2.1 Group 1: Punjabi

Group 1 consists of allocutive languages which allow only non-agreeing 2P pronouns with allocutivity. Furthermore, no featural-mismatch is allowed between the 2P pronoun and the allocutive marker. Languages like Basque, Tamil, Magahi and Punjabi belong to Group 1. We illustrate with Punjabi.

Punjabi has two 2P pronouns tuu (tai is its oblique form) and tusii (twaa is its oblique form). The form tuu can only be used for a singular non-honorific addressee while tusii can be used either for a plurality of addressees independent of their honorificity, or for a singular honorific addressee. A similar divide is seen in allocutive marking. The form ii/aa is the singular non-honorific form, while je is the plural form used both for a plurality of addressees as well as a singular, honorific addressee.

First, there is an agreement based restriction on the co-occurrence of a 2P pronoun and the allocutive marker in this language: when tusii is in the subject position and bears nominative case, it triggers corresponding agreement on the verbal complex. In such structures, the allocutive marker cannot occur. As shown in (5a), the nominative 2P pronoun tusii must trigger corresponding 2P agreement o. The allocutive marker je cannot occur in such a sentence. Contrast this with the example in (5b), where the 2P pronoun appears in the object position bearing an oblique case. Consequently, it does not trigger agreement — allocutivity is allowed in such a sentence. Note that truth-conditionally, no difference arises due to the use of the 3P person auxiliary e versus the allocutive marker je in (5b).

Secondly, when the two addressee-oriented expressions (i.e., a 2P pronoun and an allocutive marker) co-occur, they are not allowed to mismatch. The 2P plural/honorific form je can only co-occur with the 2P plural/honorific pronoun twaa-nuu, but not with the 2P singular/non-honorific form tai-nuu, as shown in (6a). Similarly, the non-honorific allocutive marker aa/ii can only occur with the 2P singular/non-honorific form tai-nuu, as shown in (6b).

2.2 Group 2: Japanese

In contrast with Group 1, Group 2 allows allocutivity with all 2P arguments, regardless of agreement and (mis)matching features. Languages such as Korean and Japanese belong to Group 2. We illustrate with Japanese.

Japanese has an honorific form of the allocutive marker -mas, which appears on the verb. When the sentence lacks a verb — for example, when it is predicated by an adjective or a noun in a copula construction, des- appears instead of -mas.² Since there is no phi-agreement in Japanese — i.e., a 2P pronoun in any structural position does not control agreement on the predicate — the allocutive marker can always be licensed regardless of whether the 2P pronoun appears in the subject or in the object position in a clause. This is shown in (7).

Moreover, Group 2 allows featural mismatches (in honorificity) between the allocutive marker and 2P pronouns. To show this, we first present the pronominal paradigm in Japanese.³ The language has numerous forms for pronouns of each person specification. Consider the paradigm of pronominal forms in Japanese in Table 1 (see also McCready 2019 for a detailed description of the pronouns). As is evident from the table, while the pronouns vary in the person feature, and in gender for 3P, there are no distinct singular and plural pronominal forms. The language employs associative markers to encode plurality. Second, for each person specification, pronouns vary based on honorific meanings. Table 2 shows an approximated honorific scale for 2P pronouns.

Keeping this in mind, we examine the (mis)matching examples in Japanese. First, consider the sentence in (8a).

As shown in (8a), an honorific 2P pronoun anata can co-occur with the honorific allocutive marker -mas. Not surprisingly, a similar matching example can obtain with sotira, another honorific pronoun, as shown in (8b).

What is interesting is that it is also possible to combine the honorific allocutive marker -mas with pronouns in the mid-honorific range in Table 2 (i.e., kimi, omae, and anta). Consider the examples in (9), where the relevant pronouns are glossed as mh (mid-honorific).

Most interestingly, honorific allocutivity can also occur with kisama and temee, which are on the far non-honorific end of the scale in Table 2, as shown in (10). We gloss kisama and temee as nh (non-honorific).

