PSY2002 SEMESTER 1 - WEEK 6 Flashcards
what can cause sensory conflict
different sense providing conflicting info about sensory stimulus, conflict need to be resolved (multisensory integration problem)
name 3 contributers to sensory uncertainty
perceptual limits (eg; visual resolution determined by spacing of photoreceptors in fovea)
neural noise (eg, synaptic noise)
cognitive resource limits (eg; attention)
give an illusions example for when sensory uncertainty can occur
Necker cube = disambiguated by adding shadow/small bar introducing occlusion cue
describe testing conflict between vision and touch (Rock, 1964) to investigate integration of visual and haptic info
create conflict via holding object but look at it via a tube
look small but feel big
found vision dominating perceived size ie visual capture
outline research into how audition can dominate vision (Sharms, 2000) using visual flash/auditory beep
report number of visual flashes
auditory beeps (1-4) play during flash
found number of beep determine reported number of visual flashes = auditory capture
what can be concluded about visual/auditory capture
no strict sensory hierarchy, different senses able to dominate over other
define modality precision hypothesis
discrepancies are always resolved in favour of more precise or more appropriate modality
modality with highest precision (with lowest uncertainty) chosen depending on task
according to modality precision hypothesis, what modality is chosen for spatial tasks
vision
according to modality precision hypothesis, what modality is chosen for temporal tasks
audition
name an issue with the modality precision or modality appropriateness hypothesis
misleading = not modality itself or stimulus that dominates but instead dominance determined by estimate, and how reliably it can be taken within a specific modality of given stimuli
instead, an estimate precision is better
Ernst & Banks studied if sensory weighting is fixed or adaptable, using judgements of height of bar and discrepancies (uncertainty) between visual and haptic (tactile) input. what did they manipulate
to show weighting changes signals reliability, manipulated reliability of visual stimulus by adding noise to display
weight changed from visual dominance with no noise added (v reliable visual info) to haptic dominance when lots of noise added
Ernst & Banks studied if sensory weighting is fixed or adaptable, using judgements of height of bar and discrepancies (uncertainty) between visual and haptic (tactile) input. what should they find
behaviour change from visual to haptic capture
vision should show that when no added noise then vision should dominate size judgement
what is PSE
PSE (point of subjective equality) height where ppt equally likely to say one bar taller/shorter than other
Ernst & Banks studied if sensory weighting is fixed or adaptable, using judgements of height of bar and discrepancies (uncertainty) between visual and haptic (tactile) input. outline set up and what ppts had to do
vision = VR, change height of bar and modify uncertainty by adding visual noise
haptics = force feedback change, change height of bar
ppts judge how high think bar is
Ernst & Banks studied if sensory weighting is fixed or adaptable, using judgements of height of bar and discrepancies (uncertainty) between visual and haptic (tactile) input. what did they find
when no added visual noise, perception of bar height biased toward visual input
when increased visual noise, perception bar height determined by visual and haptic input
when high visual noise, perception of bar heights only determined by haptic inputs
Ernst & Banks studied if sensory weighting is fixed or adaptable, using judgements of height of bar and discrepancies (uncertainty) between visual and haptic (tactile) input. outline method
presented visual bars w/ sensory conflict between haptic and visual info to compare against bar without sensory conflict and determine PSE
(manipulated sensory uncertainty of visual feedbacks)
Newell found that visual/haptic object recognition depends on orientation of object relative to observer. when do we best recognise things visually
from side we learnt
Newell found that visual/haptic object recognition depends on orientation of object relative to observer. when do we best recognise things haptically
from side fingers explore most
what does sensory cooperation between differing modalities allow
increase info gathered, give more robust estimates
outline neural model of sensory integration
observer most likely to use on-line perceptual judgements to estimate variance (its unlikely we use variance from past experiences)
each neuron has preferred orientation, respond less strong in other orientation
when stimulated population activity has clearly defined peak marking orientation and also some variance
define normative model
how problems should be solved, an optimal solution, based on theory
and can establish bounds (best that we can possibly do)
define process model
how problem actually solved (based on data)
apply normative model vs process model to finding the shortest path between home and work
normative= compare length for all route and pick shortest
process= choose between small number of salient options
outline integrated signals method
integrate vision and haptic to reduce both visual and haptic uncertainty to form integrated signal
outline integration signals in Ernst study, how it solves issues
haptic input is 50, visual is 60 = causes sensory conflict
but
combined is in-between, meaning 55 and so means the more sensory inputs means smaller uncertainty
combined estimates are biased toward visual etstimate
why is combined/integrated signal model good
- allow future numerical prediction for experiments if people are following this specific model exactly
- able to compare prediction and actual result
- normative models applied to multisens integration allow calculating optimal way to integrate info and solution is minimising sensory uncertainty after integration
- humans act in accordance with this model and therefore integrate sensory info optimally
what is MLE (maximum likely estimation) model
most efficient manner to integrate sensory info sources
first specify goal of sensory estimation and if goal is to get most reliable unbiased estimate, then variance of final estimate needs to be reduced
MLE is estimate with the lowest variance
in MLE model, what do estimate weights need to take into account
info quality
in MLE what gives most reliable estimate
optimally integrating sensory info
reliability of integrated estimate is sum of reliabilities of individual estimates
human performances follow optimal sensory integration rules. what is visual weight range from
1-0 (1 is fully just trusting our vision)
linear weighting rule to integrate signals where weights depend on signal reliability
explain how the MLE approach is bottom-up
prior knowledge not used
to include this need extending model
when do we form weighted averages of motion and disparity signals
when asked to report an objects shape,
same for texture and disparity of depth
visual perception of slant
judgement of texture-defined edges
estimating distance
when is vision most precise, and when proporioceptions most precise
when discriminating along horizontal vs depth discrