人物专栏 | Guglielmo Cinque教授访谈

点击上方蓝字关注我们

编者按

《理论语言学五道口站》（2022年第09期，总第212期）“人物专栏”与大家分享近期Isogloss期刊对Guglielmo Cinque教授的访谈，由本站成员赵欣宇、雷晨、聂简荻、郭思源、丁子意进行翻译。Guglielmo Cinque教授，意大利威尼斯大学教授，生成语法的领军人物之一，句法制图的奠基人之一。其研究领域主要为生成语法和语言类型学。

本期访谈中，Guglielmo Cinque教授首先就语言变体研究进行了多角度分析，其次就语言的普遍性以及多样性的问题给出了见解，最后对参数理论和语言变体研究将来可能面临的问题进行了阐述。

人物简介

Guglielmo Cinque教授

Guglielmo Cinque，意大利威尼斯大学语言学教授，现代最简句法领军人物之一，美国语言学会荣誉会员。Cinque教授的研究领域为生成语法和语言类型学。他与Luigi Rizzi教授是句法制图的奠基人，致力于探究语言的精细结构。

Brief Introduction

Guglielmo Cinque is a Professor of Linguistics at Ca' Foscari University of Venice. He is one of the leading figures in modern minimalist syntax. He is also an honorary member of the Linguistic Society of America. Cinque works in the fields of generative grammar and language typology. Together with Luigi Rizzi he belongs to the founding figures of cartographic syntax, a research area devoted to the fine structure of languages.

访谈内容

01.

Isogloss：在您看来，哪个更接近抽象的语言官能？“语言”、“方言”、“个人方言”还是其他？

Guglielmo Cinque教授：语言官能（或普遍语法（UG）），就是诺姆·乔姆斯基提到的I-语言（乔姆斯基在1986以后将“I”定义为“内在的、个体的、内涵的”），这一概念与“个人方言”一致，或者称之为“个体语言习惯”。我们所要做的，就是尝试从理性客观的视角重塑我们的习语系统，并将其与他人的习语进行比较。理想情况下，当“个体语言习惯”共享大部分属性时（可能不会出现共享所有属性），就是“方言”和“语言”。与“语言”相比，“方言”与“个体语言习惯”具有更多共性。但并没有原则性的依据可以区分“方言”和“语言”（这与Max Weinreich著名定义“有了陆军和海军，语言就是一种方言”相呼应）。“不同语言”之间的比较就是个体语言习惯集群之间的比较，例如意大利语和帕多瓦语、意大利语和法语、英语和日语。在严格的变体差异限制下，共享的属性少之又少。

02.

Isogloss：研究语言变体的主要优势或理由有哪些呢？

Guglielmo Cinque教授：语言变体的研究是重建UG的重要部分，这不仅是为了更精确地了解什么可以变、什么不可以变（即UG所允许的变体限制）。而且正如Richard Kayne的研究所强调的，“将这些差异视为新的、激动人心的证据来源，以描述和界定UG的原则”（Kayne 2005: 3）。在最简方案（Chomsky 1995）以后，我们面临的挑战是：是否可以把所有或大部分的变体都可以归结为不同的移位选择，还是归结为词库中功能成分发音/不发音的差异（参见Rizzi 2009, 2011），这个问题值得深究。

03.

Isogloss：您如何看待语言中“差异”与“共性”之间的联系与张力？

Guglielmo Cinque教授：我仍然认为70年代末出现的原则和参数理论是解决语言“差异”与“共性”张力的最佳模式，虽然目前来看，构成UG的原则与参数理论可能已经发生了变化，这一点我会在下文中简要谈及。

04.

Isogloss：在您看来，传统和现代方言学对语言研究的贡献有哪些？

Guglielmo Cinque教授：正如Benincà（1994）和Kayne（2005：8）所指出，方言学作为语言变体间的比较研究，它在比较句法中最接近对照实验。（如研究生物学变体所做的对照实验一样，只有对研究对象进行剖析，才能得到真实的实验结果）。

05.

Isogloss：在您看来，哪些方法可以为语言变体研究提供可靠依据？就语感判断、语料收集、实验研究、学科交叉而言，哪种方法更理想？

Guglielmo Cinque教授：原则上，我们接受任何有助于深入理解语言的依据。但我认为，有一种依据来源可靠且地位特殊，那就是语言学家对其母语作出的判断。当然，即便如此，它也不能避免错误和保证完全准确，但这种做法总体来看最理想。虽然对于某些情况，语料来源并不可靠，比如对非母语的语法现象进行描述和分析时，我们需要谨慎对待所得出的研究结论，并尽可能地控制其他可能性因素。尽管如此，至少在语法的特定领域中，该依据也可以发挥作用（如Joseph Greenberg所作）。

06.

