Previous Utterance Research Articles

IDENTIFYING THE IMPACT OF NEGATIVE FEEDBACK AND LEARNERS' RESPONSES ON ESL QUESTION DEVELOPMENT

Swain's (1985, 1995, 2000) output hypothesis states that language production is facilitative of second language (L2) learning. An important component of the output hypothesis involves pushing learners to produce appropriate, accurate, and complex language (Swain, 1993), which may occur when interlocutors provide learners with negative feedback (Gass, 1997, 2003; Long, 1996; Mackey, in press; Pica, 1994; Swain & Lapkin, 1995). When learners modify their previous utterances in response to negative feedback, learning opportunities are created by both the provision of negative feedback and the production of modified output. Consequently, it is difficult to determine how these interactional features—alone or in combination—positively impact L2 development. The current study examines the impact of negative feedback and learners' responses on English as a second language (ESL) question development, which is operationalized as stage advancement in Pienemann and Johnston's developmental sequence for ESL question formation (Pienemann & Johnston, 1987; Pienemann, Johnston, & Brindley, 1988). Thai English as a foreign language (EFL) learners (n = 60) carried out a series of communicative tasks with native English speakers in four conditions that provided different negative feedback and modified output opportunities and also completed four oral production tests over an 8-week period. Analysis of the treatment data identified the amount of modified output involving developmentally advanced question forms produced by the learners, and analysis of the test data revealed whether the learners' stage assignment changed over time. Logistic regression indicated that the only significant predictor of ESL question development was the production of modified output involving developmentally advanced question forms in response to negative feedback.I am grateful to Alison Mackey for her insightful comments on this paper and on the dissertation research on which it is based. I also thank Rhonda Oliver, Jeff Connor-Linton, Jennifer Leeman, Jenefer Philp, Ana-Maria Nuevo, and the anonymous SSLA reviewers for their valuable comments. Any errors, of course, are my own.

Perception of utterance relatedness during the first-word-period

Interactions between six toddlers (aged 1;0 to 1;6) and adults were examined to ascertain adult perceptions of toddler utterance relatedness and to determine temporal and interactional features that underlie those perceptions. Five raters made judgments regarding relatedness of the child utterances to the previous adult utterances; 251 utterances were examined. Utterances judged by adults as related occurred within 4.25 seconds of the preceding adult utterance nearly 90% of the time. This study also points to the need for using interactional categories that go beyond describing utterance relatedness, and introduces terms (i.e. coparticipatory, initiation, narrowed focus) for doing so.

Temporal change in memory of human voice

This study investigates the extent to which human subjects accurately retain memory of speaker individuality and voice quality of utterances without receiving instruction to memorize them. Experiments constituted three phases. In phase one, 27 subjects listened to two target utterances in a paired-comparison test that required only a similarity judgment of voice quality. The target utterances were one sentence read by two males with average and characteristic voice quality, respectively, who were selected from 109 males. In phase two, performed 21 days after phase one, the same subjects evaluated Japanese expressions of voice quality [Proc. ICSLP-98, No. 1005] of the two previous target utterances based on memory only and judged with a degree of confidence whether each of seven speech stimuli (two targets and five additional speakers’ utterances) was identical to one of the two targets. In the final phase, conducted immediately after phase two, the subjects repeated the same phase-two task but after listening to the two target utterances. Data processing based on the signal detection theory revealed that temporal change in memory of speaker individuality and voice quality was relatively small over a period of 21 days. Experiments over a longer period of time are now underway.

