Word production Flashcards
(94 cards)
1
Q
When did the systematic study of speaking begin, and what was the first theory based on?
A
The end of the 19th century, Wilhelm Wundt (1900)’s theory was based on introspection.
2
Q
What did Meringer & Mayer (1896) do?
A
Collected and analysed spontaneously produced word errors for the first time.
3
Q
What is lexical access a core component of?
A
Theories of speaking.
4
Q
How many words does the mental lexicon contain in a normal literate adult?
A
50-100,000 (Miller, 1991).
5
Q
How many errors do we make when speaking?
A
1-2 every 1000 words (Levelt, 2001).
6
Q
How many words does the average person produce per second?
A
2-4 words per second.
7
Q
What does the fact that Fran Capo can say 11 words per second suggest?
A
That it is a highly automatic process.
8
Q
What key book summarised all speaking research and devised a general theory?
A
Speech production process (Levelt, 1989).
9
Q
What are the three levels of Levelt (1989)’s theory of speaking?
A
- Conceptualisation
- Formation
- Articulation
10
Q
What happens in the conceptualisation level of Levelt (1989)’s theory of speaking?
A
- Determine what to say
- Form an intention
- Select relevant information in preparation for construction of intended utterance
- Product of this process is a preverbal message
11
Q
What happens in the formation level of Levelt (1989)’s theory of speaking?
A
- Translate the conceptual representation into a linguistic form
- Lexicalisation: selection of individual words
- Syntactic planning: words are put together to form a sentence
- Phonological encoding: turn words into sounds
- Product of this process is a phonetic plan
12
Q
What happens in the articulation level of Levelt (1989)’s theory of speaking?
A
Speech.
13
Q
What is important about speech errors?
A
They were the first source of theories and models and they provide evidence for the units, stages and cognitive computations involved in speech production, as some types of errors are more common than others - there is lexical bias.
14
Q
Where are speech errors collected?
A
On databases, e.g. the Fromkin speech error database.
15
Q
How can speech errors be categorised?
A
By considering the linguistic unit and the mechanism involved.
16
Q
What are the units that can be involved in speech errors?
A
Phonemic features, phonemes, syllables, words/morphemes, phrases.
17
Q
What types of errors are there?
A
Anticipations (later segment takes the place of an earlier one), perseverations (earlier segment replaces a later one), exchanges, shifts, additions, deletions, substitutions and blends.
18
Q
Give an example of a phonemic feature error.
A
“Pity the new teacher” -> “mitty the due teacher”
- [-nasal] /p/ -> [+nasal] m
- [+nasal] /n/ -> [-nasal] d
19
Q
What kinds of phonemic segment errors are there?
A
Consonants and vowels, divided further by error type.
20
Q
Give an example of an anticipation error.
A
A reading list -> a leading list.
21
Q
Give an example of a perseveration error.
A
A phonological rule -> a phonological fool
22
Q
Give an example of an exchange error.
A
A brake fluid -> a blake fruid.
23
Q
Give an example of a deletion error.
A
Speech error -> peach error.
24
Q
Give an example of a syllable error.
A
Tremendously -> tremenly (hapology)
25
Give an example of a word error.
Tend to turn out -> turn to tend out (exchange)
26
What kinds of morpheme errors are there?
Inflection and derivational.
27
Give an example of a morpheme error.
I’d forgotten about that -> I’d forgot abouten that (inflection)
28
Give an example of a grammatical rules error.
I knew about it -> …I knowed about it
29
Give an example of a phrase error.
A hummingbird was attracted by the red colour of the feeder -> the red colour was attracted by the hummingbird of the feeder
30
What did Poulisse (1999) study?
Slips of the tongue in second languages, where the first language was Dutch (L1) and the second English (L2).
31
What did Poulisse (1999) find?
L1 based slips in L2, L1 lexical substitutions, L1/L2 substitutions, and L1/L2 blends, which seem to occur language independently.
32
What four key points can be made about analysis of speech errors?
1. Not all logically possible speech errors occur
2. Units which interact in speech errors are of one and the same type
3. Units in speech errors have characteristics in common
4. Distance between source and intrusion depends on unit type
33
What does the fact that not all logically possible speech errors occur imply?
That there's a mechanism that monitors the phonological legality of the utterance.
34
What does the fact that units which interact in speech errors are of one and the same type imply?
There must be processing components that are dedicated to these types of units.
35
Elaborate on the concept that units in speech errors have characteristics in common.
Exchanged words belong to the same syntactic category (e.g. verb, noun), but don’t necessarily share phonological similarity, whereas exchanged phonemes are phonologically similar and often found in a phonologically similar environment.
36
What does the fact that units in speech errors have characteristics in common imply?
That processes that work with particular types of units are sensitive to certain context characteristics (e.g. phonology/syntactic category) and not others.
37
Describe the idea that distance between source and intrusion in speech errors depends on unit type. What does it suggest?
When phonemes are exchanged, they're often from neighbouring words, whereas word exchanges span larger distances. This distinction supports the idea that there are different levels of processing.
38
What are the characteristics of word exchanges?
- Same syntactic category
- Open-class words (N, V, adjective)
- Phonologically different
- Relatively large distance between words/phrases
- Stress pattern remains the same
39
What conclusions can be made from the characteristics of word exchanges?
Words are inserted into slots of a syntactic frame, where frames include labels for syntactic categories but not phonological information.
40
What are the characteristics of word substitutions?
Errors that occur during lexicalisation, includes:
- Semantic related speech errors
- Phonological related (malapropisms)
- Mixed errors (words are similar in terms of both semantics and phonology)
41
What are the limitations of speech error analysis according to Meyer (1992)?
- The diagnosis of sound errors depends on listener judgments.
- Problem of the ambiguity of errors - they can be classified in different ways.
- Some classes of errors are hardly ever observed (e.g. intonation, stress), but one needs to analyse them in order to fully understand phonological encoding
42
What three things do speech errors not reveal according to Meyer (1992)?
- Anything about the time course of the association of segments to positions
- Whether segments with syllables are stored in the mental lexicon or generated during phonological encoding
- Information about the order of the processes involved
43
What conclusion was made by Meyer (1992)?
Other techniques than speech error analysis are necessary for a fuller understanding.
44
What did Ferber (1991) investigate?
The reliability of speech error databases.
45
What did Ferber (1991) find?
Collector's bias in online and offline detection of speech errors.
Out of 51 speech errors, with four listeners, they found between 7-25 speech errors. 14 errors weren't detected at all and no errors were detected by all listeners.
46
What did Baars, Motley & MacKay (1975) do?
Experimentally induced speech errors using the phonological bias technique (SLIP).
Participants were asked to silently read sets of word pairs, and to read aloud the last one. Designed to induce spoonerisms, e.g. ball doze, bash door, bean deck, bell dark, darn bore.
47
What did Baars, Motley & MacKay (1975) find?
32 word-word spoonerisms (e.g. barn door), compared to just 10 word-non-word spoonerisms (e.g. bart doard).
48
What did Baars, Motley & MacKay (1975) conclude?
There is a lexical bias effect, whereby induced sound errors tend to create words and there's no lexical bias when items in the list are non-words.
We use monitoring, therefore:
- Users check the lexical status before articulation
- No monitoring when items in the list are non-words
- Non-words are easier being detected and corrected by monitor
- Non-target words have a bigger chance of being missed by the monitor than non-words.
49
What is a criticism of Baars, Motley & MacKay (1975)?
The errors were produced experimentally - external validity.
50
What did Dell and Reich study?
Lexical bias in naturally occurring errors
51
What did Dell and Reich (1981) do?
Selected speech errors from a corpus which involved phoneme exchanges.
52
What did Dell and Reich (1981) find and conclude?
Lexical bias in exchanges, anticipations and perseverations. Concluded that the mental lexicon influences the outcome of sound errors and that there are interactive stages of language production.
53
What did Hartsuiker et al. (2006) find?
