Presentation on theme: "Marslen-Wilson Big Question: “What processes take place during the period that the sensory information is accumulating for the listener” during spoken."— Presentation transcript:
Marslen-Wilson Big Question: “What processes take place during the period that the sensory information is accumulating for the listener” during spoken word recognition? (How is this different from the question Swinney was asking?) Marslen-Wilson calls his perspective temporal as opposed to functional. How are these different? How are processing written and spoken speech fundamentally different?
Spoken word recognition is incremental How do phoneme monitoring effects show this? Why are nonwords faster than words initially? What other hypothesis does this rule out? Non-words nonwords words Position of target in sequence reaction time from /t/ onset
Visual Word Recognition is not How does this graph shows that VWR is different from AWR? Non-words Position of target in sequence reaction time from /t/ onset Non words Words
The role of Context Clarification: Bottom up = information from lower level representations feeds into computations about higher level representations Top down = information from higher-level representations feeds into computations about lower level representations Question: are both directions of flow relevant for spoken word recognition?
Contextual vs. Sensory Inputs: Word Monitoring normal prose The quick brown fox jumped over the lazy dog. Syntactic prose The blue sensible timing fled through a reaction time. Random order Timing reaction the sensible blue a through fled time.
Contextual vs. Sensory Inputs normal prose RT correlates highly with word-position RT correlates poorly with word-duration Syntactic prose Random order RT correlates poorly with word-position RT correlates highly with word-duration
Time course of spoken word recognition Words are recognized after about 200 msec - earlier in many cases than acoustic-phonetic information is available. Word-monitoring task: 273 Shadowing task: 250 msec Gating procedure: similar estimate Acoustic-phonetic information narrows set of possibilities as the word unfolds.
What does this imply 1) Many word candidates activated early in word recognition 2) contextual information provides additional constraints How does this differ from a staged model of word recognition?
Evidence for early context effects Shadowing (Clarify) Normal prose faster than “syntactic prose” Word-monitoring Normal prose faster than “syntactic prose” Gating task Recognition time slower in isolation Restoration effects in shadowing task Only happened in normal prose and when mispronounced syllable(company - comsiny) was the second or third one.
Marslen-Wilson’s model Early sensory input activates multiple possible candidates (Bottom-up). The syntactic and semantic properties of these candidates are immediately assessed (top-down) MW is agnostic about time-course.
An optimal system (according to Marslen-Wilson) Pick out correct word as soon as disambiguating information is available Tressp (for tresspass) Distinguishes from tress In what sense is this system optimal? In what ways might this system not be optimal?
Evidence for immediate activation of all compatible words Nonword decision Zawritude (nonword at Zaw) Kwespidge (nonword at Kwesp) RT is dependent on the distance of the critical phoneme from word onset
Discussion Questions In the author's conclusions, he remarks that he doesn't comment on the autonomy vs interaction dispute over lexical access.. but he shows support for "a parallel system of analysis" which would seem to suggest autonomy, while his experimental research shows the interaction of context and sensory input in lexical recognition. It would seem that he has to support one of these views, and that interaction is most consistent with his findings. Which is it?
Discussion Questions How do we reconcile Marslen-Wilson & Swinney’s findings of early effects of context in lexical access? If the semantics and pragmatic context of a sentence help in identifyinga word, are Swinney type lexical ambiguities really a problem?