In 1928 Eigenmann proposed the genus Brachygalaxias for the Chilean species bullocki which Regan (1908) described as having only five rays in the ventral fins and the origin of the anal in advance of that of the dorsal. Eigenmann defined his genus as having 5–6 ventral rays, and although he gave no precise description of bullocki it would appear that this specification is applicable to this species as the genus contains no other. It has been shown elsewhere (Stokell, 1940) that in the New Zealand species prognathus the number of ventral rays ranges from 6 to 8, and that one fin of an individual may have a different number of rays from the other. The same range and a similar individual inconstancy have been noted in attenuatus, while most of the other New Zealand species have proved variable to some extent. Particulars of the species investigated are given in Table I, and micrographs of 6 and 8 rayed ventral fins from G. attenuatus are shown in Figs. 1 and 2 respectively.

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

Not only is the character useless as a generic distinction, but it is unsafe as a specific one unless the degree and frequency of variation are taken into account. While some species appear to be equally variable in all waters investigated the occurrence of variation in others appears to be a racial character associated with definite localities. The species of questionable identity included in the table occurs in Canterbury streams at altitudes exceeding 1000 feet, in most of which it appears to be constant in the possession of 7 ventral rays, but a group of nine specimens from the Selwyn and its tributary the Hawkins contains three with 8–8. The variant with 6–6 is the only specimen available from a small stream connecting Lake Grass-mere with Lake Sarah.

It is to be noted that Eigenmann has recorded a greater number (8–9) of ventral rays in maculatus from Chile than has been observed in any New Zealand species, and that this is the only instance of specific variation recorded outside the latter country. It seems probable that variation occurs also in Australian species, as different workers record different numbers in some of them. Ogilby (1896) recorded 9 in findlayi from New South Wales, and Ramsay and Ogilby (1886) described kayi from South Australia as having 8, but Regan (1905), who regarded the latter as a synonym of olidus, described both species as seven rayed.

The second definitive character of Eigenmann's genus is the anterior origin of the anal fin, this being inserted in advance of the dorsal. The relative points of origin of the dorsal and anal fins are so variable in the family Galaxiidae, and the intergradation between extreme forms is so complete that the character cannot be invested with any generic significance. The Victorian species pusillus is described as having the anal origin in advance of that of the dorsal, though less so than in bullocki, attenatus has the points of origin about opposite, fasciatus has the anal origin below the 1st–3rd dorsal ray, while in the majority of the small New Zealand species it is below the 4th–6th. A more posterior origin of the anal is recorded in some Australian species, G. findlayi being described by McCulloch (1927) as having the anal origin behind the middle of the dorsal, while in fuscus Mack (1936) it is said to be under the posterior quarter. It is to be noted that the only forms described as having the anal origin anterior to that of the dorsal are bullocki and pusillus, both of which are also described as having five ventral rays, but the association of these features does not persist in burrowsins? which, though usually possessing five rayed ventrals, has the anal origin below the 2nd–4th dorsal ray.

The third character upon which Eigenmann defined his genus is the anterior position of the anus, this being recorded as “at the beginning, or behind the beginning of the last, third of the length.” In the accompanying figure of bullocki the anus is at .66 of the standard length, while in the New Zealand species paucispondylus (Stokell, 1938) it is at .67–.68, between which and the posterior extreme (attenuates .74–.78) there exists a full series of intermediates.

A character which appears to be of more importance is the comparatively high position of the pectoral fin, which is shown in Eigenmann's figure of Brachygalaxias bullocki, but is not commented upon by him. It seems possible that an examination of the osteological characters of this fish would reveal a greater distinction between it and typical Galaxias than is manifest in its external ones.

