Romance clitic pronouns appear to the left of the verb in I and to the right of the verb in C. This alternation correlates with: (a) allomorphy, specifically lvs. zero; (b) stress shifts; and (c) reordering of the clitic string. The alternations in (a)–(c) are also observed between non-negative and negative contexts. The key points of our analysis are: (i) the lsegment is associated with definite content; (ii) interpretively, pronouns scope out of modal/non-veridical operators; (iii) syntactically, the exponent for modality/nonveridicality may have the pronoun in its domain; (iv) externalization of the lsegment is found when semantic scope (ii) and syntactic configuration (iii) are mismatched. Therefore allomorphies (including stress), far from being morphophonological quirks, contribute to the externalization of syntactico-semantic notions of non-veridicality. In dealing with clitic (re)ordering we propose a model based on the dissociation between Merge and linear order. Phrasal constituents are ordered to the right of the verb in Romance; clitics mirror them in that they are ordered to the left, while keeping the Merge relations constant. 1. BASIC EVIDENCE AND CURRENT ACCOUNTS In many Romance varieties, the preverbal and postverbal positions of clitics correlate with a number of morphophonological and order alternations. There are three main such alternations, namely: (a) segmental allomorphies; (b) stress shifts; and (c) internal reordering of the clitic string – which are illustrated in sections 1.1–1.3 below. Some segmental alternations affecting Romance clitics are phonologically conditioned. For instance, in the Corsican variety of Zonza, vocalic proclitics, as in (1a), precede only consonantal onsets; vocalic onsets are preceded by an l form, as in (1b). This condition is phonological, involving presumably the resyllabification of the l segment with the following nucleus. 1 The research reported in this article was partially supported by the PRIN 2012 project grant on Long Distance Dependencies. We thank Adam Ledgeway and Diego Pescarini for detailed comments on the first version of this work. We use a broad transcription for the dialectal data. Word stress is indicated selectively, namely when it occurs on the antepenultimate syllable or on the last syllable of polysyllabic words. For the sake of clarity we indicate stress also in instances of enclisis where stress alternations are involved. Hyphens are introduced in IPA transcriptions and in the glosses to facilitate processing of the data; following the French orthographic conventions, they are used to separate the verb and its enclitics. Transactions of the Philological Society Volume 115:1 (2017) 98–136 doi: 10.1111/1467-968X.12093 © The authors 2016. Transactions of the Philological Society © The Philological Society 2016. Published by John Wiley & Sons, 9600 Garsington Road, Oxford OX4 2DQ and 350 Main Street, Malden, MA 02148, USA. (1) a. u/a/i 'cammani him/her/them they.call ‘They call him/her/them’ b. l ani camat-u/a/i him/her/them they.have called-msg/fsg/pl ‘They called him/her/them’ Zonza In this