FUNKNET上高手过招，虽无刀光剑影，倒也另观众看得过瘾。
人物：
- Dan Everett （深入亚马逊丛林调研Piraha土著语言数十载，《别睡着，有蛇呢》作者，反乔姆斯基的代表人物之一，找到了recursion的痛点）
- Richar (Dick) Hudson（生成派知名学者，伦敦大学学院荣誉教授）
- David Adger（生成派知名学者，伦敦玛丽女王大学语言学教授）
地点：
FUNKNET，美国 Rice University 语言学系运营的邮件讨论组，该邮件组的选题范围 “dedicated to discussion and news about all aspects of functionally-oriented linguistics”，所以FUNK的意思，你懂的。
话题：
Recursion（以及Merge）
Dick 发起，引来Dan和David的激烈讨论，小编是不是可以把它叫做 3-D论战呢？? 出彩的部分是Dan就“生成派和认知功能派的区别”所列证据和相关讨论。✌️
先来看话题的引起，那是几天前的事 (邮件时间为当地时间, 有时差)：
On Mar 6, 2016, at 8:12 AM, Richard Hudson<r.hudson@ucl.ac.uk>wrote:
The recent exchange has reminded me that I’ve never understood why recursion strikes Chomsky as such a big deal in language, given that it’s such a prevalent feature of our ordinary non-linguistic cognition. We apply operations recursively all over the place. For example:
* Finding a route from A to Z, not previously traversed:
we presumably do something like this:
1. Find B such that A to B is known, and B is nearer to Z than A is.
2. Find Y such that Y to Z is known, and Y is nearer to A than Z is.
3. If B=Y, follow A – B=Y – Z
4. Otherwise repeat #1-3 with B and Y instead of A and Z.
* Finding how persons A and Z are related:
1. Find mother Bf and father Bm of A
2. Find mother Yf and father Ym of Z
3. If Bf = Yf, or Bm = Ym, classify the relationship.
4. Otherwise repeat #1-3 with Bf, Bm instead of A and Ym, Yf instead of Z.
The point is that these aren’t simply repeated events, like walking, but purposeful attempts to solve sub-problems, so we have to remember the main problem while we’re solving the sub-problems. Any creature that can do that has the capacity for doing the same in syntax, so recursion isn’t unique to language. I bet even the Piraha~ do those things?
Dick — Richard Hudson
Dan给出的回复一开始是这样的：
On 06/03/2016 13:48, Everett, Daniel wrote:
Dick, I discuss this at length in several publications. There is no problem in claiming that the FLN (Narrow Language Faculty) underwrites human language via Merge (Berwick and Chomsky 2016) while simultaneously recognizing that recursion pervades cognition.
There are several possibilities:
1. Recursion is found in multiple modules (if you believe in modules, which I do not).
2. Recursion appeared first in language and was then naturally extended to human thought more generally.
3. Recursion first appeared in human thought and then found its way to language.
Any of these is compatible with Chomsky’s theory.
My own view is that recursion is a property of human thinking, exploited as a structure-building device in most languages for the functional reason that it allows for more information-packing per sentence and clearer reference for hierarchical relationships.
For example, Peter said that Mary claimed that Bill uttered that John believed that the moon is made of green cheese.
As I have elsewhere argued, this series of thoughts is easier to process in a recursive construction than via parataxis. On the other hand, since this is not a crucial type of sentence for any language, it is not surprising that its greater efficiency is ignored by some languages which either (i) lack recursive syntactic structures and thus presumably operations; (ii) lack multiple degrees of embedding (there are languages which have embedding but, since the embedding is limited to one or two degrees, either require a stipulation on Merge or may be said, as I would claim, to lack recursion as well. Embedding is not recursion).
Dan
Dick接着做出了回应，并表达了谦逊之意：
On Mar 6, 2016, at 9:25 AM, Richard Hudson <r.hudson@ucl.ac.uk> wrote:
Thanks Dan. I’m glad to know that we agree, though I’m embarassed at my ignorance of your work. But that just shifts the question onto Merge. I’d say Merge-like operations are also pervasive in non-linguistic cognition. Would you agree?
Dick
对于Dick的进一步问题，Dan 很快又做了回应：
On 06/03/2016 14:28, Everett, Daniel wrote:
Dick, No, I don’t think Merge is all that pervasive outside of grammar. The reason is that Merge produces endocentric, binary-branching structures only and I think that those are not even always found in languages, much less in thought. I am currently writing a largish commissioned paper for the Journal of Neurolinguistics on the evolution of language and a book for Profile (UK) and Liveright/W.W. Norton in the US on language evolution.
Dan
Dan再作回应：
On 7 Mar 2016, at 11:35, Richard Hudson <r.hudson@ucl.ac.uk> wrote:
But Merge is basically just the operation of combining two objects and drawing consequences from that combination. As you know, you don’t have to assume the binary-branching structure – you could go with dependency structure, which assumes that if you combine word A with a dependent word B, then you don’t get a larger unit containing both, but you do get a modified version of A, which you can call A’. So if you combine the words /gift/ and /vase/, with /gift/ as the dependent, you get a modified mental representation of /vase, vase’/, which means not just ‘vase’ but ‘gift vase’.
It seems to me that this is exactly the same mental operation that you get outside language by putting a gift-card next to a vase, showing that it’s a gift – and when you see a man on a bike and recognise him as a cyclist; and so on and on. Merge permeates our mental life – if you separate it from the oddities of PSG. In understanding, it’s what we do when we take account of the wider context of an object; and in action, it’s what we do when we take advantage of the larger context in refining the social significance of an action. Dick
后来，直到David Adger的加入：
On Mar 7, 2016, at 7:11 AM, David Adger <d.j.adger@qmul.ac.uk>wrote:
Not quite, though of course anyone is welcome to define anything in anyway they want. Merge, for Chomsky and most people in minimalist syntax, is a set formation operation. You’re right that you can lift the binary branching restriction, or think of it as deriving from something outside of the set formation operation. I’ve argued, for example, that it’s binary because of memory limitations, others have suggested that it’s binary because that maximises efficient search. But the standard definition simply stipulates binarity. It’s also the case, as I was saying to Dan, that Merge is not necessarily intrinsically headed.
So that makes it different from your dependency formulation, where headedness is crucial. I agree that some notion of headedness, in a very general sense of maintaining cognitive focus on one element while adding to it, is plausibly cognitively general, but that’s not part of the current definition and hasn’t been for at least a decade; headedness of dependency would b e something separate, added to the creation of structure (it goes under the rubric of the Labelling Algorithm right now, and there’s quite a lot of stuff trying to work out the consequences of different ways of formulating the general issue).
So maybe Labelling is connected to the cognitively general operation you suggest, but, at least in current versions of minimalist syntax, Merge is not. It could be connected to a general grouping capacity, though. I have a frontiers paper with Peter Svenonius where we talk about some of this stuff if you’re interested. It’s athttp://journal.frontiersin.org/article/10.3389/fpsyg.2015.01421/full
Best,
David
且看Dan如何机智回应：
On 07/03/2016 13:20, Everett, Daniel wrote:
Merge began life as a binary representation because of the influence of R. Kayne’s 1984 paper, Unambiguous Paths. Chomsky has never been terribly interested in how different structural proposals match up with other cognitive modules. The focus throughout the 60 + year history of generative grammar has been on generalizations over structure-building rules. In fact, arbitrariness from the functional perspective is almost valued in the theory, because it signals the uniqueness of the Language Organ.
And this is the big divide. Recursive operations for many researchers are connected primarily to semantic operations that can take portions of or even collections of sentences in their scope. Many theories have long recognized that phrases can be endo-centric or exo-centric, so I am glad to see that Merge is coming around to reflect what many linguists have known for decades (and what Chomsky himself allowed until Ken Hale’s seminal work on dividing the S-node into IP and TP).
Ternarity is the same. Many linguists recognize it (certainly it is recognized by even some formal phonologists). If Merge comes around to this as well, so much the better for Merge. But as I said in the review of David’s book, the main thing that – to my mind – Merge and the entire approach misses is the disconnect with meaning, seeing meaning as an interface condition instead of, in generative terms, a generative component – the driver of syntax.
So all we see in this is that the Generative Semantics vs. Interpretive Semantics debates still continue, but under new labels, such as formalism vs. functionalism.When I was a committed formalist Tom Givon wrote to me as he was founding FUNKNET and asked me to participate, because he wanted “both sides” represented in discussions on this list. Now that I have done the Churchill thing and crossed over the aisle (in a way – my culture-based functionalism isn’t quite a match), I am glad to see others, such as David continuing to keep linguists of all persuasions talking to one another, via his interventions on this list.
Dan
讨论升级了哦，Hudson老先生出来感谢David和Dan：
On 7 Mar 2016, at 16:18, Richard Hudson <r.hudson@ucl.ac.uk> wrote:
Thanks Dan and David. Yes, I know that Merge is defined in terms of super-nodes, but I agree with Dan that generative linguists have revelled in arbitrariness, because that supports the uniqueness of language. But that’s helped by the assumptions of PSG, which led to RRG and HPSG as well as Minimalism. If it’s hard to think of examples of Merge outside language, because it’s hard to think of non-linguistic analogs of PSG, then maybe that’s a reason for doubting the psychological reality of Merge. And if it’s easy to think of examples like the ones I suggested, which are compatible with a dependency analysis of language, then maybe that’s evidence in favour of dependency structure.
We seem to have a choice between two basic assumptions:
1. that language has PS-based structures that are not parallelled outside language.
2. that language has dependency structures that are easy to match outside language (maybe combined with some parts of PS, such as phrase boundaries). I know that #1 is the only one on the horizon for most linguists, but the arguments for it, in contrast with #2, are very very thin; and as an explanation for language, it wins hands down.