In 1935 Scott published a revision of the family Galaxiidae, certain sections of which appear to have been based on the published descriptions of forms that had not come under his own examination. Scott's proposals include the division of the family into two subfamilies, Galaxiinae, to contain Galaxias, Neochanna, and one new genus Saxilaga, and Paragalaxiinae to accommodate the genus Paragalaxias and a second genus to be established for G. dissimilis Regan (1905). Paragalaxias was created by Scott (1934) for a fish from the Shannon River, Tasmania, which is described as having the dorsal fin much further forward than in Galaxias, biserial teeth in the jaws, and a low number of vertebrae (44). This fish is unknown to me, but so far as may be judged from the published descriptions there seems no justification for its inclusion in this family. In typical Galaxiidae the posteriorly placed dorsal is so important a character that the inclusion of a fish having the dorsal insertion approaching mid-way in the standard length would amount to a contradiction of the family definition. There appears to be a definite gap between the proposed sub-families, so far as this character

is concerned, the range of variation in New Zealand Galaxiids being less than the difference between them and Paragalaxias. The anterior extreme in the local species of Galaxias is represented by paucispondylus in which the dorsal is inserted at from .68 to .71 of the standard length, while attenuatus, with a dorsal insertion of .75–.80 represents the posterior extreme. In a small group of Neochanna the point of dorsal insertion was found to vary from .71 to .72 of the standard length. The observed range of variation in Galaxias and Neochanna is thus only .13 of the standard length, while the difference between the anterior extreme and Paragalaxias, so far as may be judged from the published figure of the latter, is .14.

A similar difference appears to exist in the number of vertebrae of the two groups, but its precise extent cannot be determined in the absence of particulars of the variation in Paragalaxias. It seems unlikely that such a fish would be constant in this character, as variation occurs in all abdominal fishes with which I am acquainted, the usual range in species of Galaxias being 3, and the recorded minimum and maximum for the genus being 51 and 65 respectively. On the assumption that Paragalaxias is variable to the extent usual in Galaxias there would still be a considerable gap between the vertebral counts of the two genera. A variation from about 44 to 65 could not, in itself, be regarded as excessive within a family, as a range almost as great (45–63) occurs -within a single genus of smelts (Retropinna). but an important difference in the latter group is that there is complete intergradation throughout the range. It will be seen from the following table that the species of Galaxias having the highest and lowest number of vertebrae have the most posterior and anterior dorsals respectively, and although there is not complete co-relation of these characters in intermediate forms, there is a definite tendency in this direction. The data presented below have been obtained from 10 or more specimens of each species excepting alepidotus, of which only three have been examined.

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

The third definitive character of Paragalaxias, namely, the biserial teeth in the jaws, could scarcely be regarded as a disqualification for admission to Galaxiidae, although the two groups are sharply differentiated by it. Notwithstanding the great variability of members of this family in many characters, including the entopterygoidal and lingual dentition, the New Zealand forms have been found to be absolutely constant in the possession of uniserial teeth in the jaws.

While these considerations appear sufficient to justify the exclusion of Paragalaxias from Galaxiidae the family to which Scott's genus should be referred is not apparent, and it would be unprofitable at the present stage to make any suggestions in this direction. It is to be noted that the whole of Scott's specimens are small, the largest having a total length of only 40.7 mm., and that the circumstances under which they were collected suggest the possibility of their being juvenile examples taken while performing some kind of migration. In the absence of any record of sexual development or of an adult state being reached the suggestion that Paragalaxias shannonensis as described is the young of a larger fish seems sufficiently tenable to necessitate a re-description based on the examination of definitely adult specimens.

The nominal species Galaxias dissimilis Regan, which Scott included in the sub-family Paragalaxiinae, has been omitted from the above considerations for the reason that the single specimen upon which it is based appears to be a deformed one. It is significant that the principal definitive character of this fish—namely, the forward insertion of the dorsal fin (at .54 of the standard length) is associated with a longer head (head in length ratio 3.5) than is recorded in any authenticated species of Galaxias. The nearest approach to this ratio that I have encountered is 3.6, which occurred in a fish having abnormally rearward ventral fins, these being inserted equidistant, between the edge of the opercle and the hypural joint. Externally this fish revealed no sign of deformity, but dissection showed that about 12 of the posterior vertebrae were coalesced. It is probable that a similar condition exists about medially in the type of G. dissimilis, and until the vertebral column of this fish is examined it does not seem worth while to give further consideration to a species which, despite the taxonomic importance that has been attached to it, has never been re-collected. It cannot be too strongly emphasised that a single specimen is not sufficient to found a species upon, even when no contradictory character such as that exhibited by the type of G. dissimilis is present. Skeletal deformity occurs occasionally in most abdominal fishes, and, when the abnormality is congenital, may be associated with modifications of non-osteological characters such as the number of scales in Salmonoids. Judged on external characters such individuals may present the appearance of new and distinctive species, as they frequently possess symmetry of form and appear to have been little incommoded by their abnormality. There is also the possibility that a single specimen may represent one extreme of a range of variation, which, in so variable a family as Galaxiidae, could render the original description highly misleading.

