This is the author’s self-archived preprint version (before proof) of the accepted article

Table 2. Four categories of strategies

Strategy groups	Specific strategies
Knowledge-based strategies	non-linguistic anticipation, inference, resorting to world knowledge, visualization, personal involvement, etc.
Language-based strategies	chunking/segmentation/salami, restructuring/changing order, transcodage/transcoding, reproduction, linguistic anticipation, parallel reformulation/substitution, morpho-syntactic transformation, transfer, etc.
Meaning-based strategies	compression/condensation/summarizing/filtering (selection of information, deletion, generalization, simplification), text expansion/addition/ elaboration (explanatory additions, additions to maintain coherence), adaptation, neutralization/evasion, omission/skipping/message abandonment, approximation/ attenuation, paraphrasing/explaining, etc.
Delivery-based strategies	décalage/time lag/extending or narrowing EVS, waiting/delaying response/tailing/stalling (waiting with fillers), repetition, use of prosodic elements (pause distribution, intonation), repair (self-correction), no repair (decision for no repair), monitoring, etc.

3. Why teach strategies to trainees?
3.1. General cognitive constraints in interpreting and the use of strategies

The immediate, multitasking nature of interpreting imposes severe cognitive demands. The possible sources of constraints force the interpreter to engage in competing activities. The factors that may cause comprehension and production difficulties include high time pressure, division of attention (between listening and note-taking in CI and between listening and production in SI), extreme speech conditions (subject novelty, high information density as is the case with less contextualized elements like figures, names, and listings and so on, high speech rate, syntactic complexity, non-native speakers with strong accents, and awkward syntactic structure, etc.), and unsatisfying working environment (poor sound quality, etc.) (Riccardi 1998: 173; Setton 1999: 35; Al-Qinai 2002: 310; Chernov 2004: 11; Gile 2009: 192; Pöchhacker 2009: 135). Meanwhile, the interpreter may have limited working memory capacity, deficient linguistic abilities, and insufficient extra-linguistic knowledge (Gile 1999a: 154, 2009: 182). Under such circumstances, the interpreter may not be able to produce high-quality target speech. In Gile’s (2009: 190) words, if the required processing capacity to interpret the source speech exceeds the interpreter’s available processing capacity at a given time in the interpreting process, problems arise.

To cope with problems encountered, the interpreter has to allocate his or her available processing capacity strategically during the comprehension of the source speech, production of the target speech, and monitoring of output. Those strategies can therefore be categorized into comprehension strategies, production strategies, and monitoring or global strategies (Riccardi 2002: 26; Gile 2009: 201; Kohn and Kalina 1996: 130-131). The use of strategies needs to be automatic so as to save cognitive capacity for complex operations (Kalina 2000: 6). Although some strategies violate the principle of faithfulness in interpreting, they help the interpreter use a minimum amount of processing effort for maximum meaning delivery, ease cognitive strain, overcome emergencies, and enhance the communicative relevance of target-language expression (Kohn and Kalina 1996: 130-131; Riccardi 1998: 174, 2002: 26; Gile 2009: 201; Lee 2013a: 180). Al-Qinai (2002: 318) argues that interpreters resort to a number of strategies that may ease the burden and improve the pace of delivery. Mizuno (2005: 750) finds that interpreters use load-reduction strategies to avoid the accumulation of untranslated information so that their memory and processing capacities will not be overloaded. For example, summarization, compression, or condensing (Sunnari 1995: 109; Dam 1996: 273; Kohn and Kalina 1996: 132; Pöchhacker 2004: 134; Chang 2005: 24), and omission (Gile 2009: 193) can be used to deal with fast delivery and message redundancy in both CI and SI. The use of omission or deletion (Viaggio 2002: 239; Gile 2009: 193) helps the interpreter cope with external difficulties, such as high delivery rate, high information density, repetition, redundancy, and strong accents. Anticipation, though more thought of as a strategy to cope with morpho-syntactic asymmetries between the source and target language, is found by Zanetti (1999: 90) to be constantly adopted during SI as a powerful tool, even between languages which are syntactically similar. The only problem is that the affinity of morpho-syntactic structures between the source and target language may make the existence of anticipation less evident.
These strategies are used to cope with constraints and are shared by all modes of interpreting. However, there are mode-specific factors that influence the use of strategies. SI, because of its higher time pressure, the interpreter’s being externally paced by the speaker, and simultaneity between listening, speaking, and monitoring, is more cognitively demanding and thus requires the use of complementary strategies that are not necessary in CI and monolingual communication (Kalina 1992: 255).
3.2. Mode-specific constraints and the use of mode-specific strategies

CI and SI share common cognitive constraints in the comprehension and production phase. Both modes involve high time pressure. Both require division of attention, between analytical listening and note-taking in CI, and between analytical listening and production in SI. In both modes, the interpreter may face extreme input conditions, such as fast delivery, high information density, strong accents, unfamiliar content, etc. Therefore, some strategies can be used in both modes, for example, addition, compression, omission, paraphrasing, repair, inference, etc.

However, CI and SI are performed under unique conditions which may require the use of unique strategies.
In SI, the constraints are more likely to drive the interpreter’s processing capacity to the point of saturation. The interpreter multitasks between three simultaneous tasks (analytical listening, producing, and monitoring) and is externally paced by the speaker. Different from CI where the interpreter has an overview of the macro- and micro-information structure of the source text in production, the interpreter’s semantic access depends on the gradual input of information units in SI (without texts), the macro-lelvel not being available at the point of production (Kalina 1994a: 221; Al-Qinai 2002: 309; Donato 2003: 102). Quick semantic access, coordination between listening and producing, and source speech interference are possible triggers of difficulties. In SI, failure in processing one segment may affect the interpreter’s available processing capacity to process incoming information. Therefore, the simultaneous interpreter is prone to cognitive overload.
Those mode-specific constraints influence cognitive resource management and determine the employment of mode-specific strategies (Kalina 1994a: 221; Hurtado Albir and Alves 2009: 63). The concurrent flow of input and output audio signals and high time pressure in SI makes it hard for the interpreter to free himself or herself from the interference of the source signal, and therefore he or she may resort to the strategy of transcoding (Kalina 1994a: 221). The simultaneous interpreter may base his or her production on anticipation when linguistic and extra-linguistic cues are available, and on segmentation wherever such cues are insufficient to anticipate what will come next (Riccardi 1998: 179; Seeber and Kerzel 2012: 232). Additionally, segmentation, salami or chunking (Meuleman and Van Besien 2009: 20-21), and paraphrasing and simplification (Kohn and Kalina 1996: 132; Chang 2005: 24) can offset the cognitive loads caused by linguistic difficulty, such as complex sentence structures in SI. Moreover, to buy more time for coordinating simultaneous comprehension and production, the interpreter may resort to waiting and stalling or the use of gap fillers (Seeber and Kerzel 2012: 232).
By contrast, the workload in CI is lower compared to that in SI. In CI, analytical listening and note-taking efforts are separate from production. The interpreter can ask the speaker to repeat or clarify what is incomprehensible. If the interpreter cannot handle both listening and note-taking because of extreme speech conditions, he or she can stop note-taking and rely on memory, thus reducing the cognitive load. The interpreter is not exposed to two channels of audio signals which may interfere with each other. Moreover, the interpreter is not externally paced by the speaker and is free to deliver at his or her own pace in the production phase. Therefore, considering the less severe workload, there are not many strategies specific to CI. In Kalina’s (2000: 7) words, strategies are not as crucial for CI as for SI. Most strategies applicable in CI also apply to monolingual communication.
However, CI is no easy task. The interpreter has to simultaneously handle analytical listening and note-taking which compete for resources. The interpreter’s note-taking is externally paced by the speaker. In addition, note-taking is a resource-consuming task, though it should have been acquired by the interpreter as a technical skill before he or she enters the industry. According to Mead (2000: 94), unsuccessful CI performances are partially attributed to inadequate note-taking skills (e.g. indecipherable symbols in note reading). Moreover, the interpreter has to store most information in memory because the notes are only for assistance. Therefore, these constraints require the use of strategies to reduce resource competition. Most strategies used in CI are also applicable in SI. If there are strategies that are more applicable in CI, visualization, a strategy used to avoid memory failure, is definitely among them. When dealing with a descriptive message, the interpreter may strengthen his or her understanding and memory of the original message by exercising his or her imagination and forming a mental picture of the content of the original speech (Chang 2005; Bartłomiejczyk 2006).
Another source of cognitive strain in interpreting, particularly SI, is the difference between the source and target languages, and between the two cultures, which may also cause the interpreter a resource-management crisis. Therefore, language- and culture-specific strategies are also necessary.
3.3. Language-pair-specific and cultural constraints and the use of language-pair- and culture-specific strategies

Whether language pairs pose language-specific constraints and affect the use of strategies is an ongoing debate (Setton 1999: 53; Donato 2003: 127). The two camps are called by Setton (1999: 53) “universalists” represented by the Paris School, and “bilateralists” represented by the Information Processing School.

