 |
Internet-Zeitschrift für Kulturwissenschaften | 17. Nr. |
Januar 2010 |
| Sektion 2.7. |
Neue Entwicklungen in der Psycholinguistik / New Developments in Psycholinguistics SektionsleiterInnen | Section Chairs: Elly Brosig (Stuttgart, Germany) |
||
Identification and Control of Multiple Languages in the Brain
Elly Brosig (Stuttgart, Germany) [BIO]
Email: brosig.emk@gmx.de
1. Introduction:
When we hear or read a sound or word sequence how do we decide to which language this verbal input belongs? - What are the cues that make us identify this sound or letter sequence as e.g. German, English, or Polish, even when we do not know the meaning? - Evidently there must be some specific phonetic, orthographic, and morphological features inherent of each language which help us identify an input as e.g. German, English, or Polish.
2. Methods:
To find out what features are responsible for identifying a language, the author developed an experiment, in which the subjects had to decide, whether 10 written and 10 orally presented words or non-words could be English.(They were not asked: „Is this an English word?“) -
At first subjects were presented the written list, which consisted of three English words that were quite rare (putrid, amanuensis, mullion), three non-words with letter sequences that were possible in English (noop, vest, fither), three non-words that were unlikely (mux, falen, kester), and one that that was morphologically impossible in English (torpir).
Thereafter, 10 words or non-words were read to the subjects. - Two infrequent English words were pronounced with English pronunciation (herb, knit). Two current English words were pronounced according to German phonetics (grimp, knock), likewise three non-words with possible English orthography (filt, mux, quade); one non-word was pronounced in English (fubble). Two French words were pronounced with the typical French pronunciation, one with the typical stress on the last syllable (tira; passé simple of „tirer“).
The words of both lists were presented in mixed order.
Subjects were 106 students, 66 of them graduate students of English; 40 were students of Germanics, but with a good school knowledge of English. The subjects had further been differentiated as to their first language. 13 of each group were found to be non-native speakers of German. None of them was an English native speaker.
3. Results:
Results revealed that in written language subjects rely primarily on orthographic and morphological cues, which was to be expected. (Does a word look English?)
As a consequence, such rare English words as „amanuensis“ (a ghost writer) were only deemed English by 42 % of the graduate English students and 33 % of the Germanic students. (See Table 1.) - „Putrid“ was known by 77 % of the English and 65 % of the German group, which shows their rather good knowledge of English. - On the whole, the students of English were generally more inclined to accept written words as possibly English. However, both groups were equally rejecting morphologically impossible words in English. The non-word „torpir“ was rejected by 88 % of the English group and even 92 % of the German group.
Spoken speech, however, seems to be largely identified by its phonetic similiarity to the target language. English words pronounced with German pronunciation were often not identified; for instance „grimp“ was not identified by 44 % of the English group and 28 % of the German group. - With „knock“ the difference was even more pronounced. Only 12 % of the English group accepted it , but 60 % of the German group. (Maybe they were less sensitive to the differences in English pronunciation.) - However, the French word „null“ with French pronunciation was almost equally rejected by about 90 % of both groups.
Figure 1: Results of language identification test
Visual presentation |
Auditive presentation |
||||||
N = 66 grad.Engl. |
N = 40 Germanic |
N = 66 grad.Engl. |
N = 40 Germanic |
||||
| 1) mux | 36 % |
25 % |
1) grimp | 44 % |
28 % |
||
| 2) falen | 24 % |
53 % |
2) herb | 85 % |
63 % |
||
| 3) putrid | 77 % |
65 % |
3) filt | 47 % |
43 % |
||
| 4) noop | 89 % |
63 % |
4) mux | 26 % |
8 % |
||
| 5) vest | 88 % |
80 % |
5) knit | 52 % |
45 % |
||
| 6) amanuensis | 42 % |
33 % |
6) quade |
8 % |
10 % |
||
| 7) fither | 89 % |
88 % |
7) fubble | 50 % |
60 % |
||
| 8) torpir | 12 % |
8 % |
8) knock | 12 % |
60 % |
||
| 9) mullion | 67 % |
58 % |
9) tira | 20 % |
18 % |
||
| 10) kester | 20 % |
35 % |
10) null |
8 % |
10 % |
||
Most interesting and a really new finding was that not only phonetic, morphological, and semantic features seem to be important for identification of spoken speech, but also prosodic ones. The word „tira“ (passé simple from „tirer“ - to pull) with the un-English stress on the last syllable was rejected by about 80 % of both groups.
