Last Fall, I conducted a survey in which the participants were asked to complete 3 cognitive tests. The purpose of this informal research project was to find out more about the X-factor that sets apart skilled translators from other bilinguals (see my previous two posts). The test results suggest that translation ability requires highly developed executive function skills.
These results were also published today in the Bulletin, the bimonthly journal of the Institute of Translation and Interpretation in the UK. The editors rewrote text in the article somewhat to suit their style and guidelines, and it is their policy to not include references. So for the sake of completeness, here is the original text complete with footnotes.
This project was inspired by a study commissioned by the FBI on what it takes to be a translator.1 The FBI employs well over a thousand translators, and given the mission of the organization, “to protect the United States from threats both international and domestic”, the quality of the translations is obviously a crucial concern. Which leads to the logical question: how do you make sure that the translators are up to the job? The FBI study set out to determine whether high scores on language comprehension tests equate translation ability. The study concluded, to the surprise of no translator anywhere, that this is not the case, and that translation ability was an additional “X-factor” that could only be assessed by means of an actual translation test.
Although the FBI study may seem a bit like kicking in an open door, it addresses a valid concern. Language comprehension tests are objective and can be graded by anyone with the answer sheet. Translation tests are trickier because more subjective, and they have to be assessed by another translator specialized in the same language pair. A large organization whose primary business is not translation cannot be blamed for trying to find the most efficient way to test for this skill.
But what if there was a way to test for objective predictors of translation ability? The invention of functional MRI (fMRI) in 1990 made it possible for researchers to measure and map brain activity, and by 2012 it had already been used in almost 250,000 studies cognitive neuroscience and other fields.2 As it happens, a handful of these fMRI studies focused specifically on what happens in the brain during translation activity. This is what these studies found:
- Translating activates different parts of the brain than learning or using a second language.3 This confirms the findings of the FBI study. It may seem counterintuitive, since you can’t translate without using language, but the phenomenon is not unique: MRI scans have shown that mathematicians use an entirely different part of the brain to think about advanced math problems than the areas associated with learning math and performing basic operations.4 Obviously the basics have to be mastered first, but just as not everyone who passes a calculus class is a mathematician, not everyone who speaks a second language is a translator.
- Although the findings of the fMRI studies on translation activity were not identical, there was significant overlap: all the studies observed activation in the anterior cingulate cortex (ACC) and the basal ganglia (in particular the caudate nucleus and the putamen).5,6
So what do we know about these areas of the brain? Once again, fMRI studies prove helpful. The ACC is involved in “high-level processing (in outcome/error monitoring and action planning”.7 The caudate nucleus and the putamen are necessary for “attention, controlled retrieval, and monitoring of information within working memory”.8 These brain functions are “cognitive mechanisms by which performance is optimized in situations requiring the simultaneous operation of a number of different processes. They are regularly called upon when sequences of responses must be generated and scheduled, and when novel plans of action must be formulated and conducted”.9
Together, these abilities are commonly referred to as executive function or executive control.
I think that the elusive X-factor referred to in the FBI study is highly developed executive control. And executive control is something that can, in fact, be tested. Various cognitive assessments have been developed over the years that focus on specific executive functions. So if translation is characterized by the activation of brain areas involved in executive control, translators should perform well on these cognitive assessments.
To test this idea, I used software from PsyToolkit10,11 to create a short online survey with three standard cognitive assessment games that are known to recruit the ACC and the basal ganglia.12,13,14
- The Stroop test
In this test, subjects are presented with a list of words for colors, and the words themselves are printed in various colors. The subjects are asked to name the color the word is printed in. The idea is that it’s much more difficult to name the color when the word itself spells out a different color, because our brains are so conditioned to read text. This difficulty is called “Stroop interference”. The more well developed the subject’s executive function (specifically inhibitory control), the fewer mistakes he or she will make.
- The Flanker task
In this test, subjects have to respond to a central stimulus (usually a letter or an arrow), which is “flanked” on either side by other stimuli that are irrelevant but still interfere with the response. Again, lower error rate means better inhibitory control.
- The Wisconsin Card Sorting test
In this test, subjects are presented with consecutive sets of cards. They have to pick a card and then they are told whether or not their selection is correct. Based on the feedback, the subject has to figure out the unspoken rule, which is to select a card based on color, shape or number. Once the subject figures out the rule, he or she will pick the right cards until the rule is changed. At that point, the subject has to figure out the new rule, and so on. The first guess is by definition random, the second guess has a 50% chance of being correct, but after that the correct answer should be clear. The index for executive control is response time and error rate for the third guess.
To compare translators with the general population, I asked translators and a control group of non-translators to complete the survey. I received responses from 331 translators and 60 non-translators. I also asked the translators to specify their years of experience and the non-translators to specify their occupation. This enabled me to see whether performance on the cognitive tests was related to years of translation experience, and how translators compared to specific professions in terms of the results. I also removed a few outlier responses (for example from respondents who failed to answer questions or didn’t complete the tests).
The mean response times and Stroop interference were slightly higher for translators than for the control group. This is not surprising, since translation by definition requires a laser focus on the meaning of text, so it will take even more effort to focus on other aspects. However, what matters here is the mean error rate, the index of inhibitory control. And we see that despite increased interference, the error rate for translators was less than half of that of the control group.
