人物专栏 | Ur Shlonsky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》（2020年第70期，总第134期）“人物专栏”与大家分享本站采编人员王竹叶对Ur Shlonsky教授进行采访的访谈录。Ur Shlonsky教授，现任瑞士日内瓦大学语言学系教授，他的研究领域包括句法理论、闪语以及罗曼语句法分析。本期访谈中，Ur Shlonsky教授向我们介绍了系表结构研究需要注意的方面、普遍语法的初始语序、牙买加克里奥语习得的研究对句法的贡献，同时，对于疑问词why的分布及其特殊性作出了解释，最后介绍了希伯来语空主语的分类，对我们非常具有启发意义。

人物简介

Ur Shlonsky教授

Ur Shlonsky，瑞士日内瓦大学语言学系教授，主要研究领域为句法理论与句法对照分析。他做了很多闪语（特别是现代希伯来语）句法方面的研究，例如小句结构、空主语、名词结构的形式与句法、关系从句以及复指代词。在罗曼语与罗曼语方言学方面，Ur Shlonsky教授也做了许多附缀化、wh移位与重构方面的研究。近期，他对wh原位现象颇感兴趣，并且正在做关于法语wh原位现象的项目。

Brief Introduction

Ur Shlonsky is professor of the Department of Linguistics in University of Geneva (Switzerland). His main area of interest is syntactic theory and comparative syntax. He has studied various aspects of the syntax of Semitic languages, particularly of Modern Hebrew, e.g., clause structure, null subjects, the form and syntax of nominal expressions, relative clauses and resumptive pronouns. He has also worked on Romance languages and Romance dialectology and has written on topics such as cliticization, wh-movement and restructuring. More recently, he has become interested in wh-in-situ phenomena and am currently involved in a research project on French wh-in-situ.

访谈内容

01.

王竹叶：以系表结构的比较研究为例，有哪些方面是我们在研究中要特别注意的呢？

Ur Shlonsky教授：句法现象的比较研究无疑是生成语法领域中最成功的研究策略之一。比较研究法使我们能够更好地描述构成跨语言句法差异的参数，并且“比较研究法试图利用这些差异作为一种新的证据来源，以普遍语法（UG）原理的表征和划分为基础，通过掌握人类语言机制的句法成分来理解所有人类语言的属性”（Kayne，2005）。

系表结构的研究也是如此。我想到了几个跨语言比较的问题。第一个问题源于系表结构的命名。什么是系动词？或者更具体地说它的语义或语法贡献是什么？例如这个句子：Bao is sad today，sad是Bao的谓词，动词is在这个句子里似乎没有任何语义作用。实际上，很多语言不使用任何系动词来表达（‘sad(Bao)’）这个命题。因此，一个相当有趣的话题就是系动词的跨语言分布问题。在有些语言中，谓语和主语简单地组合在一起组成命题，而另外一些语言如希伯来语、波兰语等的主语和谓语之间可以插入代词（人称代词或非人称代词）。在某些语言中，系动词be、代词或者空位的选择取决于命题的时态、谓词的类别或者谓词指的是恒常属性（恒常性谓语）还是阶段属性（瞬时性谓语）。我认为系词句的形式也取决于命题究竟是非主题判断句还是主题判断句。我认为在希伯来语中也是如此，考察其在不同语言中的表现也将会是非常有趣的研究。

第二个问题是经典的系动词句子和非经典系动词句子间的差别，例如句子Bao is my best friend和My best friend is Bao之间的差异。尽管这两个句子功能相同，但信息结构却截然不同。具体来说，倒装句中的系表名词短语解释为焦点。问题是为什么会这样？我们可以思考倒装句是否涉及名词性谓语前置于主语，倒装是否涉及更复杂的推导过程。这两个句子的跨语言比较对解答这一问题很有必要。

02.

