小组活动总结 | 关于两项认知语言学实证研究的讨论

语言学交流平台2018-06-08 13:26:31

来自上海的语言学交流平台


201745日星期三晚,上外英语学院硕士生李夷做了题为《隐喻:跨语言实证与应用》的报告。在报告中,主讲人介绍了隐喻在认知语言学中的一些研究成果,以两项实证研究为主(Boroditsky2001Stefanowitsch2006)。本文对此进行梳理,供赏读与讨论。

 

隐喻在认知语言学研究中有两个很有趣的现象:一是所有人都在使用隐喻,比如研究构式语法的Goldberg用隐喻来阐释构式库(construct-i-con)中不同构式的关系;研究认知语义的学者则更不用说,他们讨论的议题包括隐喻的概念化、隐喻与词汇语义的关系、跨语言隐喻比较等等。二是几乎所有人都在批评早期Lakoff等人内省法隐喻研究的不足,虽然这部分学者中很多人的研究仍然基于Lakoff的概念隐喻理论,而且不少研究结果本质上也没有与内省法的结论有明显冲突或革新。语言定量研究的发展已经为研究者提供了许多好用的工具,本着物尽其用的精神,许多种定量研究方法都被拿来替代内省的路子。有用心理实验来测试语言对隐喻思维影响的Boroditsky,有用神经实验来测试隐喻在大脑中处理过程的KutasGioraCardillo等人,更多的学者则选择了语料库作为研究方法,这也符合认知语言学基于使用(usage-based)的理论主张。这些研究方法可以说体现了认知语言学家对早期内省法隐喻研究的反思,这应当是一个进步:毕竟,通过量化实证研究得出的隐喻成果,不仅可以更好地帮助我们认识隐喻的庐山真面目,还可以让其它方向的语言研究者更好地利用隐喻来辅助他们的研究。

 

围绕隐喻进行的定量研究很多,本次报告主要选取了两项进行讨论,首先介绍的是Boroditsky2001)针对英语母语使用者(以下简称ES)和普通话母语使用者(以下简称MS)对TIME这个目标域思维进行的心理实验。

Boroditsky (2001): Does language shapetime? Mandarin and English speakers’ conception of time

聚焦现象

TIME in English: front/back terms

1a. We have a whole day ahead of us.

1b. The hard times are behind us. (TIME IS SPACE)

TIME in Mandarin Chinese: front/back  terms are also used (前,后)

3a. 在桌子站着一个学生。

3b. 虎年之前是什么年?

4a. 在桌子边站着一个老师。

4b. 大学毕业,我进入了研究生院。 (Scott, 1989)

But, Mandarin speakers also systematically use vertical metaphors  to discuss time (上,下)

esp. order of earlier or later events:

5a. 猫

5b. 个月

6a. 她了山没有?

6b. 个月 (Scott, 1989)

观察结论

Conclusion1: Both Mandarin & English  speakers use horizontal terms to talk about time. In addition, Mandarin  speakers commonly use the vertical terms &

实验1

(目的)

Study how native languages affect how  speakers think about abstract target domain TIME

实验1

(被试)

26 ES

20 MS (MS have a mean age at the onset of  English learning of 12.8 years)

实验1

(材料)

128 primes and 32 targets (TRUE/FALSE  questions)

Primes: 128 spatial scenarios (fig1&2).  Half horizontal and half vertical.

                                                           

Fig  1

Fig  2

Targets: 16 statements about the order of months. Half used the spatiotemporal terms (before & after) and half used purely temporal terms (earlier & later). All four terms were equal in  frequency and all target statements were TRUE in answer.

Fillers: 16 additional statements about months. Randomly inserted in the exp1 trials.

实验1

(步骤)

Participants were primed by spatial scenarios accompanied by a sentence description and were either horizontal or vertical (Fig1 & 2)

Then, statements about time were made to the subjects (e.g. ‘March comes before than April’ or ‘March comes earlier than  April’). Each subject finished 6 practice questions and 64 experimental  trials, each of which consists of 2 spatial prim questions followed by 1 target question about time.

Subjects answered each target question twice: once after each type of prime (horizontal or vertical). Filler trials and random order of all trials were used to ensure that participants cannot figure out the trial structure of exp1.

实验1

(结果)

Fig  3

Fig  4

ES: answered time questions faster after horizontal primes (2128 ms) than after  vertical primes (2300 ms), F (1, 25) = 13.76, p<0.01. Reaction times were also shorter for questions phrased in before/after terms (2135 ms)  than for those in earlier/later  terms (2294 ms), F (1, 25) = 8.23, p<0.1.  

