When Bilinguals Speak
Anyone observing bilinguals speaking to different people during any one day will notice quite readily that they keep to one language when they are communicating with people who do not know their other language(s). However, they may well code-switch into, or borrow from, their other language(s) when their interlocutors know the same languages and the situation is conducive to language intermingling.
Bilinguals navigate along a continuum with two endpoints - a monolingual language mode where only one language is fully active and a bilingual language mode where several languages can be active. The consequence of this is that the state of activation of bilinguals’ languages will vary from moment to moment. Numerous factors, both internal and external, control the level of activation at any given time.
Psycholinguists have developed very refined experimental procedures to show that bilingual language production is a dynamic process which can operate in different language activation states. Recently, Dutch researchers Daan Hermans, Ellen Ormel, Ria Besselaar and Janet van Hell undertook a study which shows that even the lexical similarity between the two languages known by the bilingual, under certain circumstances, can play a role in changing the level of activation of the bilingual’s languages. They did this by manipulating the presence of cognates, i.e. translation equivalents that have similar orthographic and phonological forms in two languages, such as “apple” in English and “appel” in Dutch.
The researchers asked Dutch-English bilinguals to look at pictures on a computer screen followed by a letter representing a phoneme (e.g. the letter “b” represents the phoneme /b/). The bilinguals had to decide whether the phoneme was part of the English name of the picture being presented; they did so by pushing on a “yes” or a “no” button. There were three conditions in this phoneme monitoring study:
a) In the affirmative condition, the phoneme was indeed part of the English name of the picture. For example, the picture was that of a bottle and it was followed by a “b” or a “t”; the answer was “yes” therefore.
b) In the cross-language condition, the phoneme was not part of the English name but rather of the Dutch name of the picture. For example, “f” was presented and it is part of “fles”, the Dutch translation equivalent of “bottle”. Here the answer was “no” therefore (recall that participants had to base themselves on the English name of the picture).
c) Finally, in the unrelated condition, the phoneme was neither part of the English nor of the Dutch name. For example, “p” is not part of “bottle” nor of “fles”.
Now came the subtlety permitted by a good experimental design. The pictures shown to the bilinguals were divided up into two categories: half the pictures were used in the experimental condition where there was an English name which had a noncognate translation equivalent in Dutch. Examples were English “bottle” and Dutch “fles”, English “pillow” and Dutch “kussen”, etc. The other half of the pictures were used in the filler condition. It is here that the experiments that were run by the researchers differed from one another (we will look at two of them). In the first, all the filler pictures also had noncognate names in Dutch and English. Hence, if one adds the two halves of the experiment, no picture was followed by a letter that corresponded to a sound in the Dutch name of the picture. In sum, the experiment was monolingual, both overtly and covertly.
The results they obtained in this first experiment showed that there was no difference between the cross-language condition and the unrelated condition (the two conditions of interest), be it in reaction times or in accuracy scores. Basically, the Dutch translation equivalents of the English names of the pictures were not active. In other words, the participants were in a monolingual English mode.
In their second experiment, all the authors did was to change the pictures in the filler condition. They now had cognate names in English and Dutch, such as “moon” and Dutch “maan”, “mouse” and Dutch “muis”, and so on. (Note that they did not change the pictures in the experimental condition). This time the two critical conditions (cross-language and unrelated) did produce different response latencies and accuracy scores. It took the participants more time to do the task in the cross-language condition than in the unrelated condition, and they were less accurate in the former condition. What was happening there was that the phonological representations of the Dutch picture names were activated and they slowed down the response regarding the presence of a phoneme in the English name.
Based on these findings, the authors concluded that the bilingual language production system is indeed dynamic and that it can operate in different activation states depending on a number of factors. The level of activation of the bilingual’s languages will be due to linguistic factors, such as in the above, but also psycholinguistic and sociolinguistic factors such as who you are talking to, whether you are using the “right language” to talk about the subject in question, how well you know the language you are speaking, how recently you have spoken the other language, the presence of speakers of the other language(s), and so on. The bilingual production process is wonderfully sensitive to all these factors and this promises many intriguing research findings in the years to come.