2.1Variants of sweet taste

In Korean, a remarkable number of morpho-phonological variants are available for conveying a single taste ‘sweet’, for instance, with extremely subtle differences in terms of the gradience of: (i) a given taste (literal meaning); (ii) a given situation (extended, figurative meaning); and (iii) even a speaker’s emotional attitude associated with either the taste or the situation (expressive meaning). The following list illustrates 13 frequently used variants of ‘sweet’, which is not an exhaustive list:22.Note that only variants with emotional attitudes (i.e. ETTs) are bold-faced in the list, in which ‘pos.att’ denotes ‘positive attitude; ‘neg.att’ denotes ‘negative attitude’; neutral attitude is not marked. The definition for each variant here is a literal translation of its dictionary definition (Dictionary of the National Institute of Korean Language 2008), while its emotional attitude is posited based on the author’s intuition (who is a native speaker of Korean) corroborated by empirical distributional facts such as compatibility with other expressives.

First, in the literal dimension of meaning, most variants of ‘sweet’ convey fine-grained distinction with respect to the intensity of the taste as in (iii) ta-ti-tal-ta ‘extremely sweet’, (xi) tal-khum-hata ‘very savory sweet’. Furthermore, there are some variants of ‘sweet’ for conveying mixture of tastes such as (iv) tal-khom-ssapssal-hata ‘bitter sweet.pos.att’ and (vi) tal-khom-saykhum-hata ‘savory sweet and slightly sour’.

Second, what’s also interesting in these examples is that they have undergone semantic extension, conveying a figurative meaning: the inherent meaning of sweet taste is stretched to the figurative meaning that goes beyond the taste. In both variants (vii) tal-ccakcikun-hata and (viii) tal-chakcikun-hata, for instance, the sweetness could be either about the somewhat savory sweetish taste of a food item or about the situation. The secondary dictionary meaning for these variants is thus defined as ‘content and pleased’. Likewise, the literal meaning in (x) tal-khom-hata ‘savory sweet in the strong sense’ is extended to the figurative meaning such as ‘intriguingly charming or ingratiating’ or ‘comfortable and cozy’. Such semantic extension of ‘sweet taste’ to positive figurative sense seems to be quite universal. The dictionary definition of sweet in English, for instance, includes both the figurative meaning ‘pleasing to the mind or feelings’, as in (2b–c), and the literal one ‘pleasing to the taste’, as in (2a), as its primary meaning (Merriam-Webster Dictionary 1828).

Finally, and most importantly, there are many variants of ‘sweet’ that can express a speaker’s emotional attitude in Korean. For one thing, positive attitudes of a sweet taste can be conveyed, as in (vii) tal-ccakcikun-hata ‘somewhat savory sweetish in the weak sense.pos.att’, and (iv) tal-khom-ssapssal-hata ‘bitter sweet.pos.att’. In the following example, a speaker’s positive attitude is expressed by the vowel choice in tal- and the degree of positive attitude is expressed by the consonant choice - ccakcikun (weak.pos.att) vs. -chakcikun (strong.pos.att):

Considering the inherent positivity and general preference toward the sweet taste, the concurrent inflection of positive attitude is unsurprising, as we observed in English also. The crucial difference, however, is that in Korean negative attitudes also can be reflected in the sweet taste as in (v) tal-khom-ssupssul-hata ‘bitter sweet.neg.att’, (xii) tul-khum-hata ‘somewhat sweet in unsavory/unpleasant way.neg.att’ and (xiii) tul-ccekcikun-hata ‘slightly sweet in unsavory/unpleasant way.neg.att’. The following example from a novel reveals the hero’s negative attitude toward the sweet taste of some gum:

We term this kind of positive and negative emotive variants Emotive Taste Terms (ETTs), which are bold-faced in the list of variants for ‘sweet’ above.

2.2Variants of bitter taste

Just like the taste term for ‘sweet’, ssu-ta ‘bitter’ exhibits three dimensional meaning: (i) literal meaning on the taste; (ii) figurative meaning on the situation; and (iii) expressive meaning with negative, neutral, and positive emotional variants; most frequently used variants are the following, which is also not an exhaustive list:

With ‘bitter’ also, most variants convey a fine-grained distinction with respect to the intensity of the taste as in (i) ssap-ssa-lay-hata ‘seemingly somewhat bitter’, (v) ssu-ti-ssu-ta ‘extremely bitter’, etc. Some variants of ‘bitter’ can carry mixture of tastes as in talkhom-ssapssal-hata ‘bitter sweet.pos.att’.

