salient-metaphors-of-anger-in-Indonesian-ms.Rmd

---
title: "ANGER IN INDONESIAN: Awakening a sleeping tiger"
csl: "unified-style-sheet-for-linguistics.csl"
bibliography: "anger-references.bib"
link-citations: yes
output: 
  html_notebook:
    code_folding: hide
    fig_caption: yes
    fig_width: 6
    number_sections: yes
    toc: yes
    toc_float: yes
  bookdown::word_document2:
    df_print: kable
    fig_caption: yes
    fig_width: 6
    number_sections: no
    reference_docx: "template.docx"
author: 'Gede Primahadi Wijaya Rajeg <a itemprop="sameAs" content="https://orcid.org/0000-0002-2047-8621" href="https://orcid.org/0000-0002-2047-8621" target="orcid.widget" rel="noopener noreferrer" style="vertical-align:top;"><img src="https://orcid.org/sites/default/files/images/orcid_16x16.png" style="width:1em;margin-right:.5em;" alt="ORCID iD icon"></a>'
---

<style type="text/css">
  body{
  font-size: 14pt;
}
</style>

```{r setup, include = FALSE, message = FALSE, warning = FALSE, echo = FALSE}
knitr::opts_chunk$set(fig.width = 6, 
                      fig.asp = 0.618,
                      dpi = 300,
                      echo = FALSE,
                      tidy = FALSE)

library(tidyverse)
library(happyr) # using some functions from this package
leipzig_size <- read.table(file = "data/corpus_total_size_per_file.txt", 
                           header = TRUE, 
                           sep = "\t", 
                           quote = "", 
                           comment.char = "")
source("codes/MARAH-utility-R-functions.R")
```

# Introduction {#intro}

This paper investigates the salient metaphoric and metonymic conceptualisations of the concept [anger]{.smallcaps} in standard Indonesian ([Glottocode: indo1316](https://glottolog.org/resource/languoid/id/indo1316)), the official variety of Malay used in the Indonesian archipelago. Indonesian belongs to "the Malayic subgroup of Western Malayo-Polynesian" [@tadmor_malay-indonesian_2009, 791; @adelaar_dialects_2017, 571] of the Austronesian language. Malay-Indonesian is spoken by almost 280 million speakers combined in Indonesia, Malaysia, and Brunei [@tadmor_malay-indonesian_2009, 791; @adelaar_dialects_2017, 571]. It has been the "lingua franca" in the regions and developed the colloquial varieties exhibiting great differences among themselves and with the standard language in many aspects of linguistic structures [@tadmor_malay-indonesian_2009, 793]. 

## Previous studies on emotion concepts in Indonesian {#previous-study-emotion}

Heider [-@heider_landscapes_1991] analysed emotion concepts in Minangkabau spoken in West Sumatera, and Indonesian spoken by the Minangkabau and the Javanese communities. The goals were to map the connection between emotion words and identify prototypical scenarios of the emotions related to the relevant behavioural correlates. Then, Levenson et al.’s [-@levenson_emotion_1992] study, also in Minangkabau, measured the physiological correlates associated with (positive and negative) emotions. A recent collection of papers in Fox [-@fox_expressions_2018] offers anthropological linguistic studies of emotions in the regional Austronesian languages of Indonesia.

Shaver et al. [-@shaver_structure_2001] investigated the hierarchical and family-resemblance structures of emotion lexicons. They found that Indonesian and American English exhibit similarity in conceptualising [emotion]{.smallcaps} at (i) the superordinate level (i.e., positive, and negative emotions) [@shaver_structure_2001, 215-216], and (ii) the basic-level categories. The Indonesian's terms for the basic-level categories refer to the same categories as in American English, namely *cinta* 'love', *senang* 'happiness', *marah* 'anger', *kawatir/takut* 'anxiety/fear', and *sedih* 'sadness' [@shaver_structure_2001, 218].

Linguistic studies in Indonesian/Malay reveal the so-called "psycho-collocations" [@matisoff_hearts_1986], which are lexico-semantic figurative expressions for mental activities and personhood, and which component parts consist of body-part terms. Indonesian/Malay have extensive repertoires of psycho-collocations, especially with the word *hati* 'liver', suggesting the prominence of the liver as the seat of psychological realms in the Malay world [@siahaan_did_2008;  @goddard_contrastive_2008; @oey_psycho-collocations_1990; @sather_work_2018; @fox_towards_2018, 12]. Musgrave [-@musgrave_nonsubject_2001] and Mulyadi [-@mulyadi_verba_2012] investigated the syntactic and semantic properties of Indonesian emotion predicates. Mulyadi [-@mulyadi_verba_2012] particularly is a contrastive study of emotion verbs between Indonesian and the Asahan Malay variety spoken in Tanjungbalai (Asahan), North Sumatera, Indonesia.

Indonesian emotions have also been analysed using the *Conceptual Metaphor Theory* (CMT) [@siahaan_did_2008; @siahaan_why_2015; @rajeg_metafora_2013; @rajeg_metaphorical_2019]. Siahaan [-@siahaan_did_2008] examined the cultural conceptualisations of *hati* 'liver', proposing that (i) liver divination ritual and (ii) ethno-religious belief of *hati* as the locus for the living soul underlie the conceptualisations of *hati* as the seat of emotion and cognition [see also @goddard_contrastive_2008]. Siahaan's [-@siahaan_why_2015] follow-up study discovered that 'emotion' is the predominant figurative extension of Indonesian temperature terms. Next, Rajeg [-@rajeg_metafora_2013] analysed five Indonesian basic-level emotions. He applied *Configural Frequency Analysis* (CFA) [@gries_statistics_2009] to identify emotion-specific metaphors [@kovecses_metaphor_2000, 35] and examined how these metaphors semantically distinguish the emotions. Another study by Rajeg [-@rajeg_metaphorical_2019] investigated the distinctive metaphors for [happiness]{.smallcaps} near-synonyms, combining the *MetaNet* (MN) approach [@oana_computational_2017; @petruck_metanet_2016], *Metaphorical Pattern Analysis* (MPA) [@stefanowitsch_happiness_2004; @stefanowitsch_words_2006], and *Multiple Distinctive Collexeme Analysis* (MDCA) [@hilpert_distinctive_2006]; the study reveals that metaphors strongly distinguishing happiness and joy in English [@stefanowitsch_happiness_2004; @stefanowitsch_words_2006] are also those distinguishing the Indonesian equivalences of happiness (i.e., '*kebahagiaan*') and joy ('*kegembiraan*').

## Previous studies on [anger]{.smallcaps} in Indonesian {#previous-study-anger}
  
Heider [-@heider_landscapes_1991, 57, Table 7] discovered that, in representing anger, figurative expressions (i.e., *palak* 'stifling; angry' and *panas hati* 'lit. hot liver; angry') received higher rating than the literal expression (i.e., *marah*). Heider [-@heider_landscapes_1991, 24-25] also proposed four [anger]{.smallcaps}-like clusters in Minangkabau Indonesian: (i) "anger" clusters (*naik darah* 'lit. rising blood; angry'), (ii) "anger/cruel" clusters (*bengis* 'cruel; harshness'), (iii) "anger/dislike" clusters (*gemas* 'irritated'), and (iv) "anger/trembling" clusters (*gemetar* 'trembling'). The elicited scenarios from the "anger/cruel" clusters revealed that the antecedents of anger "are hurtful acts by others, especially naughty children, and the outcomes are physical violence and verbal abuse" [@heider_landscapes_1991, 80, 116] (e.g., [§\@ref(verbal-behaviour-typebased)](#verbal-behaviour-typebased) and [§\@ref(violent-behaviour-typebased)](#violent-behaviour-typebased)). Heider [-@heider_landscapes_1991, 80] also noted that in the actual, spontaneous behaviour (rather than in the elicited behaviour), Indonesians "mask most anger, and the open expression of anger is strongly disapproved of and negatively sanctioned".

Rajeg [-@rajeg_metafora_2013, 211-214] revealed that eight metaphors are significantly attracted to *amarah/kemarahan* 'anger'. They are [controlling emotion is controlling a moving object]{.smallcaps}, [emotion is pressurised substance]{.smallcaps}, [emotion is fluid in a container]{.smallcaps}, [emotion is heated fluid in a container]{.smallcaps}, [emotion is fire]{.smallcaps}, ([intensity of]{.smallcaps}) [emotion is temperature]{.smallcaps} ([hot/cold]{.smallcaps}), ([intensity of]{.smallcaps}) [emotion is verticality]{.smallcaps} ([high/low]{.smallcaps}), and [emotion is natural forces]{.smallcaps}. Six metaphors are statistically repelled: [emotion is a possessable object]{.smallcaps}, [causing emotion is object transfer]{.smallcaps}, [emotion is an accidental motion]{.smallcaps}, [emotion is a journey]{.smallcaps}, [becoming emotion is finding an object]{.smallcaps}, and [emotion is liquid]{.smallcaps}. The statistical attraction of Indonesian [anger]{.smallcaps} to the [heat]{.smallcaps}- and [substance]{.smallcaps}-related metaphors suggests the universality and centrality of these metaphors for [anger]{.smallcaps} as found in different languages [@kovecses_concept_2000], most notably English [@stefanowitsch_words_2006; @holland_cognitive_1987]. Rajeg's [-@rajeg_metafora_2013] quantitative study complements Yuditha's [-@yuditha_indonesian_2013] introspective proposal on the specific metaphors of anger. Lastly, Rajeg's [-@rajeg_metaphorical_2014] preliminary quantitative investigation demonstrates that distinctive metaphorical constructions across five [anger]{.smallcaps} synonyms prominently highlight the Intensity of anger.

This chapter presents new approaches in the study of Indonesian [anger]{.smallcaps} from the CMT and corpus linguistic perspectives. It integrates the lexical [@kovecses_lexical_2019] ([§\@ref(typebased-analysis)](#typebased-analysis)) and corpus-based approaches [@stefanowitsch_corpus-based_2006; @stefanowitsch_gries_2006] ([§\@ref(tokenbased-analysis)](#tokenbased-analysis)) into the so-called *salience-based approach* [@kovecses_salience_2015] ([§\@ref(method)](#method)). The inclusion of metonymy ([§\@ref(typebased-metonymy-salience-results)](#typebased-metonymy-salience-results) and [\@ref(tokenbased-metonymy-salience-results)](#tokenbased-metonymy-salience-results)) and the contextual factors ([§\@ref(contexts)](#contexts)) fill the gap from previous works, focusing only on metaphor and excluding contexts in the use of metaphors.

# Methodology {#method}

```{r keywords-frequency-analyses, message = FALSE, echo = FALSE}
# load the prepared data
load(file = "data/leipzig-NON-MARAH-wlist.RData")
load(file = "data/leipzig-MARAH-wlist.RData")

# remove the PERSONAL PRONOUN suffixes
MARAH_sum_lemma <- MARAH_sum %>% 
  mutate(lemma = str_replace_all(match, "(nya|ku|mu)$", "")) %>% 
  group_by(lemma) %>% 
  summarise(n = sum(n), .groups = "drop") %>% 
  arrange() %>% 
  filter(lemma %in% c("amarah", "kemarahan", "marah"))

NON_MARAH_sum_lemma <- NON_MARAH_sum %>% 
  mutate(lemma = str_replace_all(match, "(nya|ku|mu)$", "")) %>% 
  group_by(lemma) %>% 
  summarise(n = sum(n), .groups = "drop") %>% 
  arrange() %>% 
  filter(lemma %in% c("geram", "berang", "gusar", "kegusaran", "keberangan", "kegeraman", "kekesalan", "kemurkaan", "keberangan", "kesal", "murka"))

# compare frequency of the terms
ALL_MARAH_sum_lemma <- bind_rows(MARAH_sum_lemma, NON_MARAH_sum_lemma)

## ke- -an forms
ke_an <- ALL_MARAH_sum_lemma %>% 
  filter(str_detect(lemma, "^ke.+an$")) %>% 
  filter(str_detect(lemma, "kesal", negate = TRUE)) %>% 
  arrange(desc(n)) %>% 
  mutate(perc = round(n/sum(n) * 100, 2))
ke_an_count <- ke_an$n
names(ke_an_count) <- ke_an$lemma
ke_an_chisq <- chisq.test(ke_an_count)
ke_an_chisq_residuals <- chisq.test(ke_an_count)$residuals
ke_an_pvals <- if (ke_an_chisq$p.value < 0.05 & ke_an_chisq$p.value > 0.01) " < 0.05" else if (ke_an_chisq$p.value < 0.01 & ke_an_chisq$p.value > 0.001) " < 0.01" else if (ke_an_chisq$p.value < 0.001) " < 0.001"

## root forms
root <- ALL_MARAH_sum_lemma %>% 
  filter(str_detect(lemma, "^ke.+an$", negate = TRUE)) %>% 
  filter(str_detect(lemma, "kesal", negate = TRUE)) %>% 
  mutate(lemma = if_else(str_detect(lemma, "marah"), "(a)marah", lemma)) %>%
  group_by(lemma) %>% 
  summarise(n = sum(n), .groups = "drop") %>% 
  arrange(desc(n)) %>% 
  mutate(perc = round(n/sum(n) * 100, 2))
root_count <- root$n
names(root_count) <- root$lemma
root_chisq <- chisq.test(root_count)
root_chisq_residuals <- chisq.test(root_count)$residuals
root_pvals <- if (root_chisq$p.value < 0.05 & root_chisq$p.value > 0.01) " < 0.05" else if (root_chisq$p.value < 0.01 & root_chisq$p.value > 0.001) " < 0.01" else if (root_chisq$p.value < 0.001) " < 0.001"


```


The Indonesian terms corresponding to the English *anger* are *marah* 'angry; anger'^[The English translations come from Stevens and Schmidgall-Tellings [-@stevens_comprehensive_2004].], *amarah*^[*Amarah* is the informal variant of *marah* as indicated by the official *Kamus Besar Bahasa Indonesia* (KBBI) *The Big Dictionary of Indonesian* and defined under the entry for *marah*: https://kbbi.kemdikbud.go.id/entri/marah (accessed on October 26, 2021).] 'anger', and *kemarahan* 'fury; rage; anger', the noun derivative from the root *marah*. The choice is based on Shaver et al.'s [-@shaver_structure_2001, 217] finding that *marah* emerges as the prototypical label for [anger]{.smallcaps} category in Indonesian, as it is semantically broader and commonly used in everyday Indonesian. The commonality of *(a)marah* is evident from the frequency data in the **Indonesian Leipzig Corpora** (ILC) [@goldhahn_building_2012]. The combined token frequencies of *(a)marah* [`r prettyNum(pull(filter(root, lemma == "(a)marah"), n), big.mark = ",")`] is the highest compared to the other terms identified by Shaver et al. [-@shaver_structure_2001, 218, Table 4], namely *geram* 'furious; angry' [`r prettyNum(pull(filter(root, lemma == "geram"), n), big.mark = ",")`] and *berang* 'furious; fury' [`r prettyNum(pull(filter(root, lemma == "berang"), n), big.mark = ",")`], as well as to the other near-synonyms of *marah*, namely *murka* 'wrath; anger; fury' [`r prettyNum(pull(filter(root, lemma == "murka"), n), big.mark = ",")`] and *gusar* 'angry; offended; annoyed' [`r prettyNum(pull(filter(root, lemma == "gusar"), n), big.mark = ",")`] (*X*^2^~goodness-of-fit~=`r format(round(root_chisq$statistic, 1))`; *df*=`r root_chisq$parameter`; *p*~two-tailed~`r root_pvals`). Moreover, the derived noun *ke**marah**an* [`r prettyNum(pull(filter(ke_an, lemma == "kemarahan"), n), big.mark = ",")`] is the most frequent compared to other noun derivatives based on the other roots, namely *ke**murka**an* [`r prettyNum(pull(filter(ke_an, lemma == "kemurkaan"), n), big.mark = ",")`], *ke**gusar**an* [`r prettyNum(pull(filter(ke_an, lemma == "kegusaran"), n), big.mark = ",")`], *ke**geram**an* [`r prettyNum(pull(filter(ke_an, lemma == "kegeraman"), n), big.mark = ",")`], and *ke**berang**an* [`r prettyNum(pull(filter(ke_an, lemma == "keberangan"), n), big.mark = ",")`] (*X*^2^~goodness-of-fit~=`r format(round(ke_an_chisq$statistic, 1))`; *df*=`r ke_an_chisq$parameter`; *p*~two-tailed~`r ke_an_pvals`).

The dataset for the type-based, lexical approach is culled from (i) the Indonesian WordNet (v1.0) [@bond_wordnet_2014]^[The Indonesian WordNet 1.0 is available at http://compling.hss.ntu.edu.sg/omw/cgi-bin/wn-gridx.cgi?usrname=&gridmode=wnbahasa (last accessed on 22 January 2022.)], (ii) the monolingual *Kamus Bahasa Indonesia* (the Indonesian Dictionary) (KBI) [@kamus_bi_2008], (iii) the official *Tesaurus Tematis Bahasa Indonesia* (the Indonesian Thematic Thesaurus) (http://tesaurus.kemdikbud.go.id/tematis/)^[The thesaurus is compiled and maintained by the Language Development and Cultivation Agency of the Ministry of Education and Culture of the Republic of Indonesia.], and (iv) a bilingual Indonesian-English dictionary [@stevens_comprehensive_2004]. The relevant linguistic expressions were gathered as follows. From the WordNet, KBI, and the thesaurus, the three Indonesian [anger]{.smallcaps} words (*marah*, *amarah*, *kemarahan*) were used as the search terms. This approach will retrieve expressions that contain (one of) the [anger]{.smallcaps} terms in their definition entries. These expressions were then checked for the evoked metaphorical/metonymical conceptualisations by looking at their meanings in the *Kamus Besar Bahasa Indonesia* (KBBI) (the Big Indonesian Dictionary)^[I used KBI instead of KBBI because KBBI only outputs the brief definition of the three [anger]{.smallcaps} words but not the metaphorical/metonymical expressions related to them, which can be found through the linked thematic thesaurus in the entry. In contrast, the available full PDF version of KBI allows me to do a concordance-like search throughout the PDF using the target terms, which may appear in the definition of an entry related to them. This is not possible in KBBI because the search field is only used for the headword, not the words contained in the definition of the headword. That way, we need to know *a priori* all expressions which definitions contain the [anger]{.smallcaps} words.] following the procedure of the *Metaphor Identification Procedure* (MIP) [@pragglejaz_mip_2007; @pragglejaz_mip_2010] (see below). To gather the data from the bilingual dictionary, the English terms "anger" and "angry" were searched for in the PDF version of the dictionary; this allows retrieval of the Indonesian expressions which English definitions contain the word "angry" or "anger".

```{r leipzig-corpus-information}
leipzig_size <- mutate(leipzig_size, 
                       sources = if_else(str_detect(corpus_id, "news"), "news", "web"), 
                       sources = if_else(str_detect(corpus_id, "wikipedia"), "wikipedia", sources), 
                       sources = if_else(str_detect(corpus_id, "mixed"), "mixed", sources))
leipzig_size_sources <- leipzig_size %>% 
  group_by(sources) %>% 
  summarise(total_tokens = sum(total_tokens), .groups = "drop") %>% 
  mutate(perc_sources = round(total_tokens/sum(total_tokens) * 100, 1))
```


The dataset for the token-based, corpus approach is taken from the corpus files in the ILC (total size = `r prettyNum(sum(leipzig_size$total_tokens), big.mark = ",")` word-tokens). It is chosen since, to the best of my knowledge, ILC is the only open access source to the largest collection of Indonesian texts^[The alternative is the Indonesian corpus in *Sketch Engine* (SE), which is also from online materials as in ILC. However, SE is a paid service to which the institution I work in does not have paid subscription.] and allows downloading the raw corpus files. ILC mainly consists of randomly chosen websites (`r pull(filter(leipzig_size_sources, sources == "web"), perc_sources)`% of the total size) and online news (`r pull(filter(leipzig_size_sources, sources == "news"), perc_sources)`%), followed by the Wikipedia dumps (`r pull(filter(leipzig_size_sources, sources == "wikipedia"), perc_sources)`%) and a mixture of other sources (`r pull(filter(leipzig_size_sources, sources == "mixed"), perc_sources)`%). As in MPA [@stefanowitsch_words_2006], 1000 random concordance lines were retrieved for each *marah*, *amarah*, and *kemarahan* before manually discarding the irrelevant hits (i.e., duplicates, the predicative and attributive uses of the root *marah*, and the literal uses). Next, syntactically relevant collocations of the target terms with the potential source-domain lexical units (LUs) were manually determined [@stefanowitsch_happiness_2004, 138; @sullivan_frames_2013, 3, 5], adopting the *MetaNet* (MN) approach that integrates MPA with Construction Grammar and Frame Semantics [@sullivan_frames_2013; @oana_computational_2017; see @rajeg_metaphorical_2019 for a recent application to Indonesian]. The MIP was applied to determine whether the collocation of the target terms evoke metaphorical readings. It is determined whether the collocates' contextual meaning, when co-occurring with the [anger]{.smallcaps} terms, contrasts with their more basic meaning in other contexts, such that the "contextual meaning can be understood in comparison to the basic meaning" [@rajeg_metaphorical_2019, 64; @pragglejaz_mip_2007, 3; @sullivan_frames_2013, 36]. The KBBI was used to determine the basic meaning of the collocates with reference to MIP's features of basic meaning, namely "more concrete (what they evoke is easier to imagine, see, hear, feel, smell, and taste), related to bodily action, more precise (as opposed to vague), historically older, and are not necessarily the most frequent meanings" [@pragglejaz_mip_2007, 3]. An additional diagnostic to determine the basic meaning is a question proposed by Soriano [-@soriano_conceptualization_2005, 91]: "what exactly each expression 'was literally about'?". To illustrate, consider these two examples for two different ways to convey the existence of *kemarahan* 'anger'.

(@kemarahan_terjadi) *Kemarahan Presiden Jokowi __terjadi__ saat meninjau Pelabuhan Tanjung Priok (...)* (ind-id_web_2015_3M: 1310067)^[At the end of the numbered example, the source of the example is given in the format "(corpus file name: sentence id)" as in (ind-id_web_2015_3M: 1310067).]

(@kemarahan_datang) *(...) kemarahan itu bisa saja __datang__ dalam waktu yang ditentukan (...)* (ind-id_web_2013_1M: 143884).

Example (@kemarahan_terjadi) is considered literal given the verbal collocate *terjadi* 'happen' represents an abstract event as its basic meaning and, in this collocation with *kemarahan* (another abstract domain), *terjadi* is still understood in the abstract domain of [anger]{.smallcaps} [@croft_domain_2003, 192; @sullivan_frames_2013, 9; @rajeg_metaphorical_2019, 99]. In contrast, the collocation of *kemarahan* as the subject of the verb *datang* 'come' (@kemarahan_datang), having a basic meaning in the domain of physical translational motion, induces the metaphorical "domain mapping" [@croft_domain_2003, 192] and interpretation of the verb in the [anger]{.smallcaps} domain [cf. @sullivan_integrating_2016, 147; @dancygier_figurative_2014, 135]. Such metaphoric interpretation of *datang* emerges due to a mismatch between (i) the semantic type-constraint assigned to the semantic role of its subject (prototypically an animate entity), and (ii) the filler of that role, namely an abstract entity [anger]{.smallcaps} that is literally unable to perform a translational motion [cf. @stickles_formalizing_2016, 194; @sullivan_integrating_2016, 148; @brooke-rose_grammar_1958, 1]. The principles of metaphor-domain evocation in a grammatical construction have been described extensively by Sullivan [-@sullivan_frames_2013; -@sullivan_integrating_2016]. The unclear cases as to whether the tokens are metaphorical or literal were marked as "?" in the database, which is available from https://osf.io/c3p4y/?view_only=50e70ee803fd41ec886547cf33848d49.

The identified metaphorical expressions were then grouped thematically under their metaphorical source domains, adopting the MN approach. MN formalises the central notions in CMT by (i) representing the metaphor-input domains as semantic frames, and (ii) viewing (conceptual) metaphor as unidirectional mappings from the source-domain frames to the target-domain frames (including the mappings between the frame roles) that are mediated via the grammatical constructions [@sullivan_frames_2013; @stickles_formalizing_2016; @croft_connecting_2009]. The closest English equivalence of the Indonesian source-domain LUs guides the choice for the source-domain frames in the English MN Wiki repository^[The *MetaNet* main page: https://metaphor.icsi.berkeley.edu/pub/en/index.php/MetaNet_Metaphor_Wiki. The *MetaNet* frame repository: https://metaphor.icsi.berkeley.edu/pub/en/index.php/Category:Frame] [see @lopez_distinguishing_2011 for a similar approach in Spanish]. The classification also considers the relevant categories from previous studies. The metaphorical mappings within a metaphor are postulated by making use of the available frame roles in each MN frame entry or proposed anew based on the semantics of the source-domain LUs. The metonymic source domains are determined via expressions referring to the physiological effects of emotion.

The *metaphorical salience* measure for a given metaphor in the token-based, corpus approach considers the percentages of:

a. the token frequency of a metaphor
a. the number of metaphorical linguistic expressions of a metaphor
a. the number of metaphorical mappings in a metaphor

The token frequency of a given metaphor is all occurrences of metaphorical expressions evoking the metaphor; the percentage of the total token for a metaphor is calculated from the summed tokens of all metaphors in the database. The same procedure is applied for the percentages of the number of metaphorical linguistic expressions and metaphorical mappings [@kovecses_salience_2015, 344-346].

The metaphorical salience measure in the database for the lexical, type-based approach only considers the number of types of metaphorical linguistic expressions and the number of metaphorical mappings for a given metaphor. It is because the focus of such an approach is to gather the types (not tokens) of metaphorical expressions evoking a given metaphor and the respective metaphorical mappings.

# Type-based salience analyses {#typebased-analysis}

```{r type-based-analysis-computation, message=FALSE, include=FALSE, warning=FALSE, error=FALSE}
source("codes/MARAH-type-based-analysis-code.R")
source("codes/MARAH-token-based-analysis-code.R")

# Another aspect of the token-based, data collection is extracting multiple metaphorical expressions in a single concordance line given that the target-domain term can have more than one relevant syntactic collocation [@sullivan_frames_2013, 135-138; @musolff_variation_2014; @rajeg_metaphorical_2019, 63]. Consider example (@multiple_mpattern):
# 
# (@multiple_mpattern) *(...) kerusuhan itu merupakan __hasil dari__ amarah yang __menumpuk__ (...)* (ind_news_2010_300K: 83871)
# 
# From this one citation of *amarah*, two metaphorical expressions or patterns are relevant syntactically. First, the noun phrase (NP) *__hasil__ dari amarah* 'the *yield* of anger'. In this NP, the source-domain item *hasil* 'yield' is the head while the target-domain item appears as the complement of the modifying preposition phrase (PP) *dari* 'from'. Second, the relative clause construction *amarah yang __menumpuk__* 'anger that _piles up_/_amasses_' whereby the head noun is the target-domain term while the source-domain item is the verbal head *menumpuk* of the relative clause.
```

## Metaphor {#typebased-metaphor-salience-results}

There are `r numbers2words(nrow(metaphor_typebased_salience))` conceptual metaphors from the type-based, lexical dataset. [Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based) presents the source domains of the metaphors ranked order by their degree of salience values (see the `Aggregate` column).

```{r metaphor-table-type-based}
metaphor_typebased_salience_print %>% 
  mutate(`Metaphorical source domains` = str_replace(`Metaphorical source domains`, fixed("rough, solid entity"), "substance"), # replace solid entity with substance as per the suggestion from 1st review
         `Metaphorical source domains` = str_replace(`Metaphorical source domains`, fixed("degree of control is object dimension"), "degree of controlling anger is size of patience as a container") # replace object dimension with size as per the suggestion from 1st review
         ) %>%
  knitr::kable(caption = "Source domains of [anger]{.smallcaps} (Type-based, lexical approach)", row.names = FALSE)
```

[Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based) shows `r happyr::numbers2words(length(grep(" is ", metaphor_typebased_salience_print[[1]], perl = TRUE)))` metaphors in full, both with the source and target domains, because they are the more general metaphors.

### [Marah is a fierce, captive animal]{.smallcaps} {#animal-typebased}

```{r mapping-animal, message = FALSE, include = FALSE}
mapping_animal <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "fierce")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_restrained|experiencer_restraining)"), "key", "no"))
mapping_animal_stats <- mapping_animal %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_animal_discuss <- get_mappings(metaphor_typebased_mapping, 'fierce') %>% left_join(mapping_animal %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_animal_stats)
# `r paste(paste(get_mappings(metaphor_typebased_mapping, 'fierce')[[2]], "<br><br>", sep = ""), collapse = "")` list of mapping
```


The most salient conceptualisation of anger in the lexical database is [fierce, captive animal]{.smallcaps}, manifesting `r get_salience_stats(metaphor_typebased_salience, "n_type", "fierce, captive")` metaphorical expressions and `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_mapping", "fierce, captive"))` mappings as shown below^[Throughout this paper, the linguistic expressions of each mapping are given below the respective mapping. The mappings with their linguistic expressions are presented in the decreasing order of its number of linguistic type (i.e., the mapping's type frequency), below the foundational mapping(s) of the metaphor; this allows highlighting the central mapping and main meaning focus of the metaphor [@kovecses_metaphor_2010, 140]]. The first two mappings are the foundational mappings underlying each of the metaphorical expressions.

- restrained/captive, fierce animal &rarr; anger
- entity who restrains/makes captive the animal &rarr; experiencer
- `r mapping_animal_discuss %>% slice_max(order_by = n_lu) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("^(.+)( \\<\\- )(.+)$", "\\1 &rarr; \\3")` (type=`r mapping_animal_discuss %>% slice_max(order_by = n_lu) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'aggressive')) %>% pull(LU), "* '", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'aggressive')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`. 

- `r mapping_animal_discuss %>% filter(str_detect(MAPPING, 'deliberately')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("^(.+)( \\<\\- )(.+)$", "\\1 &rarr; \\3")` (type=`r mapping_animal_discuss %>% filter(str_detect(MAPPING, 'deliberately')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'deliberately')) %>% pull(LU), "* '", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'deliberately')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_animal_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+)\\s\\<\\-\\s(.+)", "\\2 &rarr; \\1")` (type=`r mapping_animal_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(LU), "* '", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_animal_discuss %>% filter(str_detect(MAPPING, 'regulating')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+)\\s\\<\\-\\s(.+)", "\\2 &rarr; \\1")` (type=`r mapping_animal_discuss %>% filter(str_detect(MAPPING, 'regulating')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'regulating')) %>% pull(LU), "* '", mapping_animal_discuss %>% filter(str_detect(MAPPING, 'regulating')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

The main meaning focus of this metaphor is on aggressiveness of the angered person. It is reflected in the mapping [aggressive angry behaviour is aggressive animal behaviour]{.smallcaps} because this mapping is the most productive indicated by the highest number of linguistic types (`r mapping_animal_discuss %>% slice_max(order_by = n_lu) %>% pull(n_lu) %>% unique()`). The [fierce, captive animal]{.smallcaps} is also salient in the corpus approach, but differ in its central mapping and main meaning focus ([§\@ref(animal-tokenbased)](#animal-tokenbased)).


### [Marah is fire]{.smallcaps} {#fire-typebased}

```{r mapping-fire, message = FALSE, include = FALSE}
mapping_fire <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "fire$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_fire|angry-person_burning-object)"), "key", "no"))
mapping_fire_stats <- mapping_fire %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_fire_discuss <- get_mappings(metaphor_typebased_mapping, 'fire$') %>% left_join(mapping_fire %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_fire_stats)
```

The [anger is fire]{.smallcaps} metaphor is the second most salient in the type-based dataset (`r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'fire'))` types and `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'fire'))` mappings). Rajeg [-@rajeg_metafora_2013] identified a statistically significant association between [fire]{.smallcaps} and anger in Indonesian.

- fire &rarr; anger
- burning object &rarr; angry person
- `r mapping_fire_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+)\\sis\\s(.+)", "\\2 &rarr; \\1") %>% str_replace_all("the highest degree of", "intensity of") %>% str_replace(" intensity$", "")` (type=`r mapping_fire_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_fire_discuss %>% filter(str_detect(MAPPING, 'highest degree')) %>% pull(LU), "* '", mapping_fire_discuss %>% filter(str_detect(MAPPING, 'highest degree')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")` 

- `r mapping_fire_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+)\\s\\<\\-\\s(.+)", "\\2 &rarr; \\1")` (type=`r mapping_fire_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_fire_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(LU), "* '", mapping_fire_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_fire_discuss %>% filter(str_detect(MAPPING, 'causing')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+)\\s\\<\\-\\s(.+)", "\\2 &rarr; \\1")` (type=`r mapping_fire_discuss %>% filter(str_detect(MAPPING, 'causing')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_fire_discuss %>% filter(str_detect(MAPPING, 'causing anger')) %>% pull(LU), "* '", mapping_fire_discuss %>% filter(str_detect(MAPPING, 'causing anger')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

Based on the number of the linguistic expressions, the meaning focus of the metaphor is on the increased intensity of anger, which is the same in the token-based dataset ([§\@ref(fire-tokenbased)](#fire-tokenbased)).

### [Cause of anger is physical contact/harm]{.smallcaps} {#harm-typebased}

```{r mapping-harm, message = FALSE, include = FALSE}
mapping_harm <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "harm$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(person_harmed)"), "key", "no"))
mapping_harm_stats <- mapping_harm %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_harm_discuss <- get_mappings(metaphor_typebased_mapping, 'harm$') %>% left_join(mapping_harm %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_harm_stats)
```

This metaphor is a sub-case of the [emotional effect is physical contact]{.smallcaps} metaphor [see @lakoff_metaphors_1980, 50] and is the third most salient metaphor for anger with `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'physical contact'))` types and `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'physical contact'))` mappings.

- harmed/entity in contact &rarr; person('s body)
- `r mapping_harm_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_harm_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_harm_discuss %>% filter(str_detect(MAPPING, 'causing anger')) %>% pull(LU), "* '", mapping_harm_discuss %>% filter(str_detect(MAPPING, 'causing anger')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_harm_stats %>% slice_min(order_by = n_lu) %>% pull(MAPPING) %>% gsub(x = ., pattern = "-", replacement = " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_harm_stats %>% slice_min(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_harm_discuss %>% filter(str_detect(MAPPING, 'effect.of.touch')) %>% pull(LU), "* '", mapping_harm_discuss %>% filter(str_detect(MAPPING, 'effect.of.touch')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

The [physical harm/contact]{.smallcaps} metaphor focuses on the cause of anger, conceptualised as physical touching or harm to a person and/or the person’s body. In the token-based dataset ([§\@ref(harm-tokenbased)](#harm-tokenbased)), this metaphor ranks very low (rank `r get_metaphor_salience_rank(metaphor_salience, 'physical contact')`), and highlights a different aspect.

### [Marah is disease]{.smallcaps} {#disease-typebased}

```{r mapping-disease, message = FALSE, include = FALSE}
mapping_disease <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "disease$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(person_patient)"), "key", "no"))
mapping_disease_stats <- mapping_disease %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_disease_discuss <- get_mappings(metaphor_typebased_mapping, 'disease$') %>% left_join(mapping_disease %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_disease_stats)
```

The [anger is disease]{.smallcaps} metaphor is realised by `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_type", "disease"))` types and `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_mapping", "disease"))` different mappings. The productivity of the mappings suggests that the metaphor focuses on the undesirability and effect of anger as a disease.

- patient &rarr; angry person
- disease &rarr; anger (*gondok* 'goitre')
- `r mapping_disease_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% gsub(x = ., pattern = "-", replacement = " ") %>% gsub(x = ., pattern = "(.+)_(.+)", replacement = "\\2 &rarr; \\1")` (type=`r mapping_disease_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_disease_discuss %>% filter(str_detect(MAPPING, 'undesirability')) %>% pull(LU), "* '", mapping_disease_discuss %>% filter(str_detect(MAPPING, 'undesirability')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_disease_discuss %>% filter(str_detect(MAPPING, 'effect')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_disease_discuss %>% filter(str_detect(MAPPING, 'effect')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_disease_discuss %>% filter(str_detect(MAPPING, 'effect')) %>% pull(LU), "* '", mapping_disease_discuss %>% filter(str_detect(MAPPING, 'effect')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_disease_discuss %>% filter(str_detect(MAPPING, 'calm')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_disease_discuss %>% filter(str_detect(MAPPING, 'calm')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_disease_discuss %>% filter(str_detect(MAPPING, 'calm')) %>% pull(LU), "* '", mapping_disease_discuss %>% filter(str_detect(MAPPING, 'calm')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ") %>% str_replace(fixed("; to treat s.o."), "")`

In the token-based dataset, the [disease]{.smallcaps} metaphor does not make into the top-20 most salient metaphors (rank `r get_metaphor_salience_rank(metaphor_salience, 'disease')`; token=`r get_salience_stats(metaphor_salience, "n_token", "disease")`; type=`r get_salience_stats(metaphor_salience, "n_type", "disease")`; mapping=`r get_salience_stats(metaphor_salience, "n_mapping", "disease")`). This indicates that the two different methods can complement one another to highlight the salient metaphor of a target domain.

### [Marah is insanity]{.smallcaps} {#insanity-typebased}

```{r mapping-insanity, message = FALSE, include = FALSE}
mapping_insanity <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "insanity$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(angry-person_insane-person|anger_insanity)"), "key", "no"))
mapping_insanity_stats <- mapping_insanity %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_insanity_discuss <- get_mappings(metaphor_typebased_mapping, 'insanity$') %>% left_join(mapping_insanity %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_insanity_stats)
```

Kövecses [-@kovecses_metaphors_1986, 20-22] proposes that the [insanity]{.smallcaps} metaphor is one of the principal metaphors for anger [see also @holland_cognitive_1987, 204-205]. It appears to be the case for Indonesian in the type-based analysis since it ranks in the top-5 list. All expressions (type=`r get_salience_stats(metaphor_typebased_salience, "n_type", "insanity")`) evoke the following mappings.

- `r mapping_insanity %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[1] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_insanity %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[2] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(LU), "* '", mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`
  
- `r mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'uncontrolled')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'uncontrolled')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'uncontrolled')) %>% pull(LU), "* '", mapping_insanity_discuss %>% filter(str_detect(MAPPING, 'uncontrolled')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

The metaphor focuses on the control aspect. Particularly, the experiencer has become "completely incapacitated cognitively as well as in terms of behavior" [@kovecses_metaphor_2000, 74] and exhibits insane behaviours (the third and fourth mappings from the top). This implies that anger is an intense emotion forcing the experiencer to lose all controls. The [insanity]{.smallcaps} metaphor is also found in the token-based dataset but much less prominent (rank `r get_metaphor_salience_rank(metaphor_salience, "insanity")`) and has a lower number of types (`r get_salience_stats(metaphor_salience, "n_type", "insanity")`), despite the same number of mappings. Rajeg [-@rajeg_metafora_2013, 111] also found only one token for the [insanity]{.smallcaps} metaphor (*[exp]{.smallcaps} __mabuk karena__ amarah* '[exp]{.smallcaps} be _drunk with_ anger').

### [Marah is a pressurised substance in a container]{.smallcaps} {#pressurised-typebased}

```{r mapping-pressurised-substance, message = FALSE, include = FALSE}
mapping_pressurisedfluid <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "pressurised substance in a container$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_contents/located-entity|angry person\\'s body)"), "key", "no"))
mapping_pressurisedfluid_stats <- mapping_pressurisedfluid %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_pressurisedfluid_discuss <- get_mappings(metaphor_typebased_mapping, 'pressurised substance in a container$') %>% left_join(mapping_pressurisedfluid %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_pressurisedfluid_stats)
```

The [pressurised substance in a container]{.smallcaps} is part of the [substance/containment]{.smallcaps} family of metaphors and found to be statistically central for anger in Indonesian [@rajeg_metafora_2013, 211]. The metaphor is based on the lexical units (LUs) evoking two MN frames, the [pressure in a container]{.smallcaps} (see the fourth mapping from the top below) and the [(caused) upward motion]{.smallcaps} frames (the third and fifth mappings) and realises `r metaphor_typebased_salience %>% filter(str_detect(metaphor, "pressurised")) %>% pull(n_type) |> numbers2words()` types of metaphorical expressions and `r metaphor_typebased_salience %>% filter(str_detect(metaphor, "pressurised")) %>% pull(n_mapping) |> numbers2words()` mappings presented below.

- `r mapping_pressurisedfluid %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[1] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_pressurisedfluid %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[2] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_pressurisedfluid_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_pressurisedfluid_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'caused to')) %>% pull(LU), "* '", mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'caused to')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(LU), "* '", mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'losing control')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_pressurisedfluid_stats %>% filter(str_detect(MAPPING, "causing increased")) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_pressurisedfluid_stats %>% filter(str_detect(MAPPING, "causing increased")) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'causing increased')) %>% pull(LU), "* '", mapping_pressurisedfluid_discuss %>% filter(str_detect(MAPPING, 'causing increased')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

The predominant metaphorical expressions show that anger is conceptualised as the blood substance, being increased/raised due to high pressure as captured in the third and fifth mappings. These mappings reveal the main meaning focus of (caused) increased intensity of the experienced anger and are related to the generic metaphor [intensity is verticality/height]{.smallcaps} ([§\@ref(verticality-typebased)](#verticality-typebased)). The inability to control such an increase makes the person explode (the fourth mapping). In the corpus dataset ([§\@ref(pressurised-tokenbased)](#pressurised-tokenbased)), it is also ranked in the top 10.

### [(Cause of) anger is annoyance]{.smallcaps} {#annoyance-typebased}

```{r mapping-annoyance, message = FALSE, include = FALSE}
mapping_annoyance <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "annoyance$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "([(]un[)]hidden\\-object)"), "key", "no"))
mapping_annoyance_stats <- mapping_annoyance %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_annoyance_discuss <- get_mappings(metaphor_typebased_mapping, 'annoyance$') %>% left_join(mapping_annoyance %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_annoyance_stats)
```

The metaphor predominantly captures the cause of anger based on metaphorical expressions referring to various physical and psychological disturbance manifested in the `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'annoyance'))` mappings below. This metaphor is only identified in the lexical dataset.

- `r mapping_annoyance_stats %>% filter(str_detect(MAPPING, "cause.of.anger.annoyance")) %>% pull(MAPPING) %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_annoyance_stats %>% filter(str_detect(MAPPING, "cause.of.anger.annoyance")) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_annoyance_discuss %>% filter(str_detect(MAPPING, 'cause.of')) %>% pull(LU), "* '", mapping_annoyance_discuss %>% filter(str_detect(MAPPING, 'cause.of')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`


- `r mapping_annoyance_stats %>% filter(str_detect(MAPPING, "irritated.person")) %>% pull(MAPPING) %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_annoyance_stats %>% filter(str_detect(MAPPING, "irritated.person")) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_annoyance_discuss %>% filter(str_detect(MAPPING, 'irritated.person')) %>% pull(LU), "* '", mapping_annoyance_discuss %>% filter(str_detect(MAPPING, 'irritated.person')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_annoyance_stats %>%  filter(str_detect(MAPPING, "person.causing.anger")) %>% pull(MAPPING)  %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_annoyance_stats %>% filter(str_detect(MAPPING, "person.causing.anger")) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_annoyance_discuss %>% filter(str_detect(MAPPING, 'person.causing')) %>% pull(LU), "* '", mapping_annoyance_discuss %>% filter(str_detect(MAPPING, 'person.causing')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`


Several types refer to generic annoyances/disturbances (e.g., *pengacauan* 'disturbance', *gangguan* 'hindrance', *pengacau* 'agitator'); they can refer to physical or psychological annoyances and are underspecified for the degree of affectedness. The effect of *salah penerimaan* 'wrong reception' could be considered as a psychological annoyance. There are specific verbal and nominal patterns based on the noun root *bawang* 'shallot; onion'. The motivation behind the choice of *bawang* could be the irritation of the eyes (leading to the production of tears) that it causes when someone chops it. The physical irritation caused by interacting with shallot/onion is mapped onto the cause of anger.

### [Marah is bitterness]{.smallcaps} {#bitterness-typebased}

```{r mapping-bitterness, message = FALSE, include = FALSE}
mapping_bitterness <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "bitterness$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_bitterness|person-tasting)"), "key", "no"))
mapping_bitterness_stats <- mapping_bitterness %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_bitterness_discuss <- get_mappings(metaphor_typebased_mapping, 'bitterness$') %>% left_join(mapping_bitterness %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_bitterness_stats)
```

This metaphor is manifested into `r metaphor_typebased_salience %>% filter(str_detect(metaphor, "bitterness")) %>% pull(n_type) %>% numbers2words()` metaphorical expressions reflecting `r metaphor_typebased_salience %>% filter(str_detect(metaphor, "bitterness")) %>% pull(n_mapping) %>% numbers2words()` metaphorical mappings. The metaphor is based on LUs evoking the MN [taste]{.smallcaps} frame.

- `r mapping_bitterness %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[1] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_bitterness %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[2] %>% str_replace_all("-", " ") %>% str_replace_all("_", " <- ") %>% str_replace_all("^(.+)( \\<\\- )(.+)$", "\\1 &rarr; \\3")`

- `r mapping_bitterness_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING)  %>% str_replace_all("-", " ") %>% str_replace_all("([^_]+)_([^_]+)", "\\1 &rarr; \\2")` (type=`r mapping_bitterness_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_bitterness_discuss %>% filter(str_detect(MAPPING, 'undesirability')) %>% pull(LU), "* '", mapping_bitterness_discuss %>% filter(str_detect(MAPPING, 'undesirability')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_bitterness_stats %>% slice_min(order_by = n_lu) %>% pull(MAPPING)  %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_bitterness_stats %>% slice_min(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_bitterness_discuss %>% filter(str_detect(MAPPING, 'reduced')) %>% pull(LU), "* '", mapping_bitterness_discuss %>% filter(str_detect(MAPPING, 'reduced')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ") %>% str_replace_all("[()]", "")`


The third mapping has the predominant linguistic expressions. This indicates that the [bitterness]{.smallcaps} metaphor highlights undesirability of anger, reflecting the negative evaluation of it. The final mapping suggests that the bitterness could be reduced (i.e., turned into a plainer taste), which is mapped onto reducing the undesirability (and perhaps the intensity) of anger. In the corpus dataset, the metaphor is also attested but not salient as it sits in rank `r get_metaphor_salience_rank(metaphor_salience, 'bitternes')` out of the total `r nrow(metaphor_salience)` metaphors found ([Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based)). Rajeg's [-@rajeg_metafora_2013] work also did not identify this metaphor for anger in his literary corpus.

### [Marah is heated fluid in a container]{.smallcaps} {#hotfluid-typebased}

```{r mapping-hotfluid, message = FALSE, include = FALSE}
mapping_hotfluid <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "heated")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "angry-person_container"), "key", "no"),
         status = if_else(MAPPING == "anger_(heated-)fluid", "key", status))
mapping_hotfluid_stats <- mapping_hotfluid %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_hotfluid_discuss <- get_mappings(metaphor_typebased_mapping, 'heated') %>% left_join(mapping_hotfluid %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_hotfluid_stats)
```

This is a heated variant of the [contained substance]{.smallcaps} metaphor ([§\@ref(contained-substance-typebased)](#contained-substance-typebased)), incorporating the semantics of the MN [heated fluid]{.smallcaps} frame. The [heated fluid in a container]{.smallcaps} is one of the distinctive metaphors for anger in Indonesian [@rajeg_metafora_2013]. In the lexical database, this metaphor is not too salient, represented only by `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'heated fluid'))` types of metaphorical expressions, and `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'heated fluid'))` mappings, two of which are the foundational mappings.

- heated fluid &rarr; anger

- the container &rarr; angry person

- `r mapping_hotfluid_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING)  %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_hotfluid_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_hotfluid_discuss %>% filter(str_detect(MAPPING, 'heat\\-level')) %>% pull(LU), "* '", mapping_hotfluid_discuss %>% filter(str_detect(MAPPING, 'heat\\-level')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_hotfluid_stats %>% slice_min(order_by = n_lu) %>% pull(MAPPING)  %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_hotfluid_stats %>% slice_min(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_hotfluid_discuss %>% filter(str_detect(MAPPING, 'being caused to')) %>% pull(LU), "* '", mapping_hotfluid_discuss %>% filter(str_detect(MAPPING, 'being caused to')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`
  
The predominant number of metaphorical expressions reveals that the metaphor highlights the high degree of intensity. This is conceptualised as a boiling substance. The final mapping also pertains to intensity, particularly the increasing degree of heat of the substance.

### [Intensity is verticality/height]{.smallcaps} {#verticality-typebased}

```{r mapping-verticality, message = FALSE, include = FALSE}
mapping_verticality <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "verticality$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_verticality-scale)"), "key", "no"))
mapping_verticality_stats <- mapping_verticality %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_verticality_discuss <- get_mappings(metaphor_typebased_mapping, 'verticality$') %>% left_join(mapping_verticality %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_verticality_stats)
```

This is a generic, orientational metaphor based on the [(caused) upward motion]{.smallcaps} and the [verticality scale]{.smallcaps} frames. Rajeg [-@rajeg_metafora_2013, 212] identified the significant statistical association of this metaphor with anger. The metaphor is realised in `r numbers2words(pull(filter(metaphor_typebased_salience, metaphor == "intensity of anger is verticality"), n_type))` expressions evoking `r numbers2words(pull(filter(metaphor_typebased_salience, metaphor == "intensity of anger is verticality"), n_mapping))` mappings. 


- `r mapping_verticality %>% filter(status == 'key') %>% select(MAPPING) %>% distinct() %>% pull(MAPPING) %>% str_replace_all("-", " ") %>% str_replace_all("_", " <- ") %>% paste(collapse = "; ") %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; intensity of \\1")`

- `r mapping_verticality_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_verticality_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_verticality_discuss %>% filter(str_detect(MAPPING, 'upward motion')) %>% pull(LU), "* '", mapping_verticality_discuss %>% filter(str_detect(MAPPING, 'upward motion')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_verticality_stats %>% filter(str_detect(MAPPING, "less intensity")) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_verticality_stats %>% filter(str_detect(MAPPING, "less intensity")) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_verticality_discuss %>% filter(str_detect(MAPPING, "less intensity")) %>% pull(LU), "* '", mapping_verticality_discuss %>% filter(str_detect(MAPPING, "less intensity")) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

The main meaning focus is on the increasing intensity of anger (without necessarily reaching the highest end of the scale). This is evident from the most productive mapping evoking the [(caused) upward motion]{.smallcaps} frame. The least frequent mapping indicates short-lived, low-intensity anger since the linguistic expression refers to anger not reaching the endpoint of the scale (*marah __tiada sampai__* 'short-lived anger; lit. anger not reaching the endpoint of a scale').

