nROUSE Huber

1
nROUSEHuber OReilly 2003 Ellen Lau
PSYC 709
2
Cognitive problem that nROUSE focuses on

• How does brain avoid source confusion when
  parsing a continuous stream of stimuli?

Accomodation Synaptic depression
3
AccomodationNecker cube
Unamb View 1 (long) Amb Perception View 2
Unamb View 1 (short) Amb Perception View 1

• Long, Toppino, Mondin 1992

4
Accomodationsemantic satiation
should NOT diminish semantic priming effect
diminishes semantic priming effect

• spy spy spy spy spy spy spy spy spy spy spy spy
  spy spy spy spy spy spy spy spy spy

spy
bug
BEETLE
5
Accomodation2-alt forced choice task
long prime less facilitation or inihibition
effect
Short prime facilitation effect

• Huber et al 2001

6
ROUSE ? nROUSEmodeling issues

• ROUSE captured the pattern of data from Huber et
  al 2001, so why not leave it?

7
Synaptic depression
8
ROUSE ? nROUSE
9
nROUSE

• point neuron model
• model v with membrane potential update equation
• assume different levels w/ different integration
  rates
• Use lateral inhibition / attractor dynamics to
  get system to settle into a stable state
• Parameterize noise

10
nROUSEmodeling decisions

• Use simplified localist representation of
  distributed network for manageability
• Choice responding modeled with a horse-race
  decision rule
• Through simulation, we found that a horse-race
  decision rule...provides a good indication of
  residual target activation, whereas other logical
  alternatives do not.
• Point out neurophysiologically-supported
  possible solutions for handling noise

11
nROUSEmodeling parameters

• Fixed parameters inhibition, feedback,
  threshold, leak
• Free parameters rate of integration for three
  levels, rate of depletion, recovery, and noise

12
nROUSEmodeling resultsRepetition priming with
a single prime

• Testing
• core nROUSE proposal of accomodation through
  synaptic depression
• Assumption of varying levels with varying
  integration rates

13
nROUSEmodeling resultsRepetition priming with
two primes

• Testing
• perceptual effects not due to simple biasing

14
nROUSEmodeling resultsAssociative priming with
a single prime

• Testing
• motivation for having feedback between
  orthographic and lexical-semantic levels

15
nROUSE ROUSEmodel comparison - discussion

• Because nROUSE is a dynamic model in which
  processing is directly related to the sequence of
  events, it makes specific predictions as a
  function of stimulus duration. In contrast,
  different free parameters are necessary for
  different durations in the original ROUSE model.
  Therefore, nROUSE requires fewer parameters to
  handle a range of prime durations, although this
  reduction comes at the cost of increased
  complexity

16
(No Transcript)