Collapse vs. disambiguation in multiple-entry cites

djross3

Edit: the simpler way to address this might be in the related reply below. Collapse should (optionally?) first (silently) expand to the full names before collapsing. Is that possible?

--
Let's assume I have a number of references by A. Smith and B. Smith. In general, they're from different years and therefore unambiguous (by cite), but rarely overlap in the same year.

Let's say that A. Smith wrote a paper in 2000 and another in 2010. And that B. Smith wrote a paper in 2005 and another in 2010.

Now, I would consider the following cites to be unambiguous:
(Smith 2000, 2010) (=A. Smith)
(Smith 2005, 2010) (=B. Smith)

But instead, what I end up with is the very awkward:
(Smith 2000; A. Smith 2010)
(Smith 2005; B. Smith 2010)

In other words, disambiguation takes precedence over collapsing same-author, so it acts as if these are different authors. Very odd.

The best I can do is to use "suppress author" on the second cite and end up with:
(Smith 2005; 2010)
But that's the different-author delimiter, rather than the same-author delimiter.

I could manually edit the cites but I'd rather avoid that. And the other problem is that when updates are disabled (in a long document) I won't notice these oddities until the full document is refreshed, and then I'd have to 'proofread' everything.

My impression from reading the documentation is that this is because author-disambiguation takes precedence over collapsing by-year. Of course this only comes up as an issue when disambiguation is by-cite rather than always by name so that it would just be (A. Smith 2000, 2010) to begin with. But I would prefer to keep disambiguation out of it when not needed because usually in the context of writing it's clear who I'm talking about, but at least can be disambiguated in the bibliography if needed.

One solution here I'd wonder about is if the alphabetic suffixes for disambiguation could be added across multiple authors, e.g. resulting in A. Smith 2010a and B. Smith 2010b. But I don't see any option for this, I think because of the order of application of disambiguation rules. (I also realize that isn't aesthetically optimal in the bibliography, even confusing, but I'd prefer it to the odd results for in-text cites above!) [Edit: it looks like the Cell style does this. But it looks like can only be done while disambiguate-by-cite is not enabled, so it still results in odd-looking cites, see the additional comment below.]

I'm using my own custom style for this with the following setup:
<citation et-al-min="4" et-al-use-first="1" disambiguate-add-year-suffix="true" disambiguate-add-names="true" disambiguate-add-givenname="true" givenname-disambiguation-rule="by-cite" collapse="year">
However, the exact same results appear to occur for the default Elsevier - Harvard (with titles) style for example.

Any ideas? Is this a bug? A feature? Any workarounds or options? Thanks!!

djross3

Minor additional comment:

Collapse-by-author behaves a little unexpectedly when two different authors with the same name are not ambiguous due to year of publication.

Following the examples above we'd get:
(Smith 2000, 2005)
when collapsing A. Smith 2000 and B. Smith 2005.

I'd think the proper way to collapse that would be based like:
(Smith 2000; Smith 2005)
Even though that looks surprising, it does imply they are different authors. It makes more sense than collapsing them when disambiguation is by-cite only.

Suggested solution:
Collapse should first fully expand all names and see if they are the same author*. Then it would collapse them only if the full names are the same. This would solve the problem here, and also the one above.

[*Of course it wouldn't know if two authors happened to have two identical names, like common "John Smith", but there's no way to deal with that, short of annotating their affiliation, date of birth, etc. So let's just ignore that case. For the purposes of the bibliography we might as well then assume they really are the same person.]

My personal solution: until/unless the solution above can be implemented, I will use the Cell-style option of just:
<citation et-al-min="4" et-al-use-first="1" disambiguate-add-year-suffix="true" collapse="year">
This isn't perfect but it consistently disambiguates and it is generally unlikely that two authors of the same name would be cited in the same cite, so I'd rather have that error than have the one where they are not collapsed if they would otherwise need to be disambiguated.

--
Note: again, this applies to my custom style, but the same behavior can be observed in Cell or Elsevier-Harvard for example.

bwiernik

This isn’t a bug. It’s the expected behavior for styles using this form of disambiguation. Frankly, it is much easier on your reader to not make them have to realize that they need to first look at Smith 2005 in order to determine which Smith 2010 you mean.

djross3

Specifically when cited together, this seems obvious:
(Smith 2005, 2010)
Note that the reader would already be looking up 2005 if they were interested in what was cited. I'm specifically talking about only when they are appearing together in one cite.

I completely agree with you that in a paragraph the following would be confusing:
Smith (2005) said.... and then Smith (2010) said.... when they are different people, but that's a different scenario where the disambiguation would be desirable.

On the other hand, the biggest problem with this seems to be that:
(Smith 2005, 2010)
is also the form that would appear if two different Smiths are found in the bibliography but do not happen to be ambiguous for those two years. Do you not agree that this is confusing?

Would there be a problem with collapse being (silently) based on the full names, not just what is displayed?
I'd prefer to see (Smith 2005; Smith 2010) in that case, which would imply I should look for two different Smiths in the bibliography.

That would fix both problems, without, as far as I can tell, introducing anything confusing. It would still cite with disambiguating initials when not cited in the same cite as another unambiguous cite by the author.

Of course I wouldn't object if it became:
(A. Smith 2005, 2010)
That is, disambiguation could be global for the cite, when collapsing. (I imagine that could be harder to program, so I'm not specifically requesting it, but it would work too.)

In short, could an additional collapse flag be added to CSL: collapse-only-when-full-names-equivalent?