Conversational User Interfaces

Self-service systems, online help systems, web services, mobile communication devices, remote control systems, and dashboard computers are providing ever more functionality. However, along with greater functionality, the user must also come to terms with the greater complexity and a steeper learning curve. This complexity is compounded by the sheer proliferation of different systems lacking a standard user interface. Conversational user interfaces allow various natural communication modes like speech, gestures and facial expressions for input as well as output and exploit the context in which an input is used to compute its meaning. The growing emphasis on conversational user interfaces is fundamentally inspired by the aim to support natural, flexible, efficient and powerfully expressive means of human-computer communication that are easy to learn and use. Advances in human language technology and intelligent user interfaces offer the promise of pervasive access to online information and web services. The development of conversational user interfaces allows the average person to interact with computers anytime and anywhere without special skills or training, using such common devices as a mobile phone. Advanced conversational user interfaces include the situated understanding of possibly imprecise, ambiguous or incomplete multimodal input and the generation of coordinated, cohesive, and coherent multimodal presentations. In conversational user interfaces the dialogue management is based on representing, reasoning, and exploiting models of the user, domain, task, context, and modalities. These systems are capable of real-time dialogue processing, including flexible multimodal turn-taking, backchanneling, and metacommunicative interaction. One important aspect of conversations is that the successive utterances of which it consists are often interconnected by cross references of various sorts. For instance, one utterance will use a pronoun to refer to something mentioned in the previous utterance. Computational models of discourse must be able to represent, compute and resolve such cross references. Conversational user interfaces differ in the degree with which the user or the system controls the conversation. In directed or menubased dialogues the system maintains tight control and the human is highly restricted in his dialogue behavior, whereas in free-form dialogue the human takes complete control and the system is totally passive. In mixed-initiative conversational user interfaces, the dialogue control moves back and forth between the system and the user like in most face-to-face conversations between humans. Four papers in this special issue deal with conversational user interfaces that use speech as the main mode of interaction. The paper by Helbig and Schindler discusses state-of-art component technologies and requirements for the successful deployment of conversational user interfaces in industrial environments such as logistics centers, assembly lines, and car inspection facilities. It shows that the speech recognition rate in such environments is still depending on the correct positioning and adjustment of the microphone and discusses the need for wireless microphones in most industrial applications of spoken dialogue systems. Block, Caspari and Schachtl describe an innovative dialogue engine for the Virtual Call Center Agent (ViCA), that provides access to product documentation. A multiframe based dialogue engine is introduced that supports natural conversations by allowing over-answering and free-order information input. The paper reports encouraging results from a usability test showing a high task completion rate. The paper by te Vrugt and Portele describes a tasked-oriented spoken dialogue system that allows the user to control a wide spectrum of infotainment applications, like a hard-disk recorder, an image browser, a music player, a TV set and an electronic program guide. The paper presents a flexible framework for such a multi-application dialogue system and an applicationindependent scheme for dialogue processing. Noth et al. describe lessons learnt from the implementation of three commercially deployed conversational interfaces. The authors propose five guidelines, which they consider to be crucial, when building and operating telephone-based dialogue systems. One of the guidelines concerns the fact that a spoken dialogue system must react fast to any kind of user input, no matter

Lost in thought: cognitive load and the processing of addressees' feedback in verbal communication.

Two experiments tested if cognitive load interferes with perspective-taking in verbal communication even if feedback from the addressee is available. Participants gave instructions on the assembly of a machine model. In Experiment 1, cognitive load was demonstrated to be a function of the complexity of assembly steps. In Experiment 2, position of feedback (during simple vs. during complex steps) and type of feedback (question vs. ambiguous interjection) were manipulated. With simple steps, speakers' responses were a function of feedback type. Speakers responded differently to questions than to interjections. With complex steps, however, responses were a function of cognitive load. Regardless of the type of feedback, most speakers simply repeated their previous utterances.

