Reading between two languages

1
Reading between two languages

• Selective versus non-selective lexical access in
  bilingual word-recognition

2
Monolingual Model
Semantic Representation
Lexicon
3
Monolingual Model
Semantic Representation
Lexicon
ROSE
4
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
5
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
BOOT
6
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
BOOT
7
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
BOOT
8
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
BOOT
9
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
BOOT
10
Selective Access to the Lexicon
Semantic Representation
Dutch Lexicon
English Lexicon
ROSE
HOND
FLOWER
TAFEL
LAARS
BOAT
UMBRELLA
ROOS
BOOT
PARAPLU
TABLE
BLOEM
DOG
FLES
BOTTLE
BOOT
BOOT
11
Nonselective Access
12
Lexical access is nonselective
13
Selective or nonselective

• Homographs
• Sublexical characteristics
• Phonological information
• Semantic information
• Context information
• (Memory)

14
Word recognition

• Homographs (false friends)
• KIND
• BAD
• ROOM
• RAMP
• STEP

• Cognates
• HOTEL
• FILM
• WIND
• PEN
• RIB

15
Cognates

• faster than matched control words
• rt (HOTEL) lt rt (EVENT)
• in all tasks
• Lexical Decision, Naming, Identification
• Both languages always active?

De Groot et al., (in press) Dijkstra, Van
Jaarsveld, Ten Brinke (1998)
16
Homographs (false friends)

• English Lexical Decision by Spanish-English
  bilinguals
• RED (net) RON (rum)
• Only one language active?

Gerard Scarborough, 1989
17
English lexical decision with D-E bilinguals

• WORDS
• SKIN
• STEP
• FATE
• RANK
• VAGUE
• SMART
• STAG
• RAMP

• NONWORDS
• MORP
• TWEEL
• PLAM
• ZEAR
• KNAW
• LOMB
• FLOOM
• PAKE

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
18
English lexical decision with D-E bilinguals

• WORDS
• SKIN
• STEP
• FATE
• RANK
• VAGUE
• SMART
• STAG
• RAMP

• NONWORDS
• MORP
• TWEEL
• PLAM
• ZEAR
• KNAW
• LOMB
• FLOOM
• PAKE

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
19
English lexical decision with D-E bilinguals

• Results
• STEP, RANK, SMART, RAMP
• as fast as
• SKIN, FATE, VAGUE, STAG

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
20
General lexical decision with D-E bilinguals

• WORDS
• SKIN
• FEIT
• STEP
• FATE
• KRAP
• RANK
• STAG
• HEMD
• RAMP

• NONWORDS
• MORP
• TWEEL
• PLAM
• ZEAR
• KNAW
• LOMB
• NARK
• FLOOM
• PAKE

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
21
General lexical decision with D-E bilinguals

• WORDS
• SKIN
• FEIT
• STEP
• FATE
• KRAP
• RANK
• STAG
• HEMD
• RAMP

• NONWORDS
• MORP
• TWEEL
• PLAM
• ZEAR
• KNAW
• LOMB
• NARK
• FLOOM
• PAKE

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
22
General lexical decision with D-E bilinguals

• Results with Dutch words added
• STEP, RANK, RAMP
• faster than
• SKIN, FATE, STAG

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
23
English lexical decision with D-E bilinguals

• WORDS
• SKIN
• STEP
• FATE
• RANK
• VAGUE
• SMART
• STAG
• RAMP

• NONWORDS
• DORP
• TWEEL
• KANT
• ZEAR
• SLUW
• LOMB
• STOOM
• PAKE

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
24
English lexical decision with D-E bilinguals

• WORDS
• SKIN
• STEP
• FATE
• RANK
• VAGUE
• SMART
• STAG
• RAMP

• NONWORDS
• DORP
• TWEEL
• KANT
• ZEAR
• SLUW
• LOMB
• STOOM
• PAKE

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
25
English lexical decision with D-E bilinguals

• Results with Dutch words as nonwords
• STEP, RANK, SMART, RAMP
• more slowly than
• SKIN, FATE, VAGUE, STAG

Dijkstra, Van Jaarsveld, Ten Brinke (1998)
26
Homographs (false friends)

• English Lexical Decision
• Distractors nonwords (e.g. NAPE) only
• STEP SKIN
• Generalized Lexical Decision
• Distractors nonwords (NAPE) only
• STEP lt SKIN
• English Lexical Decision
• Distractors nonwords Dutch words
• STEP gt SKIN

27
Homographs (false friends)

• Homograph effects depend on task demands.
• There are situations where both languages are
  active.
• There are situations where only one language
  determines the performance.

28
Models of Language Selection

• Language modes
• (Bilingual Interactive Activation Model)
• Language activation forms a continuum
• Depending on task languages become activated or
  inhibited.

Dutch only
English only
Both languages active
Grosjean (1997), Dijksta Van Heuven (1998)
29
Bilingual Interactive Activation Model
Dutch
English
Dutch words
English words
letters
Dijksta Van Heuven (1998)
30
Explaning homograph results BIA

• English Lexical Decision
• Distractors nonwords (e.g. NAPE) only
• STEP SKIN
• Only English words are activated.
• English language node is activated.
• Dutch word representations are inhibited.
• Dutch words do not compete for selection.

31
Explaning homograph results BIA

• General Lexical Decision (Dutch words included)
• Distractors nonwords only (e.g. NAPE)
• STEP lt SKIN
• English and Dutch words are activated.
• English and Dutch language nodes active.
• E. D. representations become active.
• In homographs, fastest representation wins.

32
Explaning homograph results BIA

• English Lexical Decision
• Distractors nonwords Dutch words
• STEP gt SKIN
• Dutch words are activated by nonwords.
• Dutch language node remains active.
• Dutch word representations become active.
• In homographs Dutch rep. inhibits Eng. rep.

33
Explaning homograph results BIA

• General Lexical Decision (Dutch words included)
• Distractors nonwords only (e.g. NAPE)
• STEP lt SKIN
• In homographs, fastest representation wins.

?

• English Lexical Decision
• Distractors nonwords Dutch words
• STEP gt SKIN
• In homographs, Dutch rep. inhibits Eng. rep.

34
Models of Language Selection

• Task Schema Model
• (Bilingual Activation Verification Model)
• Both languages always active.
• Task determines what is done with activation.

Green (1998), Dijksta Van Heuven (submitted)
35
Task schema model
Task specifications Decision criteria
L1/L2
Dutch words
English words
letters
Green (1998), Dijksta Van Heuven (submitted)
36
Explaning homograph results Schema

• English Lexical Decision
• Distractors nonwords (e.g. NAPE) only
• STEP SKIN
• Task Schema
• English Word -gt Yes
• Dutch Word -gt Irrelevant
• Dutch representations of homographs do not play a
  role.

37
Explaning homograph results Schema

• General Lexical Decision (Dutch words included)
• Distractors nonwords only (e.g. NAPE)
• STEP lt SKIN
• Task Schema
• English Word -gt Yes
• Dutch Word -gt Yes
• For homographs, fastest representation (Dutch or
  English) wins.

38
Explaning homograph results Schema

• English Lexical Decision
• Distractors nonwords Dutch words
• STEP gt SKIN
• Task Schema
• English Word -gt Yes
• Dutch Word -gt No
• Response conflict for homographs

39
Models of Language Selection

• Language mode (BIA)
• Switching words from irrelevant language off.
• Task schema
• Words from irrelevant language take part in
  selection but are filtered out immediately
  afterwards.

40
Models of Language Selection

• Task schema is necessary to explain homograph
  results
• Still, we do not know what happens underneath
  the task schema
• Combination of language mode and task schema
  model is also possible.

41
Models of Language Selection

• Language Mode says Representations from
  irrelevant language are switched off.
• Homograph results show that L1 is not switched
  off when L1 words are included as nonwords.
• Proves that nonwords are not irrelevant.
• Still possible that really irrelevant language is
  switched off.

42
Models of Language Selection

• Homograph results are not conclusive whether a
  language can ever be switched off.

43
And now?

• Not seeing a homophone effect does not proof
  much.
• Words could become active and still not play a
  role in a task.
• How can we look at processes that take place
  while words are activated (not afterwards)?

44
Sublexical Characteristics

• Which role play similarities between words and
  nonwords (from the same or from different
  languages) in bilingual lexical access?

45
Lexical access (in general)
46
Lexical access (in general)
HOUSE
CAP
RAT
CAT
C
A
T
47
Lexical access (in general)
Nonwords
48
Lexical access (in general)
Nonwords
49
Lexical access (in general)

• Nonwords that share many letters with many words
  create a lot of activation in the lexicon.
• Nonwords that create a lot of activation are
  rejected more slowly.
• Nonwords with unusual letter combinations do not
  activate many words and are rejected quickly.

50
Lexical access (in general)
Neighbors

• Letter-strings that differ by only one letter
• BOAT
• GOAT BOOT MOAT BOAR BOUT BRAT
• SOAT
• SOAK SOAP BOAT GOAT MOAT SOFT
• Words and nonwords that have many neighbors
  activate many other words in the lexicon

51
L1 words as nonwords

• English Lexical Decision by Spanish-English
  bilinguals
• reject CASA as fast as reject LASA
• Only one language active?

Scarborough, Gerard, Cortese 1989
52
L2 words as nonwords

• Spanish Lexical Decision by Spanish-English
  bilinguals
• reject HOUSE as fast as reject NOUSE
• Only one language active?

Scarborough, Gerard, Cortese 1989
53
L1 words as nonwords

• English Lexical Decision by Dutch-English
  bilinguals
• reject NOOT more slowly than reject DOOT
• Both languages active?

Nas, 1983
54
L1 words as nonwords
English Lexical Decision by Dutch-English
bilinguals reject NOOT more slowly than reject
DOOT
Nas, 1983
?

• English Lexical Decision by Spanish-English
  bilinguals
• reject CASA as fast as reject LASA

Scarborough, Gerard, Cortese 1989
55
Characteristics of nonwords

• English Lexical Decision by monolinguals
• reject PFLOK
• faster than
• reject TWOUL

illegal bigram PF
contains no illegal bigram
by English German bilinguals reject PFLOK as
slow as reject TWOUL
bigram (PF) legal in German
contains no illegal bigram
Altenberg Cairns, 1983
56
Characteristics of nonwords

• Participants in Lexical Decision Tasks are
  sensitive to sublexical characteristics
• Illegal letter combinations are picked out
  immediately and used to reject nonwords
• English words contain letter combinations that
  are illegal in Spanish.
• Spanish words contain letter combinations that
  are illegal in English.
• Paricipants in Scarborough et al. could use
  letter combinations to reject words from other
  languages

57
Characteristics of nonwords

• English monolingual coulds use PF to reject
  nonword PFLOK immediately.
• English-German bilinguals could not.
• Purely English Task, no reason to consider German
  letter combinations.
• Bilinguals cannot choose to ignore letter
  combinations from irrelevant language.

58
Characteristics of words

• English Lexical Decision with D-E bilinguals
• Patriots e.g. HEAT
• many English neighbors (MEAT, BEAT, FEAT, SEAT,
  HEAD, HEAL, HEAR)
• few Dutch neighbors (HERT, HEET)
• Traitors e.g. KEEN
• many Dutch neighbors (GEEN, BEEN, PEEN, KEER,
  KEEL)
• few English Neighbor (KEEP)

Grainger Dijkstra, 1992
59
Characteristics of words

• Patriots accepted faster than Traitors

Grainger Dijkstra, 1992
60
Characteristics of words

• English Lexical Decision with French-English
  bilinguals
• Patriots (CREAM) faster than Traitors (TRADE)
• English Lexical Decision with English
  monolinguals
• Patriots (CREAM) as fast as Traitors (TRADE)

Beauvillain, 1992
61
Priming with words

• French lexical decision, F-E bilinguals
• Masked priming paradigm
• mask (XXXX), prime (WORD1), target (word2),
  decision
• French targets (e.g. AMONT)
• English primes
• Higher frequency neighbor of target (AMONG)
• No neighbor of target (DRIVE)

Bijeljac-Babic, Biardeau, Grainger, 1992
62
Priming with words

• DRIVE - amont
• faster than
• AMONG - amont
• Crosslingual inhibitory priming of
  higher-frequency neighbor

Bijeljac-Babic, Biardeau, Grainger, 1992
63
Item characteristics Conclusion

• Even in strictly monolingual tasks, similarities
  of items to words from the non-active language
  play a role.
• Participants can use unusual letter combinations
  to reject items.
• Can participants attribute special letter
  combinations to one or the other language?

64
Characteristics of nonwords

• French Lexical Decision with F-E bilinguals
• reject SOAT
• faster than
• reject NONT

Traitor Many English, few French neighbors
Patriot Many French, few English neighbors

• General Lexical Decision (Engl. French words)
• reject SOAT
• as slow as
• reject NONT

French Ohnesorg, 1996
65
Characteristics of nonwords

• Participants used typically English letter
  combinations (language cues) to reject words in a
  French lexical decision task.
• But not in a general lexical decision task (with
  French and English words).

French Ohnesorg, 1996
66
Language Cues

• French - English bilinguals
• General lexical decision (French English words)

Grainger Beauvillain, 1987
67
Language Cues

• BATH
• APPLE
• SISTER
• WING
• BAIN
• POMME
• SOEUR
• WING

• BATH
• APPLE
• SISTER
• WHIP
• BAIN
• POMME
• SOEUR
• WHIP

Grainger Beauvillain, 1987
68
Language Cues

• SOEUR - WING
• slower than
• SISTER - WING

Language-switch cost

• SOEUR - WHIP
• as fast as
• SISTER - WHIP

Language cue eliminates switch cost
Grainger Beauvillain, 1987
69
Item characteristics Conclusion

• Even in strictly monolingual tasks, similarities
  of items to words from the non-active language
  play a role.
• Participants can use language cues to identify
  the language of an item.
• Sometimes language cues help to overcome
  interference from the other language.

70
Phonological Information

• Letters can be mapped to sounds
• Readers can use this information
• Some letters are mapped to different sounds in
  different languages
• Consider EURO

71
Phonology Priming

• Masked priming (Perceptual Identification) with
  Dutch-French bilinguals
• Intra-lingual Homophones FAIN - FAIM
• Intra-lingual Controles FAIC - FAIM
• Cross-lingual Homophones PAAR - PART
• Cross-lingual Controles PAAL - PART

Bysbaert, Van Dyck, Van de Poel, 1999
72
Phonology Priming

• FAIN - FAIM better than FAIC - FAIM
• intra-lingual effect of phonological priming
• PAAR - PART better than PAAL - PART
• cross-lingual effect of phonological priming
• intra-lingual effect cross-lingual effect
• no intra-lingual effect for monolinguals

Bysbaert, Van Dyck, Van de Poel, 1999
73
Phonology Priming

• Participants were not aware of prime.
• Dutch was completely irrelevant in this task
• Dutch letter-to-sound mappings were as effective
  in a French reading task as French
  letter-to-sound mappings.
• All letter-to-sound mappings a reader knows
  become active in all reading situations.

Bysbaert, Van Dyck, Van de Poel, 1999
74
Phonology Consistency

• Monolingual
• Phonologically inconsistent words (e.g. CAVE)
• are more difficult to read aloud than
• Phonologically consistent words (e.g. CAKE)

Jared, McRae, Seidenberg, 1991
75
Phonology Consistency

• English-French bilinguals
• intra-lingual enemies CAVE HAVE
• cross-lingual enemies FAIT BAIT

Jared Kroll, 2001
76
Phonology Consistency

• English-French bilinguals
• words with intra-lingual enemies were always read
  more slowly than words without enemies.
• Words with cross-lingual enemies were only slower
  than words without enemies when a block of French
  words had preceded them.

Jared Kroll, 2001
77
Phonology Consistency

• French-English bilinguals
• did not show an effect of intra-lingual enemies.

Jared Kroll, 2001
78
Phonology Consistency

• Result for French-English bilinguals are somewhat
  strange
• Why did people with French as L1 not show an
  effect of French enemies, while people with
  French as L2 did?
• Results for English-French bilinguals suggest
  that activation of French letter-to-sound rules
  depend on reading situation.

Jared Kroll, 2001
79
Phonology Letter search task

• Monolingual letter search task
• look for letter I
• TAIP, NAIP, BRANE, PRANE
• NAIP faster detection than TAIP
• PRANE faster rejection than BRANE

Ziegler Jacobs, 1995
80
Phonology Letter search task

• Bilingual letter search task
• look for letter O
• BAUME, PAUME
• BAUME (sounds like BOOM in Dutch)
• slower rejection than in
• PAUME (doesnt sound like anything)

Brysbaert, in prep.
81
Phonology Letter search task

• Bilingual opposite to monolingual results.
• Instead of misleading participants to except the
  O, the homophone BOOM convinces them even faster
  that there is no O in BAUME.
• The task-schema at work?
• Telling the participant that Dutch words should
  not be considered?
• Inhibiting BOOM?

Brysbaert, in prep.
82
Conclusion Phonology

• Phonology becomes active in word reading
• Letter to sound mappings from one language can
  become active while reading another language.
• Unclear, whether readers adjust their dominant
  set of letter-to-sound mappings to the language
  they are currently reading.

83
Is semantic access selective?
84
Is semantic access selective?
85
Semantics while reading Dutch?
Semantic Representation
Lexicon
TAFEL
DOOS
BOOK
HOND
BOOT
BOX
BOEK
BLOEM
TABLE
FLES
BOOT
FLOWER
DOG
BOTTLE
BOOT
86
Semantics while reading English?
Semantic Representation
BOOK
TAFEL
DOOS
HOND
BOOT
BOX
BOEK
TABLE
BLOEM
FLES
FLOWER
BOOT
DOG
BOTTLE
BOOT
87
Semantic access selective?

• Bilingual Stroop task

88
Semantic access selective?

• Monolingual Stroop
• GEEL
• GROEN
• BLOUW
• ROOD
• GEEL
• ZWART

• Bilingual Stroop
• YELLOW
• GREEN
• BLUE
• RED
• YELLOW
• BLACK

Tzelgov, Henik, Leiser, 1990
89
Semantic access selective?

• Bilingual Stroop task
• RED (rood) better than BLUE (rood)
• less interference than in monolingual Stroop
• strength of interference depends on proficiency

Tzelgov, Henik, Leiser, 1990
90
Semantic access selective?

• DOG
• UMBRELLA
• POLITICS
• BRAND
• FIRE
• KIDNEY

• DOG
• UMBRELLA
• POLITICS
• DWARF
• FIRE
• KIDNEY

DeMoor Brysbaert, unpublished)
91
Semantic access selective?

• No homograph effect
• Lexical decision on BRAND equal to DWARF
• Semantic priming
• Lexical decision on FIRE faster after BRAND than
  after DWARF
• Semantic priming by inactive meaning of
  homographs!

DeMoor Brysbaert, unpublished
92
Effect of language context

• Context words

DeBruin, Dijkstra, Chwilla, Schriefers, 2001
93
Are all three items words?

DeBruin, Dijkstra, Chwilla, Schriefers, 2001
94
Are all three items words?
NOSE
ANGEL
HEAVEN
Semantically related for English reading
of homograph
DeBruin, Dijkstra, Chwilla, Schriefers, 2001
95
Are all three items words?
NOSE
ANGEL
HORSE
Semantically unrelated
NOSE - ANGEL - HEAVEN better than NOSE - ANGEL
- HORSE
DeBruin, Dijkstra, Chwilla, Schriefers, 2001
96
Are all three items words?
ZAAK
ANGEL
HEAVEN
Semantically related for English reading
of homograph
Is the semantic relation between ANGEL and HEAVEN
less strong, when preceded by a Dutch word?
DeBruin, Dijkstra, Chwilla, Schriefers, 2001
97
Are all three items words?
NOSE - ANGEL - HEAVEN ZAAK - ANGEL -
HEAVEN better than NOSE - ANGEL - HORSE
ZAAK - ANGEL - HORSE
Is the semantic relation between ANGEL and HEAVEN
less strong, when preceded by a Dutch word?
NO!
DeBruin, Dijkstra, Chwilla, Schriefers, 2001
98
Effect of language context

• Word context
• Crosslingual semantic priming with homographs is
  not influenced by language of preceding word

DeBruin, Dijkstra, Chwilla, Schriefers, 2001
99
Counterintuitive?
The smart little boy sat on the roof and tried to
spot the star.
first-language interpretations dont seem to
bother us when reading in our second
language.
100
Effect of sentence context

• MODE
• FASHION
• POLITICS
• GIFT
• POISON
• KIDNEY

• MODE
• POISON
• POLITICS
• GIFT
• FASHION
• KIDNEY

GIFT - POISON faster than MODE - POISON
Elston-Guettler Williams, submitted
101
She
Elston-Guettler Williams, submitted
102
gave
Elston-Guettler Williams, submitted
103
her
Elston-Guettler Williams, submitted
104
friend
Elston-Guettler Williams, submitted
105
an
Elston-Guettler Williams, submitted
106
expensive
Elston-Guettler Williams, submitted
107
gift.
Elston-Guettler Williams, submitted
108
POISON
Semantic priming effect disappears
Elston-Guettler Williams, submitted
109
Effect of sentence context

• Single words
• GIFT - POISON faster than MODE - POISON
• Sentences
• GIFT - POISON MODE - POISON

Elston-Guettler Williams, submitted
110
Conclusion

• Majority of the evidence from single word studies
  suggest that lexical access is initially
  non-selective.
• Information about the relevant language (language
  cues, task, instruction) can be used to discard
  words from the irrelevant language quickly after
  they have been activated.

111
Conclusion

• However, even if the lexical representations from
  the inactive language do not interfere with the
  task completion, they can still activate
  semantics
• Applying results from single word studies to
  reading in normal situations is not trivial.