l bildbenennung l wortgenerierung(z.b. nennen sie möglichst viele tiere!) l wortlesen (hund) l...
TRANSCRIPT
Bildbenennung
Wortgenerierung (z.B. Nennen Sie möglichst viele Tiere!)
Wortlesen (HUND)
Pseudowortlesen (HUNG)
Analyse von 82 Hirnaktivierungsxperimenten mit vier verschiedenenWortproduktionsaufgaben:
Estimate of probability of overlap under the assumption of a random distribution of activated regions
number of regions: 110
mean number of activated regions: r
chance probability for a region to be reportedas activated in a single experiment (p1): r/110
chance probability for a region to be reported as activated in n1 out of n experiments:
21 n1
n1
21)p(1p!n!n
n!p (with n1 + n2 = n)
Reliability criterion: p < 0.1 cut-off point in binomial distribution
Example region 1
Number of experiments: 82
Mean number of reported regions: 12.4
Reliably activated: 12 or more experiments
Reliably not activated: 4 or less experiments
Example region 2
Number of experiments: 23
Mean number of reported regions: 10.4
Reliably activated: 5 or more experiments
Reliably not activated: -
Zuverlässig aktivierte (rot) und nicht aktivierte (blau) Hirngebiete (basierend auf allen 82 Studien)
TASK ANALYSIS
Many tasks were not just word production tasks; they involved other operations as well.
For instance, when you name the picture of a horse, you not only produce the word 'horse', but you also look at the picture and recognize it. Such additional 'lead-in' operations involve the activation of additional brain regions. These should be filtered out.
That requires a systematic task analysis, a distinction between 'lead-in' and 'core' operations of word production.
Responses during Verb Generation Task
BANANATROUSERSCHAIRGLASSESTRUMPETPENCILBUTTONBIRDEARDOOR
peel, slip on, eat up, plantput on, wash, mend, buy, warmsit, build, nail, sell, work, learnclean, put on, step on, buy, seeblow, make music, put away, hear, playsharpen, break, put away, drawtear off, close, openfly, eat up, singhear, pinchopen, close, kick against
Konzeptuelle Vorbereitung
lexikalische Selektion
lexikalisches Konzept
Lemma
Wortformzugriff
Wortform
Syllabifizierung
phonologisches Wort
phonetische Enkodierung
abstraktes Motorprogramm
Artikulation
gesprochenes Wort
visuelle Objekt-erkennung
Einleitungsprozesse KernprozesseAufgabe
Worterkennung Objektvorstellung Gedächtnis etc.
visuelle Worterkennung
Graphem/PhonemKonversion
Bildbenennung
Wortgenerierung
Wortlesen
Pseudowortlesen
Selb
stmo
nito
ring
aussprechen vs. Wort “denken”
Konzeptuelle Vorbereitung
lexikalische Selektion
lexikalisches Konzept
Lemma
Wortformzugriff
Wortform
Syllabifizierung
phonologisches Wort
phonetische Enkodierung
abstraktes Motorprogramm
Artikulation
gesprochenes Wort
visuelle Objekt-erkennung
Einleitungsprozesse KernprozesseAufgabe
Worterkennung Objektvorstellung Gedächtnis etc.
visuelle Worterkennung
Graphem/PhonemKonversion
Bildbenennung
Wortgenerierung
Wortlesen
Pseudowortlesen
Selb
stmo
nito
ring
aussprechen vs. Wort “denken”
Konzeptuelle Vorbereitung
lexikalische Selektion
lexikalisches Konzept
Lemma
Wortformzugriff
Wortform
Syllabifizierung
phonologisches Wort
phonetische Enkodierung
abstraktes Motorprogramm
Artikulation
gesprochenes Wort
visuelle Objekt-erkennung
Einleitungsprozesse KernprozesseAufgabe
Worterkennung Objektvorstellung Gedächtnis etc.
