Citations not collapsing

johncolby

Hello,

I made some simple mods to the example in http://citationstyles.org/downloads/primer.html to add:

1. et al.
2. disambiguate-add-year-suffix
3. collapse

While the et al. and disambiguate are working fine, the citations are NOT collapsing. For example, I'm getting this:

(Colby et al. 2008a; Colby et al. 2008b; Colby et al. 2010)

instead of what I want:

(Colby et al. 2008a, 2008b; Colby et al. 2010)

Here's the gist with my edited file: http://gist.github.com/812042

This happens in both the test pane and in Word. I am using Zotero 2.1b5 with MacWord Integration 3.1b2. Is this a bug in the processor, or am I doing something wrong? Thanks!

John

adamsmith

are the "et. al"s in the Colby citation identical?
And why wouldn't you expect (Colby et al. 2008a, 2008b, 2010)?

johncolby

The rest of the authors that are hidden by the et al.s are different.

My mistake, (Colby et al. 2008a, 2008b, 2010) is what I expect. Thanks!

adamsmith

I can't see why this isn't working for you - it works in Zotero 2.0.9 with the old csl -
Could you try what happens when the hidden authors are identical?

johncolby

Yes, I remember this part working fine with the old csl too. I just tried testing it with all the authors edited to be the same, and now the collapsing works as expected. It gives me (Colby et al. 2008a, 2008b, 2010). Any ideas for something I can do on my end to get the correct behavior, or is this just some sort of bug? Thank you, John

adamsmith

That looks like a bug to me - I certainly don't think this should be intended behavior - maybe Frank or Rintze can have a look?

fbennett

I vaguely recall this being intended behavior (the code was written and tested about a year ago). It's comparing the full list of names under the hood, and refusing to collapse them, as a clue to the fact that the authorship of the works actually differs.

It would be easy enough to adjust the processor to permit collapsing in this case. Let me touch the CSL list first, and see what other folks think.

(Edit: In passing, I'll note that the posted CSL code handles spacing badly. The year and author should be joined with a group element and a delimiter, rather than using a space prefix. Some of us had a hot discussion a little over month ago on the topic of whether the processor should rescue styles by suppressing spaces. I was on the should-rescue side of that one, and the citeproc-js processor does work very hard indeed to prevent duplicate spaces in output. I was surprised to see that this style code actually succeeds in sneaking past my handiwork, and produces a duplicate space when the author is dropped out of the citation. As the debate about space suppression was about saving extra work by rescuing existing styles, I'll draw a line here and say that the style should be fixed, not the processor. :)

johncolby

Great...thank you both for looking into this. I've only recently started exploring the new CSL 1.0 syntax/processor, and I am very impressed with all the added functionality. Many thanks!

bdarcus

Maybe it's just that I haven't yet had my morning coffee, but my view is this is NOT a bug.

Explanation (which we can add to the CSL spec if needed):

Year suffix indices are intended to disambiguate multiple items from the same author in the same year for author-date styles.

... where "author" is understood as a representation of the complete author list independent of any subsequent et al. shortening.

E.g. the processor should be seeing something like "doe-john:smith-jane:jones-david" internally for all sorting and disambiguation operations.

I'm not really understanding how it would realistically be any different.

adamsmith

I don't quite understand your argument - you seem to be saying that el als with different second authors shouldn't be disambiguated by a year suffix?
But that's current behavior in both csl 0.8 and 1.0 and required by many styles (where names disambiguation is turned off).

So for disambiguation purposes, the processor already does ignore everything that's hidden by et al - and I think it makes sense if it did the same for collapse/sorting.
I would guess that the majority of styles would prefer if not require this practice.

bdarcus

I don't quite understand your argument - you seem to be saying that el als with different second authors shouldn't be disambiguated by a year suffix?

Yes, that's exactly what I'm saying. How is sorting of the reference list supposed to work correctly otherwise?

ajlyon

But -- what if the full author list isn't shown anywhere in the style? Nowhere. A subsequent reader will just have to guess which (Smith, et al. 2000) refers to which Smith, et al. 2000. [..] in the bibliography?

But I don't really do CSL theory, so I might be off-base here.

bdarcus

But -- what if the full author list isn't shown anywhere in the style? Nowhere. A subsequent reader will just have to guess which (Smith, et al. 2000) refers to which Smith, et al. 2000. [..] in the bibliography?

I'd say that's a bad style. :-)

The theory as I've loosely defined it is that a citation reference is an abbreviated pointer to a fuller description in the bibliography. So the suffix attaches to the bibliographic referernce, and is merely picked up by the citation reference.

adamsmith

I don't agree and I don't see any sorting issues - but if you're right, than the current 1.0 processor does this wrong, too -

(Colby et al. 2008a; Colby et al. 2008b; Colby et al. 2010)

is what the OP gets with Zotero 2.1b and different et als for the two 2008 Colby articles.

Given publication practices in the natural and life sciences, I think year suffix disambiguation for these cases is crucial - it's not uncommon to have a paper with 20 authors, then some new postdoc joins the lab or contributed an image or whatever and the next paper has the same 10 authors, but then some others - so how else would you disambiguate these citations in the text? By adding 11 authors?
(I see ajlyon just made the same point).

And I don't see any issue in sorting - the author names can still determine which of the two articles is "a" and which "b" - it does that currently and it's never been a problem.

bdarcus

So I think this is clearly something that needs substantial clarification.

I stand by my argument that, at least in the majority of the social sciences, "Jane Doe and John Smith" are treated as different authors than "Jane Doe and John Jones". This means that they get separately listed and indexed in the bibliography:

Doe, J and J. Smith (2002a) ...
——— (2002b) ...
Doe, J and J. Jones (2002a) ...
——— (2002b) ...


