Skilled memory in interpreting expertise

In the present study, the brainwave pattern of the Professionals differed from those of Students and Graduates in the congruent conditions in both the N400 and PNP time windows. Professionals had an attenuated N400 for congruent target items, signifying a lesser difficulty in retrieving these words. This neural correlate

corroborates behavioral findings in which expert interpreters were quicker at accessing lexical and semantic information (Bajo et al., 2000; Christoffels et al., 2006). As discussed in previous sections, this phenomenon could be explained by their better use of anticipation strategies, as they pre-activated the lexical and/or semantic features of the anticipated target.

The skilled memory theory, proposed by Chase and Ericsson (1982), and the long-term working memory (LT-WM) model (Ericsson & Kintsch, 1995) based on the theory, could be particularly useful in illustrating how the Professionals retrieved the congruent words more easily, as indexed by the attenuated N400.

There are three principles in the skilled memory theory (Chase & Ericsson,

81

1982). The first is called the meaningful encoding principle, which states that experts use prior knowledge to encode new domain-specific information.

Meaningful encoding creates multiple cues for later retrieval in LT-WM. Experts have acquired skills to encode relevant information in long-term memory (LTM) in a manner that allows automatic retrieval from LTM when needed. The second principle is the retrieval structure principle, which states that individuals can acquire memory mechanisms that facilitate retrieval of information. Retrieval structures are basically a set of retrieval cues that are organized in a stable structure. They are simpler and more generic forms of patterns or schemas.¹⁶ The existence of retrieval structures and patterns/schemas means that the encoded information in LTM has a vast and rich network of links to various kinds of retrieval cues. The third principle is the speed up principle, which states that encoding and retrieval processes become faster with practice. After sufficient practice and training, skilled individuals can directly access the information stored in LTM by leveraging acquired knowledge and systems of retrieval cues. Automatic retrieval eventually becomes possible.

Applying this theory to the case of the Professionals’ mental lexicon, linguistic information of the congruent target item (e.g. meaning, pronunciation, grammatical rules, etc.) is the encoded information (Ericsson & Kintsch, 1995). This information has been encoded in various methods (e.g. memorization, repetition, inference,

association, etc.) and settings (e.g. while preparing for a conference, while interpreting a particular passage, etc.), enabling the said item to be accessed in multiple ways. Each time it is accessed, the retrieval structures would be enriched, adjusted, and strengthened, facilitating the next access. When this retrieval structure is accessed multiple times to the extent that it becomes stable, lexical access for the

16 Patterns and schemas are more elaborate and specific forms of retrieval structures that were built based on prior systematic knowledge (as in a particular academic discipline) and experiences (as in the schemata of eating in a restaurant).

82

congruent item would become an automatic process (Feltovich et al., 2006). The congruent target items used in the experiment materials were terms that have

reasonably high frequencies,¹⁷ and since they were expected endings to the preceding sentence parts (as indexed by the high cloze probability during the normalizing pre-test), the semantic distance between the target and context was relatively close, as indexed by the cosine of LSA. With more than a decade of interpreting experience, the retrieval structures and schemata that could access the congruent target items stored in the Professionals’ LTM could be richly interlaced (Hoffman, 1997).

Therefore, they could be easily accessed by the meaning that was constructed by the prior context. In other words, the Professionals’ lexical access during the N400 time window could be an automatic process. Of course, it is not impossible that some Graduates and Students probably accessed the terms as easily or even automatically, as individual variability is always a factor to consider. However, as a group, it seemed that Professionals overall accessed the words with less effort.

One could question the impact or hindrance caused by such rich retrieval

structures, because if the congruent items could be accessed in so many ways, it could be possible that more time is spent on deciding which route to access. It is also possible that the prior context activated multiple retrieval cues, and may not lead the Professionals to the correct encoded piece of information in LTM after all. The Professionals did not react particularly fast in the congruent condition, so this possibility could not be ruled out. However, it has been identified that experts are very selective in accessing relevant information (Feltovich et al., 2006). Based on either the importance distribution of features extracted from the stimuli or their own internal cognitive processes, experts have learnt to separate signal from noise and

17 The average frequency of the target words was 249 (upper list: 228; lower list: 270). This number corresponded to a ranking of 2299 in the 93826 terms stored in the Sinica Corpus, which means that these terms were among the top 2.4% frequent terms in the corpus.

83

could filter out information that would not be useful to task execution. This expert judgment in selecting relevant and important information to interpret has also been demonstrated in expert interpreters (Liu et al., 2004).