Wiktionary:Votes/2013-10/Reconstructions need references

Reconstructions need references

 * Voting on:
 * Option 1:
 * Allowing appendix pages on reconstructed protoforms (e.g., ) only if they have references to sources (scholarly work) that
 * a. give the exact form, or
 * b. provide evidence that supports that form (e.g., sound changes that would create it).
 * Appendix pages on reconstructed protoforms that remain unreferenced after a certain amount of time was given for references to be added should be deleted.
 * Option 2:
 * Allowing appendix pages on reconstructed protoforms (e.g., ) only if they have references to sources (scholarly work) that
 * a. give the exact form or one differing only in orthography / notation, or
 * b. provide evidence that supports the form (e.g., sound changes that would create it).
 * Appendix pages on reconstructed protoforms that remain unreferenced after a certain amount of time was given for references to be added should be deleted.
 * Rationale: See Wiktionary talk:Votes/2013-10/Reconstructions need references. The voters only vote on the proposed action, not on the rationale.
 * Vote starts: 00:01, 30 August 2015 (UTC)
 * Vote ends: 23:59, 30 September 2015 (UTC) or later


 * Vote created: Pereru (talk) 23:41, 18 October 2013 (UTC)
 * Discussion:
 * [[Image:Wikt rei-artur3.svg|20px]] Requests_for_deletion/Others (later at Template talk:needsources)
 * [[Image:Wikt rei-artur3.svg|20px]] Beer_parlour/2015/August
 * [[Image:Wikt rei-artur3.svg|20px]] Wiktionary talk:Votes/2013-10/Reconstructions need references

Support option 1

 * 1)  Almost a "philosophical" distinction. As I personally see it, Wiktionary is not an appropriate platform for novel ideas. And there's that. (Idk, if the following is appropriate for a vote page, feel free to ignore.) An immediate prerequisite for using it as such a platform would be an assumption of boundless trust (trust of visitors and other editors alike) when trust is, in fact, one of the hardest things to build and one of the easiest to lose (and a really underappreciated topic in sociology given its huge importance.) Then there is the school of thought of what I like to call the "rat maze idea" in which people will follow paths cleverly set out by you (the "rat master"? :D), perhaps it's right but I'm instinctively opposed to this, I always assume that people are "smart" for some reason (or at least that they carry out their own private analysis whenever they see something) moreover they want to carry out their own analysis, so, that's another "philosophical" divide. A hypothetical scenario where someone is, say, synthesizing several published sources (an action that should be OK) and then keeping them secret, imo, would be pretty much just plain rudeness (a.k.a. the methodology is only for The Initiated Ones, it's not for mere mortals like you) which is, um, just kind of rude. So, that's my scratching of the surface of some "philosophical" arguments. Neitrāls vārds (talk) 20:35, 1 September 2015 (UTC)
 * , with the condition that pages must be nominated for verification or deletion before they are deleted, rather than being deleted outright. — Ungoliant (falai) 17:32, 3 September 2015 (UTC)

