Levels of description and constraining factors in meaning construction: an introduction to the Lexical Constructional Model

Content (Part I)

• 1 引言 Introduction

• 2 模型总体结构 The general architecture of the model

• 3 词汇模板 Lexical templates

  • 3.1 RRG中的逻辑结构 RRG logical structures (LS)

  • 3.2 LCM中的词汇模板 Lexical templates in LCM

  • 3.3 词汇模板的形式 The format of lexical templates

    • Examples: grasp, consider, command

  • 3.4 小结 Summary

• 4 构式模板 Constructional templates

  • 4.0 构式模板 Constructional templates

  • 4.1 Level 1 constructions

    • transitive constructions
    • caused-motion constructions

  • 4.2. Level 2 constructions

    • ‘What’s X Doing Y?’ construction
    • double be construction

  • 4.3. Level 3 constructions

    • ‘Can You Y?’ directive construction

  • 4.4. Level 4 constructions

    • ‘X Let Alone Y’ construction
    • ‘Just Because X Doesn’t Mean Y’ construction
    • ‘X So Y’ construction and ‘Y After All X’ construction

Content (Part II)

• 5 涵摄 Subsumption

  • 5.1. Lexical-constructional subsumption

    • 5.1.1. External constraints on lexical-constructional subsumption
    • 5.1.2. Internal constraints on lexical-constructional subsumption

  • 5.2. Constructional subsumption and conceptual cueing

    • 5.2.1. External and internal constraints on constructional subsumption

• 6 结论 Conclusions

PART I

2 The general architecture of the model

2.1 Overall Architecture of LCM

Figure 1. The overall architecture of the Lexical Constructional Model [LT = lexical template; CT = constructional template; CS = Conceptual Structure]

• 【词模板】〖核心语法〗

  • ↓ 涵摄 / 概念牵发
  • ↓ ← 【Level 1 内外部约束】

• 【Lv1 构式模板】〖核心语法〗

  • ↓ 涵摄 / 概念牵发
  • ↓ ← 【Level 2 内外部约束】

• 【Lv2 构式模板/概念结构】

  • ↓ 涵摄 / 概念牵发
  • ↓ ← 【Level 3 内外部约束】

• 【Lv3 构式模板/概念结构】

  • ↓ 涵摄 / 概念牵发
  • ↓ ← 【Level 4 约束】

• 【Lv4 构式模板/概念结构 - 语篇表征】

  • ↓ 概念牵发

• 【最终的意义理解】

Notions

• 词模板 - Lexical Templates (Sec. 3)

• 核心语法 - Core Grammar

• 构式模板 - Constructional Templates (Sec. 4)

  • 概念结构 - Conceptual Structures

• 涵摄 - Subsumption (Sec. 5)

  • 概念牵发 - conceptual cueing

  • 约束 - constraints

    • internal constraints
    • external constraints

2.1 Levels in LCM

• Level 1 - core module

  • 和句法紧密相关的语义推理

• Level 2 - pragmatic module

  • 语用，语言交际的浅层推理

• Level 3 - illocutionary module

  • 深层推理，包括言外之力

• level 4 - discourse module

  • 语篇方面，尤其是衔接与连贯现象 [cohesion and coherence phenomena]

• 每一层

  • 要么 subsumed [涵摄] 到 a higher-level constructional configuration [构式结构] 之中，
  • 要么 acts as a cue [线索] for the activation of a relevant conceptual structure [概念结构] that yields an implicit meaning derivation [隐含义的派生].

• 每层的推理加工 is regulated by a number of cognitive constraints [认知约束].

2.2.1 Level 1 - lexical templates

• lexical items [词项] are built into constructional representations [构式表征]

• 词项的语义结构 is specified in terms of lexical templates [词模板]

  • internal configuration [内部结构] is established on the basis of combinations of lexical functions [词汇函数] and semantic primes [语义基元],

    • (14) predicate: [SEMANTIC MODULE<lexical functions>] [AKTIONSART MODULE<semantic primes>]
    • 各自有一些 variables [变量] (or predicate arguments [谓词论元]) within their scope.
    • 这些变量和它们关联着的语义结构 fuse into constructional variables [构式变量] (or roles [角色]) and 它们关联着的语义结构 thus giving rise to core grammar representations [核心语法表征].

2.2.2 Subsumption

• The fusion process - Subsumption [涵摄]

  • is regulated by a number of internal and external constraints [内外部的约束]

    • take the form of various cognitive operations [认知操作] (Section 5).
    • are licensing factors that filter out impossible combinations of lexical items with constructions [将不可能与构式组合的词项排除掉].

2.2.3 Level 1 inferential activity

• The LCM also allows for a degree of inferential activity [推理活动] (i.e. conceptual cueing [概念牵发]) at the level of core grammar.

  • Thus, sometimes lexical-constructional subsumption [词-构式涵摄] may give rise to an underspecified representation [未明确的表征] at level 1

    • She’s ready [for the party]
    • I will [marry you]
    • The student was late [for his Mathematics class]

2.2.4 Explicature Derivation

(cf. Sperber & Wilson 1995)

• Examples like these require Explicature Derivation [显义派生/显义推导]

  • assigning correct referents [分配正确的指代对象] to pronominal forms [代词形式]
  • disambiguating [消歧] meanings in context
  • completing underspecified expressions [补全未明确的表达式]

• different from Implicature Derivation [隐义派生/隐义推导]

  • 基于 a premise-conclusion pattern [假设-结论模式] 进行推理，其假设源自于说话人的认知环境。

2.2.5 Explicature Derivation In LCM

• Explicature Derivation [显义派生] In LCM

  • Reference assignment [指代分配]

    • occurs in connection with the fusion between predicate and constructional arguments [谓词论元与构式论元之间的熔合] at level 1, which makes this phenomenon independent of actual morphosyntactic realization [实际的形态句法实现].

  • Disambiguation [消歧]

    • carried out at any point of the semantic interpretation process [语义理解过程]

      • including the pragmatic(Lv2), illocutionary(Lv3) and discourse(Lv4) levels

    • For example, ‘Can You X’ (in Sec. 4.3).

• In Relevance Theory

  • explicature-derivation [显义派生] is seen as the context-guided inferential development [上下文引导的推理过程] of the blueprint provided by the linguistic expression [语言学表达式提供的蓝图].

2.3 Level 2

• Conceptual representations at level 2

  • either result from subsumption of fully worked-out level 1 representations into level 2 constructions
  • or arise from the cued activation of low-level situational models [状况模型] (or scenarios [剧本]).
  • Or we can have both constructional subsumption and cueing [牵发] at work thus yielding a rich array of meaning implications.

2.3 Level 2 - Who do you think you’re talking to? 1/2

Consider the case of the sentence Who do you think you’re talking to?

• 构式的观点

  • a variant of the ‘What’s X doing Y’ construction (cf. Kay & Fillmore, 1999)

    • carries a heavily conventionalized implication [高度惯例化的暗示] that the addressee has acted in a way that upsets the speaker

      • (e.g. What’s the child doing messing with my computer?).

  • the do you think element

    • 很大程度上决定了 the strength of the conventional implication

      • (cf. Who are you talking to?)

    • 在构式中有稳固的地位。

• 但有些隐含义是构式机制无法解释的……

2.3 Level 2 - Who do you think you’re talking to? 2/2

• 但有些隐含义是构式机制无法解释的。

• The ‘Who Do You Think You’re X?’ configuration

  • is typically associated with everyday situations where

    • the speaker gets upset when the addressee has behaved in socially inappropriate ways that directly affect the speaker negatively.
    • the speaker may feel that he has a right to challenge the addressee’s behavior.

• The ‘What’s X doing Y’ construction

  • 并非直接 convey this special meaning implication,

  • 而是 the implication may be cued contextually

    • e.g. What’s that you’re doing? referring to a situation where it is clear that the addressee has treated the speaker wrongly

  • 或者甚至 may be obtained on the basis of level 1 resources

    • e.g. by means of developing the explicit meaning of the message, as in What are you doing talking to me like that?

2.4 Level 3 1/2

• Level 3 representations obey the same principles as their level 2 counterparts.

• The difference:

  • just the kind of scenarios [剧本的种类] involved in the cueing process [牵发过程],
  • level 2: low-level situations
  • level 3: more generic scenarios [更一般化的剧本] which specify higher level social conventions [更高级的社会惯例] applicable to many different low level situations [适用于许多不同的低级状况].

2.4 Level 3 2/2

• Who do you think you’re talking to? can be interpreted

  • as a warning for the addressee to change his course of action.
  • This speech act value [这种言语行为的价值] derives from the social convention [源自社会习俗/社会惯例]

• Explicit performative predicate [明确的施为谓词]

  • obtain the same basic speech act value constructionally

    • (e.g. I warn you that you are not addressing me appropriately)

  • although there are different usage implications [不同的使用效果]:

    • ‘Who Do You Think You’re X?’ places greater focus on the misdeed than on the warning [聚焦于失当行为而不是警告] since the warning value is only implicit.

2.5 Level 4

• Finally level 4 representations, whether constructional or cued,

  • make use of high-level non-situational frames [状况无关的框架] establishing

    • logical connections [逻辑关联] such as

      • ‘cause-effect’ [因-果关联] or ‘evidence-conclusion’ [证据-结论关联],

    • temporal relations [时间关系] such as

      • ‘precedence’ [先于关系] or ‘simultaneity’ [同时关系], or

    • conceptual relations [概念关系] such as

      • ‘similarity’ [相似关系], ‘contrast’ [反差关系], ‘conditioning’ [条件关系], and ‘concession’ [让步关系], among others.

• One single expression can be explained on the basis of the convergence of multiple discourse connections. The bomb went off; three people died. (precedence & cause-effect)

3 Lexical templates

3.1 RRG Logical Structures (LS) 1/3

• The notion of Lexical Template is a development of the Logical Structures (LS) (postulated in Role and Reference Grammar (RRG), cf. Van Valin & LaPolla 1997, Van Valin 2005)

  • divides verb classes into {states, activities, achievements, semelfactives, and accomplishments together with their corresponding causatives}. (Drawing on Aktionsart distinctions [事件分类], cf. Vendler 1967)

    • States [状态型事件] and Activities [活动型事件] are primitives [基元]

    • Accomplishments [实现型事件] and Achievements [达成型事件]

      • consist of either a State or Activity plus a BECOME or an INGR operator

    • Semelfactives [半自主型事件] encode punctual events [准时事件/即时事件], which do not result in a State.

3.1 RRG LS 2/3

Verb class	Logical structure	Example	Instantiation of LS
State [状态型事件]	predicate’ (x) or (x, y)	see	see’ (x, y)
Activity [活动型事件]	do’ (x, [predicate’ (x) or (x, y)]	run	do’ (x, [run’ (x)])
Achievement [达成型事件]	INGR predicate’ (x) or (x, y), or INGR do’ (x, [predicate’ (x) or (x, y)]	pop burst into tears	INGR popped’ (x)
Semelfactive [半自主型事件]	SEML predicate’ (x) or (x, y) SEML do’ (x, [predicate’ (x) or (x, y)]	glimpse cough	SEML see’ (x, y)
Accomplishment [实现型事件]	BECOME predicate’ (x) or (x, y), or BECOME do’ (x, [predicate’ (x) or (x, y)]	receive	BECOME have’ (x, y)
Active accomplishment [主动类实现型事件]	do’ (x, [predicate1 ’ (x, (y))] & BECOME predicate2 ’ (z, x) or (y)	drink	do’ (x, [drink’ (x, y)]) & BECOME consumed’ (y)
Causative accomplishment [致使类实现型事件]	α CAUSES ß, where α, ß are LS of any type	kill	[do’ (x, ∅)] CAUSE [BECOME [dead’ (y)]

Table 1. Inventory of RRG logical structures (cf. Van Valin 2005: 45)

3.1 RRG LS 3/3

• The LS inventory is intended to capture only those aspects of the meaning of a word that are grammatically relevant [仅捕捉词义中和语法相关的方面]

• bases its decompositional system [分解系统/分析系统] on Aktionsart distinctions [事件分类，发展自 Vendler 1967]

  • capable of accounting for some important restrictions on constructional alternations [能够解释一些构式替换的重要限制]

3.1 RRG LS - 解释替换限制 - 以 工具主语构式 为例

The LS inventory 能够解释一些构式替换的重要限制：

• the instrument subject construction [工具主语构式]

  • possible for ‘break’ verbs [“打破”类动词]

    • (3) A bat broke the window.

