Do near-synonyms occur with the same metaphors:
A comparison of anger terms in American English

Kaisa Turkkila (kaisa.turkkila@gmail.com)

Abstract

When studying metaphorical target domains by means of corpus linguistics, a problem that needs to be addressed is how to retrieve metaphorical expressions associated with that domain from a corpus. One suggested answer is metaphorical pattern analysis, which claims that we can map out all the metaphors for a target domain by choosing one word to represent the domain and by analyzing its occurrences in the corpus. The method makes the assumption, among others, that near-synonyms occur with the same metaphorical mappings. This paper tests the assumption by examining an earlier metaphorical pattern analysis of near-synonyms and by analyzing the metaphors in which anger, rage, fury, and wrath occur in the Corpus of Contemporary American English. The results show that, broadly speaking, near-synonyms do occur with the same metaphorical mappings, but not necessarily to the extent that we could always map out entire target domains with a single search word.

1.Introduction

Metaphor has been a popular research topic in linguistics especially since the publication of Lakoff and Johnson’s Metaphors We Live By (1980). For decades, linguists have studied the nature of metaphor and its significance in our language use, but only recently have they started to pay closer attention to the methods in which metaphor is studied. Some have developed procedures for identifying metaphorically used words (Pragglejaz Group 2007, Steen et al. 2010), others have come up with methods for extracting metaphors from a corpus (Partington 1997, Stefanowitsch 2006b).

One such method is metaphorical pattern analysis (MPA), which suggests that it is possible to map out all the metaphors for a target domain by choosing just one word to represent that domain and studying the metaphorical expressions in which the word occurs (Stefanowitsch 2006b: 64). An important assumption behind this method is that near-synonyms occur broadly with the same metaphorical mappings (Stefanowitsch 2006b: 101–102). It should therefore not matter whether we use the word anger or rage to study the target domain of anger; both words should produce more or less the same results. This assumption is given some support in Stefanowitsch (2006b: 96–99), where little differences are found in the metaphors in which the words happiness and joy occur. Although Stefanowitsch (2006b: 101–102) takes the results of this comparison as proof that near-synonyms produce similar results, I would like to submit the hypothesis to further tests.

In this paper, I will therefore examine a previous study that has applied MPA on a pair of near-synonyms (Ogarkova 2007) as well as analyze a set of near-synonyms—anger, rage, fury, and wrath—in the Corpus of Contemporary American English (COCA) (Davies 2008–). I will also discuss how the anger metaphors identified in this study compare with metaphors identified in earlier literature.

2.Background

Much of the linguistic work done within conceptual metaphor theory has been based on intuition rather than empirical studies (Stefanowitsch 2006a: 4, Rojo and Orts 2010: 3301), but corpus linguistics has meant a change. Large electronic corpora are usually accessed by means of words or word forms, but because metaphorical mappings are not necessarily associated with particular words, retrieving them from a corpus can be difficult (Stefanowitsch 2006a: 1–2). Stefanowitsch (2006a: 1–2) lists and evaluates three separate strategies that have been used for extracting metaphors from a non-annotated corpus (Rojo and Orts 2010: 3301).

The first strategy is to search for metaphors manually, but this limits the size of the corpora and therefore renders the representativity of the results questionable. The second strategy relies on source-domain vocabulary, which is a convenient choice because metaphorical expressions contain source-domain words by definition. We can therefore choose a source domain, list lexical items relevant to it, search for them in a corpus, and identify the metaphorical occurrences among the hits. These occurrences can then be analyzed in more detail, uncovering information about the mappings in which the source domain participates.

The source-domain oriented approach or course doesn’t work if we want to investigate mappings associated with a particular target domain (Stefanowitsch 2006a: 3). The third strategy therefore relies on target-domain vocabulary. Stefanowitsch exemplifies this third strategy with the approach in Partington (1997: 111–112), where a list of keywords is first extracted from a corpus specialized in the target domain. Potentially metaphorical words are then identified in the list, and these as well as related words are finally searched for in a more general corpus. This approach has two weaknesses: first, it is only applicable for target domains where specialized corpora exist, and second, it can only identify source domains whose vocabulary is frequent enough to appear on the keyword list (Stefanowitsch 2006a: 3).