Regarding the variants of taste terms with the inflection of the speaker’s emotional stance, what’s interesting is the fact that the majority of variants for ‘bitter’ tend to concern negative attitude, while most variants of ‘sweet’ above reflect positive attitude. Negative attitudes are reflected in variants of the bitter taste as in (v) ssu-ti-ssu-ta ‘extremely bitter.neg.att’, (vi) ssup-ssu-ley-hata ‘somewhat bitter.neg.att’, etc. The fundamental connection between bitter taste and negative attitude is also shown by the fact that even the basic taste term ssu-ta ‘bitter’ carries a negative sense in its secondary dictionary meanings:

Furthermore, as observed in the variants of ‘sweet’ above, these variants of ‘bitter’ seem to have undergone semantic extension, but, in the case of ‘bitter’, towards the opposite direction to the negative axis: the typical aversion to bitter taste is extended to a negative meaning toward the situation. This seems to be a general tendency across languages: In English, the primary definition of bitter concerns both ‘bitter acrid taste’, as in (7a), and ‘distasteful or distressing to the mind’, as in (7b) (Merriam-Webster Dictionary 1828).

Notice that even the basic term ssu-ta ‘bitter’ above conveys two types of negative figurative meaning, ‘displeased or distressed’ and ‘wretched due to illness’, as potential secondary meanings. Likewise, the secondary dictionary definitions for other variants of ‘bitter’ include ‘extremely distressed’ for (v) ssu-ti-ssu-ta, and ‘somewhat displeased or distressed’ for (vi) ssup-ssu-ley-hata.

In these variants, the negativity may concern either a literal bitter taste of a food item or a figurative bitterness of the situation. We assume, however, that the extended meaning ‘displeased’ for these variants exists in descriptive dimension, while the negative emotional attitude is supplemented in expressive dimension.

The semantic extension of the positive variants, on the other hand, does not seem as established as that of negative counterparts, since the positive meaning is not explicitly defined as the secondary meaning in the dictionary. Instead, these positive variants deliver a positive sense about the situation, as in (10), as well as the taste itself, as in (9), at the expressive level:33.The positive emotional attitude of talkhom-ssapssalum-han in (10) is further supported by the oddity of the following example with a negative variant; it is due to the conflict with the inherent positive meaning of ‘romance’: (i) # tulkhum-ssupssulum-hansomewhat.bitter.sweet.neg.att-Adn lomaynsuromance

Given the observation so far, it is tempting to conclude that the frequent association of negative attitude with bitter taste is just a natural consequence of semantic extension from the general dispreference or negative emotion associated with bitterness per se, rather than a grammatical device for systematically marking the speaker’s emotional attitude as we argue. An important piece of evidence to support our analysis of the emotional attitude as a Conventional Implicature (CI) marker, however, comes from the fact that positive attitudes also can be conveyed within the realm of bitter taste or situation as in (i) ssap-ssa-lay-hata ‘seemingly somewhat bitter.pos.att’, and (iii) ssap-ssal-hata ‘somewhat bitter.pos.att’. Recall also the positive attitude in the mixture of taste as in (iv) talkhom-ssapssal-hata ‘bitter sweet.pos.att’.

2.3Variants of salty taste

The taste term cca-ta ‘salty’ exhibits the following variants, which is once again not an exhaustive list:

Similar to the taste adjectives for ‘sweet’ and ‘bitter’, most variants for ‘salty’ convey a fine-grained distinction with respect to the intensity of the taste as in (ii) cca-ti-cca-ta ‘extremely salty’ and (iii) ccap-colum-hata ‘somewhat salty’. One interesting typical connotation with the salty taste in Korean concerns being ‘stingy’, as shown from the fact that the secondary dictionary meaning of the basic term cca-ta ‘salty’ is ‘stingy (in vulgar language)’, and that of an intensive variant means ‘extremely stingy’.

Likewise, in English, salty exhibits semantic extension toward a negative direction: besides the basic taste of containing salt as in (12a), it could also mean ‘crude’ as in (12b) (Merriam-Webster Dictionary 1828), or ‘upset’ or ‘depressing’ as in (12c) in a slang term.

Given the potentially negative connotation (of being stingy) in the basic term of salty taste in Korean, an association of negative attitude is anticipated: in the variant (vii) ccip-ccil-hata, the speaker’s negative emotional attitude may concern either the literal salty taste, meaning ‘somewhat unsavory salty.neg.att’, as in (13a), or the extended meaning concerning the situation ‘unsatisfactory work or behavior.neg.att’, as in (13b):

What is surprising about ‘salty’ in Korean, however, is that despite the inherently negative connotation of being stingy with saltiness, positive variants also exist. In (vi) ccap-ccal-hata, the positive attitude may concern either the basic taste meaning ‘somewhat savory salty.pos.att’, or extended meanings such as ‘orderly and staunch work or behavior.pos.att’ or ‘becoming substantial and rich.pos.att’. That is, the positive emotion operates in the expressive dimension, and both the literal meaning of a given taste and the figurative meaning of a given situation operate in the descriptive dimension.

2.4Derivation of taste terms

An important question arises at this point: how can the Korean language allow the flexibility of expressing such a variety of aspects with a single basic taste term? The sophisticated system for conveying extremely nuanced emotional attitudes with taste terms is built upon two main kinds of derivation. First, Korean makes extensive use of phonological alternations of consonants and vowels. Second, Korean is an agglutinative language, allowing productive and cumulative morphological permutations such as prefixation, suffixation, or reduplication. Thus the morpho-phonological system of generating extremely gradient expressions allows the speaker to gracefully deliver the delicate meaning differences regarding their emotional attitude as well as the subtle differences in the quality/quantity of taste.