### [Marah is a contained substance]{.smallcaps} {#contained-substance-typebased}

This is a short-hand label for the [emotion is substance in a container]{.smallcaps} metaphor. The metaphor is realised in `r numbers2words(pull(filter(metaphor_typebased_salience, metaphor == "anger is substance in a container"), n_type))` metaphorical expressions evoking `r numbers2words(pull(filter(metaphor_typebased_salience, metaphor == "anger is substance in a container"), n_mapping))` mappings. It is also found to be statistically associated with anger in Indonesian [@rajeg_metafora_2013, 211-212].

```{r mapping-containedfluid, message = FALSE, include = FALSE}
mapping_containedfluid <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "is substance in a container$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_contents/located-entity)") & LU != "darah muda", "key", "no"),
         status = if_else(str_detect(MAPPING, "angry-person_container"), "key", status),
         status = if_else(str_detect(MAPPING, "(anger-level_substance-fullness-level)"), "key", status))
mapping_containedfluid_stats <- mapping_containedfluid %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_containedfluid_discuss <- get_mappings(metaphor_typebased_mapping, 'is substance in a container$') %>% left_join(mapping_containedfluid %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_containedfluid_stats)
```

- `r mapping_containedfluid %>% filter(status == 'key') %>% select(MAPPING) %>% distinct() %>% pull(MAPPING) %>% .[2] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_containedfluid %>% filter(status == 'key') %>% select(MAPPING) %>% distinct() %>% pull(MAPPING) %>% .[3] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_containedfluid_stats %>% filter(str_detect(MAPPING, "located.entity")) %>% pull(MAPPING) %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_containedfluid_stats %>% filter(str_detect(MAPPING, "located.entity")) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_containedfluid_discuss %>% filter(str_detect(MAPPING, "located.entity")) %>% pull(LU), "* '", mapping_containedfluid_discuss %>% filter(str_detect(MAPPING, "located.entity")) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_containedfluid_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_containedfluid_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_containedfluid_discuss %>% filter(str_detect(MAPPING, 'channel+ing')) %>% pull(LU), "* '", mapping_containedfluid_discuss %>% filter(str_detect(MAPPING, 'channel+ing')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

The last mapping is the most productive. It indicates the main meaning focus of the metaphor, namely externalisation/expression of anger construed as releasing the substance, based on the LUs evoking the [release liquid]{.smallcaps} frame. ([§\@ref(contained-substance-typebased)](#contained-substance-tokenbased)). This focus implies the inability to keep the anger inside.

The metaphor is part of the [substance/containment]{.smallcaps} source frames family postulated in the database. The other source frames are [heated fluid in a container]{.smallcaps} ([§\@ref(hotfluid-typebased)](#hotfluid-typebased)) and [pressurised substance in a container]{.smallcaps} ([§\@ref(pressurised-typebased)](#pressurised-typebased)). The difference between the [contained substance]{.smallcaps} metaphor and the [heated fluid]{.smallcaps} or the [pressurised substance]{.smallcaps} variants is that the former lacks the heated and pressurised semantics of the frames. The intensity in the [contained substance]{.smallcaps} metaphor is construed via the quantity of the substance, and the expression of anger is captured via the externalisation of the substance.

### [Degree of controlling anger is the size of patience as a container]{.smallcaps} {#patience}

```{r mapping-dimension, message = FALSE, include = FALSE}
mapping_dimension <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "dimension$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "anger_entity"), "key", "no"))
mapping_dimension_stats <- mapping_dimension %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_dimension_discuss <- get_mappings(metaphor_typebased_mapping, 'dimension$') %>% left_join(mapping_dimension %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_dimension_stats)
```

The [size-]{.smallcaps}/[dimension-]{.smallcaps}related metaphor is based on LUs referring to the [size]{.smallcaps} frame. It represents a generic though different nuance to capture emotional control from the previous metaphors focusing on the same aspect. There are `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'dimension'))` metaphorical expressions and `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'dimension'))` mappings for the metaphor as shown below. 

- physical entity &rarr; anger

- `r mapping_dimension_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_dimension_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_dimension_discuss %>% filter(str_detect(MAPPING, 'narrow')) %>% pull(LU), "* '", mapping_dimension_discuss %>% filter(str_detect(MAPPING, 'narrow')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`
  
The metaphorical expressions reveal the main meaning focus of inability to control anger. The size/dimension of the object could be viewed as the level of patience/calmness. The inferential aspect mapped from the source frame is that when this patience metaphorically shrinks, the person is no longer calm (i.e., unable to control the emotion) but becoming emotional (i.e., angry). This is a different inference in the use of the [size]{.smallcaps} frame for the intensity level of anger, which is only found in the corpus data ([§\@ref(quantity-tokenbased)](#quantity-tokenbased)).  

### [Marah is a natural force]{.smallcaps} {#natural-force-typebased}

```{r mapping-naturalforce, message = FALSE, include = FALSE}
mapping_naturalforce <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "natural force$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "anger_natural-disaster"), "key", "no"))
mapping_naturalforce_stats <- mapping_naturalforce %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_naturalforce_discuss <- get_mappings(metaphor_typebased_mapping, 'natural force$') %>% left_join(mapping_naturalforce %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_naturalforce_stats)
```

This metaphor is based on metaphorical expressions referring to the [natural disaster]{.smallcaps} frame, a sub-case of the [harm]{.smallcaps} frame family in the MN frame repository. The metaphor is found to be strongly associated with anger [@rajeg_metafora_2013, 213]. The lexical dataset reveals only `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'natural force'))` types of expressions evoking `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'natural force'))` mappings.

- natural disaster &rarr; anger

- intensity of the natural force &rarr; intensity of anger (type=1)

  - *angin-anginan* 'windy; capricious; unpredictable'

- `r mapping_naturalforce_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1 ")` (type=`r mapping_naturalforce_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_naturalforce_discuss %>% filter(str_detect(MAPPING, 'to cease')) %>% pull(LU), "* '", mapping_naturalforce_discuss %>% filter(str_detect(MAPPING, 'to cease')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_naturalforce_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_naturalforce_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(n_lu) %>% unique()`)

  - `r paste(paste("*", mapping_naturalforce_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(LU), "* '", mapping_naturalforce_discuss %>% filter(str_detect(MAPPING, 'ceasing')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ") %>% str_replace_all("[(]", "[") %>% str_replace_all("[)]", "]") %>% str_replace_all(".calm down.+(?=subside)", "'")`

The metaphorical expressions particularly highlight the ceasing intensity of the forces, suggesting the main meaning focus on the intensity of anger. The similar theme is found in the corpus study ([§\@ref(naturalforce-tokenbased)](#naturalforce-tokenbased)). 

### [Marah is a substance]{.smallcaps} {#solid-typebased}

```{r mapping-solid, message = FALSE, include = FALSE}
mapping_solid <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "rough, solid entity$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_entity|degree.of.solidity)"), "key", "no"))
mapping_solid_stats <- mapping_solid %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_solid_discuss <- get_mappings(metaphor_typebased_mapping, 'rough, solid entity$') %>% left_join(mapping_solid %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_solid_stats)
```

This [object]{.smallcaps}-related metaphor captures the intensity of anger in terms of object's (or substance's) firmness/solidity. The metaphor is realised by `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'solid'))` metaphorical expressions and consists of `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'solid'))` mappings. 


- `r mapping_solid %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[1] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_solid %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% .[2] %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`

- `r mapping_solid_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING)  %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_solid_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`) (`r paste(paste("*", mapping_solid_discuss %>% filter(str_detect(MAPPING, 'calmness')) %>% pull(LU), "* '", mapping_solid_discuss %>% filter(str_detect(MAPPING, 'calmness')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`)

- melting ice &rarr; the reduction of anger (type=1) (*mencair* 'no longer being angry; lit. to melt')
  
The metaphorical expressions evoke two frames related to the [firmness]{.smallcaps} frame. The first one is the [lax]{.smallcaps} frame (see the third mapping from the top above), which is in a scalar opposition to the [firm]{.smallcaps} frame. The second frame is the [liquid]{.smallcaps} (the last mapping), capturing the fact that a solid entity (e.g., an ice block) can be pliable like a liquid. All metaphorical expressions focus on reduced intensity.

### [Marah is an (un)veiled object]{.smallcaps} {#hidden-typebased}

```{r mapping-hidden, message = FALSE, include = FALSE}
mapping_hidden <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "veiled object$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "([(]un[)]hidden\\-object)"), "key", "no"))
mapping_hidden_stats <- mapping_hidden %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_hidden_discuss <- get_mappings(metaphor_typebased_mapping, 'veiled object$') %>% left_join(mapping_hidden %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_hidden_stats)
```

This metaphor is realised by `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_type", "veiled"))` types of expressions, manifesting `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_mapping", "veiled"))` mappings. They evoke the [hiding]{.smallcaps} frame (the second mapping from the top below) and [caused upward motion]{.smallcaps} frame (the third mapping).

- `r mapping_hidden %>% filter(status == "key") %>% pull(MAPPING) %>% unique() %>% str_replace_all("[-]", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")`. 

- `r mapping_hidden_stats %>% slice_max(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")` (type=`r mapping_hidden_stats %>% slice_max(order_by = n_lu) %>% pull(n_lu)`)

  - `r paste(paste("*", mapping_hidden_discuss %>% filter(str_detect(MAPPING, 'hiding')) %>% pull(LU), "* '", mapping_hidden_discuss %>% filter(str_detect(MAPPING, 'hiding')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`

- `r mapping_hidden_stats %>% slice_min(order_by = n_lu) %>% pull(MAPPING) %>% str_replace_all("(.+) is (.+)", "\\2 &rarr; \\1")  %>% str_replace(fixed("i.e."), "i.e.,")` (type=`r mapping_hidden_stats %>% slice_min(order_by = n_lu) %>% pull(n_lu)`) (`r paste(paste("*", mapping_hidden_discuss %>% filter(str_detect(MAPPING, 'unveiling')) %>% pull(LU), "* '", mapping_hidden_discuss %>% filter(str_detect(MAPPING, 'unveiling')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`)

The [(un)veiled object]{.smallcaps} metaphor relates to the social expression and degree of control of the experiencer to hide or expose the anger. The focus of the metaphor, based on the predominant metaphorical expressions, is hiding the anger. The corpus study ([§\@ref(unveiled-tokenbased)](#unveiled-tokenbased)), however, reveals a different meaning focus of the metaphor.

### [Marah is heat]{.smallcaps} {#heat-typebased}

```{r mapping-heat, message = FALSE, include = FALSE}
mapping_heat <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "heat$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_heat)"), "key", "no"))
mapping_heat_stats <- mapping_heat %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_heat_discuss <- get_mappings(metaphor_typebased_mapping, 'heat$') %>% left_join(mapping_heat %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_heat_stats)
```

[Heat]{.smallcaps} is considered to be a central, embodied source frame for anger in previous works [e.g., @kovecses_metaphors_1986; @holland_cognitive_1987; but cf. @delbecque_anger_2005 for the cultural perspective based on diachronic evidence in the history of the English language]. It is proposed as the basis for two specific metaphors, namely [anger is fire]{.smallcaps} ([§\@ref(fire-typebased)](#fire-typebased) and [\@ref(fire-tokenbased)](#fire-tokenbased)) and [anger is heated fluid in a container]{.smallcaps} ([§\@ref(hotfluid-typebased)](#hotfluid-typebased) and [\@ref(hotfluid-tokenbased)](#hotfluid-tokenbased)). This study found that the generic [heat]{.smallcaps} metaphor is not salient, being realised only by `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'heat$'))` metaphorical expressions manifesting `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'heat$'))` mappings as follows:

- heat &rarr; anger

- `r mapping_heat_stats %>% filter(str_detect(MAPPING, "heated\\-entity$")) %>% pull(MAPPING) %>% str_replace_all("-", " ") %>% str_replace_all("(.+)_(.+)", "\\2 &rarr; \\1")` (type=`r mapping_heat_stats %>% filter(str_detect(MAPPING, "heated\\-entity$")) %>% pull(n_lu)`) (`r paste(paste("*", mapping_heat_discuss %>% filter(str_detect(MAPPING, 'heated\\-entity')) %>% pull(LU), "* '", mapping_heat_discuss %>% filter(str_detect(MAPPING, 'heated\\-entity')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`)

- `r mapping_heat_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(MAPPING2) %>% unique() %>% str_replace_all("(.+) \\<\\- (.+)", "\\2 &rarr; \\1")` (type=`r mapping_heat_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(n_lu) %>% unique()`) (`r paste(paste("*", mapping_heat_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(LU), "* '", mapping_heat_discuss %>% filter(str_detect(MAPPING, 'becoming')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`)

The difference between the [heat]{.smallcaps} metaphor in the lexical approach, and the [temperature]{.smallcaps} metaphor in the corpus approach is as follows. The former focuses on the characteristic of easily angered person as always hot, reflected by the linguistic expressions that do not encode any reduction of the heat. Meanwhile, the latter (i.e., [temperature]{.smallcaps}) highlights the dynamics of the intensity when being angry, namely, (i) can be reduced (i.e., cooled off) or (ii) reaching the highest temperature; this is supported by the linguistic expression referring to coolness. In the MN frame repository, [heat]{.smallcaps} is a different frame and considered as a perspective of the [temperature]{.smallcaps} frame, which covers the range of temperature scale (from hot to cold). The [heat]{.smallcaps} frame itself is also linked to the [cold]{.smallcaps} frame via "in scalar opposition to" relation.

### [Marah is a sleeping organism]{.smallcaps} {#awakening-typebased}

```{r mapping-sleeping, message = FALSE, include = FALSE}
mapping_sleeping <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "sleeping organism$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_slepping\\-entity$)"), "key", "no"))
mapping_sleeping_stats <- mapping_sleeping %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_sleeping_discuss <- get_mappings(metaphor_typebased_mapping, 'sleeping$') %>% left_join(mapping_sleeping %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_sleeping_stats)
```

This metaphor is based on the [awakening]{.smallcaps} frame in which anger is mapped onto the inactive/calm sleeping organism role in the frame [cf. @stefanowitsch_words_2006, 76-77]. There are only `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_type", "sleeping"))` metaphorical expressions evoking the following `r numbers2words(get_salience_stats(metaphor_typebased_salience, "n_mapping", "sleeping"))` mappings: 

- sleeping organism &rarr; anger
- awakening/fishing out the organism &rarr; causing anger (*__membangkitkan__ marah* 'to _arouse_ one's anger')
- awaken entity &rarr; anger near the limit (*__bangkit__ amarah* 'anger *rose*')

The metaphorical expressions suggest the causation and onset of anger. From one perspective, the [sleeping organism]{.smallcaps} metaphor could be viewed as a pre-cursor, or a part, of the [fierce, captive animal]{.smallcaps} metaphor where the awakened organism/animal has been in a fierce and aggressive behaviour (e.g., due to being awakened/disturbed). 

The [awakening]{.smallcaps} frame, and in combination with the [fierce animal]{.smallcaps} frame in [anger is fierce, captive animal]{.smallcaps}, is also reflected in the common Indonesian proverb for anger *__membangunkan__ macan tidur* 'to *awaken* a sleeping tiger'. The verb *membangunkan* 'to awaken; to wake s.b. up' and *tidur* '(a)sleep' evoke the [awakening]{.smallcaps} frame, meanwhile *macan* 'tiger' evokes the [fierce animal]{.smallcaps} frame. The proverb highlights the danger of invoking anger out of a calm state.

### [Marah is possession]{.smallcaps}

This is not a prominent metaphor in the lexical approach as it is realised by only one type (*ambil* 'take'), evoking the [gain possession]{.smallcaps} frame:

(@typebased_gain-possession) *Jangan __ambil__ marah* 'Do not _take_ the anger' (Stevens Schimdgal Tellings 2004: 30)

This example means that the person should stay calm by not taking the anger object into his/her possession, implying a negative valence of possessing an anger-object. The mappings inferred from (@typebased_gain-possession) are as follows:

- possessable object &rarr; anger
- (candidate) possessor &rarr; experiencer
- taking/gaining the possessable object &rarr; becoming angry

The low salience of the [possession]{.smallcaps} metaphor family is not surprising given that Rajeg [-@rajeg_metafora_2013, 213-214] found the metaphor's statistically significant dissociation with anger ([§\@ref(intro)](#intro)).

### [Marah is darkness]{.smallcaps} {#darkness-typebased}

```{r mapping-darkness, message = FALSE, include = FALSE}
mapping_darkness <- filter(metaphor_typebased_mapping, str_detect(CM_BROADER, "darkness$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_darkness$)"), "key", "no"))
mapping_darkness_stats <- mapping_darkness %>% filter(status == 'no') %>% group_by(MAPPING) %>% summarise(n_lu = n_distinct(LU))
mapping_darkness_discuss <- get_mappings(metaphor_typebased_mapping, 'darkness$') %>% left_join(mapping_darkness %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_darkness_stats)

#(@typebased_darkness-1) `r metaphor_typebased %>% filter(str_detect(CM_BROADER, 'darkness'), LU == "kegelapan") %>% pull(CITATIONS)` (`r metaphor_typebased %>% filter(str_detect(CM_BROADER, 'darkness'), LU == "kegelapan") %>% pull(SOURCES)`: `r metaphor_typebased %>% filter(str_detect(CM_BROADER, 'darkness'), LU == "kegelapan") %>% pull(PAGE_LOCATION)`)
```

This metaphor is based on LUs from the [darkness]{.smallcaps} frame and is evoked by `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_type', 'darkness'))` types manifesting `r numbers2words(get_salience_stats(metaphor_typebased_salience, 'n_mapping', 'darkness'))` mappings as follows:

- darkness &rarr; anger

- person(’s body) being in darkness &rarr; angry person (type=`r unique(pull(filter(mapping_darkness_discuss, str_detect(MAPPING2, "person being in the dark state")), n_lu))`) (`r paste(paste("*", mapping_darkness_discuss %>% filter(str_detect(MAPPING, 'dark.state')) %>% pull(LU), "* '", mapping_darkness_discuss %>% filter(str_detect(MAPPING, 'dark.state')) %>% pull(LU_GLOSS), "'", sep = ""), collapse = ", ")`)

Semantically, the metaphor highlights the negative assessment of anger [@stefanowitsch_words_2006, 77]. One prototypical example of the negative effect of a person being in the dark is inability to see (and by extension, to think) clearly (or rationally) (cf. [§\@ref(visual-interference-typebased)](#visual-interference-typebased)).

## Type-based salience: Metonymy {#typebased-metonymy-salience-results}

There are `r sum(metonymy_typebased_salience$n_type)` types of metonymic expressions for [anger]{.smallcaps} grouped under `r numbers2words(nrow(metonymy_typebased_salience))` (conceptual) metonymies. The metonymies are rank ordered in [Table \@ref(tab:metonymy-table-type-based)](#metonymy-table-type-based) by the number of their metonymic expressions.

```{r metonymy-category-typebased}
physiological_response_regex <- "(body heat|internal pressure|redness|breathing|visual interference|body hair|teeth biting|foaming|caressing|clenching)"
metonymy_typebased_salience_category <- metonymy_typebased_salience %>% 
  mutate(category = if_else(str_detect(metonymy, physiological_response_regex), "physiological response", "social-communicative behaviour"))

physiology_type_1 <- metonymy_typebased_salience_category %>% 
  filter(category == "physiological response") %>% 
  mutate(meto_stats = paste("[", metonymy, "]{.smallcaps} (type=", n_type, ")", sep = "")) %>% 
  pull(meto_stats)
physiology_type_1_total <- paste("*TOTAL~type~=", sum(filter(metonymy_typebased_salience_category, category == "physiological response")$n_type),"*", sep = "")
physiology_type_1 <- c(physiology_type_1, physiology_type_1_total)

social_type_1 <- metonymy_typebased_salience_category %>% 
  filter(category != "physiological response") %>% 
  mutate(meto_stats = paste("[", metonymy, "]{.smallcaps} (type=", n_type, ")", sep = "")) %>% 
  pull(meto_stats)
social_type_1_total <- paste("*TOTAL~type~=", sum(filter(metonymy_typebased_salience_category, category != "physiological response")$n_type),"*", sep = "")
social_type_1 <- c(social_type_1, rep("", (length(physiology_type_1) - 1) - length(social_type_1)), social_type_1_total)

metonymy_typebased_category_print <- tibble(`social-communicative behaviour:` = social_type_1, `physiological response:` = physiology_type_1)
```


```{r metonymy-table-type-based}
metonymy_typebased_salience %>% 
  mutate(metonymy = str_replace(metonymy, "frustated", "frustrated"),
         metonymy = paste("[", metonymy, "]{.smallcaps}", sep = ""),
         metonymy = str_replace_all(metonymy, fixed("visual interference"), "inability to see"),
         metonymy = str_replace_all(metonymy, "(?<=\\sfor\\s)(anger)(?=\\])", "marah")) %>% 
  rename(Metonymy = metonymy,
         `Types:` = n_type,
         `% of all types:` = n_perc_type) %>% 
  knitr::kable(caption = "Conceptual metonymies for [anger]{.smallcaps} (Type-based, lexical approach)", row.names = FALSE)
```


The salience of each metonymy can only be measured in terms of the number of the metonymic expressions, given the metonymies only involve a single mapping. The mapping is reflected in the naming of the metonymy in the form of [metonymic source domain for metonymic target]{.smallcaps} (or [x for y]{.smallcaps} for short).

### [Strong/aggressive verbal behaviour for anger]{.smallcaps} {#verbal-behaviour-typebased}

The type-based analysis reveals that [`r str_replace(pull(slice_max(metonymy_typebased_salience, order_by = n_type), metonymy), " for anger", "")`]{.smallcaps} is the predominant metonymy for anger in Indonesian (*X*^2^~goodness~ ~of~ ~fit~=`r round(chisq.test(metonymy_typebased_salience$n_type)$statistic, digit = 2)`, *df*=`r chisq.test(metonymy_typebased_salience$n_type)$parameter`, *p*=`r format(chisq.test(metonymy_typebased_salience$n_type)$p.value, digit = 3)`), showing the largest inventory of metonymic expressions.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "verbal")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "verbal")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

These expressions are mostly verbs encoding the manner of speaking that metonymically points to the internal, emotional state of the speaker. When someone speaks in the manner conveyed by these expressions, the hearer could infer that the speaker is angry. Several expressions glossed as ‘snarl/snap at’ could also reflect the metaphorical mapping “aggressive animal behaviour &rarr; angry human behaviour”.

### [Violent frustrated behaviour for anger]{.smallcaps} {#violent-behaviour-typebased}

Linguistic expressions evoking certain harsh and violent actions/behaviours can be metonymically used to refer to anger. This is motivated experientially in that frustration often leads to anger, which then brings about some irrational, harsh, dangerous, or violent actions. The actions can be directed to oneself or others.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "frustated")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "frustated")), LU_GLOSS), "'", sep = ""), collapse = "; ") %>% str_replace_all(fixed("typically"), "especially")`

The metonymy is a sub-case of a more generic metonymy, namely [effect of emotion stands for emotion]{.smallcaps}.

### [Aggressive visual behaviour for anger]{.smallcaps}

This metonymy represents the externalisation of anger via the aggressive forms of visual and/or facial behaviours. This is shown by the presence of facial-related body-parts (e.g., mouth, eyes, face) in several metonymic expressions, or in the inherent meaning of the expressions.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "visual behaviour")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "visual behaviour")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Body heat for anger]{.smallcaps} {#body-heat-typebased}

In the classic CMT literature [e.g., @holland_cognitive_1987], the increased body heat is one of the experiential motivations for the [heat]{.smallcaps}-related metaphor of anger, such as [anger is fire]{.smallcaps}. Such bodily experience is felt during an intense anger and becomes a metonymic source domain for construing anger.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "body heat")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "body heat")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

The metonymy is also a sub-case of the generic [effect of emotion stands for emotion]{.smallcaps}. The metonymic expressions refer to a situation where someone experiencing heat in their body (e.g., *radang* 'feverish', *kegerahan* 'stifling heat') that is used to indicate anger.

```{r unused-body-heat-example}
# (@meradangkan) *manuver-manuver Rita yang berjuang dengan bendera Partai Berkarya cukup __meradangkan__ partai-partai lainnya, khususnya partai-partai besar seperti Golkar, PPP, PKB, PDIP, PAN atau Demokrat.* (https://www.metropolitan.id/2019/04/kunjungi-peternak-puyuh-di-akhir-kampanye/; last access: 5 February 2022)
```


### [Internal pressure for anger]{.smallcaps} {#internal-pressure-typebased}

The linguistic expressions under this metonymy refer to different kinds of internal bodily pressures, such as blood pressure, muscular pressure, and swollen chest.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "internal pressure")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "internal pressure")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Lip/teeth biting/pressing for anger]{.smallcaps} {#lip-biting-typebased}

This metonymy can be related to the issue of emotional control. Given an intense internal pressure, it forces the angry persons to externalise and react due to their anger. In that case, biting/pressing the lip/teeth may accompany the persons effort to stay calm, and it is metonymically used to indicate anger.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "biting")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "biting")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Redness in the facial area for anger]{.smallcaps} {#redness-typebased}

Another physiological effect of anger is physiological change in the facial area, such as flushing or being livid. This bodily experience is used to refer to anger.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "redness")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "redness")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

As we can see, a different body part around the face can be involved, such as the ear (*telinga*).

