The Terminology of Safety

The notions of para-disciplinary languages and functional language (cf. Hoffman 1984 and Halliday/Martin 1993) are explored with reference to those situations where communication is required to ensure safety and the prevention of hazards. We seek to demonstrate that, in certain scenarios, terminological practice gains a safety-critical edge. Our approach begins with a description of the place of terminology in a postulated language of safety evidence for which we found in our corpus of safety-related texts. After presenting a case for this language of safety we overview some terminological issues germane to the use of this language in its safety-critical environment. Introduction Good documentation procedures are always important for the efficient and economic management of an organisation and its resources. Moreover there are some scenarios in which documentation assumes an additional, safety-critical, aspect, e.g. when safetyrelated messages such as warnings and guidelines are being communicated. There are a number of databases, simulation programs and knowledge based systems that are developed to avoid hazards and ensure safety. The data, the words and the knowledge in these information systems is acquired from texts and conversations and disseminated as textual output from computer systems. It appears that those involved in documenting knowledge of safety either in textual form or in computer systems, have to deal with safety-related terms that are couched in the lexicogrammar of the (natural) language of the documents. The role of terminology in a postulated ‘language of safety’ is the topic of this paper. Terminology literature often focuses on the terminology of languages used by an identifiable community like scientists, engineers, artists and sportspersons, politicians and ideologues and so on. However, there are a number of instances where authors have discussed the use of language not just within one single discipline but across a number of disciplines. These ‘para-disciplinary’ languages include the languages of ‘negotiations’, ‘commerce’, ‘morals’, ‘ethics’ and ‘pedagogics’ . The case for two such languages, the languages of ‘journalism’ and ‘administration’, has been argued for by Sager/Dungworth/McDonald (1981) by contrasting these hypothesised languages with those of ‘medicine’ and ‘ecology’. Quasi-artificial languages, like Seaspeak (Weeks et al. 1988), with their emphasis on standardised terminology and restricted syntax, also have a bearing on our discussion. To the preceding list of so-called para-disciplinary languages we would like to add, tentatively, the ‘language of safety’ that is the language used in the prevention of hazards when ensuring the safety of people in a given domain or certain sets of circumstances. Like any other language, the language of safety should have its own ‘lexicogrammar’, that is a distinct terminology with idiosyncratic morphological and syntactic behaviour (cf. Halliday/Martin 1993). In order to seek evidence for the existence of such a language we have created a corpus of safety-related texts and have examined the social institutions and interactions in which they are embedded. Historically, the Prague School’s exposition of the language of commerce (c. 1920, cited in Hoffman (1984)) can be regarded as a precursor to the modern-day study of paradisciplinary languages. Hoffman notes that the Prague School focused on the ‘peculiarities’ of lexica and grammar of the language of commerce for arguing that this indeed was an identifiable ‘language’. This has encouraged us in our endeavours. The developments in corpus linguistics, particularly in corpus-based terminology, has helped us in examining the lexico-grammatical features of texts in our corpus. We have observed unusual preponderances and suppressions of closed-class words (e.g. determiners, pronouns and conjunctions) in comparison to their distribution in general language. We have also noted an unusual distribution of the uses of modal auxiliaries such as may and should in contrast with the results reported for scientific language by Huddleston (1971) and Meyer (1990). From a philosophical, sociological and linguistic perspective the language of safety can be regarded as an extension of, yet intimately bound to, the languages of science and technology. In this way the operation of the languages reflects the conceptual and societal processes whereby scientists explicate innovative ideas which, in turn, are taken by technologists and applied for the development of new goods and services. Throughout these processes safety is an issue whether it be in the scientific laboratory, the production plant or amidst the general public using domestic appliances. We will elaborate our ideas as to how the terminological aspects of the language of safety relate to those of other specialist languages and to society and societal processes at large. Supported by evidence gleaned from our safety corpus we postulate the existence of a ‘generic safety terminology’ and consider its usage and interpretation. We consider this safety terminology to constitute a set of terms that are applied in a multitude of domains in those instances when information pertaining to the maintenance of safety is being communicated. What is of particular interest about these terms, e.g. hazard, danger and risk, is the way in which their ‘meanings’ have some constant aspects regardless of the domain in which they are being used. This, despite the fact that the nature of hazards, danger and risks varies radically between domains. Also of interest to us is the way in which the terminology of risk analysis is open to both objective (numerically based) and subjective (qualitatively based) interpretations. Good terminological practice is of importance for successful communication, ensuring that addressers and addressees share a common understanding of technical vocabulary. In certain domains however this issue assumes a safety-critical aspect. To whit an anecdote relating to the common management and maintenance of some railway lines by London Underground (LU) and British Rail (BR). In LU parlance the phrase line clear meant that the railway tracks were clear of trains and therefore safe for workmen to repair; however BR employees understood the contrary, i.e. that the lines were clear of obstructions and workers so that trains could safely pass through. Our research has been carried out under a project, Safe-DIS (Safe Design by use of Information Systems), funded by the UK