It is well known that phoneme variation gives rise to intensification effect in Korean: First, alternations of vowels observe the following rules of vowel harmony: (i) yan (bright) vowels such as /a/, /ay/, /o/ have a strong tendency to avoid co-occurrence with yin (dark) vowels such as /e/, /ey/, /u/ within the word boundary; and (ii) the vowel type in the stem (i.e. yin vs. yang) regulates the vowel type in its suffix. In taste terms, the series of yan vowels in the stem of a base taste term tends to associate with a diminutive or moderate connotation, conveying the quality of a given taste being savory or pleasant, while the alternation to yin vowels triggers the opposite attitudes of being unsavory or unpleasant. In a subtype of the tal-ta-series ‘sweet’, for instance, vowel alternations between (i) the yang vowel /a/ and the yin vowel /ɯ/ in the stem and (ii) /a/ and /e/ in the suffix, i.e. between tal-chakcikun-hata and tul-ccekcikun-hata, correlate with positive and negative emotional attitude toward the sweet taste, respectively.

Besides the vowel alternations, consonant alternations bring about additional connotation shift: in the following variants, the alternation /k/ vs. /k^h/ in the suffixes - kom vs. - khom ‘savory’ and the alternation /tɕ/ vs. /t͈ɕ/ in the suffixes - ccakcikun vs. - chakcikun ‘somewhat (un)savory’ indicate non-emphatic weak sense vs. emphatic strong sense or feel of the speaker.

In addition to the phoneme variation, morphological variation conveys nuanced meaning differences in Korean. In taste terms, several suffixes are available to trigger various connotation shifts. The series of -kom/khom/khum concerns the quality, marking the savoriness, and the series of -lay/ley/lum concerns the quantity, marking the weaker intensity, of taste. Another possible variation is reduplication along with the infixation of -ti which seems to give rise to an emphatic effect like ‘very’, as in ta-ti-tal-ta. Reduplication without -ti, tal-tal-hata, on the other hand, typically co-occurs with lum/lay/ley/l-series to mark the weaker intensity of the taste.

3.1Emotive taste terms as expressives

It is generally difficult to articulate subtle emotional attitude of a speaker with only descriptive items; speakers thus employ expressive items to convey their subjective perspective. Representative examples of expressives include epithets such as damn and bastard in English which carry expressive information concerning a speaker’s emotional attitude of the utterance. Potts (2005Potts, Christopher 2005 The Logic of Conventional Implicature. Oxford University Press, Oxford. et seq.) suggests that the expressive content can be understood as conventional implicature (CI), stating that they have “a dramatic impact on how current and future utterances are perceived” (Potts 2007aPotts, Christopher 2007a “The Expressive Dimension.” Theoretical Linguistics 33: 165–197. , 8).

Besides the epithets with negative attitude, a variety of items have been identified as expressives across languages. For instance, adverbials such as yoku and yokumo, expressing speaker’s surprise, in Japanese are analyzed as expressive elements (McCready 2004McCready, Eric 2004 “Two Japanese Adverbials and Expressive Content.” Semantics and Linguistic Theory 14: 163–178. ). Likewise, intensifiers such as sau, total, and voll in German, roughly translated as ‘totally’, are identified as expressive intensifiers (Gutzmann & Turgay 2012Gutzmann, Daniel and Katharina Turgay 2012 “Expressive Intensifiers in German: Syntaxsemantics Mismatches.” In Empirical Issues in Syntax and Semantics 9. ed. by C. Piñón, 149–166.).

Another type of expressives is found in the realm of honorification system in the sense that, as Potts (2007b)Potts, Christopher and Shigeto Kawahara 2004 “Japanese Honorifics as Emotive Definite Descriptions.” Proceedings of Semantics and Linguistic Theory 14: 235–254. notes, it reveals the speaker’s attitude. Honorifics in Japanese (Potts 2005Potts, Christopher 2005 The Logic of Conventional Implicature. Oxford University Press, Oxford.; Potts and Kawahara 2004Potts, Christopher and Shigeto Kawahara 2004 “Japanese Honorifics as Emotive Definite Descriptions.” Proceedings of Semantics and Linguistic Theory 14: 235–254. ) and Korean (Kim and Sells 2007Kim, Jong-Bok and Peter Sells 2007 “Korean Honorification: A Kind of Expressive Meaning.” Journal of East Asian Linguistics Journal 16: 303–336. ), for example, have been classified as expressives. In a similar vein, informal and formal pronouns in German have been treated as expressives also, since they can inform the addressee of the expressive setting such as the relationship between speaker and hearer (Potts 2007bPotts, Christopher and Shigeto Kawahara 2004 “Japanese Honorifics as Emotive Definite Descriptions.” Proceedings of Semantics and Linguistic Theory 14: 235–254. ). (We will return to the discussion of honorification as expressives in Section 5.5.) Furthermore, there are still other types of expressives such as ‘even’ items in Greek, and metalinguistic ‘more’ and ‘than’ particles in Greek and Korean (Giannakiodu and Yoon 2011Giannakidou, Anastasia and Suwon Yoon 2011 “The Subjective Mode of Comparison: Metalinguistic Comparatives in Greek and Korean.” Natural Language and Linguistic Theory 29 (3): 621–655. ).