According to the Paris School (Lederer 1999: 23), interpreting is a three-step language-independent process: the interpreter 1) understands a sense, 2) forgets the language forms and retains the sense (deverbalization), and 3) reproduces the sense in the other language (reverbalization). As long as the interpreter is competent in the source and target language, there should be no language-specific difficulties. Only those constraints that may impair monolingual communication affect interpreting. According to this view, the role of strategies to cope with linguistic asymmetries (e.g. between SVO and SOV languages) is minimized. The findings of Setton (1999) seem to support this argument. After analyzing the impact of asymmetrical structures in Chinese-English and German-English SI, Setton (1999: 282-283) concluded that “marked syntactic structure alone does not obstruct SI” and that the role of extralinguistic, contextual, and situational knowledge needs to be recognized.
The Information Processing (IP) theorists, however, think differently. It is believed that, though the essence of interpreting is to transmit ideas, the transmission process is made difficult by a number of factors including cognitive constraints and language- and culture-specific obstacles, which all require the use of strategies to overcome them (Gile 2005: 141).
In Donato’s (2003: 127) words, the IP theorists argue that the interpreter never entirely forgets the surface structure of the source text. Language-pair-specific constraints, such as structural asymmetries, may result in a crisis of resource management for the interpreter, and therefore require the use of language-specific strategies.
Research has demonstrated that language-pair-specific constraints affect the use of strategies. The degree of similarity between the source and target language may lower or increase the cognitive load of the interpreter. If the language pair is morpho-syntactically similar, the transfer is facilitated. For example, when interpreting from English into Italian simultaneously, the interpreter follows the surface structure of English and resorts to transcoding without much syntactic restructuring (Viezzi 1993: 100; Donato 2003: 129-130). Similarly, interpreting between English and French, two languages with parallel syntactic structures, is much easier to accomplish than that between English and German (Wilss 1978: 343).
However, if the languages involved are morpho-syntactically different, the interpreter’s processing capacity is more likely to be overloaded. The interpreter has to process larger segments before syntactic disambiguation, keep the processed information in memory, and restructure the message completely to comply with the rules of the target language (Riccardi 1998: 173; An 2009: 188; Liontou 2011: 152). For example, interpreting between SVO languages (English, French, Italian, etc.) and SOV languages (German, Japanese, Korean, etc.), and between languages with different information order (such as between Chinese and English), forces the interpreter to use the strategy of anticipation (Wilss 1978: 348; Kurz and Färber 2003: 123; Gile 2009: 174; An 2009: 188; Lim 2011: 59; Liontou 2012: 230), segmentation (Donato 2003: 129; Lee 2007: 153; Gile 2009: 205), restructuring (Wilss 1978: 343; Riccardi 1995: 216; Donato 2003: 129), waiting (Gile 2009: 204), or extension of time-lag (Donato 2003: 129).
Anticipation, for example, has been shown to be a language-specific strategy to overcome morpho-syntactical asymmetries between the source and target language (Kurz and Färber 2003). Recent studies have presented new evidence to support the hypothesis that SI between languages with asymmetrical structures involves more cases of anticipation. Donato’s (2003) results show that interpreting from German into Italian (from an SOV to an SVO language) involves far more anticipation than from English into Italian (from an SVO to an SVO language). Bartłomiejczyk (2006) has found that SI between English and Polish, two languages with similar structure, did not involve a single case of anticipation. Bevilacqua (2009) investigated whether different degrees of SOV-related difficulties required the use of language-specific strategies in SI from SOV languages to SVO languages. Although German and Dutch are both SOV languages, German is more rigid in terms of SOV structure while Dutch is less rigid and characterized by systematic violations of the SOV order. It has been found that SI from German to Italian requires more anticipation effort and longer Ear-Voice-Span regulation than that from Dutch to Italian. Liontou (2012: 229) has confirmed that, in German-Greek SI, anticipation is used where source and target syntax varies the most; specifically, in constructions which separate sentence constituents that belong together semantically.
Additionally, socio-linguistic difference also contributes to the use of specific strategies. Interpreting from a high-context and implicit source language (such as Chinese) to a low-context and explicit target language (such as English) requires more words and longer delivery time. High time pressure may force the interpreter to use the strategy of summarization or compression (Wu 2001: 84). Similar situations include different logic of expression (the unwillingness to state personal opinions of Japanese vs. the direct style of English), and the different degrees of grammatical redundancies (such as the flexible grammatical rules of Japanese vs. the more grammar-governed English, the concise nature of English vs. the more digressive pattern of Italian) (Snelling 1992: 11; Viezzi 1993: 100; Gile 2009: 197).
Moreover, lexical difference is another factor. Since the target text may not show the same cohesive ties as the source text, the interpreter may reorganize the message by changing, adding, splitting or deleting linguistic elements to comply with the target-language norms (Al-Qinai 2002: 319; Straniero Sergio: 2003: 143).
Research on interpreting directions and strategies also sheds light on the existence of language-specific strategies. It has been shown that working into the A language and into the B language requires the use of different strategies (Bartłomiejczyk 2006; Chang and Shallert 2007).
Interpreters use strategies to deal with culture-specific elements. According to Pöchhacker (2007: 140) who investigated how media interpreters working in high-stress conditions cope with culture-specific references, culture gets lost in the target text because interpreters tend to use the strategy of omission frequently. Wu (2001: 85) suggests the use of substitution, paraphrasing or omission as strategies when dealing with culture-loaded words. Another example with Japanese is given by Gile (2009: 198) who holds that deliberate ambiguity in making commitments and expressing rejections in Japanese may be less or not at all acceptable in a different culture. This may require the use of the strategy of paraphrasing, omission, or addition to adapt the content to meet the expectation of the target listeners.
Besides general cognitive constraints, mode-specific constraints, and language-pair- and culture-specific constraints, norms also contribute to the interpreter’s strategic decisions.