In this way, the honorific allocutive marker -mas/des- can not only occur with pronouns on the honorific end of the scale but also with pronouns on the mid and low honorific ends. Note that these mismatching examples in (10) are instances of genuine mismatches in honorificity, since they also have literal (and not only sarcastic) readings, as we will elaborate upon in Section 4.

To summarise the cross-linguistic landscape, in languages such as Punjabi, mismatches are completely ungrammatical. An honorific allocutive marker can only occur with an honorific 2P pronoun. By contrast, in languages such as Japanese, the honorific allocutive marker can co-occur with 2P pronouns of distinct honorific levels. This leads us to the following question: what allows varying (non-)honorific combinations of the allocutive marker and pronouns in Japanese but not in Punjabi?

3.1 Punjabi pronouns

We start with Punjabi. Unlike the Japanese pronominal paradigm in Table 1, Punjabi pronouns form a closed class with significantly fewer honorific forms. First, honorific forms are available only for 2P and 3P. There are no 1P honorifics. Secondly, there are only two levels of honorificity in Punjabi (i.e., honorific and non-honorific). This is shown in Table 3, where H stands for honorific and NH for non-honorific.

As is also evident from the table, Punjabi re-uses its phi-forms as honorific forms, as is well-known for many Indo-European languages like French, German, Spanish among others. Thus, in addition to its use as a plural 2P pronoun which is insensitive to the (non-)honorificity of the addressee, the form tusii is also used to refer to a singular honorific addressee. In contrast, the singular form tuu can only refer to a singular non-honorific addressee. Similar facts hold for the 3P plural pronoun in the language o which is used to refer to a group of people as well as to refer to a singular honorific third person. The singular 3P pronoun is only used for a non-honorific third person referent.⁵

Akin to the pronominal paradigm, the agreement paradigm in Punjabi also re-uses the phi-agreement forms. Plural pronominal forms, when used for a plural referent, trigger corresponding plural agreement. This is shown for a 2P plural honorific referent in (12a). Notably, the plural pronoun triggers plural agreement on both the verb and the predicative adjective also when used for a singular honorific addressee. This is demonstrated in (12b).

Contrast with singular agreement, which obtains only with the singular form tuu.

The same pattern is maintained with the 3P plural form, which always triggers 3P plural agreement, even when used to refer to a singular honorific 3P, as shown in (14a). Singular agreement obtains only with the singular 3P pronoun, as in (14b).

In summary, Punjabi only has a two-way distinction in honorificity available only for 2P and 3P. This is easily analyzable by positing positive and negative values of a feature such as [HON]. However, since there is no unique morphology separating the honorific feature from phi-features either in the pronominal or in the agreement make-up, it is worth asking if honorificity even constitutes a distinct feature in Punjabi or if it is purely a matter of interpretation that the plural forms can be interpreted as honorific. We argue that [HON] is a distinct formal feature in the language, where ‘formal’ is defined as follows:

Evidence for a formal encoding of honorificity in Punjabi comes from syncretism in agreement in the feminine paradigm. In Punjabi, the copula only shows person and number inflection. The verbal form inflected for aspect shows gender and number inflection. We show the perfective verbal forms for the verb ‘go’ in Table 4. Crucially, note that the table consists of an extra row to present the honorific, singular agreement forms.

As shown in Table 4, the masculine plural form gaye obtains not only for agreement with a plural subject but also for an honorific singular subject. With a feminine nominal, however, the forms differ across the feminine plural and the honorific singular usage. This is illustrated in the following examples.

As expected, a singular non-honorific feminine subject triggers singular agreement as in (16a), and its plural counterpart, regardless of honorificity, triggers plural agreement, as in (16b). With a singular honorific feminine subject, we expect the feminine plural form gayiyãã. However, the agreement form that obtains is syncretic with the masculine plural form, as shown in (16c).

The same pattern is observed across differing aspectual forms of the verb, as well as with predicative adjective. We argue that an analysis of the above-mentioned syncretism between feminine singular honorific and masculine plural agreement necessitates an honorificity based feature in morpho-syntax. To see this, consider an account where morpho-syntax lacks an honorific feature. In such a scenario, the masculine paradigm can be derived in a straightforward way. Regardless of its (non-)honorific interpretation, a nominal specified as masculine plural in syntax triggers plural agreement. This generates the same agreement form across-the-board, regardless of whether it is interpreted as masculine plural honorific, masculine plural non-honorific or masculine singular honorific. In contrast, a noun specified as masculine singular in syntax triggers singular agreement.

However, the same analysis cannot explain the feminine paradigm. Consider (18), where a noun specified as feminine singular in syntax triggers singular agreement, which is correct. However, for a feminine noun specified as plural in syntax, the account only generates gayiyãã, which is incorrect.

We propose an alternative solution which employs a (non-)honorific feature in syntax. Two revisions are proposed: first, feminine is the marked gender feature with masculine being unmarked. Secondly, in addition to their phi-specifications, the nominal and agreement forms also bear honorific specifications via a [HON] feature. In Punjabi, it can have two possible values: + or –, where the former indicates that the addressee is honorific to the speaker, while the latter shows that he/she is not. Given these two modifications, consider the following rules of Vocabulary Insertion/VI for masculine agreement forms.

For the feminine agreement forms, the following VI rules apply.

For the [f, sg, +HON] bundle, we propose that the marked feminine feature is deleted in the presence of the [sg, +HON] bundle. This is shown via the impoverishment rule in (21). The only lexical item that can be inserted to realise the subsequent feature bundle is gaye.

Thus, unless we assume that Punjabi has a distinct formal [HON] feature visible in morpho-syntax, the syncretism between feminine singular honorific agreement and masculine plural/honorific agreement cannot be explained. This feature can be positively or negatively specified, yielding a two-way honorificity distinction.

3.2 Japanese pronouns

In contrast with Punjabi, pronouns constitute an open class in Japanese — there are numerous 2P forms that vary for honorificity, as we saw earlier in Section 2.2. In this section, we show that honorificity on pronouns in Japanese corresponds to descriptive content, in sharp contrast with the Punjabi system.

Recall from Table 1 in Section 2.1.2 that pronouns in Japanese vary in the person feature, and in the gender feature for 3P. This is not unlike the better-studied Germanic and Romance linguistic systems. However, Japanese pronouns stand out with regard to the range of (non-)honorific forms they have for each person specification. For instance, in the 2nd person, the language has, at least, ten forms, which vary for honorificity. Given this large number of pronouns, we examine if honorificity in Japanese is encoded via a formal feature, making the pronoun a functional item, or if honorificity is descriptive content, making the pronoun a lexical item on a par with nouns like ‘cat’, ‘dog’ etc.

To this end, we first try to identify one (or more) suitable honorificity-related feature(s). Distinct values of the honorific features should differentiate all forms of 2P pronouns. If this is not tenable, it would indicate that Japanese pronouns are lexical items — they consist of the person feature, and honorificity as descriptive content. As is evident, a feature like [HON] employed for honorific systems like Punjabi cannot be extended directly to Japanese. This is because [HON] is limited to two values (i.e., positive/negative), which yields only two distinct pronouns. Therefore, we adopt the [status] feature from Kim-Renaud & Pak (2006). Proposed initially for Korean, the [status] feature encodes a hierarchical relation between the speaker and addressee and can have five different values (less than, less than or equal to, equal to, greater than or equal to, greater than). Ignoring the archaic forms and focusing on the following eight pronouns in Japanese for now (omae, kisama, kimi, anata, anta, sotti, temee, and sotira), we observe that the values of the [status] feature would yield 5 distinct spellout forms, which do not suffice to explain the availability of 8 distinct 2P singular forms. Furthermore, 2P pronouns in Japanese do not vary for the values of the [status] feature. As shown in Table 5, six of the eight pronominal forms end up having the same specification for [status] in that they can be used by a speaker who is in a superior position to the addressee. With sotti, the speaker can also be equal to the addressee in status, and with sotira, (s)he may be inferior.

Thus, [status] alone does not suffice to demarcate the 2P forms from each other. Let us entertain the possibility of an additional feature [politeness], which, regardless of the [status] of the speaker, encodes his/her intent to be polite to the addressee. It has two possible values: [Politeness: +/–], where + encodes the intent of the speaker to show respect and – indicates the lack of this intent. Various combinations of [status] and [politeness] would yield 10 distinct possibilities of spell-out, which should be enough to accommodate 8 pronouns. However, even this does not suffice to demarcate the pronouns, as shown in Table 6, where the top five forms end up with identical features.

We can keep trying this experiment with more honorificity-related features. However, it is worth noting that the meanings that can disambiguate 2P pronouns valued identically for [status] and [politeness] do not seem translatable into a non-descriptive feature at all. To illustrate, consider omae and anta — both forms are sensitive to status and are typically uttered when Sp > Adr, and do not encode the Sp’s intent to be polite. The distinction between these two forms is loosely related to the speaker’s gender: anta seems to be used more often by a female speaker, and omae by a male. This gender based distinction, however, is not borne out in all instances. Moreover, note that this gender distinction corresponds to the gender of the speaker, and not the individual picked by the 2P feature (i.e., the addressee). Hence, positing a gender feature in these two pronouns cannot be correct. Similarly, take kisama, which differs from temee with regard to the degree of offensiveness it brings and the specificity of the situation it can be used in (it is typically used in quarrels). Another meaning which demarcates some pronouns from others is the self-image that the speaker is trying to portray. For instance, by using kimi, the speaker attempts to project a noble/sophisticated image of him/herself, which is distinct from the use of anta, which emphasises the speaker’s humble background. Such meanings are clearly concept-denoting. Based on this discussion, we claim that honorificity in Japanese pronouns is not encoded via a formal feature, but is instead descriptive in nature. Japanese pronouns, subsequently are lexical items. All 2P pronouns bear the same formal feature (i.e., 2P) — this should only yield one 2P pronominal form. However, given the descriptive nature of honorificity (a label used to denote a range of complex meanings) also hosted on the 2P pronouns, the language allows for a large class of 2P pronouns, which are distinguished from each other via their varying honorificity related meanings.

Prima facie, this finding seems incompatible with the syntactic nature of object honorification/OH, which bears all characteristics of Agree. We make a small digression in the following subsection to show that the descriptive nature of honorificity on pronouns can be reconciled with OH.

3.3 Functional versus lexical pronouns: consequences for binding

So far, we have seen that honorificity in Punjabi pronouns is a formal feature on par with phi-features, while it is descriptive content in Japanese pronouns. Given the composition of Punjabi pronouns, which contain only formal features (person, number and honorificity), we propose that they are functional items. More specifically, we follow the analysis of Abney (1987) to propose that pronouns in Punjabi are intransitive determiners. There is no N head since the pronoun lacks descriptive content. We provide a structure for the non-honorific singular 2P pronoun tuu in (30). The DP lacks an NP, with the D head hosting all relevant features.

In contrast with Punjabi, honorificity on 2P pronouns in Japanese is treated as descriptive content. We follow Noguchi’s (1997) N-pronoun analysis of Japanese 3P pronouns to claim that Japanese pronouns are NPs. The descriptive content of 2P pronouns i.e., honorificity, is situated on N. Furthermore, following Rouveret (1991), we also situate the person feature on N. The structure is shown in (31).¹⁰

The N-pronoun analysis for 2P pronouns in Japanese predicts nominal-like behavior of 2P pronouns in Japanese, which is borne out. First, some Japanese 2P pronouns clearly show an etymological relation with common nouns. For example, temee is etymologically a N-N compound, decomposable into te ‘hand’ and mae ‘front,’ and originally means ‘front of (my) hand’ (i.e., ‘a person in front of me/your (my/your hand)’). It is natural to assume that these distinct morphemes in honorific pronouns gradually became unanalyzable to an extent that they are now seen as a single word (Vovin 2003: 95; Frellesvig 2010: 245–246). Secondly, it is well-known that 3P pronouns in Japanese, like nouns, can be modified by a demonstrative, an adjective and a relative clause (Kuroda 1965: 105; Noguchi 1997: 777). We show that just like 3P pronouns, 1P and 2P pronouns can also be modified in the language. The 2P pronoun anata can be modified by a demonstrative and an adjective, as in (32a), and by a demonstrative and a relative-clause like structure, as in (32b).