  • impossible for ‘consumption’ verbs [“消耗”类动词]

    • (4) * A teaspoon ate his soup.

• transitivity system [及物性系统] 中的异常：

  • 两个看似平行的语法结构(/语法配置)具有不同的语法特性

• 原因: different predicate-object relationships [不同的谓宾关系]

  • ‘break’ verbs denote causative accomplishments [“打破”类动词 是 致使类实现型事件]
  • ‘consumption’ verbs are active accomplishments [“消耗”类动词 是 主动类实现型事件]
  • the instrument subject construction [工具主语构式] requires a causal element.

3.2 Lexical Template in LCM - 必要性 1/3

• another causative accomplishment verb (kill):

  • (5) The gunman killed the sheriff with a six-shooter.
  • (6) A six-shooter killed the sheriff.

• 根据 its LS configuration, kill 可以参与 the instrument subject construction. 但其语法行为 is not always comparable to that of ‘break’ verbs:

  • (7) The boy broke the window into a million pieces.
  • (8) * The gunman killed the sheriff into a dead body.
  • (9) Finally, Madison broke the window open and went in.
  • (10) * Finally, the gunman killed the sheriff dead.

• This difference in grammatical behavior cannot be explained by simply looking into the LS of both verbs:

  • (11) break: [do’ (x, ∅)] CAUSE [BECOME [broken’ (y)]

  • (12) kill: [do’ (x, ∅)] CAUSE [BECOME [dead’ (y)]

• It is necessary to achieve a greater degree of refinement in the semantic representation of the items belonging to the various verb classes.

3.2 Lexical Template in LCM - 必要性 2/3

• both kill and break have a resultative ingredient [表示结果的成分] in their meaning composition (captured by the primitive concept BECOME in the LS representation).

• However,

  • killing only admits one possible result, i.e. death,

  • while breaking is compatible with a wider range of possibilities, so it can make use of the caused-motion construction [致使运动构式]:

    • (13) a. Harry broke the glass into a thousand pieces.
    • b. Julie broke the vase into little fragments.
    • c. He broke the bottle into various segments, which wound around themselves, giving the observer the illusion of motion.
    • d. I broke the windowpane into four sections.
    • e. A picture fell from the wall, breaking the glass into thousands of sharp shards.
    • f. I broke the pillar into three parts and glued them to the base.
    • g. * He broke the board into three nearly equal pieces when he tested for yellow belt in Taekwondo.

3.2 Lexical Template in LCM - 必要性 3/3

• This difference, which is grammatically relevant, is not captured by the standard LS formalisms of RRG.

• Solution: enriching semantic representations [丰富语义表示]

  • possible to predict with greater accuracy when a given lexical item may or may not take part in a construction [更精确地预测给定的词项能否进入某个构式] n-4
  • → the LCM makes use of lexical template (developed by Mairal Usón & Faber 2002, 2005, 2007)

n-4 对 primitive predicates 的细分 2/2

• 前人工作 Van Valin & Wilkins (1993), Van Valin & LaPolla (1997) and Van Valin (2005) 也提过： primitive state and activity predicates [基元性的状态和活动型谓词] 需要进一步 decomposed [分解] 。

• RRG 的做法

  • uses internal variables (Greek letters) :

    • semantic parameters that express the semantic content of a verb,

  • and external variables (Roman characters) :

    • arguments of the meaning of a verb that are grammatically relevant,

  • Example - promise (Van Valin & La Polla 1997: 551):

    • (i) [do’ (x, [express(α).to.(β).in.language.(γ)’ (x, y)])] CAUSE [BECOME obligated’ (x,w)]

n-4 对 primitive predicates 的细分 2/2

• Lexical templates 的做法

• aim to provide a finer-grained [极佳颗粒度的] semantic decomposition [语义分解] of primitive state and activity predicates [基元性的状态和活动型谓词]

  • by using a universal semantic metalanguage which consists of

    • a number of primitives (or undefinable items)
    • and a list of operators (or lexical functions)

  • that account for the semantic and pragmatic parameters relevant to the meaning of a predicate.

3.2 Lexical Template in LCM - 构成

• Lexical templates 建立在 a universal semantic metalanguage [通用语义元语言] 基础之上。

  • an inventory of primes [基元] of meaning definitions

    • obtained by extensive factorization [广泛的因子分解]

      • (i.e. the search for meaning regularities across predicates within a lexical domain) n-5

  • a catalogue of operators [运算符/操作符], which

    • express the way semantic primes combine to express the more specific hyponyms [下位词].

n-5 lexical domains

• Faber & Mairal Usón (1999), drawing on Martín Mingorance (1984, 1990, 1995),

• structure the English lexicon into a number of lexical domains [词域]:

  • EXISTENCE 存在 (be, happen);
  • CHANGE 变化 (become);
  • POSSESSION 拥有 (have);
  • SPEECH 言语 (say);
  • EMOTION 情感 (feel);
  • ACTION 行动 (do, make);
  • COGNITION 认知 (know, think);
  • MOVEMENT 移动 (go, move);
  • PHYSICAL PERCEPTION 物理感知 (see, hear, taste, smell, touch).

• Each lexical domain is identified by means of a prime [基元] in boldface [粗体].

  • be used for the formulation of the meaning of more specific lexical items,

    • e.g. feel is used to define a number of hyponyms that belong to EMOTIONS, such as scare, terrorize, terrify, petrify.

• Factorization [因子分解] allows us to identify a number of primes which mark where the semantic decompositional chain actually ends.

• A further step is to encode the differentiating properties [具有区分作用的特性] between predicates belonging to the same lexical domain [属于同一词域的谓词].

3.2 Lexical Template in LCM - 渊源 (跳过)

• The set of semantic primes that we have used

  • coincide [相符] to a great extent with those used in Wierzbicka’s Natural Semantic Metalanguage research program,

    • which has been shown to be valid in almost over a hundred languages
    • (cf. Wierzbicka 1996, 1999, Goddard & Wierzbicka 1994, 2002).

• The operators that express the conceptual syntax

  • are based on the notion of LEXICAL FUNCTION [词汇函数] as propounded in Mel'čuk’s Explanatory and Combinatorial Lexicology.

    • Lexical functions [词汇函数] have also been shown to be typologically [类型学上] valid
    • (cf. Mel'čuk 1989, Mel'čuk et al. 1995, Mel'čuk & Wanner 1996) n-6.

n-6 It is impossible to do justice to the exact details of the formalism, so we refer the reader to Mairal Usón & Faber (2002, 2005, 2007) for a detailed account of the intricacies of the metalanguage.

3.2 Lexical Template in LCM - About lexical functions (跳过)

• We have adapted Mel'čuk’s lexical functions [词汇函数]

  • originally devised to apply to the combinatorial possibilities of a lexical unit

  • now they can

    • account for lexical domain-specific relations [具有词域特异性的关系],

    • and allow the codification [编码/体系化] of those semantic parameters [语义参数] that are not visible to syntax,

      • e.g. the manner [方式], purpose [目的], means [手段], social status [社会地位], speaker’s attitude [说话人的态度], the urgency of a request [请求的紧迫性], among others (cf. Faber & Mairal Usón 1999, Mairal Usón & Faber 2007).

• It is also necessary to make it clear that

  • there is a large degree of coincidence [相符/巧合] between

    • the parameters obtained through systematic factorization [系统性的因子分解]
    • and the list of collocational functions [搭配函数] provided by Mel'čuk’s model.

• Accordingly, we have used the inventory of lexical functions

  • to formalize the set of semantic parameters that are relevant within a lexical hierarchy.

3.3.0 The Format of LT

• Lexical templates have the following format:

  • (14) predicate: [SEMANTIC MODULE<lexical functions>] [AKTIONSART MODULE<semantic primes>]

• The semantic module,

  • expresses the semantic and pragmatic parameters [语义和语用参数] that underlie predicate meaning,
  • in terms of lexical functions.

• The Aktionsart module

  • is based on RRG logical structures
  • though the terms in boldface (primes) are not those used in RRG,
  • but rather belong to an inventory of primes.

3.3.1 Example - grasp

Consider the following example from the domain of cognition and speech act verbs:

(15) grasp: [MAGNOBSTR & CULM_12[ALL] ] know’ (x, y)

• The entry for grasp

  • is a hyponym of understand, which focuses on the acquisition of knowledge,
  • with know acting as the primitive that defines the whole verb hierarchy.

• The right hand side of the template

  • contains the Aktionsart module, which is a representation of the logical structure (as formulated in Van Valin 2005) , cf. Table 1 above).
  • In the representation grasp designates [指派] a state predicate [状态谓词] with two arguments (x, y) that are related to each other by a prime [基元] that is indicated in boldface, know’.

• The left hand side of the representation

  • encodes the semantic properties [语义特性] that allow us to distinguish grasp from other predicates within the same hierarchy.

  • In order to do so, we postulate two lexical functions [词汇函数]

    • MAGNOBSTR : capture the large degree of difficulty involved in carrying out the action [实施这一行动有较大的难度]
    • CULM_ALL : capture the endpoint of knowing something (which is understanding) [行动结束于知晓某事（即理解）]

3.3.2 Example - consider

• Another example of lexical template taken from the lexical domain of cognition verbs is the predicate consider.

  • 

  • (16) consider: [LOC_in^TEMP↔₁₂ CONT] think’ (x, y)

• inherits the properties of the state primitive think and

• specifies two semantic components which focalize

  • the duration (CONT) and
  • a present time location (LOC_in^TEMP↔₁₂).

3.3.3 Example - command

• a predicate like command,

  • speech act verb

  • a hyponym of order, inherits all its properties and

  • adds its own specificity which lies in the political/military context [政治/军事语境]:

    • 

    • (17) command: [MAGN_1[PERM]23, LOC ^soc↑₍₁₎ (PLACE_TYPE: political/military)] [do’ (x, [say’ (x, y)])] CAUSE [do’ (y, z)]

  • The subscripts ( 1, 2, 3 ) codify the speaker, auditory percept, and the addressee, respectively.

  • MAGN : the action is intensified [加强] to a very high degree, thus making it more forceful [行动被大幅增强，使其更具强制性].

  • PERM (applied to the first argument) : the speaker has power over the addressee and is licensed to ask him/her to do things [言者有凌驾于听者的权力，被允许要求他/她做事].

  • LOC ^soc↑₍₁₎ : ?

  • As for the Aktionsart module, this verb designates [指派] a causative accomplishment structure [致使类实现型事件结构] that is induced [催生] by an activity such that x says something to y and this causes y to do z.

3.4 Lexical Template Summary

• In sum, lexical templates

  • provide enhanced semantic representation and consequently

    • allow us to account for those properties which go beyond those aspects of the meaning of a word that are grammatically relevant [使我们能够解释那些「超出了词义中与语法相关的部分」的属性].

• As a result,

  • we deal with very rich semantic representations
  • along the lines of a robust and typologically valid formalism. n-7

n-7 This idea of formulating rich semantic representations is very much in consonance with the major tenets of Frame Semantics. For a comparison of lexical templates and frames, see Mairal Usón & Ruiz de Mendoza (2009).

4 Constructional templates

4.0 Constructional templates

• Lexical templates

  • are but lower-level constructional characterizations.
  • share crucial features with higher-level representations, which we shall call CONSTRUCTIONAL TEMPLATES, with which they interact.

• Constructional templates

  • are non-lexical representations with a grammatical impact. n-8

  • are not limited to the propositional level of description [命题级别的描述] ,

    • i.e. the level where lexical structure (consisting of predicate-argument combinations [谓词-论元组合] ) initially interacts with grammatical structure (involving tense [时] , aspect [体] , modality [情态] , and the arrangement of sentence constituents [句子成分的排列] ).

  • Thus, we include among constructional templates configurations that would 被其他理论家视作

    • pragmatic implicature [语用含义（Lv2）] ,
    • illocutionary force [言外之力（Lv3）] or
    • discourse structure [语篇结构（Lv4）] .

n-8 about verb-specific constructions (跳过)

• [LTs VS CTs] much in line with the distinction between VERB-SPECIFIC [动词特异性的] and VERB-CLASS SPECIFIC [动词类特异性的] constructions (cf. Croft 2003: 55ff.)

• VERB-SPECIFIC constructions [动词特异性构式]

  • account for the fact that there might be verbs within a lexical class that do not license a particular construction;

  • for example, within verbs of permission let and enable cannot occur in the ditransitive construction [双宾构式].

  • For Croft the existence of verb-specific constructions indicates that

    • the dichotomy [二分] between lexical rules and abstract constructions is false.

