Natural Language and Linguistic Theory manuscript No. (will be inserted by the editor) Transforming Manner Adverbs into Subject-Oriented Adverbs: Evidence from Japanese Ai Kubota Received: date / Accepted: date Abstract It is well known that some adverbs in English, such as stupidly, cleverly and clumsily, can be interpreted as manner adverbs or agent-oriented adverbs depending on their positions in a sentence, e.g., John danced stupidly vs. Stupidly, John danced. Three approaches are possible and have been proposed for this alternation: (i) positing an agent-oriented adverb as the basic entry from which a manner adverb is derived (Ernst 2002), (ii) positing a manner adverb as the basic entry from which an agent-oriented adverb is derived (McConnell-Ginet 1982), and (iii) positing two distinct lexical entries for the two readings (Piñón 2010). I present data from Japanese which support the second approach. However, there would be a problem if we directly adopt the second approach for the Japanese data, since the adverbs that at first sight look like agent-oriented adverbs in Japanese are not truly ‘agent’-oriented, but rather ‘surface-subject’-oriented. I propose an analysis that does not suffer from this problem, by modifying an idea from McConnell-Ginet 1982 and also incorporating the notion of comparison class from Ernst 2002. The discussion extends to another class of adverbs called ‘evaluative adverbs’, such as fortunately and oddly, which show the same morphological property with surface-subject oriented adverbs in Japanese. Keywords Manner adverb · Agent-oriented adverb · Subject-oriented adverb · Evaluative adverb · Japanese · Comparison class · Passive sensitivity Ai Kubota Department of Linguistics and Germanic, Slavic, Asian and African Languages Michigan State University B-331 Wells Hall East Lansing, MI 48824-1027 Tel.: 517-353-0740 Fax: 517-432-2736 E-mail: [email protected] 2 Ai Kubota 1 Introduction Jackendoff (1972) notes that “the category ‘adverb’ has traditionally been a catch-all term” (p. 47). That being the case, the classification of adverbs is not at all a simple task. Furthermore, whether a particular classification originally developed for some language is adequate for other languages as well is another question. This paper sheds light on one such case, where a certain category of adverbs established in one language (English) has a seemingly equivalent class in another language (Japanese), but where the two categories in the two languages turn out to have some different property. More specifically, I focus in this paper on a class of adverbs called agent-oriented adverbs in English, such as stupidly, cleverly and clumsily, and the corresponding adverbs in Japanese, such as orokani-mo ‘stupidly’, kashikoku-mo ‘cleverly’, and bukiyooni-mo ‘clumsily’. Of particular interest is the meaning alternation between the so-called ‘manner’ and ‘clausal’ readings that is observed in both English and Japanese. Henceforth, I will refer to this phenomenon as the ‘manner/clausal alternation’. It has been observed that in English a class of adverbs such as stupidly, cleverly and clumsily have more than one interpretation when they appear in the auxiliary position as in (1-a), whereas the meaning is unambiguous when they appear in the sentence-initial position (1-b) or the final position (1-c) (Jackendoff 1972).1 (1) a. John stupidly danced. (Ambiguous) (i) ‘The manner in which John danced was stupid.’ (ii) ‘It was stupid of John to have danced.’ b. Stupidly, John danced. (Unambiguous) ‘It was stupid of John to have danced.’ c. John danced stupidly. (Unambiguous) ‘The manner in which John danced was stupid.’ While adverbs in the auxiliary position allow both the manner interpretation (‘the manner in which . . . ’) and the ‘clausal’, or the ‘sentential’, interpretation (‘it was stupid of John to . . . ’), adverbs in the sentence-initial position can only be construed in the clausal interpretation, and adverbs in the sentence-final position can only be construed in the manner interpretation. There are some general questions regarding this phenomenon. First of all, are adverbs of this kind lexically ambiguous? If not, how are the two interpretations derived from a single lexical source? Either way, what is the 1 Wyner (2008: 255) notes that adverbs can be interpreted as manner adverbs even in the sentence initial position, when “the overall sentence implies a contrast; that is, we use [(i)] to deny some previous assertion” (for example, in a context where the speaker wants to deny the statement “Bill kissed Jill reluctantly”). (i) Passionately, Bill kissed Jill. In this paper, I do not consider cases that involve this additional focus effect. Transforming Manner Adverbs into Subject-Oriented Adverbs 3 lexical/semantic relation between the two interpretations? Furthermore, how common is this lexical relation cross-linguistically? In this paper, I take a look at this kind of manner/clausal alternation in Japanese, in which there is a morphological distinction between manner adverbs and clausal adverbs unlike in English, and propose a compositional semantic analysis of the two readings. As shown above, in English, it is the position of the adverb that disambiguates the interpretation. In Japanese, the two readings are instead disambiguated morphologically. As Sawada (1978) notes, orokani ‘stupidly’ only has the manner reading, whereas orokani-mo in (3) with the particle mo only has the clausal reading, both regardless of the position in the sentence.2 This means that there is no ambiguity of the kind found in the English example (1-a). (2) a. b. (3) a. b. John-wa orokani odotta. John-top stupidly danced. ‘John danced stupidly.’ (Unambiguous; Manner only) Orokani John-wa odotta. stupidly John-top danced. ‘John danced stupidly.’ (Unambiguous; Manner only) John-wa orokani-mo odotta. danced. John-top stupidly ‘Stupidly, John danced.’ (Unambiguous; Clausal only) Orokani-mo John-wa odotta. stupidly John-top danced. ‘Stupidly, John danced.’ (Unambiguous; Clausal only) Morphologically speaking, it seems as if the particle mo transforms a manner adverb into the corresponding clausal adverb. (4) manner adverb + mo → clausal adverb According to Sawada (1978), this particle, as far as this phenomenon is concerned, is a kind of interjectional particle which has a function that marks the speaker’s subjective attitude toward the proposition.3 2 In Japanese, the position of adverbs within a sentence is quite free as long as they precede the verb. Focusing (by adding phonological prominence, for example) does not affect the interpretation of the adverb as far as the manner/clausal alternation is concerned. 3 The interjectional use of mo is often called eetan no mo ‘exclamatory mo’ such as in (i). (i) Kono-ko-mo zuibun ookiku-natta naa. This-child-mo a.lot big-became excl ‘This child has become so big!’ Mo is found in other uses, e.g. as an additive particle and (a part of) NPIs as shown below. (ii) a. John-mo kita. John-mo came. ‘John also came (in addition to someone else).’ 4 Ai Kubota This mo-alternation occurs systematically with adverbs such as kashikoku (-mo) ‘cleverly’, daitanni (-mo) ‘boldly’, namaikini (-mo) ‘impertinently’, teeneeni (-mo) ‘carefully; politely’, shinsetsuni (-mo) ‘kindly’, busahooni (-mo) ‘rudely’, and so on. However, this morphological alternation is found only with adverbs that have adjectival stems. Thus, for example, the manner adverb kichinto ‘neatly; decently; properly’, which does not have an adjectival counterpart, does not have the manner/clausal alternation (*kichinto-mo), thus no clausal use, even though there is nothing semantically anomalous about the would-be clausal meaning of this adverb (e.g., ‘Neatly, they gave us an extra pillow and towel.’). As far as the Japanese data are concerned, it seems natural to hypothesize that, for adverbs that display the manner/clausal alternation, clausal adverbs (which are morphologically more complex) are derived from manner adverbs. However, in the previous literature on this alternation (in English), it is still controversial whether the clausal meaning and the manner meaning share the same lexical source. For example, while McConnell-Ginet (1982) proposes to derive clausal adverbs from manner adverbs, other authors such as Ernst (2002), Rawlins (2008), Geuder (2002) pursue an opposite approach in which manner adverbs are derived from clausal adverbs. There is still one more possibility, that is, to say that these adverbs are lexically ambiguous, which is a position taken by Wyner (2008), and possibly Piñón (2010) too. However, this last approach, if it is simply assumed that the manner adverb stupidly and the clausal adverb stupidly each have distinct and unrelated lexical entries, seems to lose the insight that these two kinds of adverbs are somehow related and that this relation is not accidental but systematic and observable across languages. In the following section, I review the previous studies just mentioned and point out problems and difficulties that they face in analyzing especially the Japanese data just introduced. Then in section 3, I propose my analysis of manner/clausal alternating adverbs, building on ideas from Ernst (2002) and McConnell-Ginet (1982). In section 4, I discuss passive sensitivity, a phenomenon in which passive sentences with an adverb of a certain class have more than one interpretation, even though the corresponding active sentences with the same adverb have only one interpretation. The passive sensitivity data discussed in this section will make clear why I adopt the terminology ‘surface-subject-oriented adverb’ (instead of the more familiar ‘agent-oriented adverb’) for Japanese in this paper (see below). In section 5, I briefly discuss a class of adverbs often called evaluative adverbs, which also undergo the mo attachment in Japanese. Section 6 summarizes and concludes the paper. b. Dare-mo ko-nakatta. Who-mo come-neg.pst ‘No one came.’ I do not know whether it is possible to have a unified analysis that covers all the instances of mo. Transforming Manner Adverbs into Subject-Oriented Adverbs 5 Before moving on to the next section, let me clarify the terminology and the classification of adverbs that I assume throughout this paper. As for the English adverbs, I adopt the classification used in Ernst 2002. As for the Japanese adverbs, I suggest a slightly different terminology in order to reflect the behavioral difference between adverbs in Japanese and adverbs in English. According to Ernst (2002), the adverb stupidly in its clausal sense as in (1-b) belongs to the class of adverbs called agent-oriented adverbs, which is a subclass of subject-oriented adverbs as in (5). The adverb stupidly as in (1-c) belongs to a different class of adverbs, i.e., manner adverbs. 4 (5) Subject-oriented adverbs (Ernst 2002) a. Agent-oriented adverbs: cleverly, stupidly, wisely, tactfully, foolishly, rudely, secretly, ostentatiously, intelligently b. Mental attitude adverbs: reluctantly, calmly, willingly, anxiously, eagerly, frantically, absent-mindedly, gladly, sadly Subject-oriented adverbs are those that “express some additional information about the subject” (Jackendoff 1972: 57), as the paraphrase in (6-b) shows.5 (6) a. b. {Cleverly/Reluctantly}, John spilled the beans. John was {clever/reluctant} to spill the beans. Among subject-oriented adverbs, what Ernst calls agent-oriented adverbs are those that “indicate that an event is such as to judge its agent as ADJ with respect to the event” (Ernst 2002: 54), whereas what he calls mental attitude adverbs are those that “describe, most fundamentally, a state of mind experienced by the referent of the subject of the verb” (ibid.: 63). In what follows, I mainly use stupidly and reluctantly as representative examples of agent-oriented adverbs and mental attitude adverbs respectively. In addition to subject-oriented adverbs, there is another class of adverbs, which is called speaker-oriented adverbs, following Jackendoff (1972) and Ernst (2002). Speaker-oriented adverbs can also be considered as a kind of clausal adverbs, as they modify clausal elements rather than verb phrases. Among speaker-oriented adverbs, there are three subclasses: speech-act (honestly, frankly, roughly, etc.), epistemic (perhaps, probably, maybe, etc.), and evaluative (unbelievably, unfortunately, luckily, strangely, etc.). Examples are from Ernst (2002: 69). (7) a. Honestly, who would do such a thing? (Speech-act adverb) 4 Agent-oriented adverbs are also called “thematically dependent adverbs (TDAs)” (Wyner 1998) and “Ad-VPs” (McConnell-Ginet 1982). 5 As pointed out by one of the reviewers, the paraphrase in (i-b) is not appropriate for some of the adverbs listed in (5), e.g., *John was {secret/calm} to spill the beans. Since the adverbs secretly and calmly does not appear in Jackendoff’s (1972) list of subject-oriented adverbs, it is either the case that the paraphrase ‘SUBJ was ADJ to . . . ’ is not adequate for subject-oriented adverbs or the adverbs secretly and calmly are not supposed to be classified as subject-oriented adverbs. See also footnote 7. 6 Ai Kubota b. c. The markets will perhaps respond to lower interest rates. (Epistemic adverb) Unbelievably, she decided to buy a camel. (Evaluative adverb) I briefly discuss evaluative adverbs in Japanese in section 5 to examine whether my proposal about mo is on the right track, since evaluative adverbs are another class of adverbs in Japanese that allow the mo-attachment just like orokani-mo ‘stupidly’. As for the adverbs in Japanese, I adopt a different terminology. Specifically, I will not call the class of adverbs that includes orokani-mo ‘stupidly (clausal)’ ‘agent-oriented’ adverbs. Instead I will call them ‘surface-subjectoriented adverbs’ (SS-oriented adverbs), since these adverbs in Japanese are not strictly ‘agent’-oriented (section 3 and section 4.2), and it is misleading to call them so. As shown in the next section, previous analyses that are based on observations about agent-oriented adverbs in English cannot be directly adopted to account for the nature of SS-oriented adverbs in Japanese. 2 Previous Studies on the Manner/Clausal Alternation There are mainly three possible approaches to the manner/clausal alternation. The first approach is to posit a clausal adverb as the basic and derive a manner adverb form it. The second approach is to posit a manner adverb as the basic and derive a clausal adverb from it. The third approach is to give up deriving one from the other and simply posit two versions for each adverb that exhibits the manner/clausal alternation. In the following subsections, I will first review Ernst 2002, which takes the first approach, and then McConnell-Ginet 1982, which takes the second approach, followed by Piñón 2010, which, as far as I can tell, goes for the third view. 2.1 Ernst 2002 According to Ernst (2002), the manner and clausal readings of each adverb are closely related to each other in such a way that manner readings are verbmodifying versions of adverbs whose lexical entries are the clause-modifying adverbs that yield agent-oriented readings. He also points out two major ways in which the two readings differ. The first is that manner readings “describe some sort of external manifestation that may or may not reflect the internal reality” ?? 56]ernst2002. For example, the sentences in (8) ((2.43) in Ernst 2002), show that the post-verbal manner adverbs ‘manifest’ (or ‘show properties typical of’) stupidity and cleverness without Alice the spy actually being stupid or clever at the moment. (8) a. b. Alice cleverly answered stupidly in order to keep her identity secret. Alice stupidly answered cleverly and gave her secret identity away. Transforming Manner Adverbs into Subject-Oriented Adverbs 7 The second difference he points out – the observation which I take to be critical and which will be crucial in my analysis too – is that manner readings and agent-oriented readings differ in terms of the comparison class for events. Take a pair of examples in (9) ((2.44) in Ernst 2002). In (9-a), “she is judged rude because of the event of her leaving, as opposed to other things she could have done, most especially not leaving”, whereas in (9-b), “she is judged rude on the basis of something about her leaving – some property of her leaving that we sometimes call a manner, which distinguishes this leaving event from other possible leaving events. For example, she might have left without saying good-bye, by slamming the door, or with a few choice imprecations on her way out” ?? 57]ernst2002. (9) a. b. Rudely, she left. She left rudely. This means that, for agent-oriented readings the comparison class consists of various possible things that the subject could have done, whereas for manner readings, the comparison class is a set of more specific kinds of events (e.g., leaving events in the above example). As Ernst (2002) notes, the notion of comparison class is necessary anyway for the interpretation of gradable predicates in general such as tall, hot, cute, and so on. Since adverbs such as stupidly, cleverly, and so on, are indeed gradable (as we can form a question that specifically asks the degree, e.g., How stupidly did he dance? ), comparison classes undoubtedly play a role in the interpretations of these adverbs too. Therefore, it is desirable to adopt the notion of comparison class, which is an independently motivated notion, if comparison class is the key to distinguish the two meanings. As for the semantic representations of manner and agent-oriented adverbs and how they are related to each other, Ernst (2002) takes the agent-oriented version in (10-a) as basic (the lexical entry for the adverb rudely), and proposes an operation called the Manner Rule which converts an agent-oriented adverb into the corresponding manner adverb (10-b).6 (10) a. b. The event e warrants positing more rudeness in Agent than the norm for events. 7 The event e manifests more rudeness in Agent than the norm for Specified Events.8 Although I agree that the notion of comparison class plays a crucial role for gradable adverbs, there are remaining questions for this approach both theoretically and empirically. First, from a theoretical point of view, there is room for discussion on the adequacy of the Manner Rule. For example, 6 For the precise implementation of the Manner Rule, see Chapter 2 in Ernst 2002. If Ernst’s (2002) analysis is correct, secretly cannot really be an example of agentoriented adverbs, since the expression ‘the event e warrants positing more secret in Agent than the norm for events’ is not straightforwardly interpretable (see also footnote 5). 8 ‘Specified Events’ in (10-b) is determined by the verb phrase that the adverb modifies. 