Pragmatic verbal repetition: review and application of a new method of quantification

1 Prologue

The spark for this research was ignited by a compelling experience with a client in the speech clinic. This young adult carries a diagnosis of autism, but, like many such individuals, has few typical autistic characteristics and several non-typical ones. Early on, his penchant for verbal repetition was noted (see Appendix). Once a treatment plan was underway, the range of responses by co-participants to the repetition behavior became intriguing. These responses fell into three categories: pleasant satisfaction over a sense of good rapport; concern that repetition was interfering with communication; and annoyance. In some instances, in an effort to support the talk, the clinician repeated the repeated utterance of the client. The ability of repetition to affect the interaction in these different ways lead to such questions as these: Why was the positive response to repetition in conversational interaction with this client so easily refreshed? Why did the clinicians’ responses undergo such swings? How did it happen that the clinician also engaged in repetition? The answer must be that a certain amount of repetition by the co-participant is normal and desirable in conversation, and that at a certain point, it becomes a bit too much, and then, way too much.

At the same time, another conversation highlighted the impact of repetition in normal discourse during an exchange between one of the authors (DS) and a stranger at Broadway and Fifth Avenue in Manhattan, New York (example [1], repeated elements in italics).

These observations in disordered and normal speech led us to the scholarly literature, where, in addition to standard overviews of pathological repetition, there was supportive commentary that repetition is prevalent and rife with functionality in normal conversation. However, guidelines about how much verbal repetition normally occurs in given discourse contexts did not emerge, nor were there criteria on how to distinguish normal from abnormal repetition.

This study presents a method whereby verbal repetition phenomena can be reliably identified, measured, and characterized, so that more light might be cast on the quantity and types of repetition in normal and abnormal discourse. The paper begins with a review of repetition studies in normal and disordered speech in several domains of the language sciences. A newly devised method of measurement is applied to three texts of normal discourse and, finally, to an excerpt from the non-normal sample that stimulated the study.

2 Introduction and review

Verbal repetition has been treated with varying interest and from disparate viewpoints across the language sciences. In modern generative linguistics, given its goal of establishing the grammar of a language in its abstract representation, repetition has not been a topic of interest. Sociolinguistic research has provided numerous insights about the various kinds of verbal repetition along with descriptive commentary of their function (Johnstone 2002; Norrick 1987; Tannen 1987b; Wong 2000; Labov 1972). But systematic measures of repetition behaviors across different kinds of texts have not been undertaken. In this article, we use the term “pragmatic repetition,” to refer to verbal repetition as it occurs during normal language use, or repetition in the pragmatics of communication. We review other uses of repetition in verbal behavior.

Our goal was to clarify the properties of verbal repetition, targeting words, phrases, clauses, and sentences, as characteristic in ordinary language use. Repetition behaviors in normal language use appear in many forms and serve several functions, some of which have been anecdotally described in the sociological literature. While interest in normal repetition has grown, what remains lacking is an empirical method for reliably establishing quantities, characteristics, and functions of repetition in discourse. Repetition is a universal feature of the language arts (Tannen 1990). The Hilburn cartoon in Figure 1 depicts visual as well as verbal repetition. The cartoon shows repetition of the phrase “All very compelling”, ^[1] which in this context enhances the theme of judging the entries and contributes to the gravity of the occasion by implying hesitancy and thoughtfulness in the judgment process.

Figure 1

Argyle Sweater cartoon. Repetition of the phrase “All very compelling” in this context enhances content, contributing to the gravity of the occasion, implying hesitancy and thoughtfulness in the judgment process. The verbal repetition mirrors the visual replication.

While it is observationally the case that repeated verbal phenomena are frequent and natural in talk, it follows that notions of too much or too little repetition, relative to an established normal range in language practice, can be usefully proposed, but until now no quantitative approaches have appeared. Our purpose in this review is to bring together and discuss these studies and to expand the discourse on verbal repetition. A detailed methodology is offered as a technique for determining incidence and functions of repetition in discourse. Using this method, results from our empirical studies of verbal repetition are presented. It is believed that an understanding of verbal repetition will benefit descriptions of pragmatics in human communication. A major use of repetition is learning new information as a component of first and second language learning and classroom learning. It is a common practice to repeat when trying to memorize or rehearse new material. Rote repetition is not considered here.

3 Repeated elements: the method

4 Results

Repeated morphemes for the three texts were 205/2,035 for the screenplay, Some Like it Hot, 445/2,858 for the reality show, and 341/2,410 for the telephone conversation. Figure 2 displays the proportion of repeated morphemes out of total morphemes for the three texts.

Figure 2

Percentage of repeated morphemes in three discourse samples: the screenplay, Some Like it Hot (SLIH), the reality show (RS), and the telephone conversation (TC).

The smallest proportion of repetition occurred in the screenplay sample (9%) compared to the reality show (19%) and the telephone conversation samples (22%). Chi-square statistics were used to compare percentage of repetition across groups. Nonparametric measures were used since assumptions are being made using small samples which were not randomly selected. There were no group differences in the proportion of repeated morphemes between samples, x² (df=2, N=44)=3.59, p=0.166.

