Hi Gregor,
I agree in theory. However, in practice, this is hard to distinguish. A clear case ist that a replacement epithet does NOT change circumscription.
I don't think there is any need to distinguish anything; I think it's purely a function of logic. *If* we define the "essense" of what a taxonID represents as the circumscribed set of individuals (living, dead, and yet-to-be-born)implied to be contained within the taxon circumscription, then it seems to me that a new taxonID is needed only when that circumscribed set changes (exapnds or shrinks). Because taxa are nested sets, the placement of a circumscribed set from one parent set into another parent set doesn't affect the contents of the child set, and hence (in this definition of what is represented by a taxonID), doesn't trigger a need for a new taxonID.
I think the crux of the issue hinges on the premise of the above; that is: *are* we defining the "thing" represented by a taxonID as the circumscribed set of individuals? If so, then I think we can be confident that a new taxonID is only needed when the contents of the cirumscribed set changes. But I'm not sure we all agree that's what a taxonID is intended to represent.
But we already disagree about merging two species: I don't think is necessarily changes the circumscription, it may only correct a previous oversight (i.e. that one circumscription is a subset of the other). But who in practice establishes whether the larger circumscription is the older or the younger name.
I think we are both in agreement on this (I mis-spoke/mis-wrote in my reply to Dave). I think it depends on whether the merge changes the implied members of the circumscribed set. For example:
Suppose Smith describes Aus bus to represent all the individuals of a population in the Marshall Islands.
Then Jones comes along and describes Aus cus to represent all the individuals of a population in the Hawaiian Islands, and he lists all the morphometric differences that distinguishes his new Aus cus from Smith's Aus bus.
Then Brown comes along and decides that Aus cus is a junior synonym of Aus bus, and that Aus bus occurs in both the Hawaiian Islands and the Marchall Islands.
Clearly, in this case, we have three taxon conepts (= three different circumscribed sets of individuals, and need three taxonIDs): One for Aus bus sensu stricto (of Smith), one for Aus cus sensu stricto (of Jones), and one for Aus bus sensu lato (of Brown).
Now, suppose instead that Jones was unaware of Smith's description of Aus bus from the Marshall Islands, and he describes Aus cus based on material from the same Marshall Islands population that Smith based his description.
In this case, I agree with you -- we only have one circumscribed set of organisms, and hence need only one taxonID to represent it, despite the fact that two protonymIDs (=two type specimens) exist within that set.
Unfortunately, a common situation is more along the lines of the following:
Smith describes Aus bus to represent all the individuals of a population in the Marshall Islands.
Then Jones comes along and describes Aus cus to represent all the individuals of a population in the Hawaiian Islands, and he is *completely unaware* of the existence of Smith's Smith's Aus bus.
Then Brown comes along and decides that the populations in Hawaii and Marshall Islands are *clearly* and *unabiguously* the same species, so synonymizes the two names.
In this case, how many circumscritpions are involved? It all hinges on my use of the word "implied" in the notion of a the "implied members of the circumscribed set".
This is why the only way we're going to be able to establish RelationshipAssertions (sensu TCS) is via third-party assertions. In other words, someone is going to have to assert an opinion over whether the implied members of Smith's Aus bus would have included the population in Hawaii, and whether the implied set of Jones' Aus cus would have included the population in the Marshall Islands.
In any case, the point is I think you are right -- the synonomization of two names does not necessarily imply that a new taxonID is needed.
As to your comment:
My main concern is however changes of the higher taxon. If a species was originally placed in a genus circumscribed as leaf spots producing elliptical spores on a specific host plant, and is moved into a genus with defined conidiogenesis, we have changed the circumscription, and previous "identifications" may now be misapplied names.
Yes -- but the new taxonID is not needed because the genus name changed; the new taxonID is needed because the implied set of members of the child set changed (due to the new character-based definition). In other words, the act of changing the genus didn't change the circumscription scope; the change in the character definition of the child taxon changed in. Even if it seems these are the same thing (i.e., that the change in character-based definition was triggered by the genus change), they are still different logical actions.
The central point is that in practice, significant experience and knowledge about past practices is required, to determine whether a new taxon ID should be given or not. Computer typists will get it wrong and experts may disagree about this.
I agree that we require significant experience and knowledge about past practices/etc. to make assertions about the logical relationships between two circumscribed sets represented by two taxonID values. But not for the reason you suggest -- rather, for the reson I described above.
My conclusion is that if every change in genus name is given a new ID, the system is manageable after assignment of ID. Computers can synonymize IDs (sameAs), but they cannot retrospectively split them. :-)
Right -- and this comes back to my fundamental problem with the notion of a taxonID. In order for it to be useful, we need a very clear and shared understanding of the entity that is represented by the taxonID. What we need to do is atomize the ID's -- and this is best done through the TaxonNameUsage object (which is relatively simple to define). So in answer toy your last quote above, in my mind *every* TNU is a "potential taxon" (sensu Berendsohn), so *these* are the Ids we should be using and re-using when we define taxon concepts. My vague understanding of a taxonID is that it is established to represent a set of tnuID's. If it were that simple, then taxonID really just becomes a subclass of tnuID. However, I don't think it's that simple, because the definition of a taxonID is not limited to simply sets of tnuIDs -- it can also be established through member-specimens, morphological characters, genetic characters, etc. Although I think it's logically possible to inherit all these things *through* tnuIDs (after all, anytime anyone ever asserts that a taxon is defined by a set of specimens or characters, a new tnuID is born); we're a long way from having all that sort of information parsed and digitized (as per Kevin's earlier comment).
So, for purely *practical* reasons, I think the main justification for even defining "taxonID" (as more than just a subclass of tnuID), is that our content base for tnuID's is still small, and the inheritance of things like specimens and characters through TNU instances is not yet well-developed. For this reason -- alluded to by Kevin -- I see a practical value to having a taxonID.
But we *still* need to define what it is!
Aloha, Rich