one task may become automatic, and (b) interpreters may learn to switch attention between the tasks in a more efficient manner. Cowan’s explanation does not take into account the first hypothesis above (the extra-effort hypothesis) and instead focuses on the other two.
Some of these divided-attention exercises have received criticisms as well. For instance the technique consisting of counting backwards while simultaneously listening to a text has been criticized by Déjean Le Féal (1997, p. 617) who writes:
Although this exercise was designed as a way to make SI seem less daunting, I think that it instead makes the process appear to be more of a feat than it actually is, because SI in fact does not, at least as I see it, involve performing two different tasks at the same time. (my emphasis)
Here, Déjean Le Féal stresses the fact that although dual-tasking is involved in SI, it is not the simultaneous performing of two completely different tasks but rather two similar tasks. Therefore, concurrently performing two tasks that are different is more demanding than two tasks that are similar. She concludes that “An exercise that is meant to be introductory, but in fact introduces the student to something other than the task to be executed, would seem somewhat pointless.” (Déjean Le Féal, 1997, p. 618)
Anticipation relies heavily on the ability to make calculated guesses about what might be said later on or what might have been said previously. It is a form of compensation strategy widely used by interpreters. Good anticipation requires a wide vocabulary reservoir as well as grammatical knowledge, wide range of knowledge in diverse fields, sensitivity to the context, ability to discover relations, and the ability to analyze quickly. Typical exercises are giving trainees cloze texts, prediction exercises, asking them to complete unfinished sentences, etc. In what follows we will take a look at this most important mechanism in SI from a theoretical point of view.
In SI studies, anticipation, according to Van Besien (1999, p. 250) “refers to the simultaneous interpreter’s production of a constituent (a word or a group of words) in the target language before the speaker has uttered the corresponding constituent in the source language.”
Besides this main type of anticipation, there is an even more common type of anticipation in which the interpreter utters a constituent in the TL after the corresponding constituent has been vocalized in the SL, but in fact “so soon afterwards and at so correct a place in his own language that there is no doubt the interpreter summoned it before hearing the original” (Lederer 1978, p. 330, as cited in Van Besien, 1999, p. 251). Lederer (1981, p. 253, as quoted in Van Besien, 1999, p. 251) labels this kind of anticipation “freewheeling interpretation: at the moment that the interpreter has decided on the meaning of the speaker’s utterance, s/he listens to the speaker merely as a control, and the translation occurs within a very short delay.”
But is anticipation a phenomenon peculiar to SI or does it occur in other circumstances as well? To be able to answer this question, let us turn our attention to the following remarks by Miller (1951, p. 103, as cited in Chernov, 2004, p. 185):
If the successive units of a message are related, if the probability of a unit depends upon the units that precede it, these relations reduce the amount of information that a single unit can carry [. . .]. Contextual dependencies mean that the message source is repeating itself [. . .]. A large degree of interdependence among the successive units of a language means that parts of the message can be lost or distorted without causing a disruption of communication […]
Miller’s observation does not concern SI, but rather the process of communication through any single language. It is based on the idea of old information and new information (or theme and rheme); any stretch of language that seeks to serve the purpose of communication, strikes a balance between old information/theme (or what the utterance is about) and new information/rheme (or what is said about the theme). That is to say, not all parts of an utterance enjoy the same value information-wise. This explains why certain parts of an utterance (obviously from the thematic part) may be omitted without severely disrupting communication.
The same idea is echoed in the following paragraph from Van Besien (1999, p. 252):
Anticipation […] is also a feature of normal or monolingual language comprehension (cf. Clark & Clark 1977), but it is often claimed that interpreters are better at it (cf. Dillinger 1990; Pöchhacker 1994; see also Van Besien 1997).
Applying anticipation to simultaneous interpretation, Chernov (2004, p. 93) explains:
[…] the basic idea is that in the process of aural perception of speech, the simultaneous interpreter’s brain generates hypotheses in anticipation of certain verbal and semantic developments of the discourse. These hypotheses are based on subconscious subjective estimates of the range of probabilities within which the given verbal or semantic situation can further develop. In subsequent processes the interpreter either confirms or rejects her hypotheses by checking against critical points of the on-going discourse […]
As the interpreter starts listening to the source speech, he enters the world of the source text equipped with his knowledge reservoir (knowledge of the world, the subject matter, the speaker and the listeners, the communicative setting, the rules and norms of the SL, etc.). Making use of all this relevant wealth of knowledge as appropriate, he makes certain hypotheses about the development of the message being delivered (just as a listener does). Based on these hypotheses, the interpreter renders his interpretation in the TL. As the text keeps unfolding itself, the interpreter checks his previous anticipations and continues to make further anticipations regarding the incoming portions.
As our explanation in the previous paragraph shows, the information that interpreters use to anticipate what the speakers intend to say is of different types. Most authors have distinguished two main kinds: linguistic information or extralinguistic information (Gile, 1995; Lederer, 1978; Lederer, 1981; Seleskovitch, 1984; Wills, 1978, as cited in Van Besien, 1999). In the latter case the interpreter makes use of his general knowledge as well as knowledge of the situation. In the case of linguistic anticipation, “the interpreter predicts the appearance of a constituent on the basis of the syntactic and/or semantic information provided by the source language sentence.” (Van Besien, 1999, p. 251) Linguistic anticipation is made possible due to the existence of expectancy grammar, collocational tendencies, specific word combinations, etc.
Another point worth considering in this regard is that “Anticipation in simultaneous interpretation is generally considered as a language-specific phenomenon, i.e. it is particularly useful when source language and target language differ in their surface structures.” (Van Besien, 1999, p. 252) For instance if an interpreter is working from Persian (which is an SOV language) into English (which is an SVO language), it is obvious that, unless the sentence is short enough, he cannot afford to wait until the verb is uttered in the ST. Thus anticipation would be indispensable. The results of the study by Van Besien (1999, p. 254), who found that out of the 78 cases of anticipation, 60 were verbs, corroborate language-specificness of this phenomenon:
The fact that so many verbs were anticipated is evidence for the viewpoint that anticipation is language-specific: German and French differ in their surface structures and the position of the verb (early in French, late in German) is one of the main differences.
Anticipation can be classified as one of the comprehension strategies utilized by interpreters (Ribas, 2012). Other comprehension strategies include:
[…] segmentation, selection of information, stalling or waiting, while production strategies consist of compression, expansion, approximation strategies, generalisation, use of linguistic open-end forms, morphosyntactic transformation and the use of prosodic elements, such as pauses and intonation. (Ribas, 2012, p. 816)
Reporting the results of his study on German-French simultaneous interpretation, Van Besien (1999, p. 250) observes: “Anticipation was revealed to be a very frequent strategy, occurring every 85 seconds.” This single observation as to the occurrence frequency of anticipation among other SI strategies is enough to highlight the vitally important role it plays in making SI possible. That can very well explain why Chernov (2004, p. 91) contends “the basic mechanism making SI possible is the probability anticipation of the development of the message.” Still more interesting is the result pertaining to the analysis of the anticipated items in Van Besien’s (1999, p. 255) study:
On a total of 78 anticipations 49 are correct, i.e. they are correct translations of the source constituent that follows. In the remaining 29 cases an approximation is given. The number of correct translations is significantly higher than the number of approximations […]. In four cases the approximation was repaired after hearing the source constituent.
As to the conditions necessary for smooth interpretation to happen, Lee (2002, p. 604) holds “anticipation (Setton 1999), chunking (Barik 1975) and background knowledge will be necessary.”
Regarding the role of anticipation in interpreter training programs, Chrnov (2004, p. 199) writes “It is clear that to be productive, training methods should aim to increase the capacity for probability anticipation by increasing the subjective