Upgrade to remove ads
Only $35.99/year

Models of Language Midterm

Terms in this set (67)

Modular
Information in one domain is contained separately from another domain. Lesions would impact only that domain.
Distributed
Language knowledge is spread out, involving a combination of many different cognitive processes. Elements and processes occur in different locations.
Factors that influence whether an item is stored or rule-based
1. Frequency of occurrence
2. Predictability of form
3. Compositionality of meaning
4. Productivity
Predictability of form
Degree to which you can predict the morphology
Frequency of Occurrence
More frequent words are stored, whereas less frequent words are not.
Compositionality of Meaning
Degree to which you can predict meaning based on the meaning of its parts.
Productivity
degree to which the process can be used to create new words that have no been heard before.
Storage
Memorizing and containing all forms
Rules
Memorizing exceptions and base forms, then applying rules to produce other forms
Connectionism
Language models that can function as a computer program. More frequent words have stronger connections. Our experience with familiar words influence our experience with new words
Top-down Processing
Cognitive process that uses pre-existing to knowledge to influence how new stimuli is perceived.
Bottom-Up Processing
Cognitive process that uses sensory information exclusively to perceive new stimuli.
Innate aspects of language
-Ability to perceive highly sensitive differences in speech sounds at birth
Learned aspects of language
-Native language and its morphological rules
Speech is variable due to
1. Contextual Clues
2. Speaker Variability
3. Segmentation
Categorical Perpcetion
Categories possessed by an observer influences the observer's perception.
McGurk Effect
Uses top-down processing (knowledge of what speech sounds look like) to influence how auditory stimuli is perceived
Characteristics of infant-directed speech
-More extreme pitch variation
-Slower speech rate
-Consistent use of pauses to indicate boundaries
-Shorter sentences
-Fewer grammatical errors
-Repeat words in different sentence types
Methods to study speech perception in infants and young children
Conditioned Head-Turn Test
High-Amplitude Sucking Test
Speech Production: birth - 2mo
-Reflexive crying
-Vegetative Sounds
-Involuntary sounds (e.g., burps)
Speech Production: 2 - 4mo
-Cooing/laughter in repsonse to social interaction
-Primary vowel sounds
-Turn-taking
-Laughter
Speech Production: 4 - 6mo
- Vocal Play
- Bilabial trills
- Grunts/growls
- Some syllables w/ consonants
-Extremes of loud/soft, high/low
Speech Production: 6mo+
-Babbling
-CV Sequences
-Reduplicated babbling
-Non-reduplicated babbling
Lexicon
Mental Dictionary. List of words w/ associated information.
Lexical Decision
Deciding whether a word belongs to the lexicon or not. Use response time infer how strongly an individual believes their choice. Impacted by frequency, respond quickly to high-frequency words
Semantic Priming
Individuals are faster to respond to cases in which words are related to each other. Semantic relatedness affects RTs in semantic priming tasks.
Phonological Priming
Phonological relatedness affects RTs.
Context Effects that impact word recognition
1. Phoneme Restoration Effect
2. Top-down processing
Phoneme Restoration Effect
When a sound within a sentence is replaced with another sound, participants respond with original sound, not replacement sound.
Lexicon is organized by
Frequency, Semantic Relatedness, & Phonological Relatedness
Serial Search Model Process
1. Search access files to find the word.
2. Access files are modality specific (visual, auditory), and organized by frequency/phonetic properties
3. Information about the word from the access file indicate where to obtain word in master file
4. Search master file, organized by semantic/phonological relatedness) to obtain word
Serial Search Model Predicts
Frequency, Semantic Relatedness, and Phonological Relatedness
Serial Search Model Flaws
-No incorporation of rapidly changing context
-Doesn't incorporate sentence context
- Implausible
- Pseudowords sends system into infinite loop
Serial Search Model Benefits
-Access files are divided based on phonology and arranged by frequency
- Semantically related words are cross-referenced in master file
Logogen Model (Parallel Search)
Each word is associated with a logogen (frequency counter), resting activation level is determined by frequency, level increases as evidence accrues (context, relatedness, sounds heard), logogen fires at threshold
Logogen Model predicts
High frequency words are more quickly recognized, semantically/phonologically related words are more quickly recognized
Logogen Model Benefits
-Logogens involve timer that prevents continuous processing for pseudowords
Cohort Model
Phonetic INPUT, ACCESS (perceptual representation activates lexical items generating set of candidates), SELECTION (most likely candidate chosen), INTEGRATION (semantic and syntactic properties of chosen words are used)
Cohort Model is
Modular & utilizes bottom-up processing
Cohort Model Cons
-Requires beginning of word to be high quality
-Context can influence/override choice of candidate (occurs during integration)
-Requires you to know when a word begins (segementation)
TRACE Model
Highly interactive connectionist model, activation passes both ways. Integrates phonetic features, phonemes, words. Has inhibitory processes within a level (voicing)
TRACE differs from cohort
-Trace model can change what is activated based on later input
-Can activavate things on basis on context
Inhibition
Connectionist models can have inhibitory connections that influence elements from activating in other levels
Dual-Route Model of Written Word Recognition
Modular model: Top route processes the printed word, finds the combination of letters in the lexicon to read the word aloud. Bottom Route- processes printed word, recognizes the letters, produces the phonological sounds to read the word aloud.
Top route used for
Frequently seen words
Phonological Dyslexia
Trouble with non-words, difficulty sounding out words through letter-phonological mapping
Surface Dyslexia
Trouble with irreguarly spelled words, because while they CAN use the bottom pathway to sound out the word, it is not possible. Bad memory
Double Dissociation
When two related mental processes function independently of each other
Evidence for double dissociation
Subtypes of dyslexia
Connectionist models don't account for double disociations
Connectionist models utilize all processes concurrently
Phoneme Anticipation
The mirst of May
Phoneme Perseveration
God rest re merry gentlemen
Phoneme Deletion
Backgound Lighting
Word Substitution
Get me a fork (spoon)
Feature Assimilation
Corn on the cop
Morpheme Exchange
The buzz are beeing
Speech errors demonstrate
1. Language is productive
2. Utterances are planned before they are spoken
Errors impact syntactice level
Place wrong semantic word into syntax tree
Error impact morphemic level
Switch content morphemes (bee & buzz) but not function morphemes
Errors impact phonemic level
-consonants replace consonants
-vowels replace vowels
-onsets replace onsets