4.0 Constructional templates - levels or types

The LCM thus distinguishes the following constructional levels:

• level 1: constructions producing argument stucture characterizations [论元结构刻画];
• level 2: constructions accounting for heavily conventionalized situation-based low-level meaning implications [重度惯例化的基于状况的低级意义内涵];
• level 3: constructions that account for conventionalized illocutionary meaning [惯例化的言外意义] (situation-based high-level implications [基于状况的高级内涵]);
• level 4: constructions based on very schematic discourse structures [非常概略的语篇结构].

4.1 Level 1 constructions

4.1.0 Level 1 constructions 1/2

• Argument structure characterizations have been the main object of investigation in constructional approaches to linguistic description (e.g. Goldberg 1995, 2006).
• Our level 1 construction templates are closely related to the account of construction types originally identified by Goldberg (1995: 3–4), as shown in Table 2.

Type of construction	Example	Semantic representation
Ditransitive [双宾]	Pat sent Hill a fax	X CAUSES Y TO HAVE Z
Caused motion [致使移动]	Pat sneezed the napkin off the table	X CAUSES Y TO MOVE TO Z
Resultative [结果]	She kissed him unconscious	X CAUSES Y TO BECOME Z
Intransitive [不及物?]	The fly buzzed into the room	X MOVES Y
Conative [意动?/直接施动]	Sam kicked at Bill	X DIRECTS ACTION AT Y

![t2](LCM Figures & Examples/t2.png)

Table 2. Construction types

4.1.0 Level 1 constructions 2/2

• The list is not intended to be exhaustive [详尽无遗] ,

• but illustrates the high-level nature of the notion of construction at this explanatory level.

• Constructions thus consist of sets of arguments that

  • relate among one another on the basis of abstract predicates

    • such as CAUSE, BECOME, MOVE, and HAVE.

• We retain this basic configuration but adapt it to the requirements of the universal semantic metalanguage [通用语义元语言] that we use together with its associated Aktionsart characterizations [事件类型刻画]. As a consequence,

  • level 1 constructional templates make use of the same high-level representational mechanisms that characterize lexical templates [Lv1的CT使用与LT相同的高层表示机制],
  • except for internal variable descriptions[除了对于内部变量的描写], since these are idiosyncratic [独特的] to each verb and verb class.

• This feature of constructional templates is natural since

  • level 1 constructions are built by abstracting away [抽象出] elements common to a number of lower-level predicate classes [一些低级谓词类型的共有元素].

4.1.1 transitive constructions 1/2

• There are some verb classes that

  • linguists have traditionally classified as transitive [及物的] , where we
  • typically have an actor and an object of the action.

• We regard transitivity [及物性] in verbs

  • as the potential [潜能] of a verb (or a whole verb class) to participate in [参与] a higher-level configuration called the transitive construction [及物构式], which has the following basic constructional template:

  • (18) [do’ (x, y)]

  • In this template, we specify

    • an action (do’), an actor (x) and an object of the action (y).

4.1.1 transitive constructions 2/2

• The transitive configuration is the basic building block for other constructions, such as

  • the ditransitive [双宾] construction

    • (which adds one more constructional argument),

  • the resultative [结果] construction

    • (which adds a predicate expressing a resultant state of the object),

  • and the caused-motion [致使移动] construction

    • (which is in fact a case of resultative construction where the resultant state is a change of location).

4.1.2 caused-motion constructions

• The caused-motion construction is illustrated in (19) and (20) below:

  • (19) They kicked me out of the casino.
  • (20) Then they pushed me into my cell and locked the door.

• This construction conflates the roles of ‘affected object’ and ‘actor’ into one element of structure (the speaker in the examples above).

• It also conflates into one single predicate (‘kick’, ‘push’) two predicate values:

  • causing motion and
  • manner of causing motion.

• We propose the following constructional template for caused-motion, where the asterisk marks the optional status of an element:

  • (21) [do’ (x, y)] CAUSE [BECOME *NOT LOC (y, z)]

  • It may be noted that the caused-motion construction is built on the basis of the basic transitive pattern [do’ (x, y)].

    • This feature of the construction has consequences that will be addressed in Section 5.

4.2 Level 2 constructions

4.2.1 Level 2 constructions

• Level 2 meaning can be obtained on the basis of a combination of degrees of

  • pragmatically guided [语用引导的] and linguistically guided [言语引导的] situation-based low-level inferencing [基于情境的低层级推理].

• Pragmatically guided inferencing [语用引导的推理] has been traditionally termed IMPLICATURE [隐含义], which is arguably obtained by the application of some pragmatic principle

  • (like the maxims of the Cooperation Principle [合作原则] proposed by Grice 1975 or the Principle of Relevance [相关性原则] put forward by Sperber & Wilson 1995).

• Linguistically (i.e. lexically or constructionally) guided inferencing has been called PRESUPPOSITION [预设] (cf. Karttunen 1974).

  • Thus, She regretted getting a tattoo presupposes that the protagonist got a tattoo.

    • This is a background inference that derives from the semantic characterization of the verb regret, which suggests that the protagonist wishes that she had not performed the action (one cannot regret what one has not done).

• But the same sentence can have a number of different implicatures, which will vary with the context (e.g. ‘she won’t get a new tattoo’, ‘she was scolded by her parents’, ‘she didn’t know tattoos are permanent’, etc.).

4.2.2 Presupposition (linguistically guided inferencing)

Our account differs from traditional accounts on presupposition and implicature in pragmatics and the philosophy of language in two ways.

First, presupposition [预设] is not a pragmatic inference but a consequence of the lexical or constructional properties of semantic descriptions [预设不是语用推理，而是语义描写中词汇或构式特性的作用].

• Thus, factive predicates [叙实谓词] like ‘regret’, ‘realize’, and ‘be proud that’

  • have the property of introducing a thematic argument [主题论元] that is presented as true from the experiencer’s perspective [经验者观察为真].
  • This property is captured in their corresponding lexical templates [词汇模板] by applying the relevant set of lexical functions [词汇函数] and primitives [基元] to their second internal argument [第二个内部论元].

• Consider the lexical template for the verb regret:

  • 

  • (22) [SYMPT (sadness) INVOLV_1,2 (want) DEGRAD (do)₂ LOC_in^temp←/(become)₂ LOC_in^temp← because PROP_1,2 (true)₂ ] feel’ (x, y)

  • The template specifies that

    • there is a sentient entity that experiences sadness
    • because he/she believes that he/she had acted in the past in a certain harmful way
    • and wishes he/she had not done so.

