Chapter 2: Disciplinary context
2.2 Conference interpreting review
2.2.2 Research profile
2.2.2.1 Cognitive issues
2.2.2.1.3 Comprehension
According to Hatim and Mason (1990: 226f; 1997: 17f) and Pöchhacker (2004: 118f), in
research into language comprehension there is a basic distinction between bottom-up (input-driven) and top-down (knowledge-based) operations, both being essential for an encompassing account of comprehension since comprehension is seen as a dynamic process that largely depends on what one already knows. Thus, processing new information requires the build-up of a mental representation resulting from the interaction between the input and already existing information whether lexical, semantic, syntactic, pragmatic, encyclopaedic, etc. or linguistic and extra-linguistic knowledge (Gile 1993: 69).
Several studies have been carried out along the lines of the aforementioned fact, but they differ in approach or concepts used to describe the process. Proponents of the theory of sense (cf. e.g. Seleskovitch 1978b: 334; Lederer 1990: 53), reject research into grammar and contrastive linguistics as the sole basis for approaching language comprehension and suggest ‘cognitive complements’ (verbal, situational and cognitive contexts and world or encyclopaedic knowledge) that interact with ‘units of meaning’ (Lederer 1978: 330) to ‘make sense’ (Seleskovitch 1986a: 371), thus allowing for comprehension of the incoming message (Seleskovitch and Lederer 1995: 23; Chapter 4).
Chernov (cf. 1979; 1992; 1994) has based his SI processing model of ‘probability prediction mechanism’ on redundancy (1979: 107). In an experiment involving 11 professional interpreters, UN General Assembly speeches of 20-minute duration have been designed in such a way that the first type of utterances contains unpredictable endings and the second nonsensical sentences. More than 76% of subjects have either omitted utterances with unpredictable endings (38.75%) or translated them according to the expectations prompted by context (37.5%), and more than 89% have mistranslated (53.58%) or omitted (35.71%) the semantically unusual sentences. Chernov has concluded that message redundancy and predictability are vital for the comprehension process. To him, the semantic comprehension process is a ‘cumulative dynamic analysis’ (1979: 104f, emphasis in original) that generally covers the gradual addition of rhematic elements to the thematic ones, bridging sense gaps, linking and combining rhematic and thematic components into fewer but more complex semantic configurations, and integrating this semantic structure into the larger situational context and the knowledge of the hearer (ibid).
Van Dijk and Kintsch (1983) model of discourse analysis because it relates to building a propositional textbase, inferencing, building macrostructures and a situational model.
Mackintosh (1985) has applied the Kintsch and Van Dijk model (1978) to CCI and SI, conducting two experiments to test its application. The experiments have indicated the relevance of the model’s macrorules of deletion, generalization and construction to interpreting processes as ‘mapping rules which link the stages in the process and explain why certain things [errors and omissions] occur’ (1985: 42). However, Mackintosh argues that the level of macrostructural processing required for ‘surface text comprehension’ without recall remains to be demonstrated because it is not possible to measure how much processing is made in response to the comprehension or recall tasks (ibid).
Drawing on findings in discourse analysis, text linguistics, sociolinguistics, pragmatics, semiotics, conversation analysis and artificial intelligence, Hatim and Mason (1990) propose a textual model of discourse processing in which context is defined in terms of three dimensions: a communicative transaction (subsuming user: <idiolect, dialect>, use: <field, mode and tenor> as register variables), pragmatic actions (intentionality, speech acts, implicature, inference, etc.) and a semiotic interaction (word, text, discourse, genre, etc. as signs). These contextual dimensions determine the structure (compositional plan) and texture (cohesion, coherence, etc. devices) of the text seen in the model as ‘the ultimate unit of effective communication’ (Hatim 1997: 12). Hatim and Mason (1997) argue that, and show how, their model applies to all fields of translation activity, from literary and religious translation, through subtitling to different interpreting types, including SI, and thus interpreter training (see also Hatim 1984; 1989; Mason 1989).