For these emotionally charged items, Potts (2007aPotts, Christopher 2007a “The Expressive Dimension.” Theoretical Linguistics 33: 165–197. , 1) notes that they have “an immediate and powerful impact on the context” and “a speaker’s expressives indicate that she is in a heightened emotional state. They can tell us if she is angry or elated, frustrated or at ease, powerful or subordinated” (Potts 2007aPotts, Christopher 2007a “The Expressive Dimension.” Theoretical Linguistics 33: 165–197. , 8). Similarly, ETTs in Korean express a speaker’s heightened emotional stance which is the canonical property of expressives. When a speaker employs one of the emotional variants of the basic taste term, the effect of utterance becomes more emphatic. The lexical selection among different variants of taste terms can thus be understood as a speaker’s tactful strategy to establish their emotive stance.

Before jumping to the conclusion, let us examine ETTs with the typical properties of expressives (Potts 2005Potts, Christopher 2005 The Logic of Conventional Implicature. Oxford University Press, Oxford., 2007aPotts, Christopher 2007a “The Expressive Dimension.” Theoretical Linguistics 33: 165–197. ) to see whether our analysis of ETTs as expressives is on the right track. An ETT, as a stance marker, is expected to exhibit the hallmarks of expressive. First, ETTs exhibit independence. Independence means that the expressive content of pejoratives like bastard contributes to the expressive dimension of meaning that is detached from the semantic at-issue dimension. In Potts’ example ‘That bastard Frederic is famous’, what is asserted is ‘Frederic is famous’ (descriptive meaning), but what is additionally expressed is ‘Frederic is a bastard in my opinion’ (expressive meaning). Likewise, the at-issue meaning of the ETT ccipccilha is ‘somewhat unsavory salty’, but it simultaneously expresses the speaker’s negative attitude toward the literal dimension of salty taste:

A positive attitude, on the other hand, can be reflected in ETT toward the figurative at-issue dimension of sweet, i.e. satisfactory, situation, as shown in the following example:

Second, ETTs show nondisplaceability. Since expressives always comment on the utterance situation itself, they cannot report on past events, attitudes, or emotions, unless they are in quotations (Potts 2007aPotts, Christopher 2007a “The Expressive Dimension.” Theoretical Linguistics 33: 165–197. , 5). ETTs in Korean also exhibit such immediate impact on the context:

Here the lexical choice of the negative emotive variant ccipccilha ‘somewhat unsavory salty.neg.att’, over neutral or positive variants, indicates that the speaker expresses their currently effective negative emotion even though the propositional content concerns an event in the past. This is shown by the infelicitous continuation ‘#Now that I think about it, it was refreshing!’ with a conflicting attitude. This means that the effect of negative emotional attitude must be tied to the time of utterance.

Within the system of CI logic, the multidimensional meaning can be represented as the following: (i) the literal descriptive meaning of the sentence with the type e (at-issue type) is ‘salty’ or (ii) the figurative meaning of the sentence with the type e is ‘unsatisfactory’; and (iii) the conventionalized implicature with the type ε (CI type) is ‘I feel negatively toward the x (as it was unpleasant or unsavory).’:

Finally, as CIs, ETTs in Korean exhibit ineffability. Each variant of ‘salty’, for instance, conveys its own connotational nuance in terms of the speaker’s positive or negative emotional attitude, in addition to the taste or situation. Employing an ETT thus means that the speaker has made a deliberate choice of a particular variant among abundant morpho-phonological alternatives for denoting the same taste. This involves a complicated (albeit effortless) process of achieving a complex combination of altered consonants, vowels, and/or suffixes. We assume that this is a tactful strategy of conveying an extremely subtle yet nuanced meaning of the taste in the context, hence a paraphrase with non-expressive elements would not maintain its unique flavor of a particular variant like ccipccilhan ‘somewhat unsavory salty.neg.att’.

Given the properties of taste terms so far, we conclude that ETTs in Korean is a subtype of expressive elements, akin to the aforementioned expressive elements.

