A very useful ratio is obtained by dividing the straight-line carapace length (L) by the maximum carapace height (Ch). The resulting index in our experience can be of considerable diagnostic value (Highfield & Martin, 1989). The width of the posterior carapace marginals (MwM) compared to the median and frontal carapace width (Fw) is another very useful feature diagnostically and can also indicate the sex of a specimen (Highfield & Martin, 1989).

Field records & observations - Tunisia

Biotypes of n. African tortoises in present day Tunisia

Tunisia in the present era can be usefully divided into several climatic and geographic regions. The highest concentrations of tortoises occur in the north and along the Sahel coastal region. The annual rainfall is highest in the north >1000mm and lowest in the south <125mm average. The precipitation in the southern interior is even lower, vast areas comprising no more than arid sandy desert. No tortoises occur in such habitats.

Northern Tunisia

Generally speaking, this region comprises a series of ridges and valleys which support a dense growth of cork and holm oak forest. Much of the area is above 600m and the highest peak is Djebel Chambi at 1519m. At lower levels, particularly in coastal sites where the vegetation has been most affected by man, extensive areas of maquis and Opuntia scrubland occur supporting broom, buckthorns, lavender, thyme and similar aromatic shrubs. This biotype is in many ways typically modern Mediterranean. In the areas of high precipitation evergreen woodland predominates and it is these locations which tend to be most favoured by tortoises. These are also the areas least affected by the activities of mankind. The region of Ain Draham and Teboursouk has the highest recorded annual rainfall of any location in n. Africa (>1050mm p.a) and is ideal habitat for land tortoises in addition to providing an equally good habitat for aquatic chelonians.

Coastal Tunisia - the Sahel

The coastal plains of the Cap Bon and from Hammamet to Sousse are also typically Mediterranean in vegetation although degeneration from original primal forest is very advanced locally with extensive maquis, garique and steppe conditions. Annual precipitation varies from circa 420mm in Tunis gradually decreasing to circa 300mm in the south. The sandy soil retains moisture well however, and a good covering of plant growth is supported. The rainfall is also supplemented with regular dew deposits and sub-soil water is also available in most locations. This area supports a productive agriculture. Tortoises are found in undisturbed habitats and on the periphery of traditionally cultivated olive and citrus groves.

The south of Tunisia

In the extreme south rainfall is very unpredictable and ranges from 150mm - 200mm along the narrow coastal band but elsewhere falling to less than 125mm. The vegetation in most areas is sparse but the land does support a very considerable number of date palms which are a major agricultural commodity. Tortoises are rare in this region compared to the north, but do occur in limited numbers as noted by Anderson (1892) and confirmed by Lambert (1983 and personal communications).
Valuable historical information on tortoise distribution in Tunisia is provided by Mayet (1903), by Mosauer (1934) who lists observations in the Rades, Sidi bou Ali and Kairouan regions, by Gadeau de Kerville (1908) and by M. Blanc (1935) who notes the presence of tortoises in the Khroumirie, Cap Bon, Kelibia, Sousse, Kairouan and region of Tunis. Other interesting accounts are provided by Boulenger (1891) who cites T. graeca in the ruins of Zarzis, by Konig (1892), Olivier (1896), Anderson (1892) and by Seurat (1922 and 1927). Additional recent data is provided by Blanc (1978) and Highfield (1990).
In December 1989 and April 1990 we recorded data from some 27 tortoises in Tunisia; of these, approximately half were captive specimens (usually kept as pets by villagers). The rest were recorded in the field. In the latter case, over 120 man-hours were expended searching previously recorded sites. The average find rate was 1 tortoise per 9 man-hours searching. The best site located yielded only 4 tortoises for 18 man-hours searching. This suggests that even in established and obviously suitable habitats the population density is now very low indeed. Other researchers who have visited Tunisia recently have also reported that tortoises are now scarce, despite many sympatric species (e.g Tarentola mauritanica, Discoglossus pictus, Ophysops occidentalis, Psammodromus algirus, Natrix maura, Lacerta lepida and Bufo spp.) being located with relative ease (Bruekers, personal communications). Unfortunately, tortoises continue to be collected from these sites for (illegal) sale to foreign tourists; both Hammamet and Sousse are centres for this activity. Tortoises are also killed, dried in the sun and varnished to produce 'souvenirs' for sale to foreign visitors. Such items can be openly purchased in many tourist centres in the country. In 10 minutes of surveying the shelves of one tourist shop alone we saw more dead 'souvenir' specimens for sale than we managed to locate live specimens in two weeks of intense searching. Many of the dead tortoises on display were juveniles.
Some general details of Tunisias unique fauna of land tortoises were provided previously (Highfield, 1990a). As additional data has now been collected and analysed, a more complete picture of these remarkable creatures is beginning to emerge.

Sexual dimorphism - diagnostic characters

This is particularly striking in Tunisian coastal zone Tortoises, where females weigh considerably more than males of equivalent age, even very mature specimens of which are extremely small. Typically, a sexually mature adult male weights approximately 372g compared to an average female at 607g. The heaviest male we encountered in the entire series weighed 550g compared to the heaviest female at 750g.

Length & height

The carapace length and height is also strongly dimorphic; adult females typically measuring 134mm SCL (straight carapace length) compared to a typical adult males 117mm SCL. Males have a low vaulted carapace, typically circa 54.23mm high, whilst females have typically attain circa 71.55mm. If a measurement is taken over the curve using a flexible ruler, females average 175.45mm and males 148.18mm. If a length-height ratio is prepared by dividing the SCL by the maximum carapace height, the average index for males is 2.16 and the average for females 1.87.

Width

The transverse median width of the carapace typically approaches or is equivalent to the maximum transverse width of the posterior marginals in the case of females, but in the case of males the posterior marginals typically project beyond the transverse median carapace width by several millimeters.

Supracaudal

Another extremely useful indicator is the supracaudal shape, which in males is posteriorlaterally projected and introflexed on its anterior face. In female specimens the supracaudal is reduced by comparison and lacks the anterior ventral introflexion .

Plastral sutures

During the course of routine morphometric data analysis, quite by chance it was noted that the plastral formulas of male and female specimens differed consistently. This is interesting, as the plastral formula (or sequence of length reduction in the mid-line sutures of the plastron) is often cited as a fixed character within a taxon.
Typically, where the femoral suture is subequal to the anal suture = Female , and where the femoral suture is equal to or exceeds the anal suture = Male. The overall relative dimensions are also very different with the average for males as P5 (femoral) = 13.36mm and P6 (anal) = 11.95mm ; for females the respective dimensions are P5 = 16mm and P6 = 18.14mm reflecting the larger overall carapace size and body mass of female specimens.

Post anal gap

It was also noted that the PaG of the female specimens was invariably less than that of the males - typically 15.59mm in females compared to 18.07 in males. There is a relatively high degree of xiphiplastral kinesis in both sexes, the xiphiplastra of females not noticeably more pliant than that of males.

Egg morphology

One of the most remarkable features of the Tunisian coastal tortoise is the incredibly tiny dimensions of their eggs; Mayet (1903) measured some at just 15mm X 13mm, dimensions we confirmed in many discussions with local people. These dimensions should be compared with typical egg sizes from other tortoises .
Our information is that the females can lay up to 12 of these tiny eggs, but more often lay clutches of 6-8. This in itself is quite remarkable, as most other small tortoises (e.g T. kleinmanni, Homopus etc.) lay fewer, but larger eggs. Large clutch densities are often associated with high levels of predation, and certainly one would expect such minute hatchlings to be especially vulnerable.
It should be noted that we qualify our comments to apply only to the "Tunisian coastal tortoise" - for we have reason to believe that other entirely different tortoises also exist in separate geographic populations within Tunisia. These other tortoises almost certainly are not closely related to the tiny specimens which inhabit the coastal zone and in our opinion, must comprise separate species.
The evidence in the literature for this is intriguing. Chaigon (1904) recorded finding three ellipsoid eggs "the size of a pigeons egg" inside a tortoise which he consumed for dinner. A large female recorded from Maktar in 1890 laid a total of 7 eggs which measured 28 X 36mm (Domergue, 1899). Another set of measurements made by Gadeau de Kerville (1908) list egg dimensions from 24.5 to 31 x 31.5 to 38.5mm. The laying female measured 202mm, which is far in excess of any specimen recorded in the coastal zone. Additionally, we doubt that the small coastal zone females could carry a clutch of between 4-8 eggs of such size, and certainly could not carry as many as 12 as reported by Mayet (1903). Our discussions with local people also resulted in reports of "very big" tortoises living in some distant (north-western) inland forests. We believe that these large tortoises may be related to some of the large specimens which occur in neighboring Algeria.
We certainly feel justified in concluding that geographically separate populations which consistently each lay eggs as diverse as 15 X 13mm and 28 X 36mm are unlikely to belong to the same species as such a range of deviation is unknown within a single taxon where the standard deviation within a single taxon for circum-mediterranean tortoise species is typically <+/-2.00mm in length and <+/-1.25mm in width (Highfield, 1990a and in press).
A Malacochersus female in our own collection regularly produces eggs measuring (on average) 42mm x 29mm, however in this case the posterior lobe opening measures some 29mm long by 33mm wide.
The most significant feature of our Tunisian coastal specimens is the combination of minute egg size plus the clutch density. A combination which is unknown in any other terrestrial tortoise. The inland populations, which produce the very large egg, for the moment remain a mystery.
Further research aimed at establishing the phylogenetic relationships and identity of the inland populations is proposed.

Hatchlings & juveniles

Clearly, the egg dimensions of the coastal zone tortoises impose size restrictions upon the hatchlings. Our experience from captive breeding north African and circum-mediterranean species suggests that the carapace length of emerging hatchlings tends to be equal to or <+3.00mm greater than the maximum length of the egg for egg sizes below 30mm in length and approximately equal to or sub-equal to the longitudinal length of the egg for eggs which measure in the range 30-40mm. The excess few millimeters above the egg length in smaller eggs are accounted for by the plastral fold present on hatchlings. We have noted that in some species hatchling orientation within the egg is typically longitudinal, and in others it is typically transverse.
This would suggest that the probable hatchling size of the Tunisian coastal tortoise is in the range circa SCL 15-17mm. This should be compared to a hatchling size of circa 33mm SCL for Algerian T. whitei, circa 27mm SCL for Moroccan T. graeca, 32mm SCL for T. hermanni boettgeri and an estimated possible hatchling SCL of circa 33-35mm SCL for the (currently unknown) inland populations of Tunisia.
The smallest tortoise we found was circa 45mm SCL and weighed circa 28g. We found this tortoise in the verge of a ploughed field adjacent to a well vegetated hillside where several adult tortoises have been recorded. There were four clear growth rings, and we would estimate the age of the tortoise at 18 months to 2 years. Locating juveniles at or below this size presents obvious difficulties for searchers.
Other, somewhat older juveniles were located, photographed and recorded. These included an immature male of 78mm SCL and two young females at 96mm SCL and 67mm SCL.

African land tortoises; some diagnostic criteria at Genus level

Carapace morphology and structure

Suprapygal morphology

The structure of the pygal and suprapygal bones of the carapace of chelonians is an important diagnostic character which may be employed to distinguish between genera and to provide an indication of evolutionary descent and phylogenetic relationships (Loveridge & Williams 1957; Crumly 1984 ; Meylan & Auffenberg 1986). The following are the generally accepted pygal characters of some currently recognised African cryptodiran genera;

GEOCHELONE Fitzinger 1835

Two suprapygals, the anterior larger, bifurcating posteriorly to embrace the smaller posterior element, which (on post-Eocene forms) is crossed near its middle by the sulcus between the fifth vertebral and the supracaudal. Concerning this genus, Loveridge & Williams (1957, p. 222) remark that "earliest of these (specialisations) are a thickened, produced gular region and the peculiar pygal pattern with the first suprapygal embracing a second smaller one, both already present in Hadrianus of the Eocene" (2). Both conditions are relevant to the new tortoises from n. Africa to be described herein. Of extinct genera, a similar suprapygal structure was shared with Hesperotestudo.

(2) Note, however that Hadrianus was synonymised into Manouria by Auffenberg (1971) as "no characters differentiated one from the other".

HOMOPUS Dumeril & Bibron 1834 (3)

Two suprapygals, the anterior larger, bifurcating posteriorly to embrace the smaller element which is adjacent to, but not crossed by, the sulcus between the fifth vertebral and the supracaudal (Loveridge & Williams, 1957, p. 353).

(3) Bour (1988) has pointed out that Mertens & Wermuth (1977) cited the incorrect date (1835) for the erection of the new genus of Homopus by Dumeril & Bibron. Bour also points out that the type species of Homopus is T. areolata THUNBERG 1787 a a result of later designation by themselves (Dumeril & Bibron 1835) and not by Fitzinger (1843) as is usually cited.

TESTUDO Linnaeus 1758

Frequently a single suprapygal, if two, they are typically separated by a straight transverse suture (Loveridge & Williams, 1957, p. 255).

PSAMMOBATES Fitzinger 1835

Typically a single suprapygal, if two, then separated by a straight transverse suture (Loveridge & Williams, 1957, Pritchard, 1979). (4)

(4) However, the illustration (p. 217) provided by Loveridge & Williams (1957) of the Psammobates suprapygal area distinctly shows a very small crescent shaped enclosed area above the pygal suture and not a straight transverse suture.

CHERSINA Gray 1831

Suprapygal typically single, if at all divided then with a straight transverse suture (Loveridge & Williams, 1957).

MALACOCHERSUS Lindholm 1929

Typically a single suprapygal (Loveridge & Williams, 1957).

KINIXYS Bell 1827

Suprapygals one or two, if two typically separated by a straight transverse suture (Loveridge & Williams, 1957).

IMPREGNOCHELYS Meylan & Auffenberg 1986 (Miocene, extinct)

Two suprapygals, a rounded second suprapygal resting within a rounded notch in the pygal with the posterior sulcus of the fifth vertebral scute at its dorsal edge (Meylan & Auffenberg, 1986).
The presence or absence of the Geochelone or Homopus type pattern, that is, a twin suprapygal consisting of a smaller element enclosed by posteriorlaterally directed rami, or the presence or absence of the more primitive single form, whether or not divided by a straight transverse suture, is usually considered sufficiently diagnostic in the instance of fossil carapaces to determine genus and establish phylogenetic relationships (e.g, see Crumly 1984; Pritchard 1979; Loveridge & Williams 1957; Schleich 1984 etc).
Most authorities have advanced the opinion that the single or transversely separated twin suprapygal should be considered a primitive form, or plesiomorphic character, (in that they more closely follow that of generalised emydids), and that the 'enclosed' form should be considered an advanced feature; the present author concurs with this view which finds support in that one of the most primitive extant forms, the rare asiatic Manouria impressa (GUNTHER 1882) possesses a keystone shaped transversely sutured suprapygal in combination with a hexagonal neural series.
However, even in the case of a generally accepted single genus there are often specific differences in suprapygal structure.
Testudo hermanni GMELIN 1789 has in the authors experience typically a single (keystone shaped) undivided suprapygal, and this tortoise is often regarded (with Testudo horsfieldi GRAY 1887) as a primitive member of the genus; Testudo ibera PALLAS 1814 also typically tends to feature a single suprapygal, although divided examples are encountered; Both forms however conform to the technical definition of the genus as a whole described above and as presently accepted by most authorities.

