輪読：単語認知１・前半（関西学院大学・金澤）

Psychology of Reading (2nd ed.)
Chap. 3. Word Perception I:
Some Basic Issues and Methods
Graduate Student, Kwansei Gakuin University
KANAZAWA, Yu(金澤佑)
yu-kanazawa@kwansei.ac.jp
pp. 49-71 (-88)
JACET Study Group of Reading
@ Umeda Campus, Kwansei Gakuin University
May 25, 2014. 1:40 pm-

Content
① Introduction
② How long does it take to identify a word?
③ Is word processing automatic?
④ How does the processing of words relate to the
processing of letters?
⑤ The role of sound in the encoding of words
⑥ Processing simple and complex words
⑦ Cross-language studies of word perception
⑧ A final issue
⑨ Summary and conclusions
2

① Introduction
• Word identification/ recognition/ access/
processing/ encoding/decoding:
• The central issue to understanding reading
• A major focus of research in cognitive psychology in the
last 40 years
• The focus of this chapter:
• Word perception of (a) skilled readers (b) of English (c)
reading isolated printed words
3

Introduction: 6 general questions
1. Is word recognition all that needs to be learned?
2. Is identifying words effortful and the rest of the
reading process automatic?
3. Are words identified by accessing the sound and then
the meaning?
4. Are letters in words processed serially or are words
processed as wholes?
5. Do skilled readers learn to apply something like the
rules of spelling in a fluent way or do they learn
specific associations between visual patterns and the
sound and/or meaning of the word?
6. Does context radically affect the process of word
identification?
Chap. 10, 11
Chap. 5
4
①

Is word recognition all that
needs to be learned?
• Central problem of reading = recognizing the
printed word? :developmental perspective
• To be able to access the words of the spoken language
from the written representation = goal?
• Note!
• The process of how children learn to read their native
language
・・・is not necessarily equal to・・・
• the process of skilled reading
5
①

Does context radically affect the
process of word identification?
• No, it doesn’t.
• No crucial difference of processing between the
words in isolation and the words in context.
• It is ecologically unrealistic to assume completely
different organization for word recognition in
isolation and word recognition in context.
6
①

Introduction: what we know about
word identification in skilled readers
• Straightforward process. Approx. 250 ms./word.
Higher-order processes only when reading is difficult
• Letters → sounds → meaning
• plays a role in identification of words
• plays a part in reading process after word identification
• Common short words are processed in parallel.
• NOT that words are learned as visual templates (gestalts)
• Even longer words are NOT processed letter by letter.
• No important difference of processing between the
words in isolation and the words in context.
7
①

②
• How long does it take to identify a word?
（単語の同定にかかる時間は？）
i. Response time methods （反応時間での測定法）
ii. Brief presentation methods （短時間の提示方法）
iii. Estimates from reading text （文の読みからの推定）
iv. Physiological methods （生理学上の方法）
8

i. Response time methods, or rather,
“response execution time” method?
• Word identification + some “excess baggage” = RT
• Example: the case of a naming task (p. 51, a-d)
• 400-500 ms. to name common words
• Note it is uncertain whether semantic access is done or not.
• Categorization task is the solution?
• No. Word identification + mental judgment = RT
• 700 ms. to make categorical judgment
• Simpler and easier optimum so far: lexical decision task
• Approx. 500 ms. to make decision
• Doubt of semantic access still remains
9
②

ii. Brief presentation methods
• Control the duration of presentation! (e.g. 60 ms.)
• A problem:
• Iconic memory persists at least about 250 ms.
• A solution: masking
• Another problems:
• Transfer to visual short-term buffer? (unconscious process)
• Neural transmission time (80-100 ms.) not counted
• The eye → optic nerve → secondary visual cortex (Petersen et al., 1988)
• A simple solution: Just add them.
• Word processing = 60 + 80-100 = 140-160 (ms.)
• Supporting data: Cohen et al., (2000)
• Timing of left fusiform gyrus （紡錘状回） being activated: 180-200 ms.
10
②

11
http://faceagnosia.files.wordpress.com/2011/03/face-processing.jpg
②

iii. Estimates from reading text
• Reading rate of a typical college student
• 300 WPM, 5 words per sec.
• → 200 ms. to read a word (consistent score!)
• Note: Reading is more than word identification.”
• Also, not every word may be identified.
12
②

iv. Physiological methods
• More direct method: examine the brain itself!
• ERPs (event-related potentials)
• Temporal resolution ○, Spatial resolution ×
• Measures gross electrical activity in the brain.
• Sereno, Rayner, & Posner (1998) c.f. Figure 3.1 (p. 55)
• First peak: 50-100ms. after the onset of the word
• Noticing “something has happened”
• Negative component: 150ms. after the onset of the word
• Word identification (difference between word frequency detected)
• Second peak: 300-400ms. after onset of the word
• Making decision about which response to select
13
sum Around 150 ms. to access the meaning of a printed word
②

Final issue in introduction:
speed of word identification & word frequency
• New corpora to measure word frequency
• HAL corpus of ELP (Balota et al., 2007): US English
• CELEX corpus (Baayen et al., 1995): UK English
• Temporal difference between high-frequency
word and low-frequency word
• Lexical Decision Task: 100 ms.
• Naming Task: 30 ms.
• Fixation Time in Reading: 30 ms.
14
②

③
• Is word processing automatic? （単語処
理は自動的か？）
• Is “identification” of words unconscious?
（単語同定は無意識か？）
• Is intention to process a word important?
（単語処理の意図は重要？）
• Does word identification take processing
capacity? （単語同定は処理容量を消費する
か？）
15

Automaticity of word processing
• Identifying written words:
• Unnatural, effortful, less fluent for beginning readers
• Natural, effortless, fluent for adult readers
• Word identification is…
• hard and context is used as an aid (Goodman, 1970).
• automatic and context is not a crucial mediator.
• Criteria of “Automaticity” (Posner & Snyder, 1975)
• (a) unawareness of the process
• (b) independence from conscious control/intention
• (c) no processing capacity usage
?
16
③

Is identification of words
unconscious? – Yes, it can be.
• How to test it:
Oral protocol? –Bad idea
Recognition test? –Chance level (Balota, 1983)
Semantic priming (Meyer & Schvaneveldt, 1971) ? –the best so far
• E.g. [dog] → [cat] vs. [fan] → [cat]
• Semantic similarity between a prime word and the target word
hastens “yes” response by 30-50 ms. in LDT, smaller yet similar
effect in naming task.
• Semantic priming effect has been replicated many times.
• The size of semantic priming effect is unsusceptible to awareness.
• The meaning of a word can be “looked up” by its visual
representation without the conscious experience of perceiving the
word.
17
③

Is intention to process a word
important? -Not really.
• Intention to process a word ≠ awareness of processing it
• Word meaning can be extracted even when it is unwanted
• Stroop effect aka “Can’t avoid processing!” (Stroop, 1995)
• Color naming task: red is slower (200ms.) to name than .
• Unrelated word (e.g. ant ) is faster than red and slower than .
• Word encoding not completely automatic
• spatial alteration of a prime and the target = No priming effect
• Repetition priming research of Besner, Risko, and Sklair (2005)
18
③

Does word identification take
processing capacity? –Probably.
• Processing capacity = attentional resource
• e.g. Breathing = no attentional resource needed
• e.g. Conversation = much attentional resource needed
• Process requiring no processing capacity does not (a) slow
down or (b) interfere with other processes.
• Note: Interference between multiple tasks can be the result of
either competition for resource or competing response.
• How to test limited capacity: see whether two things can be
done at the same time as well as one.
• A task: Visual Search (Karlin & Bower, 1976)
• “Is there an animal name present on the screen?”
• Result: RT to category decision is longer about 200 ms. extra per
additional word on the screen
• The result attributable not only to identification process but also
to categorization process
19
③

④
• How does the processing of words relate to
the processing of letters? （単語処理はどのよ
うに文字処理と関連するか？）
• Letters in words are not processed sequentially
（単語における文字は逐次処理されない）
• Words are not visual templates （単語は視覚的テン
プレートではない）
• A relatively simple model of word perception （単
語知覚の比較的シンプルなモデル）
• The role of near misses in word identification （単
語同定におけるニアミスの役割）
21

words and l-e-t-t-e-r-s
• words: physical entities between the spaces which work
as units of meaning/function
• Smith (1971)
• Skilled readers identify English words as a visual pattern
• Letters plays no role in word identification
• Gough (1972)
• Words are encoded letter-by-letter serially from left to right.
• Despite the criticism, processing time data supports the view.
22
④
parallel
sequential
somewhere in between

Negation of the “sequential”
extreme
• Random letters are processed 10 ms. per letter
(Sperling, 1963).
• Recognition of letters should take less time than
recognition of words, but the contrary was true
(Cattel, 1886).
• Supported by replication study with refined procedure by
Reicher (1969).
• The response accuracy:
• Word condition ＞ Letter condition ≒ Nonword condition
• Word superiority effect
• Letters in words are actually identified more accurately
than letters in isolation
23

Negation of the “parallel”
extreme
• Is letter recognition system independent from
word recognition system? No.
• A counterexample: “pseudoword superiority effect”
• Pronounceable nonwords (e.g. MARD) as well as words
are processed more accurately than letters in
isolation.
• wOrD idEntifiCaTiOn pRoceEds laRgeLy tHroUGh
cASe aNd font-inDepeNdeNt abSTrAct lEttEr
idEnTitiEs (Coltheart, 1981, etc.).
• Word shape is not a crucial cue for word
identification.
24

