Is working memory working in consecutive interpreting?
Abstract
It is generally agreed that language interpreting is cognitively demanding; how-
ever, to date there is little evidence to indicate how working memory is involved
in the task, perhaps due to methodological limitations. Based on a full considera-
tion of key components of interpreting, two series of experiments were conducted
to explore how working memory might play a role in discourse and sentence inter-
preting. If working memory is implicated both in grammatical encoding into the
target language, and in temporary storage of the discourse content, then higher
demand in one function might compromise the other. Thus discourses that di er
in word orders between languages could increase the processing load and leave less
resource for memory maintenance, a ecting recall performance. In Experiment 1,
Chinese-English bilingual participants' memory performance was compared when
they translated passages from Chinese to English and from English to Chinese,
where the expected word order was either congruent or incongruent between source
and target. Recall was not sensitive to word order or direction of translation. Per-
haps surprisingly, memory for incongruent discourses was numerically better than
that for congruent sentences. Experiment 2 showed that interpreting trainees per-
formed just like the participants in Experiment 1 did, suggesting that memory
performance was not modulated by translation direction in pro cient translators.
Experiment 3 explored the relationship between surface form transformation and
recall. As discourse paraphrasing did not result in better recall than verbatim recall,
it was concluded that the better memory performance for incongruent discourse in-
terpreting suggested by Experiment 1 was not the result of active manipulation of
word form or word order in interpreting. Finally, a free recall task among native
English speakers showed that the incongruent discourses tested in earlier experi-
ments were intrinsically more memorable than congruent discourses (Experiment
4). Despite this confound, this series of experiments highlighted the importance of
comprehension in interpreting, but it did not rule out the role of working memory
in the task. The role of working memory in interpreting was further explored using on-line measures in
Experiments 5-8. Experiment 5 replicated a self-paced reading study by Ruiz, Paredes,
Macizo, and Bajo (2008), comparing participants’ times to read sentences for translation to
those to read them normally. The data showed that participants accessed lexical and
syntactic properties of a target language in the reading-for-translation condition when
resources were available to them. In order to explore the role of working memory in sentence
interpreting, a dual-task paradigm was used in Experiment 6. When participants' working
memory was occupied by a secondary task (digit preload), reading times were only different
numerically between congruent and incongruent sentences. Crucially, reading times decreased
as digit preload increased. Since there were no differences in the interpretations produced or in
digit recall, it appears that participants were flexible in their resource allocation, suggesting
that processing can be tuned up to optimise performance for concurrent tasks. Experiment 7
refined the procedure in the order of responses for the dual tasks but replicated the results of
Experiment 6. A closer examination of participants’ interpretation responses showed that
devices that could reduce processing load in target language production may have been
strategically employed. Finally, another set of sentences were used in Experiment 8 in an
attempt to replicate Experiment 5. A failure to replicate the earlier findings suggested that
working memory demand might differ for different syntactic structures in sentence interpreting.
All in all, this thesis shows that research in language interpreting benefits by taking a full
account of the key components of interpreting. The use of on-line measures allowed us to
take a ne-grained approach to the investigation of interpretation processes. It is proposed in
this thesis that interpreting research may gain more insight from the data by incorporating
some of the theories and methods typically used in research into language production.