THE ROLE OF METAPHORS IN DESCRIPTIONS OF EMOTIONS 
Andrew Ortony and Lynn Fainsilber 
University of Illinois at Urbana-Champaign 
Why do we use metaphors? For nearly 2000 years, the most generally accepted answer was that people 
only use metaphors for rhetorical purposes. Metaphorical language was thought to be merely ornamental -- the 
seasoning of language, exploited for effect by poets and politicians 1, as compared with the cold factual language 
of the scientist. This view, however, is now no longer accepted (see, for example, Gentner, 1982; Boyd, 1979). 
It is now assumed, at least by psychologists and linguists, that metaphors, and their close cousins, analogies, are 
important tools of cognition and communication, providing us with unfamiliar ways of conceptualizing familiar 
things, and familiar ways of conceptualizing unfamiliar things (Lakoff & Johnson, 1980; Ortony, 1979; 
Vosniadou & Ortony, in preparation). Yet, what is still assumed, rather than demonstrated, is that nonliteral 
uses of language are sometimes necessary for accomplishing such goals, rather than merely convenient or 
elegant ways of doing so. In this paper we present a sort of empirical existence proof that there are some things 
whose descriptions appear to invoke much more use of metaphorical language than others. This, while not 
establishing the necessity of metaphors, certainly is a first step. 
In theory, there are at least three communicative functions that metaphor might serve(Ortony 1975). 
First, they might allow one to express that which is difficult or impossible to express if one is restricted to literal 
uses of language. Evidence for this "inexpressibility" claim would constitute encouraging support for the 
necessity-of-metaphors view. A second possible function of metaphors is that they may constitute a particularly 
compact means of communication. Although conscious experience is continuous in form, the linguistic system 
we use to talk about it is comprised of discrete elements (lexical items). Unlike more literal forms of language, 
metaphor may enable us to convey a great deal of information in a succinct manner by obviating the need to 
isolate the predicates to be expressed into their corresponding lexical representations. Finally, metaphors may 
help capture the vividness of phenomenal experience. If metaphors convey chunks of information rather than 
discrete units, they can paint a richer and more detailed picture of our subjective experience than might be 
expressed by literal language. This we call the "'vividness" claim. 
In this paper we shall concentrate on the first and last of these possible functions. In order to do so, we 
need to examine a discourse domain for which a prima facie case can be made for supposing that literal 
language will often be inadequate and which lends itself to variations in vividness. There doubtless are many 
such domains. The one that we selected was that of internal states, in particular, emotional states. The 
literature on the linguistic expression of emotions suggests a relatively high incidence of figurative language use 
(e.g., Davitz, 1969), providing pragmatic reasons for believing that the context of (linguistic) emotional 
expression may be a profitable one within which to study metaphor production. Emotional states seemed well- 
suited for our purposes because they tend to have an elusive, transient quality that is difficult to describe using 
literal language, although, of course, they can usually be labeled using literal language. Thus, while it might be 
easy for a person to label an emotional state as, for example, "fear," it is difficult to provide a literal 
description of the quality of some particular experience of fear. Furthermore, because emotions vary in 
intensity, one might expect differential levels of vividness. 
There seem to be two possible ways in which people might try to communicate the quality of an 
emotional state. First, a speaker might use literal language to describe the events that triggered the emotional 
state and hope that the hearer correctly infers how he or she felt. For example, a person might describe the 
details of being mugged, hoping that a listener would recognize the emotional experience as the type one would 
have if one were attacked by a mugger. In such a case, the literal description would not describe the quality Of 
lWinston Churchill once renmrked: "How infinite is the debt owed to metaphors by politicians who want to speak strongly but are not 
sure what they are going to say"! 
181 
the subjective state itself but would merely identify its eliciting conditions (Ortony, Clore & Collins, in 
preparation). Alternatively, a speaker might use a metaphor in an attempt to describe the quality of an 
emotional state. For example, one might say that one felt as though one's insides were a butter churn. Here, 
the metaphorical description does represent an attempt to characterize the quality of a subjective state. 
Although we think that emotions constitute a good domain for studying metaphor production, it does not 
follow that the use of metaphorical description will be equally prevalent for different facets of emotions. 
Emotion theorists frequently attribute differential significance to the subjective experience of emotion (De 
Rivera, 1977) or to their associated actions or action tendencies (Frijda, in press). It may be that the subjective 
experience of an emotion can benefit more from a metaphorical description than the associated action or action 
tendency. Consider the subjective experience of some specific case of anger. The quality of such a subjective 
state cannot be publicly observed. In contrast, the actions to which an anger experience might give rise, for 
example, pounding one's fist on the table, are publicly observable. Thus, one might expect people to employ 
more metaphorical descriptions when trying to characterize the subjective experiential quality of emotional states 
than when trying to characterize the overt behaviors associated with such states. The intensity of emotions that 
might also be expected to influence the use of metaphorical language. It is possible that relatively mild 
emotional states are sufficiently unremarkable that speakers are more willing to settle for simply labeling them, 
whereas the vividness of intense emotional states might sometimes generate a more pressing need for detailed 
description. 
To investigate some of these issues we ran a simple experiment in order to examine the production of 
metaphors during descriptions of emotional states and events. We predicted that people would be more likely to 
use metaphors and metaphorical comparisons when describing how they felt when they were experiencing an 
emotion than when describing what they did when they experienced it. We also thought it possible that more 
metaphorical language would be used in descriptions of intense as compared to mild emotional states. The two 
hypotheses combined could be construed as predicting an interaction of description type (feelings vs actions) and 
intensity, with the intensity factor having a greater effect on feeling descriptions than on action descriptions. 
Descriptions of feeling states, which may already make use of metaphorical language, may be especially likely 
to use metaphors when the states are intense. On the other hand, it could be argued that although intense 
emotions are more vivid than less intense ones, the associated actions do not necessarily enjoy a corresponding 
increase in vividness. This is admittedly a tenuous argument, so the prediction of an interaction between 
description type and intensity is made with less confidence than the predictions of main effects for these 
variables. Finally, in the experiment to test these hypotheses, the valence of the emotions was manipulated to 
determine whether this factor has any systematic effect on metaphor use. 
Subjects were asked to describe either how they felt when they experienced certain emotions, or what they 
did when they experienced them. The emotions used included four positive ones (happiness, pride, gratitude, 
and relief) and four negative ones (sadness, fear, resentment, and shame). Note that the particular hypotheses to 
be tested do not depend in any important way on exactly which emotions are used. In addition to providing 
descriptions involving emotions of different valence, subjects were required to describe situations involving 
either very intense experiences of them or very mild ones. 
Metaphors were identified in the transcripts of interview sessions. Protocols were scored in terms of idea 
units (Johnson, 1970) because metaphors are generally better conceptualized as single ideas than as individual 
words. Metaphor production was then measured in terms of the proportion of all distinct idea units that were 
metaphorical in nature. In other words, the measure of metaphor production was the ratio of metaphor types to 
the total number of idea unit types appearing in a protocol 2. A variety of considerations led us to operationalize 
metaphor production in this manner. We were concerned that possible systematic differences in the amount of 
verbal output produced during descriptions of the different emotion-inducing events might contaminate the 
~3ther indices were also used, such as the absolute number of metaphor types and the proportion of the total number of idea units that 
were metaphorical. The choice of measure made little difference to the pattern of results 
182 
measure of metaphor production, such that high verbal output might lead to a high production of metaphor, and 
low verbal output might be associated with little metaphor use. If so, metaphor production would be a 
consequence of verbal output per se and this effect might conceal any differential use of metaphor during 
descriptions of feelings and actions. By looking at the ratio of metaphor types to the sum of both metaphorical 
and non-metaphorical idea units, the potential confounding of metaphor production and amount of linguistic 
output was partially avoided. A second concern was that subjects' tendency to repeat words and phrases during 
an oral account might artificially inflate the measurement if metaphorical tokens as opposed to types were used. 
The results showed that a significantly greater proportion of metaphors occurred in descriptions of feeling 
states (17%) than in descriptions of actions (4%) 3 . Furthermore, the mean percentage of metaphor types used in 
descriptions of intense emotions (12%) was significandy greater than in descriptions of mild ones (9%). Two 
factors (intensity and valence) interacted with the type of description (feelings versus actions). First, there was a 
significandy greater increase in metaphor production when describing the feelings associated with intense 
emotions than when describing the actions associated with intense emotions. Second, although of less 
theoretical interest, while in the description of actions there was a tendency for more metaphors to be produced 
for negative than for positive emotions, this trend was reversed for descriptions of feelings. The patterns of 
these interactions are shown in the Tables below which show the percentage of idea unit types that were 
metaphor types. 
Description 
Type 
Intensity 
Mild Intense 
Feelings 14.7 19.6 
Actions 3.5 4.5 
Valence 
Description Positive Negative 
Type 
Feelings 18.7 15.6 
Actions 3.2 4.9 
The results also revealed that there were eight times as many frozen, or dead, metaphors as there were 
novel ones. More interesting, however, is the fact the the ratio of novel to frozen metaphors was greater for 
intense emotions (12%) than for mild ones (8%), suggesting perhaps that when people are experiencing intense 
feeling states, they are more likely to generate striking and complex metaphors to explain how they feel. To the 
extent that novel metaphors are more metaphorical than frozen ones, and assuming that intense emotional states 
are more vivid than mild ones, this finding of more novel metaphors for intense emotions adds support to the 
vividness claim because it suggests a qualitative as well as quantitative increment in metaphor use. 
To summarize, we have offered evidence that metaphorical language may make it possible for people to 
convey what would otherwise be difficult or impossible to express. This seems to be the case with the quality 
of unobservable internal states like emotions, as evidenced by our results showing the predominance of 
metaphorical language during descriptions of feeling states as opposed to actions, especially when those states 
are intense. So, for example, when one of our subjects reported that he felt like "a storm was brewing inside," 
he succeeded in conveying a particular quality of his subjective experience that is richer, more vivid, and more 
specific than could have been conveyed had he merely labeled the experience as "resentment." For the most 
part, the types of metaphors that people used to describe their emodons were figurative forms that have become 
conventionalized in the English language. When novel metaphors were used, they seemed to be particularly 
evident in descriptions of intense feeling states. Taken together, our results suggest that the inclination of 
psychologists and linguists to reject the classical Aristotelian view of metaphor as merely linguistic decoration, 
in favor of a view that accords it an indispensable communicative function is empirically, as well as 
theoretically, supportable. 
3i.e., the percentage of metaphor types averaged across emotions and subjects. 
183 
References 
Boyd, R. (1979). Metaphor and theory change: What is "metaphor" a metaphor for? In A. Ortony (Ed.) 
Metaphor and thought. New York: Cambridge University Press. 
Davitz, J. R. (1969). The language of emotion. New York, N.Y.: Academic Press. 
De Rivera, J. (1977). A structural theory of the emotions. New York: International Universities Press. 
Frijda, N. (in press). The emotions. New York: Cambridge University Press. 
Gentner, D. (1982) Are scientific analogies metaphors? In D. Miall (Ed.), Metaphor: Problems and 
perspectives. Brighton, England: Harvester press. 
Johnson, R.E. (1970). Recall of prose as a function of the structural importance of the linguistic unit. Journal of 
Verbal Learning and Verbal Behavior, 29, 12-20. 
Lakoff, G. & Johnson, M. (1980). Metaphors we live by. Chicago: University of Chicago Press. 
Ortony, A. (1975). Why metaphors are necessary and not just nice. Educational Theory, 25, 45-53. 
Ortony, A. (Ed.). (1979). Metaphor and thought. New York: Cambridge University Press. 
Ortony, A., Clore, G. L. & Collins, A. (in preparation). Principia pathematica: Some cognitive foundations of 
the emotions. 
S. Vosniadou & A. Ortony (Eds.). (in preparation). Similarity and analogical reasoning. New York: 
Cambridge University Press. 
184 
