by Jean H. HUBER (1)


(1) Muséum national d'histoire naturelle, Laboratoire d'Ichtyologie générale, 43, rue Cuvier, 75231 Paris Cedex 05, FRANCE.


ABSTRACT. - Interrelationships of 25 predefined groups, belonging to the genus Rivulus and its allies, are proposed after the phylogenetic analysis of 74 external characters using PAUP 3.1.1. software. The resulting consensus tree is congruent with previous, not all, analyses and with the biogeographical distribution. The taxa Pituna and Millerichthys are confirmed valid entities, distinctive from Rivulus, but their definite systematic position needs to be corroborated by further studies. The taxa Moema and Renova form a well supported monophyletic lineage with Trigonectes, and it is proposed they are downgraded to the subgenus level, under Trigonectes. Two new subgenera of Rivulus are described, Oditichthys and Laimosemion, which possess true distinctive characters within the genus.


RÉSUMÉ. - Actualisation de la phylogénie et de la systématique des Rivulines néotropicaux appartenant au genre Rivulus et à ses alliés (Rivulinae, Cyprinodontiformes).

            Une nouvelle classification phylogénétique du genre Rivulus et de ses alliés est proposée à partir de 25 groupes prédéfinis et 74 caractères externes, en utilisant le logiciel PAUP 3.1.1. L'arbre de consensus résultant est congruent avec les analyses précédentes, mais pas avec toutes, et avec la distribution biogéographique. Les taxa Pituna et Millerichthys sont des entités valides, distinctes de Rivulus, mais leur niveau systématique doit être précisé par de nouvelles études. Les genres Moema et Renova, qui forment un groupe monophylétique solide avec Trigonectes, sont proposés d'être placés comme sous-genres de celui-ci. Deux nouveaux sous-genres de Rivulus sont décrits, Oditichthys et Laimosemion, qui possèdent des caractères tout à fait distinctifs à l'intérieur du genre.


Key-words. - Rivulus et al., South America, Phylogeny, Taxinomy.


            The study of the phylogeny and the systematics of the Cyprinodonts is today undertaken, in general, from three approaches. Each approach provides data on a different aspect; each contributes iteratively to our knowledge:

1. The external phenotype and its environment. The external phenotype comprises: firstly the traditional live and preserved color patterns and morphomeristic data, enhanced by the recent techniques of computer modeling and the radiophotographies; secondly, the detailed micromorphological characters, such as teeth morphology and organisation, cephalic sensory systems and cephalic scalation patterns; and, thirdly, the environmental observations concerning ecology, food, sexual and non-sexual behavior and distribution. The external phenotype approach uses low cost equipment and simple know-how; it serves both to improve field work and to maximise the information gained in the field. It is more appropriate for lower levels of systematics, i.e. from species to genus. It is as yet unable to phylogenetically relate heteromorphic superspecies (i.e. species complexes), either sympatric or allopatric.

2. The internal phenotype. This makes use of the osteological and anatomical characters, enhanced by the recent clearing and staining techniques, applied to these small fish. The internal phenotype approach requires low cost equipment but highly skilled know-how. It is laboratory oriented and is more appropriate for upper levels, from the genus and above. It may help also to better describe superspecies, when common minor osteological characters can be disclosed. It is as yet unable to properly assign a polarity to a number of characters.

3. The genotype. This makes use of the nuclear and mitochondrial DNA-RNA molecular techniques, permitted by the P.C.R. (polymerase chain reaction) devices revolution. The genotype approach requires presently expensive equipment and consumables, a distinct know-how, but these barriers to entry will level down in the future. It is laboratory oriented and has great potential, since it may be used to tackle all levels of the systematics. It could be argued that the genotypic approach has the unique power to determine the ideal, exact phylogeny. This concept is nowadays obviously naive: molecular phylogenies do have their own artefacts that are intensively discussed in literature. It is so new (the first publication regarding Cyprinodonts is dated 1993) that results are as yet fragmentary. The study of the genome is limited to small portions of slowly mutating genetic material and to a few species and specimens, regardless of haplotype conditions. Complementing the data obtained from the external and internal phenotypes, molecular techniques should ultimately provide an identity card for each species that can directly be attached to the holotype, even old ones.

            All three approaches are technically independent. No researcher yet is using them all together; only one or two. Obviously, the development of each of these three approaches is highly dependent on the results of the other two; the main drawback for all is the lack of proper material for study and an insufficient exchange of standardized data among researchers.

            For all three approaches, the numerization and processing of data normally uses software, such as PAUP 3.1.1 (Swofford, 1993). This permits the presentation of useful, transparent, reproducible, exchangeable results after the clear-cut polarization of each character. Among resarchers, it has promoted an even more rigorous attitude for the analysis of characters, especially the more qualitative ones that had to be numerized. The computer processing of the data is not a just number-crunching for large amounts of complex data. It allows trees of assumed relationships to be produced for each of the three approaches.

            Without the computerization, none of the three approaches may give operational results in phylogeny reconstructions, i.e. no test could be performed among these data sets:

- the "external phenotype" approach is not possible. For example, Cyprinodonts show small, grading differences in the basic morphology within large groups of related species at the genus level, together with strong variations at the population level. Therefore, traditional keys look purely artificial, over-complex, useless in practice and are soon obsolete, unless they are limited to the scope of one superspecies. For all the above reasons, we did not propose such a key in our "Review of Rivulus" (1992: 84), but instead a series of tables with diagnostic characters of each species, known or assumed to live in a given country of the distribution.

- the "internal phenotype" approach is also thought inoperative or else it has to be limited to more general characters, with poor output. As for the external phenotype, Cyprinodonts are only distinguished by minor (grading) osteological detailed characters and many of them need to be disclosed before a separation of groups may be achieved (frequent homoplasies).

- the "genotype" molecular approach simply cannot be worked out: a single portion of a gene gives several hundreds of workable base-pairs at one time. Prior to the molecular techniques, studies were limited to the descriptive analysis of the caryotype (i.e., Scheel's pioneering works), with a limited capacity of phylogenetic groupings.


The purpose of this study is to:

- address the above uncertainties and issues, with the external approach only;

- re-evaluate our previous findings (Huber, 1992) on the genus Rivulus, using PAUP;

- produce trees and check if these trees are reasonably in line with the current knowledge and can match with expected results of the other approaches (Parenti, 1981; Costa, 1989 et subseq. for the "internal phenotype"; Murphy and Collier, 1996 et subseq. and Hrbek, in prep. for the "genotype").

Before running the data on computer, we:

- reviewed all of them, possibly with new material, especially for those "odd" and "isolated" species which could not be brought into a superspecies in 1992;

- updated the data with the newest information available and especially the new species described in between;

- extended the data to the species of genera related to Rivulus in an attempt to present a more global perspective;

- tabulated all data in a matrix to be acceptable by the software for the various groups, either genera, subgenera, combined or single superspecies.




Revision and updating of our previous grouping in Rivulus


            In 1992 (:74-84), we proposed to aggregate the 110 named taxa (including synonyms) into 18 superspecies, plus 6 "odd" species and 9 "isolated" species. The 18 superspecies still hold well, except one: molecular experiments have shown that the fuscolineatus superspecies is heterogeneous (Murphy and Collier, 1996); R. fuscolineatus and R. weberi are to be linked with the isthmensis superspecies (subgenus Cynodonichthys) and separated from the species flock glaucus-uroflammeus-siegfriedi which corresponds, with the relict hildebrandi, to a second northern invasion of central America by the elegans superspecies (subgenus Vomerivulus).

