On being bilingual Marc Brysbaert Ghent University Bilingualism is everywhere • There are at least 3000 languages for 195 independent countries. • Bilingualism is especially common among economically non-dominant groups. • The “world language” usually follows economic dominance, because these people cannot be motivated to learn another language. So, what do we know about bilingualism? • I’ll start with a model summarising our knowledge some 15 years ago. • Then I’ll show that much of what we thought to be true turns out to be wrong ... • ... at least to me Kroll & Stewart’s (1994) revised hierarchical model Mental lexicon L1 Mental lexion L2 Concepts Kroll & Stewart’s (1994) revised hierarchical model • Four ideas 1. Distinction between the L1 lexicon and the L2 lexicon. 2. We have a mechanism of language control (selective access). 3. Asymmetries in the connections (in particular the idea of direct connections between L2 words and their L1 translations). 4. Conceptual knowledge is language-independent. Spivey & Marian (Psychological Science, 1999) • Started from the finding that when people hear a word, initially all words with the same sounds become activated (e.g., when people hear “beaker”, initially the word “beetle” is acitvated as well). • Shown with the visual world paradigm (e.g., Allopenna et al., 1998) speaker beaker beetle carriage Eye movements are monitored to know what the participant is looking at. “First, look at the centre of the screen.” “Now, pick up the beaker ...” “... And put it above the circle.” Participants more often look at the beetle upon hearing “beaker” than at the other two objects (Allopenna et al., 1998). Spivey & Marian (Psychological Science, 1999) 12 late Russian-English fluent bilinguals; speech processing “Poloji marku.” [“Pick up the stamp.”] marker marku Dijkstra, Timmermans, & Schriefers (JML, 2000) - Dutch-English bilinguals (unbalanced) - 3 types of homographs: * list (HF English; LF Dutch [trick]) * brand (LF English; HF Dutch [fire]) * hark (LF English; LF Dutch [rake]) - Dutch go/no-go task homographs Dutch control words HFE-LFD (list) RT % no-go 868 26% 630 4% LFE-HFD (brand) RT % no-go 589 0% 558 0% LFE-LFD (hark) RT % no-go 763 22% 706 6% “On being blinded by your other language” Van Hell & Dijkstra (2002) • Cognates: Words in two languages with the same meaning and a similar form • E.g. tomaat – tomate – tomato • Lexical decision in Dutch • Cognates vs. control words – Dutch – English cognates (appel – apple) – Dutch – French cognates (citroen – citron) Van Hell & Dijkstra Ghent Lexicon Psychology students Students French Psychology students Dutch-English cognate 499 489 559 Dutch-French cognate 519 520 585 Control word 529 541 595 English cognate effect 30* 52** 36* French cognate effect 10 21* 10 Van Assche et al. (2009) • Can the cognate effect also be found in sentence reading? • Sentences: – “Zijn jongste zoontje wou zich als een piloot verkleden voor carnaval.” [His youngest son wanted to dress up as a pilote for the carnaval party.] – “Zijn jongste zoontje wou zich als een konijn verkleden voor carnaval.” [His youngest son wanted to dress up as a rabbit for the carnaval party.] Why are we not able to control the languages that are activated? • A lot of information becomes activated automatically (balistically) as soon as we see a word. • For instance, in addition to orthographic information, phonological information is activated as well. Phonology in silent reading • Silent reading without moving the lips is a very recent skill (early 20th century?) – In 383 Aurelius Augustine expressed his surprise when he met the bishop of Milan and saw that he could read silently. “When he read”, said Augustine, “his eyes scanned the page and his heart sought out the meaning, but his voice was silent and his tongue was still.” (as cited in Manguel, 1996, p. 42). • Saenger (1997): Silent reading became possible only after spaces were inserted between the words (started from the 8th century; before, there was scriptio continua) Phonology in silent reading • Beginning readers need to sound out the words; speed of reading acquisition depends on the transparency of the letter-sound correspondences. • Sound still plays a major role in current-day expert silent reading: – Inner voice (difficult texts) – Tongue twister effect (“the purpose of the play was to please the brave prince”) – TTE exists in all languages (also with logographic writings) Phonology in silent reading • Long time discussion whether phonological coding occurred after the word recognition (addressed phonology) or as part of the recognition process itself (assembled phonology). • Now quite good evidence that phonology is assembled prelexically and together with the orthographic code activates lexical representations (i.e., weak phonological model; Ferrand, 1995; Grainger, Kiyonaga, & Holcomb, 2006; Perea & Carreiras, 2008; Rastle & Brysbaert, 2006) Phonology in silent reading: Masked priming • Strongest evidence for the contribution of assembled phonology comes from masked priming with pseudohomophones • E.g., Rastle & Brysbaert (2006) – Target: USE – Primes: yuice vs. durke (SOA = 58 ms) – Task: Lexical decision yuice USE ########### Faster lexical decision time to the target USE after the prime yuice than after the prime durke. Because the primes are nonwords, recoding must be prelexical. Phonology in silent reading: Masked priming • Evidence that phonological priming is automatic (Brysbaert, 2001; Drieghe & Brysbaert, 2002; Xu & Perfetti, 1999). • If so, you ought to find evidence for crosslanguage priming. Brysbaert et al. (JEP:HPP, 1999, 2002) • French target words preceded by “Dutch” primes • 2 types of primes * pseudohom : * graph. contr.: soer - SOURD siard - SOURD •Perceptual identification task • Task as seen by the participants : “On each trial a French word in capitals is flashed in the middle of the screen between a forward and a backward mask consisting of #######. Your task is to guess the word.” ########### BRAIN brane Nothing was said about the primes (they were not noticed). Participants were not informed about the fact that some primes were homophonic if they were read aloud according to the Dutch spelling-sound correspondences. Dutch-French stimuli FrenchDutch French mono DutchFrench homophonic (soer-SOURD) 24 41 30 graphemic (siard-SOURD) 33 34 23 phon. priming effect -9 +7 +7 Summary selective access and separated lexicons • Strong evidence against selective access (also in speech production). • Also increasing evidence against separate lexicons. Semantic access from L2 • Is it really the case that we have to understand the meaning of L2 words via L1 translation? • How long is this the case? (developmental aspect) • Duyck & Brysbaert (2004) : what about the translation of number names? Why numbers? • There is a very straightforward prediction you can make : Whenever numbers activate their meaning you find a number magnitude effect 1. No magnitude effect in simple number naming 2. Magnitude effects in number memory and number comparison Brysbaert (1995) Semantic medition in number translation • Duyck & Brysbaert (2004, experiment 1) – Dutch-French bilinguals who: 1. Name L1 number words in L1 2. Name Arabic numbers in L1 3. Translate L2 number words in L1 (backward) Digits Words The bilinguals were too proficient • Duyck & Brysbaert (2004, experiment 3) – Participants learned new, artificial (Estonian) number words 1. Name L1 number words in L1 2. Name Arabic numbers in L1 3. Translate L2 number words in L1 (backward) Summary semantic activation from L2 words • Connections between L2 words and meaning are established very rapidly when there is a clear mapping from the words to the concepts. • Now also being shown for other types of stimuli. • Also growing evidence that the type of teaching plays a role (word-list focused vs. meaning focused). Are concepts language independent? • We know that language independence is not true for episodic memory. • We know that language independence is not true for short term retention (Sahlin et al., 2005) – 120 prolific English-Spanish bilinguals – List of 120 words heard at a speed of 1 w per 3 s; half English, half Spanish – Recognition test: old vs. new – 85% in English-English, 88% in Spanish-Spanish – 13% in Spanish-English, and 17% in English-Spanish Are concepts language independent? • But could language independence be true for semantic knowledge: Conceptmaison = concepthouse = concepthuis • We don’t know yet, but there are indications that semantic memory may be much more language dependent that we currently assume – E.g., translation subtleties: a large book <> a large sister Are concepts language independent? • Blot et al. (2003): Brain storming in L1 and L2 • “What uses could people make of an extra thumb (next to the little finger)?” • English-Spanish and Spanish-English bilinguals • Ideas generated in L1 = 25 • Ideas generated in L2 = 20 • After the session, change of language • Extra ideas in L1, but not in L2 Conclusion • Go and read – Brysbaert, M., & Duyck, W. (in press). Is it time to leave behind the revised hierarchical model of bilingual language processing after 15 years of service? Bilingualism: Language and Cognition. – Available at biblio.ugent.be • Not all doom and gloom – Bilingualism increases your ability to switch tasks. – Bilingualism may postpone Alzheimer by 5-10 yrs.