Supracaudal division

Reliance upon the division of the supracaudal scute as a diagnostic character should be regarded with some caution (Highfield, 1990).

Costal scute dimensions

Geochelone is differentiated from Testudo by the third costal scute being equal to or greater in length on its outside (marginal contact) edge than the fourth; in Testudo, by contrast, the order is supposedly reversed, the fourth costal being subequal to the third (Loveridge & Williams 1957., Pritchard 1979).
This deserves brief comment, as upon checking my own records of costal scute dimensions, I find that of 84 T. hermanni boettgeri for which data exists, without exception in every case the fourth costal outer suture is at least equal to and in the vast majority of cases wider than the third. If T. hermanni is to be considered Testudo, then this character cannot be used to separate Testudo from Geochelone. This same character state was also noted on the (admittedly limited) number of T. horsfieldi available to me. The single T. zarudnyi for which I have data possesses a 'standard' Testudo construction as do all of the (several hundred) T. ibera specimens for which I have records. The few T. marginata I have recorded also possess the typical Testudo pattern. See my reservations on the status of T. hermanni and T. horsfieldi (below). Of other African genera for which I have data, in Kinixys belliana the fourth costal is invariably wider than the third; although this has much to do with the uniquely kinetic carapace found in this genus.

Costal-marginal contacts

Bour (1989) points out that in Testudo, the 2nd costal typically contacts the 5th, 6th and 7th marginals. This trait is also present in Malacochersus. In Geochelone, contact is typically limited to the 5th and 6th marginals.
T. hermanni is again somewhat problematic with regard to Costal-marginal contacts; whilst most examples maintain 5th, 6th and 7th contacts a sizable proportion (30% of T. h. boettgeri in my sample) do not, instead making only 5th and 6th marginal contacts. The few T. horsfieldi examined to date demonstrate all three contacts.

Marginal scute morphology

The sequence of reduction in the marginal scutes can provide useful diagnostic information (the marginal formula); in skeletal material this is usually discernible via the scute sulci. The position of the marginal series lateral ridge is also diagnostic in many instances.

Xiphiplastral kinesis

It is generally stated that whereas Geochelone does not demonstrate xiphiplastral kinesis, Testudo does - at least in females, where the character is often said to be sexually dimorphic. Whilst it is true that xiphiplastral kinesis is not present in Geochelone, it can hardly be described as consistently present in Testudo hermanni. I have in my possession a large number of T. hermanni carapaces and within T. hermanni boettgeri (of both sexes) the xiphiplastra exhibit absolutely no degree of movement whatsoever. The same observation has been made on living specimens, where they are every bit as immobile as those of G. pardalis or T. horsfieldi.

This conclusion is shared by Devaux (1988) who has very extensive experience of the southern French population of hermanni;

"Certaines tortues ont des plastrons articules. La tortue grecque dispose par exemple d'une partie legerement mobile a l'arrier. La tortue d'Hermann, au contraire, a un plastron entirement fixe et rigide".

An identical opinion was expressed by Olivier (1894) who also commented upon the lack of articulation in T. hermanni compared to T. graeca but perhaps due to this authors confusing usage of synonyms (he habitually employed 'T. graeca' for T. hermanni and 'T. mauritanica' for T. graeca) his perfectly valid observations appear to have been overlooked.
On this evidence, the inclusion of the character as a definitive feature diagnostic of genus or as a key to determine sexual polarity appears seriously flawed.
On a behavioural note, it appears that at least one function of xiphiplastral kinesis is protection; when stimulated from behind, tortoises with this ability rapidly draw the rear lobe of the plastron upwards thereby offering improved protection to the legs and tail area. This is best observed on T. graeca, T. whitei and our Tunisian specimens where the degree of movement is considerable (in both sexes). Plastral mobility is also of obvious benefit in small species in respect of egg laying.
It must be remembered that Loveridge and Williams (1957) cite plastral mobility as a key diagnostic feature for the genus Testudo. In addition, plastral mobility is also absent in Testudo horsfieldi; a character which until quite recently was considered evidence that they belong to a different genus Agrionemys (Khozatsky & Mlynarski, 1966). In fact, it has been suggested that both T. hermanni and T. horsfieldi may share a common ancestry in the asiatic Ogliocene-Pliocene genus Protestudo (see Chkhikvadze, 1970 & 1971); a hypothesis which I do not necessarily view with antipathy. Certainly, despite acknowledging that there are indeed many character states shared by hermanni and horsfieldi, and together (both cranial and carapacial) shared with T. ibera and T. zarudnyi, I confess to not being completely satisfied with their current taxonomic position and am of the general opinion that much more systematic research needs to be undertaken to resolve the many outstanding questions relating to their phylogeny. The issue of their lack of xiphiplastral kinesis is particularly problematic.

Comparative morphology of the neural series

The neural series are also extremely useful comparative characters, the primitive hexagonal emydine pattern being replaced by alternately quadrilateral and octagonal arrangements in more developed forms. Of extant southern African forms, Homopus has attained a unique semi-developed state in that it has departed somewhat from the primitive hexagonal pattern, being based upon hexagonal and quadrilateral elements (but none octagonal) and combines this with suprapygal elements which Loveridge and Williams (1957) describe as being in an incipient stage of the advanced (i.e Geochelone) pattern from which it differs in that the sulcus of the 5th vertebral and supracaudal does not cross the middle of the second (enclosed) suprapygal element. This latter feature is also relevant to the specimens to be described in this paper from n. Africa.

External morphology of the 1st vertebral scute

The shape of the first vertebral scute, which can generally be classified as 'rounded' or 'angular' has proved to be an extremely useful and highly reliable diagnostic character (Highfield & Martin, 1989). In particular, this character can be used to differentiate with a high degree of accuracy n. European and Eurasian forms such as Testudo ibera and T. hermanni from most forms of n. African origin such as T. graeca and T. whitei. European and Eurasian forms typically possess a more or less angular frontal vertebral scute whilst T. graeca and T. whitei typically feature a rounded form.
It is our experience that typically the sulcus of the 1st vertebral scute is very sharply defined on the underlying bony tissue and this can provide useful data even in the absence of the actual scute.
The present author was struck by the complete dissimilarity between the shape of this scute in T. graeca from that observed in European and Eurasian forms (to which T. graeca supposedly bore a sub-specific relationship) and the marked similarity between that of T. graeca and the form observed in certain other African forms particularly Geochelone pardalis.

Gular morphology

The gular region of tortoises is of considerable interest and can provide very good diagnostic data. Obviously, the most distinct forms are those encountered in Gopherus (Xerobates) spp., in Asterochelys yniphora or in Chersina angulata where the projection is developed to a very high degree.
However, the gular region and associated epiplastron is generally distinctive in most groups and can provide a number of clues of considerable phylogenetic value.

Significantly, the gular scute width of the Tunisian specimens examined typically was greater than the length; very much so in some individual specimens.
The gular scutes in many specimens were quite unlike anything ever observed in any Testudo species; one was immediately reminded of Homopus. In a high percentage of specimens, the gular scutes extended right across the anterior of the epiplastron being extremely wide and narrow.
This character state is quite remarkable, and is entirely inconsistent with that found in T. hermanni, T. ibera and other circum-mediterranean species in my experience. We found this state in two very young juveniles, several sub-adults and several adults of obvious maturity. So it is not an ontogenetic effect. When plastral photographs are compared with Testudo specimens, the difference is immediate and striking.