In this task, in which the translators were not hampered by textual interference, they were faster under both conditions with a significantly lower error rate than the control group, suggesting more effective inhibitory control.
The mean response time for translators was lower when figuring out the rule, with a steeper drop in response time when they moved from guessing to certainty, and an error rate that was 5 times lower than that of the control group.
The survey results support the idea that translators have highly developed executive function skills, as they scored consistently higher than the control group in all index categories for executive control.
To refine the results, I also compared the performance of the translators with that of the highest-scoring control participants. The controls with the highest scores worked in fields related to research/science and engineering. I found that the translator results were on par with the results for those participants. This suggests that translation and those particular fields require the same type of executive ability, applied to different problems.
Interestingly, there was no correlation between translator scores and years of translation experience. This could also be an indication that executive function is a separate, underlying trait that is a predictor for translation ability, not something that develops as a result of translation experience.
Due to the relatively small sample size and the limited scope of the assessment it would be premature to draw any definite conclusions at this point, but it does seem to me that the results are interesting enough to warrant further research to confirm the existence of this “Executive Factor”.
1 Brooks, Rachel Lunde, and Maria Brau. “Testing the Right Skill: Evidence to Support Testing Translation Ability.” Linguistica Antverpiensia, New Series – Themes in Translation Studies. Accessed January 10, 2021. https://lans-tts.uantwerpen.be/index.php/LANS-TTS/article/view/448.
2 Glover GH; “Overview of Functional Magnetic Resonance Imaging,” Neurosurgery clinics of North America (U.S. National Library of Medicine), accessed January 10, 2021, https://pubmed.ncbi.nlm.nih.gov/21435566/.
3 Price CJ;Green DW;von Studnitz R; “A Functional Imaging Study of Translation and Language Switching,” Brain : a journal of neurology (U.S. National Library of Medicine), accessed January 10, 2021, https://pubmed.ncbi.nlm.nih.gov/10581218/. – Cathy J. Price et al. – Brain (1999), 122
4 “in professional mathematicians, a specific set of areas was activated when they were thinking about math problems. […] These areas weren’t activated by non-math problems, and were not activated in the non-mathematicians who treated complex mathematical statements as gibberish.” Price CJ;Green DW;von Studnitz R; “A Functional Imaging Study of Translation and Language Switching,” Brain : a journal of neurology (U.S. National Library of Medicine), accessed January 10, 2021, https://pubmed.ncbi.nlm.nih.gov/10581218/.”).
5 Keerthi Ramanujan, Man Wai Kong, and Brendan Weekes, “An FMRI Study of Executive Control during Translation and Oral Reading in Cantonese-English Bilingual Speakers,” Frontiers in Human Neuroscience 11 (2017), https://doi.org/10.3389/conf.fnhum.2017.223.00040.
6 Price CJ;Green DW;von Studnitz R; “A Functional Imaging Study of Translation and Language Switching,” Brain : a journal of neurology (U.S. National Library of Medicine), accessed January 10, 2021, https://pubmed.ncbi.nlm.nih.gov/10581218/. – Cathy J. Price et al. – Brain (1999), 122
7 Bush et al., 2000
8 Sandra E Leh, Michael Petrides, and Antonio P Strafella, “The Neural Circuitry of Executive Functions in Healthy Subjects and Parkinson’s Disease,” Neuropsychopharmacology 35, no. 1 (May 2009): pp. 70-85, https://doi.org/10.1038/npp.2009.88, p.4
9 Sandra E Leh, Michael Petrides, and Antonio P Strafella, “The Neural Circuitry of Executive Functions in Healthy Subjects and Parkinson’s Disease,” Neuropsychopharmacology 35, no. 1 (May 2009): pp. 70-85, https://doi.org/10.1038/npp.2009.88, p.1
10 Stoet, G. (2010). PsyToolkit – A software package for programming psychological experiments using Linux. Behavior Research Methods, 42(4), 1096-1104.
11 Stoet, G. (2017). PsyToolkit: A novel web-based method for running online questionnaires and reaction-time experiments. Teaching of Psychology, 44(1), 24-31.
12 “Performance of the conventional Stroop specifically activated the anterior cingulate, insula, premotor and inferior frontal regions.” Hoi-Chung Leung et al., “Event-Related Functional MRI Study of the Stroop Color Word Interference Task,” OUP Academic (Oxford University Press, June 1, 2000), https://academic.oup.com/cercor/article/10/6/552/347510.
13 “In an fMRI experiment  we observed increased BOLD activation in the caudate, anterior cingulate cortex (ACC), and superior and middle frontal gyri during performance of the flanker task.” Rajendra D Badgaiyan and David Wack, “Evidence of Dopaminergic Processing of Executive Inhibition,” PloS one (Public Library of Science, 2011), https://www.ncbi.nlm.nih.gov/pubmed/22162756. 14 “Poor WCST performers had a reduction of the left caudate nucleus and putamen, and right total striatum when compared to 24 healthy controls.” Paolo Stratta et al., “Association between Striatal Reduction and Poor Wisconsin Card Sorting Test Performance in Patients with Schizophrenia,” Biological Psychiatry (Elsevier, May 28, 1998), https://www.sciencedirect.com/science/article/pii/S0006322397000176.