Group Interaction under Threat and High Workload: Linguistic Factors
Research Project by Manfred Krifka,
Department of Linguistics, University of Texas at Austin,
sponsored by Gottfried Daimler und Karl Benz Stiftung, Ladenburg
April 1, 1999
The purpose of the project is to identify specifically linguistic factors, that is, structural properties of language, that may lead to communication failure in situations of threat or high workload. The proposed project has two phases: Phase one will be a linguistic investigation of cockpit voice recorder transcripts from accidents and mission simulations, under collaboration with and using data provided by the Texas Aerospace Research Project (Professor Helmreich). Phase two will investigate communication failures that may be attributed to low proficiency in English of one or more participants, and possible interferences of the native language of the participants. Research will presumably focus on East-Asian languages (Chinese, Korean, Japanese). The goal of the project is to find out whether there are such genuinely linguistic factors that contribute to communication failures, in addition and in combination with better-known factors that are a result of situations of high workload or danger. If these findings are positive, this will result in proposals for specific linguistic training exercises.
A first investigation of accident reports in aviation and in particular a reading of Cushing (1994) suggests that there are a number of areas in which structural properties of language might lead to misunderstandings, especially in situations in which the cognitive resources of participants are under stress.
Vagueness of terms. Even though technical language is subject to relatively precise forms, vagueness, which contributes to the flexibility of natural language, will occur in all types of technical talk (e.g., what is a steep ascend?). Research will concentrate on cases in which vagueness leads to communication failure. If the data indicate identifyable sources of vagueness that are likely to result in misunderstanding, special training methods could be developed to try to calibrate the use of vague terms across speakers. The study of explicit markers of vagueness ("hedges") such as or so and roughly, or the use of rounded numbers, will be of particular interest.
Ambiguity. There are several sources of ambiguity, even in technically regimented language. Ambiguity can arise because words have a wide range of meanings ("polysemy", e.g. the term takeoff that was crucial in the Teneriffe accident), because two different words happen to be pronounced the same (e.g. climb to five zero vs. climb two five zero), and because there is more than one way to parse an expression (e.g. (back on) (the power) vs. (back) (on the power)). In practically all sitations, only one interpretation will be relevant, so that speakers are not even aware of the ambiguity. But there are incidents of ambiguity that lead to dangerous situations in aviation. The current project should gather a database of misunderstandings due to ambiguities. It can then try to develop proposals that would lead to a reduction of frequent types of ambiguities. Often, ambiguities may arise not because two expressions sound exactly the same, but already because they sound similar enough to be mixed up. This problem might be exacerbated for non-native speakers of English.
Pronominal expressions. Pronouns and other anaphoric devices, like ellipsis, can lead to ambiguities and to slower and less reliable understanding. The project will identify cases of misunderstandings due to the misunderstanding of pronominal expressions, and it will suggest strategies to avoid that. It is possible that making speakers aware of the potential problems with the use of pronouns is already quite helpful, and they may also be instructed in modes of speaking in which pronouns are avoided, if possible. The use of pronouns for non-native speakers of English requires special attention, as languages can vary widely in how frequently they use such strategies (e.g., heavy use of ellipsis in East Asian languages, probably more frequent use of pronouns in languages with gender systems).
Presuppositions. The information of a sentence can be divided in two parts: the part that is asserted or proferred, and the part that is presupposed and necessary to understand the proferred part in the first place. But even presupposed information may be new; for example, a sentence like we regret that landing is not possible now may be used to inform that landing is not possible, even though it asserts just that the speaker regrets this. The project can investigate how much information is typically transmitted by presuppositions, whether this rate increases or decreases in high-stress situations, and whether the use of presuppositional information transfer leads to potential problems.
Information structuring. Linguistic information is structured, roughly in parts that relate to old or known concepts, and parts that express new concepts. For example, the sentences The left wing is on FIRE and The LEFT wing is on fire express the same information (they are true in exactly the same circumstances), but they do so with different emphasis. This relates to the topic of presuppositions mentioned earlier. The structuring of information is of great importance; for example, wrong structuring is a major problem of texts of inexperienced writers. The project will have a close look at the information structuring typically employed in cockpit communication; for this, it will have to work with voice recordings, and not with transcriptions. Again, it should be of great interest to look at non-native speakers of English, as languages differ widely in the techniques they use for information structuring (for example, particles in Japanese and Korean, free word order in Slavic languages, syntactic reorganization in Romance languages).
