How and above all where did dinosaurs live ?

               Biomechanics of the Vertebral Column and Implications for the Lifestyle of Dinosaurs                                                                                         and Certain Pelycosaurs

For more detailed information refer to the corresponding article in German

Abstract : The investigation of the vertebral column of extinct vertebrates by means of a biomechanical approach yields strong indi- cations that length, distribution, and orientation of the spinous processes in the region between shoulder girdle and pelvis can be used as a reliable measure as to whether such animals were bipeds or not and whether they were adapted to live preferably on land or possibly had an amphibious lifestyle. In comparison to modern ungulate mammals, in which length and orientation of the spinous processes especially in the shoulder area can be regarded as an adaptation to the ability of running and jumping, the vertebral col- umns of vertebrates of previous epochs exhibit remarkable divergences from this design. These are considered as indications of diffe- rent lifestyles and less specialized locomotory capabilities. The most important results of these biomechanical considerations are that the dorsal sails of the pelycosaurs Dimetrodon and Edaphosaurus reflect an adaptation to quadrupedal locomotion on land. On the other hand, only few theropods and sauropods suggest a terrestrial lifestyle. Saurischian dinosaurs appear to have lived prefe- rably in moderately shallow water, whereas most ornithischians were quadrupedal land-dwellers.

1. Introduction

The spinal column represents a most important skeletal characteristic of vertebrates. Its shape is subject to remarkable differences within the classes of Reptilia and Mammalia.These differences can be regarded as a result of adaptations to loads of different mag- nitude as well as to different kinds of locomotion and attainable speeds. Details of the spinal column can yield important clues as to its main load distribution and consequently its special main task.

Using a biomechanical approach, recent investigations by Ebel et al. (1998) have demonstrated that the shape of vertebral columns is suitable to serve as an excellent criterion to obtain unmistakable indications as to whether an extinct vertebrate primarily used a bipedal or a quadrupedal gait. For the enigmatic rauisuchian reptile Ctenosauriscus koeneni (v. Huene) we could make probable that, contrary to former ideas, the long spinous processes did not serve functions such as thermoregulation or imposing, as incidentally mentioned by Krebs (1969). The investigation of this remarkable vertebral column concerning its function led us to the conclusion that it does not represent an accidental freak of nature, rather its shape follows clearly from biomechanical requirements. The directions of the upper tips of the dorsal processes pass through a common intersection. This point of intersection corresponds to the former position of the knee joint. In addition, the otherwise rare trait of a remarkable rectangular cross-section of the in part strongly curved spinous processes points to the capability of transmitting bending forces. Presumably, a facultative bipedal locomo- tion on land became only possible in Ctenosauriscus by the evolution of this extremely specialized vertebral column.                                                                                                                                                                                                                        In connection with that investigation further questions arose, namely concerning the significance of strongly elongated spinous pro- cesses in further fossil forms such as the pelycosaur Dimetrodon and certain ornithischian dinosaurs on the one hand and the possi- ble reasons of rather short spinous processes in bipedal theropod dinosaurs on the other. The explanation of the vertebral column in Ctenosauriscus as a result of an adaptation to locomotory requirements suggests the conclusion that the underlying physical prin- ciples should likewise be transferable to special developments in other forms. As will be discussed later on, the application of mechanical principles to the vertebral columns of dinosaurs and certain pelycosaurs allows an approach to the lifestyle of these extinct vertebrates from an entirely independent viewpoint which, therefore, does not necessarily lean on former suppositions and textbook opinions, mostly dating back to the nineteenth century. Such suppositions were in agreement with the restricted knowled- ge of those days, but in the meantime there has been some progress in natural sciences and applied physics. This view suggests an idea of extinct vertebrates which is quite different in some respects, but much less spectacular compared to suppositions propa- gated in various semipopular publications. Nevertheless it does not contradict observational evidence and opens most interesting aspects.

2. Problems associated with biomechanics in biology and palaeontology

Before pointing out the regularities and conclusions to be derived from an examination of vertebral columns as to their biomechanical adaptation, I would like to shed some light on problems with which palaeontology is burdened and by which it is seriously handi- capped as to its possibilities to find actually plausible functional explanations. Such problems, unfortunately, make interdisciplinary activities unnecessarily difficult, but these are nevertheless desirable.                                                                                        Although most important for all vertebrates, for a long time the vertebral column has attracted little attention only by palaeontologists. Abel (1912) barely mentioned it in his palaeobiological investigations of vertebrates. Apparently, remarkable differences of the shape of the various extinct vertebrates did not allow a detailed explanation beyond the commonly assumed carrying function. Certainly, the anatomy of the human skeleton has intensively been studied, e.g. by Pauwels (1965), but it belongs to an upright land-dwelling biped and, therefore, differs remarkably from the majority of vertebrates and is not well suitable for comparisons. Generally, aspects of biomechanical effects concerning the functional interpretation of skeletons of fossil vertebrates, unfortunately, do not play an important part in traditionally historically oriented sciences such as palaeontology, and if considered are often misinterpreted. Generally, palaeontologists rely mainly on morphological comparisons with extant forms and on functional morphology presumpti- ons. Even today only few data on vertebral columns as well as on shape and dimensions of the vertebrae of extinct animals are avai- lable (Bailey 1997). Although the anatomist H. Virchow (1914) in his studies of vertebral columns regretted a missing interdisciplinary co-operation with technically trained people and Wainwright et al. (1976) pleaded for the utilization of mechanical principles in biolo- gy, nevertheless starting points based on applied physics as well as the employment of seemingly unusual methods are sometimes regarded with a certain unpleasance, e.g. Erben (1975: 113-114), and with utter disbelief. A missing familiarity with mechanical prin- ciples can even provoke unjustified and polemic reactions such as by Pfretzschner (1999). Surely, such reservations are caused by a different educational background. For this reason biomechanical aspects of fossil and even of modern skeletons are commonly poorly understood. Yet, this is merely a consequence of the inadequate methods employed so far.                                                The term of biomechanics is utilized in several ways. In this context it is meant to express the application of principles of mechanics to the skeleton or hard parts of Recent and extinct animals.

The evolution of a bony skeleton is nature’s highly admirable response to biomechanical requirements which only created the pre- condition for vertebrates to leave the water. Apart from certain protective functions the main task of bones consists in the trans- mission of forces, in particular compressive ones. But contrary to the assumptions in bridge analogies still in vogue, bones are not especially suitable for the transmission of bending moments, although bending loads do happen. Waisted spines (Bailey 1997) point to axial loads, whereas an adaptation to strong bending forces would generally require a triangular spine shape. A remarkable exception is offered by the conspicuous rectangular cross-section of the spinous processes of Ctenosauriscus which clearly points to acting bending moments within the so-called backsail. Generally, bones serve the attachment of muscles and only make possible strong muscular tensile forces and complicated movements such as those of arms and legs. If in the course of an adaptive develop- ment remarkable skeletal modifications occur, the investigation of biomechanical reasons should, therefore, have priority. Parsimony considerations as well suggest this procedure. However, commonly such features are explained by functions which also might play a role, but surely a subordinate one, and then in most cases are not provable or disprovable. Occasionally, the missing close fami- liarity of palaeontologists or biologists with physical principles and applied mechanics leads to a mere presumption of functions and thus to the utilization of a trial-and-error method, which cannot demonstrate its plausibility. A general knowledge of functional mor- phology and a profound one of anatomy is per se not sufficient to unveil the functional background of a skeletal design. Palaeontologists and evolutionists have been dealing with historical-narrative explanations for more than 100 years (Bock 1985), although rarely successful. Seemingly reasonable conclusions based on common sense considerations can easily be misleading (Regal 1985, Ebel 1999). Objective criteria are preferable. Although the recognition of biomechanical functions and of acting forces requires some thorough experience in the field of applied mechanics, the familiarity with such methods can offer much progress in understanding the functional design of extinct vertebrates and in reconstructing their lifestyle. Since palaeontology is a branch of natural sciences, some areas are accessible to exact methods, in this case to mechanical regularities. Palaeontologists can hardly find correct answers if they confine themselves to their normally applied methods and thus exclude important aspects from conside- ration. But there is no reason for resignation as some statements might suggest, if a functional explanation cannot easily be found. Vertebrates with a seemingly curious skeleton such as the Triassic Tanystropheus did not survive for some time despite their extre- me shape, as sometimes presumed, e.g. by Sander (1994). Certainly, there is no overspecialization, and although specialized, such forms were well adapted to their food sources.