article, however, we argue for the conclusion that enclisis/proclisis segmental alternations, as well as stress alternations and reordering, are determined by syntacticosemantic conditions. Hauser et al. (2002) and Berwick & Chomsky (2011) conceive the human Faculty of Language (FL) as consisting of an inner core (the Narrow FL or FLN) which includes the recursive operations of syntax (the computational component) and of a number of components recruited by the FLN (the so-called Broad FL or FLB). The FLB includes both Conceptual-Intentional (CI) components (shared with general cognition, such as inferential and conceptual systems) and Sensory-Motor (SM) components, in practice articulatory and perceptual systems – which preside to the ‘externalization’ of the mental representations created by syntax and CI systems. In terms of this model, our thesis is that segmental and stress enclisis/proclisis alternations are not internal quirks of the SM system, but contribute to the externalization of syntactic computation and of the CI content entering into it. A crucial role in these alternations is played by the lexicon. Thus the lalternation in (1) must involve the existence of two lexical entries, namely l and the u/a/i series (see also section 2.1). The reason is that there is no general process of linsertion or ldeletion in Romance which could independently motivate it. This lexical fact is in itself arbitrary; for instance, in standard Italian, all 3rd person clitics involve l-. In other words, the child who learns Corsican or Italian must learn the list of clitics (and cannot infer it from Universal Grammar). This does not prevent us from looking for general SM (phonological) or syntactico-semantic principles that govern the distribution of lexical forms and of their allomorphs and that in innatist models, favour the acquisition of idiosyncrasies by the child who learns the language. We follow Distributed Morphology (DM; Halle and Marantz 1993), in assuming that the lexicon lists elementary units corresponding to morphemes and that the same Merge operation responsible for syntactic constituent structure also yields word formation from elementary morphemes. On the other hand, the conclusions on enclisis/proclisis alternations anticipated above can be read as supporting the idea that there is no morphological buffer component between the syntactico-semantic computational core and its SM externalization, of the type provided by such constructs as Morphological Structure and Late Insertion in DM. In sections 1.1–1.4 we will use only a very elementary syntactic model. The fundamental sentential tree is defined in Chomsky’s (1995) terms by the category V(erb) – or technically by the complex v-V – dominated by the category I(nflection) and by the category C(complementizer). Kayne (1991) proposes that proclisis on the finite verb in Romance, e.g. in Italian (2a), depends on the I position of the verb; enclisis on the infinitive, as in (2b), depends on the verb moving to C, while the clitic maintains the same position as in (2a). This is illustrated in structures (3a–b). Using enclisis/proclisis alternations as a diagnostics for 2 An anonymous reviewer mentions historical change as the sole source of explanation for the shape of the lexicon. This seems to us an oversimplification, insofar as the lexicon must present an internal organization suitable for being learnt by the children of each new generation. 3 The adoption of morpheme-based morphology automatically excludes paradigm-based approaches. This theoretical choice characterizes a larger spectrum of models than just generative ones. MANZINI & SAVOIA – ENCLISIS/PROCLISIS ALTERNATIONS 99

中文翻译：

---

浪漫中的 Enclisis/Proclisis 交替：异形和（重新）排序

Romance clitic 代词出现在 I 中动词的左侧和 C 中动词右侧。这种交替与： (a) 异形，特别是 lvs。零; (b) 压力转移；(c) 重新排序 clitic 字符串。(a)-(c) 中的交替也在非负和负上下文之间观察到。我们分析的关键点是：(i) lsegment 与确定的内容相关联；(ii) 解释性地，代词的范围超出了模态/非真实运算符；(iii) 在句法上，模态/非真实性的指数可能在其域中包含代词；(iv) 当语义范围 (ii) 和句法配置 (iii) 不匹配时，会发现 lsegment 的外部化。因此，异形（包括重音）远非形态语音学上的怪癖，有助于将非真实性的句法语义概念外化。在处理clitic（重新）排序时，我们提出了一个基于合并和线性顺序之间分离的模型。短语成分在 Romance 中排在动词的右边；clitics 反映了它们，因为它们被排序在左边，同时保持合并关系不变。1. 基本证据和当前帐户在许多罗曼语变体中，附着词的前言和后言位置与许多形态语音和顺序交替相关。存在三种主要的此类交替，即： (a) 节段异形；(b) 压力转移；(c) clitic 字符串的内部重新排序——这在下面的第 1.1-1.3 节中进行了说明。一些影响浪漫主义 clitics 的分段交替在语音上是有条件的。例如，在科西嘉岛的 Zonza 变体中，如 (1a) 中的发声前音仅先于辅音开始；如 (1b) 中所示，声乐起始以 l 形式开头。这种情况是音韵学上的，大概涉及 l 段与以下核的重音节化。1 本文中报告的研究部分得到了 PRIN 2012 远距离依赖项目拨款的支持。我们感谢 Adam Ledgeway 和 Diego Pescarini 对这项工作的第一个版本的详细评论。我们对方言数据使用广泛的转录。单词重音是有选择地指示的，即当它出现在倒数第二个音节或多音节单词的最后一个音节上时。为清楚起见，我们在涉及应力交替的 enclisis 实例中也表示应力。在 IPA 转录和注释中引入连字符以方便数据处理；遵循法语正字法约定，它们用于分隔动词及其附加词。语言学学会会刊第 115:1 (2017) 98–136 doi: 10.1111/1467-968X.12093 © 作者 2016。语言学会会刊 © The Philological Society 2016。由 John Wiley & Sons 出版，9600 Garsington Road , Oxford OX4 2DQ and 350 Main Street, Malden, MA 02148, USA。(1) 一个。你/a/i 'cammani his/her/them they.call '他们打电话给他/她/他们' b. l ani camat-u/a/i his/her/them they.have called-msg/fsg/pl '他们叫他/她/他们' Zonza 然而，在这篇文章中，我们认为 enclisis/proclisis 分段交替，以及应力交替和重新排序，由句法语义条件决定。豪瑟等人。(2002) 和 Berwick & Chomsky (2011) 将人类语言学院 (FL) 设想为由一个内核（窄 FL 或 FLN）组成，其中包括语法（计算组件）和许多组件的递归操作由 FLN（所谓的 Broad FL 或 FLB）招募。FLB 包括概念-意向 (CI) 组件（与一般认知共享，例如推理和概念系统）和感觉-运动 (SM) 组件，在实践中是发音和感知系统——它们主导着心理的“外化”由语法和 CI 系统创建的表示。就这个模型而言，我们的论点是节段性和重音包络/包络交替不是 SM 系统的内部怪癖，但有助于句法计算和进入其中的 CI 内容的外化。词典在这些交替中起着至关重要的作用。因此，(1) 中的 lalternation 必须涉及两个词条的存在，即 l 和 u/a/i 系列（另见第 2.1 节）。原因是浪漫中没有一般的插入或删除的过程可以独立地激发它。这个词汇事实本身是任意的；例如，在标准意大利语中，所有第三人称 clitics 都涉及 l-。换句话说，学习科西嘉或意大利语的孩子必须学习 clitics 列表（并且不能从通用语法中推断出来）。这并不妨碍我们寻找一般的 SM（语音）或句法语义原则，这些原则控制着词汇形式及其同形异形体的分布，以及在先天模型中，有利于学习语言的孩子获得特质。我们遵循分布式形态学（DM；Halle 和 Marantz 1993），假设词典列出了与语素相对应的基本单位，并且负责句法成分结构的相同合并操作也从基本语素中产生词形成。另一方面，上面预期的关于 enclisis/proclisis 交替的结论可以被理解为支持在句法语义计算核心与其 SM 外化之间没有形态缓冲组件的想法，DM 中的形态结构和晚期插入等结构提供的类型。在 1.1-1.4 节中，我们将只使用一个非常基本的句法模型。在乔姆斯基 (1995) 的术语中，基本句子树由类别 V(erb) 定义——或者技术上由复杂的 vV 定义——由类别 I(nflection) 和类别 C(补语) 支配。Kayne (1991) 提出 Romance 中有限动词的 proclisis，例如在意大利语 (2a) 中，取决于动词的 I 位置；enclisis 在不定式上，如 (2b) 中，取决于动词移动到 C，而 clitic 保持与 (2a) 中相同的位置。这在结构（3a-b）中进行了说明。使用 enclisis/proclisis 交替作为 2 的诊断 一位匿名评论者提到历史变化是解释词典形状的唯一来源。在我们看来，这似乎过于简单化了，因为词典必须呈现适合每一代孩子学习的内部组织。3 采用基于语素的形态学自动排除了基于范式的方法。这种理论选择表征了更广泛的模型，而不仅仅是生成模型。曼齐尼和萨沃亚 – ENCLISIS/PROCLISIS 交替 99