Word identification and reading Flashcards
what is the TRACE model McClelland & Elman (1986)
Phonemes are combinations of auditory features and time
what are the 3 layers of the TRACE model
Three layers:
Feature units
Phoneme units
Word units
what are the 3 levels of the Interactive activation model (Rumelhart & McClelland, 1982)
Levels: Feature detectors Primary visual cortex (V1) Letter detectors Word detectors
what are the connectionist models of word identification
(Early models focus mostly on reading)
Interactive activation model
The TRACE model
problems with the TRACE model
Unlike written language, time plays an important role in spoken language
This makes things really complex (as the illustration on the right shows)
Basically, detectors at the phoneme level are sensitive to the duration and order of inputs
strengths of the TRACE model
Predicts word-level top-down effects, such as phonemic restoration
Can account for almost all classical results on word recognition, including context effects
weaknesses of the TRACE model
Doesn’t have a mechanism to account for the importance of the first phoneme
Is it even falsifiable?
Can it account for anything?
maybe too general
The best models make strong positive and negative predictions
what is the Dual-route cascaded (DRC) model (Coltheart et al., 2001)
Model of reading aloud
incorporates interactive activation model
It’s just shown at a different level of abstraction
There still are nodes for each feature/letter/word/phoneme/meaning at each level
what are the 2 paths in the DRC model
Two paths to naming a word Direct route: print - lexicon Indirect route: print - grapheme-phoneme rules
Word identification during reading
Words don’t usually occur on their own on a page
How do readers manage to process 250 words per minute (that’s more than 4 words per second)?
Simple, unsatisfactory answer: they move their eyes
But maybe eye movements are the key to understanding reading?
how many letter can the fovea see
8
how many letters can the parafovea see
15
what are the 2 types of eye movement involved in reading
saccades
fixations
what are saccades
Extremely fast, “ballistic” (duration ca. 20-50 ms)
Need to be planned in advance
No visual information available (saccadic suppression)
what are fixations
Stationary periods in between saccades
Mean duration 200 – 250 ms, high variability
Visual information available
how do eyes move in reading
Readers typically fixate most words, most commonly just left of the centre (preferred viewing position, PVP). The optimal viewing position (OVP) is at the centre.
Some words are skipped
Others are refixated
Occasionally there are regressions to earlier words
the perceptual span
Moving window paradigm (McConkie & Rayner , 1975)
15 and 11 character windows to measure perceptual span
results of Moving window paradigm (McConkie & Rayner , 1975)
Note that window size is total window size
(e.g. 13 = 6 characters to the left of fixation and 6 to the right)
FL = Full line = no mask
Performance at 29 spaces is indistinguishable from full line
what is the English perceptual span
Reading is not affected if you take away letters more than 14 characters to the right and more than 3 characters to the left(in English)
The perceptual span is asymmetric in the direction of reading:
are perceptual spans language dependent
Different languages have different perceptual spans: Hebrew (read right to left): 7 left, 2 right Chinese (read left to right): 3 characters right, 1 character left
moving window in deaf readers Bélanger, Slattery, Mayberry, & Rayner (2013)
LSKD:
Low reading skill deaf readers = smaller reading window and perceptual span
SKD
Skilled deaf readers = have larger perceptual spans after 14
SKH
Skilled hearing readers = don’t show improvement in reading rate after 14
what is the Moving mask paradigm: Rayner & Bertera (1979)
masks certain number of characters to measure reading without fovea (1, 3, 9 character masks)
Eye tracking can be used to dynamically change the display as a subject moves his or her eyes: Moving window paradigm
results of moving mask paradigm for larger masks
Fixation duration increases sharply Saccade length decreases But then increases again for larger masks as people try to escape them Number of fixations decreases Reading rate decreases Naming accuracy decreases sharply
whats the main idea behind the E-Z Reader model
Word identification drives attention and eye movements
what is the E-Z reader model
Attention is allocated serially on each word from left to right
Word identification occurs in two stages, L1 and L2
Completing L1 triggers a saccade to the next word
Completing L2 triggers an attention shift to the next word
what are the 5 processing stages of the E-Z reader model
V M1 M2 L1 L2
what is the V stage of the EZ reader model
Early visual processing stage
what is the L1 stage of the EZ reader model
First stage of lexical processing
“Familiarity check”
When L1 completes, an eye movement programme to the next word is started
what is the L2 stage of the EZ reader model
Lexical access
Retrieval of meaning
When L2 completes, attention shifts to the subsequent word
another L1 begins on that word
what is the M1 stage of the EZ reader model
Labile stage of saccade preparation
Can be cancelled by a new saccade program triggered by L1 completion
Followed by M2
what is the M2 stage of the EZ reader model
Non-labile stage of saccade preparation
Cannot be cancelled by a new saccade program triggered by L1 completion
what is parafoveal processing
Parafoveal processing occurs whenever L2 is completed before M1 and M2.
On difficult words, L2 takes longer to complete, resulting in less parafoveal processing
This can explain spillover effects and reduced preview benefit
how does speed reading work
Is it possible to learn how to read faster?
Normal speed for a university student: 250 words per minute
Claimed speeds: 600-700 wpm (at least!)
Actually, speed readers are (at best) very good at skimming and making inferences.
No memory for details they haven’t fixated.
how many words can you process a minute
250 ( 4 per second)