The sub-family Galaxiinae as proposed by Scott contains the restricted genus. Galaxias, Neochanna, and a new genus, Saxilaga, which is distinguished from Galaxias by the absence of entopterygoidal teeth, and from Neochanna by the presence of ventral fins. This new genus is divided into two sub-genera, Saxilaga having six rays in the ventral fins and lingual teeth biserial, and Lixagasa having five rayed ventrals and more than two rows of teeth on the tongue. Saxilaga contains two Tasmanian species based on three specimens, and Lixagasa was provided for the accommodation of the New Zealand species G. burrowsius,

of which there appears to have been only two museum specimens in existence. I have been unable to collect this fish in the original locality (West Oxford), but have obtained specimens of a form which I have elsewhere (Stokell, 1938) referred to Phillipps' species from a small, muddy creek on Mr. J. Merrin's farm on the south bank of the Ashburton River, between Tinwald and the sea. This fish has extremely short ventral fins, the number of rays in which is variable, but usually five. Particulars of a group of ten specimens are given in Table I. The disposition of the teeth on the tongue has also been found to be inconstant, the arrangement varying from a double row typical of Galaxias to four rows as specified for Lixagasa. It is usual for two rows to persist in intermediate forms and for additional teeth to be scattered irregularly on the edge of the tongue. It is therefore necessary to suppress the sub-genus Lixagasa proposed by Scott for Galaxias burrowsius Phillipps, 1926. The entopterygoids are usually toothed, but the teeth are sub-normal and inconstant in number, and of such weak development that they are not readily detected without dissection. Particulars of a group of ten specimens are: 2–3, 1–3, 2–2, 2–3, 1–3, 3–3, 1–0. 1–2, 2–2, 1–1. It is quite possible that this fish is not specifically identical with the specimens of burrowsius obtained by Phillipps from Oxford, but even if the Oxford form should be re-collected and should prove constant in the possession of toothless entopterygoids its claim to generic separation from Galaxias would be invalidated by the intermediate arrangement occurring in the Tinwald fish. A species that is somewhat similar in this character was described by Phillipps (1940) under the name of koaro. Though never entirely without entopterygoidal teeth, this fish is inconstant in their number and degree of development. Individuals may have one tooth on each side, a row of five on each side, or any intermediate arrangement. In attenuatus the number varies from five to nine, and in fasciatus from six to ten. The observed range of 0–10 is therefore covered by the four species mentioned.

Abnormality of another kind has been recorded in prognathus, in which the entopterygoidal teeth are sometimes disposed in an irregular band or batch. Transverse pairs of teeth have been observed in both fasciatus and attenuatus.

The occurrence of four somewhat irregular rows of teeth on the tongue has been noted in a specimen which differs greatly from burrowsius and may belong to a new species or may be an aberrant example of a known one. This fish, which has eight rays in the ventral fins, was taken from the Cass River, near the West Coast road, but subsequent collecting in the same locality has failed to produce a similar example. Quite apart from the question of specific identity, the association of quadriserial teeth on the tongue with the highest observed number of ventral rays strongly discredits Scott's suggestion of sub-generic status for similar lingual dentition associated with the lowest observed number of ventral rays.