3.2The effect of emotive taste terms in context

Within Potts’ (2007aPotts, Christopher 2007a “The Expressive Dimension.” Theoretical Linguistics 33: 165–197. , (37)) system, the result of uttering an expressive is captured as an operation on the context. The context contains various parameters including an expressive setting c_s , which consists of a set of Expressive Indices (EIs). EIs are the main objects manipulated by expressive denotations, defined as “a triple <a I b>, I ∈ [−1, 1].” EIs are the essential basis for expressive domains in expressive items (e.g. damn). These indices are defined via numerical intervals I ⊆ [−1, 1], and encode the degree and orientation of the expressive. In the triple <a I b>, an individual a is at expressive level I for an individual b. The innovation of expressive intervals is its flexibility to encode the gradience in emotional stance from neutral to very positive or negative. Emotive relations are defined by manipulating I to proper subintervals of [−1, 1], in which more positive numbers represent more positive expressive relationship, and more negative numbers mean more negative expressive relationship. Since EIs are merely entities, the nuanced connotation of expressives is difficult to be perfectly translated by a paraphrase, i.e. ineffability. Instead, EIs allow the propositional implications: in <[[tom]] [−.5, 0] [[jerry]]>, we infer a proposition along the line of ‘Tom feels negatively toward Jerry.’ The numerical indices allow us to represent the emotive relation between the judge and the referent. The EI of ETTs thus needs to indicate the intensity and polarity of emotional attitude. Building on Giannakidou and Yoon (2011Giannakidou, Anastasia and Suwon Yoon 2011 “The Subjective Mode of Comparison: Metalinguistic Comparatives in Greek and Korean.” Natural Language and Linguistic Theory 29 (3): 621–655. ; Yoon 2011Yoon, Suwon 2011 ‘Not’ in the Mood: the Syntax, Semantics, and Pragmatics of Evaluative Negation. Ph.D. Thesis. University of Chicago., 2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. ), we propose the following expressive indices of ETTs:

First, the provision in (18a) shows that ETTs have variants with differing strength and polarity of emotion (weak vs. strong, positive vs. negative taste terms). Second, (18b) states that the expressive interval I may concern either buletic or epistemic state of a given taste: for instance, I may concern a buletic state if an individual expresses desirability toward a given stimulus – an ETT variant may convey negative feelings about a given taste (or situation) such as being unsavory or unpleasant, or positive emotions such as being savory or pleasant; or I may concern an epistemic state such as degree of certainty about the speaker’s perception of the taste (e.g. ssup-ssu-lum-hata ‘seemingly somewhat bitter.neg.att; somewhat displeased or distressed.neg.att’). Finally, the provision in (18ii-d) states that there are multiple subcategories of ETTs with different degrees of expressive indices. For instance, weak negative ETTs like tul-ccekcikun-hata ‘slightly sweet in unsavory/unpleasant way.neg.att’ with [−1,0], weak positive ETTs like tal-ccakcikun-hata ‘somewhat savory sweetish in the weak sense.pos.att; content and pleased.pos.att’ with [0,1], or strong positive ETTs like ta-ti-tal-ta ‘extremely sweet; extremely considerate and affectionate.pos.att’ with [.5,1], and the neutral taste term tal-ta ‘sweet’ with [−1,1].

Theoretical implications of the current proposal on the CI of ETTs then include the generalization that the function of ETTs may be incorporated as part of the grammar along with typical expressive elements. That is, ETTs are reflexes of grammaticalization of perspective and subjective mode, akin to other phenomena such as predicates of personal taste (Lasersohn 2009), mood choice (Yoon 2011Yoon, Suwon 2011 ‘Not’ in the Mood: the Syntax, Semantics, and Pragmatics of Evaluative Negation. Ph.D. Thesis. University of Chicago., 2013Yoon, Suwon 2013 “Parametric Variation in Subordinate Evaluative Negation: Japanese/Korean vs. Others.” Journal of East Asian Linguistics 22 (2): 133–166. ), or metalinguistic comparatives (Giannakidou and Yoon 2009Giannakidou, Anastasia and Suwon Yoon 2009 “Metalinguistic Functions and the Expressive Dimension in Greek and Korean: Attitude Semantics, Expressive Meaning and NPI Licensing.” In Proceedings of the 13th Sinn und Bedeutung, ed. by L. Geist, K. von Heusinger, H. Kamp, U. Klein, F. Martin, E. Onea, A. Riester, and T. Solstad, 141–156. Stuttgart: Online Publikationsverbund der Universität Stuttgart (OPUS)., 2011Giannakidou, Anastasia and Suwon Yoon 2011 “The Subjective Mode of Comparison: Metalinguistic Comparatives in Greek and Korean.” Natural Language and Linguistic Theory 29 (3): 621–655. ).44.Thus the meaning of ETTs can be understood along the lines of predicates of personal taste like fun in the sense of Lasersohn (2009). One of the hallmarks of predicates of personal taste concerns the problem of faultless disagreement: “A: Kissing is fun. B: No, kissing is not fun.” The following example shows that such faultless disagreement is also possible with Korean ETTs: (i) Context: Kim and Lee are roommates. Kim likes Mrs. Park, an old lady next door, but Lee hates her. Mrs. Park baked an apple pie for them this morning, and they are discussing the taste of the pie: Kim: phai-kapie-Nom cengmalreally tal-khom-hata.savory.sweet.in.the.strong.sense.pos.att-Decl ‘This pie is really savory sweet.’ Lee: aniya,no tul-ccekcikun-hata.slightly.sweet.in.unsavory/unpleasant.way.neg.att-Decl ‘No, it is slightly sweet in unsavory, unpleasant way.’ The crucial difference here is that, unlike predicates of personal taste, the faultless disagreement with ETTs concerns the meaning in another dimension, i.e. on the expressive level. On the semantic at-issue level, both speakers, Kim and Lee, assert the same fact that Ms. Park’s pie has a sweet taste. The disagreement concerns two speakers’ subjective perception of the emotive nature of the sweet taste.

