|
The examples start off more or less where our discussion (facilitated through Aimee) ended, see http://www.soc.napier.ac.uk/tdwg/index.php?pagename=What%20Gets%20a%20Guid The idea is to illustrate various scenarios where concepts and GUIDs might be reused. I’m also adding in some things about how to define the meanings of concepts. I hope there’s no need to stress the provisional character of these scenarios.
This is perhaps the cleanest example, and useful to introduce some terminology. What you see above are two succeeding classification systems for the walnut family Juglandaceae: one according to Manning (1978), and the other according to Stone (1997). For example, there is a Juglans sec. Manning (1978), or a Carya ovata sec. Stone (1997). Each author has his own network of vertical parent/child relationships among the internally accepted concepts. The rejected concepts (e.g. Hicoria is a rejected synonym for Carya, as both Manning and Stone mention) are not shown here, but could be pasted in as less precisely (shallowly) defined concepts, placed next to each corresponding accepted concept.
The accepted concepts all have a set of properties associated with them, represented by the solid rectangles. The genus- and family-level concepts also have children which are connected by lines. The accepted concepts are tied to GUIDs. Manning’s GUIDs start with 10x, Stone’s GUIDs start with 20x in this case. There is no reuse of concepts or GUIDs. This means, in order to understand any sort of lateral concept relationships among Manning’s (1978) and Stone’s (1997) systems, one (i.e. a third party) would have to author lateral relationships of the kind: 100 = = 200, 105 >< 205, and so on.
So, to sum up, there are three possible sorts of relationships here:
(1) internally, rejected (shallow) concepts to accepted (deep) concepts (e.g. Hicoria sec. Stone (not shown) → Carya sec. Stone (shown));
(2) vertical relationships among accepted concepts (e.g. Carya sec. Stone → C. ovata sec. Stone); and
(3) lateral relationships among accepted concepts (e.g. Carya sec. Stone → Carya sec. Manning).
I would strongly suggest that we do not worry now about vertical or lateral relationships involving concepts that an author rejects, because often their referential extensions are too ambiguous (see my paper).
So far so good, I hope. If we (i.e. the taxonomists) used this approach in the future, all the time, then concept inflation would be very high. Any kind of understanding across classifications would be localized in the lateral concept relationships. I think we can do better than that.
This is still a fairly simple but different case. Let’s assume Stone (1997) was aware of the concept approach, and wanted to be both comprehensive and conservative in his classification. (You could easily imagine Manning and Stone to instead represent two future versions of ITIS, where ITIS’ taxonomic experts were “schooled” in the concept approach.)
The idea is: for some reason, Stone (1997) wants to reference one of Manning’s concepts in a revised classification that is nevertheless mostly Stone’s “own”. Perhaps Stone (1997) never saw any specimens of J. cinerea, or just thinks Manning’s concept is close enough for all intents and purposes. In short, Stone (1997) wants to defer any expertise on J. cinerea’s referential extension to Manning (1978), but claim expertise on the extension of Juglans sec. Stone, among other things. He cites the concept labeled with GUID 103, and authors that with GUID 201. GUID 103 appears in a different context than where it was first assigned.
What’s happened here is that Stone “borrowed” expertise from Manning. No individual taxonomist can know everything once the groups reach a certain size. The deliberate “borrowing” of expertise can and must happen all the time. A knowledgeable author should be allowed (as an option) in the future to make her own calls about borrowing and claiming, depending on her expertise and stated intentions to contribute to our taxonomic knowledge. Claiming is easier and cleaner but leads to more inflation and possibly less mutual understanding than borrowing. A conservative taxonomist will claim only if something is really (deeply) new, and borrow most everything else. This minimizes inflation.
I thought it might be a fundamental issue whether (and how) we allow such “composite” classifications of cited and authored concepts to exist in our database, and how to implement this in a schema.
Ok, I stuffed a number of scenarios into this example. What is going on? First of all, Stone (1997) was interested mostly in a genus-level revision. Let’s say he examined and accepted as “close enough” all species concepts authored by Manning in 1978. As you see, all of Stone’s “deferred” species concepts have Manning’s 10x -type GUIDs. Zero inflation at this level. So far so good.
However, Stone thinks that Manning misinterpreted the proper extensions of the two genus-level concepts. Could be some serious alternative theory about how to interpret the floral structures of walnuts, or just a mix-up of herbarium sheets. As a result, the species-level concepts glabra (GUID 106) and nigra (GUID 105) - each sec. Manning (1978) - now have “switched” their respective parents. I think it’s easy to make the case that Stone should author (not just cite) these new parents. In fact that “claiming” of genus-level concepts is what he wants to do anyway in his revision. So he gets the new concepts Juglans and Carya sec. Stone, with GUIDs 201 and 202, respectively.