The remainder of Scott's taxonomic proposals consists of the division of the restricted genus Galaxias into sub-genera based solely on the number of rays in the ventral fins, Galaxias being proposed for the seven rayed forms composing the greater part of the genus,

Agalaxis for the six-rayed species of South Africa, while G. bullocki is temporarily regarded as occupying an unnamed sub-genus having five ventral rays. The variability of the number of ventral rays has been demonstrated in the early pages of the present paper (see Table I), and the reasons there advanced against the use of this character as a generic distinction are equally valid when sub-generic status is involved. Under Scott's arrangement some specimens of so well-known a species as attenuatus would require to be placed in Galaxias and some in Agalaxis, while individuals possessing eight ventral rays would be inadmissible to either. When paired structures that are variable are employed as a basis for generic or sub-generic separation there is also created the anomalous possibility of the two sides of an individual animal being referable to different genera or sub-genera. It is, therefore, necesary to suppress the sub-genus Galaxias proposed by Scott (1935, p. 88). These considerations also invalidate the subgenus Agalaxis as at present defined, but as it is possible that the South African species for which it was proposed may possess some hitherto unrecorded character of generic importance any action at the present stage would be undesirable.

In a paper which preceded Scott's, but which can be discussed here most conveniently, Whitley (1935) gave recognition to the genera Austrocobitis Ogilby and Brachygalaxias Eigenmann, and also proposed three new ones, Querigalaxias for G. dissimilis Regan, Lyra-galaxias for G. oconnori Ogilby (1912), and Nesogalaxias for G. neocaledonicus Weber and de Beaufort (1913). No definition of any of these genera is provided but the type species of each is figured. Ogilby's and Eigenmann's proposed genera have been discussed above together with the single specimen upon which the questionable species dissimilis is based. The value of the relative position of the anal and dorsal fins, which appears to be the character employed in separating Lyragalaxias from Galaxias, has also been discussed in the earlier pages of this paper. G. oconnori, as figured by Whitley (1933), has the origin of the anal below or slightly to the rear of the dorsal termination, and, so far as may be judged from the original description (Ogilby, 1912), in which, however, the anal is said to originate under the posterior third of the dorsal, this is the only unusual feature of the fish. Such an arrangement merely represents one extreme of a range of variation, and extreme forms that are connected by a full series of intermediates are no more entitled to generic distinction than are the intermediates connecting them. G. oconnori was described by Ogilby from a single specimen, and Whitley gave no indication of further material being available either when he published the first figure of the species or when he proposed its elevation to generic status.

The original description of G. neocaledonicus is not available to me, but the only distinctive feature revealed by Whitley's figure of this species is the length of the head, which is contained only three times in the standard length. As already pointed out, an abnormally long head in individual specimens is usually owing to shortening of the body brought about by coalescence of vertebrae, and in the absence of any knowledge of whether neocaledonicus is

based on a group or a single specimen the possibility of deformity must be recognised. The figure presented by Whitley obviously represents an emaciated fish, or one that has been shrunk by spirit, both of which alternatives are sxiggestive of a scarcity of specimens.

Of the three genera proposed by Whitley, one is and another appears to be provided for species represented by single specimens while particulars of the circumstances attending the creation of the third are not available. The practice of basing species on single specimens has been commented on above, and it need merely be added here that the creation of genera solely for the accommodation of such species is even more to be deprecated. Scott omitted Whitley's three genera from his own scheme of classification on the grounds that they were not yet validated by the provision of the necessary definition, and while Scott's example is followed here the grounds for omission are extended to include the utter inadequacy of an individual Galaxiid as the basis of a genus.

There remain to be considered several characters that hitherto have received little attention as possible generic distinctions. The point of insertion of the ventral fins is a character of some importance in differentiating species, but the intergradation of extreme forms is so complete that no separation of groups is possible. The observed range of variation of a number of New Zealand species is shown in Table III, together with particulars of two other characters which will be discussed below. It has been found that the point of ventral insertion becomes progressively more rearward with the increase in size of the fish, so much so that in certain species the range of a group of juvenile specimens may be wholly anterior to midway in the standard length while that of mature fish may be posterior. The following data have been obtained from ten or more adult specimens of each species excepting alepidotus, of which only three have been examined.