Gular rise and excavation

Also very notable was the degree of gular rise above the plastron; in many cases the angle of rise was very acute, the anterior epiplastra being markedly convex and in all cases was considerably in excess of that which one is accustomed to observing in circum-mediterranean species.
Internally, the degree of excavation of the gular and epiplastron may be considered an indication of advanced or primitive derivation; Manouria impressa (on the basis of the 3 carapaces available to us) entirely lacks excavation, Manouria phayrie (Anderson 1872) is also ill-developed and in T. hermanni the degree of epiplastral excavation is also usually severely restricted; in T. ibera and most n. African specimens examined by us, as well as in most Geochelone specimens, internal epiplastral excavation is typically well developed.

Cranial osteology

Cranial characters are extremely important and valuable diagnostically, and have been discussed in considerable depth by Gaffney (1979), Crumly (1982) and with regard to Testudo by Bour (1989) among other commentators. Cranial characters are however not analysed in any detail in this present paper in connection with the specimens described as further work on these is in progress; it is a complex subject which requires detailed discussion and it is intended to produce a separate paper illustrating and describing their characters at a later date. Certain cranial characters must be referred to, albeit only generally in our present context.

Prootic shape and exposure

The shape and percentage of dorsal exposure of the prootics (which are pierced by the stapedial foramen) are determined by the overlap of the parietals; this is however a character which to a greater or lesser extent is affected by the overall size of the cranium. Small specimens typically showing less prootic exposure than large specimens. In large specimens (e.g Geochelone [Aldabrachelys] gigantea, G. [Chelonoidis] elephantopus etc.) the anterior prootics may be revealed with a different shape to the posterior prootics, a feature which is clearly of diagnostic value. Most species of Geochelone have prootics which are narrow throughout their length or wider posteriorly than anteriorly (Crumly, 1982).
The degree of dorsal prootic exposure is particularly important since it is one of the principal characters employed by Loveridge & Williams (1957) to differentiate between Geochelone and Testudo. These authors considered the character so important that they italicised it in their diagnostic criteria for Testudo outlined on page 254; "prootic typically concealed dorsally and anteriorly by parietal". However, the same authors (on page 218) previously state "prootic completely concealed by the parietal" (my italics). There is an obvious difference between "completely concealed" and "typically concealed"!. Bour (1989) prefers to employ the term 'reduced' compared to Geochelone rather than concealed.
Smaller genera typically exhibit much less of the prootic dorsally than do inherently larger genera; however, there are exceptions in that Malacochersus, for example, reveals a relatively high percentage of prootic for its size.

Triturating surface of the maxillary

This character is employed frequently to diagnose Geochelone, although Bour (1989) has also pointed out specific differences between T. graeca, T. (graeca) ibera, T. Marginata, T. kleinmanni, T. hermanni and T. horsfieldi. There tends to be an inherent difference between large and small tortoises. In African genera, the triturating surface of the maxilla in G. pardalis is particulary serrated.

Supranasal squamation

Loveridge & Williams (1957) cite the presence of supranasal scales as co-diagnostic for Testudo and point out that these are also present in Malacochersus; they are absent in Geochelone.

Distribution

Until very recently the universally accepted distribution of Testudo in north Africa was that Testudo graeca LINNAEUS 1758 occurred throughout Morocco, throughout Algeria, into Tunisia and extended to Libya where it eventually connected with Testudo kleinmanni which then replaced it. One single report of Testudo graeca exists for Egypt (quoted in Lambert, 1983) although as no description or illustration is provided for this creature it is impossible to determine precisely what it was; obviously it was something other than T. kleinmanni but on our evidence, we doubt that it was indeed a true T. graeca but rather some externally similar Libyan form as yet unclassified. There is also always the possibility of an introduced specimen (see below); this grim spectre, which was surely sent specifically to haunt taxonomists, is also discussed by Loveridge & Williams (1957). My own experience of Libyan tortoises however suggests that there are indeed some which externally and superficially at least resemble T. graeca, but which on close (or osteological) examination feature character states which conclusively separate them, e.g a Geochelone-Homopus type suprapygal. For a review and illustrations of some Libyan tortoises see Schleich (1989). See also the data on our Libyan specimen from Derna discussed later.

Introductions

The natural European distribution for T. graeca L. 1758 is limited to a relatively small population in Southern Spain (Lopez Jurado et al., 1979) although secondary populations are also reported from Mallorca - certainly in the early 1980's pet shops in Palma were selling T. graeca shipped from southern Spain and one faunal checklist implies that T. graeca is extinct on the island (Kramer & Vickers, 1983). Some T. graeca from the Rif in Morocco were also deliberately introduced to the Coto Donana nature reserve in SW Spain (Valverde, 1960). Other reports indicate small populations of alleged introduced T. graeca in southern Italy, and on the islands of Sardinia and Sicily (Bruno 1970; Bruno and Maugeri, 1976). The present author has been fortunate enough to examine closely a specimen of alleged T. g. graeca from Sardinia and can state with confidence that it is not a T. graeca L. 1758 at all but is instead one of the much larger Algerian species with the remarkable suprapygal construction to be described. Apparently Sardinia was a popular place to release tortoises as various dislocated oddities have been recorded from there including alleged T. marginata (Bruno and Maugeri, 1976; Hellmich, 1962). Although this latter reference may actually concern large Algerian derived specimens which have a long history of being mistaken for T. marginata on account of their pronounced posterior marginal lateral extension (e.g see Strauch 1862).

The Mertens hypothesis

Within Testudo graeca L. 1758 Mertens (1946) alleged four sub-specific forms; the evidence presented for this however is poor, and is in one case based upon a single preserved specimen (Testudo zarudnyi NIKOLSKI 1896) and in another instance upon no specimen at all (or even an illustration of one) and an admitted complete lack of any personal knowledge of the alleged animal (Testudo floweri BODENHEIMER 1935). This latter attribution was revised in 1958 to Testudo graeca terrestris FORSKAL 1775 by Wermuth who concluded (also in the absence of any substantive evidence and again without actually ever having seen any alleged T. floweri) that the mysterious floweri form was identical to two specimens examined by himself; one from Syria and one from Derna in Libya. Unfortunately, neither of these specimens appears valid as Forskals 'terrestris' and the status of any nomenclatural act by Forskal is also highly questionable in this particular instance for reasons discussed at length in a paper shortly to be published (Highfield & Martin, in press). The entire sub-species hypothesis as proposed by Mertens for T. graeca is not accepted by the present author. Some reasons for this are discussed in a previous paper (Highfield & Martin, 1989). Suffice to say, Mertens found "only relatively minor differences in shell proportions" between n. African and Caucasian 'T. graeca'. A somewhat astonishing conclusion in view of the true level of divergence and range of sizes which actually does exist between the populations. At the same time, it is necessary to recognise that Mertens based his conclusions upon only a small handful of specimens - and his observation that n. African tortoises "maximum size....is set at 250mm" also suggests that he was unfamiliar with all of the available literature concerning this subject. 250mm is in fact the standard or average SCL of female T. whitei from Algeria, many specimens attaining 280mm and one 30 year old specimen in our own collection measures 292mm.
None of this, however, directly affects the issue under discussion here which is the general distribution of Testudo as a genus in north Africa and in Tunisia in particular. Taking the African continent as a whole, the following situation represents that currently accepted by all known authorities.
Testudo is here limited in distribution to the coastal band running from Morocco to Egypt. Some climatic and geophysical reasons for this distribution are provided by Lambert (1983). The nearest points of (extant) Geochelone approach are the Sudan, Mauritania, Ethiopia and Eritrea for G. sulcata and the Sudan and Ethiopia for G. pardalis. The closest approaches of Kinixys (as K. belliana belliana) also occur in the Sudan, Eritrea and Ethiopia. Homopus, Psammobates, Chersina and Malacochersus as genera are exclusively restricted to distribution in southern Africa.
Loveridge & Williams (1957) nominated Pseudotestudo as a new subgenus to include Testudo kleinmanni LORTET 1887. Recently, however, Bour (1989) has shown that the characters supposed for this genus (quadrate not enclosing stapes) are based upon juvenile phase cranial development within Testudo. This tortoise should therefore continue to be assigned to Testudo.

Fossil history

It is clear from fossil evidence that the distributions of these genera were once very much more extensive than today. Geochelone in particular was a very successful and widely distributed genus. Testudo was similarly widely distributed with extensive fossil evidence from Europe especially. Of Homopus we unfortunately know much less due to the extreme paucity of specimens (of any age).
For a general overview of fossil land tortoises see Auffenberg (1974), Crumly (1984), Mlynarski (1969 & 1976), Meylan & Auffenberg (1986) and Gaffney (1979). Some interesting hypotheses upon the distribution of fossil tortoises associated with changes in past climates derived from paleobotanical data are presented by Brattstrom (1961) who also provides a series of very useful maps covering periods from the Cretaceous and Eocene to the Pliocene and Recent.
Some specifically north African fossil forms are discussed by Dacque (1912), Bergounioux and Crouzel (1968) and by Bergounioux (1952 & 1954-55) who in the latter publication described Testudo semenensis from Djebel Semene in Tunisia.
Unfortunately, very little fossil material from north Africa has ever been collected and classified; for obvious reasons, chelonian material from what is now the sahara desert is virtually non-existent. Our knowledge of the evolution of land tortoises in north Africa is therefore extremely poor.
Much of what has been accepted has been done so on the basis of assumption and speculation rather than on the basis of any physical evidence which in most cases is entirely lacking.
The view that Testudo, and only Testudo remains the only extant living genus in the region has not, however been seriously challenged until now; the only exception to this being the erection of Pseudotestudo by Loveridge & Williams (1957), a conclusion and nomenclatural act now considered invalid for the reasons stated by Bour (1989) and with which the present author concurs fully.

North Africa; A tortoise Galapagos of the Mediterranean?.

If what all known authorities currently state to be true actually is true, then one would not expect to find tremendous divergence among the tortoises of the region; those of Morocco should not differ greatly from those of, for example, Tunisia, Algeria or Libya; and there should definitely be no major and consistent structural divergence from the criteria accepted for the genus as a whole, or indeed between any series of n. African specimens and a similar series of, for example, Testudo (graeca) ibera PALLAS from Europe.
Let us address this question directly. In Algeria there exists a tortoise which habitually attains quite tremendous dimensions. The largest (female) example seen by the author measured 292mm in straight line carapace length and weighed some 4,650g (Highfield, 1990 and in press). This tortoise (of which I have had the opportunity of examining very many similar specimens both alive and prepared) conforms to the type specimen of Testudo whitei BENNETT 1836. Some details of this species have been presented earlier (Highfield & Martin, 1989). This same animal lays eggs which typically measure some 33mm long x 27.5mm wide. The resulting hatchlings measure some 33mm in length.
In Tunisia, by contrast, exists at least one population of tortoises which even as fully grown (female) adults only ever reach a maximum straight line carapace length of some 150mm and which typically weigh less that 750g. These animals lay eggs which typically measure a tiny 15mm long X 13mm wide.
By almost every criteria it is possible to devise, these animals are entirely and consistently different from one another; size; markings; colour; carapace shape; cranial shape; egg size; hatchling size and internally they differ in osteological structure. Yet for 250 years they have generally been regarded as the same species, and since Mertens paper of 1946 as the same sub-species. If these are indeed to be regarded as the same species, then what of the status of Galapagos tortoises or even of Darwins famous finches?. In the latter case, identification is said to be difficult even for an expert ornithologist in some instances, so slight are the external signs of divergence and speciation (Jackson, 1985).
Furthermore, these are not the only forms which exist in the region which fail entirely to conform to diagnostic criteria for the species said to be in exclusive occupation, namely Testudo graeca LINNAEUS 1758.
The author has examined many hundreds of specimens from Tunisia, Algeria, Libya and Morocco both as living animals and as museum specimens and found very few which actually conform to the characters exhibited by the holotype of Testudo graeca. The majority do not. Of those that do, they have come exclusively from the region between Oran in Algeria and Tangier in Morocco. No T. graeca L. 1758 have been detected from either Tunisia or Libya. All the tortoises without exception seen by the author from these countries have been forms currently unrecognised by science, and yet which are remarkably different from each other in various ways. I exclude from this claim individual animals of curious structure; in every case at least two specimens of each form have been acquired, and in some cases large series studied both as living animals and as preserved specimens.
Let us now turn to those specimens; these should demonstrate a limited range of divergence, and should without question conform to the criteria already outlined above for Testudo graeca specifically and for Testudo as a genus.

They manifestly do not.

Comparison should in all cases be made to the Holotype of Testudo graeca as described below.

Testudo graeca - the holotype

The holotype of Testudo graeca LINNAEUS 1758 is founded upon the illustration which appeared on page 204 of George Edwards Natural History of Birds published in London 10 years previously (see also Bour, 1987).

The figure "represents it of its natural bigness" and depicts a well drawn and accurate carapace measuring 104mm in length and 64mm in height. The head, legs and tail were not available to the artist and these leave something to be desired in the way of accuracy, being more representative of a lizard than a tortoise. The plate is hand coloured and shows a small tortoise having a general yellow groundcolour with dark brown to black markings.

Of this specimen Edwards writes;

"I had the male and female of this species; they lived two years with me, in the garden of the College of Physicians, London. In the warm months they copulated by leaping, in the common way of most four-footed animals. I was in hopes of propagating the species, but could never see any of their eggs in the places where they scraped holes.
The iris of the eye was of a reddish hazel-colour; the lips were hard, like the bill of a bird. the head was covered with scales of a yellowish colour; the neck, hinder legs and tail were covered with a flexible skin of a dirty flesh colour that they might be the more pliable to be put forth and drawn into the shell. The fore-legs were covered with yellow scales on their outsides which are partly exposed when the legs are drawn in. The shell is round, and pretty much rising on its upper side, and flat underneath; it is divided into many compartments, or separate scales, which have furrows or creases all round them, lessening one within another to the middle part of each scale. The shell is of a yellowish colour, clouded and spotted with large and small irregular spots of dusky or black; the vent is in the tail itself, which the female turns up in coition, and the male turns his tail inward under it, which brings the vents of each to touch. It hath five claws on each foot forwards, and four on each of the hinder feet. When they apprehend danger, they draw the head, tail and legs into the shell so that they cannot be easily hurt.
This tortoise was sent to me from Santa Cruz in West Barbary by my late friend Mr. Thomas Rawlings, Merchant, who died there (Anno 1748) after some years settlement in that country".

Type Locality

The locality of "Santa Cruz in West Barbary" refers to the old fort of Santa Cruz located on a hill close to the city of Oran, Algeria.

