PROCEEDINGSOF THE

NATIONAL ACADEMY OF SCIENCESVolume 21 January 15, 1935 Number I

ON THE EVOLUTION OF THE SKULLS OF VERTEBRATESWITH SPECIAL REFERENCE TO HERITABLE CHANGES

IN PROPORTIONAL DIAMETERS(ANISOMERISM)

By WILLIAM KING GREGORYAMERICAN MUSEUM OF NATURAL HISTORY, DEPARTmENT OF COMPARATIVE AND

HuMAN ANATOMY

Read before the Academy, Tuesday, November 20, 1934

PART I. THE SKULLS OF THE MOST PRIMITIVE KNOWN FOSSILCHORDATES (OSTRACODERMS)

In earlier papers I have directed attention to the fact that during thecourse of evolution of any series of organisms two opposite processes ofdevelopment and evolution may be observed. In the first process, whichhas been named polyisomerism,l there is a budding or multiplication ofsome given part or parts, such as the teeth of sharks or the joints of thebackbone in eels. This process has been recognized by earlier authorsunder such terms as "budding," "reduplication" (Cope), "metamerism"(Gegenbaur), "rectigradation" (Osborn, in part), "aristogenesis" (Osborn,in part).

In the opposite but often simultaneous process there is some emphasisor selective action among the polyisomeres, that is, there may be a localacceleration or retardation of growth rates, causing later polyisomeresto grow larger or smaller than earlier ones; or some part or axis of oneof the units may become the focus of a new acceleration or retardation;this produces lop-sided polyisomeres or anisomeres and the process itselfis called anisomerism. In positive anisomerism we have progressiveincrease of some particular part or spot; in negative anisomerism thereverse occurs.

This process has also been recognized in part by carlier authors undersuch terms as "differentiation" (Spencer), "allometry" (Osborn, in part),"heterogony" (Pezard, Huxley, in part), "acceleration" and "retardation"(Hyatt, in part). But no one of these terms has exactly the same contentas anisomerism. Doubtless in many cases the two processes combineand intergrade but when one is familiar with typical cases the intergradesbecome gradually recognizable.

Dow

nloa

ded

by g

uest

on

July

31,

202

0

ANATOMY: W. K. GREGORY

John Ray (1627-1705), one of the fathers of systematic zoology, includedin his tables of generic differentia various proportional characteristics, asthe short muzzle and rotund head of the cats and the long muzzle of thedogs. Likewise such metrical differences in human skulls as are denotedby the terms "dolichocephalic,"

C

FIGURE 1

Essential contrasts between Agnatha(A, B) and Gnathostomata (C). A.Under side of head of ostracoderm, show-ing supposed position of mouth and gillopenings. Restoration after Stensi6.B. Under side of head shield of ostra-coderm, showing roof of oralo-branchialcavity with dorsal part of branchialskeleton. None of the gill arches are en-larged to serve as jaws. Restoration afterStensio. C. Under side of head ofpalaeozoic shark (Cladosetache), showingenlarged jaws in series with gill-arches.After Dean.

important works by Kiaer, Stensio,

"brachycephalic," "prognathous,""orthognathous," and the like werenoticed by the founders of anthro-pology and have been investigatedever since by anthropometrists.But it remained for the palaeon-tologist Henry Fairfield Osborn in1902 to visualize "dolichocephaly"and "brachycephaly" as manifes-tations of deep-seated morpholog-ical factors that affected in oppo-site ways aimost every part of thecranium, jaws and teeth of mam-mals. In the present paper I shallshow that the same processes ofdivergent evolution that have re-sultedindolichocephalyandbrachy-cephaly among Tertiary mam-mals were already at work in theoldest known chordates of the Si-lurian period.Anisomerism in the Skulls of the

Most Primitive Known Chordates.-The American Museum of Nat-ural History has received from thePaleontological Museum of Oslo,through the kindness of the lateProfessor Kiaer, the gift of afine series of ostracoderms fromSpitzbergen and Norway. Thisand other material brought to-gether by the late Dr. BashfordDean, in addition to the manyBryant and others, have afforded

me the opportunity of using these oldest known chordates in the first ofa series of papers on the evolution of changes in proportion in the skulls ofvertebrates. These studies have been carried on in the American Mu-seum with the aid of graduate students of Columbia University.The ostracoderms or pre-fishes, which flourished from Middle Ordo-

2 PROC. N. A. S.

Dow

nloa

ded

by g

uest

on

July

31,

202

0

ANA TOMY: W. K. GREGORY

vician to Upper Devonian times, were the oldest and most primitive ofall known chordates and constituted a widely varying superclass technicallycalled Agnatha, or jawless forms. Although some of them had jaws formedof dermal plates, these plates were not supported by a pair of enlargedgill-arches as they are in the oldest true fishes and other gnathostomes, orvertebrates with gill-arch jaws. This essential contrast between thesetwo great superclasses is illustrated in the chart (Fig. 1). Some of theostracoderms fulfilled all the conditions of time and anatomical character-istics that might be expected in the remote ancestors of the existing cyclo-stomes (lampreys and hag-fishes). These far-reaching conclusions restupon the labors of a long line of palaeichthyologists, notably Pander,Rohon, Powrie and Lankester, Traquair, Patten, Kiaer, Stensio and others.The record begins in the Middle Ordovician with the single tuberculatedplate named Astraspis desiderata by Walcott; it then flowered out intothe rich ostracoderm faunas of Upper Silurian and Lower Devonian times,including at least four orders and a rapidly increasing list of reputedsuborders, families, genera and species; the class comes to an end in afew hold-overs in the Upper Devonian.The head shield of cephalaspids (Fig. 2) was subject to a wide range

of anisomerism in relative length and width and in the emphasis of itspaired cornua. In short, "brachycephaly" and "dolichocephaly" in abroad sense were in full swing among these oldest known chordates. InBenneviaspis the maximum breadth of the cephalic shield measured acrossthe posterior ends of its cornua was almost twice as great as its length inthe median line (Stensio, 1932, p. 155); in Hoelaspis angulata the breadthacross the cornua is nearly two and one-half times the head shield (Stensi6,1927, p. 292). In Kiaeraspis, on the other hand, the cephalic shield as awhole is "strikingly long, much longer than broad" (Stensio, 1927, p. 297).In Hemicyclaspis murchisoni there were no projecting cornua at thepostero-lateral angles of the shield (Stensio, 1932, p. 79). In Sclerodus(Eukeraspis), on the contrary, the cornua were very long, about twiceas long as the cephalic shield (Stensio, 1932, p. 178). The nuchal spineof the shield was directed but slightly upward in Cephalaspis whitei (Stensio,1932, p. 94) but swept upward almost vertically in Cephalaspis isachseni,implying a very high body.The Anaspida were small-headed, fish-like ostracolerms with a sinuous

body. Comparative measurements show that in Cephalaspis lyelli thebody behind the head is 1.68 times as long as the head, whereas in theanaspids, to judge from Kiaer's restorations, it ranges from 5.3 to ninetimes the length of the head. The top of the small head in anaspids isnot covered with a continuous shield but with apparently secondary smallplates and scutes arranged around the median olfactory, median pinealand paired optic openings. The "upper lip" in front of the median nostril

VOL. 21, 1935 3

Dow

nloa

ded

by g

uest

on

July

31,

202

0

BO/E/1 Sp/S

E XF

KIRERWSPIS

K.TR/MEHT/4SPIS

--~------ X !

HOfELASPIS010

F X-

CE/'/-S4LASP/S

; 0. 0

SCLERODIS

TH7YIYESTES

JD/IDYMASPI/S

fitHEJrIATEL9SSP/SFIGURE 2

Anisomerism of the head shield among the cephalaspids. Scales various. A. AfterTraquair. B, D, E, F, G, H, I, J. After Stensid. C. After Powrie and Lankester.K. After Patten. G is probably less specialized than F.

Dow

nloa

ded

by g

uest

on

July

31,

202

0

ANATOMY: W. K. GREGORY

was less expanded than in the cephalaspids; it was covered by a medianrostral and paired lateral rostral scutes (Stetson). The throat or floorof the expanded oralo-branchial chamber was covered with numerous smallscutes. The eyes were relatively large and laterally placed; this fact,together with the more fusiform body, implying active pursuit of smallprey.Thus the anaspids illustrate how a predaceous, externally fish-like form

could be derived from a primitive ostracoderm by easily understoodprocesses of secondary polyisomerism and anisomerism.The order Pteraspidomorphi or Heterostraci exhibits a primitive dorso-