4.2.3 Implicature (Pragmatically guided inferencing)

Second, our proposal admits that implicature [隐含义] can have a constructional motivation [隐含义也可以由构式驱动].

• A case: ‘What’s X Doing Y?’ construction (cf. Kay & Fillmore 1999). Consider the examples:

  • (23) a. What’s the child doing?
  • b. What’s the child doing in the kitchen?
  • c. What’s the child doing in the kitchen with the carving knife?

• Since the ‘What’s X Doing Y?’ construction is idiomatic [惯用语], it is formally different from others, like the transitive, ditransitive, caused-motion, resultative, and benefactive constructions.

• In Kay & Fillmore’s analysis, it consists of a combination of compulsory [常项] (what’s, doing) and variable [变项] (X, Y) elements.

4.2.4 degrees of idiomaticity

• However, we believe that making a distinction between idiomatic and non-idiomatic constructions does not do justice to their broad range of variation.

• There are degrees of idiomaticity. Thus, for the sake of coherence [为了保持一致] with other aspects of the LCM,

  • (where we see some elements realizing others in a multiple-level parametrization process [多级参数化过程],)

  • we shall refer to unmodifiable elements in a construction as NON-PARAMETRIZABLE,

  • while other elements will be recognized as being PARAMETRIZABLE to different extents.

  • In ‘What’s X Doing Y?’,

    • doing would be distinguished as a non-parametrizable element,
    • while the X and Y elements are highly parametrizable.

4.2.5 Analyze ‘What’s X Doing Y?’ 1/5

• As Kay & Fillmore note (1999: 4), semantically, the ‘What’s X Doing Y?’ construction seems to convey the idea that

  • there is something wrong about the situation described (there is an “incongruity” component in it).

• In our view, this incongruous situation can be seen as a call for remedial action or even as a complaint from an illocutionary perspective.

• really noteworthy: the importance of the Y element to guarantee this meaning value (cf. in this connection Ruiz de Mendoza & Otal Campo 2002: 156).

• Thus, while (23a) above is ambiguous between the questioning and the directive interpretations, this is clearly not the situation with (23b) and (23c), where the Y element has been realized linguistically.

  • (23) a. What’s the child doing? - ambiguous
  • b. What’s the child doing in the kitchen? - not so ambiguous
  • c. What’s the child doing in the kitchen with the carving knife? - almost not ambiguous

4.2.5 Analyze ‘What’s X Doing Y?’ 2/5

• The greater the elaboration [详尽阐述] of the Y element, the clearer the directive or the complaint [这一命令或抱怨] interpretations [解读].

• Because the construction implies that the speaker, in being able to supply so much information about what is going on, already knows the answer to his own question.

• This implication is captured by the Y element.

  • Since (23b) and (23c) make that element explicit, they qualify as instances of the level 2 ‘What’s X doing Y?’ construction.
  • In contrast, the potential reading of (23a) as a directive act or a complaint can only be obtained by means of a pragmatic operation.

4.2.5 Analyze ‘What’s X Doing Y?’ 3/5

• That we are dealing with implicated meaning in the three examples is evidenced by the plausibility [貌似可信] of the extensions in (24) below, all of which cancel out the meaning implication that something is going wrong.

• However, note that the greater the amount of information provided by the Y element, the more difficult it is to override the conventional implication, as in (24c):

  • (24) a. What’s the child doing? Not that I’m worried. Just asking.
  • b. What’s the child doing in the kitchen? Are you teaching him to carve the turkey? That’s definitely a good idea.
  • c. Hey, what’s the child doing in the kitchen with the carving knife? Did you then make up your mind to teach him how to carve a turkey? That’s definitely a good idea.

4.2.5 Analyze ‘What’s X Doing Y?’ 4/5

• Ruiz de Mendoza & Otal Campo (2002: 158) have argued that ‘What’s X doing Y?’ is related to other constructions that also seem to convey the idea that something is wrong in the situation described. Consider:

  • (25) a. Who’s been messing with my computer?
  • b. What’s the neighbor been doing with my flowers?
  • c. Where’s he been the last 50 years, golfing with Joe Kennedy Sr.?
  • d. Why’s he been messing with my computer?
  • e. Where’s he been all his life?

• These sentences exemplify what we may call the ‘Wh-’s been Y?’ construction, which roughly conveys the same idea as ‘What’s X Doing Y?’, namely that what is being described is wrong.
• The difference is that in ‘Wh’s been Y?’ the description is provided by a constructional presupposition [构式预设], while in ‘What’s X Doing Y?’ the description is explicit.

4.2.5 Analyze ‘What’s X Doing Y?’ 5/5

• There are some important similarities, however.

• Thus, the Y element of ‘Wh-’s been Y?’ usually contains an -ing form, though there are exceptions like (25e), and the related ‘What’s X Doing Y?’ construction also has an -ing form, doing, which in Kay & Fillmore’s account is not subject to variation.

• But this part of Kay & Fillmore’s analysis may not be accurate. Consider:

  • (26) a. What’re they building?
  • b. What’re they building in there?
  • c. What’re they building in there at this time of the year?

• Examples (26b) and (26c) can have meaning implications that are comparable to those in (23b) and (23c) above. They generally point to an incongruous state of affairs.

• Also this implication is stronger in (26c), where the Y element has been specified in greater detail, than in (26b), a situation which parallels the one described above for (23c) and (23b).

• So, what we really have is a ‘Wh-interrogative’ construction whose X and Y element can experience a greater degree of parametrization than the wh-component: ‘What be-PRESENT X V-ing Y?’

4.2 Level 2 constructions - double be construction

• double be construction

  • or copula doubling construction

  • Another example of level 2 configuration with elements that can be parametrized to varying degrees

  • e.g. The thing is, is that he didn’t tell the truth.

  • (cf. McConvell (1988), Tuggy (1996), and Massam (1999) )

  • which serves to call the hearer’s attention to a given situation while asserting its truthfulness or relevance.

  • It usually takes the configuration ‘X is, is Y’,

    • where X, which is marked by a high tone, is the topic
    • and Y, which takes a low tone, is the focus.
    • While Y is a relatively unconstrained element (it can be realized by any that-clause),
    • there is a fairly limited range of options for X, normally the thing, the problem, the question, what I mean, and what happens.

4.2 Level 2 constructions - Entrenchment

• In our view,

  • the possibility of modifying some of the relatively fixed elements of the construction is evidence that

    • constructions of this kind have been obtained through a process of ENTRENCHMENT [固位] (Langacker 1999: 105) or inherent ease of activation [固有的易激活性] ,

      • which is a function of type frequency.

4.3 Level 3 constructions

4.3.1 Functional grammar accounts for illocutionary meaning 1/3

• The idea that illocutionary force is coded in language systems either by lexical or grammatical means is not new.

• ILLOCUTIONARY FORCE INDICATING DEVICES [言外之力的指示手段] (Searle 1969):

  • word order [语序], stress [重音], intonation contour [语调轮廓?], punctuation [标点], verb mood [动词语气], and performative predicates [执行谓词].

• Some functional grammar accounts (e.g. Dik 1997a/b, Halliday & Matthiessen 2004) assume that

  • illocutionary meaning is part of grammar and has to be incorporated in grammatical description only in so far as there are linguistic devices to express such meaning [言外意义是语法的一部分，必须纳入语法描写之中，只有这样，才会有语言手段来表达这类意义].

4.3.1 Functional grammar accounts for illocutionary meaning 2/3

• Thus, Halliday & Matthiessen (2004: 108),

  • in examining the clause as exchange, note that

  • grammar has a mood system

    • (declarative [陈述式], interrogative [询问式], imperative [祈使式])

  • that realizes four basic speech functions

    • (offers [提议], commands [指令], statements [陈述], and questions [提问]).

• Dik (1997b), in turn,

  • uses typological criteria [类型学标准] (Sadock & Zwicky 1985)

  • to distinguish four basic illocutionary functions coded in most languages:

    • statements [陈述], questions [询问], commands [指令], and exclamations [感叹].

  • These basic illocutions [言外之意/言内行为]

    • correspond to three sentence types (declarative, interrogative, imperative)
    • and [correspond] to exclamative constructions.

4.3.1 Functional grammar accounts for illocutionary meaning 3/3

• Dik (1997b) further argues that there are grammatical mechanisms that can be used to derive other illocutions from the basic ones. For example,

  • declarative, imperative, and interrogative sentences can be converted into requests [请求] by adding please

    • (e.g. Please, I hate this music!; Give me the book, please; Can you swim, please?),

  • declaratives into questions by means of a tag

    • (e.g. She’s a nice girl, isn’t she?),

  • and imperatives into exclamations through suprasegmental features [超音段特征]

    • (e.g. Look who’s THERE!),

  • among other possibilities.

4.3.2 Level 3 (illocutionary) constructions 1/3

• Our own view is very close to Dik’s proposal, but with the crucial difference that

  • what Dik calls grammatical derivation [语法派生] is not really so,
  • but a matter of constructional conventionalization [构式惯例化] (Ruiz de Mendoza & Baicchi 2007).

• We thus believe that there are level 3 constructions that evince the same degrees of conventionalization that we find for level 2 constructions. Witness:

  • (27) a. Can you write Morse code? - interrogative
  • b. Can you explain Einstein’s theory of relativity? - interrogative or request
  • c. # Can you write Morse code, please? - # request
  • d. Can you explain Einstein’s theory of relativity, please? - request
  • e. Can you hear the ocean? - interrogative
  • f. # * Can you hear the ocean, please? - # * request
  • g. Can you listen to me? - request

4.3.2 Level 3 (illocutionary) constructions 2/3

• ‘Can You Y?’ sentences

  • a relatively conventional way of making requests.
  • the same construction can be used to ask questions.
  • So, constructionally ambiguous.

• Grammatically, the ‘Can You Y?’ string is interrogative [询问式], so it could be argued, à la Dik [按 Dik 所说], that the request interpretation [请求解读] is derived pragmatically [通过语用得到], in contrast with example (27d), where we have grammatical derivation [通过语法得到] through the addition of please.

• But this is clearly not the case for two reasons.

  • First, (27a, e) can hardly be converted into requests,

    • not even through grammatical mechanisms, (27c, f).

  • Second, sentences like (27g) have an extremely strong default interpretation [默认解读] as requests.

4.3.2 Level 3 (illocutionary) constructions 3/3

• Evidently, the existence of default interpretations [默认解读] that are not predictable from grammatical form calls for a non-derivational account.

• Instead, we propose a constructional account of non-pragmatic illocutionary meaning [非语用的言外意义], where illocutionary constructions [言外之意构式/言内行为构式] may have compulsory [义务的/固定的] (nonparametrizable [不可参数化的]) and variable (parametrizable) elements, as was the case with level 2 constructions.

• Parametrization [参数化] is not an unrestricted process [并非不受限制].

  • Thus, the Y element in the ‘Can You Y?’ directive construction can only take verbal predicates belonging to certain lexical classes

    • (state [状态] and non-active accomplishment [非自主实现] predicates are excluded).

    • Additionally, the Y element

      • may optionally include please (Can you listen to me, please?) or beneficiary complements (Can you sing for me?)
      • as ways of either reinforcing the directive character of the sentence or simply forcing the request interpretation.

4.3.3 Non-paramatrizable are not absolute 1/2

• Finally, note that the non-parametrizable part of the construction, just like in some level 2 constructions, admits some variation with slight changes in meaning that do not alter the overall illocutionary interpretation [构式中的非参数化成分(常项)也允许轻微变化而不改变整体的言外意义解读].

• That is, “non-paramatrizable” does not mean absolutely invariable but rather realizable through a limited set of options [有限的选项集合]. Thus, we can have sentences using will you, would you and could you forms for the ‘Can You’ part of the construction:

  • (27) h. Will/would you please answer all the questions in the questionnaire?
  • i. Can/could you please answer all the questions in the questionnaire?

4.3.3 Non-paramatrizable are not absolute 2/2

• The question may now arise as to the status of ‘Can You Y?’ strings that have a purely interrogative function.
• In the LCM such strings are considered parametrizations of the Aux-NP constituents of the polar interrogative construction.
• Obviously, ‘Can You Y?’ requests have been derived by conventionalizing one form of performing such parametrization and giving it a stable speech act value.
• Again, as with level 2 configurations, the stable association of a specific interpretation [一种特定解读] with the construction is a matter of degrees of entrenchment [固位].