A statistical language model for conventional speech reflecting the previous utterance of the other participant

AbstractThis paper proposes the construction of a language model for speech recognition in speech translation. The proposed language model is intended to improve speech recognition performance by taking account of the previous utterance content of the counterpart speaker. The previous utterance content of the counterpart speaker is represented by the intermediate language which is generally used by C‐star (Spoken Language Translation Research Group). The proposed model is based on a language model that is dependent on the intermediate language representation. An experiment concerning the conversation of hotel reservations shows that the rate of incorrect word recognition is reduced by approximately 5.4% (from 14.7% to 13.9%), and the usefulness of the proposed model is demonstrated. © 2003 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 87(1): 54–62, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.10211

Early syntactic creativity: a usage-based approach

The aim of the current study was to determine the degree to which a sample of one child's creative utterances related to utterances that the child previously produced. The utterances to be accounted for were all of the intelligible, multi-word utterances produced by the child in a single hour of interaction with her mother early in her third year of life (at age 2;1.11). We used a high-density database consisting of 5 hours of recordings per week together with a maternal diary for the previous 6 weeks. Of the 295 multi-word utterances on tape, 37% were 'novel' in the sense that they had not been said in their entirety before. Using a morpheme-matching method, we identified the way(s) in which each novel utterance differed from its closest match in the preceding corpus. In 74% of the cases we required only one operation to match the previous utterance and the great majority of these consisted of the substitution of a word (usually a noun) into a previous utterance or schema. Almost all the other single-operation utterances involved adding a word onto the beginning or end of a previous utterance. 26% of the novel, multi-word utterances required more than one operation to match the closest previous utterance, although many of these only involved a combination of the two operations seen for the single-operation utterances. Some others were, however, more complex to match. The results suggest that the relatively high degree of creativity in early English child language could be at least partially based upon entrenched schemas and a small number of simple operations to modify them. We discuss the implications of these results for the interplay in language production between strings registered in memory and categorial knowledge.

Lexical cohesion in multilingual conversation

Many studies have shown intersentential codeswitching to be related to conversational structure. In this paper, I argue that insertion can be explained in these terms as well. Drawing on Halliday's and Hasan's(1976) notion of cohesive tie, I claim that insertions are a consequence of the bilingual speaker's attempt to create coherence between utterances in different languages. By repeating a lexical item from a previous utterance even if the language of interaction has changed, a speaker establishes lexical cohesion between the two utterances. Interpreting insertion as the result of lexical cohesion serves to explain some crosslinguistic characteristics of insertion, namely the dominance of nouns(as lexical cohesion is basically restricted to open class items), and the asymmetry between the languages(as lexical choice is influenced by the context in which a lexical item is used and language choice is restricted in many contexts). Furthermore, the analysis eliminates the need to distinguish between loanwords, nonce-borrowings, Furthermore, the analysis eliminates the need to distinguishbetween loanwords, nonce-borrowings, or single-item codeswitches, as a lexical item is no longer defined in relationship to the lexicon of the language in whose context it occurs, but rather by the cohesive tie in which it participates.

Spanish no, sí: Reactive moves to perceived face-threatening acts, Part II

This paper is the second installment of a two-part pragmatic analysis of a previously uninvestigated discourse particle in Spanish, no, sí. In this paper, we examine the functions of this particle in conversational interaction and argue that it represents a pragmatic response to a perceived face-threatening act, whether intended as such or not. Our pragmatic analysis demonstrates that, as a conversational management strategy, no, sí functions as an expression of an evaluative process in which the addressee interprets and reacts to the pragmatic value of the previous utterance. Following principles of politeness in Brown and Levinson (1987), no, sí is a reaction to a perceived intent, and hedges the actual response it initiates, serving as an expression of footing to seek common ground or alignment with the participant. In the framework of conversation analysis (Schegloff, 1988), it is a marker of a dispreferred first and second part in an adjacency relationship.

