The modal verbs can and may in English and Spanish

CONTENTS

Abstract

Introduction

1. Modality and Mood
1.1. The Difference between Mood and Modality
1.1.1. What is Mood?
1.1.2. What is Modality?
1.2. Epistemic Modality and Evidentiality
1.3. Root Modality
1.3.1. Deontic Modality
1.3.2. Dynamic Modality
1.4. The Root – Epistemic Distinction (as shown in CAN and MAY)
1.4.1. An Analysis of the Dichotomy Root – Epistemic

2. Modal Verbs
2.1. The Morpho-Syntactic Properties of the English Modals
2.2. The Structural Positions of Modals
2.3. The Complement of Modals
2.4. Time Reference for CAN and MAY
2.5. Scope Properties of Modal Verbs
2.6. Modality and Negation

3. A Semantic Analysis of the English Modals
3.1. The Unitary Semantic Approach vs. Ambiguity – Polysemy View
3.1.1. The Ambiguity View (Palmer)
3.1.2. The Polysemy View (Sweetser)
3.1.3. The Monosemy View (Kratzer, Papafragou)
3.2. Quantification and Scope
3.3. Modal Restrictors
3.4. Semantics for Modal Operators

4. A Pragmatic Point of View
4.1. The Pragmatics of Root Modality
4.1.1. Derivation of Root Interpretations of CAN and MAY
4.2. The Pragmatics of Epistemic Modality
4.2.1. The Metarepresentation Hypothesis
4.2.2. Derivation of Epistemic Interpretations of CAN and MAY
4.3. ‘Speech-Act’ Modality (Factual MAY)

5. Expressing Permission and Possibility in Spanish
5.1. Modality Types
5.2. A Syntactic Account
5.3. A Semantic Account
5.3.1. Complements
5.3.2. Subject Restrictions
5.3.3. Tense
5.3.4. Negation

Concluding Remarks

References

Abstract

Modality is a semantic concept that covers notions such as possibility, probability, permission, ability, volition, necessity and obligation. The class of modals is in many languages both syntactically and semantically highly irregular and unpredictable: modals frequently have idiosyncratic conjugational patterns and are subject to highly specialized syntactic rules. One of the main characteristic of modal verbs is their relatively imprecise and indeterminate meaning, their ambiguity: the same modal can be deontic (i.e. based on rules and regulations), but it may also involve processes, sets of knowledge or belief systems, and thus get an epistemic interpretation.

In order to define the class of modals or to provide a set of environments in which a modal may be correctly or appropriately used, one must refer to many levels of language: the purely syntactic environment, as well as the logical structure, the context of the utterance, the assumptions that are shared by the speaker and the addressee, the social situation assumed by the participants in the discourse, the impression the speaker wants to make on the addressee, and so on. There is also the question of the appropriate context environments, that is, the semantic-pragmatic issue. Therefore, a complete analysis of a particular modal can only be achieved by looking both at its syntactic features and at its semantic structure; in other words, the syntax of a modal verb is based on its semantics, and these two dimensions are inseparable.

Introduction

This paper attempts to provide a complete analysis of the English modals CAN and MAY (and of their Spanish equivalent – PODER), by discussing the different expressions of root and epistemic interpretations, and also from the point of view of the possibility - necessity distinction. My main perspective is that modal expressions in natural language mostly express incomplete or underspecified contents, and these contents need to be pragmatically manipulated in order to yield the rich modal concepts that formally support inference.

The theoretical background of the present paper is provided mainly by the works of Larisa Avram (for the morpho-syntactic properties and the structural position of modals), Jonny Butler (his attempt at unifying Kratzer’s semantic analysis of modals as propositional operators with Chomsky’s minimalist ideas in order to account for the scope positions), Tim Stowell (for time reference), Annabel Cormack and Neil Smith (for the negation part), and, of course, Anna Papafragou, who, on the basis of Kratzer’s original proposal that the English modals have a unitary semantic content, resolved the earlier ambiguity / polysemy claims held by Palmer and Sweetser, and designed a plausible pragmatic derivation of the modals’ various contextual interpretations.

The present paper consists of 5 chapters, and is structured as follows:

The first chapter proceeds by defining the notions of mood and modality, and establishing the differences between them: while mood is a grammatical category, modality is a semantic / pragmatic category that encodes two major types, that is, root and epistemic. The general classifications include deontic and dynamic modality in the category of root (circumstantial) reading, and alethic (logical modality) and evidentiality as subclasses of the epistemic interpretation. This initial chapter ends by establishing the distinction between the root and epistemic readings of modals, analyzed in the particular case of CAN and MAY.

The second chapter presents mainly a syntactic approach of modality: first, I point out the morpho-syntactic properties that modals (as semi-auxiliaries) share with the “primary auxiliaries” have, be and do (NICE properties), and those that differentiate them; then, I try to account for the structural positions of the modals (following Avram’s proposal): under VP (dynamic reading), in the functional domain under Tense (deontic reading), and under a node Mood2 (epistemic reading); thirdly, I show how the complement of a modal can influence its interpretation (there are complements that force an epistemic interpretation, just as others force root interpretations); further on, I present the interaction of the modals CAN and MAY with tense and aspect, the way in which modals scope in terms of their reading (epistemic necessity / epistemic possibility and root necessity / root possibility), and the way in which negation scopes over or under modality (Echoic, Polarity, Adverbial negation), affecting either the proposition (in epistemic modality), or the modality (in root modality).

The third chapter provides a semantic analysis of the modal verbs: firstly, on the basis of Papafragou’s (and Kratzer’s) proposal, I demonstrate that modals are neither lexically ambiguous (as Palmer claimed), nor polysemous (Sweetser’s theory), they have a unitary common core, and only the interpretation is pragmatic, depending on the context; secondly, starting from Kratzer’s tripartite underlying modal operators(the modal relation, the modal base and ordering source), I assume that the semantic content of modals consists of two components: a logical relation R (entailment or compatibility) and a domain D of propositions; finally, I show that the modals CAN and MAY differ in their semantic content, MAY turning out to be more general than CAN, as its D-value is unspecified, the slot in its lexical semantics is empty, and therefore it has to rely on processes of pragmatics for completion.

The fourth chapter starts by discussing the root meanings of the two modals (potentiality, ability and permission for CAN; permission and possibility for MAY) from a pragmatic point of view, it continues by showing that epistemic interpretations make use of metarepresentations in a way that root interpretations do not, and finally it presents the special category of “speech-act” modality / “factual MAY” as seen by Sweetser (1990) - an extension of deontic MAY, by Papafragou (2000) - a subcategory of the epistemic use, and by Sugiyama (2003) - an independent category which is “neither an extension of deontic MAY, nor the same as epistemic MAY, [for it] concerns not only speaker’s beliefs, but also the addressee’s beliefs.”

The fifth chapter provides a similar analysis of the Spanish modal verb PODER – the equivalent of CAN and MAY. First, I present briefly the types of Spanish modality, and then I account for a syntactic and semantic analysis, pointing out the similarities (both English and Spanish modals select for infinitival clauses, past tense scopes above root modals, but under epistemic ones) and the differences (Spanish modals do not meet the NICE properties, carry tense-marking and subject-verb agreement and can co-occur with other modals) that modals display cross-linguistically.

CHAPTER 1. Modality and Mood

1.1. The Difference between Mood and Modality

Although these two notions are related in origin, and to some extent in meaning, they represent two separate components of grammar: while mood is a grammatical category, modality is a semantic / pragmatic concept. Originally, they were both connected with the mode, manner, or fashion of saying something, but as the language advanced from Old English to Modern English, they have developed more specialized uses. Thus, the word “modal” is used in logic and philosophy to refer to propositions involving the affirmation of possibility and impossibility, existence and non-existence, contingency and necessity, and this meaning has been taken into grammar, as well.

. The topic of mood and modality (MOD) is a difficult aspect of language description because - among other reasons - the inventory of modal meanings is not stable across languages, moods do not map neatly from one language to another, modality may be realized morphologically or by free-standing words (such as adverbs), and it interacts in complex ways with other modules of the grammar, like tense and aspect. Describing MOD is all the more difficult if the attempt is to develop a unified approach that would provide cross-linguistic coverage (McShane et al.)

