LTM and Language Flashcards
(24 cards)
Hierarchical Network Model
Links and nodes that get more detailed as they go down (like a tree diagram)
Semantic Feature Comparison Model
Defining and characteristic features
Two-stage model: compare all features, if intermediate similarity compare defining features
Spreading Activation Model
A big mess that has lighting up connections between associated objects
Unique to the individual, different strengths of the connections
No hierarchical structure, supported by semantic priming
Connectionist Models - Key Features
Simple, neuron-like units connected by excitatory or inhibitory paths
“Weights” at each connection determine how strongly an incoming signal will activate the next unit. New events change connection strength by adjusting connection weight
Benefits of Connectionist Models
Based on the nervous system, not totally disrupted by damage (graceful degradation), generalised learning due to similar patterns
Repetition Priming
Presentation of a specific item more than once improves performance
Repetition creates a semantic representation of an item, transforming episodic memory into semantic memory
Feustel, Shiffrin & Salasoo
No words versus words: present an item briefly and mask it, determine the discrete threshold for being able to recognise what it was
Over repetitions, nonwords become unitised like words
Pattern of Language Development
7 months: babbling
1 year: meaningful words
2 years: multiword utterances
Levels of Language
Phoneme, morpheme, word, sentence, discourse or text
Morpheme
Smallest meaningful unit in a language (i.e. cat, or un). Can stand alone as its own word in addition to being part of a word, or can exist as only part of a word
Semantics vs Pragmatics
Semantics = combining words to capture meaning
Pragmatics = social rules of language (interpreting ambiguity, context, inferences, etc)
Spectrograms
Visual displays of auditory signals showing frequency on the y-axis and time on the x-axis, darkness represents amplitude/intensity
Benefits and Downsides of Spectrograms
Benefits: provide a visual display of speech
Downsides: not consistent, B doesn’t always sound the same. Lack of invariance
Categorical Perception
A listener’s ability to perceive differences across category boundaries, but not within a particular category
Can tell difference between /b/ and /p/ but only at certain points
Consonants are processed this way
Continuous Perception
Perceiver’s ability to perceive differences WITHIN category boundaries as well as across boundaries
Vowels are processed this way
Speech is Special Theory
Speech is special and separate from other types of cognitive and perceptual processes
Problems: Non-human animals can perceive consonants categorically, non-speech auditory signals (plucks and bows) can be perceived categorically
Word Processing Factors
Phonological relatedness
Word frequency
Context
Slowiaczek & Hamburger
Number of initial shared phonemes affects response time with words.
No shared phonemes = slow, one shared phoneme = fast, two or three shared phonemes = much slower
Neighbourhood Density Effects
Inhibitory connections between words that share initial phonemes
Phonemic Restoration
Replace one phoneme with a cough, ask subjects to identify where the cough occurred
Subjects noticed the cough, but could not identify the correct location or note that the particular sound was missing
Context clues: depending on the predicate, the subject will be interpreted differently
Parsing
Determining how words are grouped together into phrases in order to understand the meaning of the sentence
Parsing Approaches
Syntax first: analyse the sentence using syntactic info first, then use other information
Interactionist approach: semantic information influences processing from the outset, not just syntax first
Late Closure
When encountering a new word, the parser assumes that the word is part of the current phrase
Tanenhaus et al.
Sentence processing is influenced by the scene we are observing
Apple / Towel / Box task
One and two apple conditions
Support for interactionist approach because the conditions have different results and suggest that the listener is using bot syntactic information and information provided by the scene
