The concept of hedges can be probably traced back to Zadeh’s (1965) work on logic of fuzzy sets. According to Zadeh, a fuzzy set is “a class of objects with a continuum of grades of membership” (p. 338), and there are not precisely defined criteria of membership in the real physical world. In other words, most classes of objects in the natural world do not have sharp boundaries. It is highly unlikely to constitute a set of ‘beautiful women’ or ‘tall men’ in the light of the usual mathematical sense. In psychology, Rosch (1973) conducts experiments
13
to test whether category membership is perceived as a clear-cut issue or a matter of degree, and the results consistently show the latter. For example, Rosch finds that there is a well-defined hierarchy of ‘birdiness’ which emergers in the subjects’ minds, in that robins are typical members of birds, eagles are less typical, and penguins are further away from the central prototype. Her further experiment demonstrates that while being asked to respond with ‘true’ or ‘false’ to sentences of the form ‘A (member) is a (category)’, subjects tend to respond faster on typical members than on peripheral members. That is, subjects take shorter time to respond to the sentence like ‘A robin is a bird’ than to ‘A chicken is a bird’. Rosch’s work shows that category membership seems to be in accordance with speakers’ uncertain rather than clearly defined perception. Based on Zadeh’s (1965) early work on fuzzy logic and Rosch’s (1973) work on category membership, George Lakoff (1973) relates the observations of natural phenomena in the physical world to natural language concepts. He suggests that:
…natural language concepts have vague boundaries and fuzzy edges and that, consequently, natural language sentences will very often be neither true, or false, nor nonsensical, but rather true to a certain extent and false to a certain extent, true in certain respects and false in other respects (p. 458).
The correlation between the idea of fuzzy concept and the practice of natural language brings up the notion of hedges. Lakoff (1973: 471) refers to hedges as “words whose meaning implicitly involves fuzziness -- words whose job is to make things fuzzier or less fuzzy.” An experiment is conducted to examine the effect of the hedging expression sort of in the case of birdiness:
(1) a. A robin is a bird. (true)
b. A chicken is a bird. (less true than a) c. A penguin is a bird. (less true than b)
d. A bat is a bird. (false, or at least very far from true) e. A cow is a bird. (absolutely false)
(2) a. A robin is sort of a bird. (false – it is a bird, no question, about it) b. A chicken is sort of a bird (true, or very close to true)
c. A penguin is sort of a bird (true, or close to true) d. A bat is sort of a bird. (still pretty close to false) e. A cow is sort of a bird. (false)
In Lakoff’s experiment, some subjects do not distinguish the sentences (a), (b), and (c) in example (1), and tend to regard them as all being true. However, after adding the hedge sort of, as in example (2), those subjects undoubtedly distinguish the sentence (a) from the (b) and (c). The subjects can instantly recognize the difference due to the ability to make an underlying distinction in degree of birdiness, and such a distinction is what the effect of sort of depends on. In the examples, the degree of truth is
corresponding to degree of category membership. Robins are conceptualized to be the best exemplar of birdiness since they inherit the attributes, has wings, can fly, has feathers. Dirven and Verspoor’s (1998) definition may give a better idea of what the best exemplar is. They state that “the best member, called the prototypical member or most prominent member of a category, is the subtype that first comes to mind when we think of that category” (p. 16). Since subjects clearly know that robins are the
prototypical members of birds, the sentence (2a), with the predicate modifier sort of, violates the knowledge they have. They then can determinately judge it as a false statement. Lakoff’s experiment reveals that people do make the full range of
distinctions in the birdiness hierarchy, with robins judged to be the most representative member of the bird category, chickens less typical, penguins far less typical, and bats hardly at all. The following examples may better demonstrate Lakoff’s assertion that hedges serve both attenuating and reinforcing functions. That is, how hedges function to increase fuzziness, to make things fuzzier, and to decrease fuzziness, to make things less fuzzy. The practice of these functions is correlated to the concept of category
15 membership existing in human’s cognition.
(3) a. ＊A penguin is a bird par excellence.
b. A penguin is sort of a bird.
Sentence (3a) is false since the hedging expression par excellence requires the highest degree of category membership, yet a penguin is not the prototypical exemplar of birds.