Methods are available which can yield unmistakable evidence. Hertel (1963) has outlined some applicable procedures. However, such methods have barely ever been applied. The regrettable neglect is very hampering, and persisting uncertainties are reflected by the various competing hypotheses about the lifestyle of extinct vertebrates and never-ending controversial debates among the respective adherents. For example, Bailey (1997) has recently dealt with virtually the same problem as the author of this paper, namely significance and function of elongated spinous processes in the pelycosaur Dimetrodon, in the iguanodontid Ouranosaurus and in other dinosaurs, based on a comparison with Pleistocene ungulate mammals. Although he discussed a likely adaptation to biomechanical loads on the vertebral column of Ouranosaurus as well as the problematical cantilever-bridge analogy of Thompson (1942) and weighty objections against it by Slijper (1946), moreover mentioned the likewise questionable bowstring-bridge analogy, nevertheless he prefers the conclusion that the elongation of spinous processes in the forms in question is primarily linked with the formation of a hump, as present today for instance in Camelus or Bison, which he believes to have served a fat-storage function and enabled the animals to bridge long distances without sufficient food supply during seasonal migrations. Although he reveals some scepticism as to the ”new image” of pretended warm-blooded dinosaurs, nevertheless in his study he only sets arguments that seem to support a hump function against an unlikely thermoregulatory sail function. The idea that elongated spinous processes are linked with biomechanical loads corresponds to my thorough conviction, too, but the assumption of a predominant hump function cannot be substantiated. Basically, the elongation of the spinous processes in the region of the withers of mammals is entirely independent of the formation of a hump. If some mammals such as Bison store fat in this area this may be an accidental coincidence, but certainly it is not the cause of elongated processes. Although elongated and characteristically arranged spinous processes can be found in all ungulate mammals, the presence of a pronounced hump is rather a rare trait and not necessarily linked with the vertebrae of the shoulder area. As Ebel et al. (1998) have shown, it is probable that the length of the spinous processes follows exclusively from a response to mechanical loads. The hump interpretation appears as a false conclusion. A merely formal comparison of certain fea- tures in different forms by an empirical-analytical method without considering particular functional adaptations, as for instance carried out by Bailey (1997), can only describe a skeletal feature but not lead to a convincing explanation of its main function. Thorough observations are per se not sufficient to explain a feature if the physical reasons behind it are poorly understood. A low correspon- dence between paradigma and compared animal trait in morphological comparisons is often felt as unsatisfactory. However, in gene- ral this fact does not lead to an abandonment of the proposed model and searching for a better applicable one, but makes some authors assume an adaptation to further tasks. This may be true in certain cases, yet probably the original adaptation is in any case a response to one main requirement.                                                                                                                                          Comparisons of extinct animals with Recent ones concerning their presumable lifestyle have also to take account of the partly enormous time gaps between the respective occurrences and of the fact that the level of development and activity has generally grown in the course of evolution due to competition and new requirements. It is well known that life does not represent a sequence of cyclical events in evolution and does not allow exact iterations, as morphological comparisons sometimes might suggest. It is very likely that the vertebral columns of the various forms discussed in this paper reflect an evolutionary increase of motility and attainable speeds. Different epochs are characterized by different skeletal ”designs” with improving properties, but the underlying physical laws have remained unchanged. Vicious circles are inevitable if, as Kummer (1959: 38) did, the vertebral columns of large and small forms, of mammals, reptiles, and dinosaurs are investigated together in order to find a well fitting function for all types, without regard to their different ages, milieus, and possibly differing lifestyles. An explanation fitting to a Recent form must not automatically be transferred to an extinct one.

Paul (1991) regrets that many long adopted views and prejudices concerning dinosaurs based on mere speculation remain in force despite the lack of confirming evidence. Certainly, this statement could be agreed. Yet, it does not help on to replace old myths by likewise speculative new ones derived from questionable analogies with modern vertebrates and in part from inappropriate compari- sons as to metabolism and capabilities of dinosaurs with the functional design of machines. Moreover, it is fruitless, as Paul (1991) endeavours to do, to move the burden of proof to and fro the advocates of the various ideas. In the long run a scientific consensus can be achieved only by convincing evidence, but not by a majority of supporters of proposed ideas. Starting points based on incor- rect assumptions cannot succeed because of persisting unexplicable conflicts. One of these probably incorrect suppositions con- cerns the terrestrial lifestyle of all dinosaurs.

3. Criteria for the judgement of size and distribution of the main loads on vertebral columns

Nature has strictly to follow biomechanical requirements imposed by continuously acting forces in order to enable vertebrates to perform their main activities. Above all these are gravitational and inertial forces occurring in connection with locomotion. Extreme designs such as the Triassic Ctenosauriscus are particularly well suited to give an insight into the underlying physical design princi- ples. Fig.1 shows the vertebral column of Ctenosauriscus as reconstructed by Krebs (1969) which gave rise to these considerations. The enormously elongated and partly curved spinous processes are a most characteristic feature of this fossil animal. The directions of the respective resultant compressive force acting on the vertebrae and the primarily loaded leg during a stride have been added. The mechanism of this outstanding design has been explained in detail by Ebel et al. (1998).

Fig. 1. Vertebral column of Ctenosauriscus koeneni (v. Huene) as reconstructed by Krebs (1969), redrawn and modified, functional interpretation as found by Ebel et al. (1998). Since during a step the directions of the resulting force acting on the primarily stressed leg pass through this point successively, most probably this animal uti- lized a facultative bipedal gait. The long spinous processes charac- terize Ctenosauriscus unequivocally as a land-dwelling animal.

Once a functional explanation has been found for a seemingly extreme, but actually perfectly adapted form, in general corresponding principles can likewise be applied to less conspicuous ones. As will be demon- strated later on, in many further cases the shape of the vertebral column reflects biomechanical requirements in a characteristic manner. In this case the problem is not, as otherwise the normal procedure of engineers, to create an appropriate design for a certain task under given loads, but to find out for a given skeleton or a certain part of it which forces and loads were probably active and have led to a given shape and which was its primary function. Adaptations to physical requirements, that is, to static and particularly to dynamic loads of the skeleton, can easily be recognized in the shoulder region of extant ungulate mammals (Fig.2).

This trait makes them well suitable for comparisons with extinct forms to find out conformities, but also important differences and the likely rea- sons. In modern quadrupedal herbivorous mammals the length of the neural spines in the shoulder area is apparently determined by several factors:

• Size and mass of the head and the inclination of the neck during browsing on the ground. The inclination angle has an effect on the muscular force required for raising the head or for holding it in a low position.
• Size and mass of the whole animal as well as the attainable maximum and continuous speeds, because these parameters affect the magnitude of dynamic forces during running, galloping, and jumping.
• The condition of the ground on which an animal preferably moves, either on fairly soft or on hard soil. The hardness of the ground respectively its elasticity has as well an effect on the magnitude of dynamic forces during running, galloping, and jumping.

Fig. 2. Sketch of a Pleistocene elk skeleton as an example of an ungulate mammal demonstrating the orientation of the spinous processes towards the shoulder joint in this area. Length and direction of the spinous processes represent an adaptation to high motility enabling this animal of running and jumping on hard grounds.

These conditions can occur separately or in combination and thus can lead to quite different vertebral columns. For example, the length of the spinous processes in the shoulder region of ele- phants, which are unable to gallop or jump, is not determined by a high running speed but by the high weight of head and tusks. On the other hand, the remarkably long spinous processes in Bison antiquus antiquus, shown by Bailey (1997), and in further Pleistocene species can probably be traced back to big mass, high running speed as well as additionally to frozen grounds. This is likely to apply also to the elk skeleton shown in Fig. 2.

Bridge analogies discussed so far (Thompson 1942, Slijper 1946) do not offer a sufficiently exact simulation of the functional condi- tions of vertebral columns (Kummer 1959), in particular in the shoulder region, and have not commonly been accepted. Such analo- gies can only be regarded as a rough approximation (Wainwright et al. 1976). Their main weakness consists in the fact that they did not consider but static conditions. However, the skeleton of a vertebrate is not designed to serve a function at rest, rather to with-

Fig. 3. Skeleton of a Recent Bison as an example for demon- strating the effect of elongated spinous processes on the mus- cular tensile force required for the transmission of a given force to the ground. Short spinous processes would result in higher tensile forces (dashed line).                                   

stand the much more important dynamic loads during locomotion which are accompanied by larger forces of variable directions (Figs.1 and 2). Fig.3 presents a Bison skeleton as an example for schematically demonstrating the effect of a modified spinous pro- cess length on the magnitude of muscular forces. The predomi- nating oblique direction of the neural spines takes primarily ac- count of the largest forces and of their directions occurring on arri- val on the ground during jumping and galloping. As can easily be recognized, the muscular tensile force T depends on the length of the spinous processes and consequently on the angle between muscles or tendons. It increases with shortening processes if a force of given magnitude has to be transmitted to the ground. The downwards acting compressive force is composed of gravitational vertical and inertial horizontal components. Inertial forces occur in connection with locomotion and can easily be felt if it is suddenly                                                                                                 stopped by an obstacle.

Obviously, the increased length of the spinous processes in the shoulder area is advantageous for a reduction of muscular forces, but extremely long spinous processes as in Fig.3a are generally not required and do not offer overall optimum conditions. Until recently, the functionally determined orientation of the spinous processes towards the shoulder joint in ungulate mammals had not been recognized (Ebel et al. 1998). Consequently, a satisfactory explanation of this very characteristic feature of many big mam- mals was not available. On the other hand, it is very advantageous that any vertebral column reflects exactly its adaptation to mechanical loads and thus to the animal’s particular lifestyle. For example, a comparison of the vertebral columns of modern ungu- lates and climbing primates reveals aggravating differences. These can be interpreted as a result of a different locomotory behaviour.

Once having recognized the connection between main mechanical load and shape of the vertebral column in extant mammals, it becomes much easier to understand the differences present in fossil forms and to find out important indications of their presumable function in connection with lifestyle and milieu. The following skeletal features have been utilized in this paper as criteria for state- ments about gait and milieu of the regarded fossil reptiles:

• The length of the dorsal spinous processes in relation to the size of vertebrae.
• The relative length of the spinous processes compared to the adjacent ones. For example, ichthyosaurs possess relatively long spinous processes, but the uniform length distribution due to a strong longitudinal musculature along the vertebral column is a characteristic of aquatic forms.
• The length distribution of spinous processes in the region between pelvis and shoulder girdle.
• The length of the spinous processes of the cervical vertebrae compared to that of the dorsal vertebrae.
• Length and shape of the tail and its vertebrae.

There are some further traits pointing to the probable former lifestyle, apart from features of the vertebral column, which in addition can be useful. They are summerized later on in Tab. 1.

4. Biomechanical function of the vertebral column in ornithischian dinosaurs

Generally, spinous processes serve a kind of bridge function comparable to an oblique-rope bridge or even more to a jibcrane (Ebel et al. 1998). Dorsal musculature and tendons are suspended on the spinous processes, thus making possible considerable vertical components of tensile forces. These allow the transmission of weight and inertial forces to the ground (Fig. 2 and 3). Remarkably elongated spinous processes in the shoulder region represent a response to vertically acting loads in connection with a terrestrial lifestyle.

Strong hints at a terrestrial lifestyle can be found also in many dinosaurs, indicated by elongated spinous processes. The variable shape of the vertebral column of dinosaurs suggests, however, that a terrestrial mode of life cannot be taken for granted in all of such forms, as it is usually done. In addition to the criteria mentioned above concerning the functional adaptations in mammals, in extinct ancient vertebrates the eventuality must be examined whether these were definitely land-dwellers or were possibly restricted to an amphibious lifestyle. The investigation of the likely function of the elongated and uniquely arranged spinous processes in Ctenosau- riscus gave rise to the idea that they evolved as an adaptation to requirements of a facultative bipedal gait on firm soil. They can be regarded as a response to relatively large forces and appear as characteristic of a terrestrial cursorial lifestyle. So far, there were no serious doubts about the terrestrial lifestyle of most, if not all dinosaurs, e.g. Norman (1991). Of course, this statement was correct if a distinction from mere aquatic forms such as ichthyosaurs was intended. But an unequivocal proof that all terrestrial dinosaurs inha- bited the dry land has never been furnished. The very diffeent shapes of the spinous processes in many dinosaurs point to funda- mental differences as to the mechanical loads and thus to the preferred milieu. However, certain regularities are clearly discernible.

A kind of backsail is present also in certain dinosaurs. This feature is caused by elongated spinous processes. They represent a remarkable trait of most ornithischian dinosaurs and are well recognizable in forms such as Iguanodon, Ouranosaurus or the hadrosaur Parasaurolophus. As in Ctenosauriscus this trait suggests an adaptation to hard grounds. In ornithischians as well as in extant ungulate mammals elongated spinous processes appear as an important prerequisite for carrying portions of the body weight and for the transmission of comparatively large forces to the ground. Bailey (1997) has discussed at some length a corresponding suitability in Ouranosaurus. The spinous processes serve the suspension of ligaments and strong muscles respectively a lattice of ossified tendons (Norman & Weishampel 1990). These can be interpreted as a response to continuously acting forces with conside- rable vertical components. The angle between muscle fibres respectively tendons increases with a growing length of the spinous processes which, within certain limits, makes the tensile forces of tendons decrease. The arrangement of ossified tendons is in good agreement with the suspension of muscles as outlined in Fig.3. Elongated spinous processes are plainly to be expected in land- dwelling dinosaurs, because they cannot make a profit from the weight-reducing effect of the hydrostatic buoyancy in water, in contrast to amphibious forms.

Fig. 4. The iguanodontid Ouranosaurus, redrawn from Carroll (1993). The directions of the tips of dorsal spinous processes do not pass through a common intersection. This animal is quite untypically depicted as a biped in this sketch. The long spinous processes between pelvis and shoulder girdle charac- terize this dinosaur as terrestrial. The parallel orientation of elongated spinous processes indicates a quadrupedal gait. The modified claws present additional evidence of this suppo- sition. For purposes of a functional comparison a modern bridge in southern France

Moreover, the arrangement of the spinous processes along the vertebral column between pelvis and shoulder region indicates whether a dinosaur generally moved bipedally or quadrupedally. An extreme specialization as in Ctenosauriscus cannot be found in any dinosaur. Fig.4 shows a sketch of Ouranosaurus from the Lower Cretaceous as an impressive example of an ornithischian dinosaur. Despite the long hindlimbs in such forms a comparison with the arrangement of the vertebrae in Ctenosauriscus makes clear that the elongated spinous processes in ornithischian forms and their orientation do not point to an obligatory bipedal locomotion, rather they hint unequivocally at a quadrupedal gait on land. Moreover, we can assume that the extremely long spinous processes do not represent an overspecialization, but served a function requiring just this length. The more or less parallel processes with approximately equal length extend over the entire trunk to the shoulder girdle. This fact indicates that normally the forelimbs also had to transmit forces to the ground. A primarily bipedal locomotion on land should have resulted in a strong elongation as well as in a similar arrangement of the spinous processes above the hindlimbs as in Ctenosauriscus and in a decreasing length towards the forelimbs.

Maybe, the sketch of Ouranosaurus in Fig. 4 creates the impression as if an orientation of the spinous processes towards a com- mon intersection would be feasible or could be attained by small modifications of the reconstruction, thus suggesting bipedal loco- motion. This is, however, not the case. The dorsal vertebrae are not curved as those of Ctenosauriscus. The location where a few lines indicating the direction of the spinous processes cross each other does not correspond to the position of a leg joint. The impression results from an unusual flexure of the vertebral column itself in this sketch which shows Ouranosaurus reconstructed as a hypothetical biped. Although fore and hind limbs in this form are equipped with a kind of hooves or flattened claws, instead of sharp ones, this form as well as others with considerably longer hindlimbs than forelimbs is regarded by some authors as a biped, e.g. Carroll (1993). Norman & Weishampel (1990: 531) point to a likely original bipedal gait in iguanodontids, but they consider Iguano- don as well as Ouranosaurus as quadrupeds. The upright leg position in connection with elongated hindlimbs does not automatically mean bipedality, it is a prerequisite only (Charig 1972). The vertebral columns do not yield any indications that ornithischian dino- saurs might have been bipeds. Maybe they were able to move a short distance on their hindlimbs just as well as trained horses or even elephants can do. But nobody would call the latter bipeds for that reason. The feature of elongated spinous processes is not yet as pronounced in early ornithischians such as Fabrosaurus as in later forms and may characterize the transition to the dry land.

A correspondence of the vertebral processes of land-dwelling ornithischian dinosaurs to the jib crane analogy is not as obvious as in ungulate mammals. Apparently, the length is additionally determined by a stiffening effect, although a certain flexibility should have been retained. The stiffening function in these forms is indicated by a lattice-work of ossified tendons which suggests a modification and, compared to earlier vertebrates such as Ctenosauriscus, an evolutionary improvement of the dorsal musculature required for continuous quadrupedal locomotion on land. Furthermore, the design of the vertebral column makes likely that ornithischians had abandoned the sinuous reptile gait, but the stiffness of the backbone presumably did not allow a high running speed. The frequently very different length of fore and hind limbs appears to be inherited from saurischian ancestors and may have been acceptable or even advantageous for browsing on the ground. However, these length differences cannot be regarded as a true advantage for the locomo- tion of land-dwellers and are indicative of a comparatively low cursorial speed, a supposition which is supported by the length distri- bution of the neural spines along the trunk that differs considerably from that in ungulate mammals. But apparently low speeds did not involve serious problems for these animals. Presumably, there were no fast predators, as pointed out later on.

The elongated spinous processes, ossified tendons, and the modified claws in ornithischians can be regarded as a feature of a more or less pronounced motility on land. Bailey (1997) opines that Ouranosaurus was a fairly agile animal which lived under dry or sometimes even arid conditions. The remarkably long spinous processes and a relatively short tail are in accord with this statement. Furthermore, the pelycosaur Dimetrodon from the Permian, the Triassic reptile Ctenosauriscus as well as the Cretaceous theropod Spinosaurus seem to have existed under comparable environmental conditions, to judge from the reddish sediments in which they were embedded. The coincidence of terrestrial sediments and enclosed vertebrates with elongated spinous processes does not appear as a matter of chance. However, despite their adaptation to the dry land none of these forms exhibits indications in the shoul- der region that it might have been capable of fast running, galloping, or jumping. The orientation of the spinous processes in the shoulder region of modern ungulates towards the shoulder joint represents an innovation and amplification of former functions of the vertebral column required for increased motility. Alexander (1985) as well as Paul (1991) supposed a galloping ability in Triceratops. But this ability must not be derived from a similar total mass and limb bone strength as in modern mammals. The spinous process length does not support such an idea. On the other hand, in forms such as Tenontosaurus (Ostrom 1970) with relatively short spi- nous processes and a long and high tail an adaptation to the dry land is not particularly evident. Possibly, a swimming locomotion was still or again part of the normal behaviour in such forms.

The most conspicuous difference between saurischian and ornithischian dinosaurs consists in the arrangement of the pelvis. Abel (1912) regarded the modification of the pelvic bones in ornithischian dinosaurs as an adaptation in connection with the stiffening of the tail which these forms carried above the ground. However, it is more likely that this modification happened in connection with the settlement of the dry land respectively with an increasing independency of the water, and not as a consequence of a developing bipe- dality as in land-dwelling bipedal birds. Obviously, bipedal theropod dinosaurs which so far are believed to share the milieu with orni- thischians retained the saurischian pelvis, and most representatives did not modify length and flexibility of the tail. Insofar Abel’s interpretation was inconclusive and contradictory. The same applies to Abel’s explanation of the differences between the pelvis of ornithipods such as Iguanodon and heavy ornithischians such as Stegosaurus, which he interpreted as caused by a change from bi- pedal to quadrupedal gait. It appears, however, that this modification in heavy forms is caused only by the enormous weight increa- se. The modification is not fundamental but gradual.

