PlanetCatfish.com

After a great deal of introspection, it appears I am bi-polar. Are you a lumper or a splitter?

Jools

I would create/use more subspecies.

In all other animals we have thousands and thousands species that have subspecies. By fishes it looks that there aren't a lot of species with subspecies.
Or maybe I just don't know enough fishes, that's also a possibility.


So, am I a lumper or a splitter, or am I just something else?

There is a scientific difference between lumping and splitting at the generic level (i.e. the Panaque thing), and lumping and splitting at the species/population level.

We need to make the distinction here, otherwise the discussion could go awry.

Rupert,

That makes you a splitter.

For this thread, I mean at the species level then.

Jools

I'm not sure - sometimes I think "well, these two things aren't really the same, even if the paper says so" (if a paper said L128 and L200 is the same for example), and at other times I think, well, they look the same, and the distribution is close/overlapping, what makes them different species (Pterygoplichthys pardalis & P. disjunctivus comes to mind)

--
Mats

Bijn wrote:I would create/use more subspecies.

The subspecific taxon (under the biological species concept) for me, is fundamentally flawed as mechanism for practical hypothesis testing in systematics. It is not used by ichthyologists for good reason. The bird people have their own ideas, many of which are not congruent with modern methods.

I wrote this in another thread, and I think it holds true here. Many systematists are regarded as "lumpers", but this is simply due to the types of data they analyse. Even if you suspect several different species, but your data can't fully support your conclusions, you shouldn't publish new names. I do think systematists should branch out into different types of data, but it is hard and expensive, requiring lots of expensive observation in the field or equipment in the lab.

racoll wrote:Species boundaries are hypotheses subject to testing, and just because a scientist say for example describes three species of Xingu Hypancistrus based on morphometrics of preserved material, this doesn't mean that if the groups are revisited with ecological data, behavioural data, fine scale DNA data and live colour patterns etc that the boundaries would stay the same; I'm pretty sure they wouldn't. Aquarists see a lot of systematics as "lumpers", but this is largely based on the fact they only have so much information to work with (i.e. measuring a dead, preserved fish), and these data can only support so many hypotheses.

Well, for species, if a group of animals have any diagnostic characters, then it must be considered a species. If there are no such characters, it isn't.

That means that many "old" species described in the 18th and 19th century with huge distributions should really be split - burbot, for instance - and quite a few "new" species in especially europe and north america with tiny distributions - many of the Coregonus, for instance - should be lumped.

On the genus level I probably lean towards lumping because I'm lazy and don't want to learn new names, but above all I want monophyletic genera. Yes, I know, there's no such requirement in the code, but there should be - interrelationship is valuable information, subjectively perceived differentness isn't.

racoll wrote:The subspecific taxon (under the biological species concept) for me, is fundamentally flawed as mechanism for practical hypothesis testing in systematics.

If you go by the code, a subspecies is a geographically distinct population which has no diagnostic characters. If there is any overlap in distribution or any diagnosable characters, the subspecies must be recognized as species. (That doesn't mean that subspecies must be identical, you can have clinal or statistical variation)

It's a pretty safe bet that the vast majority of subspecies are invalid, and should either be synonymized or raised to full species.

MatsP wrote:Pterygoplichthys pardalis & P. disjunctivus comes to mind

It is pretty suspicious the way both species seem to be introduced together around the world. I wouldn't be terribly surprised if they're eventually found to be two phenotypes of one single species.

I just realised that last years Proceedings of the Royal Society B papers are now open access, and as I mentioned in another thread, there is a debate "raging" in the scientific community about how taxonomy and taxonomic characters are to be treated as genomic data, statistical methods and cryptic species become more central in modern biodiversity research.

This conceptual bifurcation has a direct bearing on the splitting vs. lumping argument, and arguably with fine-scale molecular data and these "objective" statistical coalescent models, we could see a lot more splitting in years to come?

So, here is the initial paper "describing" three Hemidactylus geckos by Leache & Fujita (2010) that sparked off the controversy. Here is the reply by Bauer et al. (2010), and here Fujita & Leache (2010) respond to the criticism.

There is also more fish-related background to the discussion presented here by Mooi & Randall (2010).

In recent years I started to ignore the species concept, and exhanged it for populations

As an example, Calichthys calichthys is said to come from almost the whole of south America. Tropical, subtropical - it's supposed to be the same species

BUT I would not put a fish from Colmbia in a Uruguay setting - with winters around 10-12 C. I would look for fish cought in the sub-tropics

This can be seen as an argument for splitting the species, but I do realize my arguments are purely aquaristic - and thus not scientific

Is L200 the same species as L128? Actually I don't care - I would not interbreed them, as they are different populations

Bas Pels wrote:As an example, Calichthys calichthys is said to come from almost the whole of south America. Tropical, subtropical - it's supposed to be the same species

This is a great case in point, and is exactly what Mike Noren mentioned earlier.

