- Title
- A cascade theory of automaticity
- Creator
- Heathcote, A.; Sutton, Ken; Hayes, Brett; Mewhort, Doug
- Relation
- 12th Australasian Mathematical Psychology Conference (AMPC 2005). AMPC 2005 Conference Abstracts (Adelaide, SA 5-7 December, 2005)
- Relation
- http://www.psychology.adelaide.edu.au/ampc2005/program.html
- Publisher
- University of Adelaide
- Resource Type
- conference paper
- Date
- 2005
- Description
- We propose a theory of the development of speeded responding in choice tasks. Non-decision processing is performed by a cascade of linear accumulators, with accumulation beginning a fixed time after stimulus presentation (perceptual delay). The output of the non-decision stages feed into a decision stage consisting of an accumulator representing each choice. The first choice-accumulator to exceed its criterion, where criteria vary randomly and independently from trial-to-trial between accumulators, initiates a response after a fixed motor delay. The rates of all stages also vary randomly and independently from trial-to-trial. For almost any distributions of log-criteria and log-rates, response time (RT) has a Lognormal distribution with a lower bound determined by the sum of delays. Automaticity develops because the input to the correct accumulator increases exponentially with practice due to the formation of links to the choice stage with weights that interact multiplicatively. The mean log-rate of accumulation of each choice unit is determined by its input relative to inputs to other choice units, and so the rate for the correct response approaches an upper bound of with practice. The theory provided an accurate account of the effect of consistent-mapping practice on correct RT distribution in memory search (Kramer, Strayer & Buckley, 1990), mental rotation (Sutton & Heathcote, 2003), visual search (Heathcote & Mewhort, 2003) and categorization (Palmeri, 1997), including the effects of within- and between-category similarity in the latter data set. To account for the effect of algorithmic processing the correct choice unit was provided with an input from controlled-processing that does not change with practice. In cases where the controlled-processing input is initially larger than the automatic input a sigmoid (slow-fast-slow) decrease in RT is produced, whereas a smooth exponential decrease occurs when automatic inputs dominate. Strong control inputs accounted for sigmoid speedups prominent in Rickard’s (2004) data from an alphabet-arithmetic task, as well for some participants and conditions in Palmeri and Heathcote and Mewhort.
- Subject
- response time; automaticity; non-decision stages; choice-accumulator
- Identifier
- uon:2981
- Identifier
- http://hdl.handle.net/1959.13/32281
- Hits: 3597
- Visitors: 3567
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|