MS: answered time questions as quickly after horizontal primes (2422 ms) as after vertical primes (2428 ms), F (1, 19) = 0.01, p=0.92. Like ES, MS answered before/after target questions faster after horizontal primes (2340 ms) than after vertical primes (2509 ms).

However, purely temporal earlier/later targets were solved by MS faster after vertical primes (2347 ms) than after horizontal primes (2503 ms), which can be confirmed as an interaction between prime type & target type, F (1, 19) = 4.55, p<0.05.

实验1

(讨论)

Both ES and MS answered spatiotemporal before/after questions faster after  horizontal primes than after vertical primes (fig 3&4)

However, ES and MS were differently in  terms of purely earlier/later  temporal questions:

ES faster after horizontal primes than  after vertical primes

MS faster after vertical primes than  after horizontal primes

实验2

(目的)

Further test the relationship between language experience and patterns in thinking.

(How much and in what ways does learning new languages influence one’s  way of thinking?)

实验2

(被试)

25 Mandarin-English bilinguals

With varying age at which they began to learn English (Age of English Acquisition)

实验2

(材料)

96 primes and 40 targets

Primes: half horizontal (see Fig 5) and half vertical (see Fig 6)

Fig 5


Fig 6

TRUE/FALSE questions ‘X will win’ or ‘X will lose’

Targets: 40 statements about the order of months (similar to Exp1)

Fillers: (similar to Exp1)

实验2

(步骤)

(与实验1相似)

实验2

(结果)

r=0.47, p<0.01

Vertical bias (in milliseconds) = the difference in RT between horizontal-primed targets and vertical-primed targets

实验2

(讨论)

The age of acquisition of English was positively correlated with vertical bias. The younger the subject started learning English, the less vertical bias they exhibit.

实验3

(目的)

minimize differences in nonlinguistic factors that may influence thoughts.

实验3

(被试)

70 ES

实验3

(材料)

‘Mandarin’ linguistic training to native ES:

7a. Cars were invented above fax machines.

7b. Wednesday is lower than Tuesday.

实验3

(步骤)

(与实验1,2相似)

实验3

(结果)

(提示:实验3的数据为第三组柱,即English speakers after training)

最终结论

1. Native language exerts a strong influence over how one thinks about abstract domains (TIME)

2. Such bias can be mediated by acquisition of foreign languages.

3.  Learning a new way to discuss a familiar abstract domain (TIME) can change temporarily one thinks about that domain.

 

这项实验最大的优点就在于量化了母语使用者使用隐喻思考时的反应时(reaction time),这属于心理实验中的潜伏性数据(latency data)。该数据的测量,可视化了被试者信息处理的难易程度,让我们可以比较直观地比较英语母语使用者与普通话母语使用者对TIME这一源域的思考方式异同(详见Fig 1&2)。

在结束了第一个实验的介绍后,听众与主讲人进行了简单问答。主要存在以下质疑

    如何确保这些刺激(prime)可以起到正确的启动效应。比如有听众提出,他看到实验1和实验2中的刺激并不会联想到前后和上下。由于这些图片所显示的概念间的关联性难以形式化,只能用直觉判断,这种心理学依据的可靠性是一个质疑点。

    有人质疑心理实验就像是给计算机测量温度,但仍然不知道其内部使用的算法如何。因此即使有解释力,但没有预测力,或只有较弱的预测力(比如只能以二分的形式来回答谁好或谁坏)

 

前面说到语料库是现在认知语言学研究中的热门手段,而利用语料库来研究隐喻的代表人物之一就是Stefanowitsch。在Stefanowitsch2006)这篇文章中,一种名为MPAMetaphorical Pattern Analysis)的手段被用来进一步细分隐喻,把同时具有目标域和源域关键词的隐喻表达称为MPMetaphorical Pattern),并使用目标域关键词作为检索词,在英国国家语料库(British National Corpus)内进行隐喻抽取。

Stefanowitsch (2006): corpus-based study ofEMOTION-specific metaphors

前期分析

conceptual metaphors vs metaphorical expressions

former: general mental mappings from source domains to target domains

latter: individual linguistic items instantiating these mappings

1. ARGUMENT IS WAR

SD WAR

TD ARGUMENT

a. Your claims are indefensible.

b. His criticisms were right on target.

c. He shot down all of my arguments.


2.  LOVE IS WAR

SD WAR

TD LOVE

a. He is known for his many rapid conquests.

b. He fled from her advances.

c. He is slowly gaining ground with her.