If a processor did not do that, I'd call it a bug.

Now if that doesn't work for some styles (e.g. in the sciences), then I'd suggest we add a parameter to toggle behavior (and update the test suite to deal with this).

adamsmith

But Bruce I still don't understand - how would you distinguish - in text - between the two Colby et al 2008 options? How does the reader know which abbreviated pointer points to which entry in the bibliography?

I see only two options - year suffix (i.e. Colby et al. 2008a and Colby et al 2008b) or added names (Colby, Smith, et al. 2008; Colby, Meyer, et al. 2008). Added names are already possible in csl and overrule year suffix, so if that's your concern it's already working.
But surely you can't be advocating ambiguous pointers - having two Colby et al 2008 in the text and having the reader guess?

bdarcus

Yes, I think the citation rule should be configured to add the names in that case (or, as I say, we need to change CSL to allow different rules; even conceivable we change et al rules to avoid ambiguous pointers automatically, but am not sure how that would work).

But I work in the social sciences, where long authors lists are exceedingly uncommon. So I've not personally come across this issue.

Rintze

I've posted a question regarding this on the APA blog: http://blog.apastyle.org/apastyle/2011/02/et-al-when-and-how.html

adamsmith

Yes, I think the citation rule should be configured to add the names in that case

OK - but that's already possible, so we don't need to change anything. If you have the "add-names" disambiguation rule set, it precedes the year-suffix rule.
Those also wouldn't collapse - which brings us back to the original question - thanks to Rintze for asking APA - I think the worst case here is indeed that styles disagree, and if experience is any indication they will. Maybe John Colby is following a specific style guide or has an example?

fbennett

I think the original poster's main concern was with the collapsing behavior. His style gets year suffixes on the two Colby et al. 2008 references, and that has to be right, if the style allows/requires year-suffix disambiguation. The question is over whether an invisible difference in the author list (masked by et al. truncation) should block collapsing. It does look rather odd, and I can imagine it raising a long string of user complaints and queries. On that point, we'll see what emerges from the query Rintze has put forward to the APA people.

bdarcus

The question is over whether an invisible difference in the author list (masked by et al. truncation) should block collapsing.

Yes, this is another way to boil it down: to shift focus to the et al handling, rather than the suffix generation.

johncolby

Hi all...I appreciate the great discussion. The reason I am trying to do this is for a specific style (like you guessed). Here is the style guide:

http://www.elsevier.com/locate/inca/522773/authorinstructions

They want et-al-min=1, disambiguation with year-suffix, and collapse=year. The disambiguation/collapse should happen even if the hidden authors differ (so long as the first authors are the same). The is pretty common in Elsevier journals. Thanks, John

fbennett

Great. That's all I needed to get me moving in that direction.

fbennett

I've fixed this in the processor now, and made a release of citeproc-js. The new behavior will make its way into an upcoming beta release of Zotero when the core developers pick it up.

Thanks for pressing on this.

Rintze

APA answer:

When there is more than one reference that shortens to (Smith et al., 2000), as in your example, you would cite as many author surnames as is necessary to distinguish the references (see the exception in section 6.12, p. 175). In your example, five of the six authors are the same, so the in-text citations would include all six references at each mention: (Smith, Johnson, Williams, Jones, Brown, & Davis, 2000) and (Smith, Johnson, Williams, Jones, Brown & Miller, 2000).

adamsmith

yeah - I didn't think about this should have known - APA uses add-names disambiguation, so this has been and will continue to be correct, with no bearing on the question.
I think for styles that don't use given name disambiguation, the collapse rule as described by John Colby (and implemented in csl. 0.81 and now 1.0) makes most sense - and APA's response just highlights what Bruce has been saying all along - that this isn't even relevant in most social science styles, because they would disambiguate differently.

ajlyon

I do wonder if newer data-heavy collaborative projects in the humanities and social sciences will bring such long author lists to the more fields in the coming years. Imagine Dan Cohen's work on Victorian book titles citing the entire Google Books team as coauthors...

adamsmith

if that translates into the same sums for grants that hard scientists get I'll be happy to have 98 co-authors ;-)

johncolby

Thanks, Frank! I patched the changes in your new release into my existing Zotero, and it's working great. -John

Rintze

@johncolby: there is some ongoing discussion on how to collapse cites. I checked the style guide you linked to above, but couldn't find any specific mention that "the disambiguation/collapse should happen even if the hidden authors differ (so long as the first authors are the same)". Just so we can verify that this is really what the Elsevier journals want, could you find us an example, e.g. of a paper where this is the case?

johncolby

Hi Rintze,

Sure thing. Here is one from Psychiatry Research, and another from NeuroImage.

(look for the multiple "Chang", "Moeller", and "Volkow" refs)
Alicata, D., Chang, L., Cloak, C., Abe, K., & Ernst, T. (2009). Higher diffusion in striatum and lower fractional anisotropy in white matter of methamphetamine users. Psychiatry research, 174(1), 1-8. doi:10.1016/j.pscychresns.2009.03.011

(look for the multiple "Hasan" refs)
Lebel, C., Caverhill-Godkewitsch, S., & Beaulieu, C. (2010). Age-related regional variations of the corpus callosum identified by diffusion tensor tractography. NeuroImage. doi:10.1016/j.neuroimage.2010.03.072

Here are the files too: http://dl.dropbox.com/u/4303627/outbox/cite_collapse_examples.zip

Thanks for all the continued support,

John