To overcome these limitations, Stefanowitsch (2006a: 3–4) suggests another approach, MPA. Similar approaches have been used by other scholars, for example in Tissari (2003) and Martin (2006), but Stefanowitsch (2006b: 64–65) seems to be the first one to argue that the approach exhaustively identifies the metaphors associated with a target domain.

MPA is built on the observation that there are two kinds of metaphorical expressions: ones that include target-domain words and ones that do not. Examples (1) and (2), from COCA, illustrate these types:

1. Tonya’s confession [...] sparked an anger in Jacob.
2. Bryan sat fuming in his room.

Example (1) is what Stefanowitsch calls a metaphorical pattern: it contains both target-domain vocabulary (anger) and source-domain vocabulary (spark). Example (2) is not a metaphorical pattern, because it only contains source-domain vocabulary (fuming) and no target-domain vocabulary.

Stefanowitsch (2006b: 64) suggests that we can use metaphorical patterns to investigate mappings in the following way:

We choose a lexical item referring to the target domain under investigation and extract (a sample of) its occurrences in the corpus. In this sample, we then identify all metaphorical expressions of which the search word is a part, and these expressions can then be grouped into coherent groups representing general mappings.

This procedure is what Stefanowitsch calls metaphorical pattern analysis.

3.Ogarkova’s comparison of jealousy and envy

Some studies have already applied MPA on near-synonyms, but the comparison of happiness and joy in Stefanowitsch (2006b: 96–99) seems to have been the only one to directly address the question whether near-synonyms occur with the same mappings. Rojo and Orts (2010) compare the near-synonyms crisis and recession in a collection of news texts, but because their sample sizes are very small (167 occurrences of crisis and 193 of recession) and because their corpus of financial articles from The Economist is not representative of general English, I will not analyze their results here. I will, however, concentrate on Ogarkova’s 2007 study, which applies MPA on all the occurrences of jealousy and envy in the British National Corpus.

Ogarkova treats jealousy and envy as distinct emotions and therefore as distinct conceptual domains, and if we accept this view, her results are not relevant for the purposes of this paper. However, as Ogarkova acknowledges (2007: 100), the words jealousy and envy are often used interchangeably. The dictionary definitions in Macmillan (2012) are certainly very similar:

jealousy

an unhappy feeling because someone has something that you would like or can do something that you would like to do

1. a feeling of being unhappy and upset because you think someone who you love is attracted to someone else

envy

the unhappy feeling that you have when you want very much to do something that someone else does or to have something that they have

Apart from wording, the main sense descriptions of jealousy and envy are almost identical. The only notable difference is that jealousy has a subordinate sense that is specifically related to love relationships, and this sense seems to be what motivates the treatment of jealousy and envy as separate emotions (Ogarkova 2007: 99–101). The similarity of Macmillan’s sense descriptions, however, suggests that jealousy and envy can just as well be taken to represent a single domain. Whether we treat jealousy and envy as two distinct domains seems to depend on how specific our categorization of domains is. Even if there are two domains, jealousy and envy, it could be argued that only those instances of jealousy that fit the subordinate sense description represent the domain of jealousy and the rest the domain of envy.

Ogarkova’s results show that there are altogether seven mappings that occur exclusively with either jealousy or envy (2007: 101–116). I gathered the mappings that Ogarkova has found with both jealousy and envy into a single table so that it would be easy to compare the coverage achieved with each word. Table 1 summarizes her results. The absolute numbers of hits are normalized to a sample of 1000 occurrences.

Table 1 Mappings associated with jealousy and envy in Ogarkova (2007)

Table 1 shows the 37 mappings in Ogarkova’s sample that occurred at least four times with either search word when the number of hits was normalized to a sample of 1000 occurrences. Out of those mappings, jealousy instantiated 34 (91.89 percent) and envy 33 (89.19 percent). This means that there are three mappings that did not occur with jealousy (being envious is acting with an object, envy is antidote/poison, and envy is fuel) and four mappings that did not occur with envy ((the intensity of) jealousy is high/low, acting on an emotion is acting in a location, jealousy is a spy, and jealousy is TRANSFERRING AN OBJECT).

