Transcript A Ch 21-23
Understanding Spoken Speech ● Bottom up is minimal for humans, but crucial for computers – voice recognition Understanding Spoken Speech ● Top down in crucial for humans because – dialectal differences in speech – individual differences Understanding Spoken Speech ● Top down in crucial for humans because – dialectal differences in speech – individual differences – differences motivated by phonology • allophony Understanding Spoken Speech ● Top down in crucial for humans because – dialectal differences in speech – individual differences – differences motivated by phonology • • • allophony sounds flow into each other inability to segment individual sounds Understanding Spoken Speech ● Top down in crucial for humans because – dialectal differences in speech – individual differences – differences motivated by phonology • • • – allophony sounds flow into each other inability to segment individual sounds noise, poor pronunciation make words incomprehensible, yet we fill in the blanks Understanding Spoken Speech ● Top down in crucial for humans because – dialectal differences in speech – individual differences – differences motivated by phonology • • • allophony sounds flow into each other inability to segment individual sounds – noise, poor pronunciation make words incomprehensible, yet we fill in the blanks – No boundary markers between words Understanding Spoken Speech ● How does word recognition happen? – Theory: input turned into phonemes then matched to phonemically stored words – Theory: input matched analogically to phonetically stored items – Maybe both happen to some degree at the same time Understanding Spoken Speech ● Cue to word segmentation – English alternates stress and unstressed syllables – Many words have initial stress – b d g merge with p t k word finally Understanding Spoken Speech ● Cue to word segmentation – English alternates stress and unstressed syllables – Many words have initial stress – b d g merge with p t k word finally – p t k are strongly aspirated word initially Understanding Spoken Speech ● Cue to word segmentation – English alternates stress and unstressed syllables – Many words have initial stress – b d g merge with p t k word finally – p t k are strongly aspirated word initially – some consonant clusters are not possible within words, but are OK across words • rdst not in words, but words to Frequency Effects ● Hi freq words recognized faster Frequency Effects ● Hi freq words recognized faster – Are they stored separately and searched first? Frequency Effects ● Hi freq words recognized faster – Are they stored separately and searched first? – Are they stored with more worn paths? (strength) Word meanings ● How do you find the right meaning? – At first all meanings are activated Word meanings ● How do you find the right meaning? – At first all meanings are activated – context helps – words in ambiguous contexts take longer to comprehend • • the right turn (ambiguous) the left turn (unambiguous) Word meanings ● How do you find the right meaning? – At first all meanings are activated – context helps – words in ambiguous contexts take longer to comprehend • • • • the right turn (ambiguous) the left turn (unambiguous) They didn't like the bark (ambiguous) They didn't like the barking (unambiguous) Word meanings ● All meanings are initially activated – Prime: “There were spiders, roaches, and other bugs in the room” – Target: immediately after sentence • ANT recognized faster than SEW – This shows that bug activated ant also Word meanings ● All meanings are initially activated – Prime: “There were spiders, roaches, and other bugs in the room” – Target: immediately after sentence • ANT recognized faster than SEW – • This shows that bug activated ant also ANT and SPY activated equally fast – This shows that both meanings of BUG were actived. Word meanings ● LDT – Prime: “They all rose” – Target: STOOD • faster response than unrelated LOOKED Word meanings ● LDT – Prime: “They all rose” – Target: STOOD • faster response than unrelated LOOKED – Prime: “He gave her a rose” – Target: FLOWER • faster response than unrelated PICTURE Word meanings ● LDT – Prime: “They all rose” – Target: STOOD • faster response than unrelated LOOKED – Prime: “He gave her a rose” – Target: FLOWER • faster response than unrelated PICTURE Word meanings ● LDT – Prime: “They all rose” – Target: FLOWER • faster response than unrelated word Prime: “He gave her a rose” Target: STOOD • faster response than unrelated word So, all meanings activated initially Word meanings ● LDT – Prime: auditory capt- – Target: BOAT or GUARD • both recognized faster Word meanings ● LDT – Prime: auditory capt- – Target: BOAT or GUARD • both recognized faster – Prime: auditory capti- – Target: BOAT or GUARD • guard recognized faster, but not boat Word meanings ● LDT – Prime: auditory capt- – Target: BOAT or GUARD • both recognized faster – Prime: auditory capti- – Target: BOAT or GUARD • guard recognized faster, but not boat – Prime: auditory capta- – Target: BOAT or GUARD • boat recognized faster, but not guard Word meanings ● Cohort Model – As word is perceived it initially activates all possibilities • • • • s- activates all words with sst- activates all words with st- (fewer words) sta- activates all words with staand so on until stand only activates stand Synesthesia ● Overlap between different senses – shapes associated with tastes – numbers associated with colors Synesthesia ● Overlap between different senses – shapes associated with tastes – numbers associated with colors Synesthesia ● ● If 7 is yellow and nine is blue for a synesthete then math problems solved quicker is colors correspond For non synesthete colors won't affect time Synesthesia ● A non synesthete will see panel on the left ● A synesthete will see the panel on the right Synesthesia ● Why does it happen? – Overlap between brain areas Synesthesia ● Why does it happen? – Overlap between brain areas • • green=grapheme recognition area red=color recognition area Synesthesia ● Could other overlaps explain other things? – foot fetish Synesthesia ● Could other overlaps explain other things? – perceiving vowels as colors? • Which color corresponds to [i, u, a]? Synesthesia ● Could other overlaps explain other things? – perceiving vowels as colors? • • Which color corresponds to [i, u, a]? [i]=yellow, [u]=blue/green, [a]=red Synesthesia ● Could other overlaps explain other things? – which is bouba and which is kiki? Synesthesia ● Could other overlaps explain other things? – which is bouba and which is kiki? – back vowels are round and front are jagged? Synesthesia ● Could other overlaps explain other things? – which is bouba and which is kiki? – back vowels are round and front are jagged? – b is round and k is jagged? Synesthesia ● Could other overlaps explain other things? – which is bouba and which is kiki? – back vowels are round and front are jagged? – b is round and k is jagged? – same results for English and Tamil speakers Synesthesia ● Could other overlaps explain other things? – which is bigger teetee or tootoo? Synesthesia ● Could other overlaps explain other things? – which is bigger teetee or tootoo? – which is smaller laylee or mowmoo? Synesthesia ● Could other overlaps explain other things? – which is bigger teetee or tootoo? – which is smaller laylee or mowmoo? – Why? Look at F2