phonetics exam 2

mental lexicon

every memory you have of expressions you know (words, idioms, etc.)

lexical item

an entry in the mental lexicon

lexical access

the process of identifying what word is being produced

priming

facilitation of processing one item (the target) due to previous exposure to another item (the prime)

What things prime each other?

• inflected versions of the same word (love, loved, loving)

• semantically related words (king, queen)

• words with similar pronunciation (play, plate)

• same number of syllables

• same word stress

• same syntactic category

priming effect in production

• words said for a second time are reduced - less effort put into saying the word

frequency of usage effect in perception

• frequent words are more quickly identified

• shadowed faster

• memory representation is more refreshed

frequency of usage effects in production

produced more quickly and with more coarticulation

neighborhood of a lexical item

all the items that are similar in pronunciation - the greater the similarity, the closer they are

Dense neighborhood

a lot of frequent items are similar to it

Sparse neighborhood

not many similar items, items are infrequent

• these words are processed faster and more accurately

H & H theory

there is a continuum of speech carefulness, between hyper and hypo speech

hypospeech

minimal speech gestures, minimal perceptual distinctiveness (mumbling)

hyperspeech

agonizingly careful articulation

Lombard effect

speakers speak more carefully and louder when in loud and noisy environments

Experiment: Subjects had to read citation speech and clear speech

Results: vowels in clear speech were louder and longer than in citation speech, had less coarticulation, corner vowels in clear speech were further apart, greater VOT difference. Easier to identify the words said in clear speech than citation speech, listeners can remember these items better

Experiment: Speakers produced vowels after h in isolation. Then they had to adjust synthetic vowels until they sounded how they sound in their heads, how they imagine they produce the sound

Results: Adjusted vowel formants were more hyperarticulated than in the speaker’s productions. This is the hyperspace effect

Hyperspace effect

expansion of the vowel space beyond that for quite careful citation forms

Experimental factors

the ones you’re comparing in your experiment

Background factors

all other facts that can affect your data

How to deal with background factors

• hold them constant

  • incorporate them into your data

Bottom up processing

when one figures out the bigger units on the basis of the smaller units they contain

top down processing

figuring out the smaller constituent units on the basis of bigger units that contain them

Experiment: experimenters synthesized words varying in the VOT of an initial alveolar/velar stop. In one class, the voiced consonant would form a word, and the voiceless would not, in the other the voiceless would form a word and the voiced would not. Listeners had to identify the sound as voiced or voiceless.

Results: As expected, they recognized it as voiced when the VOT was lower, but they also recognized it as voiced more when it formed a real word than when it did not.

Implications: Listeners use both bottom-up and top-down processing

Phonotactic restrictions:

generalizations about what sequences of sounds can occur in some position in an utterance, e.g. at the beginning of a syllable

Experiment: Experimenters synthesized syllables of a liquid followed by [i], differing just in F3, then preceded it by a synthesized consonant, P T S or V. Listeners had to identify what they heard.

Results: More r responses when F3 was lower (expected), but in the midrange, they resisted identifying nonreal sound sequences.

Implications: Subjects bring their phonotactic knowledge to processing, provides a check and speeds processing through parallel processing

Experiment: experimenters presented words to subjects at varying signal-to-noise ratios, and they had to identify what was being said

Results: higher S/N ratio=more accuracy. Subjects did better with real words in sentences than with nonsense words, and did better with numbers in a sequence.

Implications: we fill in context, syntactic knowledge, and knowledge of discourse coherence

Experiment: recorded a sentence and replaced one of the sounds with a cough, listeners had to identify what sound was replaced.

Results: They all got it wrong.

Implications: context provides so much information that they don’t even need to hear the word to know what it is, and top-down trumps bottom-up processing.