Deictic expressions. These are expressions that get their reference through features of the speech situation. Spatial terms are of particular interest here, as they are often related to the speaker, but can be ambiguously related to other centers as well. For example, in front may be used by a speaker to point to his or her own frontal region, to the frontal region of the hearer, or even to the region in front of a directed object, like a car. The project should investigate the use of deictic terms in cockpit communication and identify potential sources of problems.
Speech acts. Utterances can be performed in a wide variety of functions: to describe a state of affairs (assertions), to command, to permit, to ask a question, to request permission, to acknowledge understanding, to declare a new state, like an emergency, to express feelings in exclamatives, and so on. There are potential sources of misunderstanding here. For example, the elliptical utterances that are typical for pilot/tower communication can often be understood as assertions or requests. In other cases, the same sentence can be understood as an assertion or a question, perhaps just depending on subtle intonational features. The project should investigate the types of speech acts that occur, and should record misunderstandings based on miscategorizations of the speech act type of a sentence. If such misunderstandings turn out to be serious enough, it can propose ways to mark speech acts unambiguously, e.g. by a forced please or request for commands and asking for permissions. Again, the investigation of the communicative behavior of non-native speakers of English should be of particular interest.
The project will have close ties to the project of Professor Helmreich, both for practical and theoretical reasons. It will make use of the data gathered by the Texas Aerospace Research Project and of the technical expertise that is available at this institution. It will also be complementary and ancilliary to the research goal proposed there, to analyze the micro-coding of group interaction based on the notion of coherent units of communication.
The project also will have ties to the project of Professor Dietrich, which is specifically concerned with issues of linguistic performance.
I do not work on any comparable project, but my professional interest and expertise covers most of the possible linguistic factors mentioned above. In particular, I have worked on the theory of information structure and accentuation in natural languages.
All the linguistic factors mentioned under (1) have been extensively investigated. We have a relatively good handle on the notion of vagueness and ambiguity (in particular, word ambiguity and structural ambiguity), on pronominal expressions, on information structuring, on different types of deictic expressions, and on speech acts. These theories were partly developed by investigating recorded texts and transcripts, but more typically by what may be called "experiments": From a given theory, certain predictions are derived, and they are compared with respect to the intutions of speakers, or perhaps also with respect to large-scale linguistic corpora. Knowledge of these linguistic phenomena then was applied to the analysis of existing texts and of linguistic performance, for example, in language acquisition, in the acquisition of foreign languages, and in cases of language impairments. But I am not aware of systematic studies of these linguistic factors in the situation type identified in the current project, namely, situations characterized by threat or high workload.
As already indicated, the project will have two phases.
In the first phase, which is projected for one year, a doctoral student will identify in transcripts and recordings, and in close collaboration with the project of Professor Helmreich, the linguistic factors that are likely to impede communication in situations of threat or high workload. The findings of this phase will be collected in the first deliverable that will describe incidences in which linguistic factors lead to communication problems in situations of threat and high workload, and also cases in which linguistic factors lead to a dangerous situation in the first place. These linguistic factors will be described in detail, in a way that makes sense to non-linguists. The importance of the various factors will be estimated, and, if possible, suggestions for the training of communication skills will be made.
If this project phase is successful, and if ways can be found to obtain the necessarry transcripts and recordings, there will be a second phase of one year in which a doctoral student investigates the specific factors when non-native speakers of English are involved. It will be a task of the first phase to investigate whether transcripts and recordings can be obtained. Assuming that these problems are more pronounced with languages that are typologically different from English, and capitalizing on the fact that there are a number of East Asian doctoral students in the Department of Linguistics of the University of Texas at Austin, we will concentrate on East Asian languages, presumably Mandarin, Cantonese, Korean, and possibly Japanese. The results of this phase will be described in detail in a second deliverable.
Manfred Krifka
Associate Professor
Department. of Linguistics, B5100
University of Texas at Austin
Austin, TX 78712-1196, USA
Phone: (o) +1-512-471 1701, (fax) 471 4340, (h) 326 9889
e-mail: