16.5: On the nature of epistemic modality
- Last updated
- Save as PDF
- Page ID
- 138716
As mentioned in our discussion of types of modality in §16.1, the most basic distinction is between epistemic modality and all the other types. Hacquard (2011: 1486) observes that “epistemics deal with possibilities that follow from the speaker’s knowledge, whereas roots deal with possibilities that follow from the circumstances surrounding the main event and its participants.”
Epistemic modality is often said to be “speaker-oriented”,10 because it encodes possibility or necessity in light of the speaker’s knowledge. Non-epistemic modal marking reflects some facet of the circumstances surrounding the described situation or event, such as the requirements of an authoritative person or code (deontic), or the agent’s abilities (dynamic), goals (teleological), or desires (bouletic).11 Van der Auwera & Ammann (2013) use the term situational as a cover term for the non-epistemic types, which seems like a very appropriate choice; but the term root is firmly established in linguistic usage.
Epistemic modality also differs from root modality in its interaction with time reference. Epistemic modality in the present time tends to be restricted (at least in English) to states (30a) and imperfective events, either progressive (30c) or habitual (31a). Deontic modality occurs freely with both states and events, but tends to be future oriented; deontic readings are often impossible with past events (31c, 32c). Epistemic necessity (must) is typically impossible with future events (31b), which is not surprising because speakers generally do not have certain knowledge of the future. Epistemic possibility (may), however, is fine with future events (32b).
(30) a. Henry must be in Brussels this week. [epistemic or deontic]
b. Henry must write a book this year. [future; only deontic]
c. Henry must be writing a book this year. [present; only epistemic]
(31) a. Mary must attend Prof. Lewis’s lecture every week. [epistemic or deontic]
b. Mary must attend Prof. Lewis’s lecture tomorrow. [only deontic]
c. Mary must have attended Prof. Lewis’s lecture yesterday. [only epistemic]
(32) a. Mary may attend Prof. Lewis’s lecture every week. [epistemic or deontic]
b. Mary may attend Prof. Lewis’s lecture tomorrow. [epistemic or deontic]
c. Mary may have attended Prof. Lewis’s lecture yesterday. [only epistemic]
When the modal itself is inflected for past tense, e.g. had to in (33), either reading is possible; but the scope of the tense feature is different in the two readings.12
(33) Jones had to be in the office when his manager arrived. [epistemic or deontic]
Under the deontic reading, tense takes scope over the modality: the obligation for the agent to behave in a certain way is part of the situation being described as holding true at some time in the past, prior to the time of speaking. Under the epistemic reading, the modality is outside the scope of the past tense: the speaker’s knowledge now (at the time of speaking) leads him to conclude that a certain situation held true at some time in the past. As von Fintel (2006) points out, the interactions between modality and tense-aspect are complex and poorly understood, and we will not pursue these issues further here.
Papafragou (2006: 1688) describes another kind of difference which has been claimed to exist between epistemic vs. “root” modality:
It is often claimed in the linguistics literature that epistemic modality, unlike other kinds of modality, does not contribute to the truth conditions of the utterance. Relatedly, several commentators argue that epistemic modality expresses a comment on the proposition expressed by the rest of the utterance… The intuition underlying this view is that epistemic modality in natural language marks the degree and/or source of the speaker’s commitment to the embedded proposition.
However, some of the standard tests for propositional content indicate that this is not the case: both types of modality can be part of the proposition and contribute to its truth conditions. We will mention three tests which provide evidence that epistemic modality does not just express a comment on or attitude toward the proposition, but is actually a part of the proposition itself. First, epistemic modality is part of what can be felicitously challenged, as illustrated in (34).13
(34) A: Jones is the only person who stood to gain from the old man’s death; he must be the murderer.
B: That’s not true; he could be the murderer, but he doesn’t have to be.
In this mini-conversation, speaker B explicitly denies the truth of A’s statement, but only challenges its modality. In other words, B denies □p without denying p. In this respect epistemic modals are quite different from the speaker-oriented adverbs which we discussed in Chapter 11. Those adverbs cannot felicitously be challenged in the same way, because (as we argued) they are not a part of the proposition being asserted.
Second, epistemic modality can be the focus of a yes-no question, as illustrated in (35–36). In these questions the information requested concerns the addressee’s degree of certainty, not just the identity of the murderer. The wrong choice of modal can trigger the answer “No”, as in (35), showing that modality contributes to the truth conditions of the sentence. In contrast, when an inappropriate speaker-oriented adverb is added to a yes-no question, it will not cause the answer to change from “Yes” to “No” (37).
(35) A: Must Jones be the murderer?
B: Yes, he must/#is. or: No, but I think it is very likely
(36) A: Might Jones be the murderer?
B: Yes, he might/#is. or: No, that is impossible.
(37) A: Was Jones unfortunately arrested for embezzling?
B: Yes/#No; he was arrested for embezzling, but that is not unfortunate.
Third, epistemic modality can be negated by normal clausal negation, although this point is frequently denied. It is true that some English modals exhibit differences in this regard between their epistemic vs. deontic uses. With may, for example, negation takes scope over the modal in the deontic reading, but not in the epistemic reading (38). The modal must, on the other hand, takes scope over negation in both of these readings (39).
(38) Smith may not be the candidate. [epistemic: possible that not p]
[deontic: not permitted that p]
(39) Smith must not be the candidate. [epistemic: evident that not p]
[deontic: required that not p]
However, while most English modals (including must and may, as we have just seen) take scope over negation in the epistemic reading, there are a few counter-examples, as illustrated in (40–41).14
(40) Smith cannot be the candidate. [epistemic: not possible that p]
(41) Jones doesn’t have to be the murderer. [epistemic: not necessary that p]
Examples like these show that even in English, epistemic modality can sometimes be negated by normal clausal negation. Moreover, German müssen ‘must’ takes opposite scope from English must in both epistemic and deontic readings (42).
(42) a. Er muss nicht zu hause bleiben.
he must not at home remain
‘He doesn’t have to stay home.’
[deontic: not required that p; von Fintel (2006)]
b. Er muss nicht zu Hause geblieben sein. Er kann auch weggegangen sein.
he must not at home remained be he can also away.gone be
‘It doesn’t have to be the case that he stayed home (or: He didn’t necessarily stay home). He may also have gone away.’
[epistemic: not necessary that p; Susi Wurmbrand, p.c.]
Idris (1980) states that the Malay modal mesti ‘must’ interacts with negation much like its English equivalent, in particular, that negation cannot take scope over the epistemic use of the modal. Now auxiliary scope in Malay correlates closely with word order. When the modal precedes and takes scope over the clausal negator tidak ‘not’, as in (43a), both the epistemic and the deontic readings are possible. When the order is reversed, as in (43b), Idris states that only the deontic reading is possible
(43) a. Dia mesti tidak belajar.
3sg must neg study
‘He must not study.’ [epistemic: evident that not p]
(i.e., ‘I am certain that he does not study.’)
‘He is obliged not to study.’ [deontic: required that not p]
b. Dia tidak mesti belajar.
3sg neg must study
‘He is not obliged to study.’ [deontic: not required that p]
A number of authors have cited these examples in support of the claim that epistemic modality always takes scope over clausal negation.15 However, corpus examples like those in (44) show that the epistemic use of mesti is in fact possible within the scope of clausal negation.
(44) a. Inflasi tidak mesti ber-punca dari pemerintah…
inflation neg must mid-source from government
‘Inflation does not have to have the government as its source…’
(… it can arise due to other reasons as well)
[epistemic: not necessary that p]16
b. Hiburan itu tidak mesti mem-bahagia-kan, tapi kebahagiaan itu sudah pasti meng-hibur-kan.
entertainment that neg must bless/make.happy but happiness that already certain comfort/entertain
‘Entertainment does not necessarily bring happiness, but happiness will definitely bring comfort.’
[epistemic: not necessary that p]17
So we have seen evidence that epistemic modality can be negated by normal clausal negation in Malay, in German, and even in English. Once again, this is not true of evaluative or speech act adverbials: they are never interpreted within the scope of clausal negation, as we demonstrated in Chapter 11. Taken together, the three types of evidence we have reviewed here provide strong support for the conclusion that epistemic modality is a part of the propositional content of the utterance and contributes to the truth conditions.
10 Bybee et al. (1994).
11 These examples illustrate the most commonly recognized types of modality; but as von Fintel (2006) observes, “In the descriptive literature on modality, there is taxonomic exuberance far beyond these basic distinctions.”
12 Have to is used here because true modal auxiliaries in English cannot be inflected for tense.
13 Cf. Papafragou (2006: 1698).
14 The same scope holds for the “root” readings of these examples as well.
15 See for example de Haan (1997); Drubig (2001).
16 wargamarhaen.blogspot.com/201...mbat-lagi.html