4.1Compatibility between ETTs and expressive nouns

Korean makes extensive use of expressive elements: a speaker can reveal subtle emotional attitudes ranging from negative to positive emotion by means of a variety of sentential categories such as nouns, verbs, functional adverbs, case particles, and complementizers. These emotive elements can be classified into three categories: positive/honorific, neutral, and negative/antihonorific (Giannakidou and Yoon 2009Giannakidou, Anastasia and Suwon Yoon 2009 “Metalinguistic Functions and the Expressive Dimension in Greek and Korean: Attitude Semantics, Expressive Meaning and NPI Licensing.” In Proceedings of the 13th Sinn und Bedeutung, ed. by L. Geist, K. von Heusinger, H. Kamp, U. Klein, F. Martin, E. Onea, A. Riester, and T. Solstad, 141–156. Stuttgart: Online Publikationsverbund der Universität Stuttgart (OPUS)., 2011Giannakidou, Anastasia and Suwon Yoon 2011 “The Subjective Mode of Comparison: Metalinguistic Comparatives in Greek and Korean.” Natural Language and Linguistic Theory 29 (3): 621–655. ).

To begin, we can test the compatibility between ETTs and the following variants for ‘guy’ that convey different degrees of emotional attitude toward the referent (Yoon 2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. ): (i) saykki ‘bastard’ with the strong negative Expressive Index (EI) [−1,−.5]; (ii) nom or casik ‘jerk’ with the weak negative EI [−1,0]; (iii) namca ‘man, guy’ with the neutral EI [−1,1]; (iv) ssi ‘Mr./Ms.’ with the weak positive EI [0,1]; and (v) nim and pwun ‘sir, the honorable’ with the strong positive EI [.5,1].55.One might wonder how the taste term can modify these variants for ‘guy’, but it is to some extent possible via the semantic extension from the meaning of ETTs. For instance, we can describe a person’s character as stingy with ccata ‘salty’. Furthermore, emotional variants for nouns like ‘smell’ and ‘taste’ exhibit systematic compatibility patterns with the variants with corresponding EIs:

As in the Compatibility Condition Model (CCM) in Figure 1 above, the degree of shading in Table 1 reflects the degree of their compatibility: the black cells indicate the regions of high compatibility between expressive lexical category 1 and expressive lexical category 2; the dark gray cells indicate the regions of mid-compatibility; the light gray cells indicate the regions of low compatibility; and the white cells indicate the regions of incompatibility. From Table 1 we can predict that a strong negative ETT like ccip-ccil-hata ‘somewhat unsavory salty.neg.att; unsatisfactory work or behavior.neg.att’ with [−1,−.5] is most compatible with the strong negative terms such as saykki ‘bastard’, akchwuy ‘stink’, or mastaykali ‘taste.neg.att’ which has the identical index of [−1,−.5]. A medium level of compatibility is expected with the weak negative nouns like nom ‘jerk’ [−1,0]. A low level of compatibility is anticipated with neutral nouns such as namca ‘guy’, naymsay ‘smell’, or mas ‘taste’ with [−1,1]. Furthermore, both weak positive nouns like ssi ‘Mr./Ms.’ and strong ones such as pwun ‘sir’, hyangki ‘aroma’, and phwungmi ‘savoriness’ are predicted to be incompatible. On the other hand, a strong positive ETT like ccap-ccal-hata ‘somewhat savory salty.pos.att; orderly and staunch work or behavior.pos.att; becoming substantial and rich.pos.att’ with the EI [.5,1], is incompatible with strong negative terms like saykki ‘bastard’, akchwuy ‘stink’, or mastaykali ‘taste.neg.att’, or even weak negative one nom ‘jerk’. However, it is moderately compatible with weak positive noun like ssi ‘Mr./Ms.’ [0,1] and highly compatible with strong positive terms like pwun/nim ‘sir’, hyangki ‘aroma’, and phwungmi ‘savoriness’ with the matching index of [.5,1]. The compatibility paradigm is borne out in the following data, which reveals oddity only in incompatible cases:

In order to measure the degree of compatibility between two (or more) expressive elements with an Expressive Index, we adopt the simplified Compatibility Condition Index (CCI) that Yoon (2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. , (38)) suggests:

The simplified equation in (20) shows how the strength of EI affects the compatibility constraint: weak negative items, for instance, may contribute about half of the strong negative elements. In the case of ccip-ccil-hata ‘somewhat unsavory salty.neg.att’ with [−1,−.5], the EI length of which is 0.5, and nom ‘jerk’ [−1,0], the EI length of which is 1, the CCI is 50%; we thus anticipate mid-compatibility for the interfacial bonding of the two emotive lexical items.