### [Inability to see for anger]{.smallcaps} {#visual-interference-typebased}

This metonymy appears to be the experiential grounding of the [darkness]{.smallcaps} metaphor ([§\@ref(darkness-typebased)](#darkness-typebased)). All metonymic expressions capture the visual-perception interference as if the eyes are (seeing) dark; hence, inability to see and, by way of semantic extension, think rationally.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "interference")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "interference")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Breathing difficulty for anger]{.smallcaps} {#breathing-typebased}

Intense, excited anger could lead to intense and difficult breathing. The expressions referring to difficulty in breathing, due to excessive heat and/or no fresh air, are used metonymically to refer to anger.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "breathing")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "breathing")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Standing body hair for anger]{.smallcaps} {#bodyhair-typebased}

This metonymy might be more relevant to fear than anger. However, in the Indonesian database, the expression related to bodily hair can refer to anger. It is the case in the body-part where no bodily hair is expected to grow, namely the liver '*hati*'.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "body hair")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "body hair")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Caressing the chest for anger]{.smallcaps} {#caressing-chest-typebased}

The situation when someone holds onto, or caresses, their chest, as if holding something from getting out, can refer to someone's effort to stay under control (i.e., being patient) (cf. [§\@ref(lip-biting-typebased)](#lip-biting-typebased)). The metonymic expression below indicates that *dada* 'chest' could contain anger.

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "caressing")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "caressing")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

### [Clenching fist for anger]{.smallcaps} {#fist-typebased}

This metonymy can be interpreted as highlighting control or an expression of anger (e.g., ready to attack the wrongdoer with the punch/fist).

- `r paste(paste("*", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "clenching")), LU), "* '", pull(filter(metonymy_typebased, str_detect(CM_BROADER, "clenching")), LU_GLOSS), "'", sep = ""), collapse = "; ")`

## Types of the metonymies (type-based dataset) {#metonymy-category-typebased}

The metonymies in [Table \@ref(tab:metonymy-table-type-based)](#metonymy-table-type-based) can be categorised into two broader categories ([Table \@ref(tab:metonymy-category-table-typebased)](#metonymy-category-table-typebased) below): (i) social-communicative behaviours of anger and (ii) the associated physiological responses.

```{r metonymy-category-table-typebased}
metonymy_typebased_category_binomtest <- binom.test(sum(filter(metonymy_typebased_salience_category, category != "physiological response")$n_type), sum(sum(filter(metonymy_typebased_salience_category, category != "physiological response")$n_type), sum(filter(metonymy_typebased_salience_category, category == "physiological response")$n_type)))

metonymy_typebased_category_print %>% mutate(`physiological response:` = str_replace_all(`physiological response:`, fixed("visual interference"), "inability to see"),
                                             `social-communicative behaviour:` = str_replace_all(`social-communicative behaviour:`, fixed("frustated"), "frustrated")) %>%  rename(`Social-Communicative Behaviour:`=`social-communicative behaviour:`, `Physiological Response:`=`physiological response:`) %>% knitr::kable(caption = "Two types of metonymies of [anger]{.smallcaps} (type-based dataset)")
```

[Table \@ref(tab:metonymy-category-table-typebased)](#metonymy-category-table-typebased) demonstrates that metonymies referring to the physiological responses are greater in number than those for the social-communicative behaviour. However, the total number of metonymic linguistic expressions/types are significantly higher for the social-communicative category (type=`r sum(filter(metonymy_typebased_salience_category, category != "physiological response")$n_type)`) than the physiological category (type=`r sum(filter(metonymy_typebased_salience_category, category == "physiological response")$n_type)`) (*p*~Binomial~ ~two~~-~~tailed~`r pval_print(metonymy_typebased_category_binomtest$p.value)`).

# Token-based salience {#tokenbased-analysis}

```{r token-based-analysis-computation, message=FALSE, include=FALSE, warning=FALSE, error=FALSE}
source("codes/MARAH-token-based-analysis-code.R")
ttr_metaphor_tokenbased <- happyr::ttr(df = marah, schema_var = "CM_BROADER", lexunit_var = "MP")
```


```{r difference-type-based-and-token-based}
metaphor_in_corpus_only <- setdiff(pull(metaphor_salience, metaphor), metaphor_typebased_salience$metaphor)
metaphor_in_corpus_only_top20 <- setdiff(pull(slice_max(metaphor_salience, n = 20, order_by = aggregate, with_ties = FALSE), metaphor), metaphor_typebased_salience$metaphor)
metaphor_in_lexical_only <- setdiff(metaphor_typebased_salience$metaphor, pull(metaphor_salience, metaphor))
metaphor_in_lexical_only_top20 <- setdiff(metaphor_typebased_salience$metaphor, pull(slice_max(metaphor_salience, n = 20, order_by = aggregate, with_ties = FALSE), metaphor))
metaphor_in_lexical_only_aggregate <- as.data.frame(metaphor_typebased_salience)[metaphor_typebased_salience$metaphor %in% metaphor_in_lexical_only, "aggregate"]
metaphor_in_lexical_only_ranking <- rownames(as.data.frame(metaphor_typebased_salience)[metaphor_typebased_salience$metaphor %in% metaphor_in_lexical_only, ])
metaphor_in_all_database <- intersect(pull(metaphor_salience, metaphor), metaphor_typebased_salience$metaphor)
metaphor_in_all_database_top20 <- intersect(pull(slice_max(metaphor_salience, n = 20, order_by = aggregate, with_ties = FALSE), metaphor), metaphor_typebased_salience$metaphor)
binom_test_metaphor_lexical_and_corpus <- binom.test(c(nrow(metaphor_salience), length(metaphor_typebased_salience$metaphor)))

# The number of metaphors in the corpus-based analysis (`r nrow(metaphor_salience)` types) is significantly greater than the number of metaphors in the type-based analysis (`r length(metaphor_typebased_salience$metaphor)` types) (*p*~Binomial~~,~ ~two~~-~~tailed~ = `r round(binom_test_metaphor_lexical_and_corpus$p.value, digits = 4)`).
```


The token-based, corpus study reveals `r nrow(marah)` tokens of metaphorical expressions (manifesting `r nrow(metaphor_salience)` metaphors) ([§\@ref(tokenbased-metaphor-salience-results)](#tokenbased-metaphor-salience-results)) and `r nrow(metonymy_tokenbased)` tokens of metonymic expressions (manifesting `r numbers2words(nrow(metonymy_tokenbased_salience))` metonymies) ([§\@ref(tokenbased-metonymy-salience-results)](#tokenbased-metonymy-salience-results)).

## Metaphor {#tokenbased-metaphor-salience-results}

[Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based) present all metaphor types identified from corpus-based analysis. The top-20 metaphors in [Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based) will be discussed to roughly match the similar number of metaphors found in the lexical approach ([Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based)). The metaphorical mappings for each metaphor will be presented in the decreasing order of their type frequencies (i.e., the number of linguistic expressions) to ease the identification of the "main meaning focus" of the metaphor via "the metaphorical linguistic expressions that _dominate_ a metaphor" [@kovecses_metaphor_2010, 140, italics is mine].


```{r metaphor-table-token-based}

# knitr::kable(slice_max(metaphor_salience_print, n = 20, order_by = Aggregate), caption = "Top-20 source domains of [anger]{.smallcaps} ranked-ordered by their degree of metaphorical salience (token-based)")

# knitr::kable(filter(metaphor_salience_print, Aggregate > 5), caption = "Source domains of [anger]{.smallcaps} based on the token-based, corpus approach (Aggregate values > 5%)")

knitr::kable(mutate(metaphor_salience_print, `Metaphorical source domains` = str_replace_all(`Metaphorical source domains`, "luminousity", "luminosity"),
                    `Metaphorical source domains` = str_replace_all(`Metaphorical source domains`, fixed("dimension"), "dimension/size"),
                    `Metaphorical source domains` = str_replace_all(`Metaphorical source domains`, fixed("subjugator"), "superior")), 
             caption = "Source domains of [anger]{.smallcaps} (Token-based, corpus approach)", row.names = FALSE)
```

### [Marah is a contained substance]{.smallcaps} {#contained-substance-tokenbased}

```{r mapping-containedfluid-token, message = FALSE, include = FALSE}
mapping_containedfluid_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "is substance in a container$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(_located\\-substance)"), "key", "no"),
         status = if_else(str_detect(MAPPING, "experiencer_container"), "key", status),
         status = if_else(str_detect(MAPPING, "(substance.fullness.level)"), "key", status),
         MP_GLOSS = str_replace_all(MP_GLOSS, "\\sbe channelled(?!\\*)", " *be channelled*"))
mapping_containedfluid_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_containedfluid_token)
mapping_containedfluid_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'is substance in a container$') %>% left_join(mapping_containedfluid_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_containedfluid_token_stats)
```

This metaphor is shared between the corpus and lexical approaches ([§\@ref(contained-substance-typebased)](#contained-substance-typebased)). It is the most salient one in the corpus dataset but sits at the rank `r get_metaphor_salience_rank(metaphor_typebased_salience, metaphor_regex = 'is substance in a container')` in the lexical dataset ([Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based)). Below are the mappings and the metaphorical expressions for the [contained substance]{.smallcaps} metaphor.

- expelling/releasing/channelling substance out (at others) &rarr; expressing anger (type=`r get_metaphor_mapping_n_lu(mapping_containedfluid_token_stats, 'releasing')`; token=`r get_metaphor_mapping_tokenfreq(mapping_containedfluid_token, 'releasing', lu_output = FALSE)`)

  <!-- - `r mapping_containedfluid_token %>% filter(status == 'no') %>% filter(str_detect(MAPPING, "expressing anger is expelling.releasing.channel+ing substance out \\(at others\\)")) %>% count(LU, LU_GLOSS) %>% mutate(lu_gloss_n = paste("*", str_replace_all(LU, "<\\/?w>", ""), "* '", LU_GLOSS, "' (", n, ")", sep = "")) %>% arrange(desc(n)) %>% pull(lu_gloss_n) %>% paste(collapse = "; ")` -->
  - `r get_lu_gloss_n_printed(mapping_containedfluid_token, 'releasing|expel+ing', lu_var = "MP", gloss_var = "MP_GLOSS")`
  
- the presence of object in a location &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_containedfluid_token_stats, 'presence of object')`; token=`r get_metaphor_mapping_tokenfreq(mapping_containedfluid_token, 'presence of object', lu_output = FALSE)`)

  <!-- - `r mapping_containedfluid_token %>% filter(status == 'no') %>% filter(MAPPING == "experiencing anger is the presence of object in a location") %>% count(LU, LU_GLOSS) %>% mutate(lu_gloss_n = paste("*", str_replace_all(LU, "<\\/?w>", ""), "* '", LU_GLOSS, "' (", n, ")", sep = "")) %>% arrange(desc(n)) %>% pull(lu_gloss_n) %>% paste(collapse = "; ")` -->

  - `r get_lu_gloss_n_printed(mapping_containedfluid_token, 'experiencing anger is the presence of object in a location', lu_var = "MP", gloss_var = "MP_GLOSS")`
  
- increased fullness of the substance in the container &rarr; increased intensity (type=`r get_metaphor_mapping_n_lu(mapping_containedfluid_token_stats, 'increased fullness')`; token=`r get_metaphor_mapping_tokenfreq(mapping_containedfluid_token, 'increased fullness', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_containedfluid_token, 'increased fullness')`
  
- impeding flowing liquid &rarr; regulating anger (type=`r get_metaphor_mapping_n_lu(mapping_containedfluid_token_stats, 'impeding flowing')`; token=`r get_metaphor_mapping_tokenfreq(mapping_containedfluid_token, 'impeding flowing', lu_output = FALSE)`)
  
  <!-- - `r mapping_containedfluid_token %>% filter(status == 'no') %>% filter(str_detect(MAPPING, "impeding flowing")) %>% count(LU, LU_GLOSS) %>% mutate(lu_gloss_n = paste("*", str_replace_all(LU, "<\\/?w>", ""), "* '", LU_GLOSS, "' (", n, ")", sep = "")) %>% arrange(desc(n)) %>% pull(lu_gloss_n) %>% paste(collapse = "; ")` -->
  
  - `r get_lu_gloss_n_printed(mapping_containedfluid_token, 'impeding flowing', lu_var = "MP", gloss_var = "MP_GLOSS")`

- source of the liquid &rarr; cause of anger (type=`r get_metaphor_mapping_n_lu(mapping_containedfluid_token_stats, '^cause of anger')`; token=`r get_metaphor_mapping_tokenfreq(mapping_containedfluid_token, 'cause of anger', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_containedfluid_token, '^cause of anger', lu_var = "MP", gloss_var = "MP_GLOSS")`

- substance becomes hard &rarr; intense anger (type=`r get_metaphor_mapping_n_lu(mapping_containedfluid_token_stats, 'hard$')`; token=`r get_metaphor_mapping_tokenfreq(mapping_containedfluid_token, 'hard$', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_containedfluid_token, 'hard$')`

It is obvious that the corpus dataset has much greater number of linguistic types and mappings compared to the lexical dataset. Despite this difference, the metaphor in both datasets equally focuses on the externalisation/expression of anger since the metaphorical expressions of the mapping for this aspect is the most productive.

### [Marah is a fierce, captive animal]{.smallcaps} {#animal-tokenbased}

```{r mapping-animal-token, message = FALSE, include = FALSE}
mapping_animal_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "animal$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "anger_restrained-entity"), "key", "no"),
         status = if_else(str_detect(MAPPING, "experiencer/other-emotion_restraining-entity"), "key", status))
mapping_animal_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_animal_token)
mapping_animal_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'animal$') %>% left_join(mapping_animal_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_animal_token_stats)
```


This metaphor is in the top-two salient metaphors across the two approaches but differs in its main meaning focus. In the lexical dataset ([§\@ref(animal-typebased)](#animal-typebased)), the predominant expressions highlight the aggressive behaviour of the animal (type=`r get_metaphor_mapping_n_lu(mapping_animal_stats, 'aggressive')`). In the corpus dataset, the expressions predominantly evoke mapping for restraining the animal (type=`r get_metaphor_mapping_n_lu(mapping_animal_token_stats, 'restraining')`; token=`r get_metaphor_mapping_tokenfreq(mapping_animal_token, 'restraining')`); this mapping is the least prominent in the lexical dataset (type=`r get_metaphor_mapping_n_lu(mapping_animal_stats, 'restraining')`).

- restrained, fierce animal &rarr; anger

- restraining entity &rarr; experiencer (or other states)

- restraining the animal &rarr;; regulating anger (type=`r get_metaphor_mapping_n_lu(mapping_animal_token_stats, 'restraining')`; token=`r get_metaphor_mapping_tokenfreq(mapping_animal_token, 'restraining', lu_output = FALSE)`)

  - `r str_replace_all(get_lu_gloss_n_printed(mapping_animal_token, 'restraining'), fixed("controlable"), "controllable")`

- unleashing a captive &rarr; expressing and losing control over anger (type=`r get_metaphor_mapping_n_lu(mapping_animal_token_stats, 'unleashing')`; token=`r get_metaphor_mapping_tokenfreq(mapping_animal_token, 'unleashing', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_animal_token, 'unleashing')`

- aggressive animal behaviour &rarr; aggressive anger behaviour  (type=`r get_metaphor_mapping_n_lu(mapping_animal_token_stats, 'aggressive anger')`; token=`r get_metaphor_mapping_tokenfreq(mapping_animal_token, 'aggressive anger', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_animal_token, 'aggressive anger')`

- being attacked by animal &rarr; experiencing effect of anger (type=`r get_metaphor_mapping_n_lu(mapping_animal_token_stats, 'attacked')`; token=`r get_metaphor_mapping_tokenfreq(mapping_animal_token, 'attacked', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_animal_token, 'attacked')`

The difference in the main meaning focus of the same metaphor between the two approaches indicate that combining the different methods enriches the nuance in the highlighted aspect of a metaphor.

### [Marah is fire]{.smallcaps} {#fire-tokenbased}

```{r mapping-fire-token, message = FALSE, include = FALSE}
mapping_fire_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "fire$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "anger_fire|angry.person_burning.object|anger.level_fire.intensity"), "key", "no"),
         status = if_else(str_detect(MAPPING, "anger_fire") & str_detect(LU, "^<w>api<\\/w>$"), "no", status),
         MAPPING = if_else(str_detect(MAPPING, "anger_fire") & status == "no", "anger is fire", MAPPING))
mapping_fire_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_fire_token)
mapping_fire_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'fire$') %>% left_join(mapping_fire_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_fire_token_stats)
```

Similar to the [fierce, captive animal]{.smallcaps} metaphor, the [anger is fire]{.smallcaps} metaphor is prominent in Indonesian in the lexical ([Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based)) and corpus-based datasets ([Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based)).

- fire &rarr; anger (type=`r get_metaphor_mapping_n_lu(mapping_fire_token_stats, 'anger is fire$')`; token=`r get_metaphor_mapping_tokenfreq(mapping_fire_token, 'anger is fire$', lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_fire_token, 'anger is fire$')`)

- burning/burnt object &rarr; angry person

- fire intensity &rarr; anger level

- the increased degree of fire &rarr; the increased degree of anger intensity (type=`r get_metaphor_mapping_n_lu(mapping_fire_token_stats, 'highest')`; token=`r get_metaphor_mapping_tokenfreq(mapping_fire_token, 'highest', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_fire_token, 'highest')`

- igniting fire &rarr; causing anger (type=`r get_metaphor_mapping_n_lu(mapping_fire_token_stats, 'causing')`; token=`r get_metaphor_mapping_tokenfreq(mapping_fire_token, 'causing', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_fire_token, 'causing')`

- potential open fire &rarr; latent intensity (type=`r get_metaphor_mapping_n_lu(mapping_fire_token_stats, 'latent')`; token=`r get_metaphor_mapping_tokenfreq(mapping_fire_token, 'latent', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_fire_token, 'latent')`

- fire going out &rarr; ceasing anger (type=`r get_metaphor_mapping_n_lu(mapping_fire_token_stats, 'ceasing')`; token=`r get_metaphor_mapping_tokenfreq(mapping_fire_token, 'ceasing', lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_fire_token, 'ceasing')`

The [fire]{.smallcaps} metaphor in the two approaches equally focuses on the intensity aspect, particularly the increased intensity of the fire given its greatest number of metaphorical expressions compared to the other mappings. However, if the token frequency of the mapping is considered, the [fire]{.smallcaps} metaphor predominantly captures the cause of anger (token=`r get_metaphor_mapping_tokenfreq(mapping_fire_token, 'causing', lu_output = FALSE)`).

### [Marah is weapon]{.smallcaps} {#weapon-tokenbased}

```{r mapping-weapon-token, message = FALSE, include = FALSE}
mapping_weapon_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "weapon$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "anger_weapon"), "key", "no"))
mapping_weapon_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_weapon_token)
mapping_weapon_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'weapon$') %>% left_join(mapping_weapon_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_weapon_token_stats)
```

The [weapon]{.smallcaps} metaphor is only identified in the corpus approach. The main meaning focus is directing/targeting anger to the wrongdoer or other people. In terms of the token frequency of the mapping, the causation mapping is the most predominant. Despite the highest token of the causation mapping, its linguistic manifestation is less varied compared to the targeting anger mapping, having diverse types despite its lower token frequency.

- the wrongdoer or other people &rarr; target of expression of anger (type=`r get_metaphor_mapping_n_lu(mapping_weapon_token_stats, "target of expression of anger")`; token=`r get_metaphor_mapping_tokenfreq(mapping_weapon_token, "target of expression of anger", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_weapon_token, "target of expression of anger")`

- pulling the trigger &rarr; causing anger (type=`r get_metaphor_mapping_n_lu(mapping_weapon_token_stats, "pulling the trigger")`; token=`r get_metaphor_mapping_tokenfreq(mapping_weapon_token, "pulling the trigger", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_weapon_token, "pulling the trigger")`
  
- punishing with the weapon &rarr; retribution/avenge to wrongdoing (type=`r get_metaphor_mapping_n_lu(mapping_weapon_token_stats, "punishing|avenging")`; token=`r get_metaphor_mapping_tokenfreq(mapping_weapon_token, "punishing|avenging", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_weapon_token, "punishing|avenging")`
  
- weapon &rarr; anger (type=`r get_metaphor_mapping_n_lu(mapping_weapon_token_stats, "^anger is weapon$")`; token=`r get_metaphor_mapping_tokenfreq(mapping_weapon_token, "^anger is weapon$", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_weapon_token, "^anger is weapon$")`
  
- redirecting weapon to constructive goal &rarr; channelling anger (type=`r get_metaphor_mapping_n_lu(mapping_weapon_token_stats, "constructive")`; token=`r get_metaphor_mapping_tokenfreq(mapping_weapon_token, "constructive", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_weapon_token, "constructive")`

The last mapping refers to a non-prototypical model of anger, namely "constructive use" of anger [@holland_cognitive_1987, 215].

### [Marah is an adversary/opponent]{.smallcaps} {#adversary-tokenbased}

```{r mapping-adversary-token, message = FALSE, include = FALSE}
mapping_adversary_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "adversary$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "fighter[12]"), "key", "no"),
         MP = str_replace_all(MP, "\\bnnode\\b", "node"))
mapping_adversary_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_adversary_token)
mapping_adversary_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'adversary$') %>% left_join(mapping_adversary_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_adversary_token_stats)

opponent_frame_token <- count(filter(marah, str_detect(CM_BROADER, "adversary$")), SFRAME, sort = TRUE)
opponent_frame_type <- arrange(summarise(group_by(filter(marah, str_detect(CM_BROADER, "adversary$")), SFRAME), n_lu = n_distinct(LU)), desc(n_lu))
opponent_victim_token <- count(filter(marah, str_detect(CM_BROADER, "adversary$")), NOTES, sort = TRUE) %>% 
  mutate(NOTES = replace(NOTES, is.na(NOTES), "self as victim of anger"))
```

The [adversary/opponent]{.smallcaps} metaphor is absent in the lexical dataset. The metaphorical expressions evoke the [physical combat]{.smallcaps} frame (type=`r pull(slice_max(opponent_frame_type, order_by = n_lu), n_lu)`; token=`r pull(slice_max(opponent_frame_token, order_by = n), n)`). However, closer investigation on the context of the data reveal that the hostile encounter is not always between the anger-adversary and the self/experiencer. Construing anger as an adversary includes being the adversary for external, non-self entity (predominantly victim of a directed anger) (`r sum(pull(filter(opponent_victim_token, str_detect(NOTES, 'external|state')), n))` instances of the total `r sum(opponent_victim_token$n)` tokens of this metaphor). There are `r unlist(tally(filter(opponent_victim_token, str_detect(NOTES, 'external|state|^anger', negate = TRUE)), n))` instances where the victim is the self. This variation in the combatant of anger indicates that the adversarial nature of anger can not only endanger the experiencers but also social entities around them. 

- giving in/losing &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, "losing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_adversary_token, "losing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_adversary_token, "losing")`

- combating opponent &rarr; attempt to control anger (e.g., fighting, confronting) (type=`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, "combating")`; token=`r get_metaphor_mapping_tokenfreq(mapping_adversary_token, "combating", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_adversary_token, "combating")`

- being protected from the opponent &rarr; evading anger (type=`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, "evading")`; token=`r get_metaphor_mapping_tokenfreq(mapping_adversary_token, "evading", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_adversary_token, "evading")`

- winning over the opponent &rarr; managing anger (type=`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, "winning")`; token=`r get_metaphor_mapping_tokenfreq(mapping_adversary_token, "winning", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_adversary_token, "winning")`

- potential danger of the opponent &rarr; potential effect of anger (type=`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, "^potential")`; token=`r get_metaphor_mapping_tokenfreq(mapping_adversary_token, "^potential", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_adversary_token, "^potential")`
  
The main meaning foci of the metaphor are the mappings [managing anger is winning over the opponent]{.smallcaps} and [experiencing anger is giving in/losing to anger]{.smallcaps}. Both share the same number of types though the former is lexically more varied given its token frequency than the latter. That is, `r get_metaphor_mapping_ttr(mapping_adversary_token_stats, 'winning')`% of the `r get_metaphor_mapping_tokenfreq(mapping_adversary_token, 'winning')` tokens for [managing anger is winning over the opponent]{.smallcaps} are expressed by different types [`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, 'winning')`]. Meanwhile, only `r get_metaphor_mapping_ttr(mapping_adversary_token_stats, 'losing')`% of the `r get_metaphor_mapping_tokenfreq(mapping_adversary_token, 'losing')` tokens for [experiencing anger is giving in/losing to anger]{.smallcaps} are expressed by different types [`r get_metaphor_mapping_n_lu(mapping_adversary_token_stats, 'losing')`].