            Since fall 1992, 7 new species have been described, 4 in the isthmensis series and 3 "odd"; a reappraisal of the punctatus superspecies has been produced with several synonymies (Costa, 1995a); and new collections of R. christinae, R. urophthalmus s.l., R. obscurus, R. cladophorus, R. pacificus and R. limoncochae s.l. have allowed us to update and amend the detail of some groupings.

            In total, the current list of 95 valid or presumably valid Rivulus species, out of 117 named taxa, is given in table I, together with their major morphomeristical data and the assignment of each to a phylogenetic group. In a limited number of cases, the assignment is uncertain due to contradictory or to scarce results: this is labelled with a question mark.


Extension to the species of genera and subgenera, related to Rivulus to form 25 presumably monophyletic groups


The extension to the species belonging to the related genera and subgenera of Rivulus encompasses 7 named groups as defined by Costa (1990 et subseq.): Austrofundulus (2 valid or presumably valid species), Rachovia (resp. 4), Pterolebias (resp. 9), Neofundulus (resp. 5), Trigonectes (resp. 6), Moema (resp. 4) and Renova (resp. 1). They are presented in table I, similar to Rivulus. This amounts to 31 valid or presumably valid additional species, out of 37 in total (Huber, 1996). Among those 31 species, 16 have been studied, in all species groups, i.e. 52% of the total: they are listed in appendix 1. In total, the 25 phylogenetic groups encompass:

- 17 groups of Rivulus (sensu Huber, 1992). One group is a conglomerate of three superspecies, urophthalmus-limoncochae-micropus, because we were unable to definitely distinguish solid characters among them. Moreover, the ornatus and beniensis superspecies have been put together because they share many morphomeristic characters; however, no component of the beniensis is known to-date from live or even recently preserved material in good numbers. Therefore, one cannot discard the possibility that this superspecies may be put closer to the urophthalmus superspecies in the future, as a southwestern vicariant.

- 8 groups of species, in genera or subgenera related to Rivulus (sensu Huber, 1992). Each corresponds to a described genus or subgenus as listed above, except Pterolebias which is heterogeneous and has been split into two groups (Pter.1 and Pter.2), as suggested by Seegers (1987), and confirmed here by us.

Finally an outgroup, combining primitive characters of Rivulus et al., has been selected: the subgenus Cynolebias s.s. in Cynolebias, a group of large, annual, cannibalistic Aplocheilids from South America.


Analysis of the 74 characters


Up to 74 characters have been found relevant for the phylogenetic analysis of the external phenotype. They pertain to 5 types:

- Type 1: size, general body and fins shapes; dimorphism.

- Type 2: micromorphological characters concerning the sensory system, the scalation, the teeth, the fins, etc.

- Type 3: traditional morphomeristics.

- Type 4: color pattern characteristics; dichromatism.

- Type 5: ecology, behavior and food.

            No character known to have a potential diagnostic value has been discarded.  The list of the 74 characters, with their up to 5 states, is given in table II; state 0 is assumed to be primitive, states 1 to 4 being derived; no character is dependent on another, although several are associated in practice; e.g., high dorsal fin to anal fin deviation (D/A) and long predorsal distance from snout.

            The numerization of morphomeristic external characters which vary continuously has been undertaken already in many other groups of fishes, but not in Cyprinodonts. In fact, each character can be treated as a statistical variable, whatever it can be:

- continuous;

- pseudo-discontinuous (like the micromorphological characters which vary within a population, with a dominant state in a given group);

- discontinuous (like the color pattern characters which are fixed within a species, with a dominant state in a given group).

Indeed, as shown in Huber (1992), no micromorphological character or no environmental character can be considered as diagnostic for a species or a superspecies, i.e., it is not possible to identify with certainty one specimen from one known locality based on one of these characters alone. There is too much variation within a species and even within a population. However, characters have a statistically dominant state for a species, and similarly for a group of cryptic species (i.e. a superspecies); this state has a real value for building potentially phylogenetic trees. The difficulty is to have enough specimens in order to determine surely this dominant state and to fix the boundaries of the states for a continuous or a pseudo-discontinuous character.

            All character states have been treated as unordered when several states are disclosed, i.e. the lengths of 0-1, 0-2, 0-3, 0-4 are equal. This is reasonable for discontinuous characteristics, e.g., describing color patterns. It may be argued that it should not be so, for continuously varying characters; for example, 45 and above scales in LL series may be more primitive than between 35 and 45, which may be in turn more primitive than under 35. Actually, the evolution may not have been straight forward and/or jumps of states may have occurred. In addition, it is clear that for some characters, like the 4 frontal scalation types, it can only be hypothesized whether one state is more derived than another. However, in our sensitivity analysis, we have performed additional processing with all or partially ordered characters with the aim of appraising the consequences on the most parsimonious tree.

            Although our knowledge is not complete, all characters have been assigned values for each group, avoiding the "question marks" or the occurence of "no data"; assumptions have been made, knowing the similarities of character states in some sympatric phenotypes. This may lead, in the future, to some limited amendments to the matrix, along with the addition of more characters, especially when new live populations of unprospected in-between regions are made available. The advantage of a computerized analysis is that reprocessing the matrix with new or amended data is easily performed.

All characters are considered primitive (coded: 0) in the outgroup.

Hypothetically, a more primitive Cyprinodont has, relatively to its own group, the following characteristics:

- it is larger and deeper-bodied,

- has many more scales and fin rays,

- all fins are short and rounded,

- vertical fins are probably in more advanced (forward) position, close together and have a nearly equal number of rays,

- it dwells in annual biotopes,

- it exhibits a generalized pattern of a few and irregular markings (more light gold or blue, than red) with more melanism, and with sexual dichromatism,

- it is oviparous.

The group of southern species, presently assigned to Cynolebias s.s., fits this description.

            The matrix of 26 groups and 74 characters was processed using PAUP version 3.1.1 (Swofford, 1993) to produce the most parsimonious tree. This was done only through the heuristic search method, because the matrix was too big for a branch-and-bound search. The matrix is presented in table III.

The following analyses were undertaken:

- subminimal trees;

- for each tree, computations of classical homoplasy measurements as CI, RI and HI, the effect of excluding individual characters, the effect of weighting characters (by the RI index) and the effect of ordering characters;

- finally the strict consensus tree, bootstrap computation (Felsenstein, 1985), and minimal and maximal branch lengths according to optimizations of homoplasy.

Branch lengths in trees are proportional to the number of changes occurring along the branches and shown under ACCTRAN optimization. In the strict consensus tree, numbers above the branch lengths are bootstrap proportions obtained from 1000 replicates using PAUP. A scale is given to associate branch length and the number of changes.

Some phylogenetic alternatives were tested using the Mac Clade software (Maddison and Maddison, 1990) to compute the corresponding necessary extra-steps.





Two most parsimonious trees were found with the following characteristics: tree length= 271 steps; consistency index (CI)= 0.476; retention index (RI)= 0.601; homoplasy index (HI)= 0.524. Except for a single difference, both trees are identical: the relative position of Pituna is either basic to all Rivulus-like groups except Millerichthys, or is a sister group of the block "Austrofundulus-Rachovia". This uncertainty resulted in one unresolved tritomy in the strict consensus tree. The subminimal test confirmed the unequivocal nature of the disclosed configuration: for 272 steps, already 55 subminimal trees were computed, for 273 steps, over 100. The reweighting of all characters by the maximum value of the RI resolved the tritomy of the consensus tree with Pituna inserted between Millerichthys and the block "Austrofundulus-Rachovia". The strict consensus tree only is presented (Fig. 1).