3.4. Interpreting norms and strategy use

Interpreting performance is norm-based and interpreters need norms to guide them to select appropriate solutions to the problems they meet (Schjoldager 1995: 67; Shlesinger 1999: 69; Gile 1999b: 99; Pöchhacker 2004: 132; Wang 2012: 198).

Interpreting norms are values and ideas of what counts as correct and appropriate behaviors in concrete situations; norms are developed and internalized through training, professional practice, observation of peers, and socialization, and are shared by members of the interpreting community, the aim being to meet quality standards (Schjoldager 1995: 67; Chesterman 1997: 64; Garzone 2002: 110; Straniero Sergio 2003: 135).
Interpreting norms include the following: maximizing information recovery and communicative impact, minimizing recovery interference, the rule of least effort, self-protection, adequacy, explicitation in logical relationships and meaning, specificity in information content, concentration on essentials instead of completeness of rendition, faithfulness to meaning instead of literal reproduction, generalization or omission of elegant and recherché adjectives, neutralization of figurative expressions, allowance of parallel reformulation (introducing a textual unit that is contextually plausible but which is not a faithful reflection of the source language speech), the norm of condensation, fluent output, and others (Gile 2009: 211; Schjoldager 1995: 84; Moser 1996; Shlesinger 1999: 69; Straniero Sergio 2003: 170; Pöchhacker 2004: 132; Wang 2012: 198).
These norms are rules that license the use of certain strategies (Riccardi 2005: 755). The goals of strategies are to conform to the relevant interpreting norms (Chesterman 1993: 14). Governed by interpreting norms, interpreters are justified in using certain strategies to cope with problems that they meet. For example, the interpreter may use the chunking strategy to simplify the hypotactic structure of a sentence (Garzone 2002: 113), conforming to the rule of least effort. According to Straniero Sergio (2003: 139), under the pressure of frequent technical terms, strong accents, and fast delivery rate in the context of media interpreting, the interpreter may resort to emergency strategies such as summarization, omission, and generalization, conforming to the norms of generalization or omission of elegant and recherché adjectives, condensation, and neutralization of figurative expressions.
3.5. Interpreting strategies and interpreter education

Strategies, if activated automatically, help students overcome constraints and manage their processing capacity more efficiently. As agreed by Kader and Seubert (2015: 140-142), part of the didactic mission of interpreting teaching is to help trainees gradually identify, apply and then automatize strategy use, which minimizes their cognitive load and thus ensures optimal handling of challenges during the interpreting process.