Dick
中间David和Dick有来回讨论几次，这里小编略去，直接给大家上Dan的最新回应，算是一个小结：
Date: 2016-03-08 21:37:33
I agree with a good deal of what Dick and Brian have said on this, but I did want to offer a few more thoughts on the issues.
Simplicity is, as David says, a driving force behind the postulation of Merge, part of a larger program of what Chomsky infelicitously refers to as reducing syntax to “virtual conceptual necessity. I attended many of Chomsky’s lectures at MIT in 1984/85 when he was beginning to talk in detail about economy (simplicity) in syntax and throwing out lots of devices from the syntactician’s toolbox, e.g .indices, multiple movement operations, and so on. Kayne’s influence was enormous at the time, because,arguably, no one else was doing work quite as solely dedicated to understanding formal operations on structure (from the time of his PhD thesis on syntax) Chomsky at the time referred to as perhaps the best thesis ever produced at MIT (that was in response to a reporter asking whether Ross’s dissertation wasn’tin fact the best. Ross was not in Chomsky’s good graces because of Generative Semantics and Ross’s courses at MIT were things like “Relational Grammar” and “Discourse Theory,” which attracted few students in that environment. Ross’sproblem from Chomsky’s perspective was that he was interested in meaning first,form second. And that has never been the generative syntax way.)
The vital question that is raised in much of this discussion is whether syntax is the result of general cognitive principles or specific syntax principles. One common response that formalists make to the idea that general cognition can handle syntax is to point to generative analyses of very technical, highly theory-internal phenomena and say that these do not follow from any general principles, at least not obviously. This was Chomsky’s extremely successful and condescending line of attack on Piaget at their debates in the 70s. Piaget simply didn’t know how to respond to the idea that, for example, Subjacency isa principle of syntax but how could it be a general property of the mind.
I thought at the time that this was a convincing line of argument. A couple of things have made me question that, however, in the intervening years.
1. Theories like RRG can actually derive from discourse principles extraction constraints on displaced constitutents, e.g. subjacency, and Van has even show examples of subjacency in morphology, which RRG can handle.
2. There are many analyses that functionalists have developed that are unavailable to or ignored by formalists – that is, the incommensurability is two-way.
3. Analyses of Merge, though compelling to some, are quite uncompelling to others.In particular they are absolutely unappealing for the most part for people whobegin with the assumption that semantics drives form. They are often compelling, however, for some who believe that form is primary. But there is often no basis for comparison. For example, most syntacticians I know find Larson’s proposal of “little v” an interesting idea about semantics, but not acompelling idea about syntax proper. Minimalists, even those who do not use Larson’s analysis believe that whatever insight he had should be captured structurally, not semantically.
4. This leads to the situation where dialog is interesting, but rarely conclusive (shades of French philosophy). Generalists will argue that Occam’s razor favorsa general cognition approach to syntax and that even if not all facts can be accounted for, they likely will be. Formalists argue that Occam’s Razor doesnot disfavor necessary entities, only unnecessary ones and, they argue, Mergeproduces necessary entities. (Reading David’s accounts of all Merge can do, Iam surprised he hasn’t proposed it as a cure for world hunger. But that revealsmy biases.)
4. The idea that recursion is a special mutation that made human language possible(Berwick and Chomsky 2016, among others), is what I refer to as the “X-men” theory of language – all who speak are X-men, mutants who were able to conquerthe less fortunate Homo neanderthalensis folks. This may be true. Berwick and Chomsky’s discussion of evolutionarytheory is overall very good and we know, via things like lactase persistence,than genetic mutations, even ones influenced by culture as per Dual InheritanceTheory, can occur quickly. So there is nothing a priori implausible about the X-men theory. However, the *evidence* is against it, as I point out in a book to appear next year, How Language Began and in an invited article for the Journal of Neurolinguistics, later this year.
To adegree, of course, theory-internal work is the only way to go. We all work theory-internally in a sense. On the other hand, if the things we work with are exclusively the outputs of our own theory, recognized nearly exclusively byonly other adherents of the theory, then we will generally not be talking to anyone except those who share our assumptions (the general sociolinguistics principle “you talk like who you talk with”) and dialog acrosstheories will not be possible, as I say in my review of David’s book, A Syntax of Substance (http://onlinelibrary.wiley.com/doi/10.1111/aman.12251/abstract;http://ling.auf.net/lingbuzz/002351).
My own conclusion is that meaning-based and form-based analyses will continue for sometime and that linguists/philosophers/psychologists of the future will have to sort them out. There is no meaningful dialog that could result in a Hegeliansynthesis anytime soon. In the meantime, the conversations can be a wee bit helpful in helping each to better understand the other.
Dan
——————————————————————–
论辩暂告一段落。如果你感兴趣，可以电邮联系小编 info@cognitivegrammar.org，获取该邮件讨论组全部内容。另外，点击左下角“阅读原文”可以直达Rice University的 FUNKNET（需要注册）