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

A character that is associated with the point of ventral insertion is the enlargement of lateral teeth in the jaws, the forms with anterior ventrals lacking canines, and those with posterior ventrals possessing them. Reference to Table III will show that in the three New Zealand species having the most forward ventrals canines are lacking, while in the three representing the opposite extreme they are present, although the form having the most rearward ventrals has not the

greatest enlargement. In this latter character the New Zealand species fall into more distinct groups than they do in the point of ventral insertion, but the occurrence of truly intermediate forms among the Australian species renders it valueless for generic separation. Micrographs of the mandibular teeth of G. attenuates and G. fasciatus, representing the opposite extremes in the New Zealand species, are shown in Figs. 3 and 4 respectively.

The degree of development of the pyloric caeca is usually associated with the two characters already discussed but, as will be seen from Table III, one of the species dealt with is definitely contradictory. In paucispondylus and attenuatus caeca may be very poorly developed or entirely lacking, while so far as prognathns has been investigated these structures are invariably absent. The highest observed development occurs in alepidotus, but in fasciatus, which does not differ greatly in point of ventral insertion and enlargement of canines, caeca may be lacking or present only in the most rudimentary form. Photographs of the two forms are shown in Figs. 7 and 8.

It may be noted here that the number of pyloric caeca, in species showing definite development, is usually two. An exception to this rule is provided by a form which appears to be restricted to a small stream entering the Rakaia River from the north bank just below the Gorge Bridge. This fish appears to be constant in the possession of three well-developed caeca, but agrees in other characters with a widely-distributed upland species of questionable identity which has the normal number in all other waters investigated.

The present investigation of the family Galaxiidae has revealed no evidence of the existence of natural and distinct groups that could be generically differentiated excepting the long established ones of Galaxias and Neochanna. It may be suggested that even these are not distinct, and that they are connected by forms such as burrowsius?, which usually has only five ventral rays, but it must be remembered that this number merely represents the minimum in forms possessing ventral fins, and that whereas the minimum and maximum are connected by intermediate forms there is an absolute gap between burrowsius and Neochanna. If burrowsius is to be regarded as indicating the line of descent of Neochanna, from Galaxias it might be expected that a dominantly four-rayed or three-rayed form or a form lacking ventral fins but retaining vestiges of the pelvic bones would exist to indicate a further stage in the process of degeneration, but there is no evidence of the occurrence of a lower number of ventral rays in New Zealand than in countries from which Neochanna is absent. Regarded independently a low number of ventral rays does not appear to be a degenerate character, and a consideration of the group of species having the lowest number reveals a marked inconstancy in the association of this character with others that seem to be more definite indications of evolutionary status. It is also to be noted that this group includes the Chilean species bullocki, which appears to be the most highly evolved of all the Galaxiids. Widely differing forms occur also in the six-rayed group, the South African species resembling typical seven-rayed ones

very closely, while in anguilliformis (Scott, 1935, p. 106) the dorsal, anal and pectoral fin rays appear to be more modified than in Neochanna. Similar differences persist to some extent in the sevenrayed group, the Chilean species globiceps (Eigenmann, 1928) possessing apparent affinities with burrowsius of New Zealand.