However, if we use the hedge sort of to describe a penguin’s birdiness, it becomes acceptable, as in (3b). With its implication of vagueness and imprecision, the hedge sort of is entrusted to describe more accurately an untypical member of a category. The status of a penguin is then modified by the hedge sort of to be a less prototypical member in the birdiness hierarchy, and which matches human’s conceptual knowledge in that a penguin is not a typical bird but is classified as a partial member in the bird category. In this sense, the hedge sort of makes the category boundary of birdiness fuzzier. Although Lakoff (1973) initiated the notion of hedges in terms of semantics, he does point out that pragmatics plays an important part in constituting truth values of sentences. The following example extracted from Lakoff’s (1973: 474) best
demonstrates his view toward the significance of pragmatics:
(4) a. John is a bachelor.
b. John is a regular bachelor.
Sentence (4b) would not be said of a bachelor as we normally perceive, but rather it might be said of a married man who acts like a bachelor. It is because the word regular picks out certain ‘metaphorical’ properties and those properties make ‘bachelor’ deviate from its original literal meaning. In other words, regular asserts the connotations of ‘bachelor’. The connotations of words are usually recognized as part of pragmatics, and they are deemed to be irrelevant to truth values of sentences. However, the truth value of (4b) not only depends
on the literal meaning of the predicate, but also strictly on the connotation. It indicates that
“semantics cannot be taken to be independent of pragmatics, but that the two are inextricably tied together” (p. 474). Later on, Lakoff touches upon the notion on how hedges interact with performatives. He argues that if a sergeant says (5):
(5) You might want to close that window, Private Snurg.
It would be appropriate to describe such a situation by the following two conditions:
(6) a. Strictly speaking, the sergeant didn’t order the private to close the window.
b. Essentially, the sergeant did order the private to close the window.
Therefore, Lakoff indicates that hedges obviously interact with felicity conditions for utterances and with rules of conversation. He then in this same article attributes to Robin Lakoff (personal communication) the observation that certain verbs and syntactic
constructions convey hedged performatives which modify the force of a speech act, as in (7):
(7) I suppose (guess/think) that Harry is coming.
The idea of hedged performatives becomes one way of widening the concept of hedges.
Fraser (1975) follows up on Robin Lakoff’s observations and introduces the term
“Hedged Performative” (p. 187). He proposes that when certain performative verbs such as apologize, admit, and request are preceded by specific modals, such as can, should, and must, or semi-modals. such as have to, intend to, would like to, the sentences result in an attenuated illocutionary force of the speech act designed by the verb:
(8) a. I should apologize for being rude on the phone yesterday.
b. I have to admit that you have a point.
c. I must request that you sit down
17
The models and semi-models in example (8) are considered to function as hedges. After applying the model should, example (8a) is still an apology, but one less strong. Brown and Levison (1978, 1987: 145) elaborate on Lakoff’s definition of hedges and assert that hedges can modify the degree of membership of a predicate or noun phrase in a set, and “it says of that membership that it is partial, or true only in certain respects, or that it is more true and complete than perhaps might be expected.” They further develop hedging concept in the speech act. They argue that hedging acts not only on the propositional content, as considered by Lakoff (1973), or on the performative clause, as proposed by Fraser (1975), but also on the illocutionary force and speaker commitment in general. Brown and Levison incorporate Goffman’s (1967) notion of ‘face’ into their theory of politeness. They see politeness as a social reason to deviate from Grice’s (1975, 1978) conversational principles. Hedges are regarded by Brown and Levison as expressions which show that the speaker does not adhere to Grice’s (1975, 1978) maxims, and the use of hedges mainly as means of negative
politeness. In Brown and Levison’s work, both the attenuation and the reinforcement aspects of hedges are maintained though they focus more on the former.
Since hedges are pervasive in language use, Prince et al. (1982), defining hedges as linguistic means that make things fuzzier, makes a further distinction between different types of hedges: “fuzziness within the propositional content proper” and “fuzziness in the
relationship between the propositional content and the speaker” (p. 85). The former type is referred to as “approximators”, as sort of in His feet were sort of blue, and it is further divided into “adaptors” (e.g. sort of ), which modify a term to fit a non-prototypical instance, and
“rounders” (e.g. approximately), which indicate a term that is not precise. The latter type is referred to as “shields”, as I think in I think his feel were blue, and it is further divided into
“plausibility shields” (e.g. I think, I guess, probably), which are expressions related to doubt, and “attribution shields” (e.g. according to her estimates, as far as I could tell from the
mother), which are expressions attributing the responsibility of a proposition to someone other the speaker. According to Prince et al. (1982), only “approximators” affect the truth value of propositional content, whereas “shields” affect the degree of speaker commitment but leave the truth condition of proposition unchanged.