Lexical bias is also present in other languages e.g. Spanish.
54
How does Levelt's language production model (1989), distinguish between lemmas and lexemes?
Lemmas contain syntactic and meaning information, whereas lexemes contain phonological information.
55
What is WEAVER++?
An implementation of Levelt's model; a computational model developed by Levelt, Roelofs & Meyer (1999).
56
What is the WEAVER++ model generated within?
A discourse model (i.e. what one's talking about).
57
Why is the WEAVER++ model very useful?
As it explains phenomena and can be used to model how long it takes to say something.
58
What is lexicalisation?
The process whereby the thoughts underlying words are turned into sounds.
59
What are the three main questions concerning lexicalisation?
1. How many stages are involved in lexical access?
2. What information is represented at each stage?
3. Are these serial or interactive stages?
60
What are the stages involved in lexical access?
1. Lexical selection
| 2. Phonological encoding
61
Where does the evidence for two stages of lexical access come from?
- Speech errors
- Tip-of-the-tongue phenomenon
- Behavioural experiments
- Electrophysiological experiments (van Turennout, Hagoort & Brown, 1998).
62
What did Fay and Cutler (1977) state?
That there are two distinct types of whole word substitutions:
- Sematic e.g. fingers -> toes
- Form-based (phonologically related word substitutions or malapropisms) e.g. historical -> hysterical.
Word production and comprehension use the same lexicon, and phonological and semantic word substitutions are the result of mistakes in different parts of the word retrieval process.
63
What did Butterworth (1982) state?
That word retrieval is a two-stage process.
64
What is the tip-of-the-tongue (TOT) phenomenon?
The TOT state is a noticeable temporal difficulty in lexical access.
65
What evidence is there for the universality of the TOT phenomenon?
- 2yr old children (Elbers, 1985)
- Increase with old age (Burke et al., 1991)
- More common in bilinguals (Gollan & Acenas, 2004) - more potential candidates to compete.
66
How is the TOT phenomenon often investigated experimentally?
An object is described to participants and they are asked to name it.
67
What are the two main theories which explain the TOT phenomenon?
- Partial activation hypothesis (Brown, 1970)
| - Blocking or interference hypothesis (Woodworth, 1938)
68
What is the TOT phenomenon used as evidence for?
Two stages of lexical access.
69
How is the TOT phenomenon used as evidence for two stages of lexical access?
Vigliocco, Antonio & Garrett (1997) found that retrieval of grammatical gender is persevered in TOT states, therefore some syntactic information is available.
70
What support is there for Vigliocco, Antonio & Garrett (1997)'s finding?
Badecker, Miozzo & Zanuttini (1995) used evidence from an Italian person who suffered from word-finding difficulties (anomia). They had details of grammatical gender available for words that they couldn’t produce, thus they had access to lemmas but not to phonological forms.
71
What behavioural evidence is provided by Kempen & Huijbers (1983) for two stages of lexical access?
They had participants describe simple scenes and analysed the time taken before starting speaking. Found that people don’t start speaking until the content to be expressed has been fully identified, therefore production of the first word cannot take place until we have accessed all lemmas and at least the first phonological word form.
72
What behavioural evidence is provided by Wheeldon & Monsell (1992) for two stages of lexical access?
Used repetition priming and found that naming a picture is facilitated by recently having produced the name in giving a definition or reading aloud. Importantly, prior production of a homophone (weight or wait) isn’t an effective prime, which suggests that semantics are important. Concluded that the source cannot be phonologically mediated, so it must be semantic or lemma-based.
73
What question did van Turennout et al. (1998) address?
How speakers translate thoughts into words.
74
What did van Turennout et al. (1998) do?
Focussed on retrieval of distinct types of linguistic knowledge (semantic, syntactic, and phonological information). Had participants produce noun phrases e.g. red table using a LRP go/no-go paradigm.
75
What is LRP?
Lateralised readiness potential, derived from the readiness or Bereitschaft potential. Readiness potential (RP) is a movement-related brain potential that occurs before a movement is executed (Eimer, 1998).