Isogloss：目前，很多理论研究都会辅以语料库及统计/实验分析。事实上，方言学研究中也采用了实验方法和田野调查方法。您认为在此过程中，语言理论方法发挥怎样的作用?

Guglielmo Cinque教授：比较句法中的传统方言学及理论方言学的最新进展表明，理论方言学的语料多来源于传统方言学，如问卷调查法和田野调查法（参见Benincà 1994和Kayne 2001年的讨论)。最近，方言的实证/数据研究也为某些句法现象的解释提供了新思路，从Jeroen van Craenenbroeck (2014) 对荷兰方言中动词群的可能变体研究中可见一斑。

07.

Isogloss: 您一直致力于语言普遍性的探索，如，从层级性（如您1999年的专著）到范畴研究。您认为这些现象都是相同意义上的“普遍性”吗？有哪种普遍性在认知上（或语法上）有优先地位吗？

Guglielmo Cinque教授：我们可以采取许多不同的形式来探究语言的普遍性，我不认为有任何一种普遍性在认知上或语法上存在优先地位。至于乔姆斯基（1965：29）对形式普遍性和内容普遍性的区分，我更关心第二种，产生这种倾向是由于普遍语法（UG）仍然存在许多值得我们探索的部分。我自己的感觉是，最近对普遍语法（UG）内容穷追猛打的趋势低估了语法在许多方面的领域特定性。令人纠结的问题是：为什么我们只能在一种又一种语言中发现某些特定的语法区别（Cinque 2013a），但有时（如数量结构）它们编码的方式并不明显（参见Kayne 2003和Cinque 2006），其身份往往是隐藏的。

可以肯定的是，迄今为止所提出的层级性和目前被认为是普遍存在的范畴是需要修正的近似值。为了正确的概括并说明已证实的跨语言的差异，适当水平的抽象需要更细致（这意味着现在假设的成分将不得不被进一步分解）。

08.

Isogloss：最近的一些研究认为，正是多样性刻画了人类的语言，这往往意味着语言的普遍性是错误的（或者一些所谓的特定特征，如递归，并不存在于所有语言中）。这种情况是由于I-语言/E-语言的区别尚未被理解吗？还是其他原因造成的？

Guglielmo Cinque教授：在许多领域中，我们倾向于关注一种类型的标记与另一种类型的标记的区别，而不是关注所有标记的共同点。众所周知，当我们用面部识别能力时，我们更容易注意到一张脸与另一张脸的区别（具有独特的社会优势），而不是它们的共性，在语言上也是如此。最早对语言的思考（除了一些特例，所有后续的理论研究，在上世纪下半叶的认知革命之前）都集中在不同语言的存在或“同一”语言的不同阶段，而不是语言的共同点。这种倾向一直延续到了今天，就好像有人认为“多样性是人类的基本特征（因为毕竟世界上不存在两个完全相同的人）”，但在某种程度上，这种看法却容易让人们忽略人类这一物种共同的基础特征。同样的道理也适用于语言。“递归性”可能只是人类语言某一领域所特有的属性（详见Hauser, Chomsky and Fitch 2002, and Legate, Pesetsky and Yang’s 2014 recent response to Levinson 2013），通过对不同语言的探究，人们发现很多特性也是某一领域所特有的。例如，据说每一种语言的名词短语都包含指示词、量词、数词、指小词以及特殊功能的形容词，如“other”。如果有着微小的形态差异，则取决于所讨论的语言本身，如形态格、一致性或其他形态特征的差异。这些范畴是否是某一领域所特有的属性还未有定论，但却已经被我们先前提到的观测所支持，即对名词短语语法编码过程中所存有的认知差异的研究，人们总是会得到少数相同的结论(详见Cinque 2013a)。对此持有反对意见的人认为这反映的并不是语言能力，其实无关紧要。因为这即使反映人类更普遍的认知特性，也无法排除一领域的特性，这可能是普遍语法（UG）内容普遍性的具体表现。

09.

