Two Classes of Intensifiers in Mandarin Chinese: From Subjectivity to Evidentiality

Abstract

This article investigates two distinct classes of intensifiers in Mandarin Chinese, ordinary intensifiers (Class I) and subjective intensifiers (Class II). Intensifiers from Class II, such as zhēn “really”, cannot be used in the following cases: (i) interrogatives, (ii) sentences containing epistemic modals, (iii) sentences with negation or as conditional antecedents, (iv) sentences containing attitude predicates, and (v) contexts lacking firsthand experience. This paper argues that evidentiality, as a conceptual framework concerning the source of knowledge, can account for these phenomena related to Class II intensifiers. Specifically, in this study, evidentiality constraints on both the subject and the manner of the information source. The subject must be the speaker, and the information must be acquired as firsthand experience.

1. Introduction

Intensifiers, such as very, quite and extremely in English, are degree adverbs that can modify gradable adjectives big, beautiful and tasty, thereby enhancing the degrees of expressions. They cannot be applied feLicitously to non-gradable adjectives such as huge, empty and tiny [1].

Previous studies have observed that Mandarin intensifiers exhibit systematic differences ([2,3,4], a.o.). This paper further proposes that there are two classes of intensifiers in Mandarin Chinese. Class I includes ordinary intensifiers like hěn “very”, fēicháng “extremely”, and tèbié “especially”, and Class II consists of intensifiers with more restricted functions, such as zhēn “really”, hǎo “well” and duō “much”.

Class I intensifiers can appear broadly across indicative and expressive sentences. Class II intensifiers, however, cannot be used in the following cases: (i) interrogatives, (ii) sentences with epistemic modals, (iii) sentences with negation or as conditional antecedents, (iv) sentences with attitude predicates, and (v) contexts without firsthand experience. Since sentences with the intensifiers in Class II are often speaker-oriented and expressive, these intensifiers are referred to in the literature as subjective intensifiers, see [4].

The first three situations (i–iii) have been discussed in the research of Mandarin degree adverbs [2,3,4]. The properties of their combination with the attitude predicates (iv) and the requirement on firsthand experience (v) are newly identified in this paper for these intensifiers. With these additional properties, previous accounts can no longer provide a fully adequate explanation.

I propose that the source of knowledge, i.e., evidentiality [5], provides a unified explanation for these otherwise disparate restrictions. Thus, I suggest that the subjectivity is connected to evidentiality. While ref. [3] observe that the speaker is committed to the proposition, I further argue that this commitment is specifically based on the speaker’s firsthand knowledge.

This paper is organized as follows. Section 2 presents empirical observations concerning the properties of the two classes of intensifiers. Section 3 discusses previous analyses on the intensifiers in Class II. Section 4 proposes an evidentiality-based account. Section 5 concludes.

2. Properties of Two Classes of Intensifiers

To illustrate the contrast between the two classes of intensifiers, I select three representative items from each group. For Class I, the class of ordinary intensifiers, I focus on the commonly used forms hěn ‘very’, fēicháng ‘extremely’, and tèbié ‘especially’. For Class II, I examine zhēn ‘really’, hǎo ‘well’, and duō `much’1. The intensifier hǎo “well”, though literally translated as “well” or “good”, has a similar meaning of hěn “very” when used as a degree adverb. Similarly, duō, whose literal meaning is “many/much”, can also serve as an intensifier, indicating that the degree expressed by the predicate is exceptionally high2.

For convenience, in the glosses below, hǎo is glossed and translated as “well”, duō as “much”. These exemplary intensifiers are summarized as follows.

Class I: hěn “very”, fēicháng “extremely” and tèbié “especially”

Class II: zhēn “really”, hǎo “well” and duō “much”

2.1. Interrogatives

The first difference between intensifiers in Class I and Class II is the use in interrogative sentences. While the former can be used with question marker ma to form corresponding polar questions, the latter cannot, as shown in (3) and (2). In (2), all the sentences are grammatical as they are declarative sentences. However, once the interrogative particle ma is added, the sentences become ungrammatical. This property has also been discussed in [2,3,4], etc.

(1)

Class I intensifiers + (inter)3

a.

zhègè dàngāo hěn měiwèi (ma)

this-clf cake very tasty (inter)

Without ma: “This cake is very tasty.”

With ma: “Is this cake very tasty?”

b.

zhègè dàngāo fēicháng měiwèi (ma)

this-clf cake extremely tasty (inter)

Without ma: “This cake is extremely tasty.”

With ma: “Is this cake extremely tasty?”

c.

zhègè dàngāo tèbié měiwèi (ma)

this-clf cake especially tasty (inter)

Without ma: “This cake is especially tasty.”

With ma: “Is this cake especially tasty?”

(2)

Class II intensifiers + (inter)

a.

zhègè dàngāo zhēn měiwèi (*ma) 4

this-clf cake really tasty (inter)

Without ma: “This cake is really tasty.”

With ma: “Is this cake really tasty?”

b.

zhègè dàngāo hǎo měiwèi (*ma)

this-clf cake well tasty (inter)

Without ma: “This cake is well tasty.”

With ma: “Is this cake well tasty?”

c.

zhègè dàngāo duō měiwèi (*ma)

this-clf cake much tasty (inter)

Without ma: “This cake is pretty much tasty.”

With ma: “Is this cake pretty much tasty?”

Class II intensifiers are also incompatible with wh-questions. As illustrated in (3b), neither what-question shénme nor which-questions nǎge are acceptable with zhēn “really”, hǎo “well” or duō “much”.

(3)

Class I & II intensifiers in wh-questions

a.

shénme/nǎge  dàngāo hěn/fēicháng/tèbié  měiwèi?

what/which cake very/extremely/especially tasty

“What kinds of/Which cake is very/extremely/especially tasty?”

b.

* shénme/nǎge  dàngāo zhēn/hǎo/duō  měiwèi?

what/which cake really/well/much tasty

“What kinds of/Which cake is really/well/much tasty?”

Importantly, this observation applies not only for subjective predicates or predicates of personal tastes (PPTs) like beautiful, tasty and tasty, but also for objective predicates like tall, heavy and red. As exemplified below in (4), (4a) are all grammatical sentences with gāo “tall” modified by Class I intensifiers, hěn “very”, fēicháng “extremely” and tèbié “especially”, while (4b) are only acceptable in declaratives not interrogatives with gāo “tall” modified by Class II intensifiers, i.e., zhēn “really”, hǎo “well” and duō “much”.

(4)

Class I & II intensifiers + objective predicates + (inter)

a.

yuànzǐ lǐ  de shù xiànzài hěn/fēicháng/tèbié        gāo le (ma)

yard inside mod tree now very/extremely/especially tall pfv (inter)

Without ma: “The tree in the yard has been very/extremely/especially tall now.”

With ma: “Has the tree in the yard been very/extremely/especially tall now?”

b.

yuànzǐ lǐ  de shù xiànzài zhēn/hǎo/duō        gāo le (*ma)

yard inside mod tree now really/well/much tall pfv (*inter)

Without ma: “The tree in the yard has been really/well/much tall now.”

With ma, intended: “Has the tree in the yard been really/well/much tall now?”

2.2. Epistemic Modality

The second distinction concerns the interaction with epistemic modals. Class I intensifiers are fully compatible with epistemic modals, whereas Class II intensifiers are not, as illustrated in (5) and (6b). Sentences in (5) with kěnéng “might” in (6b) with yídìng “must” are all unacceptable when the Class II intensifier appears.

This restriction is noted in [3], though not discussed there in detail.

(5)

Class I & II intensifiers + might

a.

xiǎowáng kěnéng hěn/fēicháng/tèbié  cōngmíng

Xiaowang might very/extremely/especially smart

“Xiaowang might be very/extremely/especially smart.”

b.

* xiǎowáng kěnéng zhēn/hǎo/duō  cōngmíng

Xiaowang might really/well/much smart

“Xiaowang might be really/well/much smart.”

(6)

Class I & II intensifiers + must

a.

tā zuò shì yídìng hěn/fēicháng/tèbié    rènzhēn

He do thing must very/extremely/especially serious

“He must be very/extremely/especially serious about his work”

b.

* tā zuò shì yīdìng zhēn/hǎo/duō    rènzhēn

He do thing must really/well/much serious

“He must be really/well/much serious about his work”

2.3. Truth-Functional Connectives: Negation and Conditional Antecedents

The third property of intensifiers in Class II is that they are incompatible with truth-functional connectives, e.g., with negation, and with conditional connectives [3,4]5.

(7a) and (7b) are examples for negation, and (8a) and (8b) are examples for conditional antecedents. Sentences in (7a) with intensifiers from Class I are acceptable with negation scoped over, while sentences in (7b) with intensifiers from Class II are unacceptable. In this case, the negation scopes over the intensifier. When the negation is outscoped by the intensifier, the unacceptability is canceled, which will be discussed further later in this paper. Given that other negators in Mandarin like bù and méi are incompatible with the intensifiers, including Class I intensifiers6, this paper restricts the discussion of negation to cases where búshì functions as the negator.

(7)

Class I & II intensifiers + negation

a.

qiǎokèlì  búshì hěn/fēicháng/tèbié      tián

chocolate neg very/extremely/especially sweet

“Chocolates are not very/extremely/especially sweet.”

b.

* qiǎokèlì búshì zhēn/hǎo/duō      tián

chocolate neg really/well/much sweet

“Chocolates are not really/well/much sweet.”

(8)

Class I & II intensifiers + conditionals

a.

rúguǒ tiānqì   hěn/fēicháng/tèbié     lěng tā  huì  chuānshàng yǔróngfú

if     weather very/extremely/especially cold she will put on   down coat

“If it is very/extremely/especially cold, she will put on a down coat.”

b.

* rúguǒ tiānqì     zhēn/hǎo/duō       lěng tā huì chuānshàng yǔróngfú

if     weather really/well/much cold she will put on   down coat

“If it is really/well/much cold, she will put on a down coat.”

2.4. Attitude Predicates

Another distinction between intensifiers from Class I and Class II is that the former can be embedded under attitude predicates while the latter cannot. Both with the attitude predicate zhīdào “know” and with xiāngxìn “believe”, only intensifiers in Class I can be embedded, and intensifiers in Class II cannot be embedded.

(9)

Class I & II intensifiers + believe

a.

wǒ xiāngxìn zhègè  shìjiè hěn/fēicháng/tèbié    měihǎo

I believe  this-clf world very/extremely/especially beautiful

“I believe this world is very/extremely/especially beautiful.”

b.

*/?

wǒ xiāngxìn zhègè  shìjiè zhēn/hǎo/duō    měihǎo

I believe this-clf world really/well/much beautiful

“I believe this world is really/well/much beautiful.”

(10)

Class I & II intensifiers + know

a.

wǒ zhīdào xiǎolǐ  hěn/fēicháng/tèbié    cōngmíng

I  know Xiaoli very/extremely/especially smart

“I know Xiaoli is very/extremely/especially smart.”

b.

? wǒ zhīdào xiǎolǐ  zhēn/hǎo/duō   cōngmíng

I  know Xiaoli really/well/much smart

“I know Xiaoli is really/well/much smart.”

Notably, there are different degrees of unacceptability between zhīdào “know” + Class II and xiāngxìn “believe” + Class II. Although both combinations sound unnatural, constructions with xiāngxìn “believe” are less acceptable than those with zhīdào “know”.

2.5. Acquaintance Inference

It is observed that there is an acquaintance inference that requires firsthand experience when using subjective expressions, such as PPTs and sensory perceptive verbs (tastes, feels, seems) (see [8,9,10], a.o.). In the case of Mandarin intensifiers, Class II intensifiers strongly require firsthand experience of the speaker.

To illustrate, consider scenarios involving watching a movie or tasting a cake. It is unacceptable to describe the movie or the cake with the intensifier zhēn “really” if the speaker has not personally seen the movie or tasted the cake before, as shown in (11b) and (12b). By contrast, Class I intensifiers are not strictly tied to firsthand evidence and may be used on the basis of indirect information such as hearsay. Thus, the sentences in (11a) and (12a) remain acceptable even when the speaker’s knowledge derives from someone else. For instance, if Mary tells John about a good movie that he has not yet seen, John may felicitously use the Class I sentences in (11a), though the Class II sentences in (11b) would still be unacceptable.

(11)

Class I & II intensifiers + a move-watching case

a.

zhèbù  diànyǐng hěn/fēicháng/tèbié    hǎokàn, suīrán  wǒ hái  méi

this-clf movie   very/extremely/especially nice   though I still neg

kòng   kàn

spare time watch

“This movie is very/extremely/especially nice, though I haven’t had time to watch it yet.”

b.

# zhèbù diànyǐng zhēn/hǎo/duō    hǎokàn, suīrán wǒ hái méi kòng

this-clf movie  really/well/much nice    though I still neg spare time

kàn

watch

“This movie is really/well/much nice, though I haven’t had time to watch it yet.”

(12)

Class I & II intensifiers + a cake-tasting case

a.

zhège   dàngāo hěn/fēicháng/tèbié   hǎochī, suīrán wǒ hái méi

this-clf cake  very/extremely/especially tasty though I still neg

kòng  chī

spare time taste

“This cake is very/extremely/especially tasty, though I haven’t had time to taste it yet.”

b.

# zhège   dàngāo zhēn/hǎo/duō   hǎochī, suīrán wǒ hái méi kòng

this-clf cake  really/well/much tasty  though I still neg spare time

chī

taste

“This cake is really/well/much tasty, though I haven’t had time to taste it yet.”

Moreover, firsthand experience of the speaker cannot be replaced by the experiencer’s firsthand experience. Consider a scenario in which Mary and her daughter Vina have both tasted a cake, and Mary’s husband John asks, “How does Vina feel about the cake?” Mary cannot answer on Vina’s behalf using (13b) with zhēn “really”. The utterance is infelicitous because Mary herself lacks the requisite firsthand experience for zhēn. In contrast, she can partially report Vina’s opinion using a Class I intensifier such as hěn “very”, as in (13a).

(13)

Class I & II intensifiers + different representatives

a.

mǎlì:   nàge   dàngāo  hěn  měiwèi

Mary this-clf cake  very tasty

(for Vina) Mary: “That cake was very tasty.”

(for herself) Mary: “That cake was very tasty.”

b.

mǎlì:   nàge   dàngāo zhēn  měiwèi

Mary this-clf cake  really tasty

#(for Vina) Mary: “That cake was really tasty.”

(for herself) Mary: “That cake was really tasty.”

More precisely, the relevant “speaker” is the utterance context’s speaker, not the individual whose words are being reported. Consequently, Class II intensifiers such as zhēn cannot appear in indirect speech reports, as shown in (14b). A sentence like “Vina said that the cake was really tasty” or “Vina told me that the cake was really tasty” with zhēn is acceptable only under direct quotation or in free indirect discourse, as in narrative contexts (see [11] for more discussion on free indirect discourse).

(14)

Class I & II intensifiers + indirect report

a.

mǎlì:   wéinà shuō/gàosù-wǒ nàge  dàngāo hěn měiwèi

Mary Vina  say/tell-me  this-clf cake  very tasty

Mary: “Vina said/told me that cake was very tasty.”

b.

* mǎlì:   wéinà shuō/gàosù-wǒ nàge  dàngāo zhēn měiwèi

Mary Vina  say/tell-me  this-clf cake  really tasty

Mary: “Vina said/told me that cake was really tasty.”

There is also some heterogeneity among the three Class II intensifiers. To account for this, I appeal to the type–token distinction. In discussing subjective predicates, ref. [10] notes that firsthand experience may be based on encountering the same type of stimulus rather than the same token. For example, one may form a firsthand judgment about the taste of Massaman curry based on having tried the same dish at a different restaurant, even if not the exact token in question, see examples below.

(15)

a. Type: Massaman curry is tasty, I’ve tried it before at another restaurant.

b. Token: This Massaman curry is tasty, but I haven’t tried it yet.

                              examples from [10]

However, among the Mandarin subjective intensifiers, zhēn “really” emphasizes the token side rather than the type side, setting it apart from hǎo “well” and duō “much”. Take the following scenario as an example. Mary and John were shopping in the supermarket, and Mary found several packages of strawberry-flavored Oreo cookies on the shelf. She has tasted the same flavor of cookies before, and she wanted to recommend that to John, like in (16). Since Mary has firsthand experience about these cookies, according to the discussion above, it should have been acceptable to use both intensifiers from Class I and Class II. Nonetheless, the intensifier zhēn “really” is not allowed in this scenario, see (16a), while other intensifiers from Class II like hǎo “well” are acceptable, see (16b). The sentence containing zhēn ‘really’ in (16a) becomes acceptable only once Mary opens the package and tastes that specific token of the cookies at home.

