Lecture Notes Cognitive Psych March 16 & 18
Lecture Notes Cognitive Psych March 16 & 18 PSY 0422
Popular in Cognitive Psychology
Popular in Psychlogy
This 6 page Class Notes was uploaded by Lindsay Notetaker on Friday March 20, 2015. The Class Notes belongs to PSY 0422 at University of Pittsburgh taught by in Spring2015. Since its upload, it has received 61 views. For similar materials see Cognitive Psychology in Psychlogy at University of Pittsburgh.
Reviews for Lecture Notes Cognitive Psych March 16 & 18
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 03/20/15
316 Lecture 16 Language Challenges of speech Semantic start off with meaning Pragmatic how to say it adjust to listener 0 Making inferences about what the listener knows 0 Speak differently to a 15 month old than an adult Syntactic what grammar construction 0 Many different syntactically correct ways to convey meaning and we need to choose Lexical which of roughly 50000 words to use 0 Synonymous with words mental dictionary Phonological how do words sound 0 Element of sound 14000 syllables Articulatory how to physically say it Series of very fast memory retrievals Common problems in speech Disfluencies half of talking time pausing retracing hesitating o Re ects thinking and word retrieval problems Deadends tipofthetongue problems 0 Come to a dead stop and can t access a particular word that we know Speech errors many Influences on language production Memory need somewhere to start more available more likely to be used earlier Structure Memory in uences on production Just in time imperative use what s available when it s available 0 Use activated information to help us get over the challenge of what to select when Experiment suppose people listen to words and see a picture church in a lightening storm 0 Say what you see primed with associated word earlier in utterance o thunder lightening hitting the church 0 worship church being hit by lightening Strong evidence that whatever is activated in memory just before has an impact on what words we decide to use and when Use it or lose it Structure in language production Speakers generate language in syntactic chunks clauses noun amp verb phrases Evidence from pauses and hesitations 0 Mean pause length at clause boundary 1 sec 0 Shorter pause lengths within clause More evidence when speakers repeat or correct themselves they tend to repeat or correct a whole phrase repetition of units Example Turn off the stove the heater switch NOT turn off the stove off the heater switch we repeat a whole linguistic noun phrase How to study speech production Speech error corpora collections of naturally occurring and transcribed slips of the tongue such as Spoonerisms 0 Example The Lord is a shoving leopard to his ock 9 not intended Various problems with using natural speech 0 Depends on listener judgments 0 Some errors hard to classify 0 Some theoretically important ones occur rarely Speech Errors in the lab Baars Motley amp MacKay 1975 subjects are asked to silently read word pairs Occasionally they are cued to read aloud as quickly as possible 0 Priming trials ball dome beak doll bus door bell dark 0 Critical trial darn bore people say barn door 9 larger error 0 Priming trials big dutch bang doll bill deal bark dog 0 Critical trial dart board 9 lesser error The error bard doard occurs only about 10 of the time compared with 30 for the barn door set Lexical bias tendency to create words not nonwords Freud on Speech Errors Freud said speech errors arise from the concurrent action or the mutually opposing action of two different intentions Failure to censure unconscious intention results in speech error that expresses some aspect of the unconscious often sexual oedipal Cognitive Account Speech errors re ect properties of the language system They result from concurrent competing activation of certain linguistic units Failure to achieve correct speech production output allows insight into the structure of the language system and processing e g which units are concurrently active Levels of slips Errors happen at different linguistic levels 0 Sound errors rack pat and pack rat 0 Morpheme errors selfinstruct destruction and selfdestruct instruction 0 Word errors writing a mother to my letter and writing a letter to my mother Type of Errors Sound exchanges night life nife lite beast of burden burst of beaden coin toss toin coss Anticipation Errors take my bike bake my bike b sound takes over Perseveration Errors beef noodle beef needle Blends tab taxicab shruck structurechunk Two Important Speech Errors Lexical biasz tendency in phonological errors to create words rather than nonwords 0 Hold hard cash instead of cold hard cash likely o Weautiful boman instead of beautiful woman less likely Mixed errorsz tendency to produce errors that are semantically and phonologically related to the intended word 0 Apricot instead of apple share same beginning sounds Interpretation Suggests that semantic and phonological retrieval must be interacting Perhaps selection of sounds feeds back to the earlier level and biases selection OfWOFdS O o o o o o o o o o C Interactive activation model PDP model with layers of nodes corresponding to semantic features words and phonemes All connections are bidirectional if we activate a word the meaning is activated and then the sound is activated etc Word Activation spreads through the network Word node that receives the highest activation is selected Phonemes that receive the highest activation are selected Example Semantics of CAT turns on word node corresponding to cat phoneme Also gives some activation to words for related concepts Onsets Vowels Codas CCdOg DD Crat The word cat turns on the corresponding phonological nodes k ae and t Phonological nodes feed back activation to the word layer strengthening cat and turning on rat and mat Cat should be most active followed by rat which gets activation from semantics and phonology Interactive Activation Model How does this model account for the ndings Lexical bias output of the model is determined by selecting the combination of phonemes with the highest activation level 0 Likely to be one that has topdown activation from a word node Mixed errors activation ows in both directions 0 Word selection is in uenced not only by topdown semantic information but also by backward activation from the phonological level 0 Mixed errors receive activation both topdown from semantic level and bottomup from the phonological level and are thus more likely to be erroneously selected 318 Lecture 17 Language Summary of Speech production Speaking involves coordination of many processes In uenced by memory and structure Speech errors provide cues as to how speech is produced Interactive model does best job 0 Suggests interaction between semantic word and phonological stages Understanding Language Overview Route to understanding language is similar to what we learned earlier about visual input Visual Processing light patterns 9 Visual features 9 Objects 9 Scenes Language Processing sound patterns 9 Phonemes 9 Words 9 Sentences Four levels of analysis Phonology sounds put together to form words production and perception of language sounds Syntax rules for order of words and phrases sentences Semantics meaning word sentence paragraph Pragmatics rules of conversation paragraph longer 0 Meaning is relative to concepts beliefs Phonology Phonemes individual speech sounds smallest unit of speech sounds Phonemes can be combined to form words 0 Bat and Cat differ by one phoneme b and c Phonological regularities memorization or rules 0 We memorize all of these sounds and access them when necessary 0 Rules for combining sounds implicit knowledge I How to decide to pronounce plural nonwords Examples pg 294 Why is speech perception hard 0 Coarticulation parallel transmission of information multiple sounds continuous stream I We need to divide up those sounds into phonemes and identify them even though it s coming in a continuous stream Segmentation As native language speakers we easily separate the language into words WWW Hearing the information doesn t give us insight on when sound begins and ends Where we segment speech language is critical t Ex THEREDONATEAKETTLEOFTENCHIPS what meaning you get out of this depends critically on where you separate the words How do we segment Motor theory production and perception linked backwards engineering perception is innate and special to humans Auditory theory speech perception derives from the general properties of the auditory system not species specific Phonology Summary When people speak 0 the sounds of separate words blur together much more than we think 0 we use our past eXperience with words to separate the words hypothesized to be regulated by implicit rules Syntax and Semantics Both are critical in understanding language Sentences are regulated by grammar and semantics Semantics is meaning of words 0 Example turtle green animal with hard shell Grammar is a set of rules that structure language 0 Example in English the order of a sentence is subject verb object Example of phrase structure draw out S 9 NP VP Semantic Grammar Distinction The brain also makes a distinction between semantics and grammar as parts of complex language Patient Evidence Damage to Broca s area 9 grammar impairment Damage to Wernicke s area 9 semantics impairment Our brains appear to parse syntax grammar separately from semantics meaning Understanding Text Involves building a representation of the whole text You can think of this representation as a propositional representation pattern of symbols that make up a meaningful declarative sentence and often can be thought of as true and false 8 The hippie touched the debutant in the park 9 SO Relation 10 she slapped him Evidence for propositional representations Memory load experiment 0 More propositions require more memory and so harder to remember Priming experiment 0 Geese crossed the horizon as wind shuf ed the clouds o Priming is stronger within a proposition geesehorizon than across propositions geeseclouds Language and Ambiguity Types of Ambiguity Lexical ambiguity a word has more than one meaning Surface ambiguity alternative meanings have different phrase structures Underlying ambiguity alternative meanings have the same phrase structures Immediacy of processing Leads to slower or more errorprone processing with ambiguity Using eyetracking to study garden path sentences Syntax and Semantics Summary We try to understand the syntactic structure of a sentence on a wordbyword basis building propositional representations in memory Shows how language and memory come together This involves making some guesses Sometimes we are fooled when the best guess or most common choice is wrong