Follow phonotactic rules of the language. Adhere's to sonority hierarchy.
Bock & Levelt's Model of Spoken Word Recognition
Modular, rule-based, accounts for speech errors.

Message > Functional processing (lexical selection, functional assignment, word exchanges), positional processing (constituent assembly, inflection, stranding errors) > phonological encoding (sound errors)
Evidence for Bock & Level
Sheet vs. Sheep, ophonological encoding follows lexical selection.
Dell Model of Spoken Word Recognition
Connectionist & distributed, more connections = more likely to result in errors. When you want to say something, begin with lexical form then retrieve phonological and semantic features.
Evidence for Dell
Most errors were made when same vowel but real word, and fewer when it was a different vowel and non-word
Wernicke-Geshwind Spoken Word Recognition
Auditory info stimuate the primary auditory cortex, wernicke's area analyses the sound and finds it within the lexicon, lexical information is transmitted to Broca's area via AF, Broca's area develops motor plan & sends it to PMC, PMC & does it.
Wernicke-Geshwind Written word recognition
Image is analysed by V1 & Angular gyrus, angular gyrus decodes the image and matches it to a form in the lexicon located in wernicke's area, info is transmitted to Broca via AF, Broca develops motor plan to repeat word and send it to motor1, PMC does it.
Ellis & Young Model of Lexicon
Modular model, two lexicons (auditory input & visual input), each step is a processes, there is just input/output
Sets with similar terms