Nomenclature

In the main heading opposite the illustration the animal is referred to simply as 'The African Land Tortoise'; later it appears as Testudo tessellata africana minor (nomen nudum). It received the name Testudo graeca in the 10th edition of the Systema Naturae of 1758.
Of the name Testudo graeca applied by Linnaeus to a tortoise exclusively from Africa, Statius Muller (1774) explains;

"The Mosaic tortoise, Testudo graeca. The artistic placing of various coloured stones into shapes is called mosaic or musaic work and this art came from Greece to Italy 500 years ago. If one now perceives that the shell of this variety of tortoise is covered almost exclusively with square 'leaves', which have a number of hollows in the squares, making even smaller squares, one will immediately realise the origin of the name 'Greek' or 'Mosaic' tortoise".

Synonymy

The synonymy of Testudo graeca is extensive, but perhaps the best known synonym is Testudo mauritanica DUMERIL & BIBRON 1835.

Dimensions & probable age of the specimen

At 104mm this is undoubtedly a young specimen. The average adult size of male tortoises examined by the author from northern Morocco and the region of Oran is 145mm. This figure agrees closely with that of Lambert (1982) who studied tortoises in the same region. Annual growth lines are clearly visible in Edwards illustration and vary between 9 and 13 in number on the various scutes; from this, we can infer that the animal was probably approximately 10-12 years of age. The latter growth lines are fairly wide and all appear well defined. In aged specimens this is not so. The attained length of 104mm would also be consistent with an estimated age of 10-12 years in this species. Sexual activity can certainly occur at 7-8 years of age, so again this tends to confirm an age of circa 10 years for Edwards specimen.

Carapace height & curvature

Measured from the illustration this is revealed as 64mm, producing a height/length ratio of 1.64 - entirely consistent with young male T. graeca from the same locality today (typical adult male ratio range 1.70 - 1.85).

Variation of colour & morphology within T. graeca

Gross variations within the true species from the region of the type locality have not been encountered - all individuals conforming quite closely to the holotype in colouration, markings and overall morphology. Some darker individuals are encountered (especially in the north of Morocco), and in some specimens the marginal 'saw-tooth' or 'V' markings are more or less defined. However, in the principal specific characters comprising the unique frontal vertebral scute, adult length and body mass range no significant deviations have been noted in confirmed specimens from the type locality.
The holotype of T. graeca when studied closely is revealing; it depicts a tortoise bearing characters which place it firmly from the stated type locality and which can be seen, in identical form, in tortoises from that particular locality today. What is significant is that these characters seem restricted to a population of fairly limited distribution and are not seen in other specimens from different localities in north Africa.

The holotype should now be compared to the n. African specimens described below.

Specimen = Holotype Furculachelys nabeulensis n. gen. n. spp.

Carapace morphology	Holotype	Topotype
SCL:	121.00mm	120.00mm
Plastron length:	111.00mm	N/A
Carapace height:	64.50mm	62.00mm
MwM:	88.50mm	93.00mm
Mw:	85.00mm	86.50mm
Fw:	78.00mm	80.00mm
LoC:	155.00mm	160.00mm
Epiplastral excavation:	Yes	Yes
Gular length:	18.50mm	18.00mm
Gular width:	24.50mm	22.00mm
Gulars enter entoplastron:	(6.00mm)	(7.00mm)
Gular height:	15.50mm	15.50mm
Skeletal mass:	112.00g	110.00g
PaG:	22.75mm	N/A
Anal notch width:	28.50mm	N/A
Epiplastron midline length:	8.00mm	8.50mm
Entoplastron midline length:	19.50mm	21.00mm
Hyoplastron midline length:	15.00mm	17.00mm
Hypoplastron midline length:	24.50mm	24.00mm
Xiphiplastron midline length:	25.50mm	N/A
Average marginal series width:	14.14mm	14.59mm

Topotype: An adult male from the Tunisian coastal zone, precise locality unknown due to the specimen having being subject to transport. For general characters see description of Holotype, from which it differs only in possessing an additional suture in the 1st suprapygal. No other significant variation from the Holotype, other than purely allometric deviations are evident. The xiphiplastron of this specimen is missing.

Holotype: An adult male, found lying on its back dead in a forested area in the region of Nabeul, Tunisia in April 1990 by Miss M. Hill. Most body tissue had decayed, and there was evidence of rodent attack, the anterior part of the skull having been consumed. THe carapace however remained intact with the exception of a displaced pygal (which was found nearby and subsequently restored to the carapace) and the 2nd suprapygal element which is still missing. However, as the 1st suprapygal and pygal are both present, the shape and size of this are evident.

Descriptions and comparisons: This is an intriguing and highly interesting specimen. The carapace is very well preserved and fortunately we also have a large mass of data from living specimens from the same locality with which to compare it; this leaves no doubt that it is a very representative sample and is in no way abnormal for the locality.
There are 8 pairs of pleurals, and 11 pairs of marginals (peripherals). The sulci of the marginal scutes in each case cross the marginal bones at the midline between sutures with a slight posteriorly directed inclination, and the 3-4, 6-7 and 7-8 marginal sutures make perfect contact with the 1-2, 3-4 and 5-6 pleural sutures respectively; a total of 3 lateral peripheral-pleural contacts. There is only a very weak lateral ridge to the marginals with 75% above and 25% below.
The marginal scutes reduce in a series (measured at their outer edge) from 2>1>9>10>4=11=8>5>6=7 (19mm, 18mm, 17mm, 15mm, [13mm x 3], 12.5mm [11.5mm x2]), a total marginal series width of 155.50mm. If this figure is divided by 11 the result is an average marginal series width of 14.14mm. The AMSW is a very useful guide figure, being both dimorphic and, when combined with the marginal formula, specifically diagnostic.
The 7 neurals are configured 4-8-4-8-4-8-6, the latter element contacting the anterior edge of the 1st suprapygal. The suprapygal itself consists of two elements (the second missing) the former bifurcating posteriorly in the form of two rami which enclose the smaller secondary element . The sulcus of the supracaudal-fifth vertebral scute does not cross the suture of the pygal and 2nd suprapygal but coincides precisely. The pygal is a straight oblong, convex on its outer face and somewhat introflexed on its ventral face measuring 19.75mm x 15mm with a maximum tissue thickness of 6.75mm. The nuchal bone measures 17mm wide at its anterior edge, and attains a maximum width of 31mm. It measures 15mm deep. The outer edge of the 3rd costal scute-marginal contact sulcus measures 23mm and that of the 4th 20mm. Thus the 4th is subequal to the 3rd. The outer sulcus of the 2nd costal contacts the 5th, 6th and 7th marginal sulci, the 1st costal contacts the 1st, 2nd, 3rd, 4th and 5th marginal sulcci, the 3rd costal contacts the 7th, 8th and 9th marginal sulci and the 4th costal contacts the 9th, 10th and 11th marginal scutes.
The posterior suture of the entoplastron contacts the humeropectoral sulcus, but is not actually crossed by it; the entoplastron is entered by the posteriormost component of the gular sulci. The small inguinal scutes do not obviously contact the femoral scutes.
The gular region is paired, thickened and anteriorly terminates simultaneously on a plane with the nuchal zone, not projecting at all beyond it; internally, the epiplastron is quite substantially excavated.

The bridge marginals are not significantly expanded dorsally and there is but a weak lateral ridge. This character is powerful in primitive tortoises (e.g Manouria) and its reduction is considered derived. Of southern African genera, Homopus and Psammobates both possess a significant lateral ridge combined with unexpanded bridge marginals. Loveridge and Williams (1957) state that complex head squamation is to be considered primitive in that it resembles that of generalised emydines, e.g supranasals, prefrontals and frontals. Such squamation is present in both Testudo and in Malacochersus. It is absent in Geochelone. In fact, it is generally present in all north African tortoises, both Testudo and Furculachelys.