ventral asymmetry and bilateral symmetry, with lateral position of thepaired eyes. There were five primary plates covering the head, onemedian rostral, one dorsal, one ventral and two lateral or branchial plates.In some forms median and lateral posterior spikes were taken over fromthe trunk region (Bryant, 1933). Among the true pteraspids anisomerism,as shown in Bryant's material (1933, Fig. 2), was rife, producing allextremes from "mesaticephalic" to "dolicho-" and "brachycephalic"types. The rostrum varied from short in Poraspis, Cyathiaspis, Anglaspis,etc., to the sturgeon-like rostrum of Podolaspis.The mouth of pteraspids, as shown by Kiaer (1928), was somewhat

comparable to that in myxinoids and bore maxillary and mandibularplates which were not connected with the branchial arches and thereforenot strictly homologous with the complex jaws of true gnathostomes.There was a transverse row of about eight or nine gular plates on eachside, implying considerable movement of the floor of the mouth. Therewas only a single gill opening on each side at the posterior end of theshield but there were about eight pairs of large pouches arranged in an an-tero-posterior series on either side of the mid-line.The Drepanaspis type was doubtless derived from the primitive pter-

aspids by a rapid transverse widening of the head of such a primitiveanaspid as Eoarchegonaspis wardelli (Bryant). Thus the small eyes werecarried out to the lateral borders and a cylindrical shield was changed intoa broad skate-like shield. The five primary plates remain but the fieldconnecting them is covered with small irregular polygonal plates essentiallysimilar to the large scales on the body. Through these anisomerouschanges the strongly benthonic body shrinks to seventy-five per cent ofthe length of the head shield.The step from Drepanaspis to the flattened Thelodus pagei involves the

breaking-up of the head shield into placoid tubercles. Such a fragmenta-tion, or secondary polyisomerism, culminates in the Coelolepidae (Kiaer,1932), which bave the appearance at first sight of being the least differen-tiated of all ostracoderms. And yet both Stensi6 and Kiaer have cited

VOL. 21, 1935 5

Dow

nloa

ded

by g

uest

on

July

31,

202

0

ANA TOMY: W. K. GREGOR Y

many facts tending to show that such fragmentation is quite secondaryin all branches of the ostracoderms (cf. Stensi6, 1932, pp. 195, 196).The question which is the most primitive of all the known Upper Silurian

ostracoderms has been advanced toward solution by the considerationsadduced by Kiaer (1928) and by Stensio (1932, pp. 185 et seq) concerningthe morphology of the supposed naso-hypophysial duct in ostracodermsand cyclostomes. This duct in existing lampreys opens dorsally in frontof the eyes but in embryonic stages it arises ventrally; then by the hyper-trophy of the post-hypophysial fold the duct is gra.dually shifted upwardso that it opens dorsally. According to Stensio, since there was no externalmedian opening of the naso-pharyngeal duct in pteraspids, this duct,which is present in all other known vertebrates, must have been on theunder side of the head and probably immediately below the anterior tip

..........

FIGURE 3

Anisomerism of the head shield among pteraspids and thelodonts. Scales various.A. Eoarchegonaspis wardelli. After Bryant. B. Drepanaspis gemuendenesis.From Smith Woodward, chiefly by Traquair. C. Thelodus pagei. After Traquair.

of the notochord, where the hypophysial pit arises in cyclostome embryos.In the pteraspids this naso-hypophysial tube remained on the under sidebut in the cephalaspids and their successors the petromyzonts the naso-hypophysial tube was displaced on to the dorsal surface through the enor-mous hypertrophy of the post-hypophysial fold, which expands into asort of upper lip.Such an overgrowth of the prenarial portion of the cephalaspid shield

may well have been connected with the radial and peripheral growth ofthe so-called "electric nerves," which in several of Stensi6's plates (1927,pls. 45, 49) may be seen dividing and subdividing in a radial peripheralway. In any case the possession of this enormous "upper lip," analogousperhaps in sensory function to the expanded rostrum of the spoonbillsturgeons (Polyodon), was probably a point of departure for specializationsof selective value in cephalaspids. The same peripheral spreading of theupper lip would cause the dorsal position of the paired eyes. The result

6 PRoc. N. A. S.

Dow

nloa

ded

by g

uest

on

July

31,

202

0

ANATOMY: W. K. GREGORY

of this expansion of the upper lip is to produce a subcircular head shieldin the typical cephalaspid and an entirely misleading convergent resem-blance to skates and rays, in which the disc is due to expansion of thepectoral fins.