The main focus in Ogarkova’s comparison is on mappings that are statistically more significantly associated with either jealousy or envy. According to Ogarkova (2007: 117–122), there are five metaphors that are more significantly associated with jealousy (jealousy is a disease/pain/physical annoyance, jealousy is insanity/foolishness, jealousy is fire, jealousy is a weapon/sharp object, and jealousy is (hot) liquid (in a container under pressure)) and three metaphors that are more significantly associated with envy (envy is a possessed object, acting on envy is acting with an object, and being envious is becoming green in color). In a similar comparison, Stefanowitsch (2006b: 48) finds that 4 percent of metaphorical patterns distinguish between happiness and joy, which is interpreted as a positive result that suggests that focusing on one word only does not affect the results too much. In Ogarkova (2007: 117), the corresponding amount is 14 percent (8 out of 56 mappings), which seems much higher.

4.A comparison of anger, rage, fury, and wrath in COCA

4.1Materials and methods

I tested the hypothesis that near-synonyms occur with the same mappings by applying MPA on a set of near-synonyms from COCA. The analysis was divided into two major phases. The first phase consisted of a rather straightforward application of the procedure outlined above:

1. I chose four lexical items that refer to the target domain of anger: anger, rage, fury, and wrath. This set of words is identified in Stefanowitsch (2006b: 71) as containing possible candidates to represent the domain of anger.
2. I took a random sample of 500 occurrences of each noun from COCA.
3. Among these occurrences, I identified the metaphorical expressions in which anger was the target domain. The identification relied on the metaphor identification procedure (MIP) (Pragglejaz Group 2007) and on its extension MIPVU (Steen et al. 2010), where a word is considered to be used metaphorically if its contextual meaning contrasts with a more basic meaning but can be understood in terms of it.
4. I categorized the metaphorical expressions according to their source domains. When possible, I used the same categories that are used in the analysis of anger in Stefanowitsch (2006b)—many of which are taken from Kövecses’s (1998: 128–129) summary of earlier studies.

The second phase of the analysis focused on mappings that were not found with all search words. To find out whether this was due to sample size or whether it reflected real differences between the words, I did the following for each search word:

5. I drew all the occurrences of the word from COCA and stored them in a text file.
6. I listed the mappings with which the search word did not occur but that did come up with other search words.
7. I compiled a list of search strings that instantiated the missing mappings in the other samples and in Stefanowitsch (2006b). For example, for the mapping anger is a disease, the list would consist of the following strings: fester, froth, relapse, impotent, livid, sick, symptom, purge, and infect. The search strings did not have to occur as full words—infect, for example, would find words such as infectious and infected.
8. Out of all the occurrences of the search word, I selected those that included the search strings listed in step (g).
9. For the selected occurrences, I applied steps (c) and (d).

4.2Results and discussion

Table 2 summarizes the metaphors found with anger, rage, fury, and wrath; it lists the mappings that occurred at least three times in the total set of 2000 samples in the order of their overall frequency and conflates less frequent ones under the other category[1]. Columns Anger–Wrath show how many times a particular search word instantiated a particular category. The numbers of occurrences that were found with additional queries in the second phase of the analysis and not in the actual 500-word samples are shown in brackets and are not included in the total count.

Table 2 A summary of the mappings associated with anger, rage, fury, and wrath

*Bracketed amounts are not included in the total count.

Most of the mappings in Table 2 are familiar from earlier literature such as Lakoff and Kövecses’s (1987: 195–219) analysis of anger in American English and in Stefanowitsch’s (2006b: 71–78) metaphorical pattern analysis of anger in the British National Corpus. As in Stefanowitsch’s analysis, the most frequent mappings identified here represent event structure metaphors. In Stefanowitsch’s sample, the location system (anger is a location/place) accounts for 10.2 percent of metaphorical patterns and the object system (anger is a possession, anger is a moving/moved object, anger is an object, anger is an object in a location) for 56.15 percent. Here, the total percentages are quite similar, 9.89 for the location system and 60.12 for the object system, but there is some variation between the search words. The location is system is hardly at all instantiated by wrath—it accounts for 0.91 percent of the metaphorical patterns identified with the word—whereas the object system accounts for a total of 86.76 percent. The corresponding percentages with the rest of the search words are 7.43 and 58.75 with anger, 18.27 and 43.03 with rage, and 16.72 and 43.16 with fury.