Remarks: This carapace is conclusively and absolutely separated from Testudo by the bifurcating suprapygal which in many respects approaches that of Homopus; it differs from Homopus however by the possession of a series of octagonal-quadrilateral neurals. It is also separated from Geochelone by possessing a xiphiplastral hinge. The sulcus of the supracaudal and 5th vertebral intersects at the exterior suture of the suprapygals and pygal; a character state not present in post-Eocene forms of Geochelone. Further, the degree of rise or curvature on the anterior face of the 2nd (enclosed) suprapygal is quite extreme and definitely closer to that seen in Homopus than in Geochelone (especially the extant s. African G. pardalis); it is, in fact, closer to an inverted 'V' shape rather than a simple arc of a circle. Another character state inconsistent with either Testudo or Geochelone is that the gulars are paired, but typically very profoundly broader than long not only in the Type Specimen but also throughout the general population of the Type Locality, sometimes extremely so, in a manner again very reminiscent of Homopus.
Unfortunately, only a relatively small skull fragment survives, but from this it is obvious that the parietals almost completely obscure the prootics; the species is therefore defined upon the small adult body size as described above, and the genus upon the following principle combination of character states; the prootics entirely or almost entirely concealed by the parietals; the mobile xiphiplastron; the gulars typically wider than long; the two or three-part suprapygal, the outer larger element bifurcating posteriorly to embrace the second, smaller element which rises approximately to or just above the midline of the first; the first element having no transverse divisions or one single transverse division located above the anterior suture of the 2nd (enclosed) element; the epiplastron rising on its anterior face and internally excavated below the introverted gular lip; the neurals as described above.

Etymology: The generic name Furculachelys is derived from Furcula (Latin = forked) and Chelys (Latin = tortoise) indicating the principal diagnostic condition of the suprapygal. The specific name nabeulensis refers to the terra typica of Nabeul, Tunisia.

Living appearence: In life, these tortoises are very brightly coloured with a yellow groundcolour featuring black markings; the vertebral scutes feature a black central blotch, anteriorly and laterally bordered by a brown-black band. The costal scutes typically feature a centralised black dot, more or less enlarged or distinct, in some specimens suurrounded by additional irregular small black dots or blotches; the 2nd, 3rd and 4th costals are tyically bordered anteriorly and at the marginal contact with bands of black, but the 1st costal typically lacks the anterior border at the 1st vertebral contact. The marginals are typically bordered anteriorly with a moderately narrow black band at each contact; in some specimens this marking is more extensive, becoming almost triangular; in yet other specimens, the marginals feature a series of lateral dots rather than any anterior borders. The plastron normally features a large central diffuse black barking centered upon the abdominal region.
The markings of hatchlings are similar, but typically the groundcolour is a paler shade of yellow and the carapace markings tend to be browner, rather than pure black.
There are very noticeable regional variations in marking; even some neighbouring populations each possessing distinct variations upon the basic pattern. In one instance, we found two populations separated by only a few kilometers which each bore striking individual characteristics. Interestingly, every tortoise examined within those populations bore the same 'local' marking, but each population was very different from the other. A similar situation exists in respect of American Box turtle populations (Terrapene spp.)
The scales of the front legs are typically light or sandy yellow, sometimes tipped with black; the skin itself is usually sandy yellow. Dorsally, the head is usually centrally marked with a large yelow blotch which may be more or less distinct. There are typically two bright yellow supranasal scales. The hind legs may feature large spike-like scales, typically yellow in colouration, at or about the heels. There are small thigh tubercles present, occasionally paired.
The eyes are small, very black and bright. When captured, female specimens often retreat into their shells and peer tentatively out, but male specimens show a remarkable lack of fear, waving their front legs wildly in anger and frustration, and when finally released, make their escape into the nearest cover at a remarkably high speed. Two males placed together will immediately begin to fight, circling for advantage with much butting and some occasional biting.

For additional data on Furculachelys nabeulensis, see the earlier comments on these tortoises in the same authors 'Preliminary Report' (Highfield, 1990).

Tortoises, species and endemism; discussion, problems and a hypothesis

Endemic taxa at all levels (genus, species or sub-species) are those which are confined to a discrete geographical region. Divergence, variation and ultimately selective evolutionary developments are most likely to occur when individual populations are isolated from one another by geophysical or environmental factors which limit or prohibit genetic interchange between them. Islands are for obvious reasons the classic example of this mechanism.
However, islands do not represent the only circumstances under which this mechanism can operate; north Africa with its isolated djebels (mountains), massifs, plains, oases and hammadas is an equally stimulating environment for natural selection, adaptive evolution and divergence.
The required time-scale for such speciation to manifest however is problematic; nonetheless, it is a fact that many n. African and particularly Saharan mammals demonstrate sufficient morphological divergence throughout their geographical range to be accorded sub-specific status (Ranck, 1968).
One factor which is of critical importance in respect of tortoises is these animals already extremely limited potential for mobility. Wide ranging gene flow between isolated populations is rendered, by this factor alone, much less probable than would be the case with a more mobile animal. Tortoises also have very specific environmental requirements; to a tortoise, an exposed area offering no cover or retreat from the heat of the day is as impassable as the highest mountain. Such factors contribute to the continuing long-term isolation of individual populations. For all practical purposes, an isolated oases or hillside population of tortoises may as well be on a real island; they cannot leave their environment and no new genetic material can reach them.
Neo (1978) reports that heterozygous values generally tend to be higher in reptiles than in birds or mammals, however island populations are more generally homogenous (Frankel and Soule', 1981). Because of their highly specific biotypic requirements and generally isolated habitats with minimal transit of individuals between populations a high degree of genetic convergence is often found within groups of reptiles. Where a greater exchange of individuals occurs, in more easily traversed and larger habitat areas then a state of balanced polymorphism may occur between homozygous, dominant homozygous and recessive individuals (Croudace, 1989).
These factors may have combined to accelerate speciation and adaption within the n. African environment.

With respect to tortoise populations, it is worth noting the following basic criteria which tend to exist generally which prohibit interbreeding between various species;

a) They are incompatible structurally

b) They occupy different habitats

c) They have different breeding times or mechanisms

d) They are incompatible genetically

Insofar as n. African tortoise populations are concerned, it appears that all four factors are present to greater or lesser degrees.
Of the first point, structural incompatibility, it would be very difficult indeed to find a better illustration of this than is to be found in the case of Testudo whitei and the miniature tortoises of Tunisia and Libya. The possibility of successful mating occurring for physical reasons between these two groups has to be considered highly unlikely to say the least. A T. whitei male for example is on average some 350% larger by body mass than a female from Sidi Kalifa and intromission must be considered a virtual impossibility.
Of the last point, unfortunately at this stage little reliable data actually exists; however, given the range of other divergent characters present it seems no unreasonable to assume that marked genetic divergence must also exist and that it probably is sufficient to inhibit interbreeding in at least the majority of cases. Certainly, the authors own experiences in captive breeding n. African tortoises strongly suggests that unless obviously 'identical' pairs are used fertility is generally nil or at best very poor. The only consistent successes obtained have been where both parents are quite clearly very convergent in all recognisable characters.
It is hoped to present additional data on this subject in a subsequent paper; however, data already collated is sufficient to demonstrate beyond all reasonable doubt that different populations (or species) produce uniquely formed eggs, have hatchlings of different size on emergence and different early phase growth rates. This is in addition to their possessing unique visible characters such as body shape or markings which are always entirely consistent with others derived from the same population, but which are often quite unlike those of neighboring populations.
It is the opinion of this author that in n. Africa there exists an as yet indeterminate number of species of land tortoise which have in error until now been consigned under the general catch-all nomenclature 'Testudo graeca'. Most of these species have never been previously figured or described by science. Despite in some cases really quite staggering divergence from the accepted holotype and the (fortunately still extant) population of the Type Locality.
It does, however, now require the fullest and most comprehensive investigation possible in the interests of conservation. Not only that, but the lessons which could be learned of evolution generally from this situation demand that urgent action is taken to preserve this absolutely unique diversity of tortoise species; if something effective is not done, then these animals face extinction almost before they have been introduced to science such are the pressures they currently face.