A quick glance over Catalog of Fishes reveals Callichthys callichthys has about 13 synonyms!

The fact these synonyms are regarded as single species simply reflects the lack of modern taxonomic attention. Look at the revision of Auchenoglanis; where there was two species a couple of months ago, now there are eight.

Almost certainly, many of the C. callichthys synonyms should be valid.

It just takes someone to do the work...

racoll wrote: Look at the revision of Auchenoglanis; where there was two species a couple of months ago, now there are eight.

That's exactly why I am a lumper. To my (non-scientific) eye far too many species are created because for example a dorsal spot sits one millimeter to the left in comparison to that of the holotype.

Marc van Arc wrote:to my (non-scientific) eye far too many species are created because a dorsal spot sits one millimeter to the left in comparison to that of the holotype.

I see what you are saying, but it does take a fairly substantial body of evidence to convince a peer review. Single characters are rarely used.

I see what you are saying, but it does take a fairly substantial body of evidence to convince a peer review.

To play a little deveil's advocate...I have not seen anyone argue that the score of Corydoras spp described in aquarium magazines over the last 20 years are not valid spp.

The Code does not mandate peer review, does not mandate what qualifies as a legitimate species description (to any meaningful extent), but does mandate that old, hard to find non-peer reviewed descriptions based on a single example bought in an aquarium store and published before the advent of modern tools and methods actually have priority!
-Shane

Indeed, and the worst (best?) one isn't even a Cory. Synodontis galinae anyone?

I think the more I know about something, the more likely I am to think split. Also, fish breeders IMHO are much more likely splitters.

Jools

Mike wrote:Well, for species, if a group of animals have any diagnostic characters, then it must be considered a species. If there are no such characters, it isn't.

I'm a splitter, subspecies feels 1800 and have no place in modern taxonomic.

Janne

Also a splitter (mostly).
Even if different populations/species look the same they can be very different in ecology/behaviour. This is more visible in birds where some species are very difficult to distinguish visually, but are obviously different species (behaviour/song). These different species often have a different ecology as well (and in case of our hobby, different care). Why should this be different with fishes?

Shane wrote:To play a little deveil's advocate...I have not seen anyone argue that the score of Corydoras spp described in aquarium magazines over the last 20 years are not valid spp.

I can not comment on Corydoras specifically, and I do not wish to cite examples, but more often than not the "aquarist" descriptions are of so low quality (no or wrong locality, no or wrong measurements, no preserved holotype...) that they make continued work on the group difficult.

The Code does not mandate peer review, does not mandate what qualifies as a legitimate species description (to any meaningful extent), but does mandate that old, hard to find non-peer reviewed descriptions based on a single example bought in an aquarium store and published before the advent of modern tools and methods actually have priority!

IMO the fact that the minimum requirements for what constitutes a species description are so very low is a much bigger problem than the principle of seniority.

racoll wrote:I just realised that last years Proceedings of the Royal Society B papers are now open access, and as I mentioned in another thread, there is a debate "raging" in the scientific community about how taxonomy and taxonomic characters are to be treated as genomic data, statistical methods and cryptic species become more central in modern biodiversity research.

Well, firstly my impression is that "true" cryptic species, species which really are morphologically indistinguishable from eachother, are rare. I do not believe I have ever seen one. More often purported "cryptic species" are those 18th and 19th century species with enormous distributions I talked about earlier, which, when studied more closely, turn out to be complexes of a number of distinctly (and some times radically!) different species.

The idea that species should be defined based on dissimilarity of genome comes from barcoding and metagenomics projects, which generate vast amounts of data with imperfect correlation to existing taxonomy. There's still very few examples of described purely "molecular" species, but the concept is clearly gathering momentum. While I have no problem with identifying species by DNA analysis, I think it's misguided to define species by DNA analysis, because the end result will inevitably be an enormous number of species which are subjectively defined, and which can not be identified except by DNA analysis.

Mike_Noren wrote:The idea that species should be defined based on dissimilarity of genome comes from barcoding and metagenomics projects, which generate vast amounts of data with imperfect correlation to existing taxonomy. There's still very few examples of described purely "molecular" species, but the concept is clearly gathering momentum. While I have no problem with identifying species by DNA analysis, I think it's misguided to define species by DNA analysis, because the end result will inevitably be an enormous number of species which are subjectively defined, and which can not be identified except by DNA analysis.

Bring on the "pocket DNA analyzer", that we can all carry around that put a little bit of fin/scale/hair into, and immediately get a "result" as to what/who it is ... ;)

But until such a time that exists, I agree, we can't have DNA as the ONLY differentiating factor for species. You should be able to tell by at least one other way (morphometrics or for example capture location).

--
Mats

Mike_Noren wrote:Well, firstly my impression is that "true" cryptic species, species which really are morphologically indistinguishable from eachother, are rare.

True. Can't disagree with this, although imagine many more will appear when people actually start deliberately looking for them.

Mike_Noren wrote:The idea that species should be defined based on dissimilarity of genome comes from barcoding and metagenomics projects

Not sure I can agree here though. After the initial hyperbole, I don't think anyone involved with DNA barcoding is making these statements. I mean, a recent study of Romainian butterflies co-authored by Paul Hebert (the architect of DNA barcoding) is fairly modest, making no such claims of defining species or replacing taxonomy:

Dinca et al. wrote:The main goals of this method [DNA barcoding] are (i) to ensure fast and reliable species identification and (ii) to aid the discovery of undescribed species. These goals complement many potential applications related to biodiversity conservation, pest management, forensics and healthcare.

Mike_Noren wrote:...which generate vast amounts of data with imperfect correlation to existing taxonomy.

Almost all of a huge number of studies report highly congruent results with currently regarded species, often with only a handful of paraphyletic or haplotype sharing species. Where it often fails, is when the existing taxonomy is poor, or species identifications are unreliable. There is also an assumption that the existing taxonomy is perfect, which it isn't.

MatsP wrote:But until such a time that exists, I agree, we can't have DNA as the ONLY differentiating factor for species.

Mike_Noren wrote:While I have no problem with identifying species by DNA analysis, I think it's misguided to define species by DNA analysis, because the end result will inevitably be an enormous number of species which are subjectively defined, and which can not be identified except by DNA analysis.

I have no problem delimiting (notice I don't say defining) species with just DNA. I understand it won't be very user friendly for aquarists, but I don't see why science should ignore the biological reality of diversity just because our human eyes and rulers can't find a difference. There is not much difference between saying this, and that all Rineloricaria for example should be one species because they are hard to tell apart without locality data, a preserved specimen, a microscope, a lot of literature, and many years taxonomic expertise. It would certainly be no harder with just DNA.

However, the issue is how this DNA data is presented. While the Bayesian methods Leache & Fujita (2010) used are a valuable contribution, they should be used as supporting evidence only. The diagnostic characteristics should remain based on apomorphic character states. This removes a lot of the subjectivity, as the tree building methods and therefore posterior probabilities are often highly sensitive to perturbation in initial parameters and priors. Bayesian methods are great, as you can incorporate a lot of the uncertainty into the analysis, but for me it is too experimental and subject to far too much practical and theoretical change to be used on its own. The concept of diagnosis and apomorphy has been around since the 60's and there is no reason why the two cannot be used in tandem on the same data.

Jools wrote:Indeed, and the worst (best?) one isn't even a Cory. Synodontis galinae anyone?

I think the more I know about something, the more likely I am to think split. Also, fish breeders IMHO are much more likely splitters.

Jools

Well breeders are collectors so they/we like to "collect" different species rather than varieties of the same. Having said that I think by nature I am more inclined to lump but I can see the utility in splitting for conservation etc.

Hi,

I am certainly more on the splitting side of life.
Though I currently try - partly because I am predominantly dealing with as-yet undescribed variants - to avoid the term species (or put it into quotation marks).

On the practical side of fish keeping I go by "if it looks different, consider it something different", even if scientifically speaking there is no robust, quantifiable difference.

Cheers, Sandor

I would put myself down as a splitter, but a calculated one. That is to say not just because a fish looks different, whether it be colour or shape, I try and look at the whole picture before making my mind up.

A point that MattsP mentioned

You should be able to tell by at least one other way (morphometrics or for example capture location)

Capture location can be very misleading as a number factors come into play here. Many fish migrate annually, others may only follow the shallows as the swell of the huge amount of water that inundates the body of water they are in. Depending at what time of year and to a degree the river/stream/lake water level that the species type was captured, compared to where and when the fish in question had been collected. They could be hundreds of miles apart, which could lead on to think that we may have two similar species.

Another problem and one that I am more than a little convinced about is that man himself has caused his own confusion simply by collecting fish in one place and releasing them in another, either by design or accident. A case in point is C. eques described from the Rio Amazonas at Codaja and hundreds of kilometres away CW043 from North east Peru. These have proven to be one and the same species. It is easy to fall into the 'Splitter' trap. I have and so have some scientists.