Although Bailey (1997) emphasizes the low likelihood of a heat transfer function of the dorsal sail in Ouranosaurus and similar forms with elongated neural spines and discusses an adaptation of the vertebral column to acting loads in connection with quadrupedal locomotion, nevertheless he fails to abandon completely the idea of a heat transfer function. He supposes among others a heat shield and fat storage function of a hump associated with long spinous processes. The latter is a presumption for which neither con- firmative evidence nor a true analogy with modern vertebrates can be shown. The warm Mesozoic climate did not require significant fat reserves. These would even have been disadvantageous for reasons of heat dissipation. However, armours appear as an additional characteristic of certain land-dwelling dinosaurs, the Thyreophora, which actually may have served a function of heat protection. Buffrénil et al. (1986) discuss several possible functions of the plates in Stegosaurus. The consideration of biomechanical criteria does not reveal a particular function of these structures. Yet, the exact orientation of the plates is still a matter of debate. These authors prefer a function connected with heat exchange and point to the fact that the lithofacies of the Morrison Formation, which has yielded many remainders of Stegosaurus, indicates that these forms preferred relatively dry and open habitats, in contrast to other contemporary dinosaurs. Taking into account further armoured dinosaurs such as the related ankylosaurs one might think of a protective function against dehydration and perhaps increasing direct sun radiation, which became necessary after the transition to a terrestrial lifestyle in a more open habitat. Experiments by Regal (1985) utilizing lizards have demonstrated that enlarged scales can help to reduce the rate of heat uptake. Armours tend to protect mainly the upper parts of the body. By the way, it is noteworthy that only very few footprints attributable to armoured dinosaurs are known (Lockley 1993). Moreover, it is remarkable that during the Triassic age strong armours evolved respectively were enhanced also in several further groups such as the Proganochelydia (tortoi- ses) and the Aëtosauria, which may have occupied corresponding habitats, whereas representatives of most other dinosaur groups did not develop heavy armours. Presumably, as a consequence of their particular lifestyle or preferred milieu the latter did not need it. It appears rather unlikely that such armours evolved without an adaptation to a special requirement, but likewise improbable as a defence against predators, since the ultimately imperfect protection had to be payed with a substantial increase in weight and reduc- tion of motility. Coombs (1978) emphasized the low cursorial adaptation in Stegosaurus.

5. Biomechanical function of the dorsal sails in Dimetrodon and Edaphosaurus

Remarkably elongated spinous processes have their first appearance in certain pelycosaurs such as Edaphosaurus and Dimetrodon from the Upper Carboniferous and Lower Permian, whereas the vertebral columns in other representatives of the Pelycosauria remain inconspicuous. Although the shape of such sails is generally similar to Ctenosauriscus and even more spectacular, in detail arrange- ment, orientation, and cross-section of the spinous processes are very different.

Probably, the process elongation in these pelycosaurs can likewise be explained as an adaptation to increased vertically acting forces and thus to a terrestrial lifestyle. Unequivocal indications of other functions are not present. A heat regulatory function has been proposed as the most likely one among several others. However, the generally rare occurrence of backsails makes such a function unlikely. If backsails had offered important advantages they should have occurred as a wide-spread trait. A heat regulatory function would not appear particularly advantageous in relatively large animals such as these pelycosaurs as well as in Ctenosau- riscus in comparison to considerably smaller ones. Because of relatively big mass they are generally believed to be capable of easily preserving their body temperature and would need rather long time for a change in temperature. Elongated spinous processes cannot be found in contemporary small forms. Although Bailey (1997) does not reject a thermoregulatory function in Dimetrodon because he fails to recognize a biomechanical one in the delicate spinous processes, nevertheless he mentions an investigation by Haack (1986) that does not conclusively confirm an advantage of the backsail in a heat transfer function. No physical principle is known that might stimulate a skeletal modification for that purpose. The advantage of a sail would be extremely small at the start of an evolutionary modification. To be effective such a modification should be accompanied by a further function with an immediate benefit. But there is no modern reptile or mammal in which the skeleton or parts of it are determined by primary requirements of thermoregulation. Furthermore would it hardly be possible to prove such a function by evidence from the fossil material (Pfretzschner 1999). It should be kept in mind that the thermoregulatory function is based on mere speculation. Other proposed functions such as imposing or display may sometimes be involved, but certainly such functions do not represent the original and main ones. They can only become increasingly evident if a species is perfectly adapted to its milieu, and because of ample food supply and low compe- tition has a certain margin for luxury modifications which are not immediately necessary for its survival.

Mechanical considerations strongly suggest that it is not the backsail itself which serves a function but its bony structure. The extremely elongated spinous processes in Dimetrodon and Edaphosaurus, which are even more reminiscent of a dorsal sail than those of Ctenosauriscus, suggest that they also have served a function in connection with locomotion on land, because the elon- gation of the spinous processes in these forms is restricted to the trunk area between hind and fore limbs. Yet, as a whole the design does not well correspond to a bridge analogy. Contrary to Ctenosauriscus the extremely long spinous processes do not pos- sess the curved shape and rectangular cross-section required for bipedal locomotion, but they are straight and rounded. As well as in Ouranosaurus an orientation of the tips towards a distinct intersection point is not present. These differences are probably caused by a different mechanical stress of the backbone compared to Ctenosauriscus, though not a fundamentally different one. The design appears somewhat old-fashioned in comparison to later forms, and its biomechanical function is not so obvious at first view. But certainly the vertebral column did perfectly serve its function. The likely mechanical function of the dorsal sail is outlined as follows.

Fig. 5. A technical disk with its components linked by hinges as an analogue to the dorsal sails in Dimetrodon and Edaphosau- rus. Sketch of Dimetrodon, redrawn from Sander (1994). The long spinous processes in Dimetrodon and in Edaphosaurus as well mark these reptiles as terrestrial forms.

Although an exclusively mechanical reason can be supposed for the long spinous processes in Dimetrodon and Edaphosaurus, the load distribution on the vertebral column differs considerably from that of modern ungulate mammals because it is apparently uniformly allo- cated to fore and hind limbs. Moreover, the length of the delicate and very slender spinous processes is presumably not primarily determi- ned by the suspension of muscles, since according to Fig.3 an extremely large angle between the muscle fibres would not be ne-

cessary. A bridge analogy would even appear problematical, because the sinuous quadrupedal gait of these reptiles requires a pro- nounced lateral flexibility. Therefore, a somewhat different mechanism is employed. In this case the configuration can best be com- pared with a technical disk, with its parts connected by vertical hinges, as outlined in Fig.5, together with a sketch of Dimetrodon. The primary mechanical function of a disk is a stiffening effect. Obviously, the backsail of long-spined pelycosaurs also served a stif- fening function to prevent a deflection of vertebral column and trunk. This effect of the sail with its long spinous processes is under- lined by the fact that these skeletal parts are often preserved in an articulated condition (Carroll 1993). They were apparently covered by voluminous soft parts. In order to make possible an appropriate distance from the ground during quadrupedal locomotion in very shallow water or on dry soil a strong deflection of the vertebral column had to be avoided. On land this problem was intensified by the missing weight-reducing effect of the hydrostatic buoyancy. Because of relatively short limbs in Dimetrodon, which had not yet attained an upright posture, the distance between trunk and ground was very small, and there was a steady hazard of ground con- tact. Dimetrodon shared a common problem with all land-dwelling vertebrates, namely to achieve sufficient stiffness of the vertebral column on the one hand and to retain sufficient flexibility on the other. Sufficient distance from the ground was established by the stiffening effect of elongated spinous processes. Presumably, this effect was exclusively attained by mechanically stiffening the dorsal shape, but not by an enhancement of the dorsal musculature. Despite the modification of strongly elongated spinous pro- cesses, the lateral flexibility as an important prerequisite for the sinuous quadrupedal reptile gait was completely retained. Probably, the special gait was the main reason for this unique vertebral column, which has never again been realized in later forms, as well as for the roundish cross-section of the spinous processes. The delicate spinous processes can serve as indicators of comparatively low dynamic forces. Therefore, it is probable that these pelycosaurs moved very leisurely on land. Presumably, Ctenosauriscus and ornithischians moved somewhat faster. However, a comparison with extant ungulate mammals is certainly inappropriate.

Objections might be raised that long spinous processes are not present in the modern monitor Varanus komodoensis, although it is of corresponding size and lives on land (Auffenberg 1981). It is conceivable that the dorsal musculature of pelycosaurs was not yet suitable to prevent otherwise a strong deflection of the vertebral column continuously. Yet, an extrapolation from modern to extinct forms as to certain properties appears fruitless and even not allowable. Apparently, there were several possible strategies to reinfor- ce the backbone. A completely different mechanism with similar effects was realized by the ossified tendons in ornithischians. On the other hand, all modern monitors are accomplished swimmers (Steel 1996), thereby indicating that they are not completely adap- ted to a terrestrial lifestyle and possibly retain the option to return to a semiaquatic environment in case it should become necessary or offer advantages.

The effect of stiffening the vertebral column by elongation of spinous processes occurs in Ctenosauriscus, in ornithischian dinosaurs, as well as in the pelycosaurs in question. It is the common feature of these otherwise very different forms. However, in detail mecha- nical stress and materialized design are very different, as mentioned above. Obviously, this feature evolved independently several times in different forms and epochs, just as required, but in complete accordance with the respective mechanical requirements. On the other hand, long spinous processes are restricted to a few specialized species and can serve as an indication that early verte- brates were unable to settle the dry land without special problems, even after the evolution of a bony skeleton with hind and fore limbs, rather considerable additional efforts as well as favourable environmental conditions were necessary. The disappearance of these conspicuous pelycosaurs in the course of the Permian points to the fact that a continuous settlement of the dry land was not yet attractive.

Since a strong modification of the vertebral column by process elongation seems to have happened for the first time in these pelycosaurs, they may have been the first large vertebrates at all which adapted to a primarily terrestrial lifestyle. None of the vertebrates from the Devonian, Upper Carboniferous, or Permian apart from these pelycosaurs, not even the large pareiasaurs, exhibits features of the vertebral column which might unequivocally be interpreted as truely terrestrial. All of these forms exhibit relatively short spinous processes, with a particularly uniform length distribution along the trunk, pointing to a life mainly in the water, though not necessarily as swimmers. A small length of the spinous processes can be used as an independent argument confirming conclusions by Clack (1997) that tetrapod tracks do not necessarily demonstrate that the trackmakers lived on the dry land. Early tetrapods cannot automatically be regarded as terrestrial (Coates & Clack 1990). A foot-like structure could be useful as well for manoeuvring on the ground in very shallow streaming or agitated water. It is possible that such forms could leave the water for a short time but the increased loads on the vertebral column outside the water did not find a reflection in the spinous process length and, therefore, appear to have been unimportant. Probably, the completed transition from fresh water to the dry land cannot be defined solely by the presence of legs and feet (Clack 1997), and the skeletal differences between early land-dwellers and amphi- bians may be very small. Nevertheless, the vertebral column should yield some indications. Curiously enough, up to now it has never been used as a criterion to distinguish between swimmers and land-dwellers. Unfortunately, there are still many uncertainties about the origin of land-dwelling tetrapods (Clack 1997). Although it may be opposed to our feelings as land-dwellers, nevertheless it appears more probable that the limbs of the ancestral vertebrates did not evolve in connection with the settlement of the dry land. Apart from attractive food sources in shallow water, presumably an important adaptive goal was the development and perfection of lung respiration, which required a supporting skeleton and powerful forelimbs for raising the head above the water surface. Strong forelimbs are now supposed in Ichthyostega (Clack 1997), the tail of which indicates a capability of swimming locomotion. Moreover, respiration by the gills is considered still important in such forms (Coates & Clack 1991). The improved respiration could lead to a considerably increased activity level when the oxygen content of the air had attained a sufficient extent and was then an enormous advantage, even for descendants which later returned to an aquatic lifestyle. Whereas today the air contains approximately 21 % oxygen per volume, in water it is normally about 0,7 %. The lung respiration was a most important precondition for the settlement of the dry land after it had become attractive (Hölder 1989).

6. Biomechanical function of the vertebral column in sauropods

In comparison to the ornithischian dinosaurs discussed above, in saurischians the vertebral column is marked by a different design with conspicuously short spinous processes. This fact cannot be harmonized with the idea of a terrestrial lifestyle of all dinosaurs. More than one hundred years ago palaeontologists such as Cope, Marsh, and Owen have already considered that sauropods could have largely lived in the water (Coombs 1975). There were reasonable arguments for this supposition, and it was later emphasized by other workers such as Bird (1944), Erben (1975), or Ishigaki (1989). The reasons were the large overall size, relatively weak vertebral columns in relation to the strong limb bones and to the animal`s total weight, the pin-shaped teeth which were adapted to eat soft plants, and particularly the high positioned narial openings. In modern mammals the latter trait can be found only in water- dwellers such as whales, dolphins, or in hippopotamuses. In fact, the relatively short spinous processes of most of these forms, for example Camarasaurus outlined in Fig.6, do not yield unambiguous indications of an adaptation to a terrestrial lifestyle, contrary to statements in the literature, e.g. Sander (1994). Generally, the spinous processes of the dorsal vertebrae show a remarkably uniform length. Therefore, they do not well correspond to the piers of a bridge, at least much less than the vertebrae in the shoulder region of ungulate mammals. A bridge analogy is sometimes suggested, for example by Norman (1991). The spinous processes, although arranged in a double row, appear too short for the suspension of strong muscles and, consequently, for the transmission of strong vertical forces to the ground. The absence of ossified tendons (Coombs 1975) suggests that in contrast to ornithischians a strong dorsal musculature and a continuous stiffening of the backbone were not needed. On land these dinosaurs had surely some trouble in carrying their long tail and in particular the head and the long neck with its short spinous processes. But there are only very few traces of a dragged tail, if at all, although the tail design and the saurischian pelvis suggest that it could not be carried high above the ground (Abel 1912). The weight reduction in the vertebrae of most sauropods does not appear as a feature of an unusual and so- phisticated light-weight design of land-dwelling vertebrates, rather it indicates that generally the load on the spinal column was low, and the weight reduction could help to keep the animal buoyant, above all its long neck and the head. Nature aims at high efficiency, but not at extreme designs if there is no urgent need. The remarkable contrast between light vertebrae and relatively massive limb bones was interpreted by Matthew (1915) as a means to keep the animal erect while wading. The long flexible tail was suitable for an eel-like propulsion by undulation, although the attainable speed was very low. But the animals were not lost after losing ground con- tact of the feet.

                                     Position of narial openings in various sauropods

A shift of nasal openings on top of the head would only make sense if lightweight head and neck usually floated at the water surface and were supported by their hydrostatic buoyancy. The structure of the cervical vertebrae with short proces- ses and, therefore, short lever-arms suggests that the head had not to be raised (Fig. 8), respectively that it was normally held in a low position or, as I presume as the most likely reason, floated at the water surface. A recent digital recon- struction of two diplodocoid sauropods by Stevens & Parrish (1999) suggests that the necks of these forms were preferably held in a straight horizontal postu- re. An estimation of musculature available versus that one required made Alexander (1985) arrive at the presumption that the sauropod Diplodocus might barely have been able to raise its neck to a vertical position. A similar, but even more sophisticated specialization of the cervical vertebrae can be found in the Triassic reptile Tanystropheus longobardicus (Wild 1973) which probably was hardly able to raise its head above the sea level (Tschanz 1988).

Fig. 6. Camarasaurus as an example of a sauropod, redrawn from Carroll (1993). Note the comparatively short spinous processes of dorsal and cervical ver- tebrae. This dinosaur would have had some trouble to carry neck and head on land continuously. Camarasaurus does not appear to be adapted to a mere terrestrial lifestyle.

Furthermore, this attitude of neck and head appears very reasonable because it would have allowed the blood pressure to stay within the limits known from other reptiles (Ostrom 1980), whereas raising the head, in large forms up to more than 7 m above the heart level, would have required an entirely unrealistic systemic systolic pressure of more than 500 mm Hg for supplying the brain (Hohnke 1973) and a heart of almost incredible size and pump performance (Norman 1991). Such a high blood pressure cannot be found in other animals. It exceeds the normal human one by a factor of roughly four and would require a compression strength of the blood- vessels comparable to a bicycle tyre. It is well known that, for comparison, the atmospheric standard pressure at sea level equals 760 mm Hg. Since the blood pressure needed for safely supplying the brain during erection of the head above the heart level is proportional to the product of height times blood density, the pressure must rise by approximately 76 mm Hg per metre. On the other hand, the only benefit of such an enormous modification would have been the ability to erect head and neck, similar to a giraffe. This comparison is, however, inappropriate for the absolute size differences and for the resulting physical reasons. There are no indications that the heart of dinosaurs might have been more efficient than that of other reptiles or mammals. Therefore, it does not appear allowable just to postulate superior properties. It is a peculiar method of certain authors, e.g. Paul (1991), to claim that postulated advanced capabilities of dinosaurs close to those of modern mammals have not been falsified. But such capabilities have not been confirmed as well. What else can be done if arguments such as those of Hohnke (1973) are not appreciated? Ideas not in strict accordance with physical laws can never be correct. An amphibian lifestyle did not require such capabilities. The necessity of a horizontally carried neck appears as an inevitable consequence of the available anatomical and physical evidence.

Bakker’s ideas are extremely fanciful, however unfortunately, completely unrealistic.

A semiaquatic lifestyle does not necessarily mean that sauropods fed on plants in the water only and never went on land. However, in general their vertebral column does not present indications of a special adaptation to a life on the dry land. Hence, they cannot be regarded as land-dwellers. Sauropodian remainders are not restricted to fresh water deposits (Dodson 1990: 402). They occur also in marine deposits, as well as do sauropod tracks (Meyer 1990), preferably in interti- dal to supratidal shallow water sediments (Meyer 1998). This idea is also suppor- ted by the well-known track published by Bird (1944) which consists of forelimb footprints only, with the exception of a single footprint of a hindlimb where the track changes its direction. Apparently, this track was generated by a sauropod swimming in shallow water, but Lockley (1993) has doubts about this interpreta- tion. Apparently, in order to emphasize the absurdity to presume a semiaquatic lifestyle Norman (1991) mentions an alleged snorkel function of the remarkably long neck. Yet, the track described by Bird (1944) demonstrates that sauropods obviously were buoyant and, therefore, could not walk on the ground of deep water. Thus the presumption of a snorkel function is obsolete. The incomplete ossification of the limb joints (Coombs 1975, 1978) can serve as an additional argument for an amphibious lifestyle.

Just as in all other vertebrates appears the length of the spinous processes in sauropods as a reliable indicator of preferred lifestyle and environment. However, there are certain length differences between different forms. Although generally fairly short, in Dicraeo- saurus they show an elongation which also encompasses the cervical vertebrae. Yet, an adaptation to a terrestrial lifestyle or to very shallow water can clearly be recognized only in one single species, namely in the unique Amargasaurus cazaui from Argentina (Sal- gado & Bonaparte 1991), outlined in Fig. 7. This moderately large form differs remarkably from all other sauropods by its strongly elongated spinous processes. Apparently, an elongation of the spinous processes in the area between pelvis and shoulder as in Ouranosaurus was not sufficient in Amargasaurus, rather the elongation was extended to the cervical vertebrae, probably in order to enable the animal to raise its head on land. But this modification must not lead to the conclusion that this form was specialized to browse on trees. The considerably shortened neck contradicts such an idea. The differences are simply a documentation of skele- tal modifications necessary for the transition from life in water to a predominantly terrestrial lifestyle. Bailey (1997) is of the opinion that these animals lowered their head for browsing. It appears possible that such modifications indicate a shallowing of the water level of the preferred milieu which did no more allow the head and the body being continuously supported by its buoyancy. Never- theless, presumably the environment was still characterized by vegetation in shallow water.

Fig. 7. Dicraeosaurus with markedly elongated spinous processes. The elongated vertebral processes indicate that this medium-sized dinosaur was adapted to a terrestrial mode of life.

Fig. 8. Schematic relationship between a lengthening of the spinous processes in Amargasaurus, present the longest known vertebral processes known up to now in a sauropod. and the size of the muscular force required to generate a moment of definite magnitude M.

