Word Recognition Flashcards
Lexical Organization: Semantic Networks
- Words are interconnected in terms of meaning (semantic networks)
- Activating a word “spreads activation” to other semantically-related words
eg: “cake” — sweet, cupcake, sugar, dessert
Data from two methods:
1) lexical decision tasks
2) visual world eyetracking
METHOD 1: Lexical Decision Task
A task requiring participants to identify whether a stimulus target is a real word (eg. “lion”) or not (eg. “plouk”)
A semantically-related word (prime) aids in retrieving another word (the target)
TIGER LION Prime Target
Semantic Priming
The ability to recognize a target (and identify it as a word) is affected by a prime that shares its meaning.
Facilitation
Processes that “speed up” lexical retrieval
Spreading Activation
- Words are interconnected in terms of meaning (semantic web/network)
- Activating a word “spreads activation” to other semantically related words
Mediated Semantic Priming
A prime word (eg. lion) speeds up responses to a target word (eg. tiger) due to an indirect connection via some other intervening word (eg. jungle)
Decay Function
Rate at which information fades in memory, such that info that has become activated gradually returns to a baseline level of activation
- decay function reduces the transfer of activation between unrelated nodes (connectivist view)
- eg. cake —> dessert —> flour … (wheat) —> (mill) …
- connection is reduced with time
METHOD 2: Visual World Eyetracking
- Records eye movement gazes/saccades in real-time
- Eye movements are influenced by thought processes and presented stimuli
- Location of gazes = function of stimulus characteristics
Lexical Ambiguity
Homophones: same sound, different meaning
- eg. bred, bread
- eg. missed, mist
- eg. “the performer took a deep bow”; “it’s difficult to hunt with a bow and arrow”
Synonyms: same meaning, different sound (polysemous words)
- eg. small, little
- eg. “she’s got a run in her stockings”; “let’s run through the various options”
How do we retrieve the correct meaning of polysemous words?
Two proposed models for determining correct meaning of polysemous words:
1) Access and Selection Model
2) Early Selection Model
Access and Selection Model
A&S Model supports that context will bias the interpretation of an ambiguous word, so that only the intended meaning is accessed
- a “bottom-up” process: activation begins at the phonological level and spreads upwards
- processing sounds, words, and meaning in order of sentence (as it is heard)
- all potential meaning are equally active until ruled out
- efficient for unfamiliar words/meanings
Access and Selection Model (EXAMPLE)
“Sameera got her paycheque from the bank”
Activation:
Step 1) Sound : b æ ŋ k
Step 2) Words : bank (a) OR bank (b)
Step 3) Meaning : (a) “financial institution” OR (b) “land by water’s edge”
Step 4) Selection : incorrect meaning is inhibited
Early Selection Model
Early Selection posits that stimuli are filtered, or selected to be attended to, at an early stage during processing
- “top-down” process—interpretation is driven by previous knowledge
- activation begins at semantic level and spreads down, and is then reactivated from the bottom phoneme level, reconfirming meaning
Early Selection Model (EXAMPLE)
“Sameera got her paycheque from the bank”
Activation:
Step 1) Upon hearing “Sameera got her paycheque”, meaning is revealed — context clue “paycheque” begins to activate meaning
Step 2) Upon hearing “bank”, meaning is confirmed
Step 3) Activation is reactivated at the phonemic level by sensory input, spreading upwards and reconfirming meaning
Access and Selection Model EVIDENCE (*)
Cross-Modal Lexical Decision (Swinney 1979)
Research Q: looking at how we process linguistic stimuli (access and selection VS early selection)
Method:
- 2 modalities (audio & visual)
- participants listened to audio of a sentence; eg: “the man was not surprised when he found spiders, roaches and other bugs”
- bug = prime; manipulation; polysemous/ambiguous word
- visual target (word) was then presented either immediately after target word or several syllables after
- ant (context—related to intended meaning of prime)
- spy (related to unintended meaning)
- sew (unreleated)
Predictions:
Unambiguous prime (insects)
- related | ANT | reaction time: faster
- inappropriate | SPY | reaction time: slower
- unrelated (control) | SEW | reaction time: slower
Ambiguous prime (bugs) (multiple meanings)
- related | ANT | reaction time: faster
- inappropriate | SPY | reaction time: faster
- unrelated | SEW | reaction time: slower
Results:
Unambiguous prime (insects); if the visual target immediately appears after the prime
- related | ANT | reaction time: faster
- inappropriate | SPY | reaction time: slower
- unrelated (control) | SEW | reaction time: slower
Ambiguous prime (bugs); visual target presented immediately
- related | ANT | reaction time: faster
- inappropriate | SPY | reaction time: faster
- unrelated | SEW | reaction time: slower
- first, both meanings are accessed
Ambiguous prime (bugs); visual target presented a few hundred ms later
- related | ANT | reaction time: faster
- inappropriate | SPY | reaction time: slower
- unrelated | SEW | reaction time: slower
- then, the appropriate meaning is selected (the inappropriate meaning is discarded
SUPPORTS ACCESS & SELECTION MODEL (BOTTOM-UP)
Lexical Organization: Phonological Networks
Words are also connected by sound
- phonological neighbours: words that differ by only one sound
- many neighbours: cat — (m)at, (b)at, (p)at, ca(p), ca(b), ca(n) c(u)t, c(o)t …
- few neighbours: wolf — wo(ol), wo(of), (g)olf
- slower recognition for words with MORE neighbours than those with less — more neighbours means the web is more complex and activation is slowed
Onset Competitors
Words that begin with the same set of sounds
eg bee
- beetle
- beeper
- beefy
- beetroot
- beehive …
Cohort Activation Model
Marlsen-Wilson (1987)
We partially begin to process the word as soon as we hear the sounds, activating onset competitors
- b… bee… beet…
Incompatible options are then discarded as more sound input is processed
- b… (b)ag, (b)an, (b)olt, (b)ees, (b)eehive, (b)eetle…
- bee… bag, ban, bolt, (bee)s, (bee)hive, (bee)tle…
- beet… bag, ban, bolt, bees, beehive, (beet)le…
Cohort Activation Model EVIDENCE
(Allopenna et al., 1998)
Method: visual world eyetracking
- task: a word (ex. beaker) is played
- participants need to click on the corresponding visual image
- stimuli:
- target: beaker
- onset competitor: beetle
- unrelated: carriage
Results:
- people look at target and onset competitor at 200-400 ms
- evidence that people are activating onset competitor as well
- supports cohort model; we don’t wait until the end of stimuli to start processing
Cohort Activation Model: Limitations
- we won’t always hear the word correctly
- onset vs offset: do we only care about left edges of words? eg bea ker
- segmentation Issues: no breaks between words
- noise
- disruptions in speech
TRACE Model
McClelland & Elian (1986)
- connectionist model of speech perception
- by segmenting the individual sounds, phonemes can be determined from spoken words
- provides a potential explanation for cohort model limitations
Continuous Mapping: we continuously process sounds as we hear them
- sounds in the middle and ends are also activated
Rhyme Competitors: words that share the same sound endings
- eg beaker
- spea(ker), coo(ker), tal(ker), mar(ker)…
TRACE Model EVIDENCE
(Allopenna et al., 1998)
- OG study with manipulation to assess the activation of rhyme competitors
Method: visual world eyetracking
- task: a word (ex. beaker) is played
- participants need to click on the corresponding visual image
- stimuli:
- target: beaker
- onset competitor: beetle
- rhyme competitor: speaker
- unrelated: carriage
Results:
- looks to target (beaker) throughout
- effect of rhyme competitor (speaker)
- lesser extent of activating rhyme competitors (speaker) than onset competitors (beetle)
- supports TRACE model
Semantic Activation
(Yee & Sedivy, 2006)
Research Q: do words that are semantically-related to target/onset competitor also light up?
Method: visual world eyetracking
- task: a target word (ex. hammock) is played
- participants need to click on the corresponding visual image
- stimuli:
- target: hammock
- semantic onset competitor: nail
- unrelated: tissue, grasshopper
- when we hear the segments of ‘hamm…’ would there be looks to the
- hammock?
- nail? (“hamm…er”)
Results:
- participants look more to nail (than to unrelated) when hearing hammock, but effect is short-lived
- the reveals transient semantic effects for parts of words
Recognizing Written Words
Some writing systems encode meaning, others encode sound
Logographic Writing Systems
Symbols map directly to a concept/meaning
- eg Chinese Hanzi, Japanese Kanji, Korean Hanza
- some phonetics involved (tonal languages)
Syllabic Writing Systems
Symbols map directly to a syllable
- eg Japanese hiragana/katakana, Cherokee
Alphabetic Writing Systems
Each symbol maps directly to a sound
- eg English, Spanish, Greek, Russian …
Shallow (Transparent) Orthography
Correspondances between letters (graphemes) and sounds (phonemes) are one-to-one—each symbol maps to a sound
- eg Serbian, Italian, Spanish
Words are pronounced as they appear
Deep Orthography
Correspondances between letters and sounds are not one-to-one—symbols do not necessarily map to sounds (lots of irregularities)
- eg French, Hebrew
Dual Route Model of Reading
DIRECT: visual symbols directly map to word recognition—works for high frequency words
VISUAL WORD → perceive visual symbol → immediate retrieval → recognize as a word → access the meaning
ASSEMBLED: symbols are “sounded out” prior to accessing meaning—language learning/novel words
VISUAL WORD → perceive visual symbol → sound it out → recognize as a word → access the meaning
Cross Modal Lexical Decision
Research Q: does auditory stimulus have impact on visual stimulus?
Method:
- participants hear a word (prime)
- word is flashed on screen (target)
- participants must identify whether the target is a word or not
- prime: beaker
- target: glass (sem. related to beaker), insect (sem, related to beetle), stereo (sem. related to speaker)
- manipulation: timing of target presentation for [bea][ker]
- present glass at time 1 (bea)
- present glass at time 2 (ker)
Prediction:
Time 1 presentation : (bea)
- glass | faster
- insect | faster
- stereo |slower
Time 2 presentation : (ker)
- glass | faster
- insect | slower
- stereo | slower
Results:
- matched predictions
- stronger activation for onset competitors than for rhyme
- at time 2, there is an inhibitory connection due to delay in presentation, preventing activation of incorrect item
- not consistent with Allopenna (1998)—due to difference in modality/methodology?
- weaker effect of onset comp.