View
220
Download
0
Category
Tags:
Preview:
Citation preview
Parafoveal Processing InfluencesWord Frequency & Predictability
Effects on Eye Movementsduring Reading
University of Glasgow(est. 1451)
GlasgowLanguageProcessing
Christopher Hand
Sébastien Miellet
Paddy O’Donnell
Graham Scott
Emotion Word Processing:Evidence from Eye Movements
Scott, O’Donnell, & Sereno
Frequency & Predictability
• Word frequency– High-frequency (HF) words are read more quickly
than low-frequency (LF) words.– A word frequency effect (HF<LF) is used as a marker
(index) of successful word recognition (lexical access).
• Contextual Predictability– Given a prior context, words that are highly predictable
(HP) from context are read more quickly than those that are less predictable (LP).
– The temporal locus of predictability effects, lexical (interactive) or post-lexical (modular), is a matter of continued debate.
Freq X Pred: Early RT Studies
Stanovich & West (1981):Naming
700
750
800
850
900
950
1000
Incongruous Congruous
Predictability
RT
(m
s)
LF
HF
Schuberth & Eimas (1977):Lexical Decision
500
550
600
650
700
Incongruous Congruous
Predictability
RT
(m
s)
LF
HF
Additive Interactiveor
TargetCondition Context HF LFCongruous The skier was buried in the…Incongruous The bodyguard drove the…
snowsnow avalancheavalanche
TargetCondition Context HF LFBiasing Flying to its nest was a…Neutral To our surprise we saw a…
bird bird hawkhawk
1
1.2
1.4
1.6
1.8
2
Neutral Biasing
Predictability
Un
sig
ned
Vo
ltag
e (μ
V)
LF
HF
Freq X Pred: ERP Study
Sereno, Brewer, O’Donnell (2003)
N1 component: 132-192 ms
Freq X Pred: Eye MovementsAltarriba, Kroll, Sholl, & Rayner (1996)Lavinge, Vitu, & d’Ydewalle (2000)Rayner, Binder, Ashby, & Pollatsek (2001)- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -Rayner, Ashby, Pollatsek, & Reichle (2004)Miellet, Sparrow, & Sereno (in press)
Inhoff (1984)
500
550
600
650
700
Low High
Predictability
Fix
ati
on
Du
rati
on
(m
s)
LF
HF500
550
600
650
700
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Freq X Pred: Eye MovementsAltarriba, Kroll, Sholl, & Rayner (1996)Lavinge, Vitu, & d’Ydewalle (2000)Rayner, Binder, Ashby, & Pollatsek (2001)- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -Rayner, Ashby, Pollatsek, & Reichle (2004)Miellet, Sparrow, & Sereno (in press)
Inhoff (1984)
500
550
600
650
700
Low High
Predictability
Fix
ati
on
Du
rati
on
(m
s)
LF
HF500
550
600
650
700
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Freq X Pred: Eye MovementsAltarriba, Kroll, Sholl, & Rayner (1996)Lavigne, Vitu, & d’Ydewalle (2000)Rayner, Binder, Ashby, & Pollatsek (2001)- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -Rayner, Ashby, Pollatsek, & Reichle (2004)Miellet, Sparrow, & Sereno (in press)
Inhoff (1984)
500
550
600
650
700
Low High
Predictability
Fix
ati
on
Du
rati
on
(m
s)
LF
HF500
550
600
650
700
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
conducteden
français
Bugs Bunny eats lots of carrotscarrots to stay healthy.
June Cleaver always serves meat and potatoespotatoes for dinner.
LF
HF
High Predictability
June Cleaver always serves meat and carrotscarrots for dinner.
Bugs Bunny eats lots of potatoespotatoes to stay healthy.
LF
HF
Low Predictability
Rayner et al. (2004)
Rayner et al. (2004)
Limitations Remedies
Items per condition:8
Targets embedded in:single sentence
Length of context (# pre-target words):7.7 words
22
15.5
2nd of 2sentences
Bugs Bunny eats lots of carrotscarrots to stay healthy.
June Cleaver always serves meat and potatoespotatoes for dinner.
June Cleaver always serves meat and carrotscarrots for dinner.
Bugs Bunny eats lots of potatoespotatoes to stay healthy.
LF
LF
HF
HF
High Predictability
Low Predictability
Although a rugby player, Clive struggled through the crowdat the bar carrying glasses of lagerlager and bags of crisps.
Gillian was on the last mile of the women’s marathon. Shegrabbed a bottle of waterwater from a spectator and drank it.
Gillian was on the last mile of the women’s marathon. Shegrabbed a bottle of lagerlager from a spectator and drank it.
Although a rugby player, Clive struggled through the crowdat the bar carrying glasses of waterwater and bags of crisps.
LF
LF
HF
HF
High Predictability
Low Predictability
Ingrid’s boiler had suddenly broken down. Fortunately, herneighbour’s father was a plumberplumber and would be able to help.
Callum was having trouble with his homework. He asked hisuncle who was a teacherteacher to help him with the assignment.
Callum was having trouble with his homework. He asked hisuncle who was a plumberplumber to help him with the assignment.
Ingrid’s boiler had suddenly broken down. Fortunately, herneighbour’s father was a teacherteacher and would be able to help.
LF
LF
HF
HF
High Predictability
Low Predictability
Method
• Materials & Design:Frequency (HF,LF) x Predictability (HP,LP)
HF & LF targets matched pair-wise on word length (mean = 5.84 letters; range: 5-8 letters)
HF LF
HP LP HP LP
Frequency (BNC, per million) 145 145 4 4
Predictability (1-7) 6.19 4.07 6.11 3.69Cloze probability .57 .02 .50 .01
# of items 22 22 22 22
Method
• Materials & Design:Frequency (HF,LF) x Predictability (HP,LP)
HF & LF targets matched pair-wise on word length (mean = 5.84 letters; range: 5-8 letters)
HF LF
HP LP HP LP
Frequency (BNC, per million) 145 145 4 4
Predictability (1-7) 6.19 4.07 6.11 3.69Cloze probability .57 .02 .50 .01
# of items 22 22 22 22
Method
• Participants: 64 (mean age = 22; #F = 47)
• Apparatus:– Dual-Purkinje Eyetracker (Gen 5.5)– 4 characters ≈ 1o of visual angle
• Procedure:– 88 experimental passages– Yes/No comprehension questions on half the trials
WOW!
Results: Fixation Time Measures
• Early– First fixation duration (FFD)– Single fixation duration (SFD)– Gaze duration (GD)
• Later– Next forward-going fixation after target (‘spillover’)– Total Fixation Time (TT)
reject 4%skip 21%
1 fix 63%
2+ fix 12%
Single Fixation Duration: Freq x Pred
250
260
270
280
290
300
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms
)
LF
HF
26 ms
9 ms
First Fixation Duration
250
270
290
310
330
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Gaze Duration
250
270
290
310
330
Low High
Predictability
Fix
ati
on
Du
rati
on
(m
s)
LF
HF
Just then,a smug grincomes overhis face…
Ha! Told you so!
• In reading, words are initially processed parafoveally before they are directly foveated.
• Prior research indicates that both the frequency and predictability of the parafoveal word can influence its subsequent fixation duration.
• Specifically,– HF parafoveal words are subsequently fixated for less
time than LF ones (Inhoff & Rayner, 1986).– HP parafoveal words are subsequently fixated for less
time than LP ones (Balota, Pollatsek, & Rayner, 1985).
Parafoveal Pre-Processing
The sore on Tam-Tam’s rump was swollen.
invisibleboundary
* ** *
* * *
Boundary Technique
The sore on Tam-Tam’s tnam was swollen.
invalid parafoveal preview
…grabbed a bottle of waterwater from a… 1-3
…grabbed a bottle of waterwater from a… 7-9
…grabbed a bottle of waterwater from a… 4-6
…grabbed a bottle of waterwater from a… 10-12
…grabbed a bottle of waterwater from a… 13+
Launch Distance
(Poorly) Simulated Perceptual Span
…grabbed a bottle of waterwater from a… 10-12
…grabbed a bottle of waterwater from a… 4-6
…grabbed a bottle of waterwater from a… 7-9
…grabbed a bottle of waterwater from a… 1-3
…grabbed a bottle of waterwater from a… 13+
%Data Contingent on Launch Distance
1-3 (16%)
4-6 (24%)
7-9 (21%)
10-12 (10%)
13+ (5%)
skips (21%)
rejected(4%)
(15%)
(21%)
(17%)
(6%)
(4%)
#items#subjects #conditions per cond # data points
64 x 4 x 22 = 5632
# Data Points by Condition: SFD
0
50
100
150
200
250
300
350
13+ 10-12 7-9 4-6 1-3
Launch Distance (# letters)
# D
ata
Po
ints LF-LP
LF-HP
HF-LP
HF-HP
SFD: Freq x Pred x Launch
200
220
240
260
280
300
1 2 3 4 5 6 7 8
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Low HighPredictability
7-9
Low HighPredictability
4-6
Low HighPredictability
1-3
Launch Distance from Target (# letters)
• RT studies, in general, find that word frequency and contextual predictability interact.