The elongation of the cervical spinous processes in Amar- gasaurus and consequently the increasing distance bet- ween acting muscular force and axis of the vertebral column brought about a decrease in the size of muscular forces for equal efficiency, respectively the efficiency could considera- bly be enhanced with the original muscular force. This con- text follows from Fig. 8. Amargasaurus cazaui is an excel- lent example for demonstrating the skeletal differences due to physical requirements between forms living mainly in the water and those living on the dry land.                                                                                              

7. Biomechanical function of the vertebral column in theropod dinosaurs

In most theropods, popularly regarded as terrifying land-dwelling bipeds, the spinous processes are rather short. This fact leads to the question why or whether the process length in these forms was sufficient to transmit the body weight to the ground and to make possible a bipedal locomotion in connection with a predatory lifestyle on land. Traditionally, early Triassic dinosaurs, including the herbivorous prosauropod Plateosaurus, are commonly interpreted as bipeds, at least during fast running (Sander 1994). Yet, an occasional use of the forelimbs to support the body in prosauropods is considered as likely by Coombs (1978). These ideas can already be found in early considerations about the dinosaurian lifestyle, e.g. v. Huene (1907/08) or Abel (1912). Abel’s textbook, if critically read, can still yield much useful information, although some ideas cannot be correct, to judge from biomechanical conside- rations. Recent ground birds and kangaroos were regarded as possible analogues in those days. These ideas have strongly influen- ced reconstructions of theropods as to posture and gait up to the present day. A bipedal gait on land appeared obvious, as well as the head being carried as in an ostrich. Apart from these old ideas, suppositions concerning posture, gait, and cursorial speeds seem to be based largely on the very different length of fore and hind limbs as well as on morphological comparisons with modern vertebrates, in part also on trackways. Nevertheless, the reasons for this supposition remain somewhat unclear. Various fanciful attempts to reconstruct lifestyle and milieu of dinosaurs in semipopular publications cannot obscure the fact that these reptiles are still poorly understood and that the knowledge about them is very restricted.

Until recently, the shape of the vertebral column did not play an important role in considerations concerning the lifestyle. This paper suggests a different idea of theropod dinosaurs. As a logical consequence of the above considerations the conception of bipedality in terrestrial theropods does not appear well founded. Yet, the shape of the vertebral columns with short neural spines becomes easily understandable if we assume that theropod dinosaurs just as sauropods normally lived in moderately shallow water, but not neces- sarily as swimmers. Indeed, it is conceivable that the vertebral column of carnivorous theropods reflects an adaptation to such a lifestyle. By this assumption the reasons of the very different length of fore and hind limbs become clearer. The probability that these animals utilized a bipedal locomotion on land appears rather low because of the unspecialized vertebrae, and such a behaviour would not be typical. Although certain modern four-legged reptiles are capable of a two-legged locomotion (Christian et al. 1994) they cannot use this gait continuously and for long distances. The longer hindlimbs are an ancient feature, inherited from the eosuchians (R. Wild, communication by letter, 1995). On the other hand, the strong hindlimbs certainly suggest a bipedal locomotion. Yet, con- spicuous differences between the vertebral columns of theropods and of Ctenosauriscus point to the idea that theropods were not adapted to the dry land. Moreover, the differences between the spinal columns of ungulate mammals and dinosaurs hint at different locomotory capabilities. Apparently, so far useful criteria to judge the suitability of dinosaur skeletons for bipedal gait, fast locomoti- on, or a preferred milieu were not available.

By establishing analogies concerning motility and capabilities between modern warm-blooded vertebrates and extinct organisms we are intuitively inclined to overestimate the unknown properties of the latter or to associate similar environmental conditions and cor- responding lifestyles to them. In particular the traditional method of morphological comparisons with modern vertebrates for consi- derations about the lifestyle of extinct animals is hazardous and can easily lead to misinterpretations. Although in general this method can be very successful, it must however always be borne in mind that it is somewhat superficial and does not consider the deeper reasons of a special development which may result from different physical conditions or requirements. Unfortunately, results obtained in this manner are sometimes uncertain and can be highly speculative, above all if corresponding lifestyles of the compared fossil and extant forms cannot definitely be confirmed or if an originally different function has been modified and amplified during evo- lution (Ebel 1996, 1999). Since morphological comparisons are based on the supposition of a functional analogy between the re- spective forms, the recognition of a modified function in this way can, unfortunately, be precluded. For this reason an extrapolation of conclusions gained from modern vertebrates to the properties of extinct ones requires particular caution.

Acrocanthosaurus from the late Cretaceous as a rare example of a theropod with elongated spinous processes pointing to a lifestyle in very shallow water or even on land

In carnivorous theropods ranging from Compsognathus to Ty- rannosaurus a lengthening of the spinous processes in the trunk area between pelvis and shoulder can only rarely be observed. For example it occurs in Acrocanthosaurus or in Spinosaurus (Sander 1994). The latter form is, however, very poorly known. A reliable reconstruction of its skeleton seems impossible at present, except that the elongated spinous processes (Stromer 1915) point to an adaptation to terrestrial conditions. There must be a special reason for the short processes in most theropods. If such forms had really been true terrestrial bipeds, then the spinal column should show at least distant similarities, for example remarkably elongated spinous processes in the pelvis region, with Ctenosauriscus which is unique in this respect.

The large ground bird Diatryma from the Early Tertiary of the northern hemisphere shows a good adaptation to locomotion on the dry ground by elongated spinous processes above the hind limbs as well as rudimentary arms. Apparently, the direction of the spi- nous processes corresponds to the area where the respective foot is in contact with the ground.

Fig. 9 presents a reconstruction of  Tyrannosaurus as an example of a theropod dinosaur with short spinous processes. Skeletons of such forms exhibit strong differences also among each other, depending in particular on the overall size. Certainly, these differences cannot be traced back to different design principles. It is very curious and remarkable that in theropods an increa- sing size is accompanied by an increasing reduction of the forelimbs which do not appear any more suited for locomotion or a grasping function. Apparently, the forelimbs of such forms had lost their former main function, maybe in analogy to the wings of ratite birds. But surely this atrophy cannot be regarded as an adaptation to bipedality on land, as suggested between the lines by Coombs (1978).

Fig. 9. Skeleton of  Tyrannosaurus rex as an example of a theropod, redrawn from Carroll (1993). The short spi- nous processes in the region between pelvis and shoul- der girdle indicate that the stress of the vertebral column was relatively low, and therefore the animal does not appear to have been adapted to a terrestrial environment and probably lived in moderately shallow water. Note the extremely long tail.

Because of the short spinous processes in connection with a lightweight design of the dorsal vertebrae (Norman 1991, Brooks et al. 1998) we can assume that the vertical stress of the vertebral column due to static and dynamic loads was comparatively low. For land-dwelling carnivorous dinosaurs such a shape of the spinal column obviously does not corres- pond to its supposed carrying function. Whereas the spinous processes of land-dwelling ungulate mammals are remarka- bly elongated in the shoulder region, in particular in Pleistocene species adapted to frozen grounds, on the contrary the short pro- cesses of theropods suggest an adaptation to very soft grounds at least to say or to a lifestyle in moderately shallow water. The rudi- mental forelimbs are particularly difficult to understand, because it is hardly conceivable why a predator should completely abandon the use of its arms. Certainly, the development of relatively weak vertebral columns and short spinous processes as well as the adherence to the saurischian pelvis has adaptive reasons, insofar as there was no urgent need for a strong modification of the skeleton as in ornithischian dinosaurs, which extended their habits to the dry land. But all of these features can be harmonized with an amphibious lifestyle. It is well known from our own experience that moving the legs under water, either walking or paddling, is much more strenuous than on land. Therefore, the caudo-femuralis muscle had to remain an important limb retractor in these dino- saurs, and there was a strong necessity to adhere to the saurischian pelvis. Sharp claws and frequently a long flexible tail appear as additional characteristic features of such a lifestyle. Claws could be helpful for ensuring the position in streaming shallow water or for locomotion. This function may also apply to the remarkably enlarged claws in the Triassic diapsid Drepanosaurus (Pinna 1986). On land long claws would tend to be hampering, and they have actually been modified by ornithischians. A long tail could be useful for slow swimming using an undulating locomotion as Recent crocodiles do. Moreover, theropods just as merely aquatic forms have retained the gastral ribs (Norman 1991), which possibly play a protective role in contacts with the ground, e.g. for resting or breeding.

The obvious differences between dinosaurs with short and long neural spines are very likely caused by different lifestyles. Modern reconstructions of the dinosaurian mode of life do not take account of these differences, although the occurrence of tracks generated by different forms (Lockley 1993) points to different preferences as to the environment. Early ornithischians such as Fabrosaurus resemble saurischians by the presence of short spinous processes, but the new trait of ossified tendons hints at a change in loads on the spinal column and hence in lifestyle. Despite the different length of fore and hind limbs this form does not show indications of bipedality on land. Maybe, the skeletal shape reflects the transition to the dry land being in progress. From a biomechanical view- point land-dweller in this context can only mean independency of the weight-reducing effect of the hydrostatic buoyancy which repre- sents the only main difference, but a very far-reaching one, regarding the load on the vertebral column between vertebrates living on land or in the water.

Widespread flattened continental landscapes with shallow lagoons, delta basins, lakes, sluggish rivers bordered by gallery forests on the one hand and desert-like dry land with only few huntable animals on the other (Erben 1975: 274) could have been a main reason for theropods to stay predominantly in the water where fish for instance was an attractive food source. A corresponding specializa- tion is particularly well recognizable in Baryonyx with its crocodile-shaped skull (Charig & Milner 1986). Presumably, plants as the primary food of further prey were restricted to the proximity of water. Since a swimming locomotion is hardly feasible in shallow swampy regions with dense vegetation and is expensive in streaming water, bipedal locomotion using ground contact by the hind- limbs was a useful alternative, with the body being more or less submerged and supported by its buoyancy. Certainly, theropods were bipeds and well adapted to the existing climatic and environmental conditions, but their skeletal shape makes likely that they utilized this gait predominantly in moderately shallow water.