Morphology and syntax differ significantly from language to language. If, for instance, we defined modality as a category sharing the morpho-syntactic features of the English modal verbs, the definition would only suit the very same modal verbs of English, and, to a lesser extent, modals in some related languages; but for a vast majority of languages, this definition would be pointless. A semantically based definition, on the other hand, offers ground for cross-linguistic validity. Even speakers of completely different languages share the same basic experiences and needs, that is, meanings. This makes meaning more universal than a particular syntactic or morphological structure. (Narrog, 2005)

Mood and modality represent the linguistic expression of the speaker’s attitude toward an utterance or an event that may or may not take place. Subsequently, it is mainly the speaker’s perception that influences his approach, which can be either objective or speaker-oriented (subjective, participative).

1.1.1. What is Mood?

Mood has been defined as: “the grammatical category which expresses the degree or kind of reality assigned to a sentence” (Trask 1997), “a system of inflections on verbs: indicative-subjunctive-imperative” (James 1986), “a manifestation of modality” (Georgi & Pianesi 1998).

Semantically, the category of mood distinguishes between factual (related to facts in reality – the indicative mood) and non-factual (the subjunctive, the infinitive and the imperative):

I. The sentence (or notional) mood -- is a semantic evaluation of the clauses or utterances, with respect to the standard (truth conditions) constituted by assertions. Descriptions of English grammar usually recognize up to four sentence moods:

1) The indicative mood – it is also called “declarative”, because it simply declares, asserts something, it is factual, descriptive, presenting the world “as it is”; these statements express actions that actually did take place, are taking place or will take place in real time (are deictic).

E.g.: Romania entered the EU on January 1, 2007.

The interrogative mood – is included within the indicative category. It expresses questions, which usually are marked – both in written (graphically, the question mark) and in speaking (a rising intonation). E.g.: Where did you park your car ?; but this is not always the case, as there are other ways of asking questions.

E.g.: I'd like to know the train times for Sunday, please.

2) The imperative mood - expresses directives, such as orders, commands, instructions, requests, invitations etc. It may be graphically marked (exclamation mark), and in speaking it is usually observed a firm tone.

E.g.: Do not touch that button! ; Switch the appliance off and remove the plug from the socket.

The subjunctive mood – the word “subjunctive” means “placed underneath, subordinated, added at the end”, for in most languages this mood appears only in subordinate clauses. It is a non-factual mood used when the content of the clause is being doubted or supposed rather than definitively asserted (as opposed to the indicative mood), and expresses uncertainty, unreality, hypotheses, wishes.

It can have two forms:

(i) synthetic – formed with past simple or infinitive. E.g.: He acted as if he were crazy; I demand that this gate be opened; Long live the queen!;

(ii) analytic – with the auxiliary “should”. E.g.: He requested that I should go.

3) The infinitive mood – the dictionary form of a verb.

II. The verbal (or grammatical) mood – is a morpho-syntactic category encoded by means of the “mood” feature. It distinguishes between:

1) Synthetic - that is, the mood of verbs, simplified in a binary subjunctive vs. imperative classification.

2) Analytic mood - the modal verbs.

1.1.2. What is Modality?

Modal sentences cannot be understood apart from considerations of their being anchored in some social context; this is why the definition and description of modals has been one of the most pervasive and persistent problems of linguistics, spilling over as well into related disciplines such as philosophy or psychology (Zdrenghea & Hoye,1995)

“Modal expressions allow us to talk (and modal concepts allow us to think) about states of affairs which are not present in the current situation and may never occur in the actual world” (Papafragou, 2000). Bybee & Fleischman (1995) defined modality as “the semantic domain pertaining to elements of meaning that languages can express in various ways, morphologically, lexically, syntactically, or via intonation”. The Japanese linguist Heiko Narrog (2005) presented the issue of modality from a linguistic point of view, in three different ways: (i) modality as the expression of the speaker’s attitude, or the expression of his subjectivity, opinions and emotions; (ii) modality as including all linguistic expression outside the proposition; (iii) modality as the expression of realis vs. irrealis, or factuality distinctions.