76
What did van Turennout et al. (1998)'s task involve?
A syntactic-phonological classification task before a naming response. E.g. participants had to press a button if the word began with 'b' but not if it began with 's' (phonological), and this button was either with their left or right hand depending on the word's gender (syntactic).
77
How did van Turennout et al. (1998) avoid confounding variables?
Words were matched for frequency, number of syllables and word length. Each picture was presented 4 times in naming only trials and 6 times in trial that required the additional classification task, and the assignment of response types rotated across participants .
78
Describe van Turennout et al. (1998)'s Experiment 1 (and findings).
Syntactic gender determined the response hand, word-initial phoneme decision determined go/no-go. Found that participants prepared their button-pressing response with the relevant hand even in no-go trials.
79
Describe van Turennout et al. (1998)'s Experiment 2 (and findings).
Syntactic gender determined go/no-go, word-initial phoneme decision determined the response hand. Found that participants only prepared to press the button in go trials.
80
What did van Turennout et al. (1998) find?
LRP developed not only for go trials but initially also for no-go trials when syntactic gender determined the response hand.
81
What conclusions can be made from van Turennout et al. (1998)'s findings?
Syntactic gender of a noun is retrieved before its abstract phonological properties.
It takes about 40ms to retrieve a noun’s initial phoneme once the syntactic gender has been retrieved.
82
What did Schreifers, Meyer & Levelt (1990) investigate?
Whether lexical access involves two independent and serial ordered stages, where the first stage is lemma retrieval (syntactic and semantic properties) and the second lexeme retrieval (phonological word form information).
83
What paradigm was used by Schreifers, Meyer & Levelt (1990)?
Picture-word interference - participants were asked to name a picture while a distracter word was presented auditory or visually. They varied stimulus onset asynchrony (SOA) - the distracter word was presented either before (-), simultaneously, or after (+) the picture
84
Why was Schreifers, Meyer & Levelt (1990)'s study particularly useful?
Because it can be used to distinguish between models, as they make different predictions.
85
What outcome is predicted by serial models for Schreifers, Meyer & Levelt (1990)'s task?
Semantically related distractor words will result in inhibition (relative to unrelated distractor words) for early SOA, and phonologically related distractor words will lead to facilitation (relative to unrelated distractor words) for later SOA.
86
What outcome is predicted by interactive models for Schreifers, Meyer & Levelt (1990)'s task?
The effects of phonologically and semantically related distractor words are the same for different SOAs.
87
What did Schreifers, Meyer & Levelt (1990) find?
Semantic interference effect only when the word was presented first (early SOA), and phonological facilitation effect only for simultaneous and late SOAs.
88
What did Schreifers, Meyer & Levelt (1990)'s findings support?
A two-stage serial model of lexical access; mixed errors and the Lexical Bias effect must be explained by another theoretical notion such as an output editor or a monitor system.
89
What evidence is there to support a need for lemmas in a two-stage model of lexical access?
TOT data suggests that syntactic and phonological information are independent (Caramazza & Miozzo, 1997), and Italian speakers can sometimes retrieve partial phonological information when they can’t retrieve the gender of the word and vice versa - there's no correlation between retrieval of gender and phonological information.
90
What did Caramazza (1997) state about the need for lemmas in a two-stage model of lexical access?
Lemmas aren't needed, there's an interaction between the semantic network, syntactic network and phonological network.
91
Describe Dell’s (1986) model of speech production.
Highly influential, similar to IA model.
Explains speech errors, doesn't involve lemmas.
Interactive connectionist model of speech production in which semantic, lexical and phonological representations interact.
92
What do units specify in Dell’s (1986) model of speech production?
The syntactic, morphological and phonological properties of words.
93
How does activation work in Dell’s (1986) model of speech production?
Activation spreads from the sentence level to the morphological level and then to phonological level. The most active item is inserted into the current active slot of the frame (syntactic frame, morphological frame, phonology frame)
94
How does Dell’s (1986) model of speech production explain lexical bias?
Through the feedback between the phonological and lexical levels.