The remaining distinctions of Neochanna indicated in the original definition are the absence of entopterygoidal teeth, the compression of the teeth in the jaws, the short, conical, and widely spaced gill rakers and the existence of only one pyloric appendage, the whole of which have been found to be vitiated by intergradation between species of this genus and forms possessing ventral fins which are herein referred to Galaxias. That the absence of entopterygoidal teeth is not peculiar to Neochanna is already known, and although incisor-like teeth in the jaws appear to be restricted to this genus, the present investigation has shown that their occurrence is not invariable within it. A specimen of Neochanna from Waihopo, North Auckland, which appears to belong to a new species, differs from typical examples of N. apoda from Palmerston North and Wairarapa in having conical teeth in the jaws precisely as in Galaxias. Micrographs of the teeth of the two forms are shown in Figs. 5 and 6. The North Auckland specimen is further distinguished by the low number of rays in the dorsal and anal fins (ii 10 and iii 11 respectively), and the proportion of simple to branched rays, which is similar to that occurring in Galaxias. In N. apoda the number of simple rays in the forward part of the anal fin varies from x to xiii. The body of the North Auckland specimen is more compressed than in apoda, the vertebrae number 57, and the colour, in spirits, is uniform slaty-grey on the back, becoming somewhat lighter below. The ventral fins, pelvic bones and entopterygoidal teeth are entirely lacking, and there is a single well developed pyloric caecum. The latter character cannot be regarded as of any generic value as two well-developed caeca have been observed in an otherwise typical specimen of N. apoda. Similarly the form and arrangement of the gill rakers provide no distinction between the two genera. In the North Auckland specimen of Neochanna these structures are as long and closely set as in G. attenuatus, while in G. prognathus they are as short as in N. apoda. They are somewhat longer though still rather rudimentary in G. burrowsius?, while of the species in which development is definite, attenuatus has the shortest and paucispondylus the longest.

The customary view that the absence of ventral fins in Neochanna, the compression of the teeth in the jaws and the modification of the gill rakers and fin rays have been brought about by processes of degeneration is considerably weakened by the present evidence that these processes were not simultaneous ones. It is by no means clear whether the secondary features of apoda were developed subsequent to the divergence of Galaxias and Neochanna or inherited from the ancestral form, but it is apparent from the different arrangement in the North Auckland specimen that the modification of the teeth, gill rakers and fin rays was independent of the process by which Neochanna lost its ventral fins.

Dorsal fin with 12 rays (all counted) anal 12, pectoral about 14, ventral 8. Rays in anal fin finely shredded as in alepidotus. Pectoral extending about 3 of the distance from its base to the ventral, ventral extending about 4 of the distance from its base to the anal, anal originating below the seventh dorsal ray. About 43 vertebrae to rear of anal fin, length of specimen to rear of anal fin about 6 inches.

The whole of these features and also of other structural details observable in the specimen are such as occur in various combinations in present-day species, the only difference between these and the fossil specimen being that the present combination has not been noted in existing forms. Measurement of a number of species has shown that the distance from tip of snout to rear of anal fin varies from μ7 to μ8 of the total length, which ratios applied to the fossil specimen give a total length of between 7½ inches and 8½ inches. So far as I am aware the only New Zealand forms attaining this length are alepidotus and fasciatus, from both of which the fossil is excluded by its short paired fins and posterior origin of the anal. In alepitodus the pectoral extends somewhat more than μ5 of the distance to the ventral, the ventral extends μ6–μ7 of the distance to the anal, and the anal originates below the 2nd–4th dorsal ray; in fasciatus, the pectoral is somewhat longer and the anal origin is slightly further forward. The paired fins of attenuatus agree satisfactorily with those of the fossil, but the anal has a much higher number of rays (17–20), its origin is opposite that of the dorsal and its branched rays are merely forked. In lynx the point of anal origin approaches agreement with the fossil, but the number of rays in this fin is higher (14–16) and the paired fins are much too long. G. burrowsius may be dismissed on account of the low number of rays in the ventral fins (usually 5). The species paucispondylus approaches agreement in the length of the paired fins, the point of anal origin and also in the number of vertebrae, which appears to have been low in the fossil specimen, but the branched rays of the anal are merely forked as in attenuatus. The latter difference occurs also in prognathus which, moreover, is one of the smallest of the New Zealand species, the maximum total length observed being less than 3½ inches. Several forms of questionable specific identity differ in much the same manner as lynx, and none of the species of other countries as described by Regan shows satisfactory agreement, although collectively they repeat the majority of the individual features of the fossil specimen.

So far as may be judged from this fossil, in which, however, several important characters are not observable, there existed in Pliocene times a form which was con-generic with many presentday Galaxiids and of equal evolutionary status. The material with which the fossil is associated suggests that the fish was fresh-water dwelling, and while this does not necessarily constitute important evidence against the allegation of marine origin of the Galaxiidae it certainly affords no support to this view. It may be inferred from the circumstances that, with the exception of attenuatus, the species of each country