The CCI allows us to calculate the approximate degree of compatibility between variants of emotive nouns and ETTs, hence predicts the co-occurrence patterns of expressives, as summarized in Table 2, refined from Table 1 above:

The Compatibility Condition Index (CCI) indicates how natural the co-occurrence of multiple expressives is. Thus the 100% of CCI means that the strong negative ETTs like ccip-ccil-hata ‘somewhat unsavory salty.neg.att’ with [−1,−.5] most naturally and frequently occur with the strong negative nouns such as saykki ‘bastard’ or akchwi ‘stink’ which has the perfectly matching index of [−1,−.5]. The 50% of CCI predicts that it occurs less frequently with the weak negative terms like nom or casik ‘jerk’ [−1,0]. On the other hand, due to the lack of 0% regions for the neutral terms such as namca ‘guy’, naymsay ‘smell’, mas ‘taste’ [−1,1], any kind of ETTs can occur. Also, the same pattern is predicted with positive ETTs. Our prediction is borne out in the following examples in which only regions with 0% compatibility give rise to oddity:

4.2Compatibility between ETTs and expressive case markers

Korean is known for its sophisticated honorification system, in which a respectful attitude is reflected in honorific case markers like kkeyse ‘Nom.hon’, whereas a disrespectful attitude can be conveyed also by derogatory particles like ttawi ‘lit. ilk (enumerative particle)’ (Yoon 2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. ). Assuming that these (anti-)honorific markers are a subtype of expressives (Kim and Sells 2007Kim, Jong-Bok and Peter Sells 2007 “Korean Honorification: A Kind of Expressive Meaning.” Journal of East Asian Linguistics Journal 16: 303–336. ), we can test the compatibility between these (anti-)honorific case markers and ETTs.

First, the negative emotion reflected in the variant of ccip-ccil-hata ‘somewhat unsavory salty.neg.att’ renders it incompatible with honorific nominative case markers like kkeyse. Furthermore, all variants of ETTs can occur with neutral case markers like ka, and the strong positive variant like ccap-ccal-hata ‘somewhat savory salty.pos.att’ sounds odd with case markers with negative attitude like ttawi.ka. The compatibility condition between ETTs and case markers can be summarized as follows:

As such, the systematic pattern further supports Yoon’s Compatibility Condition Model (CCM). The advantage of CCM is to allow us to predict how an expressive item would interact with another expressive item, while also suggesting appropriate Expressive Indices (EI) of ETTs and case markers. For instance, Kim and Sells’ (2007)Kim, Jong-Bok and Peter Sells 2007 “Korean Honorification: A Kind of Expressive Meaning.” Journal of East Asian Linguistics Journal 16: 303–336. assumption that honorific case markers like kkeyse and kkey have an EI of [.5,1] is supported by their incompatibility with negative ETTs such as tulkhum/ccipccil/ssupssul-hata ‘sweet/salty/bitter.neg.att’ [−1,−.5]. Neutral case markers like ka and eykey with [−1,1] are immune to any emotive state, hence occur with any variant of ETT. Finally, anti-honorific case markers like ttawi-ka and ttawi-eykey with [−1,−.5] are expected to be incompatible with positive ETTs like talkhom/ccapccal/ssapssal-hata ‘sweet/salty/bitter.pos.att’ with [.5,1]. In sum, the Compatibility Condition for the expressive elements between ETTs and case markers is predicted from the numerical indices of emotional attitude of each expressive element.

4.3Compatibility between ETTs and expressive verbs

Along with the emotive nouns and (anti-)honorific case markers, the subject honorific marker -si on verbs or copulas has been analyzed as expressives with positive attitude. We assume the subject honorific marking -si on verbs is an expressive marker of the strong positive attitude with the narrow positive range of [.5,1] (Kim and Sells 2007Kim, Jong-Bok and Peter Sells 2007 “Korean Honorification: A Kind of Expressive Meaning.” Journal of East Asian Linguistics Journal 16: 303–336. ), hence incompatible with negative ETTs like tulkhum/ccipccil/ssupssul-hata ‘sweet/salty/bitter.neg.att’ with a strong negative index [−1,−.5]:66.Honorific attitude and emotional attitude have been treated together in the previous literature. It has been shown that honorific speech tends to be more emotionally constrained while non-honorific speech is more emotionally free in Japanese (Dunn 2010Dunn, Cynthia Dickel 2010 “Information Structure and Discourse Stance in a Monologic “Public Speaking” Register of Japanese.” Journal of Pragmatics 42: 1890–1911. ), and Japanese honorification has been analyzed as expressives along with other emotional kinds such as damn (Potts and Kawahara 2004Potts, Christopher and Shigeto Kawahara 2004 “Japanese Honorifics as Emotive Definite Descriptions.” Proceedings of Semantics and Linguistic Theory 14: 235–254. ). As a reviewer points out, however, it is important to note that they do not always go hand in hand. One attempt to tease them apart is found in Yoon (2015)Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. , which will be discussed in Section 4.4.

The emotional attitude, on the other hand, can be further supplemented by verbal expressives with negative attitude. Negative verbal inflections like V-peli express a negative attitude toward the content of the proposition by putting an extra emphasis on the completion of action or state; thus the speaker’s resentment is expressed on the irrecoverability (Joe and Lee 2002Joe, Jieun and Chungmin Lee 2002 “A ‘Removal’ Type of Negative Predicates.” In Proceedings of the 10th Japanese Korean Linguistics Conference. CSLI Publication.; Choe 2004Choe, Jae-Woong 2004 “Obligatory Honorification and the Honorific Feature.” Studies in Generative Grammar 14, 545–559.; Yoon 2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. ). Assuming that the negative index of the verbal inflection peli is [−1,−.5], the strong negative emotion reflected in verbal morphology explains why it is most natural with negative ETTs like tulkhum/ccipccil/ssupssul-hata ‘sweet/salty/bitter.neg.att’, tolerable with neutral ETTs like tan/ccan/ssun ‘sweet/salty/bitter.neut.att’, but odd with the positive ETTs like talkhom/ccapccal/ssapssal-hata ‘sweet/salty/bitter.pos.att’:

The compatibility pattern so far is summarized as follows:

4.4Compatibility condition, cumulative effects, and multidimensionality

The observations so far follow the Compatibility Condition (Yoon 2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. ; Yae and Yoon 2017Yae, James and Suwon Yoon 2017 “Compatibility Condition for Expressives Revisited: a Big Data-based Trend Analysis.” Language Sciences 64: 69–102. ), which assumes “the polarity (either negative or positive attitude axis) of expressives must match with one another within an utterance, while the strength (e.g., weak or strong negativity) at the expressive level does not have to be precisely identical.” The co-occurrence patterns of multiple expressives reveal how the emotional attitude in ETTs interacts with different emotive elements within the sentence. Given the facts, we conclude first that the expressive component of ETTs is sensitive to that of other elements. Second, the polarity (i.e. negative or positive axis of emotional attitude) must either agree with that of its environment, or at least be tolerable with the environment. Finally, the degree of strength of a given attitude is not required to precisely match with another attitudinal component in the sentence, but there must be sufficient overlap between EIs of two expressive elements (e.g. 25% overlap).

Regarding the question of how precisely expressive elements have an impact on one another, we need to know whether the multiple markings of expressives should be understood as a kind of attitude agreement effect, or a more accumulative effect with significant systematic strengthening. Just like the interaction between ethnic slurs and other expressives discussed in Yoon (2015)Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. , ETTs exhibit similar kind of cumulative effects. In the following example with four instances of expressive items, the collective effects of one anti-honorific element and three negative-expressive elements result an extremely strong negative attitude:

Before closing our discussion, let us briefly tackle one more important issue regarding multidimensionality, i.e. the question on the widely assumed correlation between honorific and positive emotional dimension, and antihonorific and negative emotional dimension. The question of autonomy of different expressive dimensions is a tricky issue, since the two dimensions tend to go hand in hand and their differences are rather subtle to tear apart. As Yoon (2015Yoon, Suwon 2015 “Semantic Constraint and Pragmatic Nonconformity for Expressives: Compatibility Condition on Slurs, Epithets, Anti-honorifics, Intensifiers, and Mitigators.” Special issue on Slurs at Language Sciences 52: 46–69. , 20) notes, however, each dimension seems to function independently. In the following examples with apparently conflicting multidimensional attitudes, the speaker needs to conform to the social obligation of addressing the hearer with the proper title sir but simultaneously intends to convey a negative emotional stance with bastard or how dare you:

Likewise, in the following example in Korean, the speaker is socially obliged to use honorific forms for the CEO of the company, while expresses negative emotion toward the descriptive content with the negative expressive verbal suffix peli.

Furthermore, as a reviewer correctly points out, the combination of negative ETTs and honorific markers, ccipccilhan{Brown-ssi/kyoswu.nim} in (21d,e) above seems possible in a proper context, although these honorifics generally tend to co-occur with positive or neutral variants of ETTs. (This tendency is because a speaker typically drops such an honorific title when expressing a negative emotion toward the referent, unless uttered in a sarcastic tone of voice.)

ETTs thus confirm that there can be separate expressive dimensions for emotion and honorification, just as there are distinct descriptive dimensions for literal and figurative meanings as we argue.

To summarize the discussion so far, the compatibility condition of ETTs and other expressive items importantly reveals the specific requirement on the emotive range of each item. We take this fact to argue that we can add ETTs to the list of expressives for conveying multilayered meanings.