From 1&2 we can distinguish two types of metaphorical expressions:

Those that contain target-domain items  (TD: indefensible, target, shoot down, SD: claim, criticism, argument)

Those that do not (SD: conquests, fled, advances, gain ground)

Thus, Stefanowitsch refer to items in 1 as metaphorical patterns (hereinafter, MP):

A metaphorical pattern is a multi-word expression from a given SD into which one or more specific lexical item from a given TD have been inserted.

Metaphorical Pattern Analysis  (hereinafter, MPA): retrieve a large number of instances of a TD item (claim, criticism, argument) from a corpus and exhaustively identify the metaphorical patterns that it occurs with.

MPs also exhibit paradigmatic relations between TD items & SD items:

1c. He shot down all of my arguments.

TD: argument

SD: shot down (my planes/missiles)

Conceptual metaphor: ARGUMENT IS WAR

Specific relation: argument ≈ planes/missiles

 

2a. He is known for his many rapid conquests.

TD: (null)

SD: conquests

Conceptual metaphor: LOVE IS WAR

Specific relation: (null)


The relative frequency of SD & TD items in a set of MPs may be used to assess the productivity of a certain conceptual mapping

研究个案

EMOTION metaphors (HAPPINESS/joy)

样本容量

A random sample of 1,000 hits from BNC  for each lexical item was retrieved

比较对象

Kövecses (1998)

15 metaphorical mappings for HAPPINESS

A. UP We had to cheer him up

B. BEING OFF THE GROUND  I am six feet off the ground

C. BEING IN HEAVEN  That was heaven on earth

D. LIGHT  Lighten up

E. VITALITY  He was alive with joy

F. WARM  That warmed my spirits

G. HEALTH  It made me feel great

H. AN ANIMAL THAT LIVES  WELL  He was happy as a pig in shit

I. A PLEASURABLE PHYS. SENSATION  I was tickled pink

J. FLUID IN A CONTAINER  He was overflowing with joy

K. CAPTIVE ANMIAL  His feelings of happiness broke loose

L. OPPONENT IN A STRUGGLE  He was knocked out

M. A RAPTURE/HIGH  I was drunk with joy

N. INSANITY  They were crazy with happiness

O. A NATURAL FORCE  He was swept off his feet

Why comparison:

1. Prove the exhaustivity & systemacity of MPA

2. Investigate the questionable  subtypes of conceptual metaphors proposed by previous introspection method (Kövecses, 1998)

数据分析1

906  metaphorical expressions out of 1,000 hits: comparison between corpus data & introspection

Most of the mappings identified via  introspection are also identified by MPA except HAPPINESS IS BEING IN HEAVEN, BEING HAPPING IS BEING AN ANIMAL THAT LIVES WELL, HAPPINESS IS A PLEASURABLE PHYSICAL SENSATION (questionable subtypes of conceptual metaphors)

Conceptual metaphors proposed by Kövecses (1998) account for only 7.2% of all MPs identified (Table 1)


Table  1

A large portion of MPs extracted from BNC are i. location metaphors (8.9%) and ii. object metaphors (69.31%). Another 10.9% of the total MPs are listed in Table 2

Table  2

Still, 3.6% of the total MPs are unaccounted (innovative/literary metaphors: HAPPINESS IS A BALLON, bubble of joy; HAPPINESS IS BLOOD, joy pulsate through X; HAPPINESS IS A SHARP OBJECT, stab of joy, INTENSITY OF HAPPINESS IS DEPTH, deep joy).

数据分析2

Metaphors  ignored by Kövecses (1998) ---- OBJECT metaphors:

1. CAUSING EMOTION IS TRANSFERRING AN OBJECT (e.g. X bring/give joy, X share X’s joy)

χ2=142.96,  p<0.0001

 

2. TRYING TO ATTAIN AN EMOTION IS SEARCHING FOR AN  OBJECT (e.g. X find joy in Y)

χ2=34.82,  p<0.001

 

3. INTENSITY OF EMOTION IS SIZE (OBJECT)

χ2=17.28,  p<0.05

 

4.  INTENSITY OF EMOTION IS QUANTITY (OBJECT)

χ2=15.56,  p<0.05

 

5. EMOTION IS UP

χ2=42.19,  p<0.001

LOCATION metaphors


6. EMOTION IS A LOCATION

χ2=39.96,  p<0.001

 

7. ACTING ON AN EMOTION IS ACTING IN A LOCATION

χ2=31.42,  p<0.001

数据分析3

Interaction between metaphor & lexical semantics

happiness & joy (how do the two words interact with the EMOTION metaphor HAPPINESS?)