government’s, ‘Safety-Critical Systems Programme’ (Ahmad 1995a). The principal aim of Safe-DIS is to document, archive and relay safety-related information with computer systems in order to ensure that less experienced practitioners in a field are guided safely through their work. The systematic organisation of safety terminology together with close attention to a language of safety, it appears, plays a major role in ensuring safety. ‘Language of Safety’ : An existence question. Safety is an issue in a range of scenarios, including: (i) those involving the general public such as, a walk near a cliff edge, the operation of an electrical appliance or the imbibing of medicine; (ii) the regulation of processes in particular domains such as the transportation of dangerous chemicals; and (iii) the design, implementation and maintenance of safety-critical systems, e.g. air traffic control systems and nuclear power plants. We take the view that successful communication is paramount for the maintenance of safety in all these, and other, cases. Examples of this communication include: warning shouts in situations of immediate danger; signs highlighting potential danger; instructions for the safe use of domestic appliances; the communications between humans and computers in safety-critical systems; institutional guidelines for safe practice in a particular domain; governmental health and safety policies; and academic papers describing methods for risk analysis and for hazard identification and avoidance. We hypothesise that the content and structure of the texts produced for these communications will be influenced, in part, by a shared combination of factors that will result in a common aspect to the linguistic nature of these texts. To investigate these ideas we have collated a corpus of safety-related texts representative of the multifarious scenarios in which communication has a safety-critical edge. The research reported in this paper is based on a collection of texts (see Table 1) that are concerned with safety in a range of disciplines such as: public health engineering, health and safety at work, analysis of risks in high integrity systems. Table 1 The safety corpus Text Text type Topic Word count 1 Academic text Risk analysis 6220 2 Academic text Risk analysis 8899 3 Learned journal System failure 3057 4 Government Safety Working in sewers 11778 5 Company Safety Manual Working in sewers 10820 6 Company Safety Manual Working in sewers 3126 7 Company Safety Policy Working in sewers 857 8 Company Safety Manual Working in sewers 2263 9 Safety Handbook Construction 1842 10 Textbook Hazardous materials 14993 11 Safety labels Assortment 1440 The texts include learned journal articles, technical manuals for use in design, and chapters from textbooks on drainage maintenance and the handling of dangerous substances. We have two other kinds of texts that are usually not included in various text classification schema (cf. Sager/Dungworth/McDonald, 1981): (i) public information, these texts are written by specialists for a mass audience, a genre which has close similarities with popular science and technology texts, and; (ii) safety labels these labels are gathered from a range of household goods, office equipment and other machinery designed for operation by laypersons. For the purposes of the current paper we refer to these texts collectively as the ‘safety corpus’. The texts selected for the safety corpus are presumed to share at least one common aspect which can be described as their ‘functional tenor’. We borrow this phrase from Beedham/Bloor (1989) who analysed a corpus of texts whose functional tenor, the authors claimed, was pedagogics. In the case of the texts comprising the safety corpus we would say that they share a functional tenor which is the maintenance of safety. Functional tenor can be regarded as an attribute of a para-disciplinary language: as such it is an aspect which remains constant across domains and thus contrasts with the notion of sub-languages. In seeking to identify a para-disciplinary language we find ourselves engaged in a pursuit similar to that of Swales/Bhatia’s corpus-based work on legal language (1983). More recently Kresta (1994) sought to characterise the reply as a text type in the scientific discourse of linguistics through the identification of links between the extralinguistic features relating to the communicative circumstances of the texts under investigation and the linguistic features observed in the texts. These approaches, as does ours, relate to the field of functional stylistics which, for instance, ‘demonstrates how desirable qualities in scientific writing, exactitude, simplicity, abstractedness etc. are realised by linguistic means, declarative sentences, neologisms, passive forms etc.’ (Hoffman 1984). Our analysis started with the compilation of various statistics for the safety corpus. These statistics were computed using a software package and methods developed at the University of Surrey, System Quirk; detailed results of the analyses will be presented in (Ahmad/Salway, in preparation). These statistics include both absolute and relative frequency counts for the lexical items in the corpus. Additionally a ‘weirdness coefficient’ (Ahmad 1995b) was calculated for each lexical item by dividing its relative frequency in the safety corpus by its relative frequency in the twenty million word Longman Corpus. This latter figure is taken to be representative of a word’s relative occurrence in general language. Thus, the ‘weirdness coefficient’ is a measure of how more or less frequently a lexical item appears in an analysed corpus than in ‘general language’. That is to say that a ‘weirdness coefficient’ >>1 suggests that a lexical item is occurring much more frequently in the analysed corpus than in general language. In order to illustrate the idiosyncratic distribution of closedand open-class words in safety-related texts, we contrast such a text with general language texts (c.f. the Longman Corpus) by comparing the frequency of words in the safety text with that of the same words in the Longman Corpus. For illustrative purposes, we have selected a text comprising 1849 words in total distributed amongst 573 distinct wordforms, which deals with the potential hazards faced by drainage engineers involved in the manual inspection of waste water systems (see Table 2). Table 2 Distributions of closedand openclass words in a safety text Group a: Normal closed class Group b: Suppressed closed class Token Weirdness Coeff. Token Weirdness Coeff.