Oppose option 1

 * 1)  Sounds like a good way to ensure the needless large-scale deletion of useful material. Ideally everything should be referenced, but in practical terms, Wiktionary does not have enough editors working on particular reconstructed languages to go through every single existing reconstructed entry and add references to each one by some time limit. If something unreferenced is specifically challenged, then certainly it should be overruled by scholarly sources, but for the thousands of reconstructed entries where no doubts have been expressed, forcing deletion if unsourced would do far more harm than good for the project. Disagreement resolution can be handled on a case-by-case basis, favouring material with scholarly references, without applying such an indiscriminate cudgel as this proposed policy. Vorziblix (talk) 17:22, 3 September 2015 (UTC)
 * I don't think scholarly sources should always trump Wiktionary's own reconstructions. Scholarly sources might be very outdated (Pokorny) and it wouldn't make sense to replace a current reconstruction with an old one. Secondly, scholarly reconstructions can be demonstrably wrong, as in the case of where the reconstructed ancestor for Lithuanian  is given as Proto-Baltic, which is obviously wrong; the correct reconstruction is . Likewise, the reconstruction given for  is Proto-Baltic , even though the ō > uo change is limited only to East Baltic. —CodeCat 18:33, 3 September 2015 (UTC)
 * True, many scholarly sources may be wrong, outdated, etc., but Wiktionary’s own reconstructions should surely be based on evidence provided by newer/more correct scholarly sources in such cases; even if one cannot cite the exact word forms that more correct models suggest, one can cite the models themselves and reconstruct on that basis, overruling the outdated sources. Vorziblix (talk) 19:40, 4 September 2015 (UTC)
 * @Vorziblix: Can you please clarify why this would lead to a "large-scale deletion", especially with Ungoliant's caveat applied? If each entry has to first be nominated before it is deleted, why would there be a large-scale deletion? Have you seen WT:RFV used for large-scale deletion of material that does not have attesting quotations in the Wiktionary database, as a result of WT:ATTEST? The entries that have quotations in Wiktionary to support WT:ATTEST are a tiny fraction of all entries, right? --Dan Polansky (talk) 18:21, 4 September 2015 (UTC)
 * The proposal doesn’t include Ungoliant’s caveat, and I voted before the caveat had been written here. If this is indeed applied like RFV, selectively and on a word-by-word basis, the objections above evaporate, but the proposal as it stands can easily be interpreted to mean that all protoforms will be universally deleted by a certain date if unreferenced. Consider my oppose votes to be weak supports if the caveat is included, but strong opposes if not. (Other issues still exist, but as long as »provide evidence that supports that form« is interpreted broadly enough, they are surmountable.) Vorziblix (talk) 19:40, 4 September 2015 (UTC)
 * @Vorziblix: You can place the same caveat in your vote as others did. I do acknowledge that there is a wiggle room in the phrasing of the vote that some might want to abuse, which is why these caveats that people explicitly add to their votes are probably a good thing. Note that a vote in the oppose section is a bit hard to count as a conditional support, even if it says so in the text itself. --Dan Polansky (talk) 20:01, 4 September 2015 (UTC)
 * Right, sure; amended below. Vorziblix (talk) 05:37, 5 September 2015 (UTC)
 * 1) * unless point "a." is amended to something along the lines of "give the exact form or an equivalent differing only in orthography". --WikiTiki89 18:19, 3 September 2015 (UTC)
 * 2) I oppose option 1 (the wording is too narrow) and support option 2. - -sche (discuss) 23:43, 3 September 2015 (UTC)
 * 3)  per -sche. —Μετάknowledge discuss/deeds 18:13, 5 September 2015 (UTC)

Abstain about option 1

 * 1) . I posted support to option 2. I can't figure this out quickly. Since the result on option 1 is now 2-3-0 and since option 2 looks like it could pass, I am not motivated enough to spend my energy on option 1. --Dan Polansky (talk) 10:54, 20 September 2015 (UTC)

Support option 2

 * 1) . Ungoliant's caveat goes without saying, much like WT:CFI doesn't spell out that words which are unattested have to be sent to WT:RFV before being deleted, but we all know that there's usually no way to know that a term is unattested without sending it to WT:RFV. - -sche (discuss) 20:01, 3 September 2015 (UTC)
 * 2)  --Vahag (talk) 20:57, 3 September 2015 (UTC)
 * 3)  with the same condition that Ungoliant stated above. --WikiTiki89 21:20, 3 September 2015 (UTC)
 * I'll also add that entries without references, like regular entries, cannot be nominated for RFV just for fun; this should only be done if there is legitimate doubt of the entry's accuracy. --WikiTiki89 17:47, 4 September 2015 (UTC)
 * 1)  this one too. — Ungoliant (falai) 17:48, 4 September 2015 (UTC)
 * 2)  this, and I agree that -sche is right to suppose that Ungoliant's condition an inherent part of our processes, rather than something that must be spelt out. —Μετάknowledge discuss/deeds 20:04, 4 September 2015 (UTC)
 * It doesn’t hurt to be clear, at least as long as you-know-who remains active in Wiktionary (hint: it’s the same guy who was removing subsenses, and undoing whole edits that included susbsenses, with the excuse that there was no individual discussion making it clear that subsenses were allowed). — Ungoliant (falai) 20:23, 4 September 2015 (UTC)
 * , conditional on the given caveat, as per the rest of the discussions on this page. Vorziblix (talk) 05:37, 5 September 2015 (UTC)
 * 1)  with Ungoliant's (: 17:32, 3 September 2015) and Wikitiki's conditions (hereinbefore: 17:47, 4 September 2015). — I.S.M.E.T.A. 22:57, 8 September 2015 (UTC)
 * 2)  --Daniel Carrero (talk) 18:38, 9 September 2015 (UTC)
 * 3)  —Mr. Granger (talk • contribs) 12:52, 15 September 2015 (UTC)
 * 4)  with hesitation. There may be repercussions of the wording that I do not like. Nonetheless, for unattested reconstructions (not meeting WT:ATTEST), we do need to use references in some way to support them, or else we'll fail the verifiability requirement. Importantly, the attested forms putatively derived from the reconstructed form are not attestation of the reconstructed form itself, contrary to what I have seen someone say; the gap between the evidence and the thing supported by the evidence is too big. I furthermore think that references should be used in conflict resolution about the specific rendering or mark up of the reconstructions. Any deviation from the references should have an expressly stated rationale, perhaps on the talk page of the entry. Each reference that does not use the exact form should have the form used in the reference stated. --Dan Polansky (talk) 10:32, 20 September 2015 (UTC)

Oppose option 2

 * 1)  A valuable proposal, but still subtly overstated in a few ways so that I cannot support elevating it into policy (even after the addition of Option 2).
 * The first issue is what I see as an overly lexicographical approach to proto-language appendices. Their primary purpose cannot be to document the lexicon of a reconstructed language, given that the language in question is, after all, unattested. The comparative method allows us to infer features of such languages; but rather than being primary objects of study on their own, the main motivation for reconstructing languages is to shed light on the prehistory of attested languages. This includes etymology: one of the valuable features of proto-language appendices is that they allow discussing e.g. the wider Indo-European etymological proposals of related Germanic words in a centralized fashion. Leaving a stable note like "from PG *whatever, possibly from PIE *wkod-" on mainspace entries and taking the detailed discussion to a dedicated page is much preferrable to trying to keep a dozen different etymology sections in sync, even if there are sourcing issues in figuring out what the PG form was exactly.
 * The second issue I have is over the clause for "provid[ing] evidence that supports [a] form" (perhaps nitpicky, but an issue regardless). I agree that a proto-form that has not been directly sourced from literature requires some type of sourcing on how it has been derived. Having to present all individual references on each individual appendix page would be however overkill. Moreover, internal consistency is a highly valuable trait of any proto-language reconstructions we present, which already requires handling any "reconstructions-by-Wiktionary" likewise in a centralized fashion. Thus e.g. "giving sound changes that support a form" should be rather a task for e.g. the corresponding proto-language considerations page (or perhaps a dedicated subpage or two, if we want to actually spell out sound correspondences, instead of merely listing references that give them).
 * However: I would still support a proposal (can we still add an Option 3?) that appendix pages can only present a reconstruction, as if as a headword, if it is indeed in some fashion supported by references. But a failure to do so should not lead to an outright deletion of the appendix page, only to its re-design as less proto-term-centric. Possibly key subsections such as ==== Descendants ==== or ==== Derived terms ==== could be rather included directly under a === Reconstruction === or === Etymology === header, discussing what is known and what is not of the proto-form's reconstruction.
 * --Tropylium (talk) 22:10, 3 September 2015 (UTC)
 * I think you hit the nail on the head that the main purpose of reconstructed pages is to list cognates. To delete those when the reconstructed form is unsourced is throwing away the baby with the bathwater. Then again, I suspect that some users here would even go as far as to require sources merely to state that the cognates are in fact cognates. —CodeCat 23:17, 3 September 2015 (UTC)
 * It's not directly relevant to this vote, but for the record, I do think that references might be sometimes necessary for verifying the existence of a cognate set — especially with older proto-language stages like Proto-Indo-European that have hundreds of descendants, but also to avoid late loanwords being reconstructed as common inheritance (as in, ~, both from  and not from a technically entirely plausible Proto-West Germanic **forta(z); or  ~ , both from  and not from a technically entirely plausible Proto-Finnic **paattei). --Tropylium (talk) 23:48, 3 September 2015 (UTC)
 * Oppose for the same reasons as Option 1. Vorziblix (talk) 17:37, 4 September 2015 (UTC)
 * (Consider this instead a weak support if Ungoliant’s caveat is taken as given. Tropylium’s suggestion would be better still.) Vorziblix (talk) 19:40, 4 September 2015 (UTC)
 * Indented as striken out, and removed the icon. --Dan Polansky (talk) 14:35, 12 September 2015 (UTC)
 * 1) . The consequences of this vote haven't been clarified enough. "provide evidence that supports the form (e.g., sound changes that would create it)" means what exactly? The descendants are the actual evidence, and I am already very strict with deleting any proto-pages that lack descendants. Since this vote seems to focus on sourcing rather than evidence, I'm assuming that the intention is to explain how the descendants developed from the reconstructed form, in order to demonstrate the correctness of the reconstruction. But it seems nobody has considered that this essentially puts a stop to the adding of descendants altogether, because the requirement would apply to every descendant language individually. For example, we would no longer be able to add  to the descendants of  without first creating a page explaining the sound changes from Proto-Germanic to Danish. It makes a bureaucratic nightmare out of something that should just be simple and obvious. —CodeCat 19:23, 9 September 2015 (UTC)
 * , I guess. To both. I don't really see the point of this change, that can't be done already in discussion pages and with requests for references/clarification. Our more active etymologically inclined editors are pretty good at doing this anyway. I wouldn't mind having a separate appendix page for different proto-languages that explains and discusses sound changes/developments that could be referred to in etymologies when needed, or something to that effect. Anglom (talk) 22:00, 9 September 2015 (UTC)
 * , re "I don't really see the point of this change, that can't be done already in discussion pages": the reason for this vote (which was initially drafted in 2013, as reflected in its pagetitle, but which we held off on starting for two years in the hopes that things could be worked out in discussion pages) is that in discussion pages, one of our most active etymologically inclined editors (CodeCat) has repeatedly insisted, for years now, that reconstructions don't need references and that the theories behind original research (OR) don't need to be documented or attributed to their theorists; see e.g. here and here. She went as far as to block one of our other etymology-editors who had been tagging her referenceless OR as needing references or documentation (see here). In the face of that, this vote is intended to unambiguously affirm that yes, reconstructions need to be supported by references or documented theories (sound changes, etc). - -sche (discuss) 00:18, 10 September 2015 (UTC)
 * I actually blocked Pereru because he was recreating Proto-Baltic entries after a BP discussion had agreed to turn it into an etymology-only language (i.e. POV pushing against consensus). It wasn't because of the source templates, though of course it did occur in the same context. —CodeCat 01:01, 10 September 2015 (UTC)
 * Thanks to -sche for the above statement and for providing links to the relevant discussions. --Dan Polansky (talk) 11:05, 20 September 2015 (UTC)

Abstain about option 2

 * 1)  Although I strongly support the idea that reconstructions need references, I find the language in this proposal to be a bit vague. If a source uses different orthography or notation from our own, is it clear what our interpretation would be? If a source suggests a sound change, does that pin down a specific reconstruction that it attests? Just as there is a fine line between protologisms and neologisms that we can cite, I definitely would not want to see this opened up to original research. DAVilla 05:48, 15 September 2015 (UTC)
 * @DAVilla: What is your thought on option 1? --Dan Polansky (talk) 10:56, 20 September 2015 (UTC)
 * Also abstaining, I guess. Although the wording is slightly stricter, I don't understand how "differing only in orthography/notation" doesn't "provide evidence that supports the form". In other words, if it could be visualized, the section option 1 tries to carve out is concave. DAVilla 16:36, 22 September 2015 (UTC)

Decision

 * Option 1 fails 2-3-1 and Option 2 passes 10-3-1. Note that 6 of the 10 supporters of Option 2 explicitly supported the caveat that such entries be given due process rather than deleted outright, so although that was not part of how the vote was framed, I would say that there appears to be consensus for that. —Μετάknowledge discuss/deeds 00:49, 7 October 2015 (UTC)