            Changing all characters (or only the morphomeristic ones) from unordered to ordered produced 4 equal trees of 320 steps that were very similar to each other, except in the branch lengths and the internal organization of the block "Renova- Moema- Trigonectes". The comparison of the consensus trees obtained with all unordered or all ordered characters revealed that:

- their general structures were the same;

- the basic branches and taxa were the same;

- but the groupings of lineages differed in some cases.

            The bootstrap analysis of the unordered consensus tree gave an idea of the robustness of the various branches, in addition to the lengths of the branches themselves. For example, a bootstrap value of over 70% and an assigned branch length of over 8 steps indicated a high support from the data; values between 50% and 70%, and 4 to 8 steps indicated a good support; and values below these limits reflected uncertainties in the data.

            For Rivulus et al., using the external phenotype approach only, some nodes were highly supported:

- all Rivulus groups separated from Millerichthys, Pituna, Neofundulus, Renova, Moema, Trigonectes, Pterolebias, Austrofundulus, Rachovia, i.e., all strict annual species (76%);

- the groups Austrofundulus- Rachovia (68%) and Renova- Moema- Trigonectes (94%);

- the superspecies grouping of Riv_gea and Riv_pun (74%) and of Anablepsoides and Riv_ben_orn (68%).

            On the other hand, there were some poorly supported nodes like those of Pituna, Vomerivulus, Cynodonichthys and those immediately below Rivulus (see Fig. 1).

            The sensitivity analysis, made by excluding individual characters one by one, did not change significantly the number and shape of the minimal trees or the general framework of the consensus tree. The exclusion of characters with an individual value of RI equal to zero, i.e. numbers 4, 15, 41 and 66, were undertaken for a specific heuristic search: the two resulting trees (262 steps) were only different by a limited number of branch lengths and the consensus tree was identical to the tree resulting from the unordered analysis.




Understandably, the general configuration of the consensus tree fits with the basic morphology, but it provides with much more information.


1. Supporting evidence:

- no group in the available genera is scattered all through the tree, which means that the genera, and especially the speciose genus Rivulus sensu Costa, are homogeneous (but see further, the case of Pituna).

- the species Rivulus robustus that we considered as very isolated (Huber, 1992: 83) and that Costa (1995b) moved to a new monotypic genus, closer to Cynolebiatini than to Rivulini, is indeed placed in a very primitive position in the tree, close to the outgroup. If the biogeographical hypothesis of an ancient piece of land(s) between South America and Yucatan in Mexico is confirmed in the future, then Millerichthys would be part of a proto-invasion of Central America, along with the primitive Rivulus species of Cuba and Hispaniola and the opportunistic brackish Rivulus marmoratus. We conclude from the present study that Millerichthys deserves the full generic status. This external phenotype approach is unable, though, to place that taxon as a member of the Rivulini or the Cynolebiatini. This could be undertaken when live material of Millerichthys is at last available.

- Rachovia and Austrofundulus are strongly related, the latter being considered as a subgenus (or a synonym, according to Costa, 1990a), even if they are different by several characters. They are only separated by 14 steps in the tree. In light of this, Austrofundulus should be considered as a subgenus of Rachovia, but we refrain in formally bringing this proposal since Thomerson (pers. comm.) is working on that subject in depth. That block is the closest, with Millerichthys, to Cynolebias.

- Pituna is not closely related to Rivulus as disclosed by Costa (1991) based on osteological observations and contrary to our own findings with conventional methods (Huber, 1992); however, Costa placed Pituna closer to Pterolebias and the matter should be restudied, considering the present split of that genus into two sublines.

- the monophyly of the three taxa Renova, Moema, Trigonectes is well supported. Renova is the most basal, as it is mentioned in the recent description of that taxon. However, these three taxa are very close to each other and we parsimoniously propose that Renova and Moema be considered as subgenera of Trigonectes.

- the monotypic taxon Anablepsoides is phylogenetically related to Riv_orn_ben, as disclosed by Costa (1990b) on the basis of a common osteological character; the present analysis produced an unexpected value, since we were unable to find a relationship with conventional methods (Huber, 1992: 105).

- the strictly annual species are all placed in a more basal position in the tree than the non annual or semi-annual Rivulus species; this finding is in line with the molecular results obtained recently by Murphy and Collier (1997), but not with the osteological ones by Parenti (1981) or Costa (1990a).

- the close relationship of Vomerivulus and Cynodonichthys (Huber, 1992), two subgenera of Rivulus, the latter being more basal than the former, is in line with the two hypothetical waves of recent invasion of central America, through the isthmus of Panama (Murphy and Collier, 1996). Branching Cynodonichthys with Vomerivulus as sister groups would require only 2 extra-steps out of a total of 271 steps (<1%).


2. Novelties:

- for the first time, the relationship of superspecies of the genus Rivulus is proposed and three/four major lineages are hypothesized: the "ancient Caribbic" lineage with Rivulus s.s. and Riv_oce; the "slender gracile" lineage with three sublineages; the "standard striated" line with five sublineages (including Cynodonichthys and Vomerivulus); and possibly the "spade-shaped caudal" lineage. However, it must be stressed that the branch lengths separating these lineages are all short, except one (the "slender gracile" line), and that the bootstrap values are low (30% or below).

- for the first time, the Caribbean Rivulus are brought together, although they are separated strongly by the ecology, their behaviour (marmoratus is a unique self fertilizing hermaphrodite), and the scale counts. Besides, the most primitive of the two components of Riv_oce, namely caudomarginatus appears to be southern (from Santos to Rio) and not Caribbean.

- for the first time, a global systematic tree of Rivulus and its allies is proposed that fits the biogeographical knowledge: the basal lineage of strict annuals represented by Millerichthys, Austrofundulus, Rachovia and by one component of Pituna (stellifer) are dwelling in northern South America, in the Venezuelian Llanos, which are probably their center of origin. The derived lineage of strict annuals represented by Pterolebias, Trigonectes, Moema, Renova has a dedicated distribution in the fold belts of the Amazon Basin and, further within the Basin (Moema, only), in line with the pre-Andean western coast. Its basal components are also living in the Llanos, hence the link with the first group. The three/four lineages of Rivulus are endemic to homogeneous regions, like the "ancient Caribbic" lineage in the Caribbean islands from a Venezuelian base (Murphy & Collier, 1996), the Cynodonichthys/Vomerivulus from northern Colombia with a basal form of Vomerivulus in the mountains (R. magdalenae) and probably also eastern Colombia. Likewise, the other lineages, which are often sympatric, appear to have expanded from a base in the Guyanan shield: the rest of the "standard striated" lineage has expanded into the entire Amazonia and the coastal plain of eastern Brasil; the "slender gracile" lineage into the entire Amazonia, where it has successfully invaded the Brasilian interior plateau up to Paraguay, but not the coast; and the "spade-shaped caudal" into the entire Amazonia, but nowhere else.


3. Discordances:

- Neofundulus is placed as a primitive group relative to all Rivulus-like groups, including Pterolebias and Trigonectes. This finding is not corroborated by previous osteological studies which relate it to the Trigonectes lineage only. However, no molecular study of a component of Neofundulus is available yet. Neofundulus is largely distributed in the southern part of the South American plateau savannahs (caatingas) and the confirmation of the present study finding would mean that the genus suffered extinction in northwestern South America savannahs, after having succeeded in migrating through the western fold belts of the upper Amazon basin. Another option would be that Neofundulus is derived from Pituna southern components (compacta and poranga), but this is not in agreement with the osteological evidence.