王竹叶：我们知道世界语言的语序类型主要有SOV，SVO，VSO，VOS等几种，那么在您看来UG的初始语序应该是怎样的呢？

Ur Shlonsky教授：S、V和O三种句法成分在数学上可能出现六种排列方式，即SVO、SOV、VSO、VOS、OSV、OVS。假设主语、动词和宾语的线性排列顺序不受其他因素制约的话，我们可以预测，以上六种语序在世界语言中的占比都应该大致为16.6%。但事实并非如此：语言类型学研究表明，前两种语序在世界语言中具有压倒性的优势，占据了大约80%。这并不是一种偶然。此外，我们还知道在许多语言中，主语、动词和宾语的顺序也在某种程度上预示着其他成分的排序偏好（Greenberg的普遍蕴含性）。这两个事实都为“语序受抽象句法原则制约”这一观点提供了支持，正因如此，语序方面的相关研究是理论句法研究的首要课题。

“核心方向性参数”是在二十世纪八十年代提出的，用以解释Greenberg等人发现的和谐趋势（例如，日语是核心居后的语言，爱尔兰语是核心居前的语言）。这就是说，正是不同的参数决定了一种语言究竟是右分支结构，呈现C [TP T [AuxP AUX [VP V DP]]]这样的语序；还是左分支结构，呈现[[DP V VP] AUX AuxP] T TP] C这样的语序。从这看来，核心居左的语言与核心居右的语言的语序排列似乎呈现一种镜像的关系。然而，事实证明，几乎在所有类型的句法成分（TP，CP或DP）中，核心居前的语言都比核心居后的语言具有更明显的变化趋势。例如，Greenberg研究发现，以名词短语为例，名词核心居前的语言其后置的修饰语（形容词、数字和指示词）的顺序更灵活。相比之下，名词核心居后的语言其前置的修饰语则要严格地遵循一定的排列顺序——指示词、数字、形容词、名词。而这种不对称现象对语序的“镜像”观点来说绝对是一个有力的冲击。

Kayne的线性对应公理（CS）认为，句子的层级结构也可以映射到其线性顺序上，具体来说，句子的指定语统制其后的动词，而动词则统制宾语。因此形成SVO语序，导致了在词项性质特征的微观变化中的词序变化的重述。 从“指示语-核心-补足语”这种语序出发，词汇项（通常是功能词汇的核心）决定它们的补足语是否会发生移位，以及如果发生移位，移动的成分究竟是整个短语还是该短语的核心。Cinque提出用这种方法来改变名词及其修饰语的顺序。无论是从宏观参数（即管辖核心或分支方向），还是从微观参数（即管辖移位选项和位置特征）来看，这种移位现象都已经得到了非常丰富且细致的研究。

有人则指出这种方法存在两个问题，一是其派生的复杂性；二是其中的一些派生步骤缺乏可靠的动因。很显然第一个问题并不存在，因为所谓的“复杂性”是一个依靠直觉的概念，是基于先验的判断。诚然，在评价派生时，将复杂性考虑进去是合情合理的。但我们应该明确的是首先必须界定“复杂性”这一概念本身，而后才能衡量派生的复杂程度。在对派生的评估中当然有复杂性的考虑，但首先必须对概念本身进行定义，以便确定复杂性的尺度。我们没有理由认为句法计算比其他自然现象更复杂，因为自然科学一再表明，生物本身就是一种高度复杂的现象。一个动词之所以可以和一个时态语素进行合并，要么是因为动词上升到T节点发生了“核心移位”，要么是因为动词短语清空了其词汇义内容，动词和动词短语（清除词汇义后的剩余部分）与T节点合并发生了“短语移位”。这个问题有之前的研究作为支撑，因为这两种操作在UG中都适用。

在我看来，第二个问题确实有待解决，举例来说，在得到OV语序的过程中直接宾语究竟移到了动词之前的什么位置以及这个移位的动机究竟是什么，关于这个问题我们确实需要给出合理的解释。因此，可以说，语序变异的研究方法不仅为我们后续的研究开辟了新的视野，同时也引导我们提出新奇且令人兴奋的新问题。

03.

王竹叶：基于您进行的牙买加克里奥语习得的研究，您能简要说明该研究对句法研究的贡献吗？

Ur Shlonsky教授：在和Amirand de Lisser, Stephanie Durrleman和Luigi Rizzi的共同努力下，我们发现说牙买加克里奥语的孩子与使用其它语言的孩子一样，在习得句法时经历了相同的发展过程。根据Rizzi九十年代的研究发现，早期的牙买加克里奥语根空主语、根不定式和截词这些现象，可以解释de Lisser的语料库数据。同时还研究了时态-语气-体标记（TMA）的发展。牙买加克里奥语是一种分析型语言，没有曲折形态，因此TMA标记是句法的中心。随着孩子的成长，最初的截断假设也在逐渐发展。

04.

王竹叶：why应该放在哪里，为什么why如此特殊？

Ur Shlonsky教授：Why与其他的附加疑问词不同，因为它可以与否定句连用，比如说why didn’t he do this? ，这个句子是合法的，但 *how didn’t he do this? 这个句子就是不合法的。此外，why也可以与前置（在why右边的）焦点兼容。这些语言事实促使Rizzi主张在左缘结构中，why 被外部合并到Spec/IntP的位置上去。在多年前的一篇论文中，Gabriela Soare和我都认为这些语言事实证明why在小句中的高位（高于否定）是外部合并形成的，然后移位到Spec/IntP的位置。我们认为，在罗马尼亚语这种多个wh成分前置的语言中，why依然处于原位：why必须出现在所有前置的wh成分之下（因为越过wh成分移位到指示语位置违反了相对最简性原则）。如果why在它获得语义解释的位置是外显的，那么为了使在why do you think he did this?中why的长距离识解得到解释，why的位置在原则上是固定的。how come这个短语不能得到长距离解释，但是可以出现在Spec/IntP位置，why就不可以。我与Caterina Bonan的共同研究（见于其2019年日内瓦大学博士论文）表明，一些语言既有高位的why，也有低位的why。

05.

王竹叶：自1981年乔姆斯基提出“空语类”这一概念以来，国内外学者对其进行了深入的研究。各种语言中空语类的研究也在持续不断的争论中取得了不少成果。您能具体谈谈现代希伯来语中空主语的分类吗？

Ur Shlonsky教授：在意大利语一类的语言中，空主语的丰富程度与动词的形态密切相关，而在汉语一类的语言中，空代词则与话语所指直接相关，与这两类语言不同的是，希伯来语的空主语无论是在形式许可方面还是在解释选择方面，都更为复杂。

希伯来语有三种时态上的形态变化：在过去时与将来时中，动词会有人称性数的曲折变化，而现在时中，动词只有性数的曲折变化。指称性空主语代词（与虚义代词及准指称性代词相反）只会在过去时与将来时中出现。允许空主语出现的关键特征是人称，然而人称只出现在过去时和将来时，不会出现在现在时中。可是，这并不能解释另一个分布事实——pro只能是第一或第二人称。

我相信有不同的方式来陈述这些语言事实。我说过，事实上在希伯来语中并没有第一和第二人称的空主语，并且在过去和将来时中表示一二人称性数配合的词缀蕴含了功能性核心言语行为参与者，即一二人称代词的所指成分。言语行为参与者及其对应的代词都作为动词的词缀出现。在希伯来语中，并不存在一个可以指代言语行为参与者中第三人称的成分，所以仅有第三人称空主语。综上所述，第三人称空主语作为指称代词是不合法的，我把这种现象归因于此类代词在将来时人称中的缺席。非指称性的第三人称空主语有可能是合法的，因为对于它们所指内容的本质来说，人称特征并不是必须的。

当谈到内嵌句的第三人称空主语代词时，情况变得更加复杂。在这里我们了解到，当指称空主语被主句的主语成分统治（同指）时，指称空主语才是合法的。最后，当我们回顾有关这一主题所做的相关研究时，我发现这一方面的“图景”比过去更为清晰了，然而现有的相关研究分析，包括我自己的研究在内，依旧不那么尽如人意。

English Version

01.

Zhuye Wang: Take the comparative study on copular constructions as an example, what aspects should we pay special attention to?

Prof. Ur Shlonsky:The comparative approach to syntactic phenomena is undoubtedly one of the most successful research strategies in the field of generative syntax. It allows us to better characterize the parameters that underlie cross-linguistic differences in syntax and “it attempts to exploit those differences as a new … source of evidence bearing on the characterization and delineation of the principles of Universal Grammar (UG), of the properties that, by virtue of holding of the syntactic component of the human language faculty, will be found to hold of every human language” (Kayne 2005).