Deinonychus as a particularly instructive example of semiaquatic theropods is shown in Fig.10 in its typical posture (Ostrom 1969). In the water the submerged body would assume almost automatically a near-horizontal attitude. This lifestyle did not confront the animal with the problems of balance control during bipedal locomotion in the same severity as a land-dweller. For this reason an alignment of the spinous processes to the knee joint as in Ctenosauriscus was not necessary. The tail of Deinonychus is very remarkable because it shows an apparently adaptively acquired restricted flexibility. This feature is present as well in rhampho- rhynchoids. In Deinonychus it may have served an analogous function, namely steering during swimming (Ebel 1996) in connection with a modified propulsion generated by the forelimbs.

Fig. 10. Deinonychus, a moderately large theropod with a length of about 3m and long arms from the Upper Cretaceous and Archaeopteryx from the Upper Jurassic. These theropods display skeletal similarities which may be traced back to a similar life- style, that hunting fish in the water. As in Archaeopteryx in Deinony- chus also the tail is stiffened

There are some indications which can serve as evidence that theropods actually were water-dwellers. Coombs (1980) as well as Mossman & Sergeant (1983) depict a track generated only by the hindlimb claws of a theropod floating or walking in moderately shallow water. This track indicates that theropods used a bipedal gait in the water, at least occasionally. Furthermore, several theropods have even been found in marine sediments. Up to now, this fact has not sufficiently been appreciated. Generally, such remainders are believed to have been washed in from a coast and thus to represent untypical conditions. However, the good preservation of the Bavarian specimen of Compsognathus makes unlikely that it has been washed in. Presumably, this animal as well as Archaeopteryx and pterosaurs lived in the vicinity of a nearby island, the presence of which is indicated by terrestrial plant remainders in the Solnhofen Lithographic Limestones (Viohl 1985).

The comparatively incomplete ossification of the leg joints in all saurischian dinosaurs (Coombs 1978) which did not allow but modest accelerations and a relatively short stride length (Carroll 1993) can be mentioned as an additional support of a lacking adaptation to hard grounds on land. Presumably, there was no stimulus for these animals to leave the water. The questionable advantages of a life on the dry land could not yet compensate for the disadvantages, for example the steady danger of dehydration and lack of food. It is conceivable that only a slow regression of the water level created a stimulus or a necessity for Cretaceous theropods such as Acrocanthosaurus or Spinosaurus to adapt to very shallow water or to terrestrial conditions. The short spinous processes in most theropods indicate that they moved so slowly on land, if at all, that the dynamic forces remained low. However, if they were not capable of a continuous bipedal locomotion on land, then forms with hind and fore limbs of very different length cannot have been suited either for a fast quadrupedal locomotion. Not a single dinosaurian vertebral column presents indications that these animals might have been capable of jumping and galloping, contrary to claims in the literature, e.g. by Bakker (1971) or Alexander (1985). In that case a characteristic elongation and orientation of the spinous processes towards the shoulder joint of quadrupedal forms such as ceratopsids should have been present as in quadrupedal mammals capable of galloping, because fast locomotion and the arrival on the ground following a jump is connected with increased dynamic forces which may considerably exceed the static ones. Possibly, carnosaurs such as Tyrannosaurus just as crocodiles utilized a lifestyle of lurking for prey in moderately shallow water. On the other hand, small theropods, the coelurosaurs, must be regarded as active predators as stated earlier, e.g. by Ostrom (1990: 279). However, their hunting activity should have been restricted to a shallow water environment.

8. Synopsis of characteristic features

The investigation of the vertebral column as to its likely functional adaptation in the pelycosaurs Edaphosaurus and Dimetrodon as well as in dinosaurs by means of a biomechanical approach reveals that, to judge from the different length of the neural spines, the extinct forms appear to have been adapted to various environments and lifestyles.

Table 1. Skeletal features of dinosaurs which by their presence, absence, or modification can be helpful for determining as to whether a dinosaur used a bipedal or a quadrupedal gait. Single features given in this table or a combination of several can yield strong indications of a terrestrial or amphibious lifestyle.

Concerning dinosaurs these investigations suggest the conclusion that there were semiaquatic and land-dwelling herbivorous forms on the one hand and carnivorous theropods primarily living in shallow water, apart from doubtless aquatic forms.

Table1 shows an overview of features of the vertebral column that can be used to distinguish between amphibious, land-dwelling, bipedal, or quadrupedal dinosaurs. Some further traits such as the stiffened tail in Deinonychus or modified claws present additional characteristics of a special lifestyle, but per se these are not necessarily an unequivocal trait of amphibians or land-dwellers.

9. Discussion

The above considerations demonstrate that the shape of a vertebral column contains much information about its main biomechanical load and adaption to carrying and locomotory tasks. Consequently, a biomechanical examination can yield reliable conclusions regarding preferred gait and milieu of extinct vertebrates. For forms with elongated spinous processes the presumption of a terrestrial lifestyle can be confirmed. However, with the exception of the facultative bipedal rauisuchian reptile Ctenosauriscus not a single dino- saur yields indications of an obligatory bipedality on land. Utilizing the shape of vertebral column and the length of spinous proces- ses as a measure for the load capacity of the skeleton one is confronted with the consequence that the number of land-dwelling dinosaur species was probably considerably lower than supposed so far, that the continuous settlement of the dry land appears to have been subject to considerable problems for dinosaurs, and was essentially confined to the ornithischians.

A particularly surprising result of the above considerations is that, apart from sauropods, theropods also were probably adapted to a semiaquatic lifestyle, and thus an idea of these animals is suggested that differs considerably from former suppositions which, however, were far from being beyond doubt. In contrast to sauropods a restriction of theropods to the water has never been taken into consideration, probably because of the elongated hindlimbs and bird-like feet which have been interpreted for a long time as an adaptation to bipedal locomotion on land. Since Abel (1912) did not consider this eventuality he failed to find an entirely convincing explanation for the modified pelvic bones in ornithischians. He arrived at the conclusion that the modification happened in connection with a stiffened tail and bipedality in these forms. He pointed to the similar anatomy of the pelvis in birds and ornithischians as a result of convergence. However, he could not unquestionably explain why tail and pelvis in bipedal saurischian dinosaurs remained unchanged. Different lifestyles and milieus as outlined above offer a more obvious explanation. Only after having left the water ornithischians and birds had to or could modify the pelvic bones. Apparently, in the water the occasional or regular utilization of the tail for a slow eel-like propulsion contrasted with a particular stiffening, except in some forms with arm-propelled locomotion and a modified tail function due to a specialized lifestyle such as rhamphorhynchoids, Archaeopteryx, or probably Deinonychus. Another misinterpretation concerns the statement of Abel (1912) about the modification of the feet in ornithopods which he regards as an adaptation to fast running and jumping, although he mentions that no tracks of unequivocally running or jumping dinosaurs are known. Moreover, the feet of small theropods were erroneously interpreted as an adaptation of tree climbers. The outlined differences in lifestyle appear as a more consistent explanation.

Aside from mechanical considerations that strongly point to a semiaquatic lifestyle of theropods several further facts can serve as support of this idea. The embedding conditions of theropod skeletons which are usually found in lacustrine as well as in fluviatile sediments (Barsbold & Osmolka 1990: 243), sometimes assembled with ornithischian forms in a presumable thanatocoenose, pos- sibly caused by a flood event, and even in marine deposits do not at all contradict a possible semiaquatic lifestyle. Fossils preserved in fluviatile or in terrestrial sediments anyway do not necessarily characterize the former preferred milieu, whereas in lacustrine or marine sediments the likelihood is considerably increased that enclosed complete fossil animals lived in, on, or at least close to the water. The supposition of an amphibious lifestyle is well suitabe to arrive at a better insight into theropod forms, their skeletal func- tion and particularly the forelimb function which has caused problems of understanding (Witmer 1997). This lifestyle appears as an excellent response to the prevailing Mesozoic environmental and hot climate conditions and would have offered the best chances for dinosaurs respectively their ancestors to react by adaptation to sufficiently slow sea level changes, whereas specialized land- dwellers such as Dimetrodon, Ctenosauriscus or Ouranosaurus disappeared without attributable descendants.

The climatic conditions of the Mesozoic age concerning the temperature level and its constancy (Norman 1991) as well as the envi- ronmental conditions differed considerably from those of the following Tertiary. According to Ostrom (1980) the mean annual tempe- ratures may even have been 10° C to 15° C higher than today. Thus, life in more or less shallow water would have offered the overall most favourable conditions for most dinosaurs, while during this epoch the dry land would have been associated with restrictions that hampered a continuous settlement. But I can hardly envisage that true land-dwelling reptiles could return to a marine environment. Apparently, such processes of specialization take a fast course, since transitional forms are extremely rare. Starting points for such developments must have been present in the normal faunal spectrum and should be identifiable. Yet, generally adaptations to an aquatic lifestyle are not recognized before substantial skeletal modifications such as the development of propulsive fins or reduc- tions of other skeletal features have happened.

A semiaquatic lifestyle makes easier understandable that representatives of theropod dinosaurs could specialize to diving for food or hunting fish under water. Although such a lifestyle is by no means the only conceivable specialization for theropods, in two cases it gave rise to the evolution of wings in small forms which later proceeded to flying in the air (Ebel 1996), namely Archaeopteryx, which is traced back to an unidentified theropod, and the rhamphorhynchoids, the direct ancestors of which are not definitely known. Wild (1984) regards certain eosuchians as ancestors of pterosaurs because an elongated fourth finger is already present in these forms. In addition, the evolution of a sternum in lepidosauromorphs underlines the growing importance of the forelimbs for a specialized locomotion. It is conceivable that only a developing separation of the functions of fore and hind limbs made the remarkable feature of strongly elongated hindlimbs possible, which hitherto have commonly been interpreted as a feature of bipedality on land. However, the shapes of bipedal dinosaurs and ratite birds differ considerably. Actually, bipedality on land does not appear superior to quadru- pedal locomotion. In birds it is an obligatory consequence of the modified function of the forelimbs and closely linked to the evolutio- nary history of flight. Regarding the evolutionary pathway to the origin of flight the idea of amphibious theropods appears more plau- sible than the supposition of a terrestrial theropod which returned to the water to become a fish hunter as considered as a working hypothesis by Ebel (1996). On the other hand, the fact that birds and pterodactyloids developed from underwater hunters can be regarded as corroboration that water was the normal milieu of theropods and their ancestors. At present, the lifestyle of arm-propel- led underwater hunting is utilized only by a few birds such as auks, puffins, or diving petrels. These birds have a size approximately comparable to Archaeopteryx. However, the otherwise highly specialized penguins can serve as evidence that relatives of birds no more able to fly in the air can fly under water and hunt fish. Actively flying in the air is not as easy as it may appear at first view to laymen in aerodynamics. Apart from sufficient wing area several further pre-conditions have to be fulfilled, such as sufficient flight stability and the appropriate wing movements. Surely, these conditions cannot develop from gliders, runners, or by mere chan- ce, as formerly supposed. It is noteworthy that two entirely independent approaches, the one presented in this paper and that of Ebel (1996), lead to the consistent conclusion that the ancestors of pterodactyloids and birds were probably adapted to an aquatic milieu. Furthermore, the strong skeletal differences between rhamphorhynchoids and pterodactyloids on the one hand and between sauri- schians and ornithischians on the other point to the radically different physical requirements in the water and on the dry land.

Ostrom (1995) revealed striking similarities regarding the anatomy of the forelimbs between Archaeopteryx on the one hand and the dromaeosaurid Deinonychus and some further forms on the other. Taking into account the shoulder girdle of Deinonychus, which according to Ostrom (1979) was presumably equipped with a strong musculature, it appears possible that the forelimbs in these forms were transformed to a wing-like or fin-like structure which could serve the generation of propulsion in deeper water where the hindlimbs could not maintain ground contact. During swimming the highest speed could actually be achieved with the body comple- tely submerged due to the absence of bow wave drag at the water surface. However, Ostrom’s reconstruction of Deinonychus sug- gests that the head was normally held in an elevated position and the animal, therefore, was not a specialized underwater hunter. Its mass appears too big for the high speed required. In any case a particularly light skeleton with hollow bones was very advantageous for a water-dwelling hunter, because of relatively low mass that had to be accelerated. It is likely that all coelurosaurs were able to float at the water surface and may have occupied in part the ecological niches of modern aquatic birds.

There are some indications that representatives of the ornithomimidae such as Struthiomimus with elongated arms and manus were possibly adapted to a facultative arm-propelled locomotion in the water. This may also apply to the recently discovered Unenlagia comahuensis (Novas & Puerta 1997), which is smaller than Deinonychus and shares common features with Archaeopteryx and Deinonychus. Outstanding traits of the delicate ornithomimidae are a relatively large cerebellum in comparison to ornithischians (Hopson 1980), which can be interpreted as an adaptation to complex motions, as well as an improved stereoscopic eyesight. These are favourable attributes for diving and swimming under water. Certainly all of these forms would have been too heavy for flying in the air. A limit for the development of flight capability in bird-like animals is given by a maximum mass of approximately 2 kg which allows a standing takeoff (Pennycuick 1986) and, therefore, was possible only under certain pre-conditions. But surely there is also an upper mass limit for the lifestyle of hunting in the water or below the surface utilizing propulsion generated by the arms. Since an increase in body length leads to a cubic mass increase, but only to a quadratic increase of the relevant muscular cross-sections, the attainable speed must decrease with a growing size. It is conceivable that the excess of this limit caused an increasing atrophy of the forelimbs in large theropods. The arms lost their former function, maybe in analogy to the vestigial wings of modern birds which became too heavy for flying and cannot, and need not, adapt these formerly highly specialized limbs to a new function. Moreover, it is unlikely that the ”wings” of these theropods might have been equipped with flight feathers because of the high mass that had to be accelerated and, therefore, high bending forces which would have acted on the feather shaft. The featherless wings of penguins that have secondarily evolved appear as the best device for animals of this mass category which hunt in the water.

Obviously, the similarities as to the skeletal development between Archaeopteryx and rhamphorhynchoids on the one hand and the above-mentioned theropods on the other can be traced back to common ancestors. As long ago supposed, all of these forms were closely related, respectively connected by a similar lifestyle which in the former was not yet primarily determined by flying in the air. Therefore, it appears not so astonishing that an independent evolution of the ability to fly in the air could happen several times, possibly oftener than presently known (Kaiser 2000). However, such an evolution had to be restricted to small species, as evidence confirms. Note that in Archaeopteryx the alula, an important trait of highly manoeuvrable birds, is not yet developed. A pneumatiza- tion of the skeleton has been found in many saurischians, a feature recently reported also from Archaeopteryx (Brooks et al. 1998). A close relationship is indicated by the breeding behaviour that evolved presumably early in certain theropods and, therefore, was not an invention of birds (Norell et al. 1995). Maybe, the breeding position of an Oviraptor reported by these authors can serve as an indication that the arms had a wing-like or fin-like structure and thus could serve as a device to maintain the appropriate breeding temperature.

Footprints of saurischians, frequently a socialization of sauropod and theropod tracks (Kaever & Lapparent 1976, Lockley 1993, Lockley & Hunt 1995), seemingly contradict the idea of a semiaquatic lifestyle of theropods. Although obviously bipedal dinosaur tracks are relatively abundant, the circumstances of their generation can be very different. However, occasionally a pretended bipeda- lity can be a matter of preservation or can even be feigned in some cases by forelimb traces whiped out by the hindlimbs (Paul 1991). Since a considerable moisture of the ground is a basic prerequisite for the preservation of tracks (Lockley 1987) these must have been generated in shallow water or at least in its immediate proximity. Therefore, it is difficult to state whether the trackmakers came from the dry land or from deeper water and how far they normally moved away from it. Yet, many footprints have certainly been produced under a water cover (Coombs 1980, Clack 1997), and their shape is likely to depend on the water depth, as well as on the consistence of the ground. Many footprints are remarkably shallow. There is no contradiction between a track that was generated near a shore and a semiaquatic lifestyle of the trackmaker. Lockley (1987) mentions tracks which were probably produced by predo- minantly aquatic phytosaurs. None of the known dinosaur trackway localities appears to mark a watering-place. Contrary to sauri- schian footprints those of ornithischians are comparatively rare (Lockley et al. 1998).

Despite their general adaptation to a life in moderately shallow water sauropods and theropods certainly went on land occasionally or regularly as extant crocodiles do, for example for egg deposition or eventual hatching supply. Forms such as Tyrannosaurus with rudimentary forelimbs had no alternative to utilizing a clumsy bipedal gait on land. Despite this behaviour they cannot be regarded as mere land-dwelling animals, just as little as a swan. Certainly, the erect bipedal posture of many reconstructed large theropods shown in various publications, e.g. Carroll (1993), would have been connected with problems such as an increased blood pressure (Dodson 1990: 404) and would have been possible on a restricted scale only. Moreover, this attitude appears unlikely as the typical one. The almost horizontal trunk attitude which follows from Ostrom`s reconstruction of Deinonychus or that outlined by Mossman & Sergeant (1983) appears more likely and transferable as the normal posture to other bipedal forms. This posture results also from the reconstruction of Ctenosauriscus by Ebel et al. (1998). Recent birds have overcome the problems of attitude by concentrating their mass close to the centre of gravity and thereby demonstrate an excellent adaptation to bipedality on land as well as its limits.

As a consequence of the presumed, but unproven capabilities of land-dwelling dinosaurs has the upright posture of theropods with the trunk in a near-vertical position been utilized by certain authors, e.g. Bakker (1980), as a main argument for an enhanced meta- bolism in dinosaurs which is supposed to come close to that of mammals, because the length of the hindlimbs is believed to be characteristic of high running speeds. This idea is not generally accepted and controversially discussed (Thomas & Olson 1980). The vertebral columns of dinosaurs do not yield any evidence that they might have been capable of high speeds. On the contrary, tracks confirm that all of them appear to have been rather slow (Alexander 1976, Coombs 1978). However, even if some dinosaurian representatives were warm-blooded there must be doubts whether it is allowable to use a supposed posture as a main argument to support this presumption. In that case it had to be certain at least that theropods actually were predominantly land-dwellers and were able to stabilize the presumed enhanced activity level. Life in lukewarm water would neither require endothermy nor favour it. In contrast to that supposition a semiaquatic lifestyle, with the locomotion carried out by the hindlimbs touching the ground, would have favoured an evolutionary pressure to lengthen the hindlimbs and to change the limb attitude in order to be able to bridge deeper water, presumably more than the adaptation to a terrestrial lifestyle would have done. Moreover, the possession of strong claws, particularly in theropods but present as well in sauropods, appears useful for pushing off on soft grounds or to withstand the pressure of currents and corresponding to its original function. Although big claws tend to produce a terrifying impression on human beings their suitability as a weapon in Deinonychus appears restricted because of their presence in the feet only.

On the whole, a semiaquatic lifestyle makes carnivorous dinosaurs lose much of their horrifying image, which is often enhanced by emphasizing the size of mouth and teeth in popular presentations, and makes their feasibilities appear more moderate. It is rather unlikely that these animals should have been capable of making Jurassic parks as well as Cretaceous ones unsafe.