The largest group after event structure metaphors is also the same as in Stefanowitsch (2006b: 75–77). Mappings related to the general domains of liquid and heat (anger is fire, anger is a substance in a container (under pressure), anger is an explosion, anger is a liquid, anger is fluid in a container, anger is a natural force, anger is hot fluid in a container, and anger is heat) account for 11.88 percent, which is not very far from the 13.3 percent reported in Stefanowitsch (2006b: 77). This group of mappings corresponds with Lakoff and Kövecses’s metaphor anger is heat and its more specific variants anger is the heat of a fluid in a container and anger is fire, which they suggest are the most central metaphors for anger in American English (1987: 197)—a claim that these results seem to support. Again, wrath instantiates these mappings much less frequently than the other search words: the group accounts for only 3.20 percent of the metaphorical patterns in its sample, when the corresponding percentages are 13.19 with anger, 18.27 with rage, and 15.50 with fury.

Similar variation is also apparent in the next most frequent mapping, anger is an opponent in a struggle. It accounts for 2.99 percent of the metaphorical mappings in the total sample but only 0.91 percent in the sample for wrath. Another group of metaphors, anger is aggressive animal behavior, anger is a captive animal, anger is a sleeping organism, which Lakoff and Kövecses (1987: 206–207) see as instances of a more general metaphor anger is a dangerous animal, does not display similar variation. The group accounts for 1.92 percent of metaphorical patterns identified in the total sample and 2.05 percent in the sample for wrath. These observations show that not all mappings display the same variation, but even when a mapping can be identified with wrath, it often instantiates it less frequently than the other search words.

Next, I wish to draw attention to some differences between the results of this analysis and the metaphors identified in earlier studies. The metaphor anger is a superior (His actions were completely governed by anger) is suggested in Kövecses (1998: 129) and also identified in Stefanowitsch (2006b: 74) (anger rule the day). This data set includes several patterns where the person experiencing anger has the superior position (banish rage from awareness, in command of anger, keep anger in its place, ungovernable anger, ungoverned fury) and only one pattern where anger is the superior (fury banish emotion). In the light of this data, the label anger is superior/inferior seems more appropriate for the mapping.

Table 2 includes 5 mappings that seem to be absent from the earlier literature on anger metaphors:

anger is a sound
fury pulse on X's tone, the echo of fury, quiet fury/rage, silent fury/rage, whirring fury, wrath echo

anger is a (dear) person
accommodate X’s wrath, a stranger to X’s anger, attract anger, court X’s anger, foster rage, get in touch with anger, invite X’s wrath/the wrath of X, long-cherished anger, meet ((with) the wrath of) X

anger is food
feed on rage, fresh fury, fury ripen, raw rage

anger is blind
blind anger/rage

Anger is a mask
don a mask of fury, face be contorted in a mask of rage, rage mask X

The SOUND metaphor has come up in investigations on other emotion concepts, at least in Stefanowitsch’s metaphorical pattern analysis of sadness (2006b: 88), but it does not seem to have been identified as a metaphor for anger. The person/people metaphor is not usually listed among emotion metaphors at all, but Esenova (2009) does suggest a more specific version anger is a child. One pattern in the (dear) person category (foster rage) is compatible with the child metaphor, but most patterns imply a more general or completely different relationship. food is another source domain that usually does not come up in analyses of emotion concepts, and even here some of its instances could perhaps be categorized under anger is plant. The last two mappings anger is blind and anger is a mask are specific and not very productive: both of them are instantiated by a single lexical unit (the former by blind and the latter by mask). The instances under anger is blind might also be motivated by the metonymy interference with accurate perception for anger (Lakoff & Kövecses 1987: 196–197).