Isogloss：在生成语法领域，对原则和参数理论框架下的“个性”与“共性”的研究，已经取得了丰硕的成果，但是这种研究无论是在理论层面还是在实验层面都受到很多批判。那在您看来，当今“参数理论”处于什么样的地位呢？

Guglielmo Cinque教授：我认为我们的研究离不开参数这一概念，但确实有很多问题还有待解决，如：什么是可能的参数？真的有参数化的原则吗？参数的初始位置在哪里（在狭义句法层面？在词汇层面？在句法-语音接口？还是在普遍语法之外？）。不同语言间的差异可能不是受到某一个参数的影响，而是多个参数之间相互作用的结果。例如最初的空主语（Null Subject）参数(Rizzi 1982)，如今已经被细化为多个更精确的成分(详见Nicolis 2008, also referring to Rizzi’s subsequent work)。普遍语法（UG）中一些原则，如合并（Merge）、线性对应原则（the Linear Correspondence Axiom (LCA)）,准据冻结效应（Criterial Freezing）等，可能并没有参数化，而且小句和短语的功能层级可能也并没有参数化。尽管很多（或者大部分）参数是直接由词项进行解码的（Kayne 2005, Rizzi 2009, 2011），但我认为有一部分参数仍然属于狭义句法。以传统的“语序参数”为例，根据线性对应原则，线性顺序是层级结构的函数。如果“中心语前置”语言与“中心语后置”语言中包含源于不同移位的非对称性层级结构，那么一部分参数将涉及到（内部）合并在狭义句法中的运算方式。这类参数不能完全看作是PF接口层面的外化，同时，它也不只是众多词汇编码指令综合作用的结果。有些现象出现在非一致性案例中似乎是合理的，比如一个（功能性或内容性）词项不符合“中心语前置”或“中心语后置”形式，但在某些更具一致性的案例中却是不合理的。

10.

Isogloss: 未来的几十年中，在语言及其变体研究方面，您认为我们必将面临哪些挑战？

Guglielmo Cinque教授：在我看来，有两个问题亟待解决并有望在未来让我们能更加深入地了解，一个是语序（如何分析跨语言中已证实和未证实的语序变化），另一个是词汇（“可能词”的概念以及如何处理不同语言中明显的词汇差异）。

关于第一个问题，尽管我们清楚在语言中存在着决定小句和短语成分顺序的潜在规则（详见the “head-initial” and “head-final” organization of linear order – cf. Cinque 2013b, Chapter 1），但其中存在的差异性和不一致性引起了很多困惑。更加困难的是，我们还发现了很多不同的规范语序，这些语序（或基于这些语序的层级性差异）看起来像“噪音”一样杂乱无章，同时它们没有任何语义和语用意义（为什么有些“中心语前置”语言说“he can swim well”与有些“中心语后置”语言说“he well swim can”表达的意思相同？）。

关于词汇方面，传统观点认为词汇实际上是一个藏有丰富结构和系统的特殊数据库。我认为，Kayne（2008，2012）提出了一种非常有效的句法方法，可以用来解决语言之间明显的词汇差异或者两种语言之间的类似于偶然空缺的差异。（Cinque的文章中也讨论了相关问题）。

English Version

01.

Isogloss: From your perspective, what are the relevant levels of abstractness to approach the faculty of language? The standard ones (namely “language,” “dialect,” and “idiolect”)? Others?

Prof. Guglielmo Cinque: The relevant level of abstractness to approach the Faculty of Language (or Universal Grammar (UG)), is Noam Chomsky’s notion of I-language (“I” for internal, individual, and intensional – Chomsky 1986 and subsequent work), which ultimately coincides with idiolect, or rather I-idiolect. In the best cases what we do is study and try to rationally reconstruct the system underlying our own idiolect, and compare it with the idiolects of others. “Dialect” and “language” are convenient idealizations for clusters of I-idiolects that happen to share most properties (perhaps never all properties). “Dialect” is often used to refer to a cluster of I-idiolects that share more common properties than those characterizing a “language”, but there is really no principled reason, other than that, to distinguish “dialect” from “language” (in syntony with Max Weinreich’s famous definition “A language is a dialect with an army and navy”). Comparisons between “different languages”, say Italian and Paduan, or Italian and French, or English and Japanese involve comparisons between clusters of I-idiolects that share fewer (and fewer) common properties (within severe limits to variation).

02.

Isogloss: What are the main advantages / reasons to study linguistic variation?

Prof. Guglielmo Cinque: The study of linguistic variation is an essential part of the effort to reconstruct UG, not only to have a more precise idea of what can vary and what cannot (i.e., the limits of the variation allowed by UG), but also, as Richard Kayne has shown in his work and has stressed explicitly, “to exploit those differences as a new and often exciting source of evidence bearing on the characterization and delineation of the principles of UG” (Kayne 2005: 3). Within a Minimalist perspective (Chomsky 1995 and subsequent work), one of the challenges we are confronted with is whether all or most variation can be reduced to different movement options and to differences in pronunciation/non-pronunciation of functional elements encoded in the lexicon (cf. Rizzi 2009, 2011), a question I come back to.

