Visual Word Recognition
Download
Report
Transcript Visual Word Recognition
Visual Word
Recognition
Language Use and Understanding
Class 15: Van Orden
Announcements
Final exam: May 5, 9-12 am, in Meliora
366
Note: different room/time than scheduled
Coming up:
Wed: Castles & Coltheart (MR presenting)
Gernsbacher
Mon 3/22: Fodor (Anthony presenting)
Wed 3/24: discuss results
Mon: 3/29: Elman & McClelland (Mtan
Language Comprehension
Research Q: How do we extract meaning
from the linguistic signal?
Major issues
What information is used for processing
decisions?
When does each piece of information get
integrated? (What is the time course of
processing?)
Cognitive architecture - encapsulated or
not?
Interaction of different levels
Linguistic vs. nonlinguistic sources of
information
Levels of Processing
Comprehension
Sound / speech perception
Visual and Spoken Word recognition
Representations of words
Matching of signal with words
Word segmentation
Ambiguity
Sentence Processing
Discourse Processing
Role of Prosody
Word Recognition
How do we read words?
What processes/representations are
involved?
Models of reading
Direct access mode of reading:
orthography
Phonolgical mediation
Red --> sequence of letters is linked to a
stored memory of lexical item
Red --> /rEd/ --> sounds linked to a stored
memory of lexical item
Dual-access theories - both used, but what
is relation between them?
Word Recognition
How do we read words?
What processes/representations are
involved?
Written word
Orthographic information
Phonological information
Lexical Representation
-Phonological rep.
-Orthographic rep.
A separate question…
Does reading use “rules” or “distributed
representations”?
More on this next week…
Van Orden’s Q’s:
1. Does phonological mediation occur?
2. What is time course of orthographic
and phonological influences?
3. Does word identification include a
verification process?
Phonological Mediation
What is the previously existing evidence?
Lexical Decision task: is this a word?
DYME - longer to reject
DYTE - shorter time to reject
What are some criticisms of this?
Looks only at “no” responses
Perceived familiarity may affect results
Van Orden’s task
FLOWER
TULIP
ROWS
ROBS
Task: is the second word an exemplar of
the category?
Say “yes” or “now” and then read word
Task advantages
Sound similiarity should affect “yes”
responses (as well as “no” resposnes)
Based on a semantic judgment, not just
perceived stimulus familiarity
Task allows the use of word foils (as
opposed to nonword foils)
Allows comparison of sound-alike words
(ROWS) with spell-alike words (ROBS)
Experimental manipulations
Examplar vs. sound-alike vs. spell-alike
foils (TULIP vs. ROWS vs. ROBS)
Phonological mediation predicts more errors
for ROWS than ROBS
Degree of orthographic similarity of foils
MEET for MEAT (similar spelling)
ROWS for ROSE (less similar spelling)
Dual-route model predicts effects or
orthographic similarity separate from sound
similarity, predicts difference)
Experiment 1
Homophone vs. spelling control
ROWS vs. ROBS vs. ROSE
Degree of orthographic similarity of
homophone
MEET
ROWS
Results?
Exp. 1 Results
More false positives for homophone foils
than spelling controls
More false positives to similarly spelled
homosphones than less similarly spelled
homophones
Supports all dual-access theories, but
doesn’t distinguish between different
models
Experiment 2
Are phonological effects “early” or “late”?
Method: use pattern masking, which limits
processing to early effects
Prediction of Interactive Activation model:
phonological effects won’t be disrupted by
pattern masking, but orthographic effects
will
Time course of processing
Issue of general interest in processing
Which processes are “early” or “late”?
Which precede which?
Are there techniques to tap only the early
processes?
Emphasis on early processes as most
important
Word Superiority Effect
Reicher, 1969, Wheeler, 1970
Which letter did you see in the word
presented?
D
K
********
WORD
********
********
********
FASTEST
WORD
D
ORWD
D
K
********
********
********
********
Experiment 2
Same as Exp. 1, but with pattern masking
at 150-SOA or less
Results?
Homophony results replicated
No difference on basis of spelling similarity
What does this mean?
Phonological effects are “early”
There’s some “orthographically
determined process” that is affected by
pattern masking. (I.e., the spelling effect
disappeared).
Can we conclude that dualroute models are it?
In the introduction they say that the dual-route
models predict that there will be more false
positives for those words that are more similar
(in sound and spelling) to the target word. The
results of Experiment 1 showed a difference
between similarly spelled foils and less similar
spelled foils, yet the discussion of experiment
2 says this is not evidence of the dual-route
model. How is this? (Jessica DeSisto)
Answer: not EXCLUSIVE evidence
Exp. 3 - Bypass vs.
Verification
Verirfication model:
Stage 1: Lexical access by phonological
similarity
Stage 2: Orthographic verification
Bypass model: experienced readers
bypass phonology for orthography -predicts little/no phonological mediation for
frequent words
Exp. 3
Verification model: Predicts effect of
frequency of ROSE (category exemplar)
Bypass model: Predicts effect of
frequency of ROWS (stimulus)
Manipulate frequency of each
Results: Frequency effects only for
category exemplar, not stimulus
Conclusions of Exp. 3
Do not support Coltheart’s dual process
theory, which involves the bypass
hypothesis
Support the verification model
Overall results
Exp. 1: supports both phonological
mediation and orthographic effects
Consistent with Coltheart’s dual-process model
and Interactive Activation model
Exp. 2: pattern masking eliminates
orthogrpahic but not phonological effects
Inconsistent with Coltheart’s dual-process model
Consistent with Interactive Activation and
verification models
Exp. 3: Verification and not Bypass hyp.
supported