Lecture 5: Intro to Production

Question 1

(wasn’t there for lecture so may be missing a few things)

Two sides of communication

Answer 1

• translating sound into meaning

- translating meaning into sound

Question 2

How do we speak?

Answer 2

-Decide what we want to say
 Find words that fit our message
 Figure out which order those words need to go in to properly express our message
 Figure out what the sentence structure needs to be given our message and the words we’ve selected
 Figure out how the words we’ve selected are going to sound
 Plan several muscle movements to create the appropriate series of sounds

Question 3

Comprehension vs Production

Answer 3

 Message known vs. unknown
 input comes from within us vs. input comes from other person
 We make decisions about what is said vs. have little control over
what is said
 Start with meaning…vs. start with acoustics

Question 4

Miracle of Language Production

Answer 4

 All humans acquire the language of their environment.
 By the age of about 5 or 6 we have a nearly fully functioning linguistic system.
 As literate adults, we know between 50-100,000 words
(estimates vary, but in any case it’s A LOT of words).
 We manage to produce between 2-4 words per second.
 We make only around 2 errors every 1000 words.
 Unlike in sentence processing or word recognition, we have nothing in the environment to help us. We create the output, the input comes from us. Note that production lags behind comprehension in acquisition

Question 5

Stages of Language Production

Answer 5

1. Conceptualization (what do I want to say?)
2. Formulation (how do I find the right words to express my idea? how do I order those words into well-formed sentences?)
   - >words and syntax
3. Execution/Articulation (how do I make my mouth or hands move to produce the words and sentences)
   - >sounds

Question 6

Formulation

Answer 6

 Formulation = Lexicalization + Syntactic Planning +
Phonological Encoding
 Lexicalization - Finding the words that best express
your meaning
 Syntactic Planning - Putting the words in the right
order, with the right relationships and inserting the
appropriate grammatical elements
 Phonological Encoding - assembling the sounds needed for the words and making the necessary adjustments for context, etc. This level of representation drives articulation.

Question 7

Early Developments

Answer 7

 The early models of language production (Dell, 1986;
Fromkin, 1980; Garrett, 1975) investigated the types of
errors found in naturally occurring speech. Meringer &
Mayer (1895) spearheaded this approach.
 Errors provide a window into the types of
representations used in production and the types of
computations involved.
 By observing the errors and forming generalizations
over their patterns of occurrence, you can start to
form hypotheses about what the architecture of the
system must look like.

Question 8

Speech Error Types

Answer 8

Dell 1986
 Different sized units can be involved in errors.
 Sound errors: Accidental deletion or
interchange of sounds between words.
 Background lighting ⇒ backgound lighting
 Snow flurries ⇒ Flow snurries.
 Word errors: Accidental transpositions or
replacement of words.
 Writing a letter to my mother. ⇒ Writing a mother to my letter.
 Three aces ⇒ Three kings

Question 9

Characteristics of Errors

Answer 9

 Words typically exchange with words with same syntactic features.
Syntactic Category Constraint
 Nouns with nouns, verbs with verbs, masculine nouns with masculine nouns
-A fifty pound bag of dog food ⇒ a fifty pound DOG of BAG food
-Wears a very neat suit ⇒ HAS a very neat suit
 Sounds typically exchange with other elements in the
same word or syllable position.
-Word beginnings exchange with word beginnings:
-> Consonant clusters: Snow flurries ⇒ Flow snurries.
-Endings with endings:
->Can you read the small print ⇒ cad you rean the small print
-Vowels exchange with vowels:
->bring someone of this cAlibre into the cOllege ⇒ … cOlibre into
the cOllege
 Sound exchanges do not respect the SCC

Question 10

Implication of Error results

Answer 10

Interpretation: When selecting which word to say, syntactic category is important.
 When assembling the sounds of the words, syntactic category not important
 Suggests two different stages of processing. One stage for lexical selection, where syntactic features are important, and another stage for phonological processing, where syntactic features are not
important.
 Note this logical argument!
 This observation is the basis for the Lemma/Lexeme Distinction!
 Lemmas = amodal grammatical word
 Lexeme = phonological form of the word

Question 11

Speech production models

Answer 11

 There are several competing models of speech production, based
on different types of primary data.
 Reaction times, speech errors, patient data
 Speech production researchers agree on a few things.
 We must distinguish semantic, syntactic and phonological types of information which are stored and accessed independently

How these respective stores of information are related to one another is a central question within the field.
 Does information flow freely between all three components?
 Or, does information flow only in one direction?
 Are all the boxes linked? Or are some links not direct?
 Does information flow continuously between the boxes? Or in discrete stages?

Question 12

Lexicalization

Answer 12

How we select the words we want to express
Concepts
Grammatical word (LEMMA)
Phonological Forms (LEXEMES)

But how do these stages of processing relate to each other?

Question 13

Representation for Lexicalization

Answer 13

 Concepts - non-linguistic memory representations that
encode meaning. Theses meanings can relate to words.
If they do, they are called lexical concepts.
 Lemmas - abstract, amodal, lexical representation
encoding syntactic information for a word
 Lexeme - lexical representation encoding phonological,
or sound, information for a word.
 Phoneme - individual sounds that make up the lexemes.

Lemma, Lexeme distinction
-above line has to happen before below (serial discreat level)

(diagram in notes in Levelt 2001 mentioned, look up)

Question 14

Supporting 2-stage models

Answer 14

 Early distinction between Lemma and Lexeme comes from speech
errors.
 Additional support for the distinction as well as other differences
between the models come from reaction time experiments.
 Repetition priming (Wheeldon & Monsell, 1992)
 Method - first produce word from a definition. Later name a picture
 Wait - Wait √
 Weight - Wait X
 Results
 long lasting
 lemma based, not form based