• Eye movement studies, measuring target word fixation time in normal reading, find that frequency and predictability are additive.
• When the amount of parafoveal preview is manipulated (post-hoc via launch distance), however, different patterns of data emerge.
Summary of Results
Summary of Results
• Specifically,
• Fixation duration vs. RT?– Far condition ≈ RT physically, but not behaviorally– Near condition ≈ RT behaviorally, but not physically
Launch distance Freq Pred FxP
Far (7-9)
Mid (4-6) HFHF
Near (1-3) HFHF,LFLF HFHF<LFLF
• Apparent additive effect of Freq & Pred in reading is comprised of opposing interactive effects.
Conclusions
SFD: Freq x Pred X Launch
200
220
240
260
280
300
1 2 3 4 5 6 7 8F
ixa
tio
n D
ura
tio
n (
ms
)
LF
HF
LP HP
7-9
Launch Distance from Target (# letters)
Single Fixation Duration: Freq x Pred
250
260
270
280
290
300
Low High
Predictability
Fix
ati
on
Du
rati
on
(m
s)
LF
HF
LP HP
4-6
LP HP
1-3
• Explanation 1: Frequency First
Conclusions
Single Fixation Duration: Freq x Pred
250
260
270
280
290
300
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
• Explanation 2: Floors & Ceilings
Conclusions
SFD: Freq x Pred X Launch
200
220
240
260
280
300
1 2 3 4 5 6 7 8
Fix
atio
n D
ura
tio
n (
ms
)
LF
HF
LP HP
7-9
Launch Distance from Target (# letters)
LP HP
4-6
LP HP
1-3
• Launch Site
Conclusions
Used as a tool to capturethe temporal dynamics ofparafoveal processing.
–
Demonstrated sensitivityto the temporal & spatialcontingencies intrinsic toreading.
–
Provides a more fluid ortransitional account ofconcurrent oculomotorand linguistic processing.
–
Thank y’all
• “Parafoveal-on-foveal” effects (oh no!):– When the ease or difficulty in processing a parafoveal
target initially manifests itself on the current, pre-target fixation.
Effects Before the Target?
…grabbed a bottle of waterwater from a…
location of current,pre-target fixation
Fixation Before Target: Freq x Pred
240
250
260
270
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
4 ms(marg.)
6 ms (p<.01)
Fixation Before Target: Launch Dist.
200
220
240
260
280
300
1 2 3 4 5 6 7 8
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Low HighPredictability
7-9
Low HighPredictability
4-6
Low HighPredictability
1-3
Launch Distance from Target (# letters)
FFD: Freq x Pred X Launch
200
220
240
260
280
300
320
1 2 3 4 5 6 7 8
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Low HighPredictability
7-9
Low HighPredictability
4-6
Low HighPredictability
1-3
Launch Distance from Target (# letters)
SFD: Freq x Pred X Launch
200
220
240
260
280
300
320
1 2 3 4 5 6 7 8
Fix
ati
on
Du
rati
on
(m
s)
LF
HF
Low HighPredictability
7-9
Low HighPredictability
4-6
Low HighPredictability
1-3
Launch Distance from Target (# letters)
GD: Freq x Pred X Launch
200
220
240
260
280
300
320
1 2 3 4 5 6 7 8
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Low HighPredictability
7-9
Low HighPredictability
4-6
Low HighPredictability
1-3
Launch Distance from Target (# letters)
TT: Freq x Pred
290
310
330
350
370
390
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms
)
LF
HF
Effects: F1 F2
Freq *** *** Pred *** *** FxP marg n.s.
Probability of Skipping
0.15
0.20
0.25
0.30
Low High
Predictability
Pro
bab
ility
LF
HF
Next Fwd Fix: Freq x Pred
250
260
270
280
290
300
Low High
Predictability
Fix
atio
n D
ura
tio
n (
ms)
LF
HF
Effects: F1 F2
Freq n.s. n.s. Pred marg. n.s. FxP ** marg.
Recommended