- Some taxa are in a more unstable position than others, following the change from one state to another of characters that are ambiguous for them. This is especially the case of Pituna, which may be moved as a sister group to the block Austrofundulus-Rachovia; the case of the block Riv_rec/Riv_bre, which may be split and scattered within the "standard striated" line; and the case of Riv_fre, which may be moved as a sister group of Anablepsoides. As a result, we have not included Riv_fre in the new taxon described further on, and we have refrained from naming the "spade-shaped caudal" lineage in a separate subgenus pending further studies; branching the block Riv_rec/Riv_bre further down the tree, in a derived position from Vomerivulus and Cynodonichthys, would cost 5 extra-steps out of a total of 271 (2%).





Ideally, groups are best defined when the diagnosis matches the synapomorphies, i.e. defining a group with its uniquely derived character (or several of them). Such characters exhibit a state that occurred only once during the course of evolution. This applies to 7 character states (among 6 characters: number 6, 10, 11, 31, 49, 68). Autapomorphies apart, all other character states are homoplastic. Therefore, the diagnoses proposed here are mostly based on combinations of homoplastic character states (under "other characteristics"). Biogeographical data are also given.


- The strict annual groups.

The first key observation is the separation of the present strict annual groups: Austrofundulus, Rachovia, Millerichthys (?), Pterolebias, Trigonectes, Moema, Renova, Neofundulus, Pituna. These groups share: a longer head; a larger size, except Millerichthys; a dark band through eye, limited or not to the orbit; a strong sexual dimorphism in dorsal and anal fin shapes, except Millerichthys; the absence of supracaudal ocellus in female, except Renova and in some populations or individuals in Pituna, Rachovia and Millerichthys; and a frontal scalation of D or S-type. They tend also to be deeper bodied than non annuals. For breeding, they all dive in peat (Millerichthys is unknown).

 - Millerichthys+Austrofundulus+ Rachovia+Pituna+Neofundulus. A humped (not straight) dorsal profile of head in old specimens; a not elongated body shape.

- Millerichthys+Austrofundulus+Rachovia. Very low scale count in a lateral series (LL= 30-32), minor sexual dimorphism in body shape and fin position.

- Millerichthys. Autapomorphies: 2-3 golden bands on male mid anal fin (this is reduced to spots near base in Pituna, and irregularly overall, in Austrofundulus and Rachovia) and black spots on female caudal peduncle near the base of rays. Other characteristics are: a smaller size than all annual groups, similar to the small sized Rivulus groups, as is generally the case for relict Cyprinodonts; compact with a flat front shape; a rather longer head and higher depth of body than the groups of Rivulus; a rather complete lateral sensory system, like Austrofundulus, Rachovia, Pituna and Pterolebias; and a strong sexual body dichromatism, like Pituna. The preopercular sensory system is made of reduced buttons, like Pituna and Austrofundulus. It inhabits a very small range of swamps in coastal eastern Mexico.

 - Rachovia+Austrofundulus. Synapomorphy: a unique supraorbital sensory system of irregular buttons in a continuous groove, contrary to all other groups. Other characteristics are: a compact shape of body, with a typical lyre-shaped caudal fin in male and consequently a strong sexual dimorphism at anal; a shorter interorbitar than head depth, like Pituna and unlike other annuals; an almost complete lateral sensory system, like Millerichthys, Pituna and Pterolebias; and a straight black bar through the eye and beyond. Inhabits the coastal savannah swampy areas of atlantic Colombia, Venezuela and Guiana.

- Rachovia. Rather compact, not as deep as Austrofundulus. An extended ventral fin in dominant male of some species. A short caudal length at mid-level, like Austrofundulus, but less so. Frequently, a lower orange margin at male caudal (also seen in Trigonectes et al., independently acquired).

- Austrofundulus. Autapomorphies: a heavily dark pigmented perianal zone (also seen in Anablepsoides, but differently and independently acquired); a broad black vertical bar at the border of the male caudal fin (may also be seen in Rachovia, depending on species or populations). Other characteristics are: much deeper at anal level than at caudal level (actually, the deepest at anal level, with 30% of standard length); a very long head, over one third of standard length; a short caudal length at mid-level, like Rachovia, and a high dorsal fin ray count, like Neofundulus. The preopercular sensory system is made of reduced buttons, like Pituna and Millerichthys. It shares with Cynolebias s.s. the presumed primitive character of a spine at the upper posterior part of the opercle. There is a unique combination of primitive meristic characters (high transversal count, with high advanced dorsal and anal fins).

- Pituna. Autapomorphy: a strong post-opercular upper dark blotch in male somewhat like, but not identical to, Neofundulus (independently acquired; however, due to its unstable position, it cannot be ruled out that further studies disclose that it is a synapomorphy). Other characteristics are: compact shape of body, with rounded caudal and pointed dorsal and anal fins, without extension; a shorter interorbitar than head depth, like Austrofundulus and Rachovia but unlike other annuals; an almost complete lateral sensory system, like Austrofundulus, Rachovia, Millerichthys and Pterolebias. The preopercular sensory system is made of reduced buttons, like Austrofundulus and Millerichthys. A strong sexual body dichromatism exists, like Millerichthys. Golden to light blue spots are scattered on sides and near base of anal fin of male.  Inhabits two disjunct regions: the northern Amazonian belts in the Venezuelian Llanos and their southern counterparts in the central Brazilian plateau, both swampy savannahs.

- Neofundulus+Trigonectes+Moema+ Renova+Pterolebias. Unlike all other annual groups, this supergroup is showing a sensory (supraorbital and preopercular) system which is similar to Rivulus, with well defined neuromasts in discontinuous grooves. Shape of body is not compact, but more elongated and cylindrical than the first supergroup of annuals.

- Neofundulus+Trigonectes+Moema+ Renova. Most known species have red striations on sides that fuse to broader, less numerous lines (4 or 5, versus 7 to 10) near peduncle (a color pattern also seen in the limoncochae superspecies of Rivulus, independently acquired).

- Neofundulus. Autapomorphies: a strong post-opercular upper dark blotch in male somewhat like, but not identical to, Pituna (independently acquired); no preopercular marking; a broad medium golden band at anal and a circumcaudal golden submargin (crescent pattern), surrounded by dark, in male. Other characteristics are: cylindrical, but not elongated group, contrary to Trigonectes et al. and to the non annuals. Related to Trigonectes et al., by the separated base of the ventral fins. Short predorsal length and lower D/A ratio, the dorsal fin being even more advanced than the anal fin in some populations; a high dorsal fin ray count, like Austrofundulus; and a short, rounded caudal fin. Some populations are good jumpers (like Rivulus), but they are usually bottom dwellers; loose posture in life, like most Rivulus (Huber, 1992). Inhabits the swampy areas of the caatingas and the pampas of the upper and mid Paraguayan basin in central Brazil, Bolivia, Paraguay and Argentina.

- Trigonectes+Moema+Renova+ Pterolebias. Synapomorphy: marked sexual dimorphism in pectoral fin shape.

- Trigonectes+Moema+Renova. Synapomorphy: a pointed mouth, more pronounced in Trigonectes than in Renova. Other characteristics are: large size, elongated with a straight shape of body; straight posture in life, like Riv_har and Riv_rec, independently acquired; and the caudal fin shape in male is variable but always longer than in female. Related to Neofundulus, by the separated base of the ventral fins. Strong (or less strong in Trigonectes) red reticulation on preopercle of male; red spots near the base of the anal fin (discontinuous in Trigonectes and Moema, continuous in Renova); a lower orange margin at male caudal (also seen frequently in Rachovia, independently acquired); and no golden markings on sides of male.