7 8 Ai Kubota at what point of derivation does this rule apply? Is there any restriction for the application of this rule? Is this a language-specific rule? How does the Manner Rule fit in a framework of compositional semantics? Furthermore, there is an empirical difficulty especially when we turn to the manner/clausal alternation in Japanese. Note that in Japanese it is manner adverbs that are morphologically simplex.9 mo-attachment −−−−−−−−−→ (11) manner adverb orokani ‘stupidly’ SS-oriented adverb orokani-mo ‘stupidly’ ←−−−−−−−−−−−−− Ernst’s Manner Rule If we directly adopt Ernst’s (2002) approach (i.e., if we take agent-oriented adverbs to be basic and adopted the Manner Rule), then we would have to assume that mo’s function is to cancel the application of the Manner Rule (since the mo-less version is unambiguously a manner adverb). Alternatively, we might assume that the lexical item orokani (with an underlying agentoriented meaning) is never interpretable without the obligatory application of the Manner Rule, or with the attachment of a semantically vacuous morpheme mo. But it seems highly unlikely that a lexical item that is never interpretable without an obligatory application of an additional lexical operation like the Manner Rule (which itself is already a highly unnatural situation) becomes suddenly interpretable in the presence of an overt but meaningless morpheme. Thus, as far as the Japanese data is concerned, I suggest to take the other way, that is, compositionally deriving the clausal reading via the combination of the manner adverb and mo. Although I depart from Ernst’s (2002) approach in this respect, I adopt his idea of capturing the difference between agent-oriented and manner readings via the notion of comparison class. 2.2 McConnell-Ginet (1982) According to McConnell-Ginet (1982), adverbs are always construed directly with a verb.10 In the case of manner adverbs as in (12), the adverb rudely is construed with the verb departed, and adds some meaning to the verb. On the other hand, agent-oriented adverbs in (13) are not syntactically sisters to the verb departed but to the higher abstract verb acted. (12) a. Surface structure: Louisa departed rudely. 9 This pattern in which the manner adverb is morphologically simpler does not seem to be limited to Japanese, as it is also attested in other languages such as German (e.g., klugerweise ‘cleverly’, which consists of klug ‘clever’ and weise, can only be interpreted as a agent-oriented adverb). The Manner Rule would be more convincing if there is a significant number of languages in which the morphological relation between the two adverbs is the opposite of (11). I do not know of any such language. 10 To be more precise, adverbs are actually considered as an argument of the verb and not really a modifier. Transforming Manner Adverbs into Subject-Oriented Adverbs (13) 9 b. LF: Louisa [ departed rudely ] a. Surface structure: Louisa rudely departed. (rudely with an agentoriented meaning) LF: Louisa [ acted rudely to depart ] b. In this approach, the interpretation of an adverb depends on which verb the adverb is construed with, either the main verb depart or the higher, abstract verb act. The manner adverb meaning is considered as semantically basic, and the agent-oriented reading is obtained by positing a different syntactic structure (which receives a different semantic interpretation). This idea seems to be plausible especially for the Japanese manner/clausal alternation, since it allows us to analyze the manner adverb orokani as the basic form. If this essentially syntactic analysis of the alternation is on the right track, then we might take it to mean that mo is a morpheme that signals that the adverb is located somewhere higher in the structure. In this way, we would not have to rely on a special operation like the Manner Rule. However, this approach has been criticized in the literature, especially with respect to the adequacy of a higher verb like act. For example, Geuder (2002) points out that a sentence like John departed cannot be paraphrased as John acted to depart. He also raises a question with respect to the contrast between she acted rudely and she rudely acted. In order to capture the difference between the two, the latter has to be analyzed as she acted rudely to act. However, as Geuder (2002) points out, it is not clear what act to act means. 2.3 Piñón (2010) With Geuder’s (2002) criticism in mind, Piñón (2010) argues that there is a higher verb decide instead of act. Furthermore, he argues that this higher verb decide does not exist by itself somewhere in the structure but is introduced by the agent-oriented adverb. That is, the meaning of decide is part of the meaning of agent-oriented adverbs. Thus, there are two different lexical entries for stupidly, one for the manner reading (14-a) and the other for the agentoriented reading (14-b).11 (14) a. b. Manner: Jstupidlym K= λVv,t λe.V (e) ∧ stupid(e) a Agent-Oriented: Jstupidly K= 00 decide(e , x, [λe0 .W (e0 , x)])∧ 00 λWhe,vti λxλe.∃e CAUSE(e00 , e) ∧ W (e, x) ∧ stupid(e00 ) In short, what is described as stupid in (14-a) is the event of V , where V is denoted by the verb that the adverb modifies. In (14-b), on the other hand, what is described as stupid is not the event of V but the event of x’s deciding 11 In the original proposal, the clausal argument of ‘decide’ is intensionalized. I have reproduced a simplified entry in (14) since this aspect of the analysis does not affect our discussion below. 10 Ai Kubota to do W , where W is denoted by the verb. In other words, stupidly, John danced means that the event of John’s deciding to dance was stupid but does not necessarily mean that the event of John’s dancing was. Setting aside how precisely the above denotations work compositionally and assuming that these two versions of stupidly derive manner and agentoriented meanings respectively, we are now back to our original puzzle: how are manner readings and agent-oriented readings related to each other? Should we give up deriving one of the two readings from the other and go for lexical ambiguity? Of course, it is not impossible to derive one from the other, for example, via the following operation that converts manner adverbs into agentoriented ones. CAUSE(e00 , e) ∧ W (e, x)∧ (15) λSvt,vt λWe,vt λxe λev .∃e00 (Jstupidly m K) S(e00 , λe.decide(e, x, [λe0 .W (e0 , x)]))) CAUSE(e00 , e) ∧ W (e, x)∧ =λW λxλe.∃e00 Jstupidly m K(e00 , λe.decide(e, x, [λe0 .W (e0 , x)]))) 00 00 CAUSE(e , e) ∧ W (e, x)∧ = λW λxλe.∃e decide(e00 , x, [λe0 .W (e0 , x)]) ∧ stupid(e00 ) a = Jstupidly K However, it is unclear what this operation is and how common it is. Clearly, this is not a simple type-shifting operation, but something that introduces the higher verb decide and a causal relation between the two events (x’s decision of doing W and x’s doing W ). It remains unclear how general this kind of operation is cross-categorically, within a language, and across languages. Moreover, there is another concern regarding the higher verb decide. That is, whenever there is an agent-oriented adverb, the sentence is always predicted to mean that the subject decided to do the action. What, then, about a sentence like (16), where John didn’t decide to die/fall?: (16) John-wa orokani-mo {shinda/koronda}. John-top stupidly-mo {died/fell} ‘Stupidly, John died/fell.’ The prediction is that the agent-oriented adverb orikani-mo introduces the higher verb decide and thus the sentence means that the subject decided to die. However, the sentence does not necessarily mean that John committed suicide. It simply means that it was stupid of him to have died, either voluntarily or by doing something (perhaps inadvertently) that eventually led him to die against his will. I will come back to this example at the end of section 3. The fundamental problem of this approach lies in the assumption that agent-oriented adverbs introduce a higher verb decide as part of their meanings. Not only is it problematic for accounting for data like (16) but it also loses McConnell-Ginet’s (1982) insight that adverbs like stupidly are fundamentally manner adverbs (or Ad-Verbs in her term) and that the agent-oriented reading is sololy derivable from a structural difference. Transforming Manner Adverbs into Subject-Oriented Adverbs 11 2.4 Summary In this section, we saw three approaches to the manner/clausal alternation: deriving the manner meaning from the agent-oriented (or, clausal) meaning, deriving the agent-oriented meaning from the manner meaning, and lexical ambiguity. Ernst’s (2002) analysis nicely captures the difference between the two meanings in terms of comparison classes. However, since it takes the first approach, the analysis is not suitable for Japanese, given the morphological complexity of the form associated with the clausal reading. In this respect, McConnell-Ginet’s (1982) idea that agent-oriented adverbs are originally manner adverbs is appropriate for the Japanese data, if we assume that mo is a kind of syntactic marker that forces the adverb to be located structurally higher. A problem remains, however, with respect to the higher abstract verb act, as pointed out by Geuder (2002) and Piñón (2010). A lexical ambiguity approach (Geuder 2002, Piñón 2010) cannot adequately explain why there is a tight connection between certain kinds of adverbs across languages. In the next section, I propose an analysis adopting two important insights from the previous literature. One is the idea of comparison class from Ernst 2002 and the other is the idea from McConnell-Ginet 1982 that the manner reading is the basic and the clausal reading is derived from it. Unlike McConnell-Ginet’s (1982) idea, however, I propose that the difference between the two readings is not solely due to the structural differences but also because of the effect of mo that changes the semantics of the adverbs it attaches to. 3 Deriving Clausal Readings from Manner Adverbs Let us remind ourselves what the puzzle was. In English, adverbs like stupidly and cleverly have the manner reading (‘the manner in which x did V was stupid’) and the clausal reading (‘It was stupid of x to do V ’), and disambiguation is possible by word order – sentence-final adverbs give rise to manner readings and sentence-initial adverbs give rise to agent-oriented/clausal readings. In Japanese, the two readings are instead distinguished morphologically: orokani ‘stupidly’ (manner) in the simplex form vs. orokani-mo ‘stupidly’ (clausal) with mo. Based on this morphological fact, I will pursue an approach that compositionally derives the clausal reading from the manner reading. In section 3.1, I analyze the manner adverb orokani incorporating the notion of comparison class from Ernst 2002. In section 3.2, I propose that the same manner adverb, when mo is attached, is turned into a clausal adverb. The reason why the meaning changes from a manner meaning to a clausal meaning is not because the adverb is adjoined to a head such as act, decide, nor any other kind of head that is responsible for the agent thematic role, but because its comparison class is readjusted to a broader one by adjoining mo. Thus, on this approach, examples like (16) – in which adverbs like orokani-mo induce clausal readings despite the absence of an agent – are unproblematic. 