Different frequency of mapping

TRYING TO ATTAIN AN EMOTION IS SEARCHING FOR AN EMOTION

joy: 16/1000 (3 different MPs)

happiness: 110/1000 (28 different MPs)


a. TRYING TO ATTAIN HAPPINESS  IS SEARCHING/HUNTING FOR SOMETHING

sought-after  happiness, unlooked-for happiness, pursuit of happiness, search/quest for  happiness, path/route/way to happiness, X chase (after) happiness, X be in search of happiness, X harry after happiness, X look for happiness (in X), X  search for happiness, X pursue happiness, X seek happiness, X reach out towards happiness, X snatch at happiness, X stretch out hand for happiness

 

b. ATTAINING HAPPINESS IS  FINDING/CAPTURING SOMETHING

happiness  seem within reach, X attain happiness, X find happiness (in/through/with X),  X capture/grab/recapture happiness, X reach happiness

 

c. NOT BEING ABLE TO ATTAIN  HAPPINESS IS INABILITY TO REACH SOMETHING

X  stand in way of happiness, happiness elude X, happiness be irretrievable

 

Different aspects of mapping

happiness: the search for ~, the route taken to find ~, the moment of finding ~, the possibility of not finding or reaching ~

joy: ATTAINGING HAPPINESS IS  FINDING OR CAPTURING SOMETHING

猜想

Stefanowitsch (2006): happiness is potentially a less intensively experienced emotional state and hence a more stable one and can be easily conceptualized.

Joy is more intense and short-lived, and thus can only be stumbled by chance.

This can also explain why BEING HAPPY IS UP is significantly associated with joy.

 

这项研究主要有两个意义

其一,Stefanowitsch在进行语料库隐喻检索前先介绍了比较对象,即Kövecses1998)进行的happiness概念隐喻分类,并指出了他的分类中有疑问的几条。随后的语料库分析证明,几条有疑问的分类在检索得到的数据中并不存在,且Kövecses所列举的共15happiness隐喻仅占到总数据的7.2%,这一发现可以从一个侧面说明内省法隐喻研究的局限。

其二,Stefanowitsch基于语料库的数据,分析了happinessjoy这两个词与隐喻的目标域happiness之间的互动关系,进而发现了happinessjoy在同一目标域中不同的作用。这种对隐喻-词汇语义的交叉研究作用不可小觑:我们对这一互动关系研究得越透彻,就越能发现哪些词汇与隐喻关系最为密切,也就能在庞大的语料库中更精确地找到隐喻。

 

       报告最后,针对这项实证研究同样有一些疑问

    如何通过形式标准从语料库中准确地找到想要的条目。在语料库检索中常存在这个一个问题,比如想要的条目集合为A,而通过检索式检索到的条目为集合B。如何排除B-A,并将A-B包含进来,是一个难点。排除B-A常常采用的是人工筛选,但当条目数量很大时,任务量非常重,且一旦中间出现了问题需要重做,任务量会成倍增长。将A-B包含进来一方面可以通过不同检索式的组合,另一方面则需要依赖语料库技术本身的发展,在目前阶段,A-B常常是忽略的。

    隐喻判断的标准问题,即具体哪些条目属于上述的A集合。目前都是在第一遍检索关键词的基础上再通过人的直觉判断的。但不同学者的标准在实际操作中都会不一样。因此这仍然是一个难题。

    典型的隐喻可以帮助计算机在自然语言理解过程中解读出说话人真正想表达的意思,但当隐喻的外延扩大之后,有大量的隐喻是通过上下文的词汇推断出来的,甚至出现了上文中检索joy1000个例子中906个例子都是隐喻的情况,那试问:还有什么不是隐喻呢?这个词究竟该怎么用,才不是隐喻的用法呢?

 

本文提出的疑问在此次活动的问答环节中尚未解决,有待进一步讨论,欢迎您宝贵的留言!


林先生,做有观点的语言学平台

长按二维码关注

微信号:Mister-Ling


语言学交流平台(又名“林先生”)的编辑目前来自上海各高校语言学专业,日常推送的内容包括语言学科普小品文以及各类信息(包括全国范围内的语言学讲座、会议、沙龙、征稿、课程表以及招聘等等),另有国外最新期刊文章报导。

目前我们设有两个专题板块: 语言类型学和计算语言学,更多板块正在筹建中。

语言学交流平台关注世界语言学前沿,希望将语言的各个范式放入一个大框架中,建立沟通平台,同时融合语言学理论与信息技术,使语言学理论能更好地创造价值,反哺社会。

语言学交流平台配有相应的QQ群:

主群:340047791

语言类型学:425794110

计算语言学:147393670


欢迎您的加入!