(16)

Class II intensifiers + type/token differences

a.

mǎlì:   zhèkuǎn   bǐnggān zhēn   měiwèi

Mary this-clf cookie really tasty

#(in the supermarket) Mary: “This flavor of cookies is really tasty.”

(at home, after tasting the cookie) Mary: “This flavor of cookies is really tasty.”

b.

mǎlì:   zhèkuǎn   bǐnggān hǎo   měiwèi

Mary this-clf cookie well tasty

(in the supermarket) Mary: “This flavor of cookies is very (“well”) tasty.”

(at home, after tasting the cookie) Mary: “This flavor of cookies is very (“well”) tasty.”

3. Previous Research on Subjectivity

3.1. [4]: The Subjectivity of Class II Intensifiers

Ref. [4] also recognizes the inconsistency behind a group of Class II intensifiers. Drawing on the subjectivity-calculation framework proposed by [12], Wang assigns numerical subjectivity values to these items. As such, zhēn “really” is the most subjective intensifier with the value of 0.9, and followed by duō “much” (0.62), lǎo “oldly/very” (0.57), hǎo “well” (0.53) and tài “too” (0.41) in sequence. This research examines the subjectivity in intensifiers of Class II and acknowledges the gradient nature within the same class. It is particularly evident as zhēn “really” exhibits a higher degree of subjectivity compared to other intensifiers.

Nonetheless, only the subjectivity itself cannot elucidate the underlying difference between intensifiers from Class I and Class II. Certain properties, such as the incompatibility with attitude predicates and the necessity of firsthand experience, are not encompassed in [12]’s list of subjective properties. More importantly, asserting that these intensifiers are subjective remains a descriptive generalization, differentiating one group of intensifiers from others. It does not provide an explanation for the subjective nature of these intensifiers, i.e., why they behave subjectively and where the subjectivity comes from. This gap also leaves room for the proposal advanced in this paper, which seeks to take a step back in order to explore the underlying reasons for subjectivity.

3.2. [3]: Resistance to Non-Veridicality and the Speaker-Oriented CI Expression

Ref. [3] explore the Class II degree intensifiers hǎo “well”, lǎo “oldly/very” and guài “strangely”, and argue that these adverbs are resistant to non-veridical contexts. Ref. [3] also apply a multidimensional compositional system in their analysis, which includes an at-issue dimension and a conventional implicature (CI) dimension [13,14]. Regarding hǎo “well”, they encode it in the CI dimension that the speaker is subjectively impressed by the proposition p. This approach highlights the expressive side of intensifiers in Class II, which echoes with the subjectivity mentioned above. Moreover, they also foreground the speaker-oriented attitude in their discussion. This also accounts quite well for the resistance to non-veridical contexts like negations, interrogatives, and conditional antecedents.

However, this discussion also lacks a comprehensive perspective on the property of acquaintance inference. The intensifiers in Class II like zhēn “really” present an interesting case as they demand firsthand experience from the speaker. This is not explainable with the speaker-oriented attitude only. In principle, a speaker may be entirely convinced of a proposition on the basis of belief or strong evidence, without direct experiential access. For example, even though Mary has heard the movie Titanic was very nice from her friends, Mary strongly trusts her friends’ opinions, Mary still cannot use (12b) with intensifiers from Class II. Likewise, observing her daughter eat a piece of cake and appear to enjoy it does not provide evidence robust enough for Mary to utter (14b). In both cases, indirect or inferential evidence fails to satisfy the requirement imposed by Class II intensifiers for the speaker’s own firsthand experience. The crucial criterion is that Mary herself has watched the movie or had the cake, rather than merely holding a belief or attitude about it.

Additionally, as mentioned in Section 2.4, veridical attitude predicates like zhīdào “know” are also incompatible with Class II Intensifiers. This shows that ref. [3]’s observation on the resistance to non-veridical cases is not enough in explaining the restrictions behind Class II intensifiers. While non-veridicality entails the resistance of subjective intensifiers, the resistance of subjective intensifiers does not necessarily entail non-veridical situations. This indicates that non-veridicality is a sufficient but not necessary condition for the use of Class II intensifiers.

Thus, it is insufficient to only consider the speaker-oriented attitude and non-veridicality for these subjective intensifiers; the next step is to integrate the requirement of firsthand experience and the speaker-oriented attitude mentioned above. I thus propose in the following section that evidentiality can be a possible way out for the integration.

4. A Proposal: Confirmed Knowledge and Evidentiality

4.1. Knowledge-Relevant Phenomenon

Rather than emphasizing non-veridicality, I propose that knowledge offers a more fundamental basis for distinguishing Class I from Class II intensifiers. This knowledge requirement is sententially projective.

It is observed that all the different properties between Class II and Class I are relevant to knowledge. Among these, the acquaintance inference is the most obvious one, which has some direct requirement on the source of knowledge, i.e., firsthand experience.

Moreover, both interrogatives and epistemic modality are associated with knowledge. According to refs. [15,16], asking questions can be considered as a request for knowledge (apud [17]). Ref. [18] also argues that epistemic modals relativize the truth value of a proposition to the speaker’s knowledge. In sentences with intensifiers from Class II, the speaker is fully committed to the proposition, which means that the speaker has the knowledge stored to known facts. Therefore, it is contradictory to inquire about the knowledge from addressees; it means expressing certainty in an uncertain context. In short, the inherent uncertainty of interrogatives and epistemic modals is incompatible with the strong commitment encoded by Class II intensifiers.

The incompatibility with attitude predicates is also explained. When the relevant attitude is not fully factive, as with xiāngxìn “believe” or huáiyí “doubt”, it directly contradicts the commitment encoded by intensifiers in Class II. Predicates that presuppose a more secure epistemic stance, even factive predictes, such as zhīdào “know” or míngbái “understand”, are not contradictory but instead render the sentence redundant. Given that a Class II intensifier like zhēn “really” already signals that the proposition is confirmed knowledge for the speaker, it is inappropriate to introduce the proposition under discussion again with attitude predicates. These sentences are thus pragmatically infelicitous. This also accounts for the gradient effect observed earlier. Combinations of xiāngxìn “believe” with Class II intensifiers are strongly unacceptable due to direct contradiction, whereas combinations with zhīdào “know” are merely redundant and thus somewhat less degraded.

As for truth-functional connectives, by definition, each part needs to make truth value judgments to obtain individual truth-values, which can then be calculated in the truth value of compound statement. However, in the case of intensifiers from Class II, the speaker has already committed that it is true at the time of expression. Thus, the value of the statement can no longer be hypothesized, negated, or operated in other truth-functional connectives. Take the case of negation to illustrate. In (17), the negation scopes over the intensifier, and the subjective intensifier zhēn “really” is incompatible with this pattern. Under this knowledge-relevant framework, zhēn “really” has already confirmed the speaker’s judgment on the predicate cōngmíng “smart”, thus prohibiting the negation over the committed knowledge. In (18), however, zhēn “really” is acceptable in the sentence when it scopes over the negation.

(17)

The negator scopes over the intensifier

a.

tā búshì hěn/fēicháng/tèbié    cōngmíng

He neg  very/extremely/especially smart

“He is not very/extremely/especially smart. (He can be average)”

b.

* tā   búshì zhēn/hǎo/duō    cōngmíng

He neg  really/well/much smart

“He is not really/well/much smart. ”

(18)

The intensifier scopes over the negator

a.

tā   hěn/fēicháng/tèbié    bù   cōngmíng

He very/extremely/especially neg smart

“He is very/extremely/especially unwise. (He is very stupid)”

b.

tā   zhēn/hǎo/duō    bù    cōngmíng

He really/well/much neg smart

“He is really/well/much unwise. (He is really stupid)”

In her discussion of (non)veridicality and event actualization, ref. [19] also argues that “veridicality and non-veridicality rely on the speaker’s knowledge of whether an event is at least partially actualized,” treating the speaker’s knowledge as a crucial operator (p. 44). This further motivates linking the properties of subjective intensifiers to the broader notion of knowledge. When non-veridicality and knowledge are compared, the former cannot account for cases involving veridical attitude predicates or the requirement of firsthand experience, whereas the latter straightforwardly does. Hence, the concept of knowledge is more fundamental and more explanatory than the concept of non-veridicality.