Phrasal Unit Boundaries and Organization of Turns and Sequences in Korean Conversation

This paper examines an aspect of the grammar-interaction interface with respect to how participants orient to intra-turn phrasal unit boundaries as a locus that has interactional import for turn and sequence organization in Korean conversation. Phrasal unit boundaries in Korean serve as a space within a turn in which the speaker of the turn in-progress invites the recipient to acknowledge the speaker's point expressed up-to-that-point and collaboratively display his/her understanding thereof. In a sequentially and topically 'ripe' context, such unit boundaries often constitute places where more active participation on the part of the recipient is invited in the form of collaborative completion in which the recipient co-constructs the primary speaker's on-going turn. Phrasal unit boundaries also provide interactional resources which the speaker may exploit to sequentially delete out a problem in talk, e.g., by a sort of tying operation in which a subsequently added phrasal unit ties back to the speaker's previous utterance to the effect that the intervening talk containing a problem is canceled out by way of being shown to have been an interruption of the speaker's turn in-progress. This practice points to an aspect of the way in which the 'agglutinative' grammatical process involving the use of phrasal units shapes interactional patterns as observed in the course of organizing turns/sequences and managing problems in talk.

Do young children use echoic information in their comprehension of sarcastic speech? A test of echoic mention theory

This study sets out to provide a test of echoic mention theory, which predicts that irony and sarcasm are most easily comprehended by a listener when the speaker explicitly ‘echoes’ a previous utterance or some shared norm rather than when the speaker only implicitly alludes to the same information. Children aged 6‐10 years were given stories containing either a sarcastic comment that explicitly echoed an earlier remark, a sarcastic comment that only implicitly alluded to an earlier remark, or a literal comment from a speaker. Half of the children heard the stories presented with a sarcastic intonation and half heard the stories presented with an uninflected intonation. The results of the study showed that when vocal intonation was absent, children comprehended explicit stories better than implicit stories. The findings suggest that echoic mention theory is a useful theory for describing children's developing comprehension of sarcastic speech, but is only one of many factors that play a role in the comprehension process.

Imitation in the spontaneous language of children with and without Down syndrome

This study examined the potential roles of immediate imitation in the spontaneous language of eight children with Down syndrome and eight peers matched for mean length of utterance (MLU). The typically developing group, but not the group with Down syndrome, used significantly longer imitated than spontaneous utterances. They also used expanded and modified imitations significantly more often than the children with Down syndrome, who used a relatively larger proportion of reduced and exact imitations. Thus, imitation seemed to play a different syntactic role for the groups. Several children with Down syndrome and one typically developing child imitated words selectively (i.e. they either imitated words that they did not use spontaneously or vice-versa), suggesting that imitation played a role in lexical development for these children. The most robust findings related to the pragmatic function of imitation. Imitations of both groups frequently included the final word of the previous maternal utterance. One interpretation of this finding is that imitation acted as a conversational place-holder when the children could not otherwise participate appropriately. Mothers also shortened their utterances immediately following an imitation. This finding suggests that mothers may have interpreted imitation as a result of children's incomplete comprehension.

Structural Parallelism and Discourse Coherence: A Test of Centering Theory

Three experiments investigate the effects of utterance structure and anaphoric reference on discourse comprehension. These factors are examined in the context of utterance pairs with parallel constituent structure (e.g.,Josh criticized Paul. Then Marie insulted him). Previous studies of structural parallelism have shown that an ambiguous pronoun (e.g.,him) is biased to corefer with an antecedent in the same structural position (e.g.,Paul). Of interest was whether parallelism can also influence the capacity for a pronoun to facilitate discourse comprehension and whether the centering model of discourse coherence can account for such effects. Most generally, centering predicts that a pronoun will increase coherence when it corefers with the subject of the previous utterance and that a single pronoun is sufficient to optimize local coherence. Experiment 1 addresses the interpretation of ambiguous pronouns and shows that preferred antecedents for both subject and nonsubject pronouns were entities realized in the same grammatical role. Experiment 2 examines the comprehension of utterances containing a pronoun whose antecedent occupies either a parallel or nonparallel position. The results show that utterances are coherent only when the pronoun corefers with a parallel antecedent. Experiment 3 evaluates how the presence of multiple anaphoric links facilitates comprehension and shows that more than one pronoun may be necessary to make an utterance locally coherent. Overall, the results reveal several limitations in centering theory and suggest that a more detailed account of utterance structure is necessary to capture how coreference influences the coherence of discourse.

Dutch Anaphoric Dat in Definitions

Abstract. Dutch anaphoric dat has an important function in definitory exchanges. If dat is used in the recapitulation of a non-nominal definition, it increases the coherence of the defining utterance, by tying the recapitulation to the previous utterances and by permitting the switch of the focus from definiens to defmiendum. However, dat is not really coherence-building when used for left-dislocation in core-definitions. Indeed, a left-dislocated definition cannot be more coherent than a definition without interruptions and 'gaps'. The role of dat in left-dislocation is rather to permit extra focalisation and thematisation of the definiendum in non-nominal definitions, and of the definiens in nominal definitions.

Language Learners' Interaction: How Does It Address the Input, Output, and Feedback Needs of L2 Learners?

This study asked whether L2 learners' interaction with other learners can address three of their supposed needs for L2 learning-that is, their needs for L2 input modified toward comprehensibility, for feedback focused on form, and for modification of output-in ways that interaction with native speakers (NSs) has been shown to do. To Results of the comparison revealed similarities in the types of modified input and feedback the learners and NSs offered other learners in their respective dyads and in both the type and amount of output modifications they produced. The learners received less modified input from other learners than from NSs. As expected, learners' morphosyntax conformed less to standard English than did that of NSs. This pattern was found throughout the learners' input and feedback but was far less evident in feedback formed through simple segmentation of previous utterances. The study thus indicated that interaction between L2 learners can address some of their input, feedback, and output needs but that it does not provide as much modified input and feedback as interaction with NSs does.

Speech dialogue system based on simultaneous understanding

The authors propose a novel way to implement a speech dialogue system. The method, called Simultaneous Understanding, accomplishes speech recognition, understanding and action simultaneously with the user's utterance. This makes the dialogue system more interactive, because of the following advantages: the user does not have to wait for the system response, unless he wants to see the action results on the screen. If he sees a mis-recognized entry, he can correct it before the database is accessed. The authors implemented two ticket reservation speech dialog systems, using an existing speech recognition system. One of them was based on the above method, which can accept new inputs while simultaneously analyzing the previous utterance, and the other accepts new inputs after analyzing the previous utterance. It was found that the above method improved the total average utterance accuracy by 5.4% and the total time spent to solve tasks by 4.7%. This shows that the new method is promising for increasing the interactiveness for the speech dialogue system.

Pronouns, Names, and the Centering of Attention in Discourse

Centering theory, developed within computational linguistics, provides an account of ways in which patterns of interutterance reference con promote the local coherence of discourse. It states that each utterance in a coherent discourse segment contains a single semantic entity—the backward‐looking center—that provides a link to the previous utterance, and an ordered set of entities—the forward‐looking centers—that offer potential links to the next utterance. We report five reading‐time experiments that test predictions of this theory with respect to the conditions under which it is preferable to realize (refer to) on entity using a pronoun rather thon o repeated definite description or name. The experiments show that there is a single backward‐looking center that is preferentially realized as a pronoun, and that the backward‐looking center is typically realized as the grammatical subject of the utterance. They also provide evidence that there is a set of forward‐looking centers that is ranked in terms of prominence, and that a key factor in determining prominence—surface‐initial position—does not affect determination of the backward‐looking center. This provides evidence for the dissociation of the coherence processes of looking backward and looking forward.