- Trigonectes. Autapomorphy: very extended ventral fins in male (also in Pterol1_lon, independently acquired). Other characteristics: slightly flattened body at the anal level and dorsal insertion far behind that of anal (like Moema). Inhabits the central Brazilian plateau, southerly up to the Paraguayan and Argentinian Chaco.

- Moema. Autapomorphy: acuminate pectoral fins in male. Other characteristics are: slightly flattened body at the anal level and dorsal insertion far behind that of anal (like Trigonectes). Inhabits the Amazon Basin and its belts as a northern vicariant of Trigonectes.

- Renova. More cylindrical than flattened, with a supracaudal ocellus in female. Known from a single locality in Amazonian Orinoco, north of the distribution of Moema.

- Pterolebias. Autapomorphies: very long caudal fin in dominant male, up to 70-80% of S.L.; and dominantly 8 fin rays at ventrals. Other characteristics are: large (over 8 cm), elongated species, with 2 subgroups; the first is deeper at anal level and the second more cylindrical, like Rivulus; the first has a fan-like caudal fin, and the second a lyre-tail in male. A rather complete lateral sensory system, like Austrofundulus, Rachovia, Pituna and Millerichthys.

- Pterol1_lon. Autapomorphies: very extended ventral fins in male (also in Trigonectes, independently acquired); filamentous dorsal, anal and caudal fins in male. Other characteristics are: slightly compressed deeper shape of body, with a convex dorsal profile of head; sexual dimorphism in body shape; and non extended pectoral fins in male. Distribution: all over South America, except the pacific coast and eastern Brazil; the two subgroups replace each other allopatrically, but their distributions are poorly known and may be disjunct with Pterol2_per, in-between.

- Pterol2_per. Autapomorphy: elongated pectoral fins in male. Other characteristics are: elongated and cylindrical shape of body, with a flat dorsal profile of head; and short ventral fins in male. This superspecies is made of two distinctive color types, peruensis and wischmanni et al., with a distribution along the eastern Andean lowlands in upper Amazonia.


- To the contrary, the non annual or semi-annual groups tend to show the opposite state of characters than the strict present annuals. In the tree, they are nested in a more derived position. This differentiation is clear in the tree, although biologically speaking it is known to be relative. Groups like Riv_har and Riv_rec are closer to annuals, in terms of egg development.

- The "ancient Caribbic" lineage of Rivulus. A rather deeper shape than the standard Rivulus, with the body depth greater at ventral level than at anal fin level. Both groups share the presumably primitive character (also in Cynolebias s.s.) of the lateral line scales being irregularly organised in fluctuating lines; an upper longitudinal post-opercular dark blotch; the border of the caudal fin is black in male; and they are characterized by a similar number of rays at dorsal and anal fins (A-D< 2), with a low D/A deviation for R. s.s. (+5) and high for Riv_oce (+8).

- Rivulus s.s. Fewer scales in a lateral series (LL= 34-36). Inhabits two Caribbean islands, Cuba and Hispaniola (Dominican Republic).

- Riv_oce. Autapomorphy: a dark gold-ringed supracaudal ocellus in both sexes (not gold-ringed in Riv_rec). Other characteristics are: many scales in a lateral series (the highest: LL> 45) and a very high transversal scales count (TRAV.> 12), like Austrofundulus (independently acquired). Female and male intensively melanistic, with golden spots on sides. Both sexes have a supracaudal ocellus. Inhabits the mangroves in crab holes, between Florida and southeastern Brazil.

- The "slender gracile" lineage of Rivulus. Small (less than 5 cm) size and slender shape, rather pencil-like; elongated, oval caudal fin in male; a combination of derived meristics: anal fin counts, low (10<A<12), very low scale counts laterally (LL= 30-35) and transversally, and very few scales on caudal fin of dominant specimens. This group is characterized by diversified and derived color patterns on fin margins (also internally) and by diversified color patterns on sides (always lineated in the "standard striated" lineage). The first three groups (also Riv_bre) have a short predorsal length in Rivulus, the last two a very long one.

- Riv_fre. Autapomorphy: a dark permanent longitudinal median band in female (temporary in male). Other characteristics: blue iridescence on male sides; straight posture, unlike most Rivulus, in mid-waters. Inhabits the permanent rivulets of the Guianan shield.

- Riv_gea. See the description of the new subgenus, Laimosemion, below. Inhabits the permanent rivulets of the Guianan shield and the mid-Amazon basin, up to Manaus.

- Riv_pun. Only separated from Laimosemion s.s., by the backward position of the dorsal fin (higher predorsal length and higher D/A ratio). Inhabits the southern belts of the Amazon Basin and the central Brazilian plateau and its surroundings in Bolivia, Paraguay and Argentina.

- Riv_orn_ben+ Anablepsoides. Synapomorphy: prefers edges of large rivers and lakes, contrary to other Rivulus which live in creeks. Other characteristics are: very small dorsal fin, very low transversal scales count, and very long caudal fin in male (maximum of 40% of S.L.). Costa (1995) has disclosed a common osteological character between Anablepsoides and Rivulus ornatus (Riv_orn).

- Riv_orn_ben. Autapomorphy: a very large ocellus (nearly half of the caudal peduncle) in female. Other characteristics are: shares with Anablepsoides, the extreme rear insertion of the dorsal fin and the very low combination of fin ray counts; squarish golden spots in 4-5 lines near peduncle of male, which follow a series of 7-8, and then 3-4 red lineated series of spots. Inhabits preferably margins of lakes in the Amazon Basin and its northern, western and southern belts in Colombia, Peru, Bolivia, Brazil and Paraguay.

- Anablepsoides. This subgenus is defined by 10 autapomorphies: head, broader than deep at interorbitar; dorsal insertion behind the end of the anal fin; upward orientation of pectoral fins; no sexual dichromatism and dimorphism; female and male, melanistic on belly and sides; 4 to 6 black oblique bars on sides; a low permanent longitudinal black band on sides; strict surface dweller (unable to swim downwards); and feeds on crustaceans. Other characteristics are: shares with Riv_orn_ben, the extreme rear insertion of the dorsal fin, the very low combination of fin ray counts and, for Rivulus, the longest caudal fin length (35-40% of standard length). It is also characterized by its very low depth of body (15% of standard length at anal level). Inhabits the Amazon Basin.

- The "standard striated" lineage of Rivulus. A large supergroup of medium to large size and cylindrical shape, with a strong D/A deviation (only shared by the small Anablepsoides and Riv_orn_ben). The caudal is rounded in both sexes, except in Riv_har. This supergroup, contrary to the "slender gracile" lineage, is characterized by poorly diversified color patterns on sides (10 lines of continuous or discontinuous red spots, which may be fused into 4-5 wider lines, posteriorily) and fins (which are often ornated by dark and light margins and submargins, or the reverse, and generalized spots or streaks internally).

- Cynodonichthys. Autapomorphy: a reverse pattern on caudal fin of male, with a white margin and a dark submargin. Other characteristics: the supracaudal ocellus is always present (at least in juvenile) in female. Distribution: from West of the Colombian Cordillera to eastern Mexico, both on Atlantic and Pacific slopes.

- Vomerivulus. Autapomorphy: a white to red conspicuous vertical border at caudal fin of male. Other characteristic is that the supracaudal ocellus is extremely rare in female. Distribution: similar to that of Cynodonichthys, but only up to Costa Rica; according to Murphy and Collier (1996), it corresponds to a second wave of invasion into central America, after the closing of the Panama isthmus.

- Riv_har. See the description of the new subgenus, Oditichthys, below.

- Riv_per+Riv_uro_lim_mic. Two groups of large size, with a cylindrical shape and a strongly loose posture in life. On the back, alternating dark and light blotches, similar but not identical to those of Riv_orn_ben+Anablepsoides and Riv_har. Distribution: the former, more primitive, is reclusive in highlands; the latter in lowlands of the Guyanan shield, East of the delta of the Orinoco in Venezuela to the whole Amazon basin in Brazil, Colombia, Peru; and also in the Caribbean islands of Martinique and Santa Lucia.

- Riv_per. Deep-bodied group, with high meristic counts, a shorter head, a shorter caudal fin and a more generalized pattern on sides.

- Riv_uro_lim_mic (Riv.u.l.m. in table 1). Autapomorphy: a dark margin and a light submargin in male unpaired fins. Other characteristics are: the supracaudal ocellus is conspicuous, surrounded by a golden zone and present only in females.

- Riv_san. Autapomorphy: golden vertically arranged spots on sides of males. Other characteristics are: the more slender group in the "standard striated" line, with an oval shaped caudal fin in male. Relict distribution in the coastal plain of eastern to southeastern Brazil (a southern vicariant of Riv_uro).

- The "spade-shaped caudal" lineage of Rivulus. Autapomorphy: a spade-shaped caudal fin in male. Other characteristics are: cylindrical and striated like the former supergroup, but with a rather straight posture in life; and a sexual dimorphism in dorsal, anal fins shape, unique in Rivulus, except in the species gransabanae.

- Riv_bre. Autapomorphies: dark temporary lower band in both sexes; and dark thinly fasciated sides and fins of male. Other characteristics are: small size like the "slender gracile" line. The dark band is not the same as that seen in Riv_fre; it may be homoplasic. Distribution: in coastal Guiana and surrounding highlands of Guiana and Venezuela.

- Riv_rec. Autapomorphies: extensions in ventral fins of both sexes, longer in male; a supracaudal ocellus in both sexes (but not surrounded by a golden zone, like in Riv_oce); and red reticulations on lower opercle of male. Other characteristics are: medium size; distinguished by its probable semi-annual condition; and preferring to spawn, like Riv_har, near the bottom. Distribution: presently known from the upper Amazon and Orinoco basins of Brazil, Peru and Venezuela, replacing allopatrically Riv_bre.





1- Oditichthys, a new subgenus of Rivulus with the most extreme capacities of aestivating and jumping.

- Type species: Rivulus igneus Huber, 1991. Note that the oldest taxon has not been designated intently. The species igneus combines the most primitive external characters in the group, fits the best with the diagnosis, and has been established to be semi-annual and a bottom spawner.

- Synonym: "Rivulus" Parenti, 1981 (in part).

- Derivatio nominis: from the Greek "Odites", meaning the wanderer, referring to its remarkable capacity for moving over wet land as already disclosed for Rivulus hartii in Trinidad, by Boulenger (1890) and Jordan (1923). Gender: masculine.

- Components: hartii (Boulenger, 1890); waimacui Eigenmann, 1909; holmiae Eigenmann, 1909; bondi Schultz, 1949 (a synonym of hartii); immaculatus Thomerson, Nico & Taphorn, 1991; igneus; ophiomimus Huber, 1992. Note that ophiomimus is enclosed in the group, as per 1992, because it fits with the diagnosis; however, it shows a colored preopercular plaque, like in the group of Riv_per. The other "striated" lineages are not included in the new taxon and especially the group of Riv_per; further studies are needed to position these groups in relation to the new taxon and to Cynodonichthys and Vomerivulus. Preliminary molecular results (Murphy and Collier, 1996) place Riv_per and Riv_uro_lim_mic in a monophyletic group, the only studied component of Oditichthys being unstable.

- Diagnosis: a member of the genus Rivulus as redescribed by Huber (1992). Oditichthys is best defined by two autapomorphies: 3 series of dark dots, longitudinally on posterior sides of females; and broad vertical dark bars on back and neighboring upper sides of both sexes. In addition, it exhibits a unique combination of primitive and derived characters: very large size (over 100 mm in T.L., the largest); straight posture in life (also seen in the small-sized group of Riv_fre), not loose like in most Rivulus (Huber, 1992); flat dorsal shape at frontal and predorsal levels; a longer head and longer caudal fin in male, unlike the similar group of Riv_per; a very high meristic counts in fin rays, together with the high deviation in the dorsal and anal fins insertion (D/A> 10); very high scale counts in lateral but not transversal series (unlike Riv_oce); 7 ventral rays, dominantly instead of the usual 6 rays (Parenti, 1981 and pers. obs.); a lack of pre- or post-opercular marking as in the other "striated" groups; by its semi-annual condition (7-10 weeks incubation); by its very large eggs (2.5 to 3.0 mm, twice the usual diameter); and its territorial behavior (Huber, 1992). Distribution: the Guyanan shield and Venezuela, east of Lake Maracaibo (including the off-shore islands), and the surrounding inland in the Amazon Basin.


2- Laimosemion, a new subgenus of Rivulus with striking similarities with the old world genus Aphyosemion.

Type species: Rivulus geayi Vaillant, 1899.

Derivatio nominis : from the Greek "Laimos" (throat) and "semion" (a sign or a banner), in reference to the common characteristics between this group and the African genus Aphyosemion s.l., including at the throat level the so-called "shield". Gender: neuter.

Components: geayi; strigatus Regan, 1912 (a synonym of geayi); dibaphus Myers, 1927 (a synonym of geayi); agilae Hoedeman, 1954; manaensis Hoedeman, 1961 (a synonym of agilae); probably, gransabanae Lasso, Taphorn & Thomerson, 1992 and cladophorus Huber, 1992; and nicoi Thomerson & Taphorn, 1992, which might also be a member of that group, based on morphomeristics.

The first two quoted valid species exhibit a common osteological character disclosed by Costa (1995) and also shared by Aphyosemion; bifid epipleural ribs. This is not reported in the group of Riv_pun that we would include, hesitantly, in Laimosemion. It is distinguished from Riv_gea by very few characters (see supra). A synapomorphy appears between the two lineages: the presence of red chevrons on male sides. However, this is somewhat ambiguous since these chevrons are coarser in Riv_gea and thinner anteriorily, then darker posteriorily in Riv_pun. Moreover, the female pattern is clearly subdued in Riv_gea and distinct and more melanistic in Riv_pun; our approach does not permit resolution of that uncertainty. The components of Riv_pun are: punctatus Boulenger, 1895; zygonectes Myers, 1927; apiamici Costa, 1989 (a synonym of pictus); decoratus Costa, 1989; pictus Costa, 1989; vittatus Costa, 1989 (a synonym of pictus); pinima Costa, 1989 (a synonym of pictus); modestus Costa, 1989; violaceus Costa, 1989; and the undescribed species from Cameta. The relationships of Laimosemion with the group of Riv_fre, which falls as a sister group in the tree, should be studied in-depth before the latter may also be included.

 - Diagnosis: a member of the genus Rivulus as redescribed by Huber (1992) Laimosemion is best defined by two autapomorphies: a preopercular pattern made of conspicuous red reticulations from below the eye to the opercle, like the "shield" in Aphyosemion s.l.; a dominant pattern of red chevrons on male sides (subdued in female). In addition, it exhibits a unique series of derived characters: small size (less than 50 mm T.L.); a gracile, elongated body shape with a loose posture; a dominant F-type frontal scalation; a diversified and derived color patterns of sides and internal fins, like in Aphyosemion; a faint and irregular supracaudal ocellus in female (also in western Aphyosemion); a dark temporary (not permanent like in Riv_fre), mood-driven, median band on sides of both sexes (like in some Aphyosemion); and also a mood-driven squarish post-opercular dark blotch (like some western Aphyosemion). Distribution: as given per the "slender gracile" lineage.





The present study is the first attempt to relate all groups of genera and subgenera allied to Rivulus at the same time. The processing of PAUP, after the numerization, of all data results in a consensus tree which agrees rather well with previous, independently published, morphological surveys; for the first time, a relationship between isomorphic or heteromorphic superspecies is proposed that previous methods could only at best separate.

However, even if the obtained consensus tree is stable after various sensitivity analyses, the bootstrap values of several branches are low and require further support. They need to be corroborated by the two other approaches, namely the internal phenotype and the genotype. This corroboration is likely since our results do not affect the basic tree obtained through the osteological studies. The first DNA studies also confirm the value of color patterns and morphomeristical characters for building superspecies and of biological conditions for phylogeny. Ultimately, the matrix will be useful for identification purposes in the field and new characters will help to strengthen the robustness of the tree.


Acknowledgements. - This study would not have been possible without the generous help of the Paris Museum staff, accustomed to PAUP specificities: Véronique Barriel over its first stage, Guillaume Lecointre for the initial spur and during the final and detailed processing and analysis. Both are warmly thanked.

Out of courtesy, the manuscript has been sent to the experts in the group who currently work on the two other basic angles: W.J.E.M. Costa (Rio de Janeiro) in osteology, G.C. Collier (Tulsa), J.E. Thomerson (Edwardsville) and T. Hrbek (Saint Louis) in molecular biology. Their encouragement and remarks have been valuable, together with those of the anonymous reviewers.





Boulenger G.A., 1890. - Description of two new Cyprinodontid species (Cyprinodon danfordii and Haplochilus hartii). Ann. Mag. Nat. Hist., (6)6: 169-170.

Costa W.J.E.M., 1990a. - Analise Filogenetica da Familia Rivulidae (Cyprinodontiformes, Aplocheiloidei); Classificaçao e Distribuçao da Familia Rivulidae. Rev. Brasil Biol., 50(1): 65-82; 83-89.

Costa W.J.E.M., 1990b. - Description d'une nouvelle espèce du genre Rivulus (Cyprinodontiformes, Rivulidae) de l'Amazonie Orientale. Rev. fr. Aquariol., 17(2): 41-44.

Costa W.J.E.M., 1995a. - Revision of the Rivulus punctatus species-complex (Cyprinodontiformes; Rivulidae). Ichthyol. Explor. Freshw., 6(3): 207-226.

Costa W.J.E.M., 1995b. - Pearl Killifishes. The Cynolebiatinae. Systematics and Biogeography of a neotropical Annual Fish Subfamily. : 128 p. T.F.H. Publ.

Huber J.H., 1992. - Review of Rivulus. Ecobiogeography - Relationships. 586 p. Soc. Fr. Ichtyologie Ed., Paris.

Huber J.H., 1995. - Nouvelles collections de Cyprinodontes paraguayens, avec description de 4 espèces Rivulines inédites et redécouverte d'une espèce à la localité typique jusqu'alors indéterminée. Killi Kontact (Ass. Kill. Franc. Belg.), Aug., 23(2): 24 p.

Huber J.H., 1996. - Killi-Data 1996. Updated checklist of taxonomic names, collecting localities and bibliographic references of oviparous Cyprinodont fishes (Atherinomorpha). Soc. fr. Ichtyologie, Ed. , Paris: 400 p. (in french, english, and german).

Jordan D.S., 1923. - Habits of the Trinidad Guapin, Rivulus harti (Boulenger). Copeia, 119: 69-70.

Murphy W.J. & G.C. Collier, 1996. - Phylogenetic relationships within the Aplocheiloid fish genus Rivulus (Cyprinodontiformes, Rivulidae): Implications for Caribbean and central American Biogeography. Mol. Biol. Evol., 13(5): 642-649.

Murphy W.J. & G.C. Collier, 1997. - A molecular phylogeny for aplocheiloid fishes (Atherinomorpha; Cyprinodontiformes): the role of vicariance and the origins of annualism. Mol. Biol. Evol., 14(8): 790-799.

Seegers L., 1987. - Die Gattung Pterolebias Garman, 1895, mit der Beschreibung von Pterolebias staecki. Aquar. Terr. Zeit. (D.A.T.Z.), 40(5): 199-204.

Swofford D.L., 1993. - PAUP- phylogenetic analysis using parsimony. Version 3.1.1. Illinois Natural History Survey, Champaign.



Appendix 1. List of material studied

(in addition to the Rivulus material listed in table 7 in Huber, 1992).

Cynolebias elongatus (type), NMW 76518, 1 specimen, La Plata, Argentina; Cynolebias elongatus, MNHN 1997- under registration, 1 spm., Pila, Argentina, aquarium maintained; Cynolebias monstrosus (quoted in Huber, 1995), Paraguay; Austrofundulus limnaeus, MNHN 1991-379, 4 spms, aquarium maintained; Neofundulus ornatipinnis (quoted in Huber, 1995), Paraguay; Pterolebias luelingi, MNHN 1988-1992, 35 spms, near Trinidad, Rio Mamoré, Bolivia; Pterolebias aff. longipinnis (quoted in Huber, 1995), Paraguay;Pterolebias peruensis, MNHN 1991-375, 6 spms, Padre Isla, Iquitos, Peru, aquarium maintained; Pterolebias zonatus, MNHN 1991-377, 5 spms, Venezuela, aquarium maintained. Pterolebias xiphophorus (types), FMNH 100597, 2 spms, near crossing of Rios Ventuari and Yureba, Venezuela; Rachovia brevis, unregistered, 1 spm., Venezuela, aquarium maintained; Rachovia maculipinnis, MNHN 1991-392, 17 spms, Venezuela, aquarium maintained; Rachovia pyropunctata, MNHN 1991-383, 6 spms, Venezuela, aquarium maintained; Renova oscari, MNHN 1997- under registration, 3 spms, Isla Raton, Venezuela, aquarium maintained; Rivulus birkhahni, MHNG under registration, 4 spms, Bocas del Toro, Panama; Rivulus compactus, MNHN 1997-46, 3 spms, central Brasil; Rivulus monticola, MNHN 1997-482, 2 spms, from the type locality, Ecuador; Rivulus nicoi (types), FMNH 100598, 2 spms, near the mouth of Rio Ventuari and Yureba, Venezuela; Rivulus tecminae (types), FMNH 100599, 4 spms, Rio Guayapo, Venezuela; Trigonectes aplocheiloides (quoted in Huber, 1995), Paraguay; Trigonectes balzanii (quoted in Huber, 1995), Paraguay; Trigonectes rogoaguae, MNHN 1988-1994, 10 spms, near Trinidad, Bolivia; Trigonectes (Moema) aff. staecki, UMMZ 178951, 3 spms, Karanambo Creek, Rupunini River, British Guiana; Trigonectes (Moema) piriana, NRM unregistered as "sp. NSC1", 1 spm., from the type locality, N.E. Brasil, aquarium maintained.