### [Intensity is verticality/height]{.smallcaps} {#verticality-tokenbased}

```{r mapping-verticality-token, message = FALSE, include = FALSE}
mapping_verticality_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "verticality$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(_vertical.entity|_verticality.scale)"), "key", "no"))
mapping_verticality_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_verticality_token)
mapping_verticality_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'verticality$') %>% left_join(mapping_verticality_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_verticality_token_stats)
verticality_frame_stats <- happyr::get_frames("verticality$", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

The [intensity is verticality/height]{.smallcaps} is shared between the lexical (rank=`r get_metaphor_salience_rank(metaphor_typebased_salience, 'verticality$')`) and corpus approaches (rank=`r get_metaphor_salience_rank(metaphor_salience, 'verticality$')`) in the top-10 ranking. This metaphor is based on LUs referring to several MN frame families, namely [being at a high location]{.smallcaps}, [(caused) upward/downward motion]{.smallcaps}, and [scale]{.smallcaps}, which is incorporated as a frame role in the [verticality]{.smallcaps} frame. The predominant frame in the metaphor is the [(caused) upward motion]{.smallcaps} frame (type=`r sum(filter(verticality_frame_stats, str_detect(frames, 'upward'))$type)`; token=`r sum(filter(verticality_frame_stats, str_detect(frames, 'upward'))$n)`). The [verticality]{.smallcaps} metaphor has `r get_salience_stats(metaphor_salience, "n_type", "verticality$")` types of metaphorical expression evoking `r numbers2words(get_salience_stats(metaphor_salience, "n_mapping", "verticality$"))` mappings. Together, all these frames and their LUs highlight the dynamics of the intensity as movement or position of the [anger]{.smallcaps}-entity along the vertical scale.

- vertical entity &rarr; anger

- entity being at/going up to/reaching the highest end of the scale &rarr; highest intensity (type=`r get_metaphor_mapping_n_lu(mapping_verticality_token_stats, "^highest")`; token=`r get_metaphor_mapping_tokenfreq(mapping_verticality_token, "^highest", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_verticality_token, "^highest")`

- (caused) upward motion &rarr; increasing intensity (type=`r get_metaphor_mapping_n_lu(mapping_verticality_token_stats, "caused")`; token=`r get_metaphor_mapping_tokenfreq(mapping_verticality_token, "caused", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_verticality_token, "caused")`

- low on the scale &rarr; reduced intensity of anger (type=`r get_metaphor_mapping_n_lu(mapping_verticality_token_stats, "reduced")`; token=`r get_metaphor_mapping_tokenfreq(mapping_verticality_token, "reduced", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_verticality_token, "reduced")`

- verticality scale &rarr; intensity of anger (type=`r get_metaphor_mapping_n_lu(mapping_verticality_token_stats, "is verticality scale$")`; token=`r get_metaphor_mapping_tokenfreq(mapping_verticality_token, "is verticality scale$", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_verticality_token, "is verticality scale$")`)

- balancing the scale &rarr; maintaining anger (type=`r get_metaphor_mapping_n_lu(mapping_verticality_token_stats, "balancing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_verticality_token, "balancing", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_verticality_token, "balancing")`)

The two predominant mappings refer to (i) an entity reaching/being at the peak of the scale and (ii) the lifted position of the entity along the scale, without implying whether such upward motion of the anger-entity would reach the peak.

### [Marah is confinement/impediment]{.smallcaps} {#confinement-tokenbased}

```{r mapping-confinement-token, message = FALSE, include = FALSE}
mapping_confinement_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "impediment$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_impediment|angry.person_confined.entity)"), "key", "no"))
mapping_confinement_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_confinement_token)
mapping_confinement_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'impediment$') %>% left_join(mapping_confinement_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_confinement_token_stats)
confinement_frame_stats <- happyr::get_frames("impediment$", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

This metaphor is only attested in the corpus dataset, realising `r get_salience_stats(metaphor_salience, "n_type", "impediment")` types and `r numbers2words(get_salience_stats(metaphor_salience, "n_mapping", "impediment"))` mappings. They are shown below.

- `r str_extract(get_key_mappings(mapping_confinement_token), "^.+(?=\\___)")`

- `r str_extract(get_key_mappings(mapping_confinement_token), "(?<=\\___\\s).+$")`

- being confined &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_confinement_token_stats, "being confined$")`; token=`r get_metaphor_mapping_tokenfreq(mapping_confinement_token, "being confined$", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_confinement_token, "being confined$")`

- being/moving away from a confining/impeding location &rarr; controlling self not to be angry (type=`r get_metaphor_mapping_n_lu(mapping_confinement_token_stats, "moving away")`; token=`r get_metaphor_mapping_tokenfreq(mapping_confinement_token, "moving away", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_confinement_token, "moving away")`

- escaping/getting over the confinement/impediment &rarr; ceasing anger (type=`r get_metaphor_mapping_n_lu(mapping_confinement_token_stats, "ceasing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_confinement_token, "ceasing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_confinement_token, "ceasing")`

- negatively affected motion &rarr; effect of anger (type=`r get_metaphor_mapping_n_lu(mapping_confinement_token_stats, "^effect of anger")`; token=`r get_metaphor_mapping_tokenfreq(mapping_confinement_token, "^effect of anger", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_confinement_token, "^effect of anger")`

- sunken confinement &rarr; anger (type=`r get_metaphor_mapping_n_lu(mapping_confinement_token_stats, "sunken")`; token=`r get_metaphor_mapping_tokenfreq(mapping_confinement_token, "sunken", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_confinement_token, "sunken")`)
  
The LUs of the metaphor evoke a variety of frames. Some of them are within the [motion impediment]{.smallcaps} frame family, namely [confinement]{.smallcaps} and [barrier]{.smallcaps} (sub-cases of the [motion-affecting properties of the landscape]{.smallcaps} frame), and [restraints]{.smallcaps} (a sub-case of the [motion-affecting objects]{.smallcaps} frame). Another frame is [hiding]{.smallcaps}, especially the LUs referring to enclosing/encompassing the experiencer (that is generically related to the [containment/bounded-region]{.smallcaps} frame). Several other tokens of LUs evoke the [motion with respect to an undesirable entity]{.smallcaps} frame. Semantically, the metaphor highlights the undesirability and difficulty of being in a state of anger.

### [Marah is a pressurised substance in a container]{.smallcaps} {#pressurised-tokenbased}

```{r mapping-pressurised-token, message = FALSE, include = FALSE}
mapping_pressurised_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "pressurised substance")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_pressurised.substance|angry.person_container)"), "key", "no"))
mapping_pressurised_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_pressurised_token)
mapping_pressurised_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'pressurised substance') %>% left_join(mapping_pressurised_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_pressurised_token_stats)
pressurised_frame_stats <- happyr::get_frames("pressurised substance", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

This is another metaphor attested from the two methods (rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'pressurised')` in the lexical approach and rank `r get_metaphor_salience_rank(metaphor_salience, 'pressurised')` in the corpus approach). The overall theme of this metaphor is intensity and control. The mappings and metaphorical expressions are shown below.

- `r str_extract(get_key_mappings(mapping_pressurised_token), "^.+(?=\\___)")`

- `r str_extract(get_key_mappings(mapping_pressurised_token), "(?<=\\___\\s).+$")`

- (causing) explosion of container &rarr; (intentional) loss of control over anger (type=`r get_metaphor_mapping_n_lu(mapping_pressurised_token_stats, "explosion")`; token=`r get_metaphor_mapping_tokenfreq(mapping_pressurised_token, "explosion", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_pressurised_token, "explosion")`

- suppressing the contents from going out &rarr; controlling anger (type=`r get_metaphor_mapping_n_lu(mapping_pressurised_token_stats, "suppressing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_pressurised_token, "suppressing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_pressurised_token, "suppressing")`

- increased pressure of the substance to the container &rarr; increased intensity (type=`r get_metaphor_mapping_n_lu(mapping_pressurised_token_stats, "increased pressure")`; token=`r get_metaphor_mapping_tokenfreq(mapping_pressurised_token, "increased pressure", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_pressurised_token, "increased pressure")`

One interesting difference between the [pressurised substance in a container]{.smallcaps} metaphor from the two datasets is that in the lexical approach, the main meaning focus is on the (caused) increased pressure and substance. In contrast, the corpus dataset focuses on the loss of control, which is conceptualised as the explosion of the pressurised substance (see the greatest number of types and tokens for the explosion mapping). Moreover, the attempt at control aspect is not attested in the lexical approach but only in the corpus approach (the fourth mapping).

### [Marah is possession]{.smallcaps} {#possession-tokenbased}

```{r mapping-possession-token, message = FALSE, include = FALSE}
mapping_possession_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "possession")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_possession_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_possession_token)
mapping_possession_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'possession') %>% left_join(mapping_possession_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_possession_token_stats)
possession_frame_stats <- happyr::get_frames("possession", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP") %>% mutate(frame_category = if_else(str_detect(frames, "(forced|propulsion)"), "affected motion", "possession"))
possession_framefamily_stats <- possession_frame_stats %>% group_by(frame_category) %>% summarise(n = sum(n), type = sum(type), .groups = "drop") %>% mutate(perc_n = round(n/sum(n) * 100), perc_type = round(type/sum(type) * 100))
```

This metaphor is a sub-case of the generic [object event-structure]{.smallcaps} metaphor, namely [attributed states are possessed objects]{.smallcaps} [@lakoff_philosophy_1999, 195-196, 198]. It realises `r numbers2words(get_salience_stats(metaphor_salience, 'n_type', 'possession'))` types (`r get_salience_stats(metaphor_salience, 'n_token', 'possession')` tokens) manifesting `r numbers2words(get_salience_stats(metaphor_salience, 'n_mapping', 'possession'))` mappings. The key mappings in the metaphor are (i) anger is the possession and (ii) angry person (or non-self-entity, such as utterances, events, or even God) is the possessor or carrier of the object. The entailed mappings from these two are as follows:

- having/keeping/carrying possession &rarr; feeling/experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_possession_token_stats, "keeping")`; token=`r get_metaphor_mapping_tokenfreq(mapping_possession_token, "keeping", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_possession_token, "keeping")`

- obtaining an object &rarr; being scolded (type=`r get_metaphor_mapping_n_lu(mapping_possession_token_stats, "scolded")`; token=`r get_metaphor_mapping_tokenfreq(mapping_possession_token, "scolded", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_possession_token, "scolded")`

- giving object &rarr; expressing anger to others (type=`r get_metaphor_mapping_n_lu(mapping_possession_token_stats, "expressing anger to others")`; token=`r get_metaphor_mapping_tokenfreq(mapping_possession_token, "expressing anger to others", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_possession_token, "expressing anger to others")`

- losing/giving away possession &rarr; not experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_possession_token_stats, "not experiencing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_possession_token, "not experiencing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_possession_token, "not experiencing")`

Most of the metaphorical expressions (i.e., `r pull(filter(possession_framefamily_stats, frame_category == "possession"), perc_n)`% of the metaphor token) evoke the [possession]{.smallcaps} frame family (type=`r pull(filter(possession_framefamily_stats, frame_category == "possession"), type)`; token=`r pull(filter(possession_framefamily_stats, frame_category == "possession"), n)`), namely [possession]{.smallcaps} itself and [transfer scenario]{.smallcaps}, the roles of which also make use of the [gain possession]{.smallcaps}, [lose possession]{.smallcaps}, and [giving]{.smallcaps}. The remaining tokens are LUs evoking frames within the [affected motion]{.smallcaps} family such as [bringing]{.smallcaps} (e.g., *__membawa__ [marah]{.smallcaps}* '_carry_ anger') and [object propulsion]{.smallcaps} (e.g., *__membuang__ [marah]{.smallcaps}* '_throw away_ anger').

The inference from keeping the possession refers to the experience of anger, and when the possessor/self loses the anger-possession, the self no longer feels angry. The sentential context of the pattern evoking the gaining and giving the possession is not necessarily causing the recipient to become angry akin to the giver. Instead, gaining/getting or being given anger from the possessor means that the recipient becomes the target of the angry person presumably because of the wrongdoing of the recipient/target of transfer (i.e., the recipient becomes the target of scolding, which may or may not lead to the recipient becoming angry). This inference is different from when the mapping is used in more positive emotions, such as [happiness]{.smallcaps}, where giving or getting happiness suggests causing someone to be happy [@stefanowitsch_happiness_2004; @rajeg_metaphorical_2019].


### [Intensity is quantity]{.smallcaps} {#quantity-tokenbased}

```{r mapping-quantity-token, message = FALSE, include = FALSE}
mapping_quantity_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "quantity$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_entity|intensity_entity.amount)"), "key", "no"),
         status = if_else(str_detect(MAPPING, "(intensity_entity.amount)") &
                            str_detect(LU, "(beberapa|sebagian dari)"), "no", status))
mapping_quantity_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_quantity_token) %>% 
  mutate(aspects = if_else(str_detect(MAPPING, "(more amount)"), "more", "less"),
         aspects = replace(aspects, str_detect(MAPPING, "entity.amount$"), "neutral"))
mapping_quantity_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'quantity$') %>% left_join(mapping_quantity_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_quantity_token_stats)
quantity_frame_stats <- happyr::get_frames("quantity$", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
mapping_quantity_more_or_less <- mapping_quantity_token_stats %>% filter(aspects != "neutral") %>% group_by(aspects) %>% summarise(n_lu = sum(n_lu), n_token = sum(n_token))
```

This metaphor is based on the [quantity]{.smallcaps} frame and only attested in the corpus dataset. Lakoff & Kövecses [-@holland_cognitive_1987, 200] propose that anger has a quantitative aspect reflecting its intensity, which in turn is conceptualised as an amount. The metaphor is manifested by `r get_salience_stats(metaphor_salience, 'n_type', 'quantity$')` metaphorical expressions representing `r numbers2words(get_salience_stats(metaphor_salience, 'n_mapping', 'quantity$'))` mappings. 

- more amount of entity &rarr; intensified anger (type=`r get_metaphor_mapping_n_lu(mapping_quantity_token_stats, "intensified")`; token=`r get_metaphor_mapping_tokenfreq(mapping_quantity_token, "intensified", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_quantity_token, "intensified")`

- less amount of entity &rarr; less intense anger (type=`r get_metaphor_mapping_n_lu(mapping_quantity_token_stats, "less intense")`; token=`r get_metaphor_mapping_tokenfreq(mapping_quantity_token, "less intense", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_quantity_token, "less intense")`

- lessening the amount of entity &rarr; de-intensifying anger (type=`r get_metaphor_mapping_n_lu(mapping_quantity_token_stats, "lessening")`; token=`r get_metaphor_mapping_tokenfreq(mapping_quantity_token, "lessening", lu_output = FALSE)`)

  - `r str_replace(get_lu_gloss_n_printed(mapping_quantity_token, "lessening"), fixed("/*decreased*"), "")`
  
- adding more amount of entity &rarr; intensifying anger (type=`r get_metaphor_mapping_n_lu(mapping_quantity_token_stats, "adding")`; token=`r get_metaphor_mapping_tokenfreq(mapping_quantity_token, "adding", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_quantity_token, "adding")`

- amount of entity &rarr; intensity of anger (type=`r get_metaphor_mapping_n_lu(mapping_quantity_token_stats, "entity.amount")`; token=`r get_metaphor_mapping_tokenfreq(mapping_quantity_token, "entity.amount", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_quantity_token, "entity.amount")`

There are two perspectives in this metaphor, the increased or the lessening amount of the entity. The former highlights the intensification of anger while the latter is for the reduced intensity. There are more types and tokens that reflect the intensification of anger (type=`r pull(filter(mapping_quantity_more_or_less, aspects=='more'), n_lu)`; token=`r pull(filter(mapping_quantity_more_or_less, aspects=='more'), n_token)`; see the first and the fourth mappings) than those for the reduced intensity (type=`r pull(filter(mapping_quantity_more_or_less, aspects=='less'), n_lu)`; token=`r pull(filter(mapping_quantity_more_or_less, aspects=='less'), n_token)`; the second and the third mappings).

### [Marah is heated fluid in a container]{.smallcaps} {#hotfluid-tokenbased}

```{r mapping-hotfluid-token, message = FALSE, include = FALSE}
mapping_hotfluid_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "heated fluid")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "(anger_\\(heated\\-\\)fluid|anger\\-level_fluid\\-heat\\-level|experiencer_container\\-with\\-hot\\-fluid)"), "key", "no"))
mapping_hotfluid_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_hotfluid_token)
mapping_hotfluid_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'heated fluid') %>% left_join(mapping_hotfluid_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_hotfluid_token_stats)
hotfluid_frame_stats <- happyr::get_frames("heated fluid", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

This central metaphor is shared between the two approaches (rank=`r get_metaphor_salience_rank(metaphor_typebased_salience, 'heated')` in the lexical approach, and rank=`r get_metaphor_salience_rank(metaphor_salience, 'heated')` in the corpus approach). The metaphor has higher tokens, types, and mappings in the corpus dataset.

`r str_replace(paste(paste("- ", unlist(str_split(get_key_mappings(mapping_hotfluid_token), "___ ")), "\n", sep = ""), collapse = ""), fixed("(heated )"), "")`

- heated fluid in container &rarr; high intensity of anger (type=`r get_metaphor_mapping_n_lu(mapping_hotfluid_token_stats, "high intensity")`; token=`r get_metaphor_mapping_tokenfreq(mapping_hotfluid_token, "high intensity", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_hotfluid_token, "high intensity")`

- boiling over the fluid to overflow &rarr; intensifying and expressing anger (type=`r get_metaphor_mapping_n_lu(mapping_hotfluid_token_stats, "boiling over the fluid to overflow")`; token=`r get_metaphor_mapping_tokenfreq(mapping_hotfluid_token, "boiling over the fluid to overflow", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_hotfluid_token, "boiling over the fluid to overflow")`

- evaporating liquid &rarr; reduced intensity of anger (type=`r get_metaphor_mapping_n_lu(mapping_hotfluid_token_stats, "evaporating")`; token=`r get_metaphor_mapping_tokenfreq(mapping_hotfluid_token, "evaporating", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_hotfluid_token, "evaporating")`

In both datasets, the [heated fluid]{.smallcaps} metaphor focuses on a heightened intensity of anger. One minor difference is, in the corpus dataset only, there is an attestation of a mapping for the ceasing of anger; this is conceptualised as an evaporating fluid (the last mapping above).

### [Marah is a submerged entity]{.smallcaps} {#submerged-tokenbased}

```{r mapping-submerged-token, message = FALSE, include = FALSE}
mapping_submerged_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "submerged")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_submerged_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_submerged_token)
mapping_submerged_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'submerged') %>% left_join(mapping_submerged_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_submerged_token_stats)
submerged_frame_stats <- happyr::get_frames("submerged", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

On a more general level, the metaphor could be viewed as a sub-case of the generic metaphor [existence is being here]{.smallcaps}, which has a universal perceptual basis (i.e., we know that something exists when it is physically presence and can be seen). The [submerged entity]{.smallcaps} metaphor captures the way anger exists and it is not attested in the lexical approach. The key mapping is `r get_key_mappings(mapping_submerged_token)`. From this mapping, two entailed mappings are postulated based on the semantics of the source frame LUs.

- the presence/emergence/protruding of the entity &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_submerged_token_stats, "protruding")`; token=`r get_metaphor_mapping_tokenfreq(mapping_submerged_token, "protruding", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_submerged_token, "protruding")`

- forcing the entity to emerge &rarr; causing anger (type=`r get_metaphor_mapping_n_lu(mapping_submerged_token_stats, "to emerge")`; token=`r get_metaphor_mapping_tokenfreq(mapping_submerged_token, "to emerge", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_submerged_token, "to emerge")`

The metaphor differs slightly from the [(un)veiled object]{.smallcaps} metaphor ([§\@ref(unveiled-tokenbased)](#unveiled-tokenbased)) whereby anger already exists and highlights the issue of whether the experiencer is keen to unveil the anger to others or not [cf. @rajeg_metaphorical_2019, 158]. Compared to the other metaphors discussed so far, [anger is submerged entity]{.smallcaps} is rather conventionalised in its linguistic realisation. It is manifested by small number of types [`r get_salience_stats(metaphor_salience, 'n_type', 'submerged')`] making up the high token of the metaphor [`r get_salience_stats(metaphor_salience, 'n_token', 'submerged')`]; that is, only `r get_meta_freq_profiles('submerged', type_per_token_lu, df = ttr_metaphor_tokenbased, metaphor_var = "cm_broader")[[2]]`% of the metaphor's tokens are of different types that are repeated many times.

### [Marah is a sleeping organism]{.smallcaps} {#sleeping-tokenbased}

```{r mapping-sleeping-token, message = FALSE, include = FALSE}
mapping_sleeping_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "sleeping")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_sleeping_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_sleeping_token)
mapping_sleeping_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'sleeping') %>% left_join(mapping_sleeping_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_sleeping_token_stats)
sleeping_frame_stats <- happyr::get_frames("sleeping", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

The [sleeping organism]{.smallcaps} metaphor is also attested in the lexical approach though at the lower rank (i.e., `r get_metaphor_salience_rank(metaphor_typebased_salience, "sleeping")`) ([§\@ref(awakening-typebased)](#awakening-typebased)). It is based on the [awakening]{.smallcaps} and the [caused upward motion]{.smallcaps} frames. The utilised aspect of the source frame is causing inactive/calm organisms to become awake/active/alert/aroused. The manner of awakening/alerting can be fishing for the organism via bait. In the target frame, this is mapped onto the cause of anger. This metaphor could be a pre-condition for the danger and fierceness of the organism that needs to be restrained as in the [fierce, captive animal]{.smallcaps} metaphor ([§\@ref(animal-tokenbased)](#animal-tokenbased)). 

`r paste(paste("- ", unlist(str_split(get_key_mappings(mapping_sleeping_token), "____ ")), "\n", sep = ""), collapse = "")`

- fishing/awakening the calm entity &rarr; causing anger (type=`r get_metaphor_mapping_n_lu(mapping_sleeping_token_stats, "awakening")`; token=`r get_metaphor_mapping_tokenfreq(mapping_sleeping_token, "awakening", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_sleeping_token, "awakening")`

- awaken/alerted entity &rarr; anger near the limit (type=`r get_metaphor_mapping_n_lu(mapping_sleeping_token_stats, "alerted")`; token=`r get_metaphor_mapping_tokenfreq(mapping_sleeping_token, "alerted", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_sleeping_token, "alerted")`

Like the [submerged entity]{.smallcaps}, the [sleeping organism]{.smallcaps} highlights the way anger is brought about. The conventionalised lexical unit of this metaphor is based on the root *pancing* 'lure a fish using a bite; to fish for sth.' and *bangkit* 'wake/get up'.

### [Marah is a natural force]{.smallcaps} {#naturalforce-tokenbased}

```{r mapping-naturalforce-token, message = FALSE, include = FALSE}
mapping_naturalforce_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "natural force")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_naturalforce_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_naturalforce_token)
mapping_naturalforce_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'natural force') %>% left_join(mapping_naturalforce_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_naturalforce_token_stats)
naturalforce_frame_stats <- happyr::get_frames("natural force", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

The metaphor occupies nearly similar rank in the two datasets (i.e., `r get_metaphor_salience_rank(metaphor_typebased_salience, "natural")` in the lexical approach and `r get_metaphor_salience_rank(metaphor_salience, "natural")` in the corpus approach). This metaphor predominantly consists of LUs evoking the [harm]{.smallcaps} frame family, particularly sub-cases of the [destroying]{.smallcaps} frame, namely [natural disaster]{.smallcaps} (type=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'natural disaster')), type)`; token=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'natural disaster')), n)`) and [destructive natural process]{.smallcaps} (type=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'destructive')), type)`; token=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'destructive')), n)`). The other minor frames are [non-translational motion]{.smallcaps} (type=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'translational')), type)`; token=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'translational')), n)`) and [weather]{.smallcaps} (type=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'weather')), type)`; token=`r pull(filter(naturalforce_frame_stats, str_detect(frames, 'weather')), n)`). The metaphorical expressions from these frames evoke `r happyr::numbers2words(get_salience_stats(metaphor_salience, "n_mapping", 'natural'))` mappings.

`r str_replace(paste(paste("- ", unlist(str_split(get_key_mappings(mapping_naturalforce_token), "___ ")), "\n", sep = ""), collapse = ""), fixed("(destructive )"), "(destructive) ")`

- amount of destructive force &rarr; degree of effect (type=`r get_metaphor_mapping_n_lu(mapping_naturalforce_token_stats, "degree of effect")`; token=`r get_metaphor_mapping_tokenfreq(mapping_naturalforce_token, "degree of effect", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_naturalforce_token, "degree of effect")`

- being destroyed &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_naturalforce_token_stats, "being destroyed")`; token=`r get_metaphor_mapping_tokenfreq(mapping_naturalforce_token, "being destroyed", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_naturalforce_token, "being destroyed")`

- reducing the amount of the disaster &rarr; causing anger to cease (type=`r get_metaphor_mapping_n_lu(mapping_naturalforce_token_stats, "reducing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_naturalforce_token, "reducing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_naturalforce_token, "reducing")`

- ceasing amount of force &rarr; ceasing anger (type=`r get_metaphor_mapping_n_lu(mapping_naturalforce_token_stats, "ceasing anger")`; token=`r get_metaphor_mapping_tokenfreq(mapping_naturalforce_token, "ceasing anger", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_naturalforce_token, "ceasing anger")`
  
The [natural force]{.smallcaps} metaphor focuses on the intense effect of the natural force, as evident by the productivity of this mapping. This is a more specific way of construing the harmful effect of anger than the other [harm]{.smallcaps}-related metaphor based on different frames within the same family ([§\@ref(harm-tokenbased)](#harm-tokenbased)).

### [Marah is a superior]{.smallcaps} {#subjugator-tokenbased}

```{r mapping-subjugator-token, message = FALSE, include = FALSE}
mapping_subjugator_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "subjugator")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_subjugator_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_subjugator_token)
mapping_subjugator_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'subjugator') %>% left_join(mapping_subjugator_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_subjugator_token_stats)
subjugator_frame_stats <- happyr::get_frames("subjugator", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

The [superior/subjugator]{.smallcaps} metaphor is only attested in the corpus data and based on the mapping of anger onto the Subjugator role in the [servitude]{.smallcaps} frame [cf. @kovecses_metaphor_2000, 21, 70]. Kövecses [-@kovecses_metaphor_2000, 71] argues that this metaphor involves asymmetric control-relation between the social inferior (the Subjugated role in the [servitude]{.smallcaps} frame), namely the self (or its associates), and the social superior (Subjugator), namely the emotion. The authority element is incorporated into the [servitude]{.smallcaps} frame from the structure of the [authority]{.smallcaps} frame, which is the sub-case of the [control]{.smallcaps} frame. There are `r numbers2words(get_salience_stats(metaphor_salience, col_names = 'n_mapping', 'subjugator'))` mappings in this metaphor, evoked by `r numbers2words(get_salience_stats(metaphor_salience, col_names = 'n_type', 'subjugator'))` types (`r get_salience_stats(metaphor_salience, col_names = 'n_token', 'subjugator')` tokens).

`r paste(paste("- ", unlist(str_split(get_key_mappings(mapping_subjugator_token), "___ ")), "\n", sep = ""), collapse = "") %>% str_replace(fixed("subjugator"), "superior/subjugator")`

- being subjugated &rarr; effect of and losing control over anger (type=`r get_metaphor_mapping_n_lu(mapping_subjugator_token_stats, "being subjugated")`; token=`r get_metaphor_mapping_tokenfreq(mapping_subjugator_token, "being subjugated", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_subjugator_token, "being subjugated")`

- exercising domination over the subjugator &rarr; managing anger (type=`r get_metaphor_mapping_n_lu(mapping_subjugator_token_stats, "domination")`; token=`r get_metaphor_mapping_tokenfreq(mapping_subjugator_token, "domination", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_subjugator_token, "domination")`

- inferior object of anger &rarr; angry person (type=`r get_metaphor_mapping_n_lu(mapping_subjugator_token_stats, "inferior object")`; token=`r get_metaphor_mapping_tokenfreq(mapping_subjugator_token, "inferior object", lu_output = FALSE)`) (`r str_replace(get_lu_gloss_n_printed(mapping_subjugator_token, "inferior object"), fixed("(dari)"), "dari")`)

The metaphor focuses on the emotional effect of anger (e.g., being irrational). It is construed via the degree of control and power that the authority has over the subjugated. This aspect is manifested as the metaphor's predominant mapping. There is only a small number of types that evoke the self's attempt to manage anger (i.e., exercise/compromising control).

### [Marah is a biological entity]{.smallcaps} {#biologicalentity-tokenbased}

```{r mapping-biological-token, message = FALSE, include = FALSE}
mapping_biological_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "biological entity")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_biological_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_biological_token)
mapping_biological_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'biological entity') %>% left_join(mapping_biological_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_biological_token_stats)
biological_frame_stats <- happyr::get_frames("biological entity", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```


The [biological entity]{.smallcaps} metaphor has relatively low tokens [`r get_salience_stats(metaphor_salience, "n_token", "biological")`] and sits towards the bottom of the top-20 metaphor (rank `r get_metaphor_salience_rank(metaphor_salience, 'biological')`). However, it is expressed by relatively varied metaphorical expressions (`r get_salience_stats(metaphor_salience, "n_type", "biological")` types out of `r get_salience_stats(metaphor_salience, "n_token", "biological")` tokens) and elaborate mappings [`r get_salience_stats(metaphor_salience, "n_mapping", "biological")`] as presented below. 

- the root/base/seed of an organism &rarr; anger as cause (type=`r get_metaphor_mapping_n_lu(mapping_biological_token_stats, "root\\/base\\/seed")`; token=`r get_metaphor_mapping_tokenfreq(mapping_biological_token, "root\\/base\\/seed", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_biological_token, "root\\/base\\/seed")`
  
- fruits/yields of a plant &rarr; anger as effect (type=`r get_metaphor_mapping_n_lu(mapping_biological_token_stats, "fruits\\/yields")`; token=`r get_metaphor_mapping_tokenfreq(mapping_biological_token, "fruits\\/yields", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_biological_token, "fruits\\/yields")`

- the seed/source of an organism &rarr; cause of anger(type=`r get_metaphor_mapping_n_lu(mapping_biological_token_stats, "seed\\/source")`; token=`r get_metaphor_mapping_tokenfreq(mapping_biological_token, "seed\\/source", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_biological_token, "seed\\/source")`)

-  the growth/birth of the organism &rarr; the existence of anger (type=`r get_metaphor_mapping_n_lu(mapping_biological_token_stats, "growth\\/birth")`; token=`r get_metaphor_mapping_tokenfreq(mapping_biological_token, "growth\\/birth", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_biological_token, "growth\\/birth")`)

- plants creeping out &rarr; effect of anger (type=`r get_metaphor_mapping_n_lu(mapping_biological_token_stats, "creeping out")`; token=`r get_metaphor_mapping_tokenfreq(mapping_biological_token, "creeping out", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_biological_token, "creeping out")`)

- fertilising the organism/plant &rarr; maintaining anger (type=`r get_metaphor_mapping_n_lu(mapping_biological_token_stats, "fertilising")`; token=`r get_metaphor_mapping_tokenfreq(mapping_biological_token, "fertilising", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_biological_token, "fertilising")`)

No single mapping could be considered predominant. The metaphorical expressions utilise different components of the [organism]{.smallcaps} frame family (e.g., the [plant life cycle]{.smallcaps} and [animate entity]{.smallcaps} frames) to highlight different aspects of anger. Anger can be considered as an effect, or as a cause leading to another effect. The metaphor also includes expressions referring to the cause, existence, and effect of anger.


### [Marah is an (un)veiled object]{.smallcaps} {#unveiled-tokenbased}

```{r mapping-hidden-token, message = FALSE, include = FALSE}
mapping_hidden_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "veiled object")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_hidden_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_hidden_token)
mapping_hidden_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'veiled object') %>% left_join(mapping_hidden_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_hidden_token_stats)
hidden_frame_stats <- happyr::get_frames("veiled object", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```


This metaphor is attested in the two approaches and occupies the lower rank in the salience tables (rank `r get_metaphor_salience_rank(metaphor_salience, 'veiled')` in the lexical approach, and rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'veiled')` in the corpus one). [Anger]{.smallcaps} is construed as an object that the experiencer has control over to be displayed (i.e., expressed for others to perceive), or keep it hidden, hence perceptually inaccessible by others. These two inferences, which are re-cast as the mappings in the metaphor, are based on the [hiding]{.smallcaps} and [cause to see]{.smallcaps} frames.

`r paste(paste("- ", unlist(str_split(get_key_mappings(mapping_hidden_token), "___ ")), "\n", sep = ""), collapse = "")`

- unveiling an object &rarr; expressing anger (type=`r get_metaphor_mapping_n_lu(mapping_hidden_token_stats, "expressing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_hidden_token, "expressing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_hidden_token, "expressing")`

- hiding/protecting an object &rarr; suppressing anger (type=`r get_metaphor_mapping_n_lu(mapping_hidden_token_stats, "suppressing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_hidden_token, "suppressing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_hidden_token, "suppressing")`

The focus of this metaphor is on the externalisation or "unveiling" the anger. It is shown by the higher type and token frequencies of this mapping compared to the hiding mapping. In the lexical approach ([§\@ref(hidden-typebased)](#hidden-typebased)), however, the metaphor focuses on the suppression of anger (i.e., maintaining control by "hiding" the object). Overall, the [(un)veiled object]{.smallcaps} metaphor highlights the control and expression of anger.


### [Marah is physical contact/harm]{.smallcaps} {#harm-tokenbased}

The [contact/harm]{.smallcaps} metaphor is also attested in the lexical approach and ranks in the top-3 of the salience table ([Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based)), not in the corpus-based salience table. The metaphorical expressions evoking the [physical harm]{.smallcaps} frame capture the effect of anger. It impacts both the experiencer and other entities that become the target of angry person, especially when the Harmed_entity role of the frame is filled with entity other than the experiencer of anger. The following mappings are postulated for the metaphor.

```{r mapping-harm-token, message = FALSE, include = FALSE}
mapping_harm_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "harm$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_harm_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_harm_token)
mapping_harm_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'harm$') %>% left_join(mapping_harm_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_harm_token_stats)
harm_frame_stats <- happyr::get_frames("harm$", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

`r paste(paste("- ", unlist(str_split(get_key_mappings(mapping_harm_token), "___ ")), "\n", sep = ""), collapse = "")`

- harmful effect (at others) &rarr; effect of anger (type=`r get_metaphor_mapping_n_lu(mapping_harm_token_stats, "harmful effect")`; token=`r get_metaphor_mapping_tokenfreq(mapping_harm_token, "harmful effect", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_harm_token, "harmful effect")`

- physical harm &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_harm_token_stats, "experiencing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_harm_token, "experiencing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_harm_token, "experiencing")`

In Kövecses's [-@kovecses_metaphor_2000] [force-dynamic]{.smallcaps} system, the metaphor highlights the part where the experience of anger leads to experiencer's responses (e.g., directing the harm at others). This aspect is also an important part of the [anger is a natural force]{.smallcaps} ([§\@ref(naturalforce-tokenbased)](#naturalforce-tokenbased)), [anger is a subjugator]{.smallcaps} ([§\@ref(subjugator-tokenbased)](#subjugator-tokenbased)), and [anger is a weapon]{.smallcaps} ([§\@ref(weapon-tokenbased)](#weapon-tokenbased)) metaphors.

The difference in the two approaches for this metaphor is in the main meaning focus. The predominant types in the lexical approach focus on the cause of anger. The expressions referring to physical touching or harm are used in the context of harming the candidate experiencer (i.e., causing the candidate experiencer to be angry) ([§\@ref(harm-typebased)](#harm-typebased)). In the corpus approach, however, the physical harm expressions are used to show the harmful effect when experiencing anger and when targeting that effect to others.

### [Intensity is luminosity]{.smallcaps} {#luminousity-tokenbased}

```{r mapping-luminousity-token, message = FALSE, include = FALSE}
mapping_luminousity_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "luminousity$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_luminousity_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_luminousity_token)
mapping_luminousity_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'luminousity$') %>% left_join(mapping_luminousity_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_luminousity_token_stats)
luminousity_frame_stats <- happyr::get_frames("luminousity$", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

Amongst the metaphors in the bottom five of the corpus-based salience table ([Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based)), [intensity is luminosity]{.smallcaps} has the highest token (`r get_salience_stats(metaphor_salience, 'n_token', 'luminousity$')`) but also the lowest types (`r get_salience_stats(metaphor_salience, 'n_type', 'luminousity$')`) and mappings (`r get_salience_stats(metaphor_salience, 'n_mapping', 'luminousity$')`). The metaphor is not attested in the lexical approach. Below are the mappings of the metaphor.

`r paste(paste("- ", unlist(str_split(get_key_mappings(mapping_luminousity_token), "___ ")), "\n", sep = ""), collapse = "")`

- reducing the level of light &rarr; reducing anger (type=`r get_metaphor_mapping_n_lu(mapping_luminousity_token_stats, "reducing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_luminousity_token, "reducing", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_luminousity_token, "reducing")`

The metaphor is based on LUs evoking the [darkness]{.smallcaps} frame, a perspective of the [luminosity]{.smallcaps} frame that characterises the amount of light from a light source. These frames are utilised in highlighting the ceasing intensity of anger. [Intensity is luminosity]{.smallcaps} is related to the [anger is fire]{.smallcaps} ([§\@ref(fire-tokenbased)](#fire-tokenbased)) but captures anger's intensity not via the degree of heat, but the amount of light.

### [Marah is a person]{.smallcaps} {#person-tokenbased}

```{r mapping-person-token, message = FALSE, include = FALSE}
mapping_person_token <- filter(metaphor_tokenbased_mapping, str_detect(CM_BROADER, "person$")) %>% 
  mutate(status = if_else(str_detect(MAPPING, "_"), "key", "no"))
mapping_person_token_stats <- get_metaphor_mapping_stat_typefreq(mapping_person_token)
mapping_person_token_discuss <- get_mappings(metaphor_tokenbased_mapping, 'person$') %>% left_join(mapping_person_token %>% filter(status=='no') %>% select(LU, LU_GLOSS, MAPPING)) %>% filter(!is.na(LU)) %>% left_join(mapping_person_token_stats)
person_frame_stats <- happyr::get_frames("person$", df = marah, frame_var = "SFRAME", metaphor_var = "CM_BROADER", lexunit_var = "MP")
```

This metaphor could be viewed as a soft personification of anger and is not attested in the lexical approach. It is based on LUs evoking features of the [person]{.smallcaps} frame, which is a sub-case of the [animate entity]{.smallcaps} and [sentient]{.smallcaps} frames. 

`r paste(paste("- ", unlist(str_split(get_key_mappings(mapping_person_token), "___ ")), "\n", sep = ""), collapse = "")`

- inviting a person &rarr; causing anger (type=`r get_metaphor_mapping_n_lu(mapping_person_token_stats, "inviting")`; token=`r get_metaphor_mapping_tokenfreq(mapping_person_token, "inviting", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_person_token, "inviting")`

- being welcome by/befriend a person &rarr; experiencing anger (type=`r get_metaphor_mapping_n_lu(mapping_person_token_stats, "friend")`; token=`r get_metaphor_mapping_tokenfreq(mapping_person_token, "friend", lu_output = FALSE)`)

  - `r get_lu_gloss_n_printed(mapping_person_token, "friend")`

- dismissing the invited person &rarr; reducing anger (type=`r get_metaphor_mapping_n_lu(mapping_person_token_stats, "dismissing")`; token=`r get_metaphor_mapping_tokenfreq(mapping_person_token, "dismissing", lu_output = FALSE)`) (`r get_lu_gloss_n_printed(mapping_person_token, "dismissing")`)

Unlike the other personifying metaphors (e.g., [adversary]{.smallcaps} or [subjugator]{.smallcaps}), [anger is person]{.smallcaps} captures the causation and diminishing of anger, rather than the intensity and effect of it. The metaphorical expressions refer to actions and objectives that animate as well as sentient/cognising entities such as a person typically does. These include welcoming, inviting, dismissing, and befriending or playing with someone. In a way, such actions represent anger as a superficially safe emotion to be with.

## Token-based salience: Metonymy {#tokenbased-metonymy-salience-results}

```{r metonymy-category-tokenbased}
physiological_response_regex <- "(body heat|internal pressure|redness|breathing|visual interference|body hair|teeth biting|foaming)"
metonymy_tokenbased_salience_category <- metonymy_tokenbased_salience %>% 
  mutate(category = if_else(str_detect(METONYMY, physiological_response_regex), "physiological response", "social-communicative behaviour"))

physiology_type <- metonymy_tokenbased_salience_category %>% 
  filter(category == "physiological response") %>% 
  mutate(meto_stats = paste("[", METONYMY, "]{.smallcaps} (type=", n_type, "; token=", n_token, ")", sep = "")) %>% 
  pull(meto_stats)
physiology_type_total <- paste("*TOTAL~type~=", sum(filter(metonymy_tokenbased_salience_category, category == "physiological response")$n_type),"*; *TOTAL~token~=", sum(filter(metonymy_tokenbased_salience_category, category == "physiological response")$n_token), "*", sep = "")
physiology_type <- c(physiology_type, physiology_type_total)

social_type <- metonymy_tokenbased_salience_category %>% 
  filter(category != "physiological response") %>% 
  mutate(meto_stats = paste("[", METONYMY, "]{.smallcaps} (type=", n_type, "; token=", n_token, ")", sep = "")) %>% 
  pull(meto_stats)
social_type_total <- paste("*TOTAL~type~=", sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_type),"*; *TOTAL~token~=", sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_token), "*", sep = "")
social_type <- c(social_type, rep("", (length(physiology_type) - 1) - length(social_type)), social_type_total)

metonymy_tokenbased_category_print <- tibble(`social-communicative behaviour:` = social_type, `physiological response:` = physiology_type)
```

The corpus analysis of the metonymy now includes the token frequencies for the metonymies and their type frequencies. These metonymies are rank-ordered in [Table \@ref(tab:metonymy-table-token-based)](#metonymy-table-token-based) by the Aggregate values (i.e., the sums of the values in the “% of all tokens” and “% of all types” columns for each metonymy).


```{r metonymy-table-token-based}
metonymy_tokenbased_salience_print %>% 
  mutate(Metonymy = str_replace(Metonymy, "frustated", "frustrated"),
         Metonymy = str_replace(Metonymy, fixed("speechless"), "inability to speak"),
         Metonymy = str_replace(Metonymy, fixed("visual interference"), "inability to see"),
         Metonymy = str_replace_all(Metonymy, "(?<=for )anger", "marah"),
         Aggregate = replace(Aggregate, is.na(Aggregate), "")) %>% 
  # select(-Aggregate) %>% 
  knitr::kable(caption = "Conceptual metonymies for [anger]{.smallcaps} (Token-based, corpus approach)")
```


```{r metonymy-tokenbased-typebased-comparison}
metonymy_tokenbased_typebased_shared <- intersect(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy)
metonymy_tokenbased_only <- setdiff(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy)
metonymy_typebased_only <- setdiff(metonymy_typebased_salience$metonymy, metonymy_tokenbased_salience$METONYMY)
metonymy_tokenbased_typebased_ALL <- unique(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy)
```

From the two datasets, there are total `r numbers2words(length(unique(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy)))` unique metonymies, and most of them (i.e., `r numbers2words(length(intersect(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy)))` metonymies [`r round(length(intersect(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy))/length(unique(metonymy_tokenbased_salience$METONYMY, metonymy_typebased_salience$metonymy)) * 100, digits = 1)`%]) are shared (cf. [§\@ref(metonymy-compared-both-approaches)](#metonymy-compared-both-approaches) for further comparative discussion). The relatively high number of the shared metonymy types suggests a rather consistent metonymic construal for anger in Indonesian based on data from different approaches.  

### [Strong/aggressive verbal behaviour for anger]{.smallcaps} {#verbal-tokenbased}

- `r get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = 'aggressive verbal', mapping_var = "METONYMY", metaphor_data = FALSE)`

This metonymy is the most salient in both approaches, having `r get_salience_stats(metonymy_typebased_salience, "n_type", "aggressive verbal", metaphor_data = FALSE)` and `r get_salience_stats(metonymy_tokenbased_salience, "n_type", "aggressive verbal", metaphor_data = FALSE)` types in the lexical and corpus approaches respectively.

### [Aggressive visual behaviour for anger]{.smallcaps} {#visual-behaviour-tokenbased}

- `r get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = 'visual behaviour', mapping_var = "METONYMY", metaphor_data = FALSE)`

This is another metonymy that is also salient in the two approaches, ranked in the top 3. However, the corpus dataset reveals significantly greater number of types [`r get_salience_stats(metonymy_tokenbased_salience, "n_type", "aggressive visual behaviour", metaphor_data = FALSE)`] compared to the lexical dataset [`r get_salience_stats(metonymy_typebased_salience, "n_type", "aggressive visual behaviour", metaphor_data = FALSE)`] (*p*~Binomial~ ~two~~-~~tailed~`r pval_print(binom.test(c(get_salience_stats(metonymy_typebased_salience, "n_type", "aggressive visual behaviour", metaphor_data = FALSE), get_salience_stats(metonymy_tokenbased_salience, "n_type", "aggressive visual behaviour", metaphor_data = FALSE)))$p.value)`). From the metonymic expressions above, one does not contain the target term [marah]{.smallcaps} but the sentential context in (@pandangan_liar) clearly indicates that such expression has to do with anger.

(@pandangan_liar) *__Pandangan matanya liar__ dan memancarkan amarah yang meluap.* (ind_mixed_2012_1M: 575480)

In this example, the eyes both spurt out boiled over anger and show wild gaze.

### [Redness in the facial area for anger]{.smallcaps} {#redness-tokenbased}

- `r get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = 'redness', mapping_var = "METONYMY", metaphor_data = FALSE)`

This metonymy has greater number of types (and thus ranked higher) in the corpus dataset (`r get_salience_stats(metonymy_tokenbased_salience, "n_type", "redness", metaphor_data = FALSE)` types; rank=`r get_metaphor_salience_rank(metonymy_tokenbased_salience, "redness", FALSE)`) than in the lexical one (`r get_salience_stats(metonymy_typebased_salience, "n_type", "redness", metaphor_data = FALSE)` types; rank=`r get_metaphor_salience_rank(metonymy_typebased_salience, "redness", FALSE)`) (*p*~Binomial~ ~two~~-~~tailed~`r pval_print(binom.test(c(get_salience_stats(metonymy_typebased_salience, "n_type", "redness", metaphor_data = FALSE), get_salience_stats(metonymy_tokenbased_salience, "n_type", "redness", metaphor_data = FALSE)))$p.value)`). The metonymic expressions indicate that anger causes such physiological effects of redness in the facial area of the experiencer.

### [Internal pressure for anger]{.smallcaps} {#internal-pressure-tokenbased}

- `r get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = 'internal pressure', mapping_var = "METONYMY", metaphor_data = FALSE)`

The [internal pressure]{.smallcaps} metonymy shares the same number of types between the two datasets (`r get_salience_stats(metonymy_tokenbased_salience, "n_type", "pressure", metaphor_data = FALSE)` types in the corpus dataset, and `r get_salience_stats(metonymy_typebased_salience, "n_type", "pressure", metaphor_data = FALSE)` types in the lexical one). The internal pressure predominantly goes with trembling body parts and voice.

### [Body heat for anger]{.smallcaps} {#body-heat-tokenbased}

- `r get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "body heat", mapping_var = "METONYMY", metaphor_data = FALSE)`

The [body heat]{.smallcaps} metonymy is less productive in terms of its type in the corpus dataset (`r get_salience_stats(metonymy_tokenbased_salience, "n_type", "body heat", metaphor_data = FALSE)` types) compared to the lexical one (`r get_salience_stats(metonymy_typebased_salience, "n_type", "body heat", metaphor_data = FALSE)` types).

### [Violent frustrated behaviour for anger]{.smallcaps} {#frustation-tokenbased}

- `r get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "violent frustated", mapping_var = "METONYMY", metaphor_data = FALSE)`

This metonymy appears to be less prominent in the corpus dataset than in the lexical one in terms of the number of metonymic expressions and, by consequence, the ranking. In the lexical approach, it is evoked by `r numbers2words(get_salience_stats(metonymy_typebased_salience, "n_type", "violent", metaphor_data = FALSE))` metonymic expressions (rank=`r get_metaphor_salience_rank(metonymy_typebased_salience, 'violent', FALSE)`), which is significantly greater than the number of expressions in the corpus dataset (`r get_salience_stats(metonymy_tokenbased_salience, "n_type", "violent", metaphor_data = FALSE)` types; rank=`r get_metaphor_salience_rank(metonymy_tokenbased_salience, 'violent', FALSE)`) (*p*~Binomial~ ~two~~-~~tailed~`r pval_print(binom.test(c(get_salience_stats(metonymy_typebased_salience, "n_type", "violent", metaphor_data = FALSE), get_salience_stats(metonymy_tokenbased_salience, "n_type", "violent", metaphor_data = FALSE)))$p.value)`).

### Other singleton metonymies

There are `r nrow(filter(metonymy_tokenbased_salience, n_type < 2))` metonymies that only have a single type; they are (i) [`r filter(metonymy_tokenbased_salience, n_type < 2)[[1]][1]`]{.smallcaps} (`r str_replace(get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "crying", mapping_var = "METONYMY", metaphor_data = FALSE), "\\s\\[[0-9]\\]$", "")`), (ii) [`r filter(metonymy_tokenbased_salience, n_type < 2)[[1]][2]`]{.smallcaps} (`r str_replace(get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "foaming", mapping_var = "METONYMY", metaphor_data = FALSE), "\\s\\[[0-9]\\]$", "")`), (iii) [`r filter(metonymy_tokenbased_salience, n_type < 2)[[1]][3]`]{.smallcaps} (`r str_replace(get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "teeth biting", mapping_var = "METONYMY", metaphor_data = FALSE), "\\s\\[[0-9]\\]$", "")`), (iv) [`r str_replace(filter(metonymy_tokenbased_salience, n_type < 2)[[1]][4], fixed("speechless"), "inability to speak")`]{.smallcaps} (`r str_replace(get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "speechless", mapping_var = "METONYMY", metaphor_data = FALSE), "\\s\\[[0-9]\\]$", "")`), and (v) [`r str_replace(filter(metonymy_tokenbased_salience, n_type < 2)[[1]][5], fixed("visual interference"), "inability to see")`]{.smallcaps} (`r str_replace(get_lu_gloss_n_printed(metonymy_tokenbased, mapping_regex = "visual interference", mapping_var = "METONYMY", metaphor_data = FALSE), "\\s\\[[0-9]\\]$", "")`). Two metonymies have greater number of types in the lexical dataset, namely [inability to see for anger]{.smallcaps} [`r get_salience_stats(metonymy_typebased_salience, "n_type", "visual interference", metaphor_data = FALSE)`] and [lip/teeth biting/pressing for anger]{.smallcaps} [`r get_salience_stats(metonymy_typebased_salience, "n_type", "teeth biting", metaphor_data = FALSE)`].