Thus, based on the formal v/s descriptive nature of honorificity on pronouns, 2P pronouns in Punjabi are DPs but those in Japanese are NPs. If our D versus N-pronoun analysis of Punjabi and Japanese pronouns is on the right track, we expect to see differences in their binding patterns. The functional (D/phi) versus lexical status of 3P pronouns has been tied to their binding ability in the literature (Noguchi 1997; Déchaine & Wiltschko 2002; Koak 2008, among others). For instance, pronouns in English are functional which allows them to be construed as a bound variable, as is shown in (33). Similar observation has been made for Shuswap pronouns, analysed as functional items (Déchaine & Wiltschko 2002), as in (34). In contrast, pronouns which are lexical in nature cannot be construed as bound variables. This is well-known for the Japanese 3P pronoun kare, as in (35) (Saito & Hoji 1983; Hoji 1991; Noguchi 1997 etc.).

Since the 2P pronoun in Japanese is lexical, it should not be possible to construe these pronouns as bound variables. For Punjabi, the prediction is not so clear since being a functional item is a necessary but not a sufficient condition for an item to be construed as a variable. To see what a bound variable reading of 1P/2P pronouns looks like, consider the following illustration from English in (36) (Rullman 2004; Kratzer 2009, among others).

We observe that Japanese 2P pronouns cannot obtain a bound variable reading. As shown in (37), the possessive 2P pronoun can only receive a referential interpretation. To obtain a bound reading, the reflexive zibun must be employed instead. Bound construals of 2P pronouns are also unavailable in relative clauses, (38).

In Punjabi, sentences such as ‘only you did your homework’ with a 2P possessive pronoun do not allow a bound reading; a reflexive is needed instead.

However, there are certain instances, where the bound construal of a 1P/2P pronoun is available. Consider the following example with a 1P plural pronoun, based on Rullman (2004). ¹¹

To summarise, 2P pronouns in Punjabi are functional items that may participate in syntactic binding. This makes the syntactic approach a live possibility to explain the ban on honorific mismatches in the language.¹²

However, the syntactic approach to banning mismatching honorificity (Portner et al. 2019), which requires pronouns to be bound by c, cannot be extended to Japanese. Japanese 2P pronouns are lexical items that cannot be construed as bound variables. Since Japanese pronouns cannot be bound by a syntactic c-like head, the honorific specification of pronouns is syntactically independent of that on the allocutive marker. This, in principle, makes room for honorific mismatches between 2P pronouns and allocutivity in Japanese.

4.1 Previous approaches to Japanese honorificity

In the formal semantics literature on honorificity, researchers have proposed a consistency rule to explain matching among honorific items in a clause (Potts & Kawahara 2004; Potts 2007; McCready 2014; 2019). For illustration, we consider McCready (2014), who proposes that (i) each honorific expression is assigned a particular honorific range (e.g., Hon₁), and (ii) the honorific level of the entire sentence is calculated as an average of all expressions used in the sentence, unless the intervals have empty interactions (ibid., 508), as shown below.¹³

Under this view, “high and low-level items cannot be used together, though combinations of high and mid-level items are possible, as are combinations of mid- and low-level items (ibid., 508).”

McCready (2019: Ch.7), however, proposes a sightly different view from the above analysis. While maintaining the conditional rule in (42) as a general principle for honorific expressions (McCready 2019: 32), she explicitly analyzes Japanese pronouns as not encoding intervals but something different, distinguishing them from those found in other honorific languages (e.g., Thai). She motivates this treatment based on the observation that 1P pronouns of a low politeness register and 2P pronouns of a high politeness register can coexist within the same sentence in Japanese but not in Thai (McCready 2019: 110). If we consider that both pronouns merely denote an interval in Japanese, we wrongly predict that there is no overlap in politeness register and conclude that such sentences are unacceptable. Thus, rather than directly specifying the register, she hypothesizes that pronouns “carry information relevant to register assignment” — more specifically, the speaker expectations and commitments relating to the social behavior of the conversational participants — and let the register specification be done via pragmatic inference.

