When Two Languages Compete: Evidence for
Cross-language Activation in Bilingual Production
Judith F. Kroll
Department of Psychology
Pennsylvania State University
University Park, PA 16802 USA
Acknowledgments
Collaborators:
•
•
•
•
•
•
•
•
•
•
•
Teresa Bajo
Susan Bobb
Kate Cheng
Ingrid Christoffels
Dorothee Chwilla
Albert Costa
Annette De Groot
Franziska Dietz
Ton Dijkstra
Giuli Dussias
Chip Gerfen
•
•
•
•
•
•
•
•
•
•
•
•
Tamar Gollan
David Green
Noriko Hoshino
April Jacobs
Niels Janssen
Debra Jared
Jared Linck
Pedro Macizo
Erica Michael
Natasha Miller
Maya Misra
Scott Payne
•
•
•
•
•
•
•
•
•
•
•
Pilar Piñar
Carmen Ruiz
Nuria Sagarra
Mikel Santesteban
Herbert Schriefers
Ana Schwartz
Bianca Sumutka
Gretchen Sunderman
Natasha Tokowicz
Janet Van Hell
Zofia Wodniecka
Research Support:
• National Science Foundation Grants, BCS-0111734 and BCS-0418071
• NIH Grant RO1MH62479
• RGSO Grant, Penn State University
• Language Science Research Group at Penn State University
Cognitive research on bilingualism has increased
dramatically in the past 10-15 years
 Experimental psycholinguistics contributes one approach
to understanding the nature of bilingual experience
 Bilinguals provide a model for cognitive scientists
interested in developing a universal account of how cognitive
systems develop and interact with one another
 In this talk I focus on behavioral research methods that
complement linguistic analyses and neuroimaging to
illustrate this approach
Amsterdam, Centraal Station
Dutch-English speaker
“bike”
“fiets”
Talk Outline
 Spoken production as a case to illustrate the logic of cognitive
approaches to bilingualism
 What evidence suggests that both languages are active
when even a single word is spoken?
 How deep into speech planning does that activation
extend?
 How is the activity of the two languages resolved?
 Cues to language status
 Inhibition
 Questions for ongoing research
Dutch-English speaker
“bike”
“fiets”
How does a speaker of two languages
select the words to produce?
Selective access: The intention to speak in one
language determines which candidates become
active
 The two languages are functionally separate
Non-selective access: Candidates in both languages
become active in parallel and may compete for selection
 Distinct cues to language membership may
eventually bias access for candidates in the
intended language or allow those in the
unintended language to be inhibited
Selective access: Functional separation
Dutch-English speaker
“fiets”
Non-selective access: Parallel activation and later selection
Dutch-English speaker
“bike”
“fiets”
Past bilingual production experiments converge on the conclusion of
nonselective access but come to different conclusions regarding the locus of
selection. Some argue that selection occurs at the level of the lemma or
abstract lexical representation whereas other studies suggest that crosslanguage competition extends to the phonology
Nonselective to the lemma level
Nonselective to the phonology
If highly proficient bilinguals cannot select the language they
intend to speak in advance, then the problem becomes even
more cognitively challenging for second language learners and
unbalanced bilinguals for whom the first language is far more
dominant and active than the second.
 Spoken production as a case to illustrate the logic of cognitive
approaches to bilingualism
 What evidence suggests that both languages are active
when even a single word is spoken?
 How deep into speech planning does that activation
extend?
 How is the activity of the two languages resolved?
 Cues to language status
 Inhibition
 Questions for ongoing research
Psycholinguistic Tools: Three Laboratory Production Tasks
“fiets”
picture naming
word translation
bike
“fiets”
word naming
fiets
“fiets”
 Exploit a classic interference task:
Stroop (1935): Name the color of the ink:
GREEN: “green”
GREEN: “red”
XXXXX: “blue”
Variations of the Stroop Task
Interference in picture naming
coche
How does the language of the distractor word affect the presence of
semantic interference? If production is language selective, then only
distractors in the language to be spoken should produce interference.
Semantic interference does not depend
on the language of the distractor!