12 Ai Kubota 3.1 Manner Adverb orokani ‘stupidly’ Throughout the paper, I adopt a fairly common syntactic assumption that all the syntactic arguments are base-generated within VP (or vP depending on the syntactic framework), and that the subject is later moved to Spec, TP in order to be assigned case. For the sake of simplicity, I ignore the temporal meaning and assume that T is essentially semantically vacuous. The following shows the derivation for a simple sentence John-ga odotta ‘John danced’ without an adverb. (17) TP λe.dance(John)(e) T0 λxλe.dance(x)(e) NP John-ga VP λxλe.dance(x)(e) T λP λxλe.P (x)(e) odotta After the event variable e is existentially bound, the sentence asserts that there was an event e such that e was a dancing event by John. Throughout the paper, I assume that the event variable is existentially closed at some point in the same way. When there is a manner adverb as in (18), I assume that the adverb is adjoined to VP. (18) John-ga orokani odotta. John-nom stupidly dance.past ‘John danced stupidly.’ I propose the following denotation for the manner adverb orokani ‘stupidly’. (19) JorokaniK = λPhe,vti λxλe.P (x)(e)∧stupid(e) > standard(stupid)([λe0 ∃x0 .P (x0 )(e0 )]) As shown in (19), stupid is a measure function of type hv, di – it takes an event e and returns the degree of stupidity of e. standard(stupid)([λe0 ∃x0 .P (x0 )(e0 )]) gives the standard degree of stupidity, where what counts as ‘standard’ is determined by the comparison class λe0 ∃x0 .P (x0 )(e0 ), a set of P -ing events by some individual x0 . In the case of adjectives, the comparison class is a set of individuals – of type he, ti (von Stechow 1984). For adverbs, on the other hand, I assume that the comparison class is a set of events – of type hv, ti. The way in which the meaning is expressed in (19) is based on one of the standard views on gradable adjectives, according to which the truth condi- Transforming Manner Adverbs into Subject-Oriented Adverbs 13 tions of gradable adjectives in their positive forms involve a context-dependent standard degree. To be more precise, the representation in (19) should be considered as a combination of orokani and a null degree morpheme pos (von Stechow 1984, Kennedy 1997) just like gradable adjectives such as tall, expensive, cute, and so on. The pos morpheme is responsible for the notion of standard degree and comparison class. In the case of comparatives, a comparative morpheme (more or -er in English) replaces the pos morpheme and thus there will be no standard degree determined context-dependently. Indeed, we do find a comparative form of adverbs (e.g., more stupidly), which calls for a compositional analysis extending the analysis of adjectival comparatives to adverbial comparatives. It is possible to deal with the comparative form of adverbs if we adopt the idea shared among the analyses of adjectival comparatives in general that the gradable predicate itself only contributes a mapping relation between individuals (or, here, events) and degrees and that the reference to the contextual standard (if any) is separately contributed by the empty pos operator. In this paper, however, I simply express the meaning of gradable adverbs as a combination of measure function and pos for the sake of convenience. Let us now turn to the example (18) and see how the meaning is derived compositionally. (20) TP λe.dance(John)(e)∧ stupid(e) > standard(stupid)(λe0 ∃x0 .dance(x0 )(e0 )) NP John-ga T0 λxλe. dance(x)(e)∧ stupid(e) > standard(stupid)([λe0 ∃x0 .dance(x0 )(e0 )]) VP λxλe. dance(x)(e)∧ stupid(e) > standard(stupid)([λe0 ∃x0 .dance(x0 )(e0 )]) AdvP λP λxλe. P (x)(e)∧ stupid(e) > standard(stupid)([λe0 ∃x0 .P (x0 )(e0 )]) VP λxλe.dance(x)(e) odotta orokani T λP λxλe.P (x)(e) 14 Ai Kubota The derived meaning is as follows: there was an event e such that e was a dancing event by John, and the stupidity of e is greater than the standard stupidity, where the standard is determined by the comparison class that consists of various kinds of dancing events involving some individual x0 as the dancer. Since the comparison class is restricted to a specific kind of event (in this case, dancing events) what counts as stupid is based on this restricted set, e.g. {x0 ’s dancing with precise steps, x0 ’s dancing with a smooth hand movement, x0 ’s dancing with his eyes rolled up into his head, . . .}. This is similar to the sentence ‘John did a stupid dance’ in which the adjective stupid directly modifies the noun dance. Just like gradable adjectives such as small and big as in a small elephant and a big mouse, what counts as stupid depends on the comparison class that is either explicitly mentioned or implicitly understood in the discourse. A small elephant could be bigger than a big mouse, in which case the individual is “small for an elephant” or “big for a mouse”. Similarly, the event that John has engaged in does not have to be stupid in all respects and John himself does not have to be stupid, but what he did was a stupid way or kind of dancing.12 3.2 Clausal Adverb orokani-mo ‘stupidly’ With the same denotation proposed in (19) for the manner adverb orokani ‘stupidly’, it is possible to derive the clausal reading when the particle mo is adjoined first to adjust the comparison class appropriately. I propose the denotation of mo as in (21). The first argument A is supposed to be supplied by an adverb of type hhe, vti, he, vtii. (21) JmoK = λAhhe,vti,he,vtii λP λxλe.A(λx0 λe0 .∃Q[Q(x0 )(e0 )])(x)(e) ∧ P (x)(e) This is, in some sense, like the additive particle also – it introduces an additional predicate Q, in addition to the predicate P denoted by the VP.13 Mo denotes a function from type hhe, vti, he, vtii to hhe, vti, he, vtii – it takes an adverb as an argument and returns a function of the same type. Thus, when it combines with the adverb orokani, it changes the adverb in the following way: 12 In this connection, I should note that there are proposals that view manner modification as event kind modification (Landman and Morzycki 2003, Anderson and Morzycki 2014, Gehrke 2014). It would be interesting to see if it is possible to integrate the notion of comparison class and event kind modification, a task which I leave for future research. 13 Recall from above that mo has this additive function as one of its meanings (see footnote 3), but I will set aside further discussion on a more comprehensive analysis of mo. Transforming Manner Adverbs into Subject-Oriented Adverbs (22) 15 JmoK(JorokaniK) λAhhe,vti,he,vtii λP λxλe. 0 λPhe,vti λx00 λe00 . P 0 (x00 )(e00 )∧ = A(λx0 λe0 .∃Q[Q(x0 )(e0 )])(x)(e) stupid(e00 ) > standard(stupid)([λe000 ∃x000 .P 0 (x000 )(e000 )]) ∧P (x)(e) ∃Q[Q(x)(e)] =λP λxλe ∧stupid(e) > standard(stupid)(λe0 ∃x000 .∃Q[Q(x000 )(e0 )]) ∧P (x)(e) I suggest that the clausal adverb orokani-mo is adjoined to T0 . The following represents the derivation for the clausal meaning of orokani-mo. (23) TP λe.∃Q[Q(John)(e)] ∧stupid(e) > standard(stupid)(λe0 ∃x000 ∃Q[Q(x000 )(e0 )])∧ dance(John)(e) NP John-ga T0 λxλe.∃Q[Q(x)(e)]∧ stupid(e) > standard(stupid)(λe0 ∃x000 ∃Q[Q(x000 )(e0 )]) ∧dance(x)(e) AdvP λP λxλe.∃Q[Q(x)(e)]∧ stupid(e) > standard(stupid)(λe0 ∃x000 ∃Q[Q(x000 )(e0 )]) ∧P (x)(e) Adv orokani mo T0 λxλe.dance(x)(e) VP λxλe.dance(x)(e) odotta The derived meaning is as follows: there was an event e such that e is a dancing event by John. In addition, there is another way Q to describe e – for example, it could be ‘John’s moving’. However, since it only mentions the existence of such a predicate Q, it is semantically trivial. The stupidity of e is greater than the standard stupidity, where the standard is determined by the comparison class that consists of a set of events e0 . The comparison class is not restricted to the specific kind of events like various kinds of dancing but it is a set of events e0 such that one could have involved in that event in some way or another. Importantly, the representation in (23) does not entail that John the subject is stupid. It simply says that the event that John danced was stupid considering other things that one could have done. I believe that this is correct because by saying “Stupidly, John danced”, it does not mean that T λP λxλe.P (x)(e) 16 Ai Kubota John is stupid and he danced. What is expressed in (23) is more like “That John danced was stupid”.14 In this way, I adopt the idea of comparison class from Ernst 2002 and let mo transform a manner adverb into the corresponding clausal adverb by adjusting the comparison class of the adverb. This analysis departs from his analysis, however, in that instead of employing the Manner Rule to transform a clausal meaning into a manner meaning, I derive the clausal meaning compositionally from the manner meaning. In this respect, the present proposal is similar in spirit to McConnell-Ginet’s (1982), although it is different from her analysis as well, since it is not dependent on an abstract verbal head such as act or decide that is responsible for introducing an agentive argument of some kind. As far as the Japanese data is concerned, the present analysis is more appropriate in view of cases in which no agent is arguably involved in the event expressed by the sentence. It is commonly argued that the subject of unaccusative verbs such as die, fall, arrive and, so on, is actually base-generated as the complement of the verb just like direct objects, and is assigned the patient or theme role instead of the agent role. The argument is then promoted to the surface subject position because of the case assignment reason. Adverbs like orokani-mo can appear in a sentence with an unaccusative verb, as already pointed out and repeated below. (16) John-wa orokani-mo {shinda/koronda}. John-top stupidly-mo {died/fell} ‘Stupidly, John died/fell.’ If we analyze clausal adverbs like orokani-mo ‘stupidly’ as strictly agentoriented adverbs, (16) will be predicted to be unacceptable. The proposed analysis does not rule out examples like (16), since what the adverb is oriented to is not what ensures the agentive subject but simply T0 , which is typically considered as a sister to the surface subject. One might wonder why clausal adverbs have to be located at T0 and not VP. It is true, if we restrict our attention to sentences in the active voice, that we can derive exactly the same meaning in the end by assuming that orokani-mo ‘stupidly (clausal)’ is a VP modifier (assuming that tense does not directly affect the meaning of the adverb). However, as I will show in section 4, it becomes clear why it is better to have SS-oriented adverbs with mo at the T0 level and not within VP when we turn to the behavior of SS-oriented adverbs in Japanese in passive sentences. The prediction is that if clausal adverbs in Japanese (i.e., SS-oriented adverbs) are indeed adjoined to T0 and are always oriented to the surface subject, then they are always oriented to the surface subject (the patient or theme argument in passive sentences) and never to the 14 In connection to footnote 5, this paraphrase can handle adverbs such as secretly as in secretly, John danced. While it does not make sense to say “it was secret of John to have danced”, the alternative paraphrase “that John danced was secret” may come close to what it is intended to mean. Transforming Manner Adverbs into Subject-Oriented Adverbs 17 underlying subject (the agent argument). As I show in the next section, this is indeed the case in Japanese. 4 SS-Oriented Adverbs and Passive Sensitivity In this section, I compare SS-oriented adverbs (or agent-oriented adverbs in English) and mental attitude adverbs with respect to the issue of passive sensitivity. After introducing the basic facts about this phenomenon observed both in English and in Japanese, I show that Japanese SS-oriented adverbs like orokani-mo ‘stupidly’ are not passive sensitive – they can only be oriented to the surface subject – unlike English agent-oriented adverbs. I also point out that there is a clear contrast between SS-oriented adverbs and mental attitude adverbs. I propose that the reason is because the mo-attached adverbs are always T0 -adjoined whereas the mo-less ones are attached lower in the structure. 4.1 Passive sensitivity It has been observed that some subject-oriented adverbs such as reluctantly, intentionally, and deliberately are passive sensitive (McConnell-Ginet 1982, Wyner 1998). (24) a. b. Kim reluctantly hit Sandy. →Kim was reluctant. Sandy was reluctantly hit by Kim. (i) →Sandy was reluctant. (ii) →Kim was reluctant. (Active; unambiguous) (Passive; ambiguous) While the active sentence (24-a) has only one interpretation, namely, one in which the subject (Agent) is reluctant, the passive sentence (24-b) has two possible interpretations: either the surface subject (Patient) was reluctant or the underlying subject (Agent) was. According to McConnell-Ginet (1982), the reason for this ambiguity in passive sentences is that there are two possible structural positions for the adverb to attach to in the passive construction. The idea is that in passive sentences, the passive auxiliary be may be interpreted as the ‘active’ be as proposed by Partee (1977) or the predicate act in Dowty 1979. In such a case, there are two verbs (or VPs) for the adverb to modify – either the lower verb hit or the higher verb be. This is illustrated in (25) below. 18 (25) Ai Kubota a. S NP VP2 Sandy V VP1 was b. ADV V reluctantly hit PP by Kim S NP VP2 Sandy ADV V VP1 reluctatnly was hit by Kim When the adverb reluctantly is construed with the lower verb (VP1) as in (25-a), it is oriented to the subject (Agent) of hit, whereas in case the adverb is construed with the higher verb (VP2) as in (25-b), it is oriented to the subject of active be, which is the underlying Patient of the verb hit. Depending on the attachment site, the interpretation of the adverb changes, specifically with respect who is reluctant. By contrast, in active sentences like (24-a), no ambiguity arises, since there is no passive auxiliary that the adverb can be construed with.15 Passive sensitivity is also observed in Japanese just like in English. (26) a. John-wa Mary-o iyaiya dakishimeta. John-top Mary-acc reluctantly hug.past ‘John reluctantly hugged Mary.’ (Unambiguous) →John was reluctant. b. Mary-wa John-{ni/niyotte} iyaiya dakishime-rareta. Mary-top John-{by/by} reluctantly hug-passive.past ‘Mary was reluctantly hugged by John.’ (Ambiguous) (i) →John was reluctant. (ii) →Mary was reluctant. 15 For Wyner (1998), the ambiguity is not explained in terms of structural differences, but by assuming that there are two kinds of passive be: one that is semantically vacuous and the other that assigns a volitional meaning to the subject. In this view, however, not only do we need two kinds of be, but also two different ways to interpret the main verb: one with volition and the other without. Transforming Manner Adverbs into Subject-Oriented Adverbs 19 So it seems that the same explanation holds for the Japanese case as well. However, not all adverbs behave in this way. As I show below, some subject-oriented adverbs, including orokani-mo ‘stupidly’, do not exhibit passive sensitivity. 4.2 Mental attitude adverbs vs. SS-oriented adverbs In the previous subsection, we have only considered passive sensitivity with respect to the subject-oriented adverb reluctantly. Although both reluctantly and stupidly are examples of subject-oriented adverbs, they are said to belong to different subcategories of subject-oriented adverbs – stupidly is a representative of agent-oriented adverbs and reluctantly is that of mental-attitude adverbs – as mentioned in section 1. According to Ernst (2002), the two types of subject-oriented adverbs differ in terms of their semantic representations. Schematically, the interpretation of agent-oriented adverbs generally looks like (27), whereas for mental attitude adverbs, the interpretation looks like (28). (27) The event e warrants positing more PADJ in Agent than the norm for events. (28) The event e is {accompanied by/intended with} a greater degree of ADJ by Experiencer (subject) than the norm for Experiencer. Setting aside the subtle difference among the expressions ‘warrants positing’ and ‘accompanied by/intended with’, the major difference between the two types of subject-oriented adverbs seems to be the kind of comparison classes. For mental attitude adverbs, the comparison class is “restricted to the experiencers and is not determined by events” (Ernst 2002:pp. 62). Ernst (2002) notes that mental attitude adverbs such as reluctantly also tend to have two interpretations just like agent-oriented adverbs – the clausal (mental attitude) reading (29-a) and the manner reading (29-b). In (29-a), the subject is the experiencer, thus she is indeed reluctant. On the other hand, in (29-b), the subject may not necessarily be reluctant – she might just be showing such a behavior but she could be happy in her heart. (29) a. b. Mindy has reluctantly been going to accounting classes. Mindy has been going to accounting classes reluctantly. This seems to suggest that despite the difference between the two types of subject-oriented adverbs (mental attitude and agent-oriented adverbs), the mechanism of manner/clausal alternation is essentially the same. However, the difference between mental attitude adverbs and SS-oriented adverbs in Japanese seems to be more distinct compared to the difference between mental attitude adverbs and agent-oriented adverbs in English with respect to manner/clausal alternation and passive sensitivity. First, with Japanese SS-oriented adverbs, we saw that the manner/clausal alternation involved mo attachment, whereas mental attitude adverbs in Japanese are not compatible with mo attachment. Since mental attitude adverbs in Japanese do not have 20 Ai Kubota the manner reading, as shown in (30-a), the only way to express the manner meaning is to employ a related but distinct modifier specifically referring to the manner such as iyagatta yoosu-de ‘in a reluctant way’ in (30-b). (30) a. John-wa iyaiya gakkoo-ni kayotteiru. (#demo, John-top reluctantly school-to commute.npst (But hontoo-wa gakkoo-ga suki-da.) actually school-nom favorite-cop.npst) ‘John reluctantly goes to school. (#But, actually, he likes school.)’ b. John-wa iyagatta yoosu-de gakkoo-ni kayotteiru. (demo, John-top reluctant manner-in school-to commute.past (But hontoo-wa gakkoo-ga suki-da.) actually school-nom favorite-cop.npst) ‘John went to class reluctantly. (But, actually, he likes school.)’ In English, on the other hand, there is no morphological distinction between the manner and clausal readings. Secondly, while both mental attitude adverbs and agent-oriented adverbs are passive sensitive in English, only mental attitude adverbs, but not SSoriented adverbs, are passive sensitive in Japanese. In English, not only mental attitude adverbs but also agent-oriented adverbs seem to be passive sensitive as shown below.16 (31) Mary was stupidly hugged by John. a. (??)It was stupid of Mary to have been hugged by John. b. It was stupid of John to have hugged Mary. By contrast, SS-oriented adverbs are never passive sensitive in Japanese. The adverb orokani-mo ‘stupidly’ in the passive construction can only target the surface subject and never the underlying subject, and this judgement is quite clear among native speakers. (32) Mary-wa orokani-mo John-ni dakishime-rareta. Mary-top stupidly-mo John-by hug-passive.past ‘Stupidly Mary was hugged by John.’ a. It was stupid of Mary to have been hugged by John. b. NOT: It was stupid of John to have hugged Mary. As shown above in (26), it is not the case that passive sensitivity is not observed at all in Japanese – it is observed with mental attitude adverbs. At this point, it is worth noting that mental attitude adverbs in Japanese are generally incompatible with mo, unlike SS-oriented adverbs, e.g. iyaiya ‘re16 However, it is less clear compared to mental attitude adverbs whether they are really passive sensitive according to the native speakers I consulted with. While some speakers accept both readings in (31), others report that it is difficult to interpret it as (31-a), hence the marking ‘(??)’. Transforming Manner Adverbs into Subject-Oriented Adverbs 21 luctantly, unwillingly’, shibushibu ‘grudgingly, reluctantly’, shikatanaku ‘involuntarily’, yorokonde ‘happily’, shinpaisooni ‘anxiously’, ochitsuite ‘calmly’.17 If this observation is correct, then what does the presence/absence of mo tell us about the contrast between mental attitude adverbs and SS-oriented adverbs in Japanese? Specifically, first, why do SS-oriented adverbs, but not mental attitude adverbs, undergo the clausal/manner alternation and the mo attachment accordingly? And secondly, why do mental attitude adverbs (i.e. those that do not have the mo particle) show passive sensitivity but not SS-oriented adverbs? As for the first question, one possible solution is to propose that the comparison classes for mental attitude adverbs are fixed, for example as in (33), so that no matter where the adverb shows up in the structure – whether it is adjoined to VP or to T0 – the comparison class is never readjusted. Thus, adverbs are always interpreted as mental attitude adverbs and not as manner adverbs, in which case the comparison class would consist of a set of some specific kind of events that is denoted by the VP. (33) JiyaiyaK = P (x)(e)∧ λP λxλe. reluctant(e) > standard(reluctant)([λe0 .Experiencer(e0 ) = x]) In order for an adverb to be shifted to the corresponding manner adverb, its comparison class needs to be adjusted to a set of some specific kind of events. However, since it is already fixed as a set of events that an individual x with sentience is involved in as an experiencer, there is no way it can be interpreted as a manner adverb. As for the second question, I suggest that SS-oriented adverbs with mo are only allowed to attach to T0 . This explains why it is always oriented to the surface subject. In order for the adverb to be oriented to the underlying subject, it needs to be adjoined to the lower VP (i.e. below the higher VP) according to the previous approaches (McConnell-Ginet 1982, Wyner 1998). My suggestion is that this is ruled out since the adverb obligatorily adjoins to T0 . If this idea is on the right track, then one might wonder if there are any other kinds of adverbs in Japanese that undergo the mo attachment and whether this explanation still holds. In the next section, we briefly look at one such case – a class of adverbs called evaluative adverbs. 5 Evaluative Adverbs Let us step back at this point and see where we are. So far, we have mainly discussed subject-oriented adverbs such as stupidly – especially those that are called agent-oriented adverbs in English and what I call SS-oriented adverbs in Japanese (section 3). We also compared them to mental attitude adverbs, 17 To my knowledge, the only exception is nesshinni-mo ‘eagerly’. 22 Ai Kubota the other subkind of subject-oriented adverbs, such as reluctantly (section 4). I showed that, in Japanese, SS-oriented adverbs behave quite differently from mental attitude adverbs with respect to the clausal/manner alternation and passive sensitivity. Coincidentally, both phenomena involved the existence/absence of mo. First, I observed that SS-oriented adverbs in Japanese, which happen to be typically attached by mo, are not passive sensitive: passive sentences with SS-oriented adverbs are not ambiguous, contrary to English passive sentences with agent-oriented adverbs. On the other hand, mental attitude adverbs in Japanese, which happen to be mo-less are passive sensitive: the passive sentences with mental attitude adverbs are ambiguous, as it is also the case in English passive sentences with mental attitude adverbs. Based on this fact, I proposed that the existence of mo indicates that the location of adverbs have to be T0 or somewhere external to VP. The purpose of this section is to examine whether this proposal is on the right track. To do so, I now turn to another class of adverbs called ‘evaluative adverbs’, such as fukooni-mo ‘unfortunately’ in the following example, koounni-mo ‘fortunately’, arigataku-mo ‘thankfully’, kimyooni(-mo) ‘oddly’, and mezurashiku(-mo) ‘uncommonly’. (34) Fukooni-mo taifuu-ga shima-o chokugeki-shita. Unfortunately typhoon-nom island-acc direct.hit-did ‘Unfortunately, a typhoon hit the island.’ As one may notice from the example above, this is another class of adverbs in Japanese that also involves mo. Although I will not propose an explicit analysis of evaluative adverbs in this paper, I will examine the semantic property of evaluative adverbs and the semantic contribution of mo. In particular, I will show that the idea that mo-attached adverbs are always located somewhere higher in the structure such as T0 is compatible with existing analyses on evaluative adverbs such as Sawada (1978) and Potts (2005). Before reviewing their analyses, let us briefly look at the overall semantic properties of evaluative adverbs. 5.1 Semantic properties of evaluative adverbs As I introduced in section 1, citing Ernst (2002), and also repeated below, evaluative adverbs are not classified as subject-oriented adverbs but as a subtype of speaker-oriented adverbs. (35) a. Subject-oriented adverbs: (i) Agent-oriented adverbs (stupidly, cleverly, rudely, etc.) (ii) Mental attitude adverbs (reluctantly, calmly, willingly, etc.) b. Speaker-oriented adverbs: (i) Speech-act adverbs (honestly, frankly, roughly, etc.) (ii) Epistemic adverbs (perhaps, probably, maybe, etc.) (iii) Evaluative adverbs (unbelievably, unfortunately, luckily, etc.) Transforming Manner Adverbs into Subject-Oriented Adverbs 23 Evaluative adverbs “represent the speaker’s evaluation of some state of affairs according to how good or bad it is (luckily, unfortunately), how (ab)normal it is (normally, strangely, curiously, surprisingly), its desirability (ideally, preferably), or a wide range of other criteria (e.g., for significantly, absurdly, conveniently, astonishingly, etc.)” (Ernst 2002: 76). While many of the adverbs that are categorized as evaluative adverbs are mo attached adverbs, it is not the case that all of them are, e.g., ainiku(*-mo) ‘unfortunately’, narubeku(*-mo) ‘preferably’, futsuuni(*-mo) ‘normally’. Adverbs that are not compatible with mo are those that are not morphologically derived from adjectives, as it is also the case with SS-oriented adverbs. For the purpose of this section, I will focus on those that are derived from adjectives and for which mo is obligatory. Unlike the mo-attachment on SS-oriented adverbs, however, the mo-attachment on evaluative adverbs is in some cases optional. Morimoto (1994) reports that there are three morphological types for evaluative and subject-oriented adverbs, with respect to the presence of mo. (36) a. b. c. Not compatible with mo ainiku(∗-mo) ‘unfortunately’ Mo is optional saiwai(ni(-mo)) ‘fortunately’, kinodokuni(-mo) ‘poorly’, kanshinni(-mo) ‘impressively’, mezurashiku(-mo) ‘uncommonly’, unwaruku(-mo) ‘unluckily’ Mo is obligatory fushigini-mo ‘mysteriously’, kannyooni-mo ‘generously’, zannenni-mo ‘unfortunately; sadly; to one’s regret/disappointment’, hinikuni-mo ‘ironically’, orokani-mo ‘stupidly’ According to Morimoto (1994), mo is obligatory for the adverbs that have a corresponding manner adverbial meaning, thus mo’s function is to distinguish evaluative adverbs from their corresponding manner adverbs. However, this explanation is not adequate as the following examples show. (37) odotta. a. *John-wa fushigini John-top mysteriously danced Intended: ‘John danced mysteriously.’ b. *John-wa hinikuni hanashita. John-top ironically spoke Intended: ‘John spoke ironically.’ c. *John-wa zannenni koronda. John-top unfortunately tumble.over Intended: ‘John tumble over in an unfortunate way.’ If Morimoto’s (1994) explanation is correct, the sentences above should be grammatical and all of them should represent manner meanings, because they are the adverbs that are supposed to take mo obligatorily in their evaluative uses as listed in (36-c). However, these examples are unacceptable, as Morimoto 24 Ai Kubota (1994) notices herself, indicating that they cannot be used as manner adverbs, at least by just taking mo away. At the moment, I do not have an alternative explanation for the reason of mo-optionality with some evaluative adverbs. What is clear, though, is that the morphological process of mo-attachment for adverbs that are derived from adjectives is a phenomenon that is not limited to SS-oriented adverbs but also commonly found in evaluative adverbs. As the examples in (37) show, most evaluative adverbs in Japanese do not have the manner adverbial meaning. Instead of being interpreted as manner adverbs, most of them typically appear as part of small clauses of some kind with verbs such as omou ‘to think’, kanjiru ‘to feel’, and kikoeru ‘to be heard’. The -ni endings of the following adverbs indicate that they are categorically adverbs rather than adjectives, in which case they should take either the na ending or a copular verb followed by a complementizer (e.g., fushigi-da-to omou ‘think that it is mysterious’). (38) a. b. c. omou. Kono-dekigoto-o fushigini This-incident-acc mysteriously think ‘I think this incident mysterious.’ Kono-hanashi-wa hinikuni omowareru. This-story-top ironically seem ‘This story seems ironic.’ kanjirareru. Kare-no fusanka-ga zannenni He-gen absence-acc unfortunately feel.can ‘I feel unfortunate of his absence.’ In English, on the other hand, many adverbs have both manner and evaluative meanings such as oddly, appropriately, and luckily. The examples in (39) and (40) below are from Ernst (2002) and Potts (2005) respectively. (39) a. b. Oddly, Carol was dancing. Carol was dancing oddly. (Evaluative) (Manner) (40) a. b. Luckily, Willie won the pool tournament. Willie won the pool tournament luckily. (Evaluative) (Manner) The difference between the two sentences in (39) is that “[i]n [(39-a)], on the one hand, there is a fact about an event of dancing by Carol that is odd; that is, it is odd that there is such an event, considering among other relevant facts in context. The manner reading in [(39-b)], on the other hand, says that there is a (Spec)Event [i.e., specific event] of dancing by Carol, which is odd compared to other dancing events” (Ernst 2002: 76). Thus, the difference between the two meanings can be explained in terms of comparison classes. For evaluative adverbs, the comparison class consists of propositions (facts), whereas for manner adverbs, it consists of a specific kind of events as mentioned before (i.e., various kinds of dancing in the above example). Whether it can be judged odd or not is determined by the relevant comparison class in the discourse. Transforming Manner Adverbs into Subject-Oriented Adverbs 25 As Ernst’s (2002) description above indicates, it seems that evaluative adverbs modify something clausal or propositional, whereas manner adverbs modify VPs. The idea that evaluative adverbs modify something clausal or propositional seems to be quite common among other researchers, although details may differ. In the following, I review a couple of analyses on evaluative adverbs. 5.2 Sawada (1978) Sawada (1978) uses various sentential operators to show that evaluative adverbs (along with other adverbs that he calls ‘sentential adverbs’) are not part of the propositional level (or what he calls the ‘propositional stratum’) but are included in a higher level (or the ‘attitudinal stratum’ in his words). This is based on the observation that sentential adverbials are not included in the scope of question (41-a), negation (41-b), imperative (41-c), and sentential pronominalization (41-d) (Quirk et al 1972, Schreiber 1971). (41) a. *Does he fortunately know about it? b. Obviously, he doesn’t want us to help him. (obviously > NEG, *NEG > obviously) c. *Country road, fortunately take me home. d. A: Clearly, Hitler was a madman. B: That’s false. (That = ‘that Hitler was a mad man’) Since the scope of question, negation, imperative and sentential pronominalization is supposed to be limited to the propositional level, sentential adverbials, which are argued to belong to a higher level, are not included in the scope of such operators. This seems to be true for SS-oriented adverbs and evaluative adverbs in general. The examples (42)-(45) correspond to the pattern in (41). (42) ?*{Orokani-mo/Fukooni-mo} John-wa odotta-no-desu ka? {stupidly/unfortunately} John-top danced-n-cop Q ‘Stupidly/Unfortunately, did John danced?’ (43) {Orokani-mo/Fukooni-mo} John-wa odora-nakatta. {stupidly/unfortunately} John-top dance-neg.past ‘Stupidly/Unfortunately, John didn’t dance.’ (stupidly> NEG, *NEG > stupidly) (44) *{Orokani-mo/Fukooni-mo} odori-nasai. {stupidly/unfortunately} dance-imperative ‘Stupidly/Unfortunately, dance!’ (45) A: John-wa {orokani-mo/fukooni-mo} Mary-o dakishimeta. John-top {stupidly/unfortunately} Mary-acc hug.past ‘Stupidly/Unfortunately, John hugged Mary.’ 26 Ai Kubota B: Sore-wa uso-da yo. That-top false-cop sfp ‘That’s false.’ (Sore = ‘that John hugged Mary’) In contrast to SS-oriented adverbs and evaluative adverbs, mental attitude adverbs and manner adverbs can be included in the scope of question, negation, imperative, and pronominalization, as shown in (46)-(49). (46) {Iyaiya/orokani} John-wa odotta-no-desu ka? {reluctantly/stupidly} John-top danced-n-cop Q ‘Did John danced reluctantly/stupidly?’ (47) {Iyaiya/orokani} John-wa odora-nakatta. {reluctantly/stupidly} John-top dance-neg.past ‘John didn’t dance reluctantly/stupidly.’ (stupidly> NEG, NEG > stupidly) (48) {Iyaiya/orokani} odori-nasai. {reluctantly/stupidly} dance-imperative ‘Dance reluctantly/stupidly!’ (49) A: John-wa {Iyaiya/orokani} Mary-o dakishimeta. John-top {reluctantly/stupidly} Mary-acc hug.past ‘John reluctantly/stupidly hugged Mary.’ B: Sore-wa uso-da yo. That-top false-cop sfp ‘That’s false.’ (Sore = ‘that John reluctantly/stupidly hugged Mary’) This seems to suggest that mental attitude adverbs and manner adverbs in Japanese, unlike SS-oriented adverbs, belong to a propositional level. It is worth noting that the existence of mo coincides with the contrast above. That is, the adverbs with mo are external to the propositional level, whereas those without mo are part of the propositional level. However, just from this observation, it is not clear what semantic difference there is between the adverbs with mo and those without it, nor how to formalize the semantic contribution of mo. As it was mentioned by one of the reviewers, the distinction between ‘propositional stratum’ and ‘attitudinal stratum’ by Sawada (1978) is reminiscent of Potts’s (2005) idea of ‘at-issue meaning’ and ‘CI meaning’. In the next subsection, I will turn to Potts’s (2005) analysis of evaluative adverbs. 5.3 Potts (2005) Potts (2005) notices the importance of the comma intonation accompanied by evaluative adverbs. For example, the adverb luckily with the comma intonation as in (50) is interpreted as an evaluative adverb, whereas the one without the comma intonation as in (51) is interpreted as a manner adverb. Transforming Manner Adverbs into Subject-Oriented Adverbs (50) a. b. c. Luckily, Willie won the pool tournament. Willie, luckily, won the pool tournament. Willie won the pool tournament, luckily. (51) a. b. Willie luckily won the pool tournament. Willie won the pool tournament luckily. 27 According to Potts (2005), evaluative adverbs (among other adverbs that he calls ‘supplementary adverbs’) introduce multidimensional meanings by contributing a CI proposition, but manner adverbs do not. He proposes the interpretation of the adverb luckily as follows. It shows that the adverb luckily has two possible extensional realizations. (52) luckily { λf λx.lucky(f (x)) : hhea , ta i, hea , ta ii λp.lucky(p) : hta , ta i Without the comma intonation, luckily only contributes an at-issue meaning. When the comma intonation is involved, the comma intonation (comma below) converts the manner adverb luckily in (52) into an evaluative adverb. (53) comma λP.P : hhta , ta i, hta , tc ii Potts (2005) presents the following two analyses: the one with the manner adverb luckily (54), and the one with the evaluative adverb luckily (55). (54) lucky(win(the(tournament))(willie)): ta lucky: hta , ta i (55) win(the(tournament))(willie): ta (win(the(tournament))(willie)): ta • comma(lucky)(win(the(tournament))(willie)):tc comma(lucky): hta , tc i win(the(tournament))(willie): ta lucky:hta , ta i The only difference between the two structures above is whether there is comma involved or not. In (54), the meaning is one-dimensional and there is no CI meaning involved. In (55), on the other hand, the meaning is multidimensional: the at-issue meaning simply says that Willie won the tournament, and the CI-meaning says that it is lucky that Willie won the tournament. Importantly, comma is essentially an identity function (λP.P ) as shown in 28 Ai Kubota (53). Therefore, the at-issue meaning in (54), and the CI-meaning in (55) are basically the same. Apparently, then, the only difference between the manner adverbial meaning and the clausal (evaluative) meaning is whether the adverbial meaning is included in the at-issue meaning, in which case it is interpreted as a manner adverb, or in the CI meaning, in which case it is interpreted as an evaluative or clausal adverb. While the idea that manner adverbs and evaluative adverbs make meaning contributions to different levels or dimensions of meaning can explain some of the differences between the two classes of adverbs, this alone is not enough for explaining other semantic differences between the two. For example, consider a case in which there are two adverbs, luckily and unluckily, one being a manner adverb and the other being an evaluative adverb. The following example is from Potts (2005) (although he does not give an explicit analysis for it). (56) Unluckily, Willie luckily won the pool tournament. Following the derivations (54)-(55), the sentence should be analyzed as follows. (57) lucky(win(the(tournament))(willie)): ta • comma(unlucky)(lucky(win(the(tournament))(willie))):tc comma(unlucky): hta , tc i lucky(win(the(tournament))(willie)): ta unlucky:hta , ta i The adverb luckily, which is a manner adverb in the given example, appears not only in the at-issue meaning but also in the CI meaning, and the CI meaning has both lucky and unlucky. This apparently is a contradiction, since comma is essentially an identity function (λP.P ) as in (53), which leads us to expect that the CI meaning is interpreted as something like ‘it is unlucky that it is lucky that Willie won the tournament’. In order to interpret the sentence correctly, we need the adverb luckily to be interpreted as a manner adverb and not as an evaluative adverb. However, it is not clear how lucky in the CI meaning of the example (57) is to be interpreted as a manner adverb, whereas lucky in the CI meaning of the example (55) is interpreted as an evaluative adverb. This problem arises because, as it is, Potts’s analysis does not take into account the difference between manner meanings and evaluative meanings. In other words, in addition to the idea that manner adverbs and evaluative adverbs make meaning contributions to different dimensions of meaning, it is also necessary to consider subtler semantic difference between the two adverbs, such as the different comparison classes involved as in the analysis I have proposed above. Importantly, for Potts (2005), evaluative adverbs are derived from manner adverbs and what makes them evaluative adverbs is the existence of comma Transforming Manner Adverbs into Subject-Oriented Adverbs 29 (the comma intonation). Thus, from this perspective, Japanese mo can be thought of as a morpheme that converts manner adverbs into evaluative (or other CI contributing) adverbs. If so, it is interesting to see from a crosslinguistic perspective that, in one language, the conversion is marked with intonation, while in another language, there is a specific overt morpheme that marks the conversion. Furthermore, from a theoretical point of view, it is also interesting to see if it is possible to explain how the comparison class is determined with the dimension difference. However, to adopt Potts’s (2005) analysis for Japanese evaluative adverbs, it is necessary to adjust the system so that it can explain the fact that evaluative adverbs in Japanese do not have manner adverbial meanings even without mo (see the examples in (37)). 5.4 Summary In this section, I showed that the mo attachment on adverbs is not limited to SS-oriented adverbs but is also found in a subclass of speaker-oriented adverbs, namely evaluative adverbs. The upshot is that the presence of mo seems to indicate that the adverb is located higher in the structure compared to mental attitude adverbs which happen to be mo-less. Importantly, the proposed analysis in section 3 does not force mo to transform the adverb strictly into SS-oriented adverbs. Rather, it simply readjusts the comparison class of the adverb and let the adverb be located higher in the structure. This leaves room for extending the analysis further to deal with evaluative adverbs briefly discussed in this section. 6 Conclusion In this paper, I introduced the phenomenon of manner/clausal alternation with particular adverbs in English and in Japanese. Among the previous analyses of this phenomenon in English (Ernst 2002, McConnell-Ginet 1982, Piñón 2010), none of them was readily adoptable for the Japanese data. Ernst’s (2002) analysis, which derives manner meanings from clausal meanings, was not suitable, because in Japanese the clausal adverbs are morphologically complex (they are the combination of manner adverbs and mo). McConnell-Ginet’s (1982) analysis, which assumes the abstract verb act that is responsible for agentivity, was not suitable either, because clausal adverbs like orokani-mo ‘stupidly’ are not strictly agent-oriented adverbs but rather surface-subject(SS)-oriented adverbs and furthermore can appear with apparently agent-less verbs such as sinu ‘to die’. I proposed a compositional analysis that derives the SS-oriented meaning from the combination of manner meaning and mo. In my analysis, mo has a semantic effect (i.e. readjusting the comparison class of the adverb) and a syntactic effect (i.e. forcing the adverb to be T0 -adjoined). This proposal was supported by the fact that SS-oriented adverbs in Japanese, unlike agentoriented adverbs in English, do not have the agent-oriented reading in passives, 30 Ai Kubota i.e., they are never passive sensitive. To further examine if this proposal is correct, I looked at another class of adverbs, namely evaluative adverbs, which also involves the mo-attachment. I consulted two analyses, Sawada (1978) and Potts (2005), on the contrast between manner adverbs and evaluative adverbs, and concluded that the proposal about mo is not incompatible with them, although their main idea that manner adverbs and evaluative adverbs make meaning contributions to different levels or dimensions of meaning is not enough for explaining the semantic difference between the two adverbs. This paper has three main contributions: (i) it captures parallel properties between adverbs and adjectives by adopting the notion of comparison class, (ii) it proposes a hybrid analysis of the previous studies (the notion of comparison class from Ernst 2002 and the idea of deriving clausal adverbs from manner adverbs in McConnell-Ginet 1982) so that it naturally explains a morphological fact in Japanese adverbs, and (iii) it accounts for the nature of SS-oriented adverbs in Japanese that look like agent-oriented adverbs in English, and also explains the phenomenon of manner/clausal alternation that is observed with both SS-oriented and agent-oriented adverbs. From a cross-linguistic perspective, this paper sheds light on the uniformity and diversity of the classification of adverbs across languages. For example, while there are some phenomena that are cross-linguistically observed, such as manner/clausal alternation and passive sensitivity, on the other hand, there are cross-linguistic differences: while the manner/clausal alternation involves morphological marking (mo) in Japanese, it is intonational marking (the comma intonation) in English. Furthermore, while clausal adverbs such as stupidly in English at first sight seem to have equivalent adverbs in Japanese (orokani-mo), it turns out that such adverbs in Japanese are strictly surface-subject-orietend unlike the corresponding adverbs in English. Since this paper focuses on the manner/clausal alternation of agent/SS-oriented adverbs, extending the analysis to other classes of adverbs (e.g., evaluative adverbs) is left for future research. Acknowledgements I would like to thank Thomas Ernst, Berit Gehrke, Yusuke Kubota, Marcin Morzycki, the audiences at the Workshop on Modification (with & without modifiers), and an NLLT reviewer for extensive and helpful comments and discussion. References Anderson C, Morzycki M (2014) Degrees as kinds. Natural Laguage and Linguistic Theory This volume Dowty D (1979) Word Meaning and Montague Grammar – The Semantics of Verbs and Times in Generative Semantics and in Mongague’s PTQ. D. Reidel Publishing Company Ernst T (2002) The Syntax of Adjuncts. Cambridge University Press Gehrke B (2014) Different ways to be passive. Natural Language and Linguistic Theory This volume Transforming Manner Adverbs into Subject-Oriented Adverbs 31 Geuder W (2002) Oriented adverbs: issues in the lexical semantics of event adverbs. PhD thesis, Universität Tübingen Jackendoff R (1972) Semantic Interpretation in Generative Grammar. MIT Press Kennedy C (1997) Projecting the adjective: The syntax and semantics of gradability and comparison. PhD thesis, UC Santa Cruz Landman M, Morzycki M (2003) Event-kinds and the representation of manner. In: Antrim NM, Goodall G, Schulte-Nafeh M, Samiian V (eds) Proceedings of the Western Conference on Linguistics (WECOL) 2002, California State University, Fresno, vol 14, pp 136–147 McConnell-Ginet S (1982) Adverbs and ligical from: A linguistically realistic theory. Language 58(1):144–184 Morimoto J (1994) Hanashite no shukan o arawasu fukushi ni tusite [On Adverbs That Represent the Speaker’s Subjectivity]. Kuroshio, Tokyo Partee B (1977) John is easy to please. In: Zampolli A (ed) Linguistic Structures Processing, North-Holland Publishing, pp 281–312 Piñón C (2010) What to do with agent-oriented adverbs. hanout Potts C (2005) The Logic of Conventional Implicature. Oxford University Press Quirk R, Greenbaum S, Leech G, Svartvik J (1972) A Grammar of Contemporary English. Longman Group United Kingdom Rawlins K (2008) Unifying ‘illegally’. In: Johannes Dölling THZ, Schäfer M (eds) Event Structures in Linguistic Form and Interpretation, Mouton de Gruyter, pp 81–102 Sawada H (1978) Nichi-eego bunfukushi-rui no taishoo gengogaku-teki kenkyuu [A contrastive study of japanese and english sentence adverbials: From the viewpoint of speech act theory]. Gengo Kenkyu 74:1–36 Schreiber PA (1971) Some constraints on the formation of English sentence adverbs. Linguistic Inquiry 2(1):83–101 von Stechow A (1984) Comparing semantic theories of comparison. Journal of Semantics 3:1–77 Wyner AZ (1998) Subject-oriented adverbs are thematically dependent. In: Rothstein S (ed) Events in Grammar, Kluwer Academic Publishers, Dordrecht, pp , 333–348 Wyner AZ (2008) Towards flexible types with constraints for manner and factive adverbs. In: McNally L, Kennedy C (eds) Adjectives and Adverbs: Syntax, Semantics, and Discourse, Oxford University Press
© Copyright 2026 Paperzz