Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
17 Cards in this Set
- Front
- Back
|
Close Shadowers
|
Shadowing: Sometimes we have the ability to predict what the next word will be and are able to recognize it despite many things that degrade speech signals
Close Shadowers: Some people are able to recognize a word very fast from the target onset (beginning of a word-250 milliseconds) Often words can be recognized before the bottom-up information or contextual cues can even been proccessed |
|
|
Low vs. High Predictability sentences
(In relation to contextual cues and shadowing) |
Depending on what we know about the world and the subject being discussed, the predictibilty of words in a sentence can vary depending on how much context is given. For example:
- I went to the library to get a book about ______ - I went to the dentist to get my teeth_______ -I went to the Honda Dealereship to buy a new _______ It all depends on prior knowledge and level of context |
|
|
The Cohort Model (Bottom-Up proccessing)
*also how bottom-up and top-down processing are used in the cohort model |
-This model explains that the first phonological unit of a word from a speech signal activates ALL WORDS in the listeners mental lexicon that begin with that sound
-The so called "Cohort": this is a mental list of all the potential target words that brain considers to be the target word -bottom-up signaling continues to play a role. As more of the signal is processed bottom-up eliminates more words out of the cohort that no longer match the progressing signal Shadowing studies: - in this study, mispronounced target words were always 3 syllables long and varied in errors among the 1st, 2nd and 3rd syllable - subjects rarely corrected mispronunciations of the first syllable -listeners used bottom-up stratagey for the first syllable (the cohort) - Words are ONLY removed, never added - bottom up activates inital set of cohort and words that will never be added to the cohort -top-down and context cant add any words but it can eliminate words from the cohort that are inconsistent from the top-down information -high-level context (top-down) can also keep words active in the cohort- even if they start to differ from the bottom-up signal ex: asfinaut would be kept (because of high context) |
|
|
Top-Down Proccessing (context)
|
-Representations on one level (the words) influence the lower level of the smaller compositional units (phonemes)
-word level ---> phonological level= top-down processing - The Ganong Effect is an example: when we hear a degraded signal our brain is guided by mental representations to identify the word for us - our lexical knowledge affects our speech perception - We use prior knowledge when understanding what we heard - There is no such thing as a "perfect signal" There is always going to be some kind of "noise" interference -Phoneme restoration effect is an example |
|
|
Bottom-Up Processing (Cohort)
|
- the representation of a word is completely contructed from the signal
-It is "signal-driven", the signal determines the recognition of of the phonological units - the more signal we hear, the more we can identify the word -Promblem: word within words |
|
|
Word Frequency
(can be considered top-down processing) |
This influences word recognition
- this is simply a measure of how many time we hear a particular word in everyday speech (come & go vs. par & fragment) - word recognition is fast for relatively frequent words -less frequent words require a bit more bottom-up in order to be recognized |
|
|
Phonological "neighborhoods"
(can be considered top-down processing) |
influences word recognition
-this refers to words a listener knows that are phonologically "close" to a target word - words are considered to be phonologically neighbors IF they differ by just one phonological unit -ex: "Cat" vs. "Bat" -e: "SCar" vs. "CHar" -ex: "cAre" vs. "cOre" - not based on spelling, but on SOUND |
|
|
Neighborhood Density
|
- refers to how many phonological neighbors a target word has relative to each other
-ex: "caat" lives in a DENSE neighborhood (many words phonologically similar) -ex: "elephant" lives in a SPARCE neighborhood (not many other similarly phonological words) |
|
|
Effects of: WORD FREQUENCY
|
- if listeners can only partially recover the phonological information in a words so that it makes it ambiguous between two competing words, we usually and should pick the more frequent word because it is probably being the one used
|
|
|
Effects of: CONTEXT
|
-Context acts a cue
- context during the process of word recognition adds robustness and allows more efficient, rapid processing of activating and identifying the correct word, sometimes before the word is even done being processed |
|
|
Word Frequency: (Lexical Decision):
|
-requires subjects to push "yes" if it they hear a real word, and "no" if it is a non-word
- subjects are asked to do this as quickly as possible the point: - word recognition should be faster for more frequent words IF frequency does affect word recognition |
|
|
Word Frequency: (Cross-modal priming)
|
-subjects hear a word (the PRIME)
- and then subjects see either a related or non-related word (the PROBE) the cross-modal part: - 2 sense modalities- vision, and hearing -subjects will recognize the PROBE faster if the prime and probe are semantically related (like dog and bone) - the probe is subconsciously activated partially by when the prime is heard |
|
|
What makes a word easier or harder to recognize?
|
- the frequency of a word (highly frequent words)
- the contextual clues -prior knowledge -frequency of competitors - the word has fewer phonological neighbors (SPARSE) - and if those neighbors (competitors) are low frequency words -noise - these all affect the speed of which, and the difficulty of recognizing a word |
|
|
Effects of Phonological Neighborhoods
|
the cohort model incorporates the notion of COMPETITORS:
- competitors are all the words that are active in the cohort at any given time -these words are all COMPETING to selected as the target/recognized word -the addition of things like noise make identifying a word much harder |
|
|
Trace Model vs. Cohort Model
|
Trace Model:
- this models human speech perception and word recognition - computer based -allows researchers to examine from moment to moment what words in the model are being "considered" - the cohort model and trace model are SIMILAR because they both include words that are competing for recognition - SO it theorizes that competitors can be "activated" as more of the word is perceived -in contrast of the cohort model that hypothesizes that words are eliminated base recognition of the initial consonant sound - the trace model allows a word to become active even if the beginning of the word is inconsistent with the beginning of a target word (bottom up signaling) -as long as there is sufficient match to the signal somewhere else in the word it predicts that words that rhyme with the target word will become somewhat active even if the initial sounds are inconsistent with the signal: - ex: "tar" might become active during the process of the word "car" -"cat" could not be confused with "bat" in the cohort model but it could be confused in the trace model ONGOING INTERACTION BETWEEN BOTTOM-UP and TOP DOWN PROCESSING - the activating more words as more of the signal is hear (B-U) - keeping words active even if they are inconsistent with the beginning of the word due to you be able to match it to similar words like ones that rhyme (or phonological neighbors) |
|
|
what is the phoneme restoration effect?
(in relation to top-down processing) |
- the fact that we can still determine a degraded signal, this is an example of this
- say the [s] in "legislature" is replaced by static noise or a cough - listeners are aware of this noise and can also report the missing phonological unit - but our minds restore the [s] in their perception due to top-down processing *this same process would not happen if it was a non-word |
|
|
extra stuff
|
*cascades: rapid eye movements fixating on objects that are present that a speaker is talking about
*eye-tracking: helps researchers possibly predict what words are active as a listener listens to speech |