A similar two-way distinction of hedges is proposed by Hübler (1983), who
distinguishes linguistic indetermination into “understatements” (expressions of “phrastic”
indetermination) and “hedges” (expressions of “neustic” indetermination), under the cover term “understatement”. Hübler presumes that there are oppositions to all sentences, and hedging devices are utilized by the speaker to prepare for possible opposition from the hearer.
This underlying assumption manifests the interactional aspect of hedging phenomena, and shows that Hübler’s framework is pragmatic by nature. “Understatements” concern the propositional content of a sentence and its correspondence with the outside world, whereas
“hedges” concern the speaker’s attitude to the hearer regarding the proposition. Thus, the understatement a bit as in the sentence It is a bit cold in here modifies the “phrastic”, that is the propositional content, while the hedge I suppose as in It is cold in Alaska, I suppose, modifies the “neustic”, that is the speaker’s attitude toward the utterance (p. 11). Hübler’s (1983) division of hedges is found to be quite similar to that of in Prince et al.’s (1982).
Prince et al.’s “approximators” corresponds to Hübler’s “understatements,” and their
“plausibility shields” corresponds to Hübler’s “hedges” (Markkanen and Schröder 1997, Varttala 2001, Fraser 2010). Caffi (1999, 2007) lately extend the bipartite distinction of Prince et al. (1982) and Hübler (1983) to a tripartite model.
In her framework of mitigation devices, Caffi (2007: 16) ascribes the concept of mitigation, a synonym of attenuation, “is a result of a weakening operation on one of the interactional parameters.” Mitigation serves to smooth interactional management in that it
“reduces risks for participants at various levels, e.g. risks of self-contradiction, refusal, losing
19
face, conflict, and so forth” (Caffi 1999: 882), and it is argued to work in a “multi-layered and multi-dimensional way, simultaneously affecting a plurality of linguistic levels and interactional dimensions” (p. 883). Based on different scopes of mitigation, Caffi distinguishes mitigating mechanisms into “bushes”, “hedges”, and “shields”. The three domains where these mitigating devices operate on are the proposition (and within it,
reference and predication), the illocution (e.g. illocutionary verbs, expressions modifying the strength of the illocution), and the deictic origin of the utterance (e.g. the I-here-now of the utterance). In “bushes”, the focus of the mitigating device is on the propositional content, which is typically made less precise. “Bushes” then refer to lexical expressions which reduce precision of the propositional content; therefore, it corresponds with Prince et al.’s (1982)
“approximators” and Hübler’s “understatements”. An example extracted from Caffi’s Italian corpus of doctor-patient and psychotherapeutic interaction demonstrates how bushes (e.g.
diminutive suffix –ino) work to attenuate the speaker’s utterance (in the following examples, the capital D stands for the doctor):
(9) D. le do uno sciroppino da prendere.+due eucc
I’ll give you ((lit. I give you)) a cough syrup+DIM to take.+two table sp
(Caffi: 2007: 99)
The doctor utters example (9) as he is about to write a prescription. The diminutive suffix – ino in sciroppino (lit. ‘a cough syrup+DIM) makes the reference act in the prescription fuzzy.
With its semantic feature [-serious], the diminutive suffix –ino functions to reduce both the severity and the unpleasantness of the therapeutic prescription. It not only downgrades the imposition on the patient to comply with the prescription but also indirectly mitigates her possible worries about her pathology. In “hedges”, the scope of mitigation centers on the illocution, i.e. on illocutionary force indicating devices, and the speaker’s commitment to the propositional content. The notion of Caffi’s “hedges” covers Prince et al.’s (1982)
“plausibility shields” and Fraser’s (1975) “hedged performatives”. “Hedges” are illustrated in the following examples:
(10) D. […] io le proporrei se vuole una medicina apposta per vedere se riescxx – se riesco a farla dormire=
[…] I’d propose to you if you want a special medicine to see if I caxx – if I can make you sleep= (Caffi: 2007: 102)
(11) D. Probabilmente è: - dove c’è l’attacapanni – probabilmente è una::
conseguenza di
un problema intestinale: che è cominciato con I’influenza eh:?
probably it is – where the clothes-stand is – it is probably a consequence of an intestinal problem: that began with the flu eh:? (Caffi 2007: 103)
io le proporrei (‘I’d propose to you’) in sentence (10) is a hedged performative since the verb in the conditional mood is a weaker form than the performative expression ‘I propose”. In the doctor-patient relationship, although the doctor acts a strong role as the healer, he formally downgrades his directive to a proposal which is up to the patient to accept. In (11), the mitigation, which works on the aspect of the illocution, displays the speaker’s epistemic commitment to the propositional content. Probabilmente (‘probably’) is employed to weaken the speaker’s degree of certainty about the proposition. In other words, the doctor
downgrades his diagnosis to a hypothesis, “nothing more than an attempt to trace abductively the temporal and causal sequence of facts” (p. 103). “Shields” are devices to avoid self-ascription to the utterance and shift responsibility to another source, which is comparable to Prince et al.’s (1982) “attribution shields”. In “shields”, according to Caffi (2007: 106), “the act is not mitigated by explicit linguistic devices, but rather it is dislocated by means of backgrounding and de-focalizing strategies, or even by the deletion of the utterance source.”
“Shields” appear to realize some strategies of displacement, which associate with the three components of deictic origin, the “I”, “here” and “now”. “Shields” in the “I” component are cases where the act is ascribed to someone else or cases where the author is simply deleted, as
21
in impersonal constructions or agentless passive constructions. The other two means of replacement are interconnected since the displacement involves the “here and now”
component of the utterance. “Shields” in the “here and now” component are cases where a narrative replaces a sequence focusing on the present context. Caffi claims that these
“shields” can either work locally, at the level of a single utterance, as in (12), or globally, as textual strategies, as in (13). Example (12) illustrates a ‘non-I’, a non-ego strategy, whereas (13) displays a “I, not-here, not now’ mitigating device (C stands for the client and T stands for the therapist).
(12) D. c’è un’i:perplasia estrogenica – c’è scritto qui.
there’s an estrogenic hyperplasia – it is written here. (Caffi 2007: 107)
(13) C. e poi: dall’85 I due anni diciamo in cui: stavo male ma non: per altri versi ma non Così male come sto adesso sono comunque riuscita a continuare a avorare mi hanno cambiato la respo*nsabilità+
T. *adesso sta male? +in questo momento?
C. be’ adesso io sono qua e sono entrata per degli accertamenti: ulteriori=
T. =no dico adesso ora qui.
C. sì.
T. sta male?
C. be’ – certo.
T. perché parla con me?
C. no: non mi dà nessun fastidio parlare con lei.
C. and then from 85 to 87 the two years when: let’s say I felt bad but not in other respects but not so bad as I feel now anyway I managed to keep working they changed my du*ties+:
T. *now are you feeling bad?+ at the moment?
C. well now I’m here and I’ve come in for further tests=
T. =no I mean now at this very moment here.
In (12), the mitigation focuses on the utterance source, the deictic origin. By shifting responsibility to some other impersonal source, the doctor is able to weaken his personal commitment to the diagnosis. Furthermore, the utterance makes no reference to either interlocutor: it is not only a ‘not-I’ but also a ‘not-you’ statement. Example (13) is a transitional case between deictic shields. This interaction demonstrates a ‘narrativization’
shield, i.e. a global strategy of de-actualization o f a topic. In this specific context (the first session in a course of psychotherapy), the topic is the patient’s present state; however, with the process of de-actualization2, a present feeling is shifted to the past and is told in a rather detached way than being enacted. Caffi (1999: 905, 2007: 118) concludes that “mitigators mitigate because they manage speech act responsibility in different ways.” “Bushes” and
“hedges” are scalar devices which work along a scale of degrees of epistemic commitment to the proposition and degrees of endorsement of the illocution respectively, whereas shields are yes-no devices (e.g. ‘I’/‘not I’, ‘now’/‘non-now’) which work directly on the core of the utterance act, the deictic origin, the formal support of subjectivity.