## Types of the metonymies (token-based dataset) {#metonymy-category-tokenbased}

From `r nrow(metonymy_tokenbased_salience_category)` metonymies, `r nrow(filter(metonymy_tokenbased_salience_category, category == "physiological response"))` metonymies represent physiological responses (type=`r sum(filter(metonymy_tokenbased_salience_category, category == "physiological response")$n_type)`; token=`r sum(filter(metonymy_tokenbased_salience_category, category == "physiological response")$n_token)`), and `r nrow(filter(metonymy_tokenbased_salience_category, category != "physiological response"))` are of the social-communicative behaviour types (type=`r sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_type)`; token=`r sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_token)`).

```{r metonymy-category-table-tokenbased}
metonymy_tokenbased_category_binomtest_type <- binom.test(sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_type), sum(sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_type), sum(filter(metonymy_tokenbased_salience_category, category == "physiological response")$n_type)))

metonymy_tokenbased_category_binomtest_token <- binom.test(sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_token), sum(sum(filter(metonymy_tokenbased_salience_category, category != "physiological response")$n_token), sum(filter(metonymy_tokenbased_salience_category, category == "physiological response")$n_token)))

metonymy_tokenbased_category_print %>% 
  mutate(`social-communicative behaviour:` = str_replace_all(`social-communicative behaviour:`, fixed("speechless"), "inability to speak"),
         `social-communicative behaviour:` = str_replace_all(`social-communicative behaviour:`, fixed("frustated"), "frustrated"),
         `physiological response:` = str_replace_all(`physiological response:`, fixed("visual interference"), "inability to see")) %>% 
  rename(`Social-Communicative Behaviour:` = `social-communicative behaviour:`,
         `Physiological Response:` = `physiological response:`) %>% 
  knitr::kable(caption = "Two types of metonymies of [anger]{.smallcaps} (token-based dataset)")
```

As in the lexical dataset ([§\@ref(metonymy-category-typebased)](#metonymy-category-typebased)), the number of metonymies in the physiological response category are greater than the ones in the social-communicative behaviour. However, the social-communicative behaviour category in total still exhibits significantly greater number of types (*p*~Binomial~ ~two~~-~~tailed~`r pval_print(metonymy_tokenbased_category_binomtest_type$p.value)`) and tokens (*p*~Binomial~ ~two~~-~~tailed~`r pval_print(metonymy_tokenbased_category_binomtest_token$p.value)`) than the physiological response category. Therefore, the lexical and corpus-based approaches converge along these distributional aspects of the metonymies.

# Comparison of the results in the lexical and corpus-based approaches

The following sub-sections summarise several points of quantitative and qualitative comparisons between the findings from the two approaches.

## Differences in the number of conceptual metaphors

```{r number-of-metaphors-in-the-two-approaches}
n_cmetaphor_corpus <- nrow(metaphor_salience)
n_cmetaphor_typebased <- nrow(metaphor_typebased_salience)
chisqtest_n_cmetaphor <- chisq.test(c(n_cmetaphor_corpus, n_cmetaphor_typebased))

# difference in the number of metaphors attested only in either approach
n_cmetaphor_lexical_only <- length(metaphor_in_lexical_only)
n_cmetaphor_corpus_only <- length(metaphor_in_corpus_only)
chisqtest_n_cmetaphor_in_either_approach <- chisq.test(c(n_cmetaphor_corpus_only, n_cmetaphor_lexical_only))

```


The obvious difference between the lexical and corpus-based approaches is the potential of the latter to provide a significantly greater number of conceptual metaphors (`r nrow(metaphor_salience)` metaphors) than the former (`r nrow(metaphor_typebased_salience)` metaphors) (*X*^2^=`r round(chisqtest_n_cmetaphor$statistic, digit = 2)`, *df*=`r chisqtest_n_cmetaphor$parameter`, *p*~goodness~~-of-~~fit~`r pval_print(chisqtest_n_cmetaphor$p.value)`).

From the total `r happyr::numbers2words(length(metaphor_typebased_salience$metaphor))` metaphors identified in the type-based, lexical approach ([Table \@ref(tab:metaphor-table-type-based)](#metaphor-table-type-based)), `r round((length(metaphor_in_all_database)/length(metaphor_typebased_salience$metaphor)) * 100, digits = 1)`% (i.e., `r numbers2words(length(metaphor_in_all_database))` metaphors) are also attested in the corpus, token-based approach ([Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based)). These shared metaphors are: `r paste(paste(paste("[", metaphor_in_all_database, "]{.smallcaps}", sep = "")[-length(metaphor_in_all_database)], collapse = ", "), ", and ", paste("[", metaphor_in_all_database[length(metaphor_in_all_database)], "]{.smallcaps}", sep = ""), sep = "")`. Of these shared metaphors, only `r numbers2words(length(metaphor_in_all_database_top20))` (`r round(length(metaphor_in_all_database_top20)/20 * 100)`%) appear in the top-20 list in [Table \@ref(tab:metaphor-table-token-based)](#metaphor-table-token-based).


The `r numbers2words(length(metaphor_in_lexical_only))` metaphors identified only in the lexical approach (ordered decreasingly by their salience) are `r paste(paste(paste("[", metaphor_in_lexical_only, "]{.smallcaps} (", metaphor_in_lexical_only_aggregate, "%; rank=", metaphor_in_lexical_only_ranking, ")", sep = "")[1:length(metaphor_in_lexical_only)-1], collapse = ", "), ", and ", paste("[", metaphor_in_lexical_only, "]{.smallcaps} (", metaphor_in_lexical_only_aggregate, "%; rank=", metaphor_in_lexical_only_ranking, ")", sep = "")[length(metaphor_in_lexical_only)], sep = "") %>% str_replace_all(fixed("rough, solid entity"), "a substance") %>% str_replace_all(fixed("degree of control for anger is object dimension"), "degree of controlling anger is the size of patience as a container")`. The remaining `r round((length(metaphor_in_corpus_only)/nrow(metaphor_salience)) * 100, digits = 1)`% (`r length(metaphor_in_corpus_only)` types) of the token-based metaphors are absent from the lexical dataset, which is a significant difference (*X*^2^=`r round(chisqtest_n_cmetaphor_in_either_approach$statistic, digit = 2)`, *df*=`r chisqtest_n_cmetaphor_in_either_approach$parameter`, *p*~goodness~~-of-~~fit~`r pval_print(chisqtest_n_cmetaphor_in_either_approach$p.value)`). The fact that several metaphors are only attested in either approach and that the corpus-based approach offers a higher likelihood to identify more metaphors, showcase the call for combining the two approaches for more comprehensive coverage.

## Differences in the salience of the shared metaphors {#shared-salience-compare}

Another quantitative difference can be inspected among the shared metaphors. Namely, they can differ in their salience values and thus ranking. [Figure \@ref(fig:salience-ranking-comparison)](#salience-ranking-comparison) visualises these differences.

```{r salience-ranking-comparison, fig.width = 10, dpi = 600, fig.cap="Comparison of ranking and salience of the shared metaphors between the two approaches/methods."}
metaphor_salience1 <- metaphor_salience |> 
  mutate(metaphor = str_replace_all(metaphor, "anger", "marah"))
df_token_based <- metaphor_salience1 |> 
  mutate(ranking = 1:nrow(metaphor_salience),
         metaphor = factor(metaphor, 
                           levels = metaphor_salience1$metaphor),
         ranking = fct_rev(factor(ranking)))
metaphor_typebased_salience1 <- metaphor_typebased_salience |> 
  mutate(metaphor = str_replace_all(metaphor, "anger", "marah"))
df_type_based <- metaphor_typebased_salience1 |> 
  mutate(ranking = 1:nrow(metaphor_typebased_salience1),
         metaphor = factor(metaphor, 
                           levels = metaphor_typebased_salience1$metaphor),
         ranking = fct_rev(factor(ranking)))
df_both <- select(df_token_based, metaphor, aggregate, ranking) |> 
  mutate(method = "corpus_token_based") |> 
  bind_rows(select(df_type_based, metaphor, aggregate, ranking) |> 
              mutate(method = "lexical_type_based")) |> 
  mutate(method = factor(method, 
                         levels = c("lexical_type_based", 
                                    "corpus_token_based")))
df_shared_metaphor <- df_both |> 
  filter(metaphor %in% str_replace_all(metaphor_in_all_database, 
                                       "anger", "marah"))

## visualisation =====
df_shared_metaphor |> 
  ggplot(aes(x = method, 
             y = ranking,
             group = metaphor)) +
  geom_path(aes(colour = metaphor),
            alpha = .7) +
  theme_light() +
  geom_text(data = filter(df_shared_metaphor, method == "lexical_type_based"),
            aes(label = str_replace(metaphor, "(^marah is | of marah)", ""), size = aggregate)) +
  geom_text(data = filter(df_shared_metaphor, method == "corpus_token_based"),
            aes(label = str_replace(metaphor, "(^marah is | of marah)", ""), size = aggregate)) +
  labs(y = "Salience rank",
       colour = "Shared metaphors",
       x = "Methods",
       size = "Aggregate value",
       caption = "Font size represents the aggregate values while the lines capture the ranking dynamics.")
ggsave("results/fig-1.png", 
       width = 10,
       height = 6.5,
       dpi = 600,
       unit = "in")
ggsave("results/fig-1.tiff", 
       width = 10,
       height = 6.5,
       dpi = 600,
       unit = "in",
       device = "tiff")
```

Clear differences can be seen. For instance, [substance in a container]{.smallcaps} is the most salient in the corpus, token-based approach while the same metaphor is only at rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'is substance in a ')` in the type-based approach. A similar increase in salience from the type-based approach is shown by the [possession]{.smallcaps} and [intensity is verticality]{.smallcaps} metaphors in the corpus, token-based approach. In addition to salience, and hence, ranking increase, there are categorically specific metaphors whose salience values drastically drop in the corpus, token-based data (i.e., from within the top-10 position in the type-based list to outside the top-10 position in the token-based data); they are [physical contact/harm]{.smallcaps} (from the rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'physical contact')` to `r get_metaphor_salience_rank(metaphor_salience, 'physical contact')`), [disease]{.smallcaps} (from the rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'disease')` to `r get_metaphor_salience_rank(metaphor_salience, 'disease')`), [insanity]{.smallcaps} (from the rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'insanity')` to `r get_metaphor_salience_rank(metaphor_salience, 'insanity')`), and [bitterness]{.smallcaps} (from the rank `r get_metaphor_salience_rank(metaphor_typebased_salience, 'bitterness')` to `r get_metaphor_salience_rank(metaphor_salience, 'bitterness')`). Finally, [Figure \@ref(fig:salience-ranking-comparison)](#salience-ranking-comparison) also shows that [fierce, captive animal]{.smallcaps} and [fire]{.smallcaps} are consistently amongst the most salient metaphors in the two approaches, suggesting their centrality in the cognitive model of [anger]{.smallcaps} in Indonesian; there are also metaphors that relatively consistently occupy the lower half of the top-20 table (e.g., [natural force]{.smallcaps}, [sleeping organism]{.smallcaps}, and [(un)veiled object]{.smallcaps}).

Empty labels at rank 4 and 5 in the corpus-based panel in [Figure \@ref(fig:salience-ranking-comparison)](#salience-ranking-comparison) are filled with different metaphors not attested in the type-based, lexical approach, namely [weapon]{.smallcaps} and [adversary/opponent]{.smallcaps} respectively. What is qualitatively interesting is that the focus on overall control over the emotion presented by the [insanity]{.smallcaps} metaphor in the type-based approach is also captured using a different image by the [adversary/opponent]{.smallcaps} metaphor. Similarly, additional metaphor at rank 10 on the corpus, token-based panel attested only in the corpus-based approach, namely [intensity is quantity]{.smallcaps}, expands the intensity focus together with other metaphors in the top-10 list (i.e., [verticality]{.smallcaps} and [substance]{.smallcaps}-related metaphors). In the next section, such a qualitative difference in main meaning focus, due to quantitative difference (i.e., mapping productivity), between the shared metaphor will be discussed.

 
## Differences in the main meaning focus of the shared metaphors

Even though the two approaches share a number of metaphors, the main meaning focus of these metaphors can differ across the two approaches. This is summarised in [Table \@ref(tab:main-meaning-focus-table)](#main-meaning-focus-table) for the shared metaphors in the top 20 list.

```{r main-meaning-focus-table}
tribble(
  ~Metaphor, ~`Lexical, type-based approach`, ~`Corpus, token-based approach`,
  "[fierce, captive animal]{.smallcaps}", "aggresiveness of the angry person ([§\\@ref(animal-typebased)](#animal-typebased))", "regulating anger (i.e., restraining the animal) ([§\\@ref(animal-tokenbased)](#animal-tokenbased))",
  "[physical contact/harm]{.smallcaps}", "cause of the anger ([§\\@ref(harm-typebased)](#harm-typebased))", "harmful effect to the experiencer/angry person and to the others/external target ([§\\@ref(harm-tokenbased)](#harm-tokenbased))",
  "[pressurised substance in container]{.smallcaps}", "causing increased pressure and substance ([§\\@ref(pressurised-typebased)](#pressurised-typebased))", "loss of control (i.e., exploding container) ([§\\@ref(pressurised-tokenbased)](#pressurised-tokenbased))",
  "[(un)veiled object]{.smallcaps}", "regulating anger (i.e., concealing the object) ([§\\@ref(hidden-typebased)](#hidden-typebased))", "expressing anger (i.e., unveiling the object) ([§\\@ref(unveiled-tokenbased)](#unveiled-tokenbased))"
) |> 
  knitr::kable(caption = "The main meaning focus of the shared metaphors in the top 20 list.")
```

Such qualitative differences stem from quantitative assessment of the productivity of the metaphorical mappings for the metaphors (i.e., the number of different linguistic expressions manifesting a given mapping) [@kovecses_metaphor_2010, 140]. Another related difference between the two approaches is that for a given shared metaphor, certain mapping is only identified in one of the approaches. This could surely affect the salience of the metaphor. For instance, the mapping expressing the controlling of the substance/emotion is not identified for the [pressurised substance]{.smallcaps} metaphor in the lexical approach but only in the corpus, token-based approach. The same is true for the [heated fluid]{.smallcaps} metaphor, in which an additional mapping unidentified in the lexical approach is present in the token-based approach, namely the ceasing (i.e., decreased intensity of) anger construed as an evaporating liquid.

The two approaches overall should be viewed as complementary in enriching the characterisation of the main meaning focus of a given metaphor.

## Metonymies {#metonymy-compared-both-approaches}

As to the metonymy, most (i.e., `r round(length(metonymy_tokenbased_typebased_shared)/length(metonymy_tokenbased_typebased_ALL) * 100, 2)`%) of the `r numbers2words(length(metonymy_tokenbased_typebased_ALL))` metonymies from the two approaches combined are shared (as discussed in [§\@ref(tokenbased-metonymy-salience-results)](#tokenbased-metonymy-salience-results)); they are: `r paste(paste(paste("[", metonymy_tokenbased_typebased_shared, "]{.smallcaps}", sep = "")[-length(metonymy_tokenbased_typebased_shared)], collapse = ", "), ", and ", paste("[", metonymy_tokenbased_typebased_shared[length(metonymy_tokenbased_typebased_shared)], "]{.smallcaps}", sep = ""), sep = "")`. 
This high percentage of the shared metonymies indicates some convergences between the two approaches in terms of the **types** and the salient metonymies for anger. Similar to the metaphor data, there are metonymies that appear only in the lexical, type-based data (i.e., `r paste(paste(paste("[", metonymy_typebased_only, "]{.smallcaps}", sep = "")[1:length(metonymy_typebased_only)-1], collapse = ", "), ", and ", paste("[", metonymy_typebased_only, "]{.smallcaps}", sep = "")[length(metonymy_typebased_only)], sep = "")`) and in the corpus, token-based data (i.e., `r paste(paste(paste("[", metonymy_tokenbased_only, "]{.smallcaps}", sep = "")[1:length(metonymy_tokenbased_only)-1], collapse = ", "), ", and ", paste("[", metonymy_tokenbased_only, "]{.smallcaps}", sep = "")[length(metonymy_tokenbased_only)], sep = "") %>% str_replace(fixed("speechless"), "inability to speak")`). 

As in the salience comparison of the metaphors ([§\@ref(shared-salience-compare)](#shared-salience-compare)), the metonymy also shows dynamics in their salience ranking across the two approaches (see [Figure \@ref(fig:salience-ranking-comparison-metonymy)](#salience-ranking-comparison-metonymy)).

```{r salience-ranking-comparison-metonymy, fig.width = 11, dpi = 600, fig.cap="Comparison of ranking and salience of the shared metonymies between the two approaches/methods."}
meto_token_salience <- metonymy_tokenbased_salience_category |> 
  mutate(METONYMY = str_replace_all(METONYMY, "anger", "marah"),
         ranking = 1:nrow(metonymy_tokenbased_salience_category),
         METONYMY = factor(METONYMY, 
                           levels = str_replace_all(metonymy_tokenbased_salience_category$METONYMY, "anger", "marah")),
         ranking = fct_rev(factor(ranking)))
meto_type_salience <- metonymy_typebased_salience_category |> 
  mutate(metonymy = str_replace_all(metonymy, "anger", "marah"),
         ranking = 1:nrow(metonymy_typebased_salience_category),
         metonymy = factor(metonymy, 
                           levels = str_replace_all(metonymy_typebased_salience_category$metonymy, "anger", "marah")),
         ranking = fct_rev(factor(ranking)))
df_meto_both <- meto_token_salience |> 
  select(metonymy = METONYMY, aggregate, ranking) |> 
  mutate(method = "corpus_token_based") |> 
  bind_rows(meto_type_salience |> 
              select(metonymy, aggregate = n_perc_type, ranking) |> 
              mutate(method = "lexical_type_based")) |> 
  mutate(method = factor(method, 
                         levels = c("lexical_type_based", 
                                    "corpus_token_based")))

df_meto_shared <- df_meto_both |> 
  filter(metonymy %in% str_replace_all(metonymy_tokenbased_typebased_shared,
                                       "anger", "marah"))

## visualisation =====
df_meto_shared |> 
  ggplot(aes(x = method, 
             y = ranking,
             group = metonymy)) +
  geom_path(aes(colour = metonymy),
            alpha = .7) +
  theme_light() +
  geom_text(data = filter(df_meto_shared, method == "lexical_type_based"),
            aes(label = str_replace(metonymy, "for marah$", ""), size = aggregate)) +
  geom_text(data = filter(df_meto_shared, method == "corpus_token_based"),
            aes(label = str_replace(metonymy, "for marah$", ""), size = aggregate)) +
  labs(y = "Salience rank",
       colour = "Shared metonymies",
       x = "Methods",
       size = "Aggregate value",
       caption = "Font size represents the aggregate values while the lines capture the ranking dynamics.")
ggsave("results/fig-2.png", 
       width = 11,
       height = 6.5,
       dpi = 600,
       unit = "in")
ggsave("results/fig-2.tiff", 
       width = 11,
       height = 6.5,
       dpi = 600,
       unit = "in",
       device = "tiff")
```

It is shown that [strong/aggressive verbal behaviour for anger]{.smallcaps} tops the metonymy tables in the two approaches, but has higher salience value in the corpus, token-based approach (cf. the different font sizes). The prominence of [strong/aggressive verbal behaviour for anger]{.smallcaps} metonymy corroborates Heider's [-@heider_landscapes_1991, 80, 116] finding that verbal abuse is one kind of reaction of anger (in addition to physical violence).

Another interesting dynamics is the noticeable rise of salience and ranking of the [redness in the facial area for anger]{.smallcaps} metonymy in the corpus, token-based data. Together with the second-most prominent [aggressive visual behaviour for anger]{.smallcaps} metonymy, [redness in the facial area]{.smallcaps} highlights the visibility aspect of anger in the corpus, token-based data. Lastly, [Figure \@ref(fig:salience-ranking-comparison-metonymy)](#salience-ranking-comparison-metonymy) also shows a drop in ranking and salience of metonymies in the corpus, token-based approach, namely [violent frustrated behaviour]{.smallcaps} (the second most salient in the type-based approach), [lip/teeth biting/pressing]{.smallcaps}, and [visual interference]{.smallcaps}. In terms of the broader types of the metonymies, we have seen in [§\@ref(metonymy-category-typebased)](#metonymy-category-typebased) and [§\@ref(metonymy-category-tokenbased)](#metonymy-category-tokenbased) that the Social-Communicative Behaviour type is still more prominent (in terms of the number of metonymic expressions and token frequency) than the Physiological Response type, which, nevertheless, is represented by higher number of metonymies.

# Metonymical basis of [anger]{.smallcaps} metaphors {#metonymic-basis}

Three of the five physiology-based metonymies suggested by Lakoff and Kövecses [-@holland_cognitive_1987, 197] motivate the two principal metaphors for anger: [anger is heat]{.smallcaps} and [anger is heated fluid in a container]{.smallcaps} [see @rajeg_metafora_2013, 211-213, for the quantitative evidence for the prominent status of these anger metaphors in Indonesian]. These three metonymies are [body heat for anger]{.smallcaps} ([§\@ref(body-heat-tokenbased)](#body-heat-tokenbased) and [\@ref(body-heat-typebased)](#body-heat-typebased)), [internal pressure for anger]{.smallcaps} ([§\@ref(internal-pressure-typebased)](#internal-pressure-typebased) and [\@ref(internal-pressure-tokenbased)](#internal-pressure-tokenbased)), and [redness in the facial area for anger]{.smallcaps} ([§\@ref(redness-typebased)](#redness-typebased) and [\@ref(redness-tokenbased)](#redness-tokenbased)). In addition to these, there are other physiology-related metonymies ([Table \@ref(tab:metonymy-category-table-typebased)](#metonymy-category-table-typebased) and [Table \@ref(tab:metonymy-category-table-tokenbased)](#metonymy-category-table-tokenbased)) motivating other metaphors. The list below shows the relevant metaphors (from the two approaches) motivated by the physiology-based metonymies.

- [Body heat for anger]{.smallcaps} could motivate these metaphors:

  - [anger is heat]{.smallcaps}
  - [anger is fire]{.smallcaps}
  - [anger is heated fluid in a container]{.smallcaps}
  - [intensity is temperature]{.smallcaps} (corpus dataset; outside the top-20 salient metaphors)

The [heated fluid in a container]{.smallcaps} may also be motivated by the [internal pressure for anger]{.smallcaps} metonymy [@holland_cognitive_1987, 197], given that internal heat pressure could trigger the boiling over of the heated fluid.

- [Redness in the facial area for anger]{.smallcaps} could motivate these metaphors:

  - [anger is fire]{.smallcaps}
  - [intensity is luminosity]{.smallcaps}

- [Caressing the chest for anger]{.smallcaps} ([§\@ref(caressing-chest-typebased)](#caressing-chest-typebased)), [lip/teeth biting/pressing for anger]{.smallcaps} ([§\@ref(lip-biting-typebased)](#lip-biting-typebased)), [internal pressure for anger]{.smallcaps}, and [breathing difficulty for anger]{.smallcaps} ([§\@ref(breathing-typebased)](#breathing-typebased)) could motivate these metaphors:

  - [anger is contained substance]{.smallcaps}
  - [anger is pressurised substance in a container]{.smallcaps}
  - [anger is heated fluid in a container]{.smallcaps}
  - [anger is heat]{.smallcaps}

Caressing the chest represents a gesture that could be interpreted as an attempt to hold the pressing anger inside (hence, an attempt to show patience). The biting/pressing of the lip/teeth could also indicate holding back pressure and pain. Difficulty to breath could be related to raising internal pressure pushing up the lungs.

- [Inability to see for anger]{.smallcaps} motivates the [anger is darkness]{.smallcaps} metaphor.

The social-based metonymies represent what Kövecses [-@kovecses_where_2015] suggests as non-embodied contexts and could motivate these metaphors:

- [Strong/aggressive verbal behaviour for anger]{.smallcaps} and [Aggressive visual behaviour for anger]{.smallcaps}

  - [anger is fierce, captive animal]{.smallcaps}
  - [anger is weapon]{.smallcaps}
  - [anger is (un)veiled object]{.smallcaps}

- [Violent frustrated behaviour for anger]{.smallcaps}; [crying for anger]{.smallcaps}

  - [anger is insanity]{.smallcaps}
  - [anger is adversary/opponent]{.smallcaps}
  - [anger is subjugator]{.smallcaps}

The qualitative analyses above offer some motivation for several metaphors in the top-20 tables of the two approaches. These metaphors are based on characteristics that are metonymically associated with anger.