Judging from the multitude of diverging definitions that have been given to modality, one can easily see that it is a highly controversial issue. In the introduction of his article “On Defining Modality Again”, Heiko Narrog (2005) states that “there is hardly any grammatical category (…) under the label of which a wider range of phenomena has been studied”. As Alan King (1997, apud Narrog 2005) put it, “undeniably, modality is a subject of considerable complexity (…) and is poorly defined despite considerable recent progress”. Some authors emphasize the relativity of definitions: for instance, van der Auwera (1998) says that “modality and its types can be defined and named in various ways. There is no one correct way”. The Japanese linguist Hiroshi Kudo (1989, apud Narrog 2005) is even more drastic in his statements: “modality has become the dustbin of grammatical categories”.

The controversy regarding modality does not end with the problem of defining it, because the classification of modal verbs seems to be just as complicated a matter. Different authors have structured modality and its meaning in different ways, creating patterns that bear their names, but cannot be generally accepted. For instance, Rescher (1968) imagined a division in 8 modalities: alethic, epistemic, temporal, boulomaic, deontic, evaluative, causal, likelihood, whereas Coates (1983) designed a scheme of 12 modalities: strong obligation, weak obligation, permission, volition, prediction, ability, dynamic possibility, epistemic possibility, strong inference, weak inference, hypothesis, quasi-subjunctive. As for the modal meanings, Palmer (1990) figured 8 of them: (i) epistemic possibility (may), (ii) epistemic necessity (must), (iii) epistemic W/S (will); (iv) deontic possibility (may, can), (v) deontic necessity (must), (vi) deontic W/S (shall); (vii) dynamic possibility (can), (viii) dynamic W/S (will).

The one categorization that accounts for most of the approaches is:

MODAL VERBS^[1]

illustration not visible in this excerpt

Deontic Dynamic

Subject- Neutral

oriented

Bybee (1985), proposes the supercategory of agent-oriented modality (the equivalent of root modality), as opposed to speaker-oriented (speech acts that aim at getting something done: imperatives, optatives, permissives) modality, and applying to all modalities in which conditions are predicated on an agent (obligation, desire, ability, permission and root possibility).

Five years later, in 1990, Palmer introduces a third term, namely the subject-oriented modality (the equivalent of dynamic modality only), as he thought that “epistemic and deontic modality relate to the speaker”, they are concerned with the speakers’ judgments and desires, and they should be separated from the dynamic modality, which refers to the will of the subject, rather than the opinions (epistemic) or attitudes (deontic) of the speaker (and addressee). Later, he divides dynamic modality, as well, into subject-oriented (e.g.: I can ride the bike) and neutral (= “it is possible / necessary for”).

1.2. Epistemic Modality and Evidentiality

The word comes from the Greek “episteme” (meaning “knowledge”), therefore Lyons (1977) asserts that “it is concerned with matters of knowledge and belief”, Coates (1983) states “it is concerned with the speaker’s assumptions or assessment of possibilities and indicates the speaker’s confidence (or lack of confidence) in the truth of the proposition”, James (1986) calls it “theoretical manner of representation” or “theoretical modality”, and Jacobsson (1994) – “truth-oriented, attitude modality”. According to Kratzer (1991), “in using an epistemic modal we are interested in what else may or must be the case in our world, given all the evidence available. Epistemic modality is the modality of curious people like historians, detectives and futurologists. A historian asks what might have been the case, given all the available facts.”

It is extrinsic or “extra-propositional”, expressing the speaker’s attitude towards the content of a proposition; Biber (1999) – to whom epistemic modality includes dynamic modality - points out two typical structural correlates of extrinsic modal verbs: the subject is usually non-human, and the main verb usually has a stative meaning. The commitment of the epistemic statements goes from strong to weak factuality: “It is certain / probable / likely / possible that he …”

Epistemic modality can be expressed by (i) adjectives or adverbs of modality (certain, likely), (ii) propositional attitude verbs (know, believe, think, doubt), (iii) modal verbs (must, might, could, need).

