1Posner AA vs Aa RT Data2Posner: Reaction Times toDifferent Pairs3LanguageThe Mental LexiconHow is information about words stored in thebrain? – “Mental Lexicon”, semanticinformation, syntactic information, word forms.One or several?Production vs. comprehensionInput - outputOrthographic - phonological4LanguageWordsHow many are there?Problems for counting: inflected forms,separate senses, dialect, scientificnomenclature, slang, addition of word-formingelements, etc.5LanguageWordsWords:Cambridge Encyclopedia of the English Language: >1,000,000.Oxford Companion to the English Language: > 1,000,000.Vocabulary:Shakespeare’s plays: 29,066 distinct forms, incl. propernames., est. 18,000-25,000.18-month old: ~50 – two years: ~300, 3-year old:~1,000 (adding 3/day)16-year old: 10,000-12,000.College grad.: 20,000-25,000.6LanguageThe Mental LexiconSemantic network: representation of word meanings.Semantic priming studies:prime-target: “chair-table”unrelated: “chair-nurse”pseudoword pairs: “chair-flark”7LanguageLexical Semantic NetworksNodes: concepts, wordsLinks: connection strengths reflect the amountof conceptual or semantic relatedness betweennodes.Stimulus presentation activates one (or several)nodes, activation spreads along linksProblems:How are links determined?What are neural equivalents of nodes and links(single neurons or cell assemblies)?8LanguageLexical Semantic NetworksDistributed processing models:Feature-based representation, input is astimulus vector and network “settles” towards atarget semantic vector.9PET data: HierarchicalExperiment10LanguageNeural Basis of the Mental LexiconEvidence from brain lesions (dissociations between namingof non-living and living things, selective deficits insemantic memory after stroke etc.)Study by Damasio et al., 1996:Propose that lexical knowledge is organized among threeseparate neural systems:1. Conceptual content (meaning, semantic features)2. Phonological elements (sounds)3. Modality-independent lexical knowledge (left temporallobe)11LanguageNeural Basis of the Mental LexiconFirst part of study: correlate neuroanatomy and clinicalresults (category-specific naming deficits, semantic featuresare intact, 127 patients).12Language13LanguageNlaDefects in word retrieval for single categoriespersons animalstools14LanguageNeural Basis of the Mental LexiconSecond part of the study: PET activation (9subjects).15LanguageNla16LanguageWritten InputRecognizing written input is a problem of visual recognition.Models by Selfridge (1959) and McClelland/Rumelhart (1981)171819LanguageWritten InputEvidence for specialized visual extrastriateareas dealing with letter strings (Petersen et al.,1990)Distinction between high-level vision andsemantic processing?“word-form area”? – lesions result in “purealexia”Remember, written language is ~6,000 yrs old.20LanguageWritten InputPetersen et al., 199021LanguageWritten InputStudy by Puce et al., 1996Nonsense, non-pronouncable letterstrings - minimize semanticprocessing (but keep visual letterforms intact)22Written InputStudy by Puce et al., 1996Faces: yellowLetter strings: pink23Verb Generation in PET24Effects of Practice25Two Pathways for Verb Generation26Mediation of Colors27Mediation of Nouns28LanguageSpoken Input29LanguageSpoken