4. Interpretation:
These results led the author to the assumption that visual (orthographic), phonetic, prosodic, and morphological features must be properties of a language tag, defining a word for instance as German, English, or Polish. Such a language tag is a property of each word, as are semantic and syntactic features. However, it seems to be in closer connection to phonetic, prosodic, morphological, and visual features than to semantic and syntactic ones. We can usually identify a letter or sound string as German, English, or Polish, even if we do not know the meaning or syntax. - Thus, the language tag seems to have a modular function, with the phonetic, prosodic, visual and morphological features forming a neural network.
This, by the way, would be substantiation of Willem Levelt’s theory of a division between the „lemma“ and the „lexeme“ of a word (Levelt, 1989). Levelt assumes that a word consists of a „lemma“, being the semantic and syntactic properties, and a „lexeme“ consisting of the phonological and morphological properties of a word, and that these two parts are processed separately in language production. According to Levelt, a mental concept we want to express in language production is first compared to the semantic part of the word. If this is successful, the phonetic and morphological part will become activated.
Of course, identification and attribution to a certain language is only possible, if a person has at least a vague knowledge of that language.- Understanding is basically a process of pattern recognition between a new input and an already existing pattern stored in the brain.
5. Control of multiple languages in the brain
Closely connected to the identification of a language is the problem of control, because after a new verbal input has been identified, it has to be encoded into its proper language system. This will be no problem, if the new input is of the native language, which is unconsciously activated all the time. - Input of a foreign language, however, will have to be consciously processed to be directed into its proper system. (Control is a conscious process.) The more proficient a speaker is of a foreign language, the less conscious such a process can be.
It becomes difficult, however, with mixed bilingual input, when two different language systems have to be activated at the same time. Such mixed language in- or output causes interferences in most people and takes more time. Simultaneous interpreters need much training to perform their job properly.
6. Experiments
To test the behavior of bilingual subject when confronted with mixed language input, the author had undertaken in former projects a cross-bilingual association test (Brosig, 1999) and a cross-bilingual Stroop experiment (Brosig, 2004), in which subjects had to switch so rapidly between languages (German and English) that it could be assumed that both language systems had been activated at the same time. The results confirmed the supposition that reaction times as well as thr number of mistakes increased significantly.
6.1. Mixed language memory test
To further explore the problem of control over different languages, the author developed a
mixed language memory test with the aim, to find out, whether mixed words of two languages that were presented alternately would be stored into their proper system, even when the subjects’ attention was not specifically directed to this task. The assumption was that this might show true unconscious bilingual processing. For this purpose the author developed the following experiment:
Three wordlists, one in German, one in English, and one mixed, with 12 nouns each, were orally presented to 56 subjects, most of whom had German as first language, with the instruction to remember as many items as possible. They were explicitly not told to remember the language of the items. Th idea was that they should not exercise too much control in storing the words, but do it more or less automatically.
Subsequently, subjects were shown 27 picture cards with items from the three lists, but in a different order, and had to recall the items in the proper language. To avoid primacy and recency effects, the first and last two items of each list had been omitted. But, as a control, one of the pictures showed an item that had not been in the lists. - In the test most subjects startled at this item and 40 % stated that it had not been in the lists.
6.2. Results
As expected, the number of mistakes was smallest from the German list (7,1 %), somewhat higher from the English list (9,6 %), and highest (10,5 %) from the mixed list, which means that almost 90 % were remembered correctly by the subjects. But even when they could not
name an item, yet most of them recalled in what language that item had been, or answered with a semantically related word in the same language, e.g. „watch“, instead of „clock“.
7. Conclusions
These findings substantiate the assumption of a separate language tag as a property of each word that may be recalled, even when the other features are forgotten. - This modular function of a language tag made the author develop a model of the process of identification, in which the language tag is operating as a control switch between a verbal input and ist final concept (the meaning). (See Figure 2).
Figure 2: Model of processing language identification
When we hear a verbal sound sequence, the neural connections of its phonological, prosodic, and morphological features become activated, and the language tag decides - by screening these features - to which language this input belongs. - This will lead to the proper lexical representation and finally to the meaning of the concept. In the case of a written input, the language tag will screen the orthographic and morphological features, and, if possible semantic ones. - Of course, it will be only possible to arrive to the final meaning, if this particular sound or letter sequence is already stored in the brain. Otherwise, the language can be identified, but not the meaning.
By these phonological, prosodic, and morphological features we also identify the foreign accent of a speaker of our own language.
References:
2.7. Neue Entwicklungen in der Psycholinguistik / New Developments in Psycholinguistics
Sektionsgruppen | Section Groups | Groupes de sections
|
 |