Descriptive information was also obtained for formal factors related to repetition: unit, localness, preservation, source, and formulaicity (see Table 1). For unit, repetition most frequently occurred at the word or phrase level. In the screenplay sample 60% of repetitions occurred in words. According to chi-square data, x² (df=2, N=126)=11.62, p=0.003, the other samples had significantly fewer word repetitions, with 34% of the reality show sample and 32% of the telephone conversation samples comprised of word repetitions. Significant differences were also noted comparing the proportion of phrase repetitions between groups, x² (df=2, N=118), 6.83, p=0.033. Proportion of phrase repetition was higher for the reality show sample (45%) and the telephone conversation sample (47%) and lower for the screenplay sample (26%). Lower numbers of repetitions occurred at the clause and sentence levels across groups. Clause repetition occurred most frequently in the telephone conversation sample (9%), with similar proportions to the reality show sample (8%). Clause repetition occurred least frequently in the screenplay sample (4%). Clause and sentence repetitions were not significantly different.

Data on localness of repetition revealed differences between groups for the proportion of immediate, x² (df=2, N=118)=23.61, p=0.000, and distant repetitions, x² (df=2, N=96)=13.00, p=0.002, but not for delayed repetitions, x² (df=2, N=86)=3.65, p=0.161. Immediate repetitions occurred frequently in the reality show sample (56%) and screenplay sample (47%), but less frequently in the telephone conversation sample (15%). However, distant repetitions occurred frequently in the telephone conversation sample (48%) and to a lesser extent in the screenplay sample (28%) and reality show sample (20%) (see Figure 3).

Figure 3

Measures of localness: percentages of repetitions that were immediate, delayed, or distant in three discourse samples: the screenplay, Some Like it Hot (SLIH), the reality show (RS), and the telephone conversation (TC).

Preservation of the initially produced form (prototype) was compared between groups. Variants of the initial prototype were most frequently identical, with the telephone conversation sample having the most identical (81%) and the screenplay and reality show samples having slightly lower proportions of identical productions (63% and 60%, respectively). There was no difference between groups with regard to preservation, x² (df=2, N=204)=3.79, p=0.150.

Results for the source of the repetition (whether repetitions were more likely to be of self or another speaker) did not yield significant contrasts.

With regard to proportion of repetitions identified as formulaic, there was a trend toward differences between groups, x² (df=2, N=187)=5.91, p=0.052. Out of the total repeated morphemes, the most formulaic repetitions were found in the telephone conversation sample (75%). Formulaic repetitions were also demonstrated in high proportion in the reality show sample (64%) and to a lesser extent in the screenplay sample (48%).

4.1 Results for functional categories

Repetitions were categorized based on one of three discourse functions: maintain form, enhance content, or foster social aspects of conversation. Form repetitions did not occur in the screenplay sample (0%) most likely due to literary artifice and editing for accuracy and maintaining flow of speech. Form repetitions occurred most frequently in the telephone conversation sample (40%) and to a lesser extent in the reality show sample (14%), yielding a significant difference between groups, x² (df=2, N=55)=43.02, p=0.000.

A significant difference was also noted between groups regarding the proportion of repetitions used to enhance content of talk, x² (df=2, N=117)=17.59, p=0.000. Content repetitions occurred most frequently in the screenplay sample (58%). They were less frequent in the reality show sample (38%) and least frequent in the telephone conversation sample (21%). Significant differences between groups were not demonstrated for use of repetition for the purpose of socialization, x² (df=2, N=139)=0.619, p=0.734. Proportions were similar across groups regarding repetition for socialization in the screenplay (42%), telephone conversation (45%), and reality show (49%) samples (see Figure 4).

Figure 4

Functions of repetition by discourse sample: the screenplay, Some Like it Hot (SLIH), the reality show (RS), and the telephone conversation (TC).

5 Discussion

In this paper, the intent was to position normal verbal repetition in the larger context of sociolinguistics, second and child language acquisition, and communicative disorders by providing a brief review of relevant perspectives arising from these disciplines. The term “pragmatic repetition” was coined to distinguish repetition used for pragmatic purpose from repetition featured in classification of aphasia, learning new material, and childhood imitation. Motivated by a desire to seek benchmarks for normal and abnormal repetition, a method was designed to quantify repetition in verbal discourse using the morpheme as elemental unit of measurement.

On applying this method to three different discourse texts, it was seen that repeated morphemes made up an average of 15% of total morphemes with a range between 9% and 19%. This finding reveals the actual presence of repetition in naturalistic (or simulated naturalistic) speech, and implies that although human language has the potential to be infinitely creative, in actual practice, it also contains a considerable amount of material which has been previously uttered.

Across the three texts, approximately 68% of the total repetitions were identical to the original utterance, with the highest number, 81%, appearing in the nonscripted telephone conversation. Further, it was seen that large proportions of the repetitions were formulaic expressions: in the telephone conversation transcript, 78% of the repetitions were of formulaic expressions. Together with the fact that formulaic language itself has been estimated to make up about 25% of discourse, the position for all original utterances in naturalistic speech continues to erode (Pinker 1995; Van Lancker Sidtis 2004). The substantial presence of formulaic speech (averaging 25%) as well as repetition of both novel and formulaic utterances (about 17.5% for two spoken texts) add up to a considerable proportion of discourse which is not newly created in these samples.