Dillinger (1994) has conducted an experiment on eight experienced and eight novice interpreters to determine if there are any potential differences in comprehension processes between the two groups, but found no significant evidence of interpreting-specific comprehension skills. Thus, he suggests that comprehension is the application of an innate skill, which is a natural consequence of bilingualism, in more unusual situations because there has been only insignificant evidence (17%) of superiority of trained interpreters (1994: 166/185). Pöchhacker (2004: 120) attributes this failure to reasons of experimental design. However, although Dillinger (1994: 156) acknowledges that his study concerns
comprehension only and that the importance of production should not be ignored, strangely enough, he uses these results to conclude that interpreting is an innate skill and thus suggest that the selection of students is in fact more crucial than training (1994: 185).
Dillinger’s results have been questioned by the results of a number of experimental studies (cf. Padilla et al. 1995; Bajo et al. 2000) that have suggested that training and experience in interpreting have been found to be essential for developing such skills as fast and accurate reading abilities, faster access to lexical and semantic storage, larger working memory capacity and more efficient use of such capacity. Besides, bilinguals have not shown any superiority of comprehension skills, which also suggests that comprehension is not due to linguistic abilities gained by knowledge of another language (Bajo et al. 2000: 140f).
Isham (1994) has found ‘evidence for and against’ deverbalisation because in an experiment nearly half of the subjects (nine professional interpreters) have shown a superior verbatim recall of the most recent sentence, translating on a sentence-by-sentence basis and focusing attention on the surface form of the SL sentences. However, the recall of the other half of interpreters has been inferior, indicating no distinction between the clause or sentence boundary conditions with interpreters’ attention being diverted away from the surface form, thus using the meaning-based approach (ibid: 205f). Isham has (ibid: 208) concluded that interpreting ‘from English into French’ is not a single process, but involves at least two, form- and meaning-based approaches. However, research is needed to determine whether the interpreters’ choice of either approach is a strategic one or not (ibid).
Incidentally, Setton’s (cf. 1998; 1999) bottom-up cognitive-pragmatic analysis, which is described by Pöchhacker and Shlesinger (2002: 178) as ‘one of the most significant advances in interpreting research at the close of the twentieth century’ and draws on RT, cognitive semantics, mental models and speech-act theory, supports the knowledge-based approach to comprehension and confirms the claims of the theory of sense concerning a big role for extralinguistic factors in the comprehension process and irrelevance of language- specific factors or language-pair processing (2.2.2.1.6.2.4 The anticipation strategy).
2.2.2.1.4 Memory
temporary storage system that is distinguished from a permanent storage one goes back to the mid-twentieth century. However, Darò (1997) suggests that the hypothesis of dualistic models of memory actually goes back to the late nineteenth century, citing James (1890) as the first to distinguish between ‘primary and secondary memory’, the former for storage of past knowledge and experience while the latter for information attracting current attention (1997: 623, emphasis in original). These correspond to the modern terms of long-term memory (LTM) and STM, respectively, the latter being later referred to as working memory (Darò ibid; Pöchhacker 2004: 121).
The state-of-the-art model of working memory has been proposed by Baddeley (cf. e.g. Baddeley 1990; 2000; Baddeley and Logie 1999). This model comprises multiple specialised components of perception that allow humans to relate new information to the knowledge already stored in their LTM. The specialised components include a central executive that serves as a supervisory, regulatory or control system that performs several functions such as connecting the two slave systems, focusing and switching attention, etc. The two slave systems are the phonological loop that is responsible for storing and processing speech-based information and the visuospatial sketchpad that is responsible for processing images (Baddeley and Logie 1999: 28).
LTM is likewise viewed as a multifaceted system comprising various subsystems, each being responsible for one type of information storage and can be functionally independent from the other subsystems (e.g. Cohen and Squire 1980; Squire and Zola-Morgan 1991; Darò 1997: 623). A distinction is made between declarative (or explicit) memory and
procedural (or implicit) memory where the former refers to events and facts directly accessible to conscious recollection (knowing that) while the latter is implicitly contained within learned skills and cognitive operations (knowing how). Explicit memory in turn consists of semantic memory for storing encyclopaedic knowledge and episodic memory for recall of one’s personal experience or past events. Implicit memory comprises motor and cognitive skills and habits, conditioning and non-associative learning (Darò ibid: 623f).
2.2.2.1.4.1 Working memory and interpreting
In an experimental study, Gerver (1974a) has found that simultaneous listening and speaking can hinder retention of the material listened to when speaking, but that recall
becomes better when a complex form of information processing is involved in the simultaneous and listening tasks because results have indicated that subjects’ recall has been better after listening, followed by SI and shadowing, respectively.
This finding was later supported by Lambert (e.g. 1989) who has obtained evidence of poor recall when listening and speaking simultaneously because the results have indicated higher recognition scores after listening, followed by CCI, SI and shadowing, respectively. Lambert explains that during listening no vocalisation is required and thus attention is devoted only to one task (ibid: 89). Recall after CCI has been higher because CCI involves complete overt rehearsal of the text, longer exposure to information, visual cues provided by the notes and oral feedback provided by the delivery. Thus, CCI represents a deeper level of processing than SI. SI is deeper than shadowing because the translation factor, which is absent in shadowing, requires more processing of the stimulus and thus longer duration can be regarded as a distinctive factor. Thus, Lambert hypothesises that listening and CCI involve deeper processing than SI and shadowing because whereas vocalisation is not required in listening and only required in CCI after the original has been delivered, simultaneous vocalisation is required in SI and shadowing. In terms of the effort and attention, during CCI and listening, attention and effort are devoted completely to the processing task whereas they are shared with other tasks in SI and shadowing (ibid: 90).
Isham (1994) has found that simultaneous interpreters’ verbatim recall of the final clause has been poorer from English to French than from English to American Sign Language. He (ibid: 204) suggests that the difference between the two groups lies in the fact that sign language interpreters have been exposed to onespoken language at a time, but simultaneous interpreters working between English and French have been in fact exposed to two spoken
languages. Thus, he has attributed the latter’s inferior recall to ‘phonological interference’ that is ‘caused by the two speech streams encountered by this group alone’ (ibid). Isham (2000) has concluded that phonological interference causing poor verbatim recall is not a matter of storage but of process. In other words, it is not due to two speech streams but to interpreters’ inability to ‘refresh the trace for phonological form through subvocal rehearsal’ because interpreters are engaged with speaking while listening (ibid: 147).
after SI has been confirmed, at least partially, by Darò and Fabbro (e.g. 1994; Darò 1997) who have suggested that other factors such as divided attention and translation processes also significantly reduce recall in that they contribute to a reduction in the capacity of the working memory, which in turn reduces long-term memorisation of translated material in SI. This they attribute to a mechanism of ‘articulatory suppression’ of ‘subvocal rehearsal’ in the phonological loop of working memory (Darò 1997: 625).
The existence of the mechanism of articulatory suppression has been confirmed in a series of experimental studies by Padilla et al. (1995) and Bajo et al. (2000) who have found that interpreters’ recall and comprehension have been superior to those of student interpreters and non-interpreters (Bajo et al. 2000: 132). It has been suggested that this superiority might be related to the assumption that interpreters use the capacity of their working memory more efficiently because they are less affected by the disruptive effect of the mechanism of articulatory suppression. Interpreters’ recurrent exposure to simultaneous listening and speaking could be an ideal exercise on working memory, which explains the decrease of the disruptive effect of the articulatory suppression mechanism (ibid: 140f).
2.2.2.1.4.2 Long-term memory (LTM) and interpreting
According to Darò (1997: 626), reduced LTM abilities (to recall previously translated material) in interpreters may be attributed to two factors. The first is the need to divide attention between the complex and simultaneous tasks. This need for attention-sharing intervenes in the work of the central executive of the working memory in that its capacity becomes overloaded when trying to achieve all these tasks. The second is an unconscious strategy developed by interpreters to control superfluous material since most of the information that interpreters deal with may not be relevant to their personal life.
However, Darò (ibid) uses the results of some experimental studies (cf. Spelke et al. 1976; Hirst et al. 1980) which have suggested that complex cognitive tasks requiring attention- sharing improve with training and specific exercises to suggest that implicit skill components such as concurrent listening and speaking as in SI, simultaneous listening and note-taking as in CCI, divided attention, etc. can also be improved by means of implicit exercising of the components of the whole process, a view that is also supported by recent experiments (cf. Padilla et al. 1995 and Bajo et al. 2000).
Darò (ibid: 626f) also suggests that semantic memory in interpreters and encyclopaedic knowledge may increase constantly through repeated exposure and practice, but that implicit (memory) skills are exposed to decay because of ageing, and this is detrimental particularly to interpreters because without these implicit procedural and highly automatised skills, interpreting can hardly be achieved.