Simple model of word perception
• Computer simulation of parallel encoding model
• Paap, Newsome, McDonald, & Schvaneveldt (1982)
• Different detectors in different stages
• Feature det. → Letter det. → Word det. (Case & Font Det.)
• Detector = neuron
• Activated when input reached the threshold
• One input can activate multiple detectors.
• Activation is gradual, not all-or-nothing, and the detector most strongly
activated is selected.
• Explanation of “word superiority effect”
• Letters in isolation: identified by only “Letter det.”
• weaker activation (say, 50%)
• Letters in words: identified by both “Letter det.” & “Word det.”
• → higher level of activation (80%)→ more accurate letter recognition
25
Figure 3.3

Parallel Letter Recognition Model.
Paap, Newsome, McDonald, &
Schvaneveldt (1982)
26

Interactive Activation Model.
McClelland & Rumelhart (1981) 27
Feature Det.
Letter Det.
Word Det.

Simple model of word
perception (cont.)
• What in the Paap et al. model correspond to word
perception and letter perception?
• The simplest possibility:
• Perception occurs if the excitation in any detector exceeds a
certain threshold (say, 75%).
→ Word perception may preceed letter perception!
e.g. unawareness of mispelling
• What now for the “pseudoword superiority effect” ?
• Paap et al. is applicable.
• The mechanism: “When a letter string appears, it not only
excites the identical lexical entry but also ‘neighbors’ of it.”
• E.g. pseudoword MARD has the following neighbors:
• ward, mark, mare, maid, etc.
28
Figure 3.4

The role of near misses in
word identification
• Uncertain aspects of word identification
• Orthographic level
• The effect of the orthographic neighbors
• The number & the frequency of them
• To time and accuracy
• The effect of errors in letter order
• Phonological level
• The effect of phoneme
29

NEIGHBORS, inhibitory and facilitative
in word-encoding
• An orthographic neighbor
• a letter string of equal length to a target word but with one
letter substituted
• Deletion neighbors (e.g. wad) & addition neighbors (e.g. waned)
not included in the narrow sense
• Having more neighbor & Neighbor being higher-frequency
• lengthening the verification stage? → inhibitory
• active feedback to the letters? → facilitative
• Both effects contend, whichever ending up being overt.
• e.g. Having single neighbor which is higher-freq. than the word
• facilitation ＜ inhibition
• e.g. Having many neighbor which is lower-freq. than the word
• Facilitation ＞ inhibition
30

The role of letter position
• Measurement of neighbor effect: Lexical Decision task
• Words with more neighbors → shorter RT
• Words with more higher-freq. neighbors → longer RT
• Related issue: “slot coding” assumption being wrong
• Rihgt lettres in the worng oredr can be prcoesesd easily!
• Two alternate types of models proposed
• “Absolute coding” model (Gomez, Ratcliff. & Perea, 2008)
• slot coding + perceptual uncertainty
• experi[m]ent, [m] can be coded as Position 7, 6, 8, 5, 9.
• “Relative coding” model (Mozer, 1983)
• Order information comes from encoding short sequences of
letter position
31
http://w
ww.mrc-
cbu.cam.
ac.uk/pe
ople/ma
tt.davis/
Cmabrid
ge/

Transposed letters (TL)
• Possible schemes in “relative coding” model
• Adjacent bigram in a word
• _e, ex, xp, pe, er, ri, im, me, en, nt, t_
• Trigram
• _ex, exp, xpe, per, eri, rim, ime, men, ent, nt_
• Open bigram pairs (SERIOL model; Whitney, 2001)
• p_m ∈ experiment
• The effect of transposed letters (TL) on word encoding
• Absolute coding view: clam is more similar to cram than to calm.
• Difference of two letters vs. one letters in their absolute order
• Relative coding view: clam is more similar to calm than to cram.
• Calm is a transposed word of clam and shares the consisting letters.
32

Transposed letters (TL; cont.)
• Transposed letter neighbors cause interference.
• Experimental paradigm 1: LDT nonword data
• TL condition (transposition; judge & jugde)
vs.
• RL condition (replacement; judge & jupte)
• TL condition took more time and was more erroneous (Perea et al., 2005).
• Experimental paradigm 2: Masked priming
• Foster & Davis (1984) http://www.u.arizona.edu/~kforster/priming/masked_priming_demo.htm
• Masked prime is presented 40-60 ms. with fixed location.
• RT: TL prime-target pairs ＞ RL prime-target pairs
• supporting relative coding view
• Difference between vowels and consonants
• Word identification: Consonants are more important.
• Position-wise: Vowels are more important.
• Blurred line between orthographic effect and phonological effect
33

The reference book
• Pollatsek, A. (2012). Word perception I: Some
basic issues and methods. In, K. Rayner, A.
Pollatsek, J. Ashby, and C. Clifton Jr.
Psychology of reading: 2nd edition (pp. 49-88).
New York: Psychology Press.
34

輪読：単語認知１・前半（関西学院大学・金澤）

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to 輪読：単語認知１・前半（関西学院大学・金澤） (20)

Recently uploaded (20)