Epistemic modals encode general possibility / probability, or necessity, as in “It may rain tomorrow” (“it is possible that it will rain tomorrow”) or “I must have left my book on the desk” (“it is necessarily the case that I have left my lecture notes on the streetcar”). Futurate will and shall are also epistemic modals rather than future tense markers. Also, the epistemic MAY and CAN may have readings of politeness.

We may speak of: (i) epistemic necessity - logically entailed by what is known, practical inference, deduction; (ii) epistemic possibility - compatible with what is known, speculation, prediction; (iii) evidentiality – the evidential basis for what is said.

While an epistemic modal evaluates evidence and assigns a confidence measure to the speaker’s utterance (which can be high, diminished, or low), an evidential merely asserts that there is evidence to back up the speaker’s utterance, but refuses to interpret it in any way.

According to Palmer (1986), epistemic modality and evidentiality are “propositional modality”, an irrealis category dealing with the degree of commitment on the part of the speaker to the speech utterance. He designates four ways in which a speaker may indicate that he is not presenting what he is saying as a fact, but rather:

(i) he is speculating about it
(ii) he is presenting it as an induction: He must be rich.
(iii) he has been told about it: I was told that he was rich.
(iv) it is a matter only of appearance, based on the evidence of (possibly fallible) senses: I saw that he was rich.

Thus, Palmer shows that (i) deals with pure epistemic modality, and (ii) - (iv) deal with evidentiality, namely induction / inference, hearsay, and sensory evidence. However, the relation between truth/certainty and evidentiality is not straightforward; rather, one may say that evidentials are used whenever the speaker wishes to state a fact that occurred beyond doubt, but whose causes are not known to him. As far as the truth value of the sentence is concerned, it can be generally accepted that direct evidence (e.g. visual and auditory evidence, i.e. epistemic modality) is more believable than indirect evidence (e.g. inference and hearsay, i.e. evidentiality).

1.3. Root Modality

The term seems to have been coined by Hoffman in his 1976 “Past Tense Replacement and the Modal System”.

Although the root modals represent the non-epistemic sense of modals, they can also express what is possible or necessary, but with reference to some non-featural notions such as capability or permission, rather than pure probability. Jacobsson (1994) called it “agent-oriented, influence modality”, and James (1986) – “practical manner of representation”, or “practical modality”, that refers to “powers of volition” and makes a representation that the world has to match.

Using a circumstantial (=root) modal, we are interested in the necessities implied by or the possibilities opened up by certain sorts of facts. Circumstantial modality is the modality of rational agents like gardeners, architects and engineers. An engineer asks what can be done given certain relevant facts. (Kratzer 1991) This kind of information is generally supplied contextually.

In contrast to the extrinsic epistemics, the root modality has been called “intrinsic modality” since it forms part of the semantic content of the proposition. The unity of root modality is shown by the syntactic patterns in which it appears: usually an animate subject, an agentive verb and often a passive infinitive.

The most frequently discussed kinds of root modals are deontic and dynamic.

1.3.1. Deontic modality

The word comes from the Greek “deon”, meaning “duty”, therefore it refers to concepts like obligation, permission, prohibition.

Deontic modality is discourse-oriented non-epistemic modality, concerned “with the necessity or possibility of acts performed by morally responsible agents” (Lyons 1977), “with obligation and permission” (Trask 1997). It indicates whether the proposition expressed by the sentence is obligatory or permissible according to some normative background such as law, morality, convention, etc.

It is essentially performative and deals with permission or obligation, as in “You may have the last slice of pizza” or “Everyone who has been summoned for jury duty must report to the courthouse”. By using a deontic modal, the speaker may actually “give (or refuse) permission, lay an obligation or make a promise” (Palmer 1990). Diachronically, deontic meanings come to acquire epistemic meanings, but, unlike epistemics, deontics refer to acts, not propositions.

Deontic modality can be subdivided into (i) directives (deontic possibility: “You may leave”; deontic necessity: “You must leave”), (ii) commissives (promises, undertakings: You shall be rewarded),(iii) imperatives,(iv)others: volitives, evaluatives.

Another subdivision is:

(i) possibility (permission): “You may leave”;
(ii) necessity (obligation): “You must go”;
(iii) volition: “He won't go”.