Fig. 1.- Strict consensus of two equiparsimonious trees of 271 steps (C.I.= 0.476, R.I.= 0.601) obtained using PAUP 3.1.1 and based on 74 externals characters. Numbers above branches are bootstrap proportions obtained from 1000 replicates.


Table I.- List of the presumably valid species, with their assignment to each superspecies, with the number of studied specimens and the following major morphomeristic data: Maximum length in mm, mean and standard deviation of dorsal fin rays, anal fin rays, lateral scales and predorsal length as a % of SL, mean only of the D/A deviation, of head length, of depth at anal level, and maximum caudal fin length, as a % of SL The index is computed by the formula "(D+A+D/A)*LL".


Table II.- List of the 74 characters and their up to 5 states, for Rivulins.

1- Max. size: large, over 8 cm=0; medium=1; small, under 5 cm= 2

2- General body shape: deep=0; compact=1; elongate=2; slender=3

3- Body shape at Anal level: depressed=0; cylindrical=1; slightly compressed=2

4- Body depths at Anal and at peduncle levels: marked difference=0; similar=1

5- Interorbital width: shorter than head depth=0; similar=1; greater than head depth=2

6- Mouth shape: rounded=0; pointed=1

7- Mouth opening: oblique upwards=0; oblique downwards=1

8- Front shape: oval, convex=0; flat=1

9- Predorsal shape: oval, convex=0; flat=1

10- Longitudinal dorsal profile of head: strongly curved or humped=0; linear=1

11- Frontal supraorbital sensory system: over 2 dozens of small buttons=0; 6 reduced buttons in a continuous groove=1; conspicuous neuromasts, isolated or in a discontinuous groove=2

12- Pre-opercular sensory system: none=0; reduced to irregular buttons in a groove=1; two exposed neuromasts, only=2

13- Lateral sensory system: present, with pores/buttons=0; absent or much reduced=1

14- Frontal scalation dominant pattern: unorganized smaller scales=0; no specific pattern (S-type)=1; D-type=2; E-type=3; F-type=4

15- Opercular shape: with a posterior upper spine=0; without=1

16- Head length in % S.L. on average: over 29%=0; between 29% and 24%=1; under 24%=2

17- Predorsal length in % S.L. on average: under 70%=0; between 70% and 75%=1; over 75%=2

18- Bases of Ventrals separated: no=0; yes=1

19- Average Dorsal fin count: over 11=0; between 11 and 7=1; under 7=2

20- Dorsal fin insertion well behind the level of the last Anal ray: no=0; yes=1

21- Average Anal fin count: over 15=0; between 15 and 12=1; between 12 and 10=2; under 10=3

22- Average D/A deviation: under 5=0; between 5 and 7=1; between 7 and 10=2; over 10=3

23- Average Ventral fin count: 6 rays=0; 7 rays=1; 8 rays=2

24- Orientation of Pectoral insertion: vertical (backwards)=0; oblique (upwards)=1

25- Average LL scales count: over 45=0; between 45 and 35=1; under 35=2

26- Scales near base of Dorsal and Anal fins: yes=0; no=1

27- Side scales irregularly positioned in lines along sides: yes=0; no=1

28- Maximum number of scales beyond hypural plate in old specimens: over 6=0; between 6 and 3=1; under 3=2

29- Average transversal scales count (TRAV): over 13=0; between 13 and 8=1; under 8=2

30- Sexual dimorphism in body shape and fin position: yes=0; no=1

31- Sexual dimorphism in Pectoral fins shape: no or weak=0; yes, marked=1

32- Sexual dimorphism in Ventral fins shape: no or weak=0; yes, marked=1

33- Sexual dimorphism in Dorsal/Anal fins shape: no or weak=0; yes, marked=1

34- Sexual dimorphism in Caudal fin: no or weak=0; yes, strong=1

35- Pectoral Fin in male: rounded=0; elongate=1; acuminate=2

36- Ventral fin extensions: no=0; yes, short, in male only=1; yes, long, in male only=2; similar in both sexes=3

37- Dorsal/Anal fin shape in male: rounded=0; acuminate=1; with streamers=2

38- Caudal fin shape in male: rounded, short=0; truncate posteriorily=1; spade-like=2; oval, elongated=3; fan-shaped=4

39- Caudal fin extension in male: none=0; lyre-shaped=1; with filaments, overall=2; with specifically located longer rays (central, lower)=3

40- Sexual dichromatism: striking (distinctive)=0; strongly subdued=1; weakly subdued=2; none=3

41- Female intensively melanistic: no=0; yes=1

42- Dark band through eye: angular=0; straight, beyond eye=1; straight, within eye=2; none=3

43- Pre-opercular lower marking: no=0; yes, similar to sides=1; yes, distinctive from sides=2; yes, dark and silvery blotches=3

44- Pre-opercular red upper oblique lines or reticulations: no=0; yes, vague or recessive=1; yes, conspicuous=2

45- Post-opercular marking (wound): no=0; yes, upper and vertically shaped=1; yes, upper and oval laterally=2; yes, median and rounded=3; yes, lower=4

46- Dark (temporary) longitudinal broad band on sides (mood dependent): none=0; median=1; low=2

47- Caudal ocellus presence: no=0; female, only/juvenile=1; adult, both sexes=2; irregular, depending on populations=3

48- Diameter of Caudal ocellus, half of peduncle: no=0; yes=1

49- Red chevrons on male sides: no=0; yes=1

50- Dark thinly fasciated pattern on male sides and fins: no=0; yes=1

51- Four to six black oblique bars on sides: no=0; yes=1

52- Melanism on belly of both sexes: no=0; yes=1

53- Golden markings on male sides: no=0; round/oval=1; squarish=2

54- Golden median band at male Anal: no=0; yes, one=1; yes, two or three=2

55- Golden series of large spots near basis of male Anal: no=0; yes=1

56- Series of longitudinal red spots on mid sides of males: no=0; yes, 4-5=1; yes, 7-10=2

57- Series of 3 longitudinal dark spots near peduncle of females: no=0; yes=1

58- Marginal and submarginal bands on male Caudal: no=0; dark, then light=1; reverse=2

59- Conspicuous red/orange broad band at lower Caudal fin of male: no=0; yes=1

60- Black spots on female peduncle, at the basis of Caudal rays: no=0; yes=1

61- Dark and light bars, alternatively, on back and upper sides (ladder pattern): no=0; yes=1

62- Vertical border on male Caudal (resp., crescent shaped): no=0; dark=1; light=2

63- Circumcaudal light median band on dark in male (also Anal): no=0; yes=1

64- Dark conspicuous markings near Male Anal base: no=0; discontinuous spots=1; line=2

65- Pectoral fins heavily fasciated or dotted in Male: no=0; yes=1

66- Perianal and genital papillae, heavily pigmented: no=0; yes=1

67- Dark lower lip (temporary or permanent): no=0; yes=1

68- Ecology: ephemeral=0; creeks=1; rivers/lakes=2; brackish/marine=3

69- Swimming: open water=0; near surface=1, at surface=2

70- General posture: standard=0; straight, pencil-like=1; loose (Rivulus-like)=2

71- Behavior: no jumper=0; jumper=1

72- Annualism: strict=0; intermediate=1; optional (aestivating)=2

73- Breeding: dives=0; over bottom=1; egg-strander=2

74- Food: fish=0; insects=1; aquatic preys=2


Table III.- Matrix of 26 groups (the outgroup and the 25 Rivulin groups) and 74 characters.



Additional reading of this author's publications ? Select any ONLINE article.