Spoken Word Recognition

Front 1

segmental information

Back 1

-distinguishes speech sounds
-specifies which sounds are in an utterance, and must be primary determinant of speech recognition

Front 2

invariance problem of spoken word recognition

Back 2

→ the problem that the highly variable speech signal must be mapped into discrete lexical representations (parsing!)

Front 3

prelexical stage

Back 3

a representation of the segmental content in the input speech signal is constructed for use in subsequent lexical access

Front 4

mismatch between the input and stored lexical knowledge

Back 4

-mismatch on an initial phoneme (e.g. Dutch word honing (honey) changed into woning (dwelling) appears to be enough to block lexical access
-presentation of a prime such as honing facilitated responses to the target big (bee) but woning had no such effect

Front 5

more facts about mismatch

Back 5

-segmental mismatch is more disruptive of lexical access in word-initial than word-final position

-this is because of the temporal nature of speech → a word with a final mispronunciation is, prior to the arrival of that mispronunciation, consistent with the intended word, but a word with an initial mispronunciation is immediately put at a disadvantage

-The more dissimilar the mismatching sound is to the sound in the word’s correct pronunciation, the more disruptive that sound is in lexical access

Front 6

suprasegmental information

Back 6

-specifies the prosodic characteristics of words

• Specifies how the prosodic structure of words (i.e., their syllabification and lexical-stress pattern) and the position of a word in the intonational structure of an utterance
• Incorrect stress pattern in a word fragment disrupts lexical access

Front 7

cross-linguistic differences in the role of lexical stress

Back 7

o Not all languages make lexical-stress distinctions, and speakers of those languages cannot use stress info
o Japanese speakers use the pitch-accent patterns of words in lexical access, and Mandarin speakers use tone info
o Languages could have free stress, or fixed-stress
• Listeners of fixed-stress languages are likely to use stress info less in lexical access, since it is predictable and does not enhance lexical distinctions
• For free-stress languages, there is variability in how much value stress info appears to play
• *MAIN POINT→ stress info is used in lexical access when it is available, but its usage appears to depend on its informational value

Front 8

cross-linguistic differences in finding word boundaries

Back 8

• English listeners are more sensitive to the rhythmic distinction between strong syllables (containing full vowels) and weak syllables (those with reduced vowels) → have picked up on the fact that content words in English tend to begin with strong syllables
• Speakers of French and English don’t show this sensitivity because these languages do not make metrical distinction
• Because the rhythm in Japanese is based on mora (subsyllabic structure), Japanese listeners use moraic info in segmentation

Front 9

phonotactic restrictions -> finding word boundaries

Back 9

-knowledge about the phonotactic restrictions on syllable structure on a language (i.e., that the sequence /mr/ in English cannot occur within a syllable) could be used to indicate the location of likely word boundaries

Front 10

multiple lexical activation

Back 10

multiple lexical hypotheses appear to be considered simultaneously as speech is heard
• E.g: words which differ in their onsets or their offsets from the word in the input (e.g. speaker and beetle given beaker) are considered in parallel with what ultimately proves to be the correct hypothesis
• Evidence: words beginning in the same way as other words are jointly considered, as are onset-embedded words
• Thus, as the number and frequency of similar-sounding words increases, word recognition becomes harder
o As the number of words in the lexical neighborhood increases, competition gets fiercer, and recognition is delayed
o The size and nature of the priming effect arises from word or word-fragment primes depends on the number of words beginning in the same way as those primes

Front 11

Uniqueness Point

Back 11

-the point as one moves left to right through a word at which the info in the signal uniquely specifies that word.
• Many short words do not become unique before they end

Front 12

phoneme identification experiment

Back 12

-subjects are presented with phonemes and have to say what they heard (forced choice) like pit or pet

-the perception of vowels is continuous, with each vowel appearing to shade gradually into the next

Front 13

Voice Onset Time

Back 13

-the interval between the release of the closure and the beginning of the voicing associated with the vowel sound --> the acoustic correlate of the distinction between voiced, voiceless and aspirated plosives lies in this interval

Front 14

studying VOT between /p/ and /b/

Back 14

IF a test stimuli fall on opposite sides of the boundary, the subjects are very accurate in their IDs. If the stimuli fall on the same side of the boundary, subjects' responses are like they are guessing

-indicates that hearers can discriminate the phonetic categories voiced vs voiceless well but cannot hear the differences within these categories

-categorical perception--> the hearer perceives in terms of categories rather than in terms of minute gradations of sound