The study of copular constructions is no exception. Several questions of cross-linguistic comparison come to mind. The first question stems from the name given to this construction. What is a copula, and, more specifically, what is its semantic or grammatical contribution? Sad is predicated of Bao in the sentence Bao is sad today and the verb is does not seem to be doing any semantic work here. Indeed, a great many languages express the proposition ‘sad(Bao)’ without any copular verb. So, one issue of substantial comparative interest is the distribution of copulas across languages. In some languages, the predicate and its subject can be simply strung together to form a proposition. In others, a pronoun (personal or impersonal) appears between the subject and the predicate (Hebrew, Polish and many others). In some languages, the choice between using a copula (the verb be), a pronoun or nothing depends on the tense of the proposition, the category of the predicate or whether the property that the predicate refers to is permanent property (individual-level) or a temporary one (stage-level). My feeling is that the form of copular sentences also depends on whether the proposition is a thetic or a categorical judgement. I believe this is the case in Hebrew and it would be interesting to see what happens in other languages.

A second set of issues relates to the differences between so-called canonical (or subject-initial) copular sentences like Bao is my best friend and inverted ones, like My best friend is Bao. Although the two are truth-functionally equivalent, their information structure is quite different. Specifically, the post-copular Noun Phrase in the inverted sentence is interpreted as focus. The question is why this should be the case. One should also try to understand whether the inverted sentence involves fronting of the predicate nominal to subject position over the subject Noun Phrase or whether inversion involves a more complex derivation. Crosslinguistic comparison of the two would be extremely useful in answering this question.

02.

Zhuye Wang: We know that the word order can be SOV, SVO, VSO, VOS and so on. What should the initial word order of UG be like in your mind?

Prof. Ur Shlonsky: There are six mathematically possible permutations of S, V and O, namely SVO, SOV, VSO, VOS, OSV, OVS. If the linearization of subject, verb and object were unconstrained, we would expect, given a representative sample of languages, that each order would be manifested by about 16.6 % of languages. This is not what we find: Typological research has shown that there is an overwhelming preference for the first two orders, around 80%. This is unlikely to be accident. Moreover, we know that in many languages, the order of subject, verb and object implies ordering preferences of other constituents (Greenberg’s Implicational Universals). Both these facts lend support to the idea that word order is constrained by abstract syntactic principles and it is therefore a prime topic for research in theoretical syntax.

The Head Directionality Parameter was proposed in the 1980’s to account for the harmonic tendencies that Greenberg and others discovered (e.g., the fact that Japanese is head-final in all constituents, while, say, Irish, is head-initial). The idea was that a parametric choice determines whether a language is right-branching, yielding the order C [TP T [AuxP AUX [VP V DP]]] or left-branching and yields [[[DP V VP] AUX AuxP] T TP] C. This way of seeing things implies that ‘left-headedness’ is the mirror image of ‘right-headedness’. It turns out, however, that in just about every constituent (TP, CP, or DP), there is a strong tendency for more variation in word orders in the head-initial schema than in the head-final one. Thus, for example, Greenberg observed that N-initial orders admit more variation in the order of (postnominal) modifiers – adjectives numerals and demonstratives, than N-final ones, where the order is much more rigid – Demonstrative, Numeral, Adjective Noun. This asymmetry is completely unpredicted in a ‘mirror-image’ view.

Kayne’s Linear Correspondence Axiom (CS) which states that hierarchical structure maps onto linear order: Specifiers c-command verbs which c-command their objects, hence SVO, has led to a restatement of word order variation in terms of micro variation in properties of features of lexical items. Starting from a specifier-head-complement order, lexical items (typically heads from the functional lexicon) determine whether their complement moves and if it does whether the whole phrase moves or only its head. Cinque has developed this approach for variation in the order of the noun and its modifiers. This shift of emphasis, from macroparameters governing headedness or the direction of branching to microparameters governing movement options and the characterization of the positions targeted by such movement, has engendered very rich and empirically detailed research.

Critics of this approach point to two problems, namely, the complexity of derivations that it engenders and the lack of motivation for some of the derivational steps. The first problem seems to me to be spurious: the notion of ‘complexity’ invoked is an intuitive notion, anchored in an a prioristic judgement. There surely are considerations of complexity in the evaluation of derivations, but the notion itself must first be defined in order for a scale of complexity to be defined. There is no reason to believe that syntactic computation is any less complex than other natural phenomena and natural science has shown, again and again, that organisms are highly complex phenomena. A verb can appear merged with a Tense morpheme either because V has raised and adjoined to T (head movement) or because VP has been emptied of all lexical material but the verb and the (remnant) VP merges with T (phrasal movement). The question here is empirical since both operations are available in UG.