I have several images and details of Corys that may or may not be given CW-numbers in the future, but I need to be more than 75% certain they are different before doing so.

Lumping is also a debatable area, Again with Corys, just look at C. aeneus. As it stands scientifically speaking this species is found in just about every major river system in South America outside of Chilli and most of Argentina. I can list at least ten that I along with others are convinced are different species.

Ian

racoll wrote:Not sure I can agree here though. After the initial hyperbole, I don't think anyone involved with DNA barcoding is making these statements.

Those projects are still the reason the idea at all gets discussed now.

Almost all of a huge number of studies report highly congruent results with currently regarded species, often with only a handful of paraphyletic or haplotype sharing species. Where it often fails, is when the existing taxonomy is poor, or species identifications are unreliable. There is also an assumption that the existing taxonomy is perfect, which it isn't.

True, that seems to be the case among eels, for instance. However, molecular species concepts also tend to fail in groups undergoing a radiation, such as rockfishes and victoria cichlids, and where there is still significant interbreeding.

I have no problem delimiting (notice I don't say defining) species with just DNA. --- However, the issue is how this DNA data is presented. --- The diagnostic characteristics should remain based on apomorphic character states.

That is a lot better than, effectively, phenetics, and I have no theoretical problem with it, but I am still not convinced it is worthwhile to name a group if the only distinguishing character is that it has, say, a T instead of a G in position 386 of COI, or a unique allele in a fragment analysis. If we're talking something more substantial, like chromosome count or inversions, then OK, but I admit I strongly favor species which are identifiable without access to a laboratory. Even if it means ignoring some existing diversity.

Coryman wrote:just look at C. aeneus. As it stands scientifically speaking this species is found in just about every major river system in South America outside of Chilli and most of Argentina.

I find it hard to imagine there is anyone who do think aeneus is one single species. Corydoras sensu lato definitely needs revision, but its size and history makes it a difficult group to work with, and there's no shortage of groups in need of revision.

Mike_Noren wrote:but I am still not convinced it is worthwhile to name a group if the only distinguishing character is that it has, say, a T instead of a G in position 386 of COI

I don't think anyone is suggesting this level of differentiation, but multiple diagnostic characters across several genes (both nuc and mt) I think is good enough evidence. As a rule of thumb, say two percent is used for mtDNA, then this is 13 sites across a 650bp marker.

A molecular clock and coalescent analysis too I think is a good idea, even to get a rough idea of evolutionary independence.

Mike_Noren wrote:However, molecular species concepts also tend to fail in groups undergoing a radiation, such as rockfishes and victoria cichlids, and where there is still significant interbreeding.

As do morphological hypotheses.

racoll wrote:I don't think anyone is suggesting this level of differentiation, but multiple diagnostic characters across several genes (both nuc and mt) I think is good enough evidence. As a rule of thumb, say two percent is used for mtDNA, then this is 13 sites across a 650bp marker.

Any such firm cut-off would be purely subjective. I suppose one could do a cluster analysis to delimit groups, but then you're probably better off with just a Bayesian tree.

A molecular clock and coalescent analysis too I think is a good idea, even to get a rough idea of evolutionary independence.

I don't think molecular clock estimates are ever a good idea. They're at best an extremely rough estimate obtained from a simulation using a model known to be grossly inaccurate, but they have an unfortunate tendency to be viewed as observations.

As do morphological hypotheses.

Yes. Interestingly molecular and morphological analyses have a tendency to fail in the same way, for the same reasons, in the same groups.

Mike_Noren wrote:Any such firm cut-off would be purely subjective.

Mike_Noren wrote:I am still not convinced it is worthwhile to name a group if the only distinguishing character is that it has, say, a T instead of a G in position 386 of COI

Agreed, but I was just explaining that rather than relying on this single nucleotide, a level of for example 1-2% derived from other similar studies, would give you about 6.5-13 sites to work with, and could be used as a rule of thumb (i.e. a basis for further hypothesis testing), not a strict criterion.

This would be just the same in morphology - i.e. not relying on a single weak character to describe each and every slightly different population as new, but rather using a range of characters in congruence with other studies of similar taxa.

Mike_Noren wrote:I don't think molecular clock estimates are ever a good idea. They're at best an extremely rough estimate obtained from a simulation using a model known to be grossly inaccurate, but they have an unfortunate tendency to be viewed as observations.

Again, it's another technique not to be used entirely on its own, but rather among a suite of potential methods to test alternative hypotheses.