Psy1302 Psy1302 Psychology of Psychology of LanguageLanguage
Lecture 14 & 15Lecture 14 & 15
Speech ProductionSpeech Production
Comprehension vs. Comprehension vs. ProductionProduction
Comprehension vs. Comprehension vs. ProductionProduction
SNP VP V NP
<the> <dog> <chased> <the> <cat>
/th/.../uh/.../d/.../ah/.../g/.../ch/...etc.
Creating SentencesCreating Sentences
Our brain does not store all sentences Our brain does not store all sentences we might ever need to produce.we might ever need to produce.
We must construct and plan our speech We must construct and plan our speech using our knowledge of languageusing our knowledge of language
The main issue of The main issue of speech productionspeech production concern the processes by which units concern the processes by which units come to be selected and then come to be selected and then combined in a particular order.combined in a particular order.
Studying Speech Production Studying Speech Production
HOW?HOW? Can we find evidence that we build structure on the Can we find evidence that we build structure on the
fly? fly? – Yes...Yes... e.g. e.g. Slips of the TongueSlips of the Tongue
Much of our initial knowledge of speech production Much of our initial knowledge of speech production comes fromcomes from– Slips of the tongueSlips of the tongue– Tip of the tongue phenomenonTip of the tongue phenomenon– DisfluenciesDisfluencies
Slips of the TongueSlips of the TongueFreudian SlipsFreudian Slips
Presidential Slip during campaign (Reported in Newsweek, 1992): I don’t want to run the risk of ruining what is a lovely recession.“reception” not “recession”
Slips of the TongueSlips of the TongueMalapropismsMalapropisms
Webster definition:Webster definition: the usually unintentionally the usually unintentionally humorous misuse or distortion of a word or phrasehumorous misuse or distortion of a word or phrase
Origin:Origin: slips named after Mrs. Malaprop (mal à propos), slips named after Mrs. Malaprop (mal à propos), a fictional character in a Richard Sheridan play (The a fictional character in a Richard Sheridan play (The Rivals).Rivals).
O, he will O, he will dissolvedissolve my mystery! my mystery!
He was a man of great He was a man of great statuestatue.. Thomas Menino, Boston mayorThomas Menino, Boston mayor Republicans understand the importance of Republicans understand the importance of bondagebondage
between a mother and child.between a mother and child. Dan Quayle, Vice PresidentDan Quayle, Vice President
http://www.fun-with-words.com/malapropisms.html
Slips of the TongueSlips of the TongueSpoonerismsSpoonerismsWebster definition:Webster definition: a transposition of usually a transposition of usually
initial sounds of two or more words (as in initial sounds of two or more words (as in tons tons of soilof soil for for sons of toilsons of toil) )
Origin:Origin: slips named after Rev. William Archibald slips named after Rev. William Archibald SpoonerSpooner
You have You have hhissed all my issed all my mmystery lectures.ystery lectures.
He is a He is a shshoving oving lleopard to his flock.eopard to his flock.
Three cheers for our Three cheers for our ququeer old eer old ddean!ean!
Anglican PriestDean of Oxfordhttp://www.fun-with-words.com/spoon_history.html
Slips of the TongueSlips of the Tonguesciencescience
Knowing which slips are possible Knowing which slips are possible and which are not constrains and which are not constrains theories of productiontheories of production
Models of speech production need Models of speech production need to account for these regularities to account for these regularities in slipsin slips
Slips of the handSlips of the hand
Newkirk, Klima, Penderson & Bellugi Newkirk, Klima, Penderson & Bellugi (1980)(1980)
Corpus of 131 errors in ASL Corpus of 131 errors in ASL – 77 videotaped77 videotaped– 54 reported observations54 reported observations
Errors like slips of the tongueErrors like slips of the tongue– ExchangesExchanges– AnticipationsAnticipations– PerseverationsPerseverations
Digression…
Types of ErrorTypes of Error MisorderingMisordering
– SubstitutionSubstitution ExchangeExchange AnticipationAnticipation PerseverationPerseveration
– AdditionAddition Anticipatory additionAnticipatory addition Perseveration additionPerseveration addition
– ShiftShift– DeletionDeletion
Noncontextual errorNoncontextual error– SubstitutionSubstitution– AdditionAddition– DeletionDeletion– Blend (word level)Blend (word level)
ObservationsObservations
Exchanged segments are from the Exchanged segments are from the same levelsame level
Implies you would never hear:Implies you would never hear:– Phoneme Level with Word LevelPhoneme Level with Word Level
““The The clcl is is marketmarketosed.”osed.” (the market is (the market is closed)closed)
ObservationsObservations Exchanged segments tend to be from the Exchanged segments tend to be from the
same kind of segmentsame kind of segment– Consonant onset with consonant onsetConsonant onset with consonant onset– Vowel with vowelVowel with vowelEtc…Etc…– Verb with verbVerb with verb– Noun with nounNoun with noun
Implies you would never hear:Implies you would never hear:– Vowel with ConsonantVowel with Consonant
““HaHauuow thow thlld”d” (hallow thud) (hallow thud)
– Onset with RhymeOnset with Rhyme ““UdUdallow thallow thudud” ” (hallow thud)(hallow thud)
ObservationsObservations
Sound substitutions Sound substitutions – Often close to each otherOften close to each other– Not necessarily similar in grammatical Not necessarily similar in grammatical
category and often similar in soundcategory and often similar in sound
ImpliesTYPICAL ERROR:I saw you fight a liar in the back quad, in fact you have...
UNCOMMON ERROR:I saw you light a fire in the yack quad, in fact boo have...
ObservationsObservations
Sound substitutions Sound substitutions – Often close to each otherOften close to each other– Not necessarily similar in grammatical category Not necessarily similar in grammatical category
and often similar in soundand often similar in sound Word substitutions can cross phrasal Word substitutions can cross phrasal
boundariesboundaries– Often far apart, crossing phrasal boundariesOften far apart, crossing phrasal boundaries– Often same grammatical category and dissimilar Often same grammatical category and dissimilar
in soundin sound Independence of stem morphemes from Independence of stem morphemes from
derivational morphemederivational morpheme
SNP VP V NP
<the> <dog> <chased> <the> <cat>
/th/.../uh/.../d/.../ah/.../g/.../ch/...etc.
Stages of AssemblyStages of Assembly
Language production requires assembling Language production requires assembling multiple levels of linguistic structure accurately multiple levels of linguistic structure accurately and fluently, in real time.and fluently, in real time.
Three levels:Three levels:– ConceptualizationConceptualization
– FormulationFormulation
– ArticulationArticulation
SNP VP V NP
<the> <dog>
/th/.../uh/.../d/.../ah/.../g/.../ch/...etc.
<chased><cat>
The dog chased the cat.
FormulationFormulation
What did the speech errors tell us about What did the speech errors tell us about formulation?formulation?
Separation between accessing semantics/syntax Separation between accessing semantics/syntax (meaning/grammar) and phonology (meaning/grammar) and phonology (pronunciation) of word(pronunciation) of word
FormulationFormulation
Distributional properties of errors suggest Distributional properties of errors suggest Grammatical Encoding stageGrammatical Encoding stage
– Puts words in order Puts words in order – Sounds irrelevant Sounds irrelevant – Syntactic relations relevant Syntactic relations relevant – Wide scope planning Wide scope planning
Phonological Encoding stage Phonological Encoding stage – Puts phonemes in order Puts phonemes in order – Sounds are relevant Sounds are relevant – Syntax is irrelevant Syntax is irrelevant – Narrow scope planning Narrow scope planning
Language production requires assembling Language production requires assembling multiple levels of linguistic structure accurately multiple levels of linguistic structure accurately and fluently, in real time.and fluently, in real time.
Three levels:Three levels:– ConceptualizationConceptualization
– FormulationFormulation
– ArticulationArticulation
GrammaticalEncoding
PhonologicalEncoding
Single Word Single Word ProductionProduction Lemma Lemma retrieval: retrieval:
– select a word that matches needed select a word that matches needed meaning and grammatical category meaning and grammatical category
Lexeme Lexeme retrieval: retrieval: – retrieve the sound of a word retrieve the sound of a word
p. 111-113 of Carroll
Digression…
Why might you believe in a Why might you believe in a distinction between Lexeme distinction between Lexeme and Lemma?and Lemma?
Tip of the tongueTip of the tongue– Can retrieve Can retrieve
lemma without lemma without lexemelexeme
know the know the meaning, first meaning, first letter, syllables, letter, syllables, and stress pattern and stress pattern but can’t but can’t generate the generate the word!!!word!!!