To see the implementation of her system for Japanese, consider the following example:

For the at-issue meaning, we obtain the proposition that the addressee is the father of the speaker (father(a_c, s_c)). For the expressive dimension, we have three pieces of information; (i) the 2P pronoun anata indicates that the speaker makes public an expectation that they will behave formally or somewhat formally toward the addressee (mid₂ ∨ high₂); (ii) the 1P pronoun ore indicates that the speaker makes public that they do not expect formal behavior from the addressee toward them (low₁); and (iii) the allocutive marker denotes an politeness interval for a high register (e.g., 0.6 to 1.0). Thus, we have the set of meanings as shown below.

In the above example, low₁ and mid₂ ∨ high₂ are semantic objects of the same type, and they differ from the semantic type (Hon = [.6, 1)). Since the meanings low₁ and register (Hon = [.6, 1)) are different semantic objects, there is no conflict in meaning. Despite being of the same type, low₁ and mid₂ ∨ high₂ are also compatible, because the former is the expectation about the addressee’s behavior toward the speaker, and the latter is about the speaker’s behavior toward the addressee. Under such an analysis where mismatching expressive meanings of two semantic objects that are of distinct types (1P pronoun and allocutive marker above) does not cause conflict, one may predict that mismatching instances between 2P pronouns and allocutive markers are also accepted. But McCready (2019) observes that a sentence as in (45a) is infelicitous: it lacks a literal reading, and can only be interpreted sarcastically. To explain this, she proposes that the low₂ and Hon = [.6, 1) are in fact comparable, and have no overlapping range in intervals (McCready 2019: 120).¹⁴

She explains: “the fact of a high register implies that the speaker should behave formally, given ordinary social obligations […]; this is not compatible with the content of [temee], which indicates an absolute lack of commitment to formal behavior by the speaker and in addition insults the addressee […] (ibid. 120–122).”

If this is the case, why are the mismatching examples in (10) (repeated below in (46)) felicitous? This comparison is the subject matter of the next subsection.

4.2 Literal readings in mismatching sentences

In McCready’s work, the mismatch between a 2P pronoun and the allocutive marker is at the heart of a non-literal/sarcastic reading. Although the sarcastic interpretation is not detailed upon in her work, we entertain a possible sarcastic interpretation, naturally derived from her analysis. Consider the following example from English¹⁵:

In the above example, the meaning contributed by ‘beautiful’ (which is arguably not-at-issue, ‘beautiful’ being non-restrictive here) is the opposite of that contributed by the use of ‘ugly’ in the first part of the speaker’s utterance. This makes the whole utterance inconsistent at the level of literal meaning. However, the use of ‘ugly’ clarifies that the speaker is using ‘beautiful’ sarcastically, not being committed to its truth (opting out from the Gricean Quality maxim). The sentence is thereby consistent under a sarcastic reading.

On a par, the use of -mas/des- in the same sentence that contains a non-honorific pronoun such as kisama or temee, as in (45a) leads to inconsistency at the level of literal meaning since one item is treated as an honorific expression and the other as non-honorific. Like the use of ‘ugly’ and ‘boring’ in (47), the use of temee shows that the speaker is not observing the Maxim of Quality for the moment. That is, although the speaker uses -mas/des-, she is not committed to its truth. Instead, -mas/des- is being used sarcastically with its inverted meaning to convey the lack of politeness.

Although plausible at first, this line of reasoning is not correct. It predicts that in the case of a mismatch, the meaning of one of the honorific items (typically, the one positively specified for honorificity, i.e., -mas/des-) is inverted. However, under the sarcastic reading for the sentence in (45a), it is the meaning of ‘genius’ that is inverted. Both honorific items retain their meaning — the 2P pronoun temee retains its non-honorific meaning, and -mas/des- continues to indicate speaker’s politeness. This reading is provided in (48). The same facts obtain with a mismatching structure containing kisama and -mas/des-.