This marked anisomerism of the anterior part of the head shield is thedominant feature of the cephalaspid, and it no doubt conditioned alsothe sub-circular arrangement of the gill-pouches on the under side of theshield.The foregoing application of the principles of polyisomerism and ani-

somerism to the study of the evolution of the oldest known chordates natu-rally brings up the question of the origin of Amphioxus and other pre-chordates. In the light of present evidence it appears that Amphioxusrepresents a very highly specialized and in many respects simplifiedderivative of some such ostracoderms as the Anaspida. Extreme secondarypolyisomerism is seen in the structure of the gill-arches and in the multi-plication of the myomeres, while sharp negative anisomerism has reducedthe brain almost to the vanishing-point. The nakedness of the bodymight also be expected in highly specialized modern derivatives of theoriginally short-bodied and well armed ostracoderms.Summary and Conclusions.-(1) The most primitive known ostra-

coderms were probably such forms as Poraspis and Palaeaspis; thesehad already advanced far from the possibly proto-echinoderm ancestorsof the chordates; however, they retained no evident trace of quinque-radiate symmetry but were already dorso-ventrally anisomerous andbilaterally polyisomerous.

(2) The head shield of the pteraspids, involving a many-layered skin,was divided into five primary plates: one median rostral, two lateral orbranchial, one median dorsal and one ventral. Both the head shield as awhole and the individual plates were subject to marked anisomerism inthe divergent derivatives of the primitive pteraspid. At one extremestands the long and narrow Podalaspis, at the other the well roundedProtaspis perlatus Bryant.

(3) The thelodonts and coelolepids may be regarded as excessivelywidened pteraspids derived perhaps from a Drepanaspis-like ancestor butwith fragmented or secondarily polyisomerous exoskeleton.

(4) The naso-hypophysial opening of cephalaspids, which is on thedorsal surface of the head shield in front of the eyes, reached its position,according to Stensio, in the same way as does the naso-hypophysial openingof the recent lampreys (petromyzonts). That is, it probably arose inthe embryo on the upper surface of the buccal cavity but opened down-ward. Through the great expansion of the post-hypophysial fold, thenaso-hypophysial pit became displaced forward, so that what was originallyits lower end came to open on the top of the head.

VOL. 21, 1935 7

Dow

nloa

ded

by g

uest

on

July

31,

202

0

ANATOMY: W. K. GREGORY

(5) The great expansion, or positive anisomerism, of the oral marginof the head shield of cephalaspids may have been associated with radialgrowth and secondary polyisomerism of the so-called "electric nerves,"the canals of which, as preserved in many of Stensi6's specimens, are seento divide and subdivide according to a regular system.

(6) Anisomerism of the head shield within the order Osteostraci pro-duced on one hand extremely wide short head shields, as in Benneviaspis,and on the other, the long narrow shield of Hemicyclaspis. Fragmentationof the shield led to the tuberculate exoskeleton of Thyestes and Sclerodus.The pectoral cornua, absent in Hemicyclaspis, became exceedingly long inSclerodus (Eukeraspis). In the vertical axis most cephalaspid headshields were low and broad but in Cephalaspis isachseni the nuchal crestswept steeply upward into a high body.

(7) The anaspids seem to have been derived from primitive cephal-aspids, as held by Stensi6, but by secondary polyisomerism of body seg-ments and marked anisomerous reduction of the head shield they weretransformed into relatively long, slender and sinuous fishes of rapid flightand probably predaceous habit. Meanwhile the head shield becamemuch fragmented and the throat covered with small scutes.

(8) The two existing cyclostome orders, petromyzonts and myxinoids,retain many qualitative, deep-seated morphological characters fromancestral ostracoderms but have carried still further the secondary poly-isomerism of body segments, so that the body may be more than twelvetimes as long as the head. Myxinoid embryos, according to Stensio, aremore primitive in retaining the inferior position of the naso-hypophysialduct and so may have been derived from the pteraspids; the petromyzontshave shifted the naso-pharyngeal duct on to the top of the head and retainpart of the cephalaspid head shield or expanded "upper lip."

(9) Amphioxus may be a greatly degraded anaspid ostracoderm whichhas completely lost its head shield and suffered marked anisomerousgrowth of the notochord, with secondary polyisomerism of the branchialbasket. The brain has become extremely small. Its nakedness is to beexpected in a specialized derivative of primitively armored forms.

1 Gregory, William K., "Polyisomerism and Anisomerism in Cranial and DentalEvolution among Vertebrates," Proc. Nat. Acad. Sci., 20, 1-9, January (1934).

8 PROC. N. A. S.

Dow

nloa

ded

by g

uest

on

July

31,

202

0