The series of studies described in this paper began in an attempt to explore underrepresented aspects of figurative language (FL), namely, how FL is produced by speakers rather than comprehended by readers or listeners and how it operates in natural rather than laboratory contexts. Such studies seemed warranted for a number of reasons. First, much of the research had treated FL as a contextually-independent phenomenon and other research embedded it within contrived contexts, yielding somewhat conflicting results. Second, most laboratory research treated FL as a modular device rather than investigating its relationship to the conceptual system. As such, FL was assumed to be somewhat infrequent, optional rather than efficient or helpful way of speaking, and largely ornamental. Finally, FL activity was limited to what was heard or read and not how it was spoken. By observing naturally produced language, it is possible to better understand the context of its use, to gain insight into the conceptual and pragmatic functions for which it was used, and to see if any aspects of its production varied from what was already understood about its comprehension. The texts for the first three studies include three college lecture courses (Corts & Pollio, 1999; Corts, 1999) and religious speech collected during a number of sermons (Corts, 1999). The final study-- an examination of production in phenomenological interviews-- has just been completed.

At the outset of these studies, it was necessary to develop a methodology for identifying figurative language from a corpus and quantifying changes in production rate. This was accomplished by having trained raters read through transcripts of a sample of speech (in this case, a collection of college lectures) and identifying the number of figurative phrases within each sentence. The sentences were then analyzed by the means of a centered moving average procedure and plotted on a curve. By examining the cumulative output curve and frequencies of the moving averages, regions of particularly high or low figurative output could be identified. In addition, gestures, which heretofore had not been studied alongside figurative language, were coded and subjected to the same moving average analysis. This methodology allows researchers to identify regions of high and low output by manipulating several basic parameters (this procedure is explained in more detail in Corts & Pollio, 1999; Corts, 1999).

Once analyzed in this way, converging evidence from the four separate studies suggested that the natural production of FL occurs in two modes; one which involves concentrated "bursts" while the other appears to be more evenly distributed throughout the corpus. Throughout this series, it has become apparent that this pattern is consistent across a number of contexts and speakers and that the distinction in these two streams is not limited to production rate. Descriptions of the two streams are described by comparing figures within bursts to those falling outside bursts.

Within each corpus, there are a varying number of bursts in which the production of figurative language increases sharply for a matter of two or more sentences and then returns to a somewhat consistent, but significantly lower rate until the corpus comes to a close or another burst is produced. Aside from the increase in production rate, figures of speech produced within bursts were examined on three further dimensions: novel/frozen or idiomatic, topical/ structural (pertaining to the subject under discussion or the act of speaking or lecturing), and metaphorical coherence/independence. Across all speech samples, bursts were significantly more likely to include a collection of novel metaphors that were coherent with a single conceptual metaphor and which represented the topic under consideration by the speaker. Outside of bursts, where FL was produced at a much slower and more evenly distributed rate, figures of speech were significantly more likely to be frozen, they were divided almost evenly between topical and structural, and were as often as not independent metaphors (metaphors unrelated to other figures through conceptual coherence). Finally, in two of the series of studies, the corpus involved a videotaped record. In both studies, it was noted that gestures occasionally would burst alongside metaphor. While gestures are not necessary during figurative language production, it was interesting to note that they often co-occur, and in such cases, the gestures represent the same metaphor as is presented in speech.

While the purpose of these studies is primarily descriptive, they provide some insight into the issues described at the outset of this paper.

(1) It appears that many figures of speech do occur outside of metaphorical context, but these more often are frozen figures of speech and do not seem to address key points in the communication. Instead, most of these figures are used to comment on the structural aspects of the speech, e.g. the speaker's intentions to change topics or to "move on." What is perhaps most important, however, is to realize that when highly novel metaphors are presented, they are presented in the context of a burst with a series of coherent, self-supporting phrases which are intended to explain the conceptual metaphor. Thus, we might generalize that many laboratory studies have little relevance to the manner in which figurative language is actually encountered and comprehended.

(2) In relation to the conceptual system, several interesting points of data emerged. First, the idiomatic figures of speech in the structural stream seem to be lexicalized and therefore, have synonyms and readily available paraphrases (e.g. "These Einsteins" could be replaced with "These geniuses"). Thus, their use seems somewhat arbitrary. Bursts, however, are centered around the important topics of the lecture and the metaphors, as conceptual metaphors, could not be replaced or paraphrased without some effect on the meaning of the speech (e.g. "Drug abuse is a game, not a disease" is more difficult to paraphrase). In bursts, it appears that there is less choice about the vehicle of the metaphor sense the figure of speech represents the topic rather than describes it. Thus, it is likely that this topical stream, including most bursts, originates at a conceptual level. This proposal, however, is difficult to test with the current data.

(3) Finally, the fact that gestures often represent conceptual metaphor constituting a burst is an interesting fact to consider, although it is difficult at present to make any strong generalizations. This observation, along with the realization that novel metaphors tend to be produced in bursts, allows researchers to understand more about the comprehension of metaphor as well since these properties are underrepresented in the experimental literature.