03.

Isogloss: How do you conceive the relation / tension between linguistic variation and linguistic uniformity throughout the years?

Prof. Guglielmo Cinque: I still think the Principles and Parameters approach that emerged in the late ‘70’s is the best model to resolve the tension between linguistic variation and linguistic uniformity, even if our current conception of the principles and especially of the parameters that make up UG may have changed; a point I briefly return below.

04.

Isogloss: In your opinion, what are the contributions of dialectology (both traditional and present-day studies) to the study of language?

Prof. Guglielmo Cinque: As Benincà (1994) and Kayne (2005: 8) have pointed out, the important contribution of dialectology, understood as a comparative investigation of closely related language varieties, is the fact that it is the closest we can get to a controlled experiment in comparative syntax (much as variation in the biological world comes close to constitute a controlled experiment in a domain where real experiments cannot be carried out without destroying the objects to be studied).

05.

Isogloss: What are the relevant sources to obtain evidence to study language and its variation (speakers’ own competence, corpora, experiments, non- linguistic disciplines, etc.)? Is any of them potentially more relevant than the others?

Prof. Guglielmo Cinque: While there should in principle be no preclusion to the kind of evidence one can use to deepen our understanding of language, in my mind one type of evidence has a privileged status in representing a more reliable source of data: the judgments a linguist can produce concerning his or her own native language. Of course, even this gives no protection from errors nor guarantee of success, but is perhaps the best approximation to the ideal experimental setting. In many cases we have to make do with less reliable sources, such as grammatical descriptions or analyses of single topics of languages that are not native to the linguists describing or analyzing them, so that any conclusion based on these must be treated with caution and subjected to further control whenever possible. In spite of that, even this kind of sources may offer insights, at least in certain areas of grammar (as the tradition stemming from Joseph Greenberg’s work has shown).

06.

Isogloss: Much current theoretical research is complemented with corpora and statistical / experimental analyses. In fact, dialectology also resorts to experimental and field work methods, traditionally. What do you think is the position of theoretical approaches to language in such scenario?

Prof. Guglielmo Cinque:Theoretical approaches to language have much to gain from traditional methods of dialect data gathering, like questionaires and field work by specialists, as the recent advances in comparative syntax within both traditional and theoretically informed dialectology have demonstrated (see again Benincà 1994 and the remarks in Kayne 2001). More recently, experimental/statistical analyses of dialect data are also yielding new and significant insights for the theoretical treatment of specific syntactic phenomena, like the possible variation of verb clusters in Dutch dialects recently studied by Jeroen van Craenenbroeck (2014).

07.

Isogloss: Part of your own work has been devoted to the study of universals. Different phenomena fall within such label, from hierarchies (as in your 1999 monograph) to categories. Are all of these phenomena "universals" in the same sense? Is there any type of universal that has cognitive (or grammatical) priority over the others?

Prof. Guglielmo Cinque: The search for universals of language may take many different forms and I do not think there is any that has cognitive or grammatical priority over others. With respect to Chomsky’s (1965: 29) distinction between formal and substantive universals, I have myself been more concerned with the second type, an inclination induced in part by the conviction that we still have to uncover many building blocks of UG. My own feeling is that the recent trend to impoverish the content of UG underestimates the domain-specificity of many aspects of grammar. The nagging question is: why do we find only certain distinctions, and the same ones, encoded grammatically in language after language (Cinque 2013a), sometime (as with numeral classifiers) in ways that are not immediately obvious (cf. Kayne 2003 and Cinque 2006)? Identity is often hidden.

It is certain that the hierarchies that have so far been proposed and the categories that are currently taken to be universal are approximations in need of amendment. Most probably the appropriate level of abstraction to capture the correct generalizations and to account for the attested cross-linguistic variation is more fine-grained (which means that the elements now being assumed will have to be further decomposed).

08.

Isogloss: Some recent studies argue that it is diversity what truly characterizes human language, often implying that the universal nature of language is wrong (or that some allegedly specific traits, such as recursion, is not present in all languages). Is this scenario a residue of the fact that the I-language / E-language distinction has not been understood? Is it something else?

Prof. Guglielmo Cinque: In many domains we tend to concentrate on what differentiates one token of a type from another rather than on what all of the tokens have in common. To take a well-known case, when we use our face recognition capacity, we are more prone to note what differentiates one face from another (with distinctive social advantages) rather than what is common to each. Much the same has happened (and is still happening) with language. The earliest reflections on language (and, with some notable exceptions, all subsequent theorizing until the cognitive revolution of the second half of last century) concentrated on the existence of different languages or on different stages of the “same” language, rather than on what languages have in common. This bias persists to the present day (in the work you are alluding to). It’s as if one were to say that it is “diversity what truly characterizes humans” (after all there are no two identical humans), but this would clearly miss the fundamental identity, at some level, of the individuals of our species. The same appears true of language. Not only specific traits like recursion (see Hauser, Chomsky and Fitch 2002, and Legate, Pesetsky and Yang’s 2014 recent response to Levinson 2013) may be a domain-specific property of language; many more things are found in language after language, which may be domain-specific. It is striking for example that every (reasonably described) language is reported to have in its nominal phrase demonstratives, quantifiers, numerals, diminutives, special functional adjectives like ‘other’, etc. (if there are minor morphological differences these depend on independent properties of the languages in question, like the presence of morphological Case, agreement, or other morphophonological idiosyncracies). The domain-specificity of these categories remains an open question, but is possibly supported by the already mentioned observation that of all the cognitive distinctions that one could expect to be grammatically encoded in a nominal phrase, one keeps finding the same few ones (cf. Cinque 2013a). The objection that they presumably reflect non linguistic capacities is beside the point (why those and not others?). Their being possible reflections of more general cognitive properties does not exclude their being domain-specific, i.e. their belonging to the substantive part of UG.

09.

Isogloss: Within the Generative Enterprise, the research stemming from the Principles and Parameters framework has proven very fruitful to study both variation and uniformity. However, this trend has been subject to much criticism, on both theoretical and empirical grounds. In your opinion, what is the status of “Parameter Theory” nowadays?

Prof. Guglielmo Cinque: My feeling is that there is no escape from parameters, but surely many questions remain open. What is a possible parameter? Are there parameterized principles? Where is the locus of parameters (narrow syntax, the lexicon, the PF interface, outside of UG)? Many differences between languages are possibly the consequence of the interplay of different (relatively small) parameters rather than of single large scale parameters. A possible example of this is the original Null Subject parameter (Rizzi 1982), which has been decomposed into finer-grained components (cf., e.g., Nicolis 2008, also referring to Rizzi’s subsequent work). UG principles (Merge, the Linear Correspondence Axiom (LCA), Criterial Freezing, etc.) are plausibly not parameterized, and neither are, possibly, the functional hierarchies of the clause and of its phrases. While many (perhaps most) parameters may be directly encoded in items of the lexicon (Kayne 2005, Rizzi 2009, 2011), my feeling is that some may still belong to narrow syntax. Take the traditional “word order parameter”. If (as dictated by the LCA) linear order is a function of hierarchical structure and if “head-initial” and “head-final” languages involve non symmetrical hierarchical structure derived by different types of movements, then part of the parameter is how (internal) Merge operates in narrow syntax. The parameter cannot be entirely a matter of externalization (at the PF interface). Equally, it cannot simply be the effect of the accumulation of many lexically encoded instructions. While this is plausible for cases of non consistency, in which a single (functional or substantive) lexical item deviates from the general “head-initial” or “head-final” pattern, the same does not seem plausible for the more consistent cases.

10.

Isogloss: What are the challenges that we will have to address in the following decades when it comes to study language and its variation?

Prof. Guglielmo Cinque: To my mind, two questions that remain to be addressed and hopefully to be better understood in the next future concern word order (how to analyze attested and unattested word order variations across languages) and the lexicon (the notion of “possible word” and how to handle apparent lexical differences among languages).

Concerning the first issue, although one can make out the existence of a great underlying ground plan governing the order of clausal and phrasal elements in the languages of the world (approximately, the “head-initial” and “head-final” organization of linear order – cf. Cinque 2013b, Chapter 1) the degree of differences, and inconsistences, is bewildering. What is even more baffling is the fact that the different canonical word orders that one finds (or the hierarchical differences at their basis) look like “noise”, in the sense that they serve no semantic or pragmatic purpose (why should some “head-initial” language say “he can swim well” and some “head-final” one “he well swim can” to convey the same thought?).

Concerning the lexicon, the traditional idea that it is essentially a repository of idiosyncrasies may in fact hide much more structure and systematicity. In my view, a promising syntactic approach to apparent lexical differences among languages or what look like accidental gaps in some language as opposed to another is the one suggested in Kayne (2008, 2012) (see also Cinque to appear on some related questions).

往期推荐

本文版权归“理论语言学五道口站”所有，转载请联系本平台。

编辑：闫玉萌 赵欣宇 雷晨 

排版：闫玉萌 赵欣宇 雷晨 

审校：李芳芳 田英慧