Deontic modality excludes ability (physical and mental) and desire - these are categorized as dynamic, though they typically have expression similar to that of permission and obligation. However, if a deontic modal sentence expresses the speaker’s desire, it can take an individual-level stative predicate. (e.g.: “My blind date must be tall” can mean that the speaker hopes that his or her blind date is tall, but it does not hold the obligation to be tall. Rather, it can express the speaker’s belief that the blind date being a tall person is a necessary condition on the state of the world).

The modal base of a deontic modal sentence does not have to include all the propositions that are known to the speaker (Roberts 1989). It could be a subset of what the speaker knows, or it could even be empty. By associating such a modal base with deontic modality, we can account for the fact that speakers do not necessarily believe that the situation is realizable. E.g.: “John must eat fish. But he won’t.”

Deontic modal sentences have truth values. They assert that there is an obligation or a case of permission in the current world; therefore, they are true if there is indeed such an obligation or permission, and false if there is no such obligation or permission. E.g.: in “Be tall”, the deontic modality returns a set of alternate worlds in which the proposition “you are tall” can become true. But being tall is an individual-level stative predicate expressing a property that cannot be changed by an individual. Subsequently, if the situation is to hold, it will already be doing so at the time of utterance; if the situation does not hold, it will never do so in the future. Thus, the only possible world-time pairs in which “you are tall” can become true are alternate world-time pairs contemporaneous with the moment of utterance.

1.3.2. Dynamic Modality

The word comes from the Greek “dynamis”, meaning “strength, power”. The term was first suggested by von Wright in 1951, in his “An Essay in Modal Logic”.

Dynamic modalities are concerned with properties and predispositions of persons referred to in the clause, especially by the subject NP. In the literature, dynamic modality is traditionally associated with abilities or potentials (i.e. the meaning predominant in the modal can). It also refers to physical necessity (“need to”, “have to”) or possibility and is “concerned with the volition of the subject of the sentence” (Jacobsson 1994). Nevertheless, the dynamic modality is not subjective (unlike epistemics and deontics), hence is less centrally modal. E.g.: “I can speak Spanish”, “She’ ll come, if you ask her”, “He has to come tomorrow”, “You can smoke in here” (“it is possible for you to smoke in here”).

Dynamic modality seems less of a unified category than epistemic and deontic modality; it has been subdivided into: (i) ability (“I can play tennis”); (ii) power (“Oil will float on water”); (iii) futurity (“I will/shall be 23 tomorrow”); (iv) prediction (“You will feel better after this medicine”); (v) habit (“Whenever he had a problem, he would come and talk to me”).

A subtype of dynamic modality is the “situational dynamic modality”, which involves possibilities or necessities, not of the agent participant in the state of affairs, but inherent in the situation or state of affairs as a whole. E.g.: “It can rain for weeks here”. It is opposed to “participant internal modality” on the one hand, and to epistemic modality on the other.

1.4. The Root – Epistemic Distinction (as shown in CAN and MAY)

As Sjef Barbiers observes in his “An introduction to modality and its interaction with the verbal system” (2002), a great deal of work on modality in generative syntax has concentrated on the distinction between root and epistemic interpretation of modal verbs. Roughly, epistemic interpretations involve a speaker-oriented, or, in the case of embedded clauses, matrix-subject oriented qualification or modification of the truth of a proposition, while root interpretations involve the will, ability, permission or obligation to perform some action or bring about some state of affairs (cf. Palmer 1986). Sentences often are ambiguous between the two readings. In many Germanic and Romance languages, the same set of modals can have both the epistemic and the root interpretations.

Epistemic and root readings of modals differ from each other in terms of how they interact with tense and aspect. Hofmann was the first to make the distinction between the two uses of modals in his 1976 work “Past Tense Replacements and the Modal System”. He noticed that the epistemic reading of modal verbs can appear in several syntactic patterns:

- Can co-occur with the progressive aspect:

E.g.: She may be sleeping downstairs.

- Can co-occur with the perfect infinitive (the complement has a past reference time): E.g.: She may have already left.

- There is no restriction on the subject (animate / inanimate):

E.g.: Susan / the apple may have fallen from the tree.