visuelle Worterkennung
Graphem/PhonemKonversion
Bildbenennung
Wortgenerierung
Wortlesen
Pseudowortlesen
Selb
stmo
nito
ring
aussprechen vs. Wort “denken”
Konzeptuelle Vorbereitung
lexikalische Selektion
lexikalisches Konzept
Lemma
Wortformzugriff
Wortform
Syllabifizierung
phonologisches Wort
phonetische Enkodierung
abstraktes Motorprogramm
Artikulation
gesprochenes Wort
visuelle Objekt-erkennung
Einleitungsprozesse KernprozesseAufgabe
Worterkennung Objektvorstellung Gedächtnis etc.
visuelle Worterkennung
Graphem/PhonemKonversion
Bildbenennung
Wortgenerierung
Wortlesen
Pseudowortlesen
Selb
stmo
nito
ring
aussprechen vs. Wort “denken”
Schematische Darstellung des Ergebnisses der Meta-Analyse von 82 Hirnaktivierungsstudien
Indefrey, P. and Levelt, W.J.M. (2004) Cognition
Picture naming
Study Hernandez Hernandez De Bleser Vingerhoets Rodriguez-FornellsYear 2000 2001 2003 (non-cognates) 2003 2005Journal BrainLang Neuroimage Neuroimage Neuroimage JOCNNo. Subj. 6 6 11 12 11Method fMRI fMRI PET fMRI fMRIactive condition cov picture naming cov picture naming cov picture naming cov picture naming onsetdec on picturescontrol condition rest rest fixation scrambled pictures fixationadditional tasksL1 Spa Spa Fle Dut SpaL2 Eng Eng Fre Fre/Eng GerL2 onset <5 <5 10 10.3/13.5 3L2 duration 18 >16 8-11 17/14 >20L2 proficiency high high high good balancedL2 use L2 dominant L2 dominant n.g low/high L2 dominant
found in L2 onset
L2 proficiency L2 exposure
De Bleser 2003 10 good – very good ?
Vingerhoets 2003 10-14 mixed low/high
not found in
Rodrigues- Fornells 2005
3 balanced dominant
Hernandez 2001 <5 high dominant
Hernandez 2000 <5 high dominant
mean no. areas: 1.4
Picture naming
Stronger activation in L2 as compared to L1
Conclusions
L1 and L2 word production recruit the same set of areas (but only at the group level).
No areas are more strongly recruited in L1.
The left posterior inferior frontal gyrus may be recruited more strongly in L2 speakers with late L2 onset and/or low proficiency.
This region contains L1 specific but no L2 specific sites that are necessary for word production.
This region is the most likely candidate for post-lexical phonological encoding (syllabification) in L1 word production.
Left posterior IFG optimized for L1 phonotactic constraints and less efficient for early L2?
The cognitive architecture of listening to language
speech signal
interpretation
decoding
segmenting
speech code
phonemes, syllables
phonological processing
word recognition
syntactic analysis
thematic analysis
integration with other knowledge sources
Study Stimulus #
Belin 1998 200ms frequency transition, 60/min 1
Belin 1998 40ms frequency transition, 60/min 2
Belin 1999 synthetic diphthong, 6/min 3
Binder 2000 tones, different frequencies, 90/min 4
Bookheimer 1998 pseudowords, 9/min 5
Celsis 1999 syllables, 180/min 6
Celsis 1999 tones, 500 + 700Hz, 180/min 7
di Salle 2001 tones, 1000Hz, 6/min 8
Engelien 1995 environmental sounds, 10/min 9
Fiez 1996 pseudowords, 60/min 10
Fiez 1996 words, 60/min 11
Giraud 2000 vowels vs. expecting vowels, 120/min 12
Holcomb 1998 tones, 1500Hz + lower tones, 30/min 13
Jäncke 1999 tones, 1000Hz, 60/min 14
Lockwood 1999 tones, 500 + 4000Hz, 60/min 15
Mellet 1996 words, 30/min 16
Mirz 1999 music 17
Mirz 1999 sentences 18
Study Stimulus #
Mirz 1999 tones, 1000Hz 19
Mirz 1999 tones, 1000 + 4000Hz 20
Mirz 1999 words 21
Müller 1997 sentences, 12/min 22
Petersen 1988 words, 60/min 23
Price 1996 words, 40/min 24
Price 1996 words, different rates 25
Suzuki 2002a words, 60/min 26
Suzuki 2002b tones, 1000Hz, 60/min 27
Thivard 2000tones with spectral maxima, 60/min
28
Warburton 1996 words, 4/min 29
Wise 1991 pseudowords, 40 or 60/min 30
Wong 1999 reversed sentences, 30/min 31
Wong 1999 sentences, 30/min 32
Wong 1999 words, 30/min 33
Wong 2002 reversed words, 15/min 34
Wong 2002 sentences, 12/min 35
Wong 2002 words, 15/min 36
Indefrey & Cutler, 2004
Studies comparing auditory stimuli to silent baseline conditions
Study Stimulus vs. control stimulus #
Benson 2001 CVC > CV > V 1
Binder 1996 words vs. tones 2
Binder 2000 pseudo vs. tones 3
Binder 2000 reversed words vs. tones 4
Binder 2000 words vs. tones 5
Giraud 2000 amplitude modulated noise vs. noise 6
Giraud 2000 sentences vs. vowels 7
Giraud 2000 words vs. vowels 8
Hall 2002 frequency modulated vs. static tone 9
Hall 2002 harmonic vs. single tone 10
Jäncke 2002 syllables vs. 350 ms white noise bursts 11
Jäncke 2002 syllables vs. steady state portion of vowel 12
Jäncke 2002 syllables vs. tones 13
Müller 2002 90% 1000Hz + 10% 500Hz vs. 1000Hz 14
Mummery 1999 words vs. signal correlated noise 15
Price 1996 words vs. reversed words 16
Schlosser 1998 sentences vs. unknown language 17
Scott 2000 sentences vs. rotated sentences 18
Thivard 2000 frequency transition vs. stationary tone 19
Indefrey & Cutler, 2004
Studies comparing auditory stimuli to simpler auditory stimuli
Listening to speech without an additional task induces extensive bilateral temporal activation but no reliable activation of Broca’s area.
Summary
With increasing linguistic complexity of stimuli, the distance of activation maxima from the primary auditory cortex increases; particularly in the left hemisphere.
It seems to be the highest linguistic processing level that leads to the most significant activation difference compared to a silent control.
Summary
The left hemisphere shows a clearer stimulus-specific differentiation of activation maxima.
Areas that seem to be especially related to (post-) lexical and sentence level processing can be identified.
Summary
bilateral posterior STG: phonology
left posterior STS: lexical phonology
left anterior STS: possibly lexical and sentential prosody, possibly lexical and sentential meaning
Summary
Neuroimaging studies on syntactic processing
Approach A: Syntax versus no syntax
The cat is chasing the mouse. (Rest)The cat is chasing the mouse. &%$#@ &%#$@ %2#The cat is chasing the mouse. dgjrt hgjtrdf frt fpg hgrfbdwThe cat is chasing the mouse. mouse the chasing cat is the
ACTIVATION CONTROL
Advantage: Syntactic parsing not subtracted out
Disadvantage: Activated areas may be related to
nonsyntactic processing components
Neuroimaging studies on syntactic processing
Approach B: More syntax versus less syntax
The cat are chasing the mouse.
The cat is chasing the mouse.
The cat is chasing the mouse.
The mouse is chased by the cat.
The cat is chasing the mouse.
The cat is chasing the rat.
The mouse that the cat chased stole the cheese.
The cat chased the mouse that stole the cheese.
ACTIVATION CONTROL
Advantage: Nonsyntactic processing components well controlled
Disadvantage: Syntactic parsing (in part) subtracted out
How many studies must agreeon a certain area?
number of regions = 110
mean number of activated regions per experiment = 5.1
chance probability for a region to be reported as activated in a single experiment: p1 = 5.1 / 110 = 0.046
chance probability for a region to be reported as activated in n1 out of n experiments:21 n
1n
121
)p(1p!n!n
n!p (with n1 + n2 = n)
Sentences vs. control below sentence level
Syntactically more vs. less demanding sentences
Semantically more vs. less demanding sentences
53 57 51Reading Listening Reading Listening Reading Listening
23 33 42 18 37 15Note: some studies reported sentence reading and listening data
Hagoort & Indefrey (2014)
Sentences vs. control below sentence level
Syntactically more vs. less demanding sentences
Semantically more vs. less demanding sentences
53 57 51Reading Listening Reading Listening Reading Listening
23 33 42 18 37 15Note: some studies reported sentence reading and listening data
Hagoort & Indefrey (2014)
Interim summary
Compared to low-level control conditions, sentence processing activates left posterior inferior frontal (BA 44, 45, 47) and left temporal cortex
Sentence listening activates bilateral temporal cortices
For passive sentence listening or word-level control conditions BA 44 (pars opercularis) is no longer reliably found
> understanding simple sentences may not involve (detectable) syntactic processing
Condition1: Sentences
Der rote Kreis stößt die grüne Ellipse weg.
(The red circle pushes the green ellipse away.)
Condition 2: Noun phrases
roter Kreis, grüne Ellipse, wegstoßen
(red circle, green ellipse, push away)
Condition 3: Single words
Kreis, rot, Ellipse, grün, wegstoßen
(circle, red, ellipse, green, push away)
All conditions at slow (6/min) and fast (8/min) rate.
Sentences vs. Single Words
Activation maximum at -60,14,12
Indefrey et al. (2004) Brain & Language
Activation maximum at -54,6,10
Indefrey et al. (2001) PNAS
S and NP production vs. control (W)
Indefrey, Hellwig, Herzog, Seitz & Hagoort (2004) Brain & Language
Sentences vs. control below sentence level
Syntactically more vs. less demanding sentences
Semantically more vs. less demanding sentences
53 57 51Reading Listening Reading Listening Reading Listening
23 33 42 18 37 15Note: some studies reported sentence reading and listening data
Hagoort & Indefrey (2014)
Syntactically demanding Violation Inflection
The test is being explain/explained* Grammatical category The dance is being not too seriously rehearsal/rehearsed* (Cooke et al., 2006)
Ambiguity He noticed that landing planes frightens some new pilots. (Rodd et al., 2010)
Complexity Relative clauses: The reporter who the senator attacked admitted the error. The reporter who attacked the senator admitted the error. (Just et al. , 1996) Non-canonicity: The red book John gave to the professor from Oxford. John gave the red book to the professor from Oxford. (Ben-Shachar et al., 2004)
Semantically demanding Violation Selection restrictions:
Dutch trains are sour. World knowledge: Dutch trains are white. (Hagoort et al. 2004)
Ambiguity The reporter commented that modern compounds react unpredictably. (Rodd et al., 2010)
Complexity Metaphor: A sailboat is a floating leaf. (Diaz & Hogstrom, 2011) Metonymy, coercion, causal relationships Africa is hungry/arid. (Rapp et al., 2011) The novelist began/wrote the book. (Husband et al., 2011) The boys were having an argument. They became more and more angry./They began hitting each other. The next day they had bruises. (Kuperberg et al., 2006) Indirect question/reply, Irony: Did you like my presentation?/ How hard is it to give a good presentation? It is hard to give a good presentation. (Bašnáková et al., 2013) Ann promised to keep her party dress clean. She came home covered in mud. Her mom said:”Thanks for staying so clean.” (Eviatar & Just, 2006)
Summary
Both syntactic and semantic compositional processing recruit frontal and temporal regions
In both frontal and temporal cortex there is a gradient with syntactic processes activating more dorsal regions and semantic processes activating more ventral regions.
this pattern speaks against reducing syntactic processing to some aspect of semantic processing
Frontal and temporal cortex activations dissociate for violations this finding supports a division of labor with frontal cortex subserving
compositional processing as such and temporal cortex having a role in the retrieval of lexically stored semantic and syntactic information
Understanding non-literal meaning, in particular ‚speaker meaning‘ requires the recruitment of additional regions, such as the medial prefrontal cortex, supporting non-linguistic Theory-of-Mind processing.
Sentence comprehension
Study Chee Hasegawa Luke Frenck-Mestre RueschemeyerYear 1999 (exp1) 2002 2002 2005 2005 (exp2, corr sent)Journal Neuron Neuroimage HBM Neuroreport HBMNo. Subj. 15 10 7 6 18/14Method fMRI fMRI fMRI fMRI fMRIactive condition sentence reading sentence listening VP reading overt sentence readingsentence listeningcontrol condition fixation fixation fontsizedec consonant strings restadditional tasks comprehension probe verification synt or sem dec judgmentL1 Chi/Eng Jap Chi Eng Ger/RusL2 Chi/Eng Eng Eng Fre GerL2 onset <6 12 >10 >12 n.g.L2 duration 13-20 14 10-21 >15 5L2 proficiency high high high(selfrating) high n.g.L2 use daily high(in L2 env.) moderate? high (in L2 env.) high (in L2 env.)
Narrative comprehension
Study Perani Perani Perani Nakai Nakada VingerhoetsYear 1996 1998 1998 1999 2001 2003Journal Neuroreport Brain Brain Neurosc letters Neurosc Research NeuroimageNo. Subj. 9 9 12 4 10 12Method PET PET PET fMRI fMRI fMRIactive condition story listening story listening story listening story listening paragraph reading story readingcontrol condition reversed speech reversed speech reversed speech rest false fonts pseudoword listsadditional tasksL1 Ita Ita Spa/Cat Jap Jap/Eng DutL2 Eng Eng Spa/Cat Eng Eng/Jap Fre/EngL2 onset 7 10 2 n.g. 10 10.3/13.5L2 duration 14-25 9-40 17-25 n.g. >10 17/14L2 proficiency moderate high high n.g. high goodL2 use low daily daily n.g. n.g. low/high
found in L2 onset
L2 proficiency
L2 exposure
Hasegawa 2002 12 high high
Luke 2002 >10 high ?
not found in
Nakai 1999 ? ? ?
Perani 1998 10 high high
Perani 1998 2 high high
Perani 1996 7 moderate low
Vingerhoets 2003 10-14 mixed low/high
Nakada 2001 >10 high ?
Rüschemeyer 2005 ? ? high
Chee 1999 <6 high high
Frenck-Mestre 2005 >12 high highmean no. areas: 2.4
Sentence listening / reading
Stronger activation in L2 as compared to L1
found in L2 onset
L2 proficiency
L2 exposure
Nakai 1999 ? ? ?
Rüschemeyer 2005 ? ? high
Luke 2002 >10 high ?
not found in
Perani 1998 10 high high
Perani 1998 2 high high
Perani 1996 7 moderate low
Vingerhoets 2003 10-14 mixed low/high
Nakada 2001 >10 high ?
Hasegawa 2002 12 high high
Chee 1999 <6 high high
Frenck-Mestre 2005 >12 high highmean no. areas: 2.4
Sentence listening / reading
Stronger activation in L2 as compared to L1
found in L2 onset
L2 proficiency
L2 exposure
Nakai 1999 ? ? ?
Luke 2002 >10 high ?
not found in
Perani 1998 10 high high
Perani 1998 2 high high
Perani 1996 7 moderate low
Vingerhoets 2003 10-14 mixed low/high
Nakada 2001 >10 high ?
Rüschemeyer 2005 ? ? high
Chee 1999 <6 high high
Frenck-Mestre 2005 >12 high highmean no. areas: 2.4
Sentence listening / reading
Stronger activation in L2 as compared to L1
found in L2 onset
L2 proficiency
L2 exposure
Rüschemeyer 2005 ? ? high
Luke 2002 >10 high ?
not found in
Nakai 1999 ? ? ?
Perani 1998 10 high high
Perani 1998 2 high high
Perani 1996 7 moderate low
Vingerhoets 2003 10-14 mixed low/high
Nakada 2001 >10 high ?
Chee 1999 <6 high high
Frenck-Mestre 2005 >12 high highmean no. areas: 2.4
Sentence listening / reading
Stronger activation in L2 as compared to L1
found in L2 onset
L2 proficiency
L2 exposure
Nakai 1999 ? ? ?
Hasegawa 2002 12 high high
not found in
Perani 1998 10 high high
Perani 1998 2 high high
Perani 1996 7 moderate low
Vingerhoets 2003 10-14 mixed low/high
Nakada 2001 >10 high ?
Rüschemeyer 2005 ? ? high
Chee 1999 <6 high high
Frenck-Mestre 2005 >12 high high
mean no. areas: 2.4
Sentence listening / reading
Stronger activation in L2 as compared to L1
found in L2 onset
L2 proficiency
L2 exposure
Hasegawa 2002 12 high high
Luke 2002 >10 high ?
not found in
Nakai 1999 ? ? ?
Perani 1998 10 high high
Perani 1998 2 high high
Perani 1996 7 moderate low
Vingerhoets 2003 10-14 mixed low/high
Nakada 2001 >10 high ?
Rüschemeyer 2005 ? ? high
Chee 1999 <6 high high
Frenck-Mestre 2005 >12 high highmean no. areas: 2.4
Sentence listening / reading
Stronger activation in L2 as compared to L1
Conclusions
L1 and L2 sentence level comprehension recruit the same set of areas.
Within this set of areas, there are to date no reliable activation level differences between L1 and L2 narrative comprehension.
Stronger L2 activation of the left posterior IFG may be found for speakers with late L2 onset when additional judgment tasks are used.
The left posterior IFG is the most likely candidate area for syntactic processing.
Left posterior IFG optimized for L1 syntax and less efficient for L2?
(taking into account word production data) Left posterior IFG optimized for L1 compositional processes and less efficient for L2?
in NL Lessons started to learn NL
TS April 03 school (6 hours/week) Febr 04
CX Jan 04 school (6 hours/day) Jan 04
ZY Jan 04 school (6 hours/day) Jan 04
HQ Jan 03 self study Jan 04
CJ April 02 school (6 hours/week) Jan 04
JX March 03 with a colleague (3 hours/week) Febr 04
Longitudinal study: Participants
Longitudinal Study: Methods
Test battery at 3, 6, 9, 15, 18, and 24 months
Behavioural testing ‘nonverbal’ intelligence test (Raven Progressive Matrices) handedness test (only at 0 months) standard Dutch proficiency test language questionnaire syntactic judgment test
fMRI experiment on syntactic processing
ERP experiment on semantic and syntactic violations
S: Het blauwe vierkant wordt door de gele cirkel weggestoten. (correct)Lan2 fang1kuai4 bei4 huang2 yuan2quan1 tui1zou3.The blue square is pushed away by the yellow circle.
De gele cirkel wordt door het blauwe vierkant weggestoten. (incorrect)Huang2 yuan2quan1 bei4 lan2 fang1kuai4 tui1zou3 The yellow circle is pushed away by the blue square away
W: (correct)
(incorrect)
vierkantfang1kuai4square
cirkelyuan2quan1circle
blauwlan2se4blue
geelhuang2se4yellow
cirkelyuan2quan1circle
cirkelyuan2quan1circle
geelhuang2se4yellow
blauwlan2se4blue
wegstotentui1zou3push away
wegstotentui1zou3push away
Conclusions
In L2-comprehension, both frontal and temporal areas show syntactic-processing related activation similar to L1 already after a few months of exposure.
The emergence of this activation precedes the ability to detect number or gender violations, but follows the ability to detect at least some types of word order violations.
So far, there does not seem to be a correlation between behavioral data and the enhanced hemodynamic response to sentences compared to word lists.