One of the reasons for teaching them is that strategies highlight the problems in rendering the source speech into the target speech and their solutions. As Riccardi (2005: 753) believes, strategies point to typical challenges resulting from the interpreting process and indicate what solutions have been applied by the interpreter. Therefore, they can be used as a tool to orient students’ learning and to devise exercises that help them automatize specific interpreting solutions. A high level of automation, although challenging, should be the ultimate goal of strategy training, because only when strategies can be activated automatically can students cope most efficiently with problems that they may encounter in interpreting (An 2009: 206; Lee 2013b: 26). A similar view is held by Bartłomiejczyk (2006: 151) who argues that “successful repeated use of a specific strategy leads to automation” and that “automated strategic processes reduce the cognitive load of interpreting.” Trainers may group strategies into “general interpreting strategies, independent of the language pair used”, or “language-pair-specific strategies, taking into account solutions imposed by structural and lexical diversities of the languages used” (Riccardi 2005: 765). Then exercises can be devised to help trainees automatize the use of strategies before using them to overcome constraints in certain interpreting tasks. For example, the strategy of adequate condensing or compression may be taught to students to help them cope with limitations in memory and with time pressure, enabling them to make conscious use of memory instead of leaving out elements at random (Dam 1993: 311). The strategy of segmentation and tailing may also be taught to students to deal with complex structures and high speed of delivery in SI (Meuleman and Van Besien 2009: 32).
In addition, differences in strategy use between novice and expert interpreters can be compared for pedagogical purposes. According to Riccardi (2002: 26), comparison of professionals’ and students’ performances shows to what extent their interpreting behavior differs and in what direction it should develop over time. Evidence of novice-expert difference in strategy use is plentiful. Firstly, it has been found that seasoned interpreters resort to effortful strategies enabling listeners to follow their delivery with minimum effort, while trainees resort to easy and effortless strategies which result in a fragmented output (Sunnari and Hild 2010: 38). Secondly, professionals do not repair minor errors as frequently as trainees who are over-concerned with slight errors, thus negatively affecting their performance (Kalina 1994b: 229-231). Thirdly, experienced interpreters adopt strategies involving compression or condensing, either at lexical or syntactic level, filtering out redundant elements, whereas trainees have not yet mastered such strategies and fail to produce high-quality output (Sunnari 1995: 118; Chernov 2004: 113). Fourthly, experts rely more on macro-level or knowledge-bound strategies and move from the known to the unknown in comprehension, while novices favor low-level micro-contextual or language-bound strategies and focus firstly on the unknown (Sunnari 1995: 118; Moser-Mercer 1997: 257; Riccardi 2002: 26). Fifth, unlike professionals who can use segmentation successfully as a strategy, novices tend to tail the source-text sentence and thus commit more errors (Kalina 1998: 193). Sixth, professionals tend to use omission and are more selective in terms of what to interpret and what not to interpret than novices (Kalina 1998: 193).
Moreover, the ability to apply appropriate strategies is a necessary component of interpreter competence which is the basis and goal of interpreting teaching (Kalina 2000). As one of the most important elements in translator competence, strategic competence encompasses procedural knowledge (know-how) to ensure the efficiency of translation process and solve problems encountered during that process (PACTE 2011: 319). Though strategic competence is important for both translation and interpreting, strategy application is more typical of and crucial for high-quality performance in interpreting than in translation (Kalina 2000: 7). This is because of the existence of more interpreting-specific constraints, which require resource-crisis management and which force the interpreter to make strategic decisions. According to Kalina (2000: 7), the strategies that should be treated as part of interpreting competence include comprehension strategies (segmentation of input, anticipation, inference, accessing previously-stored knowledge, building relations between stored and new information, etc.), production strategies (restructuring, paraphrasing, condensing, expanding, the use of prosodic or non-verbal features, etc.), and global strategies (memorizing the input, adapting one’s mental model, monitoring of output, repair of errors, etc.).
Lastly, strategy use needs to be taught consciously. The learning of strategies involves the development of declarative knowledge (know-what) into procedural knowledge (know-how). To help novices acquire strategies used by professionals and reach a high level of automation (Kalina 2000: 5; An 2009: 206), teachers need to treat strategies as crucial teaching content. According to Anderson’s (1995) description of expertise acquisition, novices firstly develop declarative knowledge about a skill, subsequently strengthen the connections among the various elements required for successful performance of the skill through trial-and-error, and finally complete the learning of the skill by automatizing the procedures. Such a shift from conscious application to unconscious and automatic application of strategies, and from declarative knowledge to procedural knowledge, if assisted by teachers, can results in more efficient interpreting by students.