurophysiol72:1337–1356.

Zimmermann E. 1990. Differentiation of vocalisations inbushbabies and the significance for assessing phyloge-netic relationships. J Zool Syst Evol 28:217–239.

Zimmermann E, Bearder SK, Doyle GA, Andersson AB.1988. Variations in vocal patterns of Senegal andSouth African lesser bushbabies and their implica-tions for taxonomic relationships. Folia Primatol(Basel) 51:87–105.


Documents

The Evolution of Human Speech: The Role of Enhanced ... Evolution of Human Speech: The Role of Enhanced Breathing Control ANN M. MACLARNON* AND GWEN P. HEWITT School of Life Sciences,