4.4 Level 4 constructions

4.4.1 ‘X Let Alone Y’ 1/2

• Level 4 constructions:

  • capture conventional implications that go beyond the clause level [捕捉超出小句级别的惯例化隐含义].

• One well-known example: the ‘Let Alone’ construction

  • e.g. Would anyone buy this garbage, let alone eat it? ; see Fillmore, Kay & O’Connor 1988

  • can be symbolized as ‘X Let Alone Y’

  • has two highly variable constituents (X and Y) related by let alone

  • sets up an entailment relation between the two constituents that it relates.

    • In this relation, the state of affairs depicted in Y is considered less likely to happen than the one in X [Y事件发生可能性小于X].
    • In turn, X either expresses or implicates a negative situation [X暗示负面情况], i.e. one in which a state of affairs is portrayed as not being the case [和现况不符的事况].
    • The kind of discourse relationship described has the pragmatic consequence [语用后果] of placing emphasis (through an exaggerated contrast [通过夸大差异]) on the Y element [对Y元素加以强调].

4.4.1 ‘X Let Alone Y’ 2/2

• ‘X Let Alone Y’ has

  • two highly variable elements

    • X is negative
    • Y provides a contrast on the bases of a scale

  • one element with a low range of parametrization possibilities: the coordinating conjunction [起协调作用的连词] can sometimes be replaced by

    • never mind

      • e.g. A lot of poor parents barely have time to see their kids, never mind cook for them.

    • and much less

      • e.g. Few will ever see an Aston Martin up close, much less drive one.

4.4.2 ‘Just Because X Doesn’t Mean Y’

• ‘Just Because X Doesn’t Mean Y’

  • Another level 4 construction that admits a degree of variation in some of its compulsory elements

  • e.g. Just because we live in Berkeley doesn’t mean we’re left wing radicals.

  • used to indicate that the second constituent does not necessarily follow from the first (Holmes & Hudson 2000).

  • Note: not easily modifiable without doing violence to the construction

    • * Because we live in Berkeley doesn’t mean we’re left wing radicals

  • while ‘Doesn’t Mean’ does admit some variation

    • Just because we live in Berkeley is no reason to think we’re left wing radicals ;

  • (cf. Bender & Kathol 2001, for similar considerations).

4.4.3 ‘X So Y’ and ‘Y After All X’ 1/2

• In our view, the ‘Just Because X Doesn’t Mean Y’ construction

  • is a highly conventionalized parametrization [惯例化的参数化] of a more generic evidence-conclusion/conclusion-evidence pattern [更一般的「证据-结论模式」或「结论-证据模式」] generally expressed through discourse connectors [语篇连接词] such as so and after all.

• Consider the following re-phrasings of one of the examples above:

  • (28) a. They live in Berkeley. They must be left wing radicals. - [evidence-conclusion]
  • b. They must be left wing radicals. They live in Berkeley. - [conclusion-evidence]
  • (29) a. They live in Berkeley, so they must be left wing radicals.
  • b. They must be left wing radicals; after all, they live in Berkeley.

• in (28a) and (28b) the connections are implicit and need to be derived inferentially.

• in (29a) and (29b), they have been made explicit through discourse connectors [语篇连接词] , which thus acquire a clear constructional potential.

4.4.3 ‘X So Y’ and ‘Y After All X’ 2/2

• We have two reverse constructions ‘X So Y’, and ‘Y After All X’, where X and Y can be either positive or negative statements, with various meaning effects:

  • (30) a. They don’t live in Berkeley, so they can’t be left wing radicals.
  • b. They can’t be left wing radicals; after all, they don’t live in Berkeley.
  • c. They live in Utah, so they can’t be left wing radicals.
  • d. They can’t be left wing radicals; after all, they live in Utah.
  • e. They don’t live in Utah, so they could be left wing radicals.
  • f. They could be left wing radicals; after all, they don’t live in Utah.

• Relevance theorists (e.g. Blakemore 1987, 1988, 2002, Blakemore & Carston 1999, Wilson & Sperber 1993) have studied discourse connectors (among them so and after all) in quite some detail.

4.4.4 Conceptual and Procedural Encoding in Relevance Theory

• Relevance Theory makes a distinction between CONCEPTUAL and PROCEDURAL ENCODING [概念性编码和程序性编码].

  • Conceptual Encoding: deals with the creation of coherent schematizations of experience [经验的连贯性图式].
  • Procedural Encoding: deals with the creation of instructions or constraints on pragmatic processing [语用加工中的指示或约束].

• Discourse connectors are a case of procedural encoding

  • since they tell us the kind of inferential process that the hearer is expected to follow.
  • ‘X So Y’ - the hearer is expected to think of X as the evidence for the conclusion in Y;
  • ‘Y After All X’ - (reversed reasoning pattern).

• This proposal has the advantage of recognizing the inferential dimension [推理的范围/尺度] of discourse connectors.

• However ...

4.4.5 Discourse Connectors exploiting Models or Frames

• However, we do not think that they are procedural any more than other aspects of interpretation at other levels, including what we have called subsumption.

• What discourse connectors do is activate high-level cognitive models (or frames) like the EVIDENCE FRAME (Ruiz de Mendoza 1999: 23), we can define as follows:

• (31) A state of affairs X is evidence of the existence of another state of affairs Y if thinking of X as true involves thinking of Y as true as well.

• So and after all exploit the Evidence Frame in converse ways by parametrizing differently the X and Y elements in it.

PART I 总结

PART I 总结

• 从激进构式主义的角度将语义/语用理解整合到统一的体系之中。

• 四个层级：字面义，语用义，言外义，语篇义。

• 形式化方案：词汇模板和构式模板。

  • 一方面可能过于复杂；
  • 另一方面在Lv2以上说明不够细致（可能在PART II有所阐述）。

• 围绕若干构式，修正/改进了许多前人的理论观点。

  • 预设不是语用推理，而是来自词汇或构式的属性。
  • 习语性不是二分的，而是有程度的，称为参数化的程度。
  • 区分语法派生和构式惯例化。
  • 构式常项也可以变化。
  • 语篇连接词 [Discourse connectors] 的编码并不是程序性的 [Procedural]。

语言知识资源方面

• 词汇模板

  • 词汇函数库
  • 基元词库
  • 词域集

• 构式模板

  • 分层级的构式库

THANKS