But it doesn’t end here! Instead of rattling all the way up a/the tree with his new genus concepts, Stone decides to “leave the family-level as is.” What he’s trying to express here, is that in spite of his lower-level rearrangements, Manning’s concept of Juglandaceae is still accurate enough. It contained certain lower-level concepts before, and does so now. No “need” for Stone to author a new family and cause concept inflation. Here’s another way to look at this (contentious) issue. It is my last and most extreme example for now:
Ok, now Stone accepts all previous species- and genus-level concepts (Juglans and Carya) published by Manning, with their corresponding GUIDs. He discovers a new species, concept paliurus sec. Stone (1997), which gets its new GUID. It turns out to be outside of Manning’s (reasonable) limitations for Juglans and Carya. Therefore, it is placed in a new genus Cyclocarya sec. Stone (1997), with a new GUID 210.
Question: is there a sense in which Stone could still - should he so desire (it’s an option, not a must) - consider Juglandaceae sec. Manning (1978) to be the parent of Cyclocarya sec. Stone (1997)?
Answer: ostensively - no! Manning never looked at Cyclocarya, and in any case failed to connect it as a child to his concept of Juglandaceae. But: intensionally, yes, it’s at least possible! What if Manning’s property-based definition of Juglandaceae was so precise and accurate that it clearly applies to the properties observed in Cyclocarya? Based on these properties, one can only conclude that Manning - whatever else he might have thought - would have had to identify Cyclocarya as a member of the walnut family.
I’m raising an important relation here, upon which much of our mutual understanding in taxonomy rests. The question is: had a previous expert speaker seen this entity X (which she actually never did), how would she have classified it, and why? Together with the issue of borrowing vs. claiming expertise, these kinds of judgments make up a lot of the referential continuity among classifications and uses of taxonomic names, as we humans (inductively) perceive it and rely on it in conversations and all non-taxonomic publications.
To express this continuity, should Stone (1997) wish to do so, he would have to reinterpret Manning’s Juglandaceae in a certain way:
What I’ve done here is also in Figs. 7-10 and in Table 4 of the manuscript. The disconnect between the lines in the hierarchy is supposed to indicate that their referential extensions are entirely specified through their properties, i.e. the solid rectangles. The lines basically don’t count. Stone has been given a choice here to use only a part of Manning’s full definition of Juglandaceae, the part that talks about the shapes of walnuts, their unique leaves and floral organ arrangements, etc. Under that purely intensional definition, Cyclocarya sec. Stone (1997) is subsumed under Manning’s (1978) Juglandaceae concept (and GUID 100), even though Manning never saw or mentioned Cyclocarya. Manning’s property-based definition is still useful and precise enough after Cyclocarya had been discovered - or at least that’s what Stone wants to express.
As I suggested before, there still must be a way in which Stone makes Manning’s Juglandaceae “his own,” though without authoring a new concept of Juglandaceae. He could only author a concept status assignment (which says in essence: Stone’s accepts Manning’s Juglandaceae). That assignment would somehow reference Cyclocarya. The “includes” (>) relationship among (Manning’s) Juglandaceae and (Stone’s) Cyclocarya is purely intensional.
What we’re moving towards in this contentious case is the issue of permitting succeeding authors to untie existing parent/child connections, if (and maybe only if) the parent concepts also have intensional aspects to them, which would allow the taxonomist to reduce concept inflation by citing those parents as acceptable. They are acceptable only if one allows their ostensive (child-pointing) and intensional (property-inferring) aspects to be separated. The motivation for this is to minimize concept inflation, and maximize concept reuse. The more concepts are properly reused, the closer we get to creating a language that’s maximally superior to names. This is a “parsimonious” concept approach - only author concepts when both the intensions and ostensions are new, otherwise cite existing ones. Note that example 4 makes a flexible “stopping rule” possible - stop authoring parents whenever you think there are congruent intensions.
I think something like the AtomicTaxa (which might be renamed as IntensionalConcepts) - concepts without the ostensive chains (also without vouchers) - might be introduced in the database schema to isolate those intensional meanings from the ostensive ones. I’m fairly sure that the TDWG TES need not have this now, but ultimately the (optional!) approach hinted at above may offer more services to experts than the existing schemas. At least I hope it’s worth thinking about that.
Attachments:
|