This shows that a mismatching 2P pronoun and allocutive marker are in principle compatible under their sincere meanings in a clause. Instead, it is the presence of ‘genius’ which gives rise to sarcasm, potentially because it denotes the extreme end of a normatively-loaded scale, and is thereby a natural target for sarcasm. Sarcastic inversion applied to ‘genius’ like items contributes a value at the scale’s extreme other end (e.g., Camp 2012).

If we are on the right track and sarcasm is not due to the mismatching pronoun and allocutive marker, we make three predictions: (a) we should optionally obtain a literal reading for (45a) (pace McCready 2019), as in the English sentence You are a genius, where ‘genius’ can be interpreted both literally and sarcastically, (b) a sarcastic reading should also obtain with a matching 2P pronoun and -mas/des- in the presence of an evaluative predicate such as ‘genius’ and (c) replacing genius with a non-evaluative predicate in a mismatching sentence should yield literal readings with more ease. All of these predictions are borne out.

First, let us begin with a context where (45a) can be used literally.¹⁶ Consider a speaker who is generally hostile towards X and addresses him via temee. In a specific situation, the speaker is completely amazed by the achievement of X: in such a situation, the utterance of the sentence in (45a) by the speaker truly means that the addressee is a ‘genius’. She continues to use temee because she cannot change the convention used up until then. While there is some weakening in the negative nuance conveyed by temee, it can still not be interpreted as polite. The meaning of -mas/des- as a politeness marker is maintained as well.

Secondly, a sarcastic reading also obtains with the predicate ‘genius’ in a sentence containing two expressions that match in honorificity, such as anata and -mas/des-, as shown in (50). Imagine a context where the addressee did something foolish. Instead of accusing the addressee directly, the speaker can utter the following sentence with ‘genius’ to convey the opposite meaning.

Finally, we show that replacing the evaluative predicate with a non-evaluative one in a mismatching sentence yields literal readings freely. Consider our examples in (51a), which contain a non-evaluative predicate such as ‘understand’, which is standardly not taken to evoke normative scales. The sentence in (51a) is a mismatching example, with temee and des- in the presence of ‘understand’. In contrast, the matching sentence in (51b) with the same predicate contains anata and des-.

If the mismatch between temee and -mas/des- is truly incompatible at the level of literal meaning, we should only obtain a sarcastic reading for (51a), where the meaning of -mas/des- is inverted. Such meaning inversion should not be obligatory in (51b), which contains two matching honorific items. However, if the sarcastic reading in (45a) is due to other reasons (e.g., meaning of ‘genius’) and not due to the mismatch between temee and -mas/des-, (51a) should allow a literal interpretation. We obtain a literal reading for (51a). It is used felicitously in a quarrel between two people of a high social status (e.g., two officials in an office). Imagine a context where the speaker is an official who is upset with the addressee (who is also an official) because she blames him for the decrease in income despite her considerable efforts in the project. Due to being particularly upset with the addressee, the speaker uses temee to indicate hostility. However, since the speaker wants to appear decent in view of her high status, she also uses des- with its literal meaning. The same interpretation obtains for des- in (51b) where the pronoun is matching (i.e., honorific).

In summary, this section has shown that in structures with mismatching honorific items, sarcastic readings may seem prominent in the presence of evaluative predicates. However, literal readings are available too. In Table 7, we summarise the availability of literal readings with different combinations of 2P pronouns and allocutivity with different predicates.

The obvious question that arises is as follows: what allows the sincere co-occurrence of the honorific allocutive marker with a non-honorific 2P pronoun in the pragmatic component of Japanese?

4.3 Proposal

To explain the grammatical co-occurrence of kisama with -mas/des- within the same sentence, we propose that:

The first bullet in (52a) is what is already assumed in McCready (2019). (52b) constitutes the major revision that we propose. Given (52a) and (52b), meanings of 2P pronouns and allocutivity are not only distinct but also incomparable. This means that nothing should prevent non-honorific 2P pronouns and honorific allocutivity from coexisting within a single sentence. As long as we can find an accommodating context, we can felicitously use mismatching sentences, as in (46).