The second problem, illustrated, for example, by the identification of the landing site of a direct object moved above V to render the order OV and the motivation for this movement, seem to me to be real empirical questions that we need to find answers for. So, one could say that the approach to word order variation in terms of the LCS opens up new horizons for research, leading us to ask novel and exciting questions.

03.

Zhuye Wang:  As you are doing the research project on the acquisition of Jamaican creole, could you please tell us briefly in which respective does it contribute to syntax?     

Prof. Ur Shlonsky: In joint work with Tamirand de Lisser, Stephanie Durrleman and Luigi Rizzi (emanating from de Lisser’s Geneva PhD dissertation), we have tried to show that child speakers of Jamaican Creole go through the same developmental stages in the acquisition of syntax as children speaking other languages. We showed that early Jamaican Creole has root null subjects, root infinitives and that the truncation approach to these phenomena, elaborated by Rizzi in work from the Nineties, can account for the corpus data elicited by de Lisser. We have also studied the development of tense-mood-aspect TMA markers. Jamaican Creole is an analytic language and has no inflectional morphology. The role of TMA markers is therefore central to its syntax. We have shown that these develop gradually, as the child’s initial truncation hypothesis erodes as she or he matures.

04.

Zhuye Wang: Where’s ‘why’ and why is ‘why’ so special?

Prof. Ur Shlonsky: Why is different from other wh adjuncts in that it is compatible with negation, compare why didn’t he do this? and *how didn’t he do this? Why is also compatible with fronted focus (to its right). These facts led Rizzi to argue that why is externally-merged in the left periphery, in Spec/IntP. In a paper from several years ago, Gabriela Soare and I argued that these facts are compatible with the view that why is externally-merged in a high position in the clause (higher than negation) and then moved to Spec/IntP. We argued that in Romanian, a multiple wh fronting language, why remains in situ: It must appear below all fronted wh elements (because movement over them to Spec/IntP would violate Relativized Minimality). We also argued that if why were externally-merged in the position in which it is interpreted, then it would be criterially-frozen in that position so that the long construal of why in e.g., why do you think he did this? could not be explained. The expression how come, which cannot be long-construed is merged in Spec/IntP, but not why. In recent work with Caterina Bonan (see also her Geneva 2019 PhD dissertation), we show that some languages have both a ‘high’ why and a ‘low’ one.

05.

Zhuye Wang: Since Chomsky put forward the concept of empty category in 1981, scholars at home and abroad have conducted in-depth studies on it. The study of empty category in various languages has also made some achievements in the ongoing debate. Could you talk about the classification of null subjects in modern Hebrew specifically?

Prof. Ur Shlonsky: Unlike languages like Italian, in which null subjects are tightly linked to (the richness of) verbal morphology, or languages like Chinese, in which null pronouns are directly bound to discourse referents, Hebrew null subjects manifest a complex picture, both in terms of their formal licensing and in terms of their interpretive options.

Hebrew has three morphological tenses: Verbs in past or future are inflected for person, number and gender while verbs in the present tense are only inflected for number and gender. Referential null subject pronouns (as opposed to expletive and quasi-referential ones) are only possible in the past and future tenses. It appears that the crucial feature licensing a null subject is [person] – absent in the present tense and present in past and future. This, however, does not account for another distributional fact, namely, that pro can only be first and second person. 

I believe that there is a slightly different way of stating these facts. I have argued that there are actually no first and second person null subjects in Hebrew and that the agreement affixes in the past and future first and second person encode the functional head Speech Act Participant (SAP) to which the first and second person pronouns cliticize. SAP and its incorporated pronouns appear as affixes on the verb. There is no equivalent to SAP for third person and so the only null subject in Hebrew is third person. As stated above, third person null subjects are not licit as referential pronouns, a fact which I attribute to the absence of a person feature on such pronouns. Third person null subjects are possible when they are non-referential, and this is because a person feature is not necessary to license them or identify their content.

The picture becomes even more complex when we look at third person null subjects in embedded contexts. Here we see that referential null subjects are licit when they are c-commanded (and hence co-referential) with the subject of the matrix clause. When I look at the work that has been done on this topic, I see that the empirical picture is now clearer than it was in the past, but the analyses, including my own, are still unsatisfactory.

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：马晓彤 王竹叶 訾姝瑶 闫玉萌 陈金玉 王平

排版：马晓彤 闫玉萌 高洁

审校：王丽媛 李芳芳