The root readings of modals, on the other hand, do not evince any of the characteristics above:

- Cannot occur in the progressive:

E.g.: *He can be singing now (no ability interpretation)

- Cannot occur with the perfect infinitive:

E.g.: *You may have gone, if you wish.

- There are selectional restrictions imposed on the subject (only +animate):

E.g.: *My car may leave now.

Although the difference was originally a syntactic one, it has been generally adopted in transformational treatments. In their epistemic meanings, the modals express the speaker’s state of knowledge, belief, or opinion about the proposition. In their deontic senses, the modals modify the surface structure subject of the sentence, indicating his ability, volition, obligation, etc.

Barbiers (2002) states some other differences between root and epistemic modals, such as: (i) epistemic modals are one-place predicates, taking the entire proposition as their complement, whereas root modals are two-placed predicates, involving relations between the subject and the rest of the clause; (ii) root modals are control structures and epistemic modals are subject raising structures; (iii) epistemics are generated higher than the root modals (possibly functional positions for the formers, and lexical positions for the latters); (iv) epistemic modals undergo movement at LF, and root modals don’t; (v) the epistemic interpretations select a larger complement than the root interpretations.

Collins (1974, apud Zdrenghea & Hoye, 1995) considers CAN a deontic modal when used in utterances expressing either permission or ability, whereas Pullum and Wilson (1977) regard CAN as sometimes being ambiguous between a deontic and an epistemic interpretation and state that “Elephants can kill crocodiles” can be interpreted as either of “Elephants have the ability to kill crocodiles” (dynamic), or “It can happen that an elephant kills a crocodile” (epistemic). On the other hand, Steele (1975, apud Zdrenghea & Hoye, 1995) analyses CAN quite differently: in “She can swim a mile”, the meaning of the modal is “ability”, and cannot even be considered as expressing modality since it does not indicate the possibility to swim, but rather the potential of the subject. For Steele, CAN is only a modal when it expresses permission. Against this, Antinucci and Parisi (1971, apud Zdrenghea & Hoye, 1995) argue that such “non-modal” uses of CAN are in fact special instances of epistemic modality in which the speaker expresses a deduction based on properties internal to the subject of the sentence.

These contradicting views appear because the term “modal” is used sometimes to refer to a syntactic category, and sometimes to a semantic one. The same is true of labels such as “deontic” and “epistemic”, despite the fact that there is no straightforward relationship between the semantic functions and their semantic realizations (Zdrenghea and Hoye, 1995).

The alleged polysemy of CAN is a function of the contexts in which it occurs, and the above disagreement may be resolved by postulating a basic meaning of CAN. Thus, Palmer (1974) describes the uses of CAN as of ability, of sensation and characteristic, but Perkins (1994, apud Zdrenghea & Hoye, 1995) argues that by adding the expression “at times”, all the features stated by Palmer support a characteristic interpretation. Moreover, he shows that in “He’d make a good confidence trickster – he can tell awful lies with the most innocent expression”, the modal is rather an ability CAN than a characteristic CAN. Palmer further claims that the characteristic CAN often has a derogatory sense – e.g.: “She can be very catty at times”, but Perkins offers another counterexample in “She can be very charming at times”.

Therefore, the contribution of CAN to the meaning of a sentence seems to be, roughly, to relate the event referred to in the propositional content to some external set of circumstances which is not explicitly identified, but only presupposed; the use of CAN involves a set of laws or principles according to which the relationship between circumstances and event can be interpreted. E.g.: In “John can swim”, one can consider a set of circumstances which includes a previous occasion on which John demonstrated his ability to swim; used in this sense, CAN expresses dynamic possibility. Also, it is possible to use CAN in the sense of static possibility, that is, permission, “You can go now”, “Can I borrow this pen, please?”, and respectively, lack of permission, “If you don’t eat your meat, you can’t have pudding”. There are also cases where CAN may be regarded as having an epistemic sense: “Cigarettes can seriously damage your health”. Perkins argues that it is also possible to regard such examples as expressing dynamic modality. The semantic parallel which exists between dynamic, deontic and epistemic uses of CAN helps to explain why a single form is used in a service of different meanings.