Do the near-synonyms anger, rage, fury, and wrath occur with the same metaphorical mappings? Table 2 shows that more or less the same metaphorical mappings can be identified with anger, rage, and fury—provided that the sample size is large enough. The only exception is the mapping anger is a weather phenomenon, which I was unable to find with the search word rage. This should not be a major problem for MPA, because the mapping status of anger is a weather phenomenon can be questioned for at least two reasons: first, even when the additional searches are included, the mapping is very infrequent, and second, the metaphorical patterns could also be accounted for by the more general mapping anger is a natural force. In fact, Stefanowitsch (2006b: 74) does not list weather phenomenon as a mapping for anger (he does so for other emotion concepts) but subsumes the example climate of anger under the natural force mapping. If we treated the weather phenomenon and natural force mappings as one category, we could say that anger, rage, and fury are associated with all the same metaphorical mappings that exceed the threshold level of three occurrences in the total sample.

The discussion above has already shown that wrath stands out in the extent to which is participates in many mappings, but it also stands out in terms of coverage. Even after additional searches, there are six mappings that were not found with wrath: anger is a disease, anger is an organism, anger is a superior/an inferior, anger is a plant, anger is insanity, and anger is a mask.

The overall results show that all the metaphorical mappings for anger in COCA can more or less be mapped out with the search word anger, rage, or fury, but not with wrath. I do not know the exact reason why wrath stands out, but I can say the following. First, wrath is by far the least frequent of these words in COCA. Its relative frequency per million words is 3.59, which does not differ that much from the frequency of fury (7.74), but which is considerably lower than that of rage (16.61) and anger (35.08). Second, the Corpus of Historical American English (COHA) (Davies 2010–) shows that the relative frequency of wrath per million words in that corpus has decreased from the 1810s to the 2000s much more than that of the other words. The decrease in the relative frequency of wrath is as much as 93.66 percent, whereas rage has decreased 73.98 percent, fury 69.27 percent, and anger 21.20 percent. This change is also reflected in the fact that both dictionaries used in this analysis for metaphor identification, Macmillan (2012) and Longman (2012), tag wrath as formal.

5.Conclusions

This paper set out to investigate whether near-synonyms occur with the same metaphorical mappings and whether we can make generalizations about a whole conceptual domain from the metaphors with which a single representative word occurs. The discussion of Ogarkova’s (2007) results shows that the answer may not be as straightforward as assumed in Stefanowitsch (2006b)—her comparison of jealousy and envy in BNC shows that the corpus contains 4 metaphorical mappings that are specific to jealousy and 3 metaphorical mappings that are specific to envy. In this study, the variation is not as great: depending on how fine-grained our categorization of metaphorical mappings is, anger, rage, and fury produce either an identical list of metaphorical mappings or an identical list with the exception of one mapping that does not occur with rage. Wrath, on the other hand, clearly stands out; there are as many as 6 mappings with which wrath does not occur.

There are two possible conclusions that can be made from this result: either wrath is not a near-synonym of anger, rage, and fury, or near-synonyms do not always appear with the same metaphorical mappings. Even if the latter conclusion is true, the consequences for MPA are not necessarily severe. The method calls for a representative word, and there are at least two reasons why wrath is not as representative as anger and rage, or even fury.

First, Stefanowitsch (2006b: 71) takes the frequency of a word as a sign of its representativeness. The total occurrences of anger, rage, fury, and wrath in COCA amount to 5752, and wrath accounts for only 6 percent of these, whereas fury accounts for 18 percent, rage for 20 percent, and anger for 55 percent. In terms of frequency, wrath is clearly the least representative word and anger the most representative one.

Second, we might assume that the more general a word’s meaning is, the more representative it is of its domain. Figure 1 below shows that wrath is the most specific of these words, both in terms of its logical denotation (‘a feeling of very strong anger that usually does not last very long’) and the register in which it is used (formal). In this respect, too, anger is the most representative word; it is superordinate to all the three other words.

In fact, near-synonyms occurring with the same metaphorical patterns would only be a problem for MPA when less representative words occur with mappings with which more representative words do not occur. This is clearly not the case with anger, rage, fury, and wrath in COCA, but at least according to the results in Ogarkova (2007), it seems to be the case with jealousy and envy in BNC. If we take jealousy and envy to represent the same domain, it seems that a single representative word might not be enough to map out the all the mappings that structure a target domain. MPA might therefore benefit from further explorations into the metaphors in which near-synonyms occur, and as suggested in Ogarkova (2007: 109), including other word forms than nouns might also add to the mappings that can be uncovered.

Figure 1: Lexical relations between anger, rage, fury, and wrath derived from Macmillan (2012)

Bibliography

Davies, Mark (2008–): The Corpus of Contemporary American English: 425 million words, 1990–present, http://corpus.byu.edu/coca/ (last accessed 31/7/2013).

Davies, Mark (2010–): The Corpus of Historical American English: 400 Million Words, 1810–2009, http://corpus.byu.edu/coha/ (last accessed 31/7/2013).

Esenova, Orazgozel (2009): “Anger metaphors in the English language”, in: Tissari, Heli (Ed.): Approaches to language and cognition (Studies in Variation, Contacts and Change in English 3), Helsinki.

Kövecses, Zoltán (1998): “Are there any emotion-specific metaphors?”, in: Athanasiadou, Angeliki/Tabaskowska, Elzbieta (edd.): Speaking of emotions: conceptualization and expression, Berlin and New York, 127–151.

Lakoff, George/Johnson, Mark (1980): Metaphors we live by, Chicago.

Lakoff, George/Kövecses, Zoltán (1987): “The cognitive model of anger inherent in American English”, in: Holland, Dorothy/Quinn, Naomi (edd.): Cultural models in language and thought, New York, 195–221.

Longman (2012): The Longman Dictionary of Contemporary English Online, http://www.ldoceonline.com/ (last accessed 31/7/2013).

Macmillan (2012): Macmillan Dictionary, http://www.macmillan dictionary.com/ (last accessed 31/7/2013).

Martin, James H. (2006): “A corpus-based analysis of context effects on metaphor comprehension”, in: Stefanowitsch, Anatol/Griers, Stefan Thomas (edd.): Corpus-based approaches to metaphor and metonymy, Berlin and New York, 214–236.

Ogarkova, Anna (2007): ““Green-eyed monsters”: A corpus-based study of metaphoric conceptualizations of jealousy and envy in Modern English”, in: metaphorik.de 13, 87–147.

Partington, Alan (1997): Patterns and meanings: Using corpora for English language research and teaching, Amsterdam and Philadelphia.

Pragglejaz Group (2007): “MIP: A method for identifying metaphorically used words in discourse”, in: Metaphor and Symbol 22, 1, 1–39.

Rojo Lopez, Ana María/Orts Llopis, María Ángeles (2010): ”Metaphorical pattern analysis in financial texts: Framing the crisis in positive or negative metaphorical terms”, in: Journal of Pragmatics 42, 3300–3313.

Steen, Gerard J./Dorst, Aletta G./Herrmann, J. Berenike/Kaal, Anna A./Krennmayr, Tina/Pasma, Trijntje (2010): A method for linguistic metaphor identification: from MIP to MIPVU, Amsterdam and Philadelphia.

Stefanowitsch, Anatol (2006a): “Corpus-based approaches to metaphor and metonymy”, in: Stefanowitsch, Anatol/Griers, Stefan Thomas (edd.): Corpus-based approaches to metaphor and metonymy, Berlin and New York, 1–16.

Stefanowitsch, Anatol (2006b): “Words and their metaphors: A corpus-based approach”, in: Stefanowitsch, Anatol/Griers, Stefan Thomas (edd.): Corpus-based approaches to metaphor and metonymy, Berlin and New York, 63–105.

Tissari, Heli (2003): LOVEscapes: Changes in prototypical senses and cognitive metaphors since 1500, Helsinki.

Appendices

Appendix 1: Metaphorical mappings associated with anger found in the first phase of the analysis

Appendix 2: More metaphorical mappings associated with anger found in the second phase of the analysis

Appendix 3: Metaphorical mappings associated with rage found in the first phase of the analysis

Appendix 4: More metaphorical mappings associated with rage found in the second phase of the analysis

Appendix 5: Metaphorical mappings associated with fury found in the first phase of the analysis

Appendix 6: More metaphorical mappings associated with fury found in the second phase of the analysis

Appendix 7: Metaphorical mappings associated with wrath found in the first phase of the analysis

Appendix 8: More metaphorical mappings associated with wrath found in the second phase of the analysis