Digression…
Picture Naming TasksPicture Naming Tasks Name that pictureName that picture Sometimes with Sometimes with PrintPrint or or AudioAudio
DistractorDistractor Vary Vary Stimulus Onset AssynchronyStimulus Onset Assynchrony
(SOA)(SOA)
SOA
timelineNegative SOA
0 msPositive SOA
Digression…
Picture Naming TasksPicture Naming Tasks Name that pictureName that picture Sometimes with Sometimes with PrintPrint or or AudioAudio
DistractorDistractor Vary Vary Stimulus Onset AssynchronyStimulus Onset Assynchrony
(SOA)(SOA)
SOA
timeline
goat
0 ms-150 ms
Hear:
Digression…
Picture Naming TaskPicture Naming Task Name that pictureName that picture Sometimes with Sometimes with PrintPrint or or AudioAudio
DistractorDistractor Vary Vary Stimulus Onset AssynchronyStimulus Onset Assynchrony
(SOA)(SOA)
SOA
timeline
goat
0 ms 150 ms
Hear:
Digression…
Schriefer, Meyer, & Schriefer, Meyer, & Levelt (1990)Levelt (1990)
Semantic distractor: (e.g. goat for sheep)Semantic distractor: (e.g. goat for sheep)Inhibition occurs at SOA = -150ms (Before presentation of picture)Inhibition occurs at SOA = -150ms (Before presentation of picture)
Phonological distractor: (e.g. sheet for sheep)Phonological distractor: (e.g. sheet for sheep)Facilitation occurs between SOA = 0 to 150 ms (After presentation of Facilitation occurs between SOA = 0 to 150 ms (After presentation of
picture)picture)No facilitation at SOA = -150 msNo facilitation at SOA = -150 ms
““goat” activates Goat Lemma competes with Sheep Lemma for goat” activates Goat Lemma competes with Sheep Lemma for selection, causing inhibition.selection, causing inhibition.
““sheet” activates sounds and is similar in sound to “sheep”, sheet” activates sounds and is similar in sound to “sheep”, facilitating production.facilitating production.
Suggest phonological encoding follows lexical selectionSuggest phonological encoding follows lexical selectionFinding is consistent with model we are going to seeFinding is consistent with model we are going to see
Digression…
Language production requires assembling Language production requires assembling multiple levels of linguistic structure accurately multiple levels of linguistic structure accurately and fluently, in real time.and fluently, in real time.
Three levels:Three levels:– ConceptualizationConceptualization
– FormulationFormulation
– ArticulationArticulation
GrammaticalEncoding
PhonologicalEncoding
FunctionalProcessing
PositionalProcessing
Garrett’s ModelGarrett’s Model
FunctionalProcessing
PositionalProcessing
GrammaticalEncoding
LEMMA
LEXEME
Planning a sentencePlanning a sentence
She handed him a broccoli.She handed him a broccoli.
Message Level – Intended meaning
AGENT
THEMERECIPIENT
ACTION
LEMMA RETRIEVAL
Feminine PronominalMasculine PronominalVegetable floretAct of Transferring
POSSIBLE ERRORS?SEMANTIC SUBSTITUTIONe.g. BROCCOLI CAULIFLOWER
AGENTRECIPIENTTHEME
ACTION
she
him
broccoli
hand
FUNCTIONAL ASSIGNMENT
• VERB ARGUMENTS• CASE ASSIGNMENTSPOSSIBLE ERRORS?
WRONG CASE ASSIGNMENTe.g. Female pronoun-nominative (SHE), Male pronoun-dative (HIM) Female pronoun-dative (HER), Male pronoun-nominative (HIM).