Analysis of differences between the three discourse samples studied here revealed similarities and differences. The screenplay sample had the smallest percentage of repetition. This lower incidence is likely due to the scripted nature of the text, lacking the spontaneity found in actual conversational speech. Differences were also found for the repetition variables representing form and function. For the screenplay and the reality show samples, repetitions were most likely to occur immediately following the target word or phrase. In the telephone conversation sample, repetitions were least often immediate and most often distant. The telephone conversation sample confirms the importance of distant repetitions. The distant repetitions appeared to perform social functions, including bonding between subjects and mutual endorsement of the selected verbal repertory. Examination of data on preservation of repetition form revealed high proportions of identical repetitions. Repetition of clauses and sentences occurred in similar proportions across samples. Differences were noted, however, in the proportion of word and phrase repetitions between groups. Once again, the two unscripted samples (reality show and telephone conversation) presented with similarities distinguishing them from the scripted sample (screenplay). Phrase repetitions were used most frequently in the unscripted samples, with word repetitions close behind. In the scripted sample, most repetitions occurred at the word level. These results provide useful information for identifying characteristics which distinguish scripted dialog from actual conversational discourse.

A higher percentage of self-repetition was seen in the telephone conversation sample, involving only two speakers, whereas the reality show and screenplay exhibited more repetition of other. For example, in the telephone conversation sample, stereotypical utterances were repetitively used by one speaker, possibly to perform the social function of asserting self in the conversation. In contrast, speakers also repeated expressions initiated by other speakers, likely performing the function of bonding, in the sense of acknowledging the choice of vocabulary of the other speaker. In this sample, each speaker repeated expressions used by the other. This type of repetition may be said to convey endorsement of the choice of words of each speaker, signaling mutual affirmation. This process can also be seen as attributing mental states, as in the theory of mind in human interaction. Example (8) provides instances of this phenomenon from the telephone conversation sample. Repetitions were both distant and delayed and constituted formulaic expressions (pretty crazy, it’s crazy).

In the screenplay, repetitions were used to enhance form of talk significantly less than the other two samples. Form repetitions were used most frequently in the unedited and unscripted telephone conversation sample. Repetitions to enhance the content of speech were least likely to occur in the telephone conversation. This emphasizes the casual nature of the conversation and reinforces the notion that during casual conversation, informational content may be less important than the social aspects of the interaction. In contrast, in the screenplay and reality show samples, content repetitions were more prevalent. This is likely related to storylines that need to be conveyed for entertainment purposes. Repetitions to enhance content of talk also included repetitions for emphasis, agreement, and echo questions as described by Parker and Pickeral (1985).

All three samples used high proportions of repetition for the function of socialization and percentages were similar across groups. In the reality show and telephone conversation samples, repetition for socialization was used more frequently than for form or content. The screenplay also used a high proportion of repetition for socialization. This was largely due to the comedic nature of the script. Several of the repetitions fell into the subcategory of humor, while the socialization repetitions in the reality show and telephone conversation samples fell mainly into the subcategories of bonding and assertion of self as well as acknowledgment of the other.

A trend toward differences was noted in the proportion of repetitions categorized as formulaic. In the telephone conversation sample 75% of repetitions were formulaic, in the reality show sample 64% of repetitions were formulaic, and in the screenplay sample, 48% of repetitions were formulaic. It is likely that the spontaneous, informal nature of the telephone conversation sample increased the use of stereotypical expressions.

The findings of this exploratory study have helped establish repetition as a vital and measurable pragmatic component of language. The similarities and differences between samples allow for a better understanding of pragmatic repetition during conversational discourse. Incidence of repetition was similar across samples. Similarities between samples were most evident when comparing the reality show and telephone conversation which were unscripted. Many of the differences between samples arise from the status of Some Like It Hot as a screenplay. This work is in agreement with previous studies (Tannen 1982, Tannen1989) showing how scripted text is uniquely different from actual spontaneous productions. Overall, the findings of this study provide a substantive foundation for future quantitative research on pragmatic repetition.

6 Postlogue

This research responded to a need to establish normal incidence and functionality of verbal repetition. The need arose from a relationship of a speech-language pathologist with a client who conversed amicably and cooperatively with elementary cognitive competence, but who repeated his own and the clinician’s utterances to his detriment. Repetition in client’s speech sample given in Appendix I was easily quantified: of 117 morphemes in the excerpted sample, 84 were repeated, yielding a proportion of 72% repetition, well above the values (averaging 15%) reported for the three normal samples used in this analysis. The goal of the methodology described here is to allow language scientists to evaluate repetition behaviors in discourse of all kinds: normal (extending to various styles of spoken and written text), developmentally anomalous, and neurogenically disordered, and to identify key characteristics and functions of this important component of human communication.