# Cultural model of [anger]{.smallcaps} {#cultural-model}

The cultural model of anger in Indonesian converges with the five-stage prototypical cultural model of anger (and emotions in general) in the cognitive linguistic literature [@holland_cognitive_1987, 210-214; @kovecses_where_2015, 84-85]. The aspects highlighted by the metaphorical mappings [@kovecses_metaphor_2000, 40-46] in the model correspond to the semantic dimensions of emotion from psychological studies [see @soriano_2013 that discusses (di/con)vergence between metaphorical meaning foci of anger and the dimensions of emotions]. For instance, the Intensity aspect [@kovecses_metaphor_2000, 41] corresponds to the dimension of Arousal [@soriano_2013, 415-416]. The Negative/Positive (or [Un]desirability) aspect [@kovecses_metaphor_2000, 44] corresponds to the dimension of Valence/Pleasantness [@soriano_2013, 417]. The cultural model of the concept [anger]{.smallcaps} in Indonesian unfolds as follows:

Stage 1: cause of anger (offending event)

The Causation aspect is an addition to Kövecses's [-@kovecses_metaphor_2000] list of aspects [see @soriano_2013, 416]. It provides the reason why anger exists. This aspect is predominantly construed as metaphorical events (see the list below, such as igniting fire, disturbing dormant animal, pulling the trigger), or a constant state of affairs (e.g., annoyance) [ @soriano_2013, 416; see also @heider_landscapes_1991, 81, finding that the cause of anger can be indicated by the "annoyance" cluster that he proposes]:

- 

    - annoyance
    - disturbing/harming dormant animal
    - touching/harming a person
    - pulling the trigger
    - igniting fire
    - inviting a person
    - seed/source of an organism

Stage 2: anger exists

The existential stage of anger contains three aspects: Intensity, Evaluation, and Harm. They could be metaphorically captured in two different ways: (i) separately by different mappings from a single metaphor (e.g., the [fierce, captive animal]{.smallcaps} metaphor can capture the Intensity and Harm aspects) [@soriano_2013, 415-418], or (ii) exclusively by one metaphor (e.g., the [darkness]{.smallcaps} metaphor only captures the negative Evaluation aspect). It is at this stage that anger exerts force on the experiencer (i) to become emotional, (ii) to experience the harmful effect of the force and exhibiting various physiological responses as shown by the metonymies, and eventually (iii) to attempt a retribution/reaction [@holland_cognitive_1987, 214; @kovecses_where_2015, 84-85] (see stage 5 below).

- aspect: Intensity

    The prototypical model of emotion postulates that, at this stage, the intensity tends to escalate and reach the limit the self can control [@kovecses_where_2015, 85]. The following mappings capture the Intensity:

    - fullness of, and increase pressure to, the substance
    - high degree of fire, and level of heat in the fluid
    - more quantity
    - aggressive behaviour and the attack of the animal
    - degree of effect of physical harm
    - amount of natural force and strength of effect
    - (extent of the) insane behaviour

- aspect: Evaluation 

    The Evaluative aspect of anger is prototypically negative [@soriano_2013, 416]. This includes the danger and unacceptability of the angry reaction to others and the self. They are metaphorically construed as:

    - bitter taste
    - being in the dark
    - intrinsic negative effect of the disease
    - strong and destructive natural force as an intrinsically negative force
    - affected/impeded motion and being confined

- aspect: Harm

    The Harm aspect relates to the psychological harm to the self and others. It is construed as the effect of physiological harm via the [emotional harm is physical damage]{.smallcaps} metaphor [@kovecses_metaphor_2000, 46], which is inherited by the more specific metaphorical mappings as follows:

    - danger of the adversary's attack
    - harmful effect of physical harm, disease, and destructive natural process
    - aiming weapon at others
    - insane behaviour
    - harm of the attack by the aggressive animal

Stage 3: attempt at control

Following the force-dynamic schema of emotion metaphors [@kovecses_metaphor_2000, 62-63], the rational self demonstrates a force tendency towards calmness. In doing so, the self attempts to resist and/or counterforce the emotional forces, including regulating the felt intensity and/or the expression (losing control of) emotion [@soriano_2013, 417]. This stage is captured by these mappings:

- 

    - restraining captive animal
    - suppressing/impeding substance or fluid from going out
    - hiding the object
    - combating the opponent
    - avoiding confinement/impediment
    - maintaining sane/calm behaviour

If all the above attempts were successful and anger ceased before the self lost control, it would represent the non-prototypical model of anger scenarios: "Successful suppression", "Spontaneous cessation", and "Terminating event" [@holland_cognitive_1987, 215].

Stage 4: lacking and losing control

When the force of anger is too strong to control, the experiencer eventually loses control. That aspect is captured by the following metaphorical mappings:

- 

    - insane behaviour
    - reduced dimension of an object (shortage of emotional control/patience/calm)
    - being subjugated as an inferior object
    - losing/giving in to the adversary
    - unleashing the fierce, captive animal
    - unveiling the hidden object
    - explosion of the pressurised container
    - boiling over the heated fluid to overflow
    - being ravaged by the destructive natural force

The aspects Difficulty to cope and Passivity are also represented in this stage since loss of control over forces implies difficulty in counter-forcing them, resulting in the self being passive to the force (e.g., being subjugated, losing, being ravaged).

Stage 5: act of retribution

This stage represents the Desire to act (to others) aspect: a reaction as "a result of the emotion itself" [@kovecses_metaphor_2000, 45; @soriano_2013, 417], rather than the desire to possess anger. Metaphorical expressions from the following mappings differently capture construal of expressing and directing the anger towards the wrongdoer (retribution) or other parties.

- 

    - releasing/channelling out the fluid at others
    - aiming the weapon at others as targets, and punishing others with the weapon

Lakoff and Kövecses [-@holland_cognitive_1987, 214] propose that "redirected anger" (i.e., non-retributive to the wrongdoer) is a non-prototypical extension. Below are examples of the redirected anger from the above mappings: 

(@chanelling_out_model) `r marah %>% filter(str_detect(CM_BROADER, 'substance in a '), str_detect(NODE_SENTENCES, 'melampiaskan.+musala')) %>% pull(NODE_SENTENCES)` (`r marah %>% filter(str_detect(CM_BROADER, 'substance in a '), str_detect(NODE_SENTENCES, 'melampiaskan.+musala')) %>% pull(FILES)`: `r marah %>% filter(str_detect(CM_BROADER, 'substance in a '), str_detect(NODE_SENTENCES, 'melampiaskan.+musala')) %>% pull(SENT_ID) %>% as.character()`)

(@retribution_model) `r marah %>% filter(str_detect(CM_BROADER, 'weapon'), str_detect(NODE_SENTENCES, 'sasaran.+Persiku')) %>% pull(NODE_SENTENCES)` (`r marah %>% filter(str_detect(CM_BROADER, 'weapon'), str_detect(NODE_SENTENCES, 'sasaran.+Persiku')) %>% pull(FILES)`: `r marah %>% filter(str_detect(CM_BROADER, 'weapon'), str_detect(NODE_SENTENCES, 'sasaran.+Persiku')) %>% pull(SENT_ID) %>% as.character()`)

The connection between the metaphor and the stages of the model tends to be one to many, as demonstrated by Soriano [-@soriano_2013] and earlier by Lakoff and Kövecses [-@holland_cognitive_1987]. A metaphor does not exclusively contribute to one stage, but plausibly to different stages of the model. This is because the metaphorical mappings for a given metaphor refer to different aspects corresponding to certain stages in the model. For instance, the [fire]{.smallcaps} metaphor has the mapping for the cause of anger (igniting the fire) (stage 1), the intensity of anger (high degree of heat) (stage 2), and intrinsically the danger of high extent of fire to the self and others (enacted at stage 5). Similarly, metaphorical mappings within the [fierce, captive animal]{.smallcaps} metaphor consist of, among others, restraining the animal captive (an attempt at control; stage 3), unleashing the captive animal (loss of control; stage 4), which can manifest in different reaction to the self or others (stage 5), and the attack of the animal to the victim that has implication to the harm aspect of anger [cf. @holland_cognitive_1987, for the contribution of the [heated fluid]{.smallcaps} metaphor to the different stages of the model].

# The role of context {#contexts}

## Situational context {#situational-context}

Kövecses [-@kovecses_metaphorcreativity_2019, 217-218; -@kovecses_metaphormindcontext_2019] proposes three sub-types of the situational context: the physical environment, the social situation, and the cultural context.

Geographically, Indonesia is the largest archipelago in the world, lying across the Equator and between the Indian and Pacific oceans. It has a relatively even tropical climate year-round with two seasons, dry and wet. Geologically, Indonesia is located at the "Pacific Ring of Fire" (PRoF) [@britannica_ring_2020; @masum_pacific_2019]. The PRoF refers to a geographical area with "high volcanic and seismic activities within the edges of the Pacific Ocean" [@masum_pacific_2019, 1], meaning that it is "a site for frequent earthquakes and powerful volcanic eruptions" [@masum_pacific_2019, 1]. These two natural forces could be accompanied with the tsunami (warning) [cf. @reid_history_2015]. There are more than 100 active volcanoes from the west to the eastern parts of Indonesia [@masum_pacific_2019, 3; @siebert_volcanoes_2011]. 

Such salient geographical and geological landscapes of Indonesia could motivate the use and comprehension of these metaphors: [natural force]{.smallcaps}, [fire]{.smallcaps}, [heated fluid]{.smallcaps}, and [temperature/hot-cold]{.smallcaps} (e.g., *kemarahan berubah __dingin__* 'anger becomes *cold*', *aku __kepanasan__ amarah* 'I *suffer the heat of* anger'; also in [§\@ref(heat-typebased)](#heat-typebased)). This study and Rajeg [-@rajeg_metafora_2013, 110-111] found singletons of expressions specifying anger as (lava in a) volcano, a sub-case of [heated fluid]{.smallcaps} metaphor: _**gunung api** kemarahan_ '*volcano* of anger' ([§\@ref(hotfluid-tokenbased)](#hotfluid-tokenbased)), _kemarahan seperti **lahar**_ 'anger be like *lava*', _**memuntahkan lahar** amarah_ '*puking the lava* of anger', and _amarah bagai **kepundan gunung**_ 'anger be like the *(lava) crater of mountain*' [@rajeg_metafora_2013, 110-111]. Moreover, the PRoF environment could motivate the [natural force]{.smallcaps} metaphor, evoked predominantly by expressions from the [natural disaster]{.smallcaps} and [destructive natural process]{.smallcaps} frames ([§\@ref(naturalforce-tokenbased)](#naturalforce-tokenbased)) [see also @fox_body_2018, 137, who found the construal of anger as a storm arising in the person in Rotenese of Eastern Indonesia; and @rajeg_metafora_2013, who found a statistical association between [natural force]{.smallcaps} metaphor and anger]. Another piece of evidence to link anger and its negativity to the destructive natural force comes from Reid's [-@reid_history_2015] historical study, which identified several uses of the phrase *gempa bumi* 'earthquakes lit. quake earth' and related expressions from the Malay Bible in the [*Malay Concordance Project*](https://mcp.anu.edu.au) [@reid_history_2015, 66]. These expressions are used to represent earthquakes as manifestations of the wrath of God, and omens (following venerable Islamic and pre-Islamic local beliefs); Reid [-@reid_history_2015, 66] further notes that description about earthquakes appear prominently in the chronicles of Java and Bali, suggesting the long-standing awareness of such context of Indonesia physical environment. 

From the socio-cultural aspect, Indonesian tends to emphasise non-individualistic/group affiliation and social solidarity. This reflects Indonesia's deeply rooted socio-cultural legacy called *gotong royong* 'lit. work together' [@simarmata_gotong_2020]. Such social-cultural consideration gives rise to the importance of empathy and the concealment of emotion in the social context [@heider_cultural_2011, 10-11; @fox_body_2018, 137]. To maintain social harmony, concealment of emotion and reconciliation should be the primary goal, rather than hostile confrontation and aggressive expression. Such regulation of emotion is also related to politeness that can benefit social interaction [@prastuti_adaptation_2020]. This is consistent with Heider's study on Minangkabau and Javanese Indonesian speakers who "mask most anger, and the open expression of anger is strongly disapproved of and negatively sanctioned" [@heider_landscapes_1991, 80]. This context motivates the use of metaphors highlighting the attempt at emotional control and expression, such as the [(un)veiled object]{.smallcaps} metaphor [see also @matsuki_1995, 149, for similar socio-cultural context in Japanese to suppress their anger]. In Rotenese of Eastern Indonesia, the [(un)veiled object]{.smallcaps} metaphor is part of "ritual admonitions" offering counsel on anger (and other emotions such as fear, shame, and doubt), where the addressee is told not to reveal anger in the neck [@fox_body_2018, 136]. In addition to the metaphor, social-communicative behaviour metonymies, especially those denoting the expressive behaviours of anger, could also be motivated by the socio-cultural aspect.

## Discourse context {#discourse-context}

```{r surrounding-discourse-code}
marah %>% filter(duplicated(SENT_ID), USE == "metaphor") %>% pull(SENT_ID) -> dup_sent_metaphor
marah %>% filter(duplicated(SENT_ID)) %>% pull(SENT_ID) -> dup_sent_metaphor_metonymy
mixed_metaphor_patterns <- marah %>% filter(SENT_ID %in% dup_sent_metaphor) %>% arrange(desc(SENT_ID), CM_BROADER) %>% select(SENT_ID, FILES, NODE_SENTENCES, MP, MP_GLOSS, CM_BROADER) %>% group_by(SENT_ID, FILES) %>% mutate(n_metaphor_per_sentence = n_distinct(CM_BROADER)) %>% ungroup() %>% group_by(SENT_ID, FILES) %>% mutate(n_pattern_in_sentence = n()) %>% ungroup()
mixed_metaphor_metonymy_patterns <- marah %>% filter(SENT_ID %in% dup_sent_metaphor_metonymy) %>% arrange(desc(SENT_ID), CM_BROADER) %>% select(SENT_ID, FILES, NODE_SENTENCES, MP, MP_GLOSS, CM_BROADER)
```

One aspect of the discourse context is the "surrounding discourse", that is, how the cotext/linguistic context of the previous discourse influence, or is related to, the production of metaphor in the following discourse [@kovecses_metaphorcreativity_2019, 218-219]. One example is the use of similar conceptual metaphors in multiple constructions (e.g., noun phrases and argument-structure constructions) that are combined into a single utterance (typically at the sentence level) [cf. @sullivan_frames_2013 for the interplay between grammatical constructions and metaphor].  

(@discourse_ctx_fire) _Wajahnya yang riang menjadi **kemerah-merahan** karena **nyala** **api** kemarahan yang **membakar** dadanya_ (ind_mixed_2012_1M: 376378).

Example (@discourse_ctx_fire) illustrates the combination of a metonymic pattern (in the main clause) and two metaphorical expressions for [anger is fire]{.smallcaps} (within the subordinate clause marked with *karena* 'because'). The use of [fire]{.smallcaps} metaphor in the noun phrase _**nyala api** kemarahan_ '*flame of the fire of* anger' could influence the forgoing discourse in the relative clause construction _kemarahan yang **membakar** dadanya_ 'anger that *burns* the chest'. In the context of the sentence, the two metaphorical expressions emphasise the specific reason for the metonymy [redness in the facial area for anger]{.smallcaps} used in the preceding, main clause. Let us see another example of [anger is substance in a container]{.smallcaps} ([§\@ref(contained-substance-tokenbased)](#contained-substance-tokenbased) and [§\@ref(contained-substance-typebased)](#contained-substance-typebased))


(@discourse_ctx_substance) _Jika amarah tersebut tak **tersalurkan** atau ternetralisasi dengan permintaan maaf atau jiwa besar mengikhlaskan, yang akan terjadi selanjutnya adalah amarah itu **mengendap** dan semakin **mengeras** dalam batinnya._ (ind_newscrawl_2016_1M: 343033)

Sentence (@discourse_ctx_substance) contains more than one metaphorical expression that indicates a sequential, cause-and-effect scenario all within the same metaphorical frame, namely the [contained substance]{.smallcaps}. The sentence talks about the conserving effects of anger-substance if not being channelled out in the form of apology and big-heartedness. This shows how the use of a given metaphor in the preceding discourse indicating an inability to channel out and neutralise anger, which is the cause, could influence the reasoning for the ensuing discourse indicating the effects of such cause using the same metaphor (i.e., becoming solidified and hardened inside).

## Conceptual-cognitive context {#conceptual-context}

One aspect of the conceptual-cognitive context hinted at by Kövecses [-@kovecses_metaphorcreativity_2019, 219-220; -@kovecses_metaphormindcontext_2019] is the **conventionality** of a given metaphor:

i.  the conventional association between a given source and target domains in the metaphorical conceptual system, and

i.  the conventionalised source-domain expression used to talk about certain aspects of the target domain. 

For the first point, Kövecses mentions [life is a journey]{.smallcaps} and [argument is war]{.smallcaps} metaphors with the conventional relationship between their source and target domains in the long-term memory of the speakers. In the case of anger in Indonesian, Rajeg [-@rajeg_metafora_2013] has identified which source domains are statistically associated with anger (compared to the other emotions) ([§\@ref(previous-study-anger)](#previous-study-anger)). This statistical association between the two domains could be assumed to reflect their conventional association, which needs further experimental verification. In addition to using statistical association to assess a metaphor's conventionality, Lakoff and Turner [-@lakoff_more_1989, 55] propose that a conceptual metaphor is considered conventionalised in language when "it underlies a range of everyday linguistic expressions". This proposal suggests the role of type frequency (i.e., the number of linguistic-types manifesting a metaphor) in measuring conventionality.

For the second point, Kövecses argues that the verb "defend" from the [war]{.smallcaps} domain is a conventionalised expression to express the meaning 'supporting an argument' (as in "*defending* an argument") in the target domain of [argument]{.smallcaps}. From the corpus-based perspective, the conventionalised linguistic expressions for a given aspect of, or meaning in, the target domain could be identified from the expression's frequency of occurrences in the corpus (e.g., how many times does a given source-domain LU, compared to the other, is used for a given aspect or meaning of the target) [see @stefanowitsch_words_2006, 68]. As an illustration, the most frequent (and potentially the default choice of) LU to express the causation aspect of anger via the [fire]{.smallcaps} frame ([§\@ref(fire-tokenbased)](#fire-tokenbased)) is based on the verb root *sulut* 'kindle/ignite', as in _**menyulutkan** [marah]{.smallcaps}_ '*kindle*/*ignite* anger' (45 tokens), rather than _**memantik** [marah]{.smallcaps}_ '*light up* anger' (3 tokens) or _**mencetuskan** [marah]{.smallcaps}_ '*ignite/light up* anger' (2 tokens).

Another type of conceptual-cognitive context is ideology, particularly religious ideology. Religious ideology and belief in Indonesia are very strong. They have long been an integral part of Indonesian daily, socio-cultural life, since it is listed as the first principle in the nation's ideology *Pancasila* ('The Five Principles'), namely belief in the Supreme God. From the religious and spiritual perspective, anger is generally viewed negatively and dangerous [@bhide_anger_2007; @faizah_how_2021; @minhas_spiritual_2017]; this provides a plausible grounding for metaphors focusing on the negative valence and danger, namely [adversary]{.smallcaps}, [fierce, captive animal]{.smallcaps}, [subjugator]{.smallcaps}, [disease]{.smallcaps}, [darkness]{.smallcaps}, [insanity]{.smallcaps}, [evil creature]{.smallcaps} (e.g., *massa sudah __dirasuki__ amarah* 'the masses has been *possessed* by anger'), and [deceiver]{.smallcaps} (e.g., *__tipuan__/__godaan__ amarah* '*trickery*/*enticement* of anger') [@soriano_anger_2003, 118 also argues for the religio-cultural influence of Christianity on construing anger as [devil]{.smallcaps} in Spanish]. For instance, as a Balinese Hindu, in my elementary-school religion subject, I was taught that anger is one of the six enemies, the so-called *Sad Ripu* ('Six Enemies'), that humans need to manage or even defeat. This is a religious basis for the use of [adversary]{.smallcaps} and [subjugator]{.smallcaps} metaphors in construing anger [see also @faizah_how_2021 for the view from Islam; and @minhas_spiritual_2017, from the multi-religion perspective]. Moreover, the devastating effects of anger have been illustrated in various parts of the famous Hindu epic, the *Mahābhārata* [@bhide_anger_2007]. Another example is the religious-ideological view of anger as a disease, which is reflected in the notion of "Spiritual healing or medicine" [@minhas_spiritual_2017]. Such a healing can also be used, among others, for "anger management" and (literally) can "prevent the impact of disease by reducing bad feeling or anger" [@minhas_spiritual_2017, 168].

## Bodily context {#bodily-context}

```{r bodily-context}
non_bp <- c("jiwa", "pikiran", "batin", "ekspresi", "raut", "akal", "kalbu", "mental", "tatapan", "diri")
bp_meta_freq <- marah %>% 
  filter(!BODY_PART %in% non_bp, !is.na(BODY_PART)) %>% 
  count(BODY_PART, CM_BROADER, sort = TRUE)
bp_freq <- bp_meta_freq %>% 
  group_by(BODY_PART) %>% 
  summarise(n = sum(n)) %>% 
  arrange(desc(n)) %>% 
  mutate(eng = c("liver", "eyes", "chest", "heart", "crown of the head", "face", 
                 "head", "brain", "body", "mouth", "hand(s)", "body")) %>% 
  select(BODY_PART, eng, n)
```

The bodily context is related to the locus of emotion, which is in turn rooted in the cultural situation and pre-modern religious practices in Indonesia. This study found that *hati* 'liver' is the predominant body-part term (`r pull(filter(bp_freq, BODY_PART == "hati"), n)` tokens) used as locus of emotion in a variety of metaphors (*X*^2^=`r round(chisq.test(bp_freq$n)$statistic, digit = 2)`, *df*=`r chisq.test(bp_freq$n)$parameter`, *p*~goodness~~-of-~~fit~`r pval_print(chisq.test(bp_freq$n)$p.value)`)^[The other body-part terms used in the metaphorical expressions are as follows (together with their token frequencies): `r bp_freq %>% filter(BODY_PART != "hati") %>% mutate(bp_eng_freq = paste("*", BODY_PART, "* '", eng, "' (", n, ")", sep = "")) %>% pull(bp_eng_freq) %>% paste0(., collapse = ", ")`.]. Siahaan [-@siahaan_did_2008] proposes that the liver divination ritual and the ethno-religious belief of *hati* 'liver' as the locus for the living soul could be the underlying cultural motivations for the importance of *hati* in the conceptualisation of emotion and cognition in Indonesian. This shows the sociocultural situatedness of embodiment in metaphorical conceptualisations of anger.

# Conclusions {#conclusions}

From combining the lexical (type-based) and corpus linguistic (token-based) approaches, this study reveals the dynamics (i.e., convergence and divergence) in the metaphors and metonymies of [anger]{.smallcaps} and their degree of salience. Metaphors attested in the lexical and the corpus-based approaches may differ in several respects. One example is in the different main meaning focus of the same metaphor in the two approaches, due to different productivity and presence of a given metaphorical mapping (e.g., the [pressurised substance in a container]{.smallcaps} and [fierce, captive animal]{.smallcaps} metaphors). Another example is the attestation of metaphors only in either approach. Such differences would also result in different degree of salience and ranking of the same metaphors.

This study also found that the metaphors strongly associated with anger in Rajeg's [-@rajeg_metafora_2013] corpus-based study ([§\@ref(previous-study-anger)](#previous-study-anger)) also make into the top-20 salience tables (i.e., the family of the [pressurised and heated substance]{.smallcaps}, the [heat/fire]{.smallcaps}, and the [natural force]{.smallcaps} metaphors), even though the relative salience of the metaphors in this study is contrasted for their occurrences with anger only (rather than with other emotions as in Rajeg [-@rajeg_metafora_2013] and Stefanowitsch [-@stefanowitsch_words_2006]). This study also offers new insights regarding the salient metonymies for anger in Indonesian that are not discussed in previous studies ([§\@ref(previous-study-anger)](#previous-study-anger)). Overall, the salience-based analyses offer a different perspective in capturing the prominence of metaphors and metonymies for anger by taking into account the central components of metaphor and metonymy in CMT, namely the extent of mappings and the linguistic expressions. 

Furthermore, the Indonesian prototypical cultural model of anger is largely correspond to the five-stage model proposed in previous work on anger in American English [@holland_cognitive_1987], and capture relevant semantic dimensions of emotions proposed in the psychological literature [@soriano_2013]. This study also identified examples of non-prototypical model of anger, such as the redirected anger and the metaphor suggesting concealment. 

Finally, this study argues for the different contexts as the motivating factors in the uses of several metaphors of anger. These contexts include the geographical and geological landscapes of Indonesia, the deeply rooted socio-cultural and religious beliefs of Indonesian, the conventional metaphors in the memory, and the socio-cultural situatedness of embodiment for the bodily locus of anger. The contextual dimension of metaphor usages is the missing aspect of previous works on anger metaphors in Indonesian.

# References {-}