shehim
hand
FUNCTIONALPROCESSING
broccoliindefiinite
POSSIBLE ERRORS?STRANDING
He ordered up ending some broccoli.SHIFTS – often inflections NOT root
She was hand himming some broccoliSuggests processing of inflectional (and derivational) morphology at this level
POSITIONALPROCESSING
Errors and StagesErrors and Stages
Intended Message:Intended Message: She handed him some broccoliShe handed him some broccoli
Likely ErrorLikely Error He handed her some broccoliHe handed her some broccoli
Unlikely ErrorUnlikely Error Her handed he some broccoliHer handed he some broccoli Him handed she some broccoliHim handed she some broccoli
Common ThemesCommon Themes
Garden Path Theory (when we talked about Garden Path Theory (when we talked about comprehensioncomprehension)?)?And notions of:And notions of:– Modularity Modularity – Informational Encapsulation Informational Encapsulation
(e.g., Syntactic Parser: access to grammatical function (e.g., Syntactic Parser: access to grammatical function categories, but not thematic information in the initial parse)categories, but not thematic information in the initial parse)
Garrett’s Model (when we talked about Garrett’s Model (when we talked about productionproduction)?)?Same ideas of Modularity and Information Encapsulation:Same ideas of Modularity and Information Encapsulation:– Discrete ProcessingDiscrete Processing– Functional Processing – lemmasFunctional Processing – lemmas
(access to grammatical function but not phonological structure)(access to grammatical function but not phonological structure)– Positional Processing and Phonological Processing – Positional Processing and Phonological Processing –
lexemeslexemes(access to phonological structure but not grammatical function)(access to phonological structure but not grammatical function)
Issues:
Discrete-stage processingStrict Feedforward(Completion of one stage before the next)
Cascading processing(Partial information sent to the lower level)
Interactive processingFeedback(Lower level affect higher level)
Are the stages discrete or cascading? Are the stages discrete or cascading?
Production IssuesProduction IssuesLevelt et al. (1991)Levelt et al. (1991)
lemma level
lexeme level/sheep/
STAGE 1
STAGE 2
GOAT SHEEP
/goat/??
Are Are stages discrete or stages discrete or cascading? cascading? How do we test?How do we test?
Does sheep prime goal?
?
lemma level
lexeme level /sheep/
STAGE 1
STAGE 2
goat
/goat/
/goal/ /sheet/
SHEEP
Discrete ProcessingDiscrete Processing
lemma level
lexeme level /sheep/
STAGE 1
STAGE 2
goat
Does sheep prime goal?
/sheet/
Discrete Processing says NO!
SHEEP
lemma level
lexeme level /sheep/
STAGE 1
STAGE 2
goat
/goat/
/goal/ /sheet/
Cascading ProcessingCascading ProcessingDoes sheep prime goal?
Cascading Processing says YES!
SHEEP
lemma level
(Semantic)
lexeme level
(Phonology)
/sheep/
goat
/goat/
/goal/ /sheet/
/sheep/
goat
/sheet/
CASCADING DISCRETE
Cascade or Discrete?Cascade or Discrete?
sheep sheep
Primary Task: Name the PicturePrimary Task: Name the Picture
Secondary Task: Lexical DecisionSecondary Task: Lexical Decision
Naming: 600 ms
150 ms 125 ms 325 msV Lem Lex
Lexical decision:goal or goat or sheet or mukle(button yes/no-rt)
Mediated Priming Mediated Priming ParadigmParadigm
Does sheep prime goal? Levelt et al. (1991): No.
lemma level
couch
lexeme level/couch/
STAGE 1
STAGE 2
sofa
/sofa/
/soda/
Peterson & Savoy (1998): Yes it does: couch primes soda via sofa
sheep – goat: categorical associates
sofa – couch: near synonyms
Peterson & Savoy Peterson & Savoy (1998)(1998)
Production IssuesProduction Issues
Are the stages interactive? Are the stages interactive? (Levelt, no; Dell, yes)(Levelt, no; Dell, yes)
lemma level cat
lexeme level /cat/
Levelt
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels
Codas
Dell
Gary Dell’s ModelGary Dell’s Model
Like the TRACE modelLike the TRACE model
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels
Codas
Interactive processingFeedback(Lower level affect higher level)
Message: Cat
Message: Some swimmers sink.
A 2-step Interactive Model of A 2-step Interactive Model of Lexical Access in ProductionLexical Access in Production
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Semantic Features
Adapted from Gary Dell, “Producing words from pictures or from other words”
Activate semantic features Activate semantic features of CATof CAT
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Semantic Features
Adapted from Gary Dell, “Producing words from pictures or from other words”
1. Lemma Access: Activation 1. Lemma Access: Activation spreads through networkspreads through network
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
Activation after 8 Activation after 8 stepssteps
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
1. Lemma Access: 1. Lemma Access: Most active word from proper Most active word from proper category is selected and linked to category is selected and linked to syntactic framesyntactic frame
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
NP
N
2. Phonological Access: 2. Phonological Access: Jolt of activation is sent to selected Jolt of activation is sent to selected wordword
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
NP
N
2. Phonological Access: 2. Phonological Access: Activation spreads through Activation spreads through networknetwork
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
NP
N
2. Phonological Access: 2. Phonological Access: Most activated phonemes are Most activated phonemes are selectedselected
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
Syl
On Vo Co
Errors (top-down)Errors (top-down)Semantic: Shared features activate Semantic: Shared features activate semantic neighborssemantic neighbors
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
NP
N
Errors (bottom-up)Errors (bottom-up)Phoneme-word feedback activates Phoneme-word feedback activates formal neighborsformal neighbors
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
NP
N
Errors (top-down & bottom-Errors (top-down & bottom-up)up)neighbors activated by both top-neighbors activated by both top-down & bottom-up sourcesdown & bottom-up sources
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
NP
N
Errors (top-down & bottom-Errors (top-down & bottom-up)up)Selection of incorrect phonemesSelection of incorrect phonemes
FOG DOG CAT RAT MAT
f r d k m ae o t g
Onsets Vowels Codas
Adapted from Gary Dell, “Producing words from pictures or from other words”
Syl
On Vo Co
Interactive or Interactive or Discrete?Discrete?Bad Dean Dad Bean
Back Deal Dack Beal
Other:DAD
Other:BEAN
Target:BAD
Target:DEAN
C1/b/
C2/d/
V1/ae/
V2/i/
C3/d/
C4/n/
Target:BACK
Target:DEAL
C1/b/
C2/d/
V1/ae/
V2/i/
C3/k/
C4/l/
Lexical Bias EffectLexical Bias Effect Words in the lexicon influence sound Words in the lexicon influence sound
substitutionssubstitutions
Experimental Data:Experimental Data:– Probabilities calculated from speech error corpusProbabilities calculated from speech error corpus
Sound substitutions resulting in words is higher than Sound substitutions resulting in words is higher than chancechance
– Inducing speech errors in laboratory using the Inducing speech errors in laboratory using the Speech Error Generation ParadigmSpeech Error Generation Paradigm
Sound substitutions resulting in words is more likely to Sound substitutions resulting in words is more likely to happen than those not resulting in wordshappen than those not resulting in words
Bad Dean Dad Bean; Dad Dean; Bad BeanBack Deal Dack Beal; Dack Deal; Back Beal
Speech Error Generation Speech Error Generation ParadigmParadigm
Instructions:Instructions:– You will see word pairs on the screen.You will see word pairs on the screen.– Read the words to yourself silently,Read the words to yourself silently,– But be prepared to say the words out But be prepared to say the words out
loud.loud.– When you see “????????” on the When you see “????????” on the
screen,screen,– Say the last word pair out loud as Say the last word pair out loud as
quickly as possible.quickly as possible.
YELL NOTSEED REAPSAME ROPE????????
LAMB TOYLOOM TENTLET TANKTIME LINE????????
BID MEEKBUD MEEKBIG MENMAD BACK????????
BALL DOZEBASH DOORBEAN DECKBELL DARKDARN BORE????????
BIG DUTCHBANG DOLLBILL DEALBARK DOGDART BOARD????????
Speech Error Generation Speech Error Generation ParadigmParadigm(Dell 1986)(Dell 1986)
Additionally: Speech Rate Manipulation3 groups of participants:500 ms700 ms1000 ms
What kind of error?
Speech Error Generation Speech Error Generation ParadigmParadigm(Dell 1986)(Dell 1986)
Repeated Repeated PhonemePhoneme
(Same (Same Vowel)Vowel)
Non-Non-Repeated Repeated PhonemePhoneme
(Diff Vowel)(Diff Vowel)
Word Word OutcomeOutcome
Bead Dean Bead Dean Deed Deed BeanBean
Bad Dean Bad Dean Dad BeanDad Bean
No Word No Word OutcomeOutcome
Beak Deal Beak Deal Deak BeakDeak Beak
Back Deal Back Deal Dack BealDack Beal
HIGH error rate
LOW error rate
Predictions?Predictions?