50
40
*
*
30
Magnitude
of Se mantic 20
Interference
10
0
-10
L1
L2
Language of Distractor
Data from Costa & Caramazza (2000) for Spanish-English bilinguals
naming pictures in Spanish (L1) with distractors in Spanish (L1) and in
English (L2)
Another approach: Effects of language mixture: If language
production is fundamentally selective, then requiring both
languages to be active should disrupt picture naming performance
Cued Picture Naming: Language of naming depends on an auditory cue
(Kroll, Dijkstra, Janssen, & Schriefers, in preparation)
Logic of Cued Picture Naming:
Mixed conditions: Name the picture in English if
you hear the high tone and in Dutch if you hear the
low tone
Force activation of both languages
Blocked conditions: Name the picture in English
(or Dutch) if you hear the high tone and say “no” if
you hear the low tone
Activation of the nontarget language is optional
Cost of Language Mixing in Cued Picture Naming:
Dutch-English Bilinguals (Kroll et al., in preparation)
Overall Mixed vs. Blocked Naming Latencies
Mean Naming Latency (ms)
900
Mixed
Blocked
850
800
750
700
650
600
L1
L2
Language
These results suggest that L1 is normally active during lexicalization into
the L2. Requiring L1 to be active does not affect L2 picture naming
performance.
The cost to L1 resembles the effects of language switching on L1.
Miller (2001): What happens to picture naming when the language
of a probe word naming task is in the L1 or L2?
1
.
?
“frog”
2
.
frog
“frog”
3.
kikker
“kikker”
Mean Picture Naming Latency (ms)
Same basic result: Little effect on L2 when L1 is required to
be used; performance appears nonselective in both cases. For
L1, there is a significant cost when L2 is required to be active
800
L1 Probe
L2 Probe
750
700
650
600
550
500
L1
Language of Picture Naming
L2
 Spoken production as a case to illustrate the logic of cognitive
approaches to bilingualism
 What evidence suggests that both languages are active
when even a single word is spoken?
 How deep into speech planning does that activation
extend?
 How is the activity of the two languages resolved?
 Cues to language status
 Inhibition
 Questions for ongoing research
How far into speech planning does parallel activation extend?
cognate status
of the picture’s
name
 An indication of the level at which the nontarget alternative is active
 If the phonology of the target alternative is available, then we might predict
facilitation in picture naming due to the overlap across languages
Simple picture naming in L2 and by monolinguals:
Costa, Caramazza, & Sebastián-Gallés (2000)
cognate facilitation
710
700
690
680
Mean
Naming
670
Latency (ms) 660
650
640
630
Cognate s
Noncognates
Bilinguals
Bilinguals:
Monolinguals:
Monolinguals
Catalan-Spanish speakers naming in Spanish (L2)
Native Spanish speakers naming in Spanish (L1)
Cognate facilitation suggests that the other language is
active to the level of the phonology
Effect of Cognate Status in Cued Picture Naming:
Dutch-English Bilinguals (Kroll et al., in preparation)
Naming Pictures in L2 with Cognate Names in L1:
Blocked vs. Mixed Conditions
100
Blocked L2 Cog
interference
Mixed L2 Cog
60
Magnitude of cognate
facilitation in naming pictures
in the L2 under mixed and
blocked language conditions.
40
20
0
-20
facilitation
Magnitude of Cognate Effect (ms)
80
-40
-60
-80
-100
0
500
1000
SOA
For L2 there is very little consequence of whether language is mixed
 Spoken production as a case to illustrate the logic of cognitive
approaches to bilingualism
 What evidence suggests that both languages are active
when even a single word is spoken?
 How deep into speech planning does that activation
extend?
 How is the activity of the two languages resolved?
 Cues to language status
 Inhibition
 Questions for ongoing research
The results for naming cognate pictures suggest that the activity
of the nontarget language reaches the level of the phonology.
But can bilinguals exploit language cues in the nature
of the event that initiates production to minimize
cross-language influences?
More likely to be a
a fiets than a bike?
Even more likely to be
a fiets?
