dc.contributor.author
Hofmann, Markus J.
dc.date.accessioned
2018-06-07T18:43:16Z
dc.date.available
2011-11-15T10:07:18.551Z
dc.identifier.uri
https://refubium.fu-berlin.de/handle/fub188/5354
dc.identifier.uri
http://dx.doi.org/10.17169/refubium-9553
dc.description
Acknowledgments VI Zusammenfassung VIII Summary XIV General Introduction 1
Extending the theoretical base: Interactive Activation Models 2 The original
interactive activation model and the identification of letters 3 The Multiple
Read-Out Model and the recognition of words 5 Dual-route models and
phonological representations in visual word recognition and reading aloud 11
How to model (semantic) associations during word recognition? 17 Overview of
the present studies and their methods 22 Study 1: Sub-lexical frequency
measures provided by corpus analyses 22 Study 2: Word frequency, lexicality
and optical imaging 24 Study 3: Modeling electrophysiological responses to
conflicting lexical representations 29 Study 4: Affective connotation of
lexical representations, ERPs, and pupillometry 31 Study 5: Modeling
associations between lexical representations and Receiver Operation
Characteristics 34 Study 1: Sub-lexical frequency measures for orthographic
and phonological units in German 41 Introduction 42 Grain sizes, domains,
databases, and measures 46 Grain sizes: Syllable, dual unit, or single unit 47
Processing domains: Orthography or phonology 49 Databases: Lemma or word form
50 Measures: Type or token 51 Method 53 Results 55 Syllable frequencies of the
lemma database 56 Syllable frequencies of the word form database 56 Dual unit
frequencies of the lemma database 57 Dual unit frequencies of the word form
database 58 Single unit frequencies of the lemma database 58 Single unit
frequencies of the word form database 59 Discussion 60 Study 2: Differential
activation of frontal and parietal regions during visual word recognition: An
optical topography study 65 Introduction 67 Methods 71 Participants 71
Materials 71 Experimental procedure 72 Data acquisition 73 Data analysis 73
Results 76 Behavioral results 76 fNIRS (optical topography) 78 Discussion 82
The lexicality effect 83 The word frequency effect 85 Optical topography as a
tool to investigate word recognition 86 Appendix 90 Study 3: Conflict
monitoring engages the mediofrontal cortex during nonword processing 91
Introduction 92 Methods 95 Participants 95 Materials 95 Procedure 95 Data
acquisition 96 Data analysis 97 Results 98 Behavioral 98 ERPs 98 sLORETA 98
Discussion 101 Conclusion 103 Study 4: Affective processing within 1/10th of a
second: High-Arousal is necessary for early facilitative processing of
negative but not positive words 105 Introduction 107 Method 112 Participants
112 Materials 112 Procedure 113 Data acquisition 114 Data analysis 114 Results
117 Behavioral 117 ERPs 117 sLORETA 118 Discussion 122 Study 5: Remembering
words in context as predicted by an Associative Read-Out Model 127
Introduction 129 Does associative spreading activate ‘false memories’? 130 Can
each item’s signal be detected in an explicit memory task? 134 Simulation
methods: The AROM and its predictions 137 Experimental methods: Testing the
AROM’s predictions 140 Participants 140 Corpus 140 Stimuli 140 Procedure 143
Experimental and modeling results 144 Discussion 148 Conclusions 155 General
Discussion: Summary and outlook 159 Sub-lexical frequencies 160 Sub-lexical
frequency measures constrained the interpretation of effects! 160 Can the
matching of global features be replaced by specific ones? 161 Word frequency
and optical imaging 163 Optical imaging revealed greater “neural activations”
to low frequency words! 163 What is “neural activation” in the IFG? 164
Lexical conflicts 167 Lexical conflicts predicted behavioral data and ACC
activation! 167 Does associative-semantic competition predict IFG activation?
170 Affective word features 171 Affective lexical features elicited behavioral
and ERP, but no pupil dilation effects! 171 Can semantic cohesiveness account
for affective effects? 173 Associative-semantic representations 176 Modeling
associations between the word stimuli of an experiment predicted false and
veridical memories! 176 Going beyond measurement models of familiarity and
recollection? 179 The rebirth of a mental lexicon: How to answer the challenge
of fixing the structure of time? 182 Does the mind construct semantic
taxonomies from associations? 186 Conclusions 189 References 191 Erklärung I
Curriculum Vitae III Wissenschaftlicher Werdegang III Lehre III Vorträge III
Poster V Publizierte Konferenzbeiträge und Buchkapitel VI Gutachtertätigkeiten
VI Zeitschriftenartikel VII
dc.description.abstract
This dissertation investigated visual word recognition based on the
theoretical framework of interactive activation models (IAMs, McClelland &
Rumelhart, 1981). Study 1 provided sub-lexical frequency measures for German,
which were used as control variables in the Studies 2, 4, and 5. Study 2 was
the first of three studies using the lexical decision task. The optical
imaging results showed that words elicit greater neural responses than
nonwords in the left inferior parietal gyrus, which suggests a role of this
region during the integration of orthographic, phonological and semantic
representations. Greater activations for word stimuli in the superior frontal
gyrus can be interpreted in terms of decision-related processes during visual
word recognition. Moreover, rare words elicited greater neural activation than
common words in the left inferior frontal gyrus. This word frequency effect
suggests a role of this region during the selection of a semantic
representation from many co-activated semantic candidates. Study 3 used an IAM
to calculate the conflicts between orthographic representations, and set this
model-generated measure of lexical competition into a direct relation with an
event-related potentials (ERP) negativity between 400 and 600 ms post-
stimulus. The electric sources of the ERP-conflict effects were attributed to
the anterior cingulate cortex. The model accounted for a significant portion
of item-level variance in reaction times, error rates and mean amplitudes.
Study 4 showed that positive and high-arousal negative words elicit response
facilitation and an early ERP effect between 80 and 120 ms post-stimulus, when
compared to neutral words. The ERP-effect in high-arousal negative words was
source-localized in the left fusiform and middle temporal gyri. The latter
finding may be explained by the larger amount of associative relations of
affective words. Study 5 captures associative relations between words for
IAMs. Two words were defined as 'associated', when they co-occur significantly
often together in the sentences of a large corpus. This corresponds to Hebbian
learning: Items being repeatedly presented together are likely to be
associated. The results of a recognition memory task showed that learned and
non-learned words elicit greater 'yes' response rates when they provide a
larger amount of associated items in the stimulus set. The co-occurrence
statistics were further used to implement associations between words in a
contextual representation layer. This IAM-model predicted which word is
recognized with which probability on an item-level. Because many of the most
strongly co-activated words revealed a semantic relation to the presented word
(e.g., synonymy), the resulting 'Associative Read-Out Model' is the first IAM
with a fully implemented semantic representation layer.
de
dc.description.abstract
Diese Dissertation untersucht die visuelle Worterkennung auf der theoretischen
Grundlage des 'Interactive Activation Models' (IAM, McClelland und Rumelhart,
1981). Studie 1 stellt sub-lexikalische Häufigkeitsmaße für das Deutsche zur
Verfügung: Orthographische und phonologische Silbenfrequenzen, Bigramm-
Biphonemfrequenzen, sowie Buchstaben- und Phonemfrequenzen. Solche Maße dienen
in den Studien 2, 4, und 5 als Kontroll-Variablen. Studie 2 ist die erste von
drei Studien, welche die visuelle Worterkennung mit der lexikalischen
Entscheidungsaufgabe untersucht. Die optischen Bildgebungsbefunde zeigen, dass
Wörter höhere neuronale Aktivierungen im linken inferioren Parietallappen
auslösen, was auf die Rolle dieser Region bei der Integration
orthographischer, phonologischer und semantischer Repräsentationen hinweist.
Die höheren Aktivierungen für Wörter im superioren Frontallappen weisen auf
die entscheidungsrelevanten Prozesse der Worterkennung hin. Seltene Wörter
lösten höhere Aktivierungen im linken inferioren Frontallappen im Vergleich zu
häufigen Wörtern aus, was die Beteiligung dieser Region bei der Auswahl einer
semantischen Repräsentation aus verschiedenen konfligierenden Kandidaten
nahelegt. Studie 3 setzt simulierte Konflikte zwischen orthographischen
Repräsentationen bei der Verarbeitung von Nichtwörtern in eine direkte
Beziehung zu einer Negativierung des ereigniskorrelierten Potentials (EKP)
zwischen 400 und 600 ms. Die Quellen dieser Aktivierungen wurden im anterioren
cingulären Cortex verortet. Das Modell klärte signifikante Varianzanteile für
Reaktionszeiten, Fehlerraten und EKP-Negativierungen auf. Studie 4 zeigt, dass
positive und hocherregend negative Wörter Antworterleichterungen im Vergleich
zu neutralen Wörtern und eine frühen EKP-Negativierung zwischen 80 und 120 ms
nach Reizdarbietung auslösen. Der EKP-Effekt hocherregend negativer Wörter
konnte im linken fusiformen und im mittleren temporalen Gyrus verortet werden,
was dafür spricht, dass affektiv konnotierte Wörter mehr assoziativ verknüpfte
Wörter aufweisen. Studie 5 erschließt assoziative Verknüpfungen zwischen
Wörtern für IAMs. Zwei Wörter wurden als 'assoziiert' definiert, wenn sie in
den Sätzen eines großen Satzkorpus signifikant häufig gemeinsam auftreten.
Dies entspricht der Hebb'schen Lernregel: Wörter, die häufig gemeinsam
auftreten, sind wahrscheinlich assoziiert. Die Ergebnisse einer
Wiedererkennens-Gedächtnisaufgabe zeigen, dass gelernte und nicht-gelernte
Wörter mehr 'Ja-'Antworten auslösen, wenn sie eine größere Anzahl assoziierte
Wörter im Reizmaterial aufweisen. Die ko-okkurrenzstatistischen Maße wurden
benutzt, um eine kontextuelle Modell-Repräsentationsebene mit assoziativen
Verbindungsgewichten auszustatten. Das Modell sagt auf dem Item-level voraus,
welches Wort mit welcher Wahrscheinlichkeit wiedererkannt wird. Da viele der
am stärksten assoziierten Wörter zum präsentierten Wort eine semantische
Verknüpfung aufweisen (z.B. Synonymie), ist das so gewonnene 'Associative
Read-Out Model', das erste IAM mit einer semantischen Repräsentationsebene.
de
dc.format.extent
XVIII, 227, IX S.
dc.rights.uri
http://www.fu-berlin.de/sites/refubium/rechtliches/Nutzungsbedingungen
dc.subject
interactive activation model
dc.subject
sub-lexical frequency measures
dc.subject
word recognition
dc.subject
associative spreading activation
dc.subject.ddc
100 Philosophie und Psychologie::150 Psychologie
dc.title
Setting letters and words into context
dc.contributor.contact
mhof@zedat.fu-berlin.de
dc.contributor.firstReferee
Prof. Dr. Arthur Jacobs
dc.contributor.furtherReferee
Prof. Dr. Lars Kuchinke
dc.date.accepted
2011-11-04
dc.identifier.urn
urn:nbn:de:kobv:188-fudissthesis000000025784-3
dc.title.subtitle
An associative read-out model
dc.title.translated
Auf der Suche nach dem Sinn des Lesens
de
dc.title.translatedsubtitle
Buchstaben und Wörter im Kontext
en
refubium.affiliation
Erziehungswissenschaft und Psychologie
de
refubium.mycore.fudocsId
FUDISS_thesis_000000025784
refubium.mycore.derivateId
FUDISS_derivate_000000010214
dcterms.accessRights.dnb
free
dcterms.accessRights.openaire
open access