One is left to speculate about where these tortoises actually came from. The hypothesis advanced by Loveridge and Williams (1957) is that;

"the separation of the African and east European races has not been of long standing, it is natural to infer that formerly a continuous population extended across southern Europe whence the invasion of Western north Africa occurred via Spain and not across Egypt. In this view, the southern Spanish population of T. graeca would be a relict one, persisting after the general extinction of the species in western Europe".

At the risk of oversimplifying the situation, the overall view held by the present author is that T. ibera, T. zarudnyi, T. hermanni, T. horsfieldi and T. marginata are all descended from entirely different (northern-asiatic) roots than the majority north African populations, which, I believe it can be demonstrated (most especially by the suprapygal construction), have much more in common with extant s. African forms than with any extant European form. I also hypothesise that it may well be necessary to consider T. horsfieldi and T. hermanni separately as despite the cranial evidence for their belonging to Testudo many other powerful character states tend to disassociate them.
On this topic, which cannot be discussed at length here, the evidence of Kirsche (1984) is often cited; the hybridisation of T. horsfieldi with T. hermanni does not prove that horsfieldi is Testudo, but does support a close relationship between the two. A view which I also share.
Various aspects of divergence between the African graeca and its alleged European sub-species have been described by other authors, most notably by Bour (1989) who analysed cranial material and by Obst and Ambrosius (1971) who studied serological evidence. Considered with other evidence both sets of results must call into serious question Mertens' (1946) hypothesis of the sub-specific status of the European tortoises. The suprapygal evidence presented here is in any event overwhelming and conclusive; these tortoises cannot possibly be Testudo graeca and cannot possibly be closely related to either T. hermanni or to T. ibera, T. marginata or T. zarudnyi.
If a close phylogenetic link between northern and southern African forms is hypothesised (and the suprapygal evidence in particular would strongly support this), then one does not have to look vary far to find one possible reason why they are now separate; the encroachment of the Sahara within the continent.
Until only comparatively recently in evolutionary terms, large areas of what is now barren desert were lush savannah and thick forest. 1,000,000 years ago the climate of n. Africa was essentially tropical. 50,000 years ago mankinds ancestors discovered fire and began to use axes to clear the forests. 10,000 years ago the Aterian Caucasoid Proto-Hamite culture inhabited the region of Gafsa in Tunisia and extended their cultures influence (Capsian Man) as far south as Kenya - although some authorities express doubts regarding the precise chronology and extent of population in this era (Balout, 1981).
The cutting of wood for fires, the increased burden placed upon the land by grazing herds and the localised changes in the flora and climate this brought about is what made the barren Sahara we now know. These changes are advanced throughout the entire Mediterranean region (Le Houerou, 1980).
Details apart, we can be certain that as little as 5,000 years ago the Sahara was very different from today and supported a much wider and more typically southern African fauna; "the Sahara was an area with rich populations of wild animals and thriving life" (Rzoska, 1984).
A more detailed account of the creation of the Sahara, and the destruction of forest and wildlife this entailed is presented by Hugot (1974). What is clear is that the diversity of animal life present was truly remarkable, and that most of it was what we would today associate more with southern Africa rather than northern Africa; Hippo, elephant, gazelle, ostrich, lion, panther, giraffe, antelope, leopard and a host of other game. Some of these creatures were indeed only annihilated from north Africa within the last 100 years (the last lion and panther were shot in Tunisia at Babouch, near to Hammamm Bourgiba) and it was not until 1983 that the last Atlas Leopard was presumed shot in Morocco; fortunately, there remains a faint possibility that at least one small group may still survive (Haddane, 1989). Scullard (1974) points out that forest elephants grazed at the foot of the Atlas mountains as recently as 480 B.C. and occurred from the Mediterranean to the Cape of Good Hope in south Africa.
If this biotype was suitable for such creatures, why not tortoises?. And were they Geochelone-Homopus like or Testudo-like?. The only certain thing is that their bones now lie buried under the desert sands which destroyed and claimed their habitat. Many species which previously inhabited the central Sahara have become extinct since the Pleistocene. But are some of the n. African coastal tortoises of today their descendants?.
Tortoises are of course widespread in southern Africa, with a rich diversity of forms (e.g. Geochelone, Kinixys, Homopus, Psammobates, Chersina and Malacochersus). Some of which occur sympatrically. In north Africa only two forms, Testudo graeca (and that as an alleged subspecies), and a more distantly related form, T. kleinmanni are said to exist ; a somewhat anomalous and remarkable state of affairs given the previously alluded to faunal richness of the region and its undoubted ability to support such creatures.
If one considers the now extinct general zoological fauna of the region, and the affinity that expressed with southern African wildlife, and takes into account the ideal tortoise habitats which (not so long ago) existed in the Maghreb (and still exist today in many places), the alleged paucity of n. African tortoise genera and species is all the more remarkable.
One simple but surprising answer could be that these n. African populations have never been systematically studied by experienced specialist taxonomists in modern times. Their range of diversity has been overlooked and ignored. Grotesque mistakes have been made such as classifying all small tortoises without exception as 'juveniles', and all large specimens as 'old' or, in the last century, as species which do not actually occur in the region (e.g, writing of 'Testudo graeca', actually T. whitei, Boulenger writing in 1891 commented; "Old specimens have been taken for the allied T. marginata Shoepff, s. (syn) campanulata STRAUCH (by Gervais and Lallemant), the habit of which appears to be restricted to Greece").

Conclusions

The hypothesis that there is only Testudo graeca graeca L. 1758 in n. Africa, and that Testudo ibera PALLAS 1814 and Testudo zarudnyi NIKOLSKI 1898 are merely sub-species which are closely related to it is quite simply, on this evidence, no longer tenable. By direct implication, any conservation proposals based upon such a hypothesis are now revealed as hopelessly inadequate. The conservation situation is far worse than anyone has previously suspected, with many unique forms of limited distribution present rather than one widespread homogenous form; the problem of preservation is acute such are the threats facing all land tortoises in this present age.
The damage done to n. Africas tortoise fauna by the commercial pet trade was already acknowledged as extremely serious; this new evidence of a multiplicity of species groups amplifies the probable effects of that trade to nothing less than catastrophic. One widely distributed species obviously has a much better biomass recovery potential than numerous isolated individual species of limited zoogeographic distribution.

There are, to summarise, several possible answers to the problem of why such a diverse and rich tortoise fauna (as indicated by the specimens described here) should exist in the region;

They may represent isolated relict groups derived from the once extensive, but now extinct, chelonian fauna of what is now barren desert but was once semi-tropical forest which covered the region.

They may represent comparatively recently evolved forms derived from genetic stock left behind as the forests retreated, only the more suitable forms adapting and surviving by natural selection. The main objection to this however would be the time-scale involved which is by currently accepted standards far too short to permit such a process and a further objection is raised by the well established suprapygal pattern they demonstrate which clearly links them with Homopus, Geochelone and Hesperotestudo if not Hadrianus.

They may have arisen by invasion from some other direction, e.g Europe or from Asia via the Middle East, but this would require an incredible feat of convergent evolution to produce the suprapygal patterns discovered which so closely parallel those of several extant s. African forms (Geochelone and Homopus) and which are unknown in extant European stock.

Acknowledgements