Target:DEAN
Target:BAD
C1/b/
C2/d/
V1/ae/
V2/i/
C3/d/
C4/n/
Target:Bead
Target:Dean
C1/b/
C2/d/
V2/i/
C3/d/
C4/n/
Bad Dean Dad Bean
Bead Dean Deed Bean
Other:DAD
Other:BEAN
Other:DEED
Other:BEAN
LOWEST error rate
HIGHEST error rate
Word OutcomeWord OutcomeRepeated Phoneme
Repeated Phoneme
No Repeated Phoneme
No Repeated Phoneme
Non-Word Outcome Non-Word Outcome
>
>
> >
Speech Rate and Speech Rate and ErrorsErrors
1.1. ____ speech rate, more errors____ speech rate, more errors500 ms: 112; 700 ms: 89; 1000 ms: 55 errors500 ms: 112; 700 ms: 89; 1000 ms: 55 errors
2.2. Full exchanges occur at _____ speech rateFull exchanges occur at _____ speech rate500 ms: 39 (35%); 700 ms: 26 (29%); 1000 ms: 7 (13%)500 ms: 39 (35%); 700 ms: 26 (29%); 1000 ms: 7 (13%)
3.3. Lexical bias effect should increase for _____ speech Lexical bias effect should increase for _____ speech raterate
500 ms: 51 (46%); 700 ms: 54 (61%); 1000 ms: 34 (62%)500 ms: 51 (46%); 700 ms: 54 (61%); 1000 ms: 34 (62%)
4.4. Repeated phoneme effect should increase with Repeated phoneme effect should increase with _____ speech rate_____ speech rate
500 ms: 51 (46%); 700 ms: 48 (54%); 1000 ms: 29 (53%)500 ms: 51 (46%); 700 ms: 48 (54%); 1000 ms: 29 (53%)
faster
faster
slower
slower
Corpus EstimatesCorpus Estimates
Dell & Reich (1981)Dell & Reich (1981) Asked naïve students to collect speech Asked naïve students to collect speech
errors for a montherrors for a month From corpus take 2 word speech From corpus take 2 word speech
errors errors (e.g., pitch fork (e.g., pitch fork fitch pork) fitch pork)
– Calculate percentage of sound exchanges Calculate percentage of sound exchanges that resulted in wordsthat resulted in words
– Compare percentage to estimated chance Compare percentage to estimated chance that an exchange would result in a word.that an exchange would result in a word.
Estimating ChanceEstimating Chance
Estimating ChanceEstimating Chance
First position(Pitch Fitch)
Second position(Fork Pork)
60
50
40
30% r
esul
ting
in w
ords
(e.g. Pitch Fork Fitch Pork)
* Data for anticipation and perseveration similar: % resulting in words is higher than chance estimates.
Actual Data
Chance Estimate
ResultsResults
Interactive or Interactive or Discrete?Discrete?Distributional properties of errors suggest Distributional properties of errors suggest Grammatical Encoding stageGrammatical Encoding stage
– Puts words in order Puts words in order – Sounds irrelevant Sounds irrelevant – Syntactic relations relevant Syntactic relations relevant – Wide scope planning Wide scope planning
Phonological Encoding stage Phonological Encoding stage – Puts phonemes in order Puts phonemes in order – Sounds are relevant Sounds are relevant – Syntax is irrelevant Syntax is irrelevant – Narrow scope planning Narrow scope planning
??? Do phonological factors influence rates of word substitution?
Broccoli Cauliflower(NO SOUND SIMILARITY)
Present Pressure(SOUND SIMILARITY)
According to discrete processing, selection of lexical items should not be influenced by sound similarity.Q: are word substitutions with sound substitutions greater than chance?
Dell & Reich (1981) Dell & Reich (1981) continuedcontinuedFrom same corpus created by naïve students, From same corpus created by naïve students,
found 289 word substitutionsfound 289 word substitutions Determine type of word substitution: Determine type of word substitution:
– semantic or othersemantic or other Divide the word and target up by phoneme Divide the word and target up by phoneme
segments and calculate for each segment segments and calculate for each segment whether the sound matcheswhether the sound matches
Compare with chance estimate that is Compare with chance estimate that is based on proportions of phonemes at each based on proportions of phonemes at each segmentsegment
Things to think aboutThings to think about
New data mustered to support the New data mustered to support the interactive view.interactive view.
But could the discrete processing But could the discrete processing view account for the new data as view account for the new data as well? well?
Could we save Garrett’s model by a Could we save Garrett’s model by a fix?fix?– E.g. self-monitoring?E.g. self-monitoring?