The Internal Structure of Extended Turn-at-talk within an Academic Context: Native Graduate Spanish, Non-native Graduate Spanish, and Non-native Beginning Spanish Speaker Data

within an Academic Conte : Native Graduate Spanish, Non-native Graduate Spanish, and Nonnative Beginning Spanish Speaker Data * Christina Makara, Southeastern Massachusetts University Turn-taking is an importa t discursive skill that speaker accepts, evaluates, agrees, disagrees, or fo eign la guage learners must ma ter f they are fights back, where s in pro pective metat be able to inter ct succes fully outside comment, speaker uses co trast, ampl fication, classroom. A turn can consi t of a ngle u t rexemplification, exp anat on, evaluation, or ex ance in response to a previous utterance. If one is ploration of causes to relate what is to come with to engage in deeper, more involved conversawhat has gone before (20). tion, however, then one must be able to take the Communication tasks inherent within the floor and extend an utterance through context of a prospective metacomment have amplification, contrast, causes, consequences, been discussed by Gumperz (9) as a way of alternatives, and evaluation. According to examining oral discourse. Several second lanKramsch (24) there are four components of the guage rhetorical studies have been conducted internal organization of a turn-at-talk within a that use communication tasks such as contrast, natural discourse setting: amplification, explanation, evaluation, and ex1. Metacomment and paraphrase ploration of causes to discriminate between the 2. Expanding through amplification; generaliwritten performance of native speakers and that zation; restrictionlcontrast examples; of non-native speaking students (Jacobs; Connor; examination of causes, consequences, Scarcella; Connor and McCagg). As yet, no alternatives research has explored to what extent non-native 3. Internal linking and structuring by announcleamers of Spanish use these types of tasks ing future points or by returning to during extended talking. Data from such reprevious points search could suggest areas for curriculum devel4. External linking to a previous point made by opinent, oral assessment, and strategy panner. improvement during extended turn-taking. Kramsch distinguishes between retrospective d prospective metacomments used within Purpose of t e Study a turn-at-talk. In a retrospective metacomment a The study was designed to examine extended *Members are invited to send material to: Dr. Douglas K. turn-at-talk across three different levels of colBenson, Dept. of Modern Languages, EisenhowerHall 104, lege Spanish speakers in order to yield inforKansas State Univ., Manhattan, KS 66506. Maximum mation about performance ability in a foreign length is 15 double-spaced pages, in either 1985 MLA or language on a variety of communicative tasks: linguistics format. Please send a typed original and a copy for each paper submitted, and include a self-addressed elaboration, background, circumstance, restateenvelope and loose stamps to cover return mailing of the ment (descriptive tasks), and purpose, manuscript. evaluation, cause, and metacommunication (in-

Building Cohesion Through Collaborative Narration

Although self-disclosure has been shown to be an important verbal means for conveying and strengthening solidarity in very close relationships, little attention has been paid to other ways in which talk is used to convey and develop solidarity among friends. This paper shows how talk is used to build cohesion by examining collaborative talk, in which an utterance supports or ratifies the previous utterance in some manner. Although collaborative talk occurs in a number offorms, collaborative narratives are of particular interest because the very strategies that are required to produce a coherent narrative also lead to a greater perception and feeling of solidarity among group members. Adolescent females use a variety of strategies when telling personal narratives jointly. Some of the strategies reflect sociolinguistic rules of narratives, such as dividing the different functions of description and evaluation among the different speakers. Other strategies reflect grammatical rules of narratives, such as maintaining coherence through repetition and conjunction. Girls also use sequencing rules to establish coherence when the speaker changes. In all cases the strategies help to enhance solidarity among the narrators.

Improving speaker consistency in an automatic speech recognition framework

Inconsistencies in speaking pose a major obstacle to the effective use of automatic speech recognition (ASR) systems. In the present study, we sought to improve speaker consistency in ASR by means of instructions and feedback. The instructions were specific suggestions for effective interaction with ASR devices. Feedback was given by a visual barometer-like display, representing the similarity between an utterance and previous utterances. In two experiments, naive speakers were tested on isolated words using speaker-dependent templates. In the first experiment, subjects who were given instructions demonstrated greater speaker consistency than those who did not receive instructions. In contrast, feedback had very little effect on performance. Instructions then were revised, based on an extensive analysis of the types of errors made by speakers, and were tested in a second experiment. Results of this experiment showed a further improvement in recognition performance; the new-instructions group had recognition accuracies that were 6% higher than the no-instructions group.