LING330: Quiz #3 Flashcards
Mapping from acoustics to perception: linear or non-linear?
Equal changes in frequency/amplitude don’t end up as equal responses in the perceptual system
What’s the jnd?
Just noticeable difference
A young healthy ear can tell the difference between a few hertz at lower decibels but once it gets over 1000 hertz you can’t tell the difference
Scientists will often transfers raw frequency values into a logarithmic scale mathematically that better matches perceptual sensitivities. What are some of these scales?
Semitone scale
The bark scale
The Mel scale
The erb scale
The Bark scale
Divides frequency range of human hearing into 24 bands (each step from one band to the next sounds about equal)
Logarithmic part: difference in hertz between each bark gets larger as the range gets higher but below 500hz=nearly linear
At higher frequencies, we need a bigger difference in values to perceive a change
Normalization
We perceive all speech frequency/timing values as relative
Our brains use info in other parts of signal to normalize the stimulus (interpreting values relative to other values not to an external standard)
+
Same item can be heard differently when in different contexts (depends on the sounds around it)
Adjusts our expectation of what sounds should sound like using extraction of social and personal info
Explains why we can perceive consonants/vowels as the same in diff articulatory contexts
What is VOT?
Voice onset time
In a categorical perception experiment involving VOT, what was discovered?
Listeners=good at distinguishing sounds that are in DIFF categories, bad at distinguishing sounds in SAME category (even if acoustic difference is the same)
What affects category boundary perception?
Language experience (depends on the language you speak)
What’s special about human speech perception?
Processed by diff cognitive system than other acoustic data like music and non-linguistic sounds
**weakened when found that other domains use categorization such as visual perception and animals also do this
Cue integration
Brain takes multiple pieces of info that it receives and uses its perception of them to create a single object (ex: phoneme)
**pieces of info separated in time and extracted at diff points in auditory pathway when listening to words or sentences
What role does the auditory cortex play?
Puts pieces of phonetic info back together again after they’ve been extracted by other parts of the brain
Describe top-down processing
Acoustic cues come into brain from the BOTTOM UP but
Predictions made in brain for speech processing are TOP DOWN
Brain is PRIMED to hear certain sounds/words in a specific context
Responsible for PHONEME RESTORATION
**proof of word-level prototypes are stored
Phoneme restoration
Experiment where sound segment is replaced by a non-speech sound; sound segment will still be perceived even if missing because it is obvious through context
**proof of word-level prototypes are stored
What are prototype models?
Traditional answers of “to what linguistic representation is the speech signal mapped?” are based on this
Assumes that ideal version of a linguistic unit is stored in memory
Accounts for categorical perception
Diffs in detail are not noticed unless they cause a mapping to diff stored representation
Abstract/idealized version of linguistic unit stored in memory, incoming symbol checked against it
Detailed info discarded
How do feature detectors work? (Prototype models)
Pick up distinct acoustic events and match them to segment characteristics (different between languages)
Segments identified -> assembled into words
Cues before certain segments that help us predict
Proposition that units of perception=DEMI SYLLABLES (CV or VC combos) or WHOLE SYLLABLES
Motor theory of speech perception
Against view of stored word-level prototypes
Argue that units of perception are articulatory gestures themselves
Sound patterns=medium by which we hear but articulatory movements that caused the sounds=OBJECTS HEARD
Learn ability=mirror neurons (same neurons fired when seeing someone doing something and when doing it yourself)
How do the motor theory of speech perception theorists explain how it is learned?
Babies listen to their own utterances which teaches them what ACOUSTIC PATTERNS go with what SPEECH GESTURES
NEURAL LINK between auditory and motor cortex
Arguments against motor theory of speech perception
- we can perceive speech that no vocal tract could create (sine wave speech)
- brain injury affects production and perception separately
Prototype models vs exemplar models
Prototype: assume that match to linguistic unit is made -> detailed info in the signal=thrown out and speech processing continues on abstract level
Exemplar models: very detailed memory traces are retained and referenced
Episodic traces=linked together at multiple levels into multi-dimensional groupings
Detailed and specific, not abstract like prototype model
Categorization=label + details of specific instances
New tokens are compared against stored tokens, not an abstract ideal
How categorization differs in exemplar models
Stimulus SIMILAR to exemplar: faster and more accurate categorization; category label is reinforced + boundaries sharpen
Stimulus NOT similar: category may shift -> language change
Support for exemplar models
- supported by psycholinguistic research that says details do matter + connections exist at multiple levels
- consistent with frequency effects (how often you hear a sound makes a difference)
- priming
Commonly used method for conducting a discrimination experiment
ABX or oddball task Stimuli A, B and X A and B=always diff stimuli X=always matches A or B Listener has to indicate whether X matches A or B
What are humans better at: discrimination or identification?
Neither (same)
Perception differences between plosives and vowels
Plosives: categorically perceived (plosives=short/dynamic)
Vowels: perceived on a continuum (vowels=relatively long but when shortened, perceived categorically lol plosives)
Two response measures that show that listeners do perceive diffs among members of the same category
Reaction time: used as a measure of amount of processing (the more ambiguous a stimulus, the more time it takes for a subject to process it + time to perform a certain reaction will be longer)
Goodness rating: shows how good an example of a given item is, as judged by a subject
Technique to test babies on speech perception (specifically plosives)
High amplitude sucking paradigm (HAS)
Babies sucking rate INCREASES when presented with stimuli that is new/unfamiliar to them
Pacifier wired to a transducer
When sucking rate decreases=habituation
Infant perception results suggest what?
Suggests that categorical perception observed in adults is NOT due to linguistic experience throughout life, rather PRESENT AT BIRTH
Two competing hypotheses of speech is special debate (whether or not human speech perception uses mechanisms that evolved especially for processing speech)
Special mechanism account (SMA): speech perceived as result of specialized, speech-specific mechanisms that exist only in humans
General mechanism account (GMA): speech perception based on general auditory and cognitive mechanisms and doesn’t require innate speech-specific processes
Evidence in the “speech is special” debate
1- perception of speech and non-speech sounds by humans (non-speech can be created to resemble speech in amplitude and frequency; tests claim that speech-specific mechanisms ARE required)
2- perception of speech in humans vs animals
The study of non-native speech perception
Cross-language speech perception
Two major models of L2 phonetic category acquisition
1- the perceptual assimilation model: model of non-native speech perception by naive listeners that aren’t familiar with the target language (non-native contrasts perceived in terms of how similar they are to phonological categories in native lang)
2- speech learning model: focuses on attainment of L2 pronunciation and says that errors in L2 production have perceptual basis; doesn’t specify whether relation between sounds in L1 and L2 is acoustic or auditory in nature
SLM predicts that the process of equivalence classification prevents L2 learners from making a new category for similar (not new) sounds (affects pronunciation)
Perceptual difficulty that adult listeners has in L2 due to what factors?
- L1/L2 phoneme inventories
- specific acoustic cues to mark phoneme distinctions
- allophonic distribution of phonetic segments in L1 and L2
- age
- amount of exposure to L2
- quality of L2 input
- how much they use the L1 and L2
- *perceptual system=plasticity that allows for phonetic categories to be changed
Five factors that affect perception
1- non linearity 2- normalization 3- cue integration 4- processing 5- language exposure
Is mapping of acoustic signal and/or perception linear?
Mapping of acoustic signal=linear
Perception=non linear
Equal changes in frequency (pitch) and amplitude (loudness) are not matched by equal responses in our perceptual system
Are auditory frequency scales logarithmic?
Yes!
200 hz is twice as high as 100 hz
1 octave is also perceived as twice as high
But a two octave increase (ex: double the pitch/performed frequency) = 4 x f0!
Perceptual scales of loudness vs duration
Loudness: we perceive sounds between 2-4 KHz better than very high or very low pitch sounds; phons/sones
Duration: words in isolation tend to be longer than those in connected speech; perceived as same duration
In terms of VOT, categorical perception means that English listeners switch from categorizing sound as [d] to [t] when?
At a specific point!
Special mechanism vs general mechanism account
Special mechanism account: speech perception is because of speech specific mechanisms
General mechanism account: speech perception is based on general auditory and cognitive mechanisms
Testing non-human perception on chinchillas
Electric shock at one end of continuum Drinking water at other end Remaining members of continuum added Categorical perception similar to humans **also conducted on monkeys and quails
Conclusions about basic speech perception
It’s not only in humans
It’s not speech specific
Result of general auditory mechanisms
2 theories of perception in L2
Sounds inventories of L1 and L2 affect which within-category distinctions are made
1- SLM (speech learning model)
2- PAM (perceptual assimilation model)
PAM (perceptual assimilation model)
-non native contrasts perceived in terms of gestural similarity to L1 phonological categories
-difference in leaning new vs similar phones:
Single category vs two category assimilation
Categorized (perceived as instances of L1) and Uncategorized (nothing like L1)
SLM (speech learning model)
Focused on articulation
Doesn’t specify articulation vs acoustics perception
New vs similar phones
English speakers learn French [y] but not the dental [t]