Definitely Dutch!
Bringing the problem into the lab…
Compare translation to picture naming: In translation there is a cue
to the language present in the nature of the input
picture naming
word translation
“fiets”
bike
“fiets”
don’t speak English!
Miller & Kroll (2002): Translation Stroop: What happens
when the distractor is in the language of the input?
English
Spanish
If there is a cue in the event that initiates speech planning, then
a distractor in the input language should not influence production.
Stroop interference in translation only when the distractor
appeared in the language of the word to be spoken, unlike
the Stroop interference observed for picture naming
1450
1400
Mean
Translation 1350
Latency (ms)
1300
Related
Unrelated
1250
Output
Input
Language of the Distractor
A cue in the language input allowed production to proceed selectively!
Does the language of a sentence context function as a cue?
Schwartz (2003): Take words that have been shown to elicit activation of both
languages and put them in full sentence context.
Cognates with identical/similar orthography but similar or different phonology:
English
piano
base
Spanish
piano
base
Cross-language
phonology
Similar [+p]
Different [-p]
620
+P
-P
600
580
Mean
Naming
560
Latency (ms)
540
520
500
Type of Cognate
Out of context: Facilitation for naming cognates in L2
when the phonology converges from L1 to L2
RSVP: Method for naming words in sentence context
(250ms/word)
+
home
The
boy
Who ran
home?
ran
Follow along with sentence.
Say red word out loud.
Answer questions when asked.
RSVP: Rapid Serial Visual Presentation
home
for
dinner.
Schwartz (2003): Does the facilitation for naming cognates
disappear in sentence context? If the sentence provides a cue to
language membership, then no cognate effects should be observed
+P
-P
720
700
Type of Sentence
High const raint
Low constraint
Example
The composer sat at the bench
and began to play t he piano as
the light s dimmed.
High
680
Mean
Naming
Latency (ms) 660
640
620
Type of Cognate
As we wa lked through the room
we not iced there was a large
piano by the window.
720
+P
-P
700
Result: Sentence constraint but
not language per se eliminates
the cognate effect
Low
680
Mean
Naming
Latency (ms) 660
640
620
Type of Cognate
Naming in the L2
What cues effectively reduce cross-language competition?
 Language-specific information in the event that initiates speech planning
(Miller & Kroll, 2002)
 Convergence between language-specific information and meaning
(Schwartz, 2003, Van Hell, 1998)
What cues do not reduce cross-language competition?
 The language of a sentence context itself (Schwartz, 2003; Van Hell, 1998)
 The intention to use one language only (Van Hell & Dijkstra, 2002)
 Instructions (Dijkstra et al., 2000)
 Cross-language script differences (Hoshino & Kroll, 2005)
 Spoken production as a case to illustrate the logic of cognitive
approaches to bilingualism
 What evidence suggests that both languages are active
when even a single word is spoken?
 How deep into speech planning does that activation
extend?
 How is the activity of the two languages resolved?
 Cues to language status
 Inhibition
 Questions for ongoing research
Inhibition: Do bilinguals inhibit one language to speak the other?
Some recent research suggests not:
Costa & Santesteban (2004): Highly balanced Spanish-Catalan bilinguals
in Barcelona do not appear to require inhibition when switching between
languages
Three laboratory approaches to examine how inhibition may
operate when L2 speakers are immersed in the L2
 Simulated immersion in the lab
 Actual immersion in an L2 environment
 Forced L2 immersion within the L1 environment
Simulated immersion in the laboratory
Kroll, Michael, & Sankaranarayanan (1998):
Attempted to simulate one aspect of an immersion environment in a
training study in the laboratory by providing unique cues to the L2.
What’s helpful about immersion?
 Suppression of L1
 Unique cues to L2
Teach American college students who know no Dutch or German 40
new Dutch words by associating them to their English translations
or to pictures of the objects to which they refer.
Training conditions for monolingual English speakers:
The twist: The pictures were sometimes shown in a
noncanonical orientation
The noncanonical
orientation may:
1. Provide a unique
cue to the L2 word
and/or
2. Inhibit the process
of retrieving the
L1 name
?
At test: Translate words from English to Dutch or name
pictures in Dutch. There is typically a cost to naming
pictures of noncanonical objects. The question here is
whether that cost will be observed for learners who
acquired the new L2 concepts by associating them to the
noncanonical objects.
What is the cost of noncanonicality at test?
[noncanonical-normal orientation]
400
300
Cost of
Noncanonical
Pictures (ms)
Word Training,
Picture Test
Picture Training,
Picture Test
Picture Training,
Word Test
200
100
0
-100
-200
Canonicality Effect
This is the remarkable result: Faster to later translate from English
to Dutch if trained on the noncanonical pictures!
Real L2 immersion during study abroad: Linck & Kroll (2005)
Native English speakers at intermediate levels of Spanish
spending a semester abroad in Salamanca compared to a
control group of classroom learners with comparable L2 study
Performed a translation recognition task (De Groot, 1992;
Sunderman, 2002) in which they had to decide whether a
word in English was the correct translation of a word in Spanish.
hombre
man
?
“yes”
Using materials from Sunderman (2002), the critical
distractors in the experiment consisted of word pairs
that were not the correct translation:
mano-man:
Critical items:
lexical form related
hambre-man: translation related
mujer-man: semantically related
Control items:
casa-man:
unrelated
[Interference = Critical RT - Control RT]
Results: Control learners show interference for all distractor
types but immersed learners show only semantic interference
120
100
80
Magnitude of 60
Interference
40
(ms)
Controls
Immersed
20
0
-20
Lexical
Form
Translation
Semantic
Distractor Type
In the immersion environment the L1 appears to be suppressed!
Forced immersion in an L1 environment: Jacobs, Gerfen, & Kroll (2005)
Native English speakers at intermediate levels of Spanish in a summer
domestic immersion program or in classroom study only.
The two groups appear similar in overall proficiency in Spanish as the L2:
Self-assessed proficiency in L1 and L2 (10 pt. Scale):
Classroom: L1 = 9.6
Immersed: L1 = 9.3
L2 = 6.3
L2 = 6.1
Percent correct rejections in Spanish lexical decision (is the string of
letters a real word in Spanish?)
Classroom:
Immersed:
65.9%
68.7%
Two groups performed a Spanish word naming experiment with cognates and
lexically/phonetically matched controls
Naming Latencies
VOT
38
620
660
640
620
600
Cognate s
Controls
Mean Duration (ms)
680
Mean VOT (ms)
Mean Naming Latency (ms)
700
Articulatory Duration
34
30
26
Cognate s
Controls
580
22
Learners
Learners Immersed
600
580
560
540
520
Learners
Learners Immersed
Cognate s
Controls
Learners
Learners Immersed
Immersed learners are faster to name Spanish words and like proficient
English speakers of Spanish, speak Spanish words as more Spanish-like
than the classroom learners and show effects of cross-language cognate
status in planning but not executing L2 speech.
Again, in this immersion environment it appears that L1 is suppressed!
Conclusions
The results of these studies support recent claims that lexical access
is language nonselective in spoken production.
There is significant activation of the L1 when speaking L2 for even
highly proficient bilinguals. That activation may produce crosslanguage competition.
The ability to negotiate that competition may come in part from cues
that reliably signal L2 and in part from the ability to inhibit irrelevant
information.
This work holds promise for developing a principled account of the
factors that constrain cross-language activation to allow fluent
performance in a single language but also code switching between
languages.
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
 Questions for ongoing research
 How do language-specific factors influence the degree of
cross-language activity?
 What are the implications of bilingual research for cross
linguistic analyses?
 How is the L1 affected by proficiency in the L2? At the level
of the lexicon, the phonology, and the syntax?
 How do the grammatical constraints that characterize codeswitching modulate cross-language activity?
 What are the consequences of age and context of
acquisition and language maintenance?
 What are the cognitive consequences of cross-language
competition?
 What is the neural basis of bilingual performance?
Back to Amsterdam…
Things could be worse in New York!
Thank you!
Descargar

Document