4.2. Restriction on Evidentiality: The Subject and the Manner

Since we unify all the properties under the concept of knowledge, these subjective intensifiers can be regarded as more restricted on the knowledge in the conversation. Crucially, this restriction only has relation with the speaker not to the addressee. These features point towards an analysis in which that restriction of Class II intensifiers arise from the restriction on the evidentiality, the source of knowledge [5].

It is not rare in the literature to connect subjectivity with evidentiality. Ref. [20] claims that subjectivity involves the speaker’s evaluation of the quality of the evidence for that quantification, and this dimension can be labeled as evidentiality. He further argues that “subjectivity should not be defined in the traditional terms of the quality of the evidence for an (epistemically qualified) state of affairs, but rather in terms of the question whether the evidence (and the conclusion drawn from it) is only available to the speaker or is rather more widely known (including to the hearer) (p. 399).” Ref. [21] (p. 359) follows [20]’s line of argument, and concludes that “subjective evidentiality occurs when the evidence lies solely with the S[peaker]/W[riter].”

As for subjective intensifiers in Class II, the specific case in this study, I propose that evidentiality functions in two ways, the subject of evidence, and the manner of evidence. The subject of evidence requires that the knowledge must come from the speaker. This point has been explored in [3], although [3] mainly frame this in terms of the speaker’s attitude and belief. On the other dimension, the manner of evidence requires that the knowledge must be firsthand knowledge. As illustrated in Section 2.5, speakers cannot use intensifiers like zhēn “really”, hǎo “well” or duō “much” to express their opinions when they have not had relevant firsthand experience before. This restriction cannot be satisfied by indirect knowledge or secondhand knowledge from others.

With respect to the manner, evidentiality with a requirement similar to zhēn’s has been observed as immediate evidentials, which indicates the observation made at the immediate moment [22,23]. The intensifier zhēn actually has this in its sensitivity to type-token difference. The intensifier zhēn “really” asks for immediate firsthand knowledge, which means the evidence should be attained almost at the same moment of using the intensifier zhēn “really”.

Recall the case of cookies in (16). Packages of cookies on the shelf cannot provide immediate experience of their taste. Although Mary has tasted the same brand and same flavor of cookies before, that earlier experience is firsthand but not immediate anymore. The faded previous experience cannot work here for the situation in the supermarket. Consequently, she cannot felicitously use zhēn “really” while standing in the supermarket. She may only use it when she is eating or has just eaten that flavor of cookies.

Moreover, subject of evidence and manner of evidence are interconnected yet independent constraints. For instance, even when the subject has firsthand access, the construction remains unacceptable if the speaker lacks immediate evidence, as exemplified in the report case (13) and the indirect discourse case (14).

In sum, these subjective intensifiers exhibit a stricter evidential requirement than ordinary intensifiers. This requirement can be disentangled in two independent aspects: (i), the subject of the source of the information must be the speaker, and (ii), the manner of the source of the information must be firsthand experience rather than indirect experience. The speaker’s emotive attitude [3] as well as subjectivity [4] can be regarded as the consequence of these requirements on evidentiality. In other words, it is the requirement on the evidentiality that makes the speaker fully committed to the content.

Based on the above analysis, I propose that subjectivity inherently originates from the source of knowledge, as suggested by [20]. Considering the case of zhēn, it appears that the stronger an intensifier’s degree of subjectivity, the more stringent its evidential requirements become. The gradient and heterogeneous nature of Class II intensifiers thus remains a topic for further investigation.

5. Conclusions

This paper distinguishes two kinds of intensifiers in Mandarin, ordinary intensifiers (Class I) and subjective intensifiers (Class II). Building on the works of [2,3,4] on the subjective intensifiers in Mandarin, this article first proposes two additional properties for these intensifiers, i.e., their incompatibility with attitude predicates and their requirement for firsthand experience. As such, the previous theoretical frameworks based on subjectivity [4] and non-veridicality [3] fail to fully account for all the properties of Class II intensifiers. This paper argues that the subjectivity associated with these intensifiers actually originates from their evidentiality, the source of knowledge. In this context, evidentiality necessitates that sentences with Class II intensifiers be speaker-oriented and requires the speaker to have the firsthand experience of the proposition.