MAY and MIGHT can also be used in a dynamic sense, and in this case they felicitously substitute CAN and COULD, with very little change of meaning, if any. “The mountain top may also be climbed from other points” – here, MAY can be paraphrased as “possible for”, and has clearly a dynamic sense; the deontic reading would not be correct here.

In “I'd go further; there are things outside what may be called normal sexual intercourse”, Palmer argues that MAY is certainly not a deontic MAY, although it might seem to be accountable for in terms of “what one may be permitted to call” . An epistemic interpretation is possible: “it may be that it is called”. Alternatively, there may be blending between “can be called” and epistemic “may be that it is” . However, MAY is preferred to CAN, because of the general epistemic quality of the expression.

Generally, CAN does not express possibility in constructions with verbs in the simple present. It is apparent, however, that the meaning of possibility is regularly associated with the simple present of the main verb in construction with both CAN and COULD when the main verb is copulative: “If it’s baked just right, carrot cake can taste very good” (possibility) and “…carrot cake could taste very good” (remote possibility). Some copula verbs in construction with action adjectives which are similar to action verbs when compared to stative verbs require the deontic reading of capability: “Sometimes, he can appear strong if he wants to”. With other adjectives, however, the reading must follow the epistemic meaning of possibility: “Sometimes, he can seem so tired in the mornings”. The adjectives which force the meaning of CAN and COULD as possibility might be named “stative” adjectives (shy, weak, sick, feeble), since they are similar to stative verbs, and since they since to imply that the subject of the sentence is passive rather than active.

Nevertheless, there are cases in which the same modal is used in two different clauses, and moreover it occurs in the same surface position, and still the interpretations are different:

E.g.: Everybody must get stoned = “Everybody is required to get stoned”

Everybody must have got stoned = “It is a necessary assumption that everybody got stoned (give what I know, believe, assume)”

Thus, the distinction deontic (in the first example) - epistemic (in the second one) proofs that the interpretive class that a modal falls into in a particular instance is determined by more than the PF form or position of the modal. Subsequently, it can be concluded that the expressions “epistemic modals” and “root modals” are shorthand for “modals receiving an epistemic / root interpretation” (cf. Butler 2003).

1.4.1. An Analysis of the Dichotomy Root – Epistemic

Various explanations have been put forward for the root / epistemic distinction. Picallo (1990, apud Butler 2003) suggests that the difference is determined at the level of insertion (merge): epistemics are merged somewhere within the IP level, and roots, somewhere within the VP. McDowell (1987, apud Butler 2003) claims that is LF that is relevant, with epistemics appearing in C at that level, and roots in VP. Brennan (1997, apud Butler 2003) also makes a similar claim.

All these accounts more or less hinge on the difference being determined syntactically. However, it has been argued that it is rather determined in the lexicon, particularly in the early generative literature: for instance, Ross (1969, apud Butler 2003) claimed, influentially at the time, that epistemics are lexically one-place (intransitive) predicates, corresponding to raising verbs, while roots are two-place (transitive) predicates, corresponding to control verbs. Two-place predicates assign two theta-roles: an internal theta-role for the infinitive and an external subject theta-role, whereas one place predicates assign only one theta-role, i.e. the internal one, to the infinitive. Similarly, Roberts (1985, apud Wurmbrand 1999) following Zubizarreta (1982) argues that epistemic modals do not assign a subject theta-role, but deontic modals assign an adjunct theta role to the surface subject. In opposition to these approaches, Susi Wurmbrand argues in her 1999 paper “Modal Verbs Must Be Raising Verbs” that both root and epistemic modal constructions are represented by a raising structure rather than a control structure.

Also, a third approach has been proposed, namely that the root – epistemic distinction is determined contextually in the semantic / pragmatic component and does not reflect a lexical or syntactic difference (Kratzer 1981, 1991; Papafragou 1998, 2000)

[...]

---

^[1] Alethic and volitional modalities have been suggested as further major types: alethic is a subcategory of the epistemic reading which shows “pure” logical modality, i.e. it doesn’t relativise the modal to any particular kind of facts, but rather considers it objectively; volitional modality is a subcategory of circumstantial (=root) modality in which the ideal world is the one corresponding to the relevant desires of the person to whom volition relates.