Please enable JavaScript.
Coggle requires JavaScript to display documents.
PRODUCTION I: Speech errors and experimental evidence - Coggle Diagram
PRODUCTION I: Speech errors and experimental evidence
Tension
Top down v. bottom up processes
i.e. structure driven v. word driven during mental composition
While these two approaches are often posed as diametrically opposite it is actually the case that the two are not mutually exclusive and often a combination of the two processes may be in use.
Focus on speech errors to reveal processes behind speech production
3 Stages of Production
Message Formation
analogous to locutionary forces in an analysis of speech acts (Searle, 1969)
The contents of messages are sometimes treated as propositions
Symbolically, propositions have a predicate (an abouter, often expressed with a verb and other modifiers that qualify aboutees) and arguments (one of which is the aboutee that the predicate modifies.
We avoid propositional terminology and instead talk about messages in terms of aboutness and the products of categorization, the concepts that play various aboutness roles. The message is a
prelinguistic representation of aboutness relations among concepts.
Structural scaffolding
messages provide information relevant to assembling words and syntactic relations
the constraints from message contents must span cognitive chasms of sorts. The chasms have inspired terminology like rift (Levelt, 1993) in word retrieval and syntactic cleft (Bock, Loebell, & Morey, 1992) in structure-building
speakers do not always formulate complete messages before the mechanisms of language production begin
(Brown-Schmidt & Tanenhaus 2006)
Notions/conceptualisation
what a speaker intends to communicate
They are by definition
non-linguistic
Structure-driven production
the majority of errors occur within syntactically sound utterances
Such conclusions were first made by scholars such as
Fromkin in 1971
and Garrett this is because even when words and their locations and utterances is wrong the syntax is right
Faye 1982
They found that syntactic errors, in large collections, constitute only about
2%
of all the errors recorded, of these only around 0.3% were ones that contained clear structural violations
some errors contain simple switches between constituent phrases that have the same grammatical properties hence when switched the utterance still syntactically makes sense
the car on top of the bike instead of “the bike on top of the car “
Speech errors such as this point towards a production where structure is first decided and then a specific words are slotted into the structural scaffold which has been constructed (
Shattuck-Hufnagel’s early “slots and fillers” model)
Attraction
is where verb number agrees not with the number of the subject but with the grammatical number of another noun phrase in the sentence eg ‘he eat it’ instead of ‘He eats it’
provides evidence for structural account of language processing and speech production
Word-driven production
Gleitman et al. (2007)
PP's presented with images where alternative actors were cued using almost imperceptible attentional cues
he found a clear impact of potential queueing on the choice of an initial phrase often the actor that was queued would be the subject phrase i.e. queued reference were more likely to be mentioned first
This is credible support for a word driven view of the transition between messages and utterances (Bock and Ferrero)
certain words have a valency value which tells you how many arguments a verb requires; if a speaker selects above with a valency of two this will entail two arguments = words driving structure
Hybrid approach
Kutchinsky (2009)
replicated the Gleitman et al. result that queued actors tended to be mentioned earlier then uncued actors
However, for the readily interpreted events the results were different: attentional queuing had little effect on early mention
when events were hard to interpret, early attention to an object in the event elicited a word-driven production pattern, but in easy-to-interpret events, the cue had no consistent impact on the starting point.
:red_cross: not the context in which the majority of utterances occur
as they often come from messages which the speaker wishes to convey rather than the speaker describing their surroundings
"Word- and structure-driven sentence formulation are far from mutually exclusive" (syntactically speaking p.10)
Problems in production
Between stages, particularly between the message and the lexicon
Links may automatically activate words, but sometimes too weakly for identification, sometimes too many at the same time (Cutting & Ferreira, 1999), and sometimes none at all (Burke, MacKay, Worthley, & Wade, 1991).
Early models were based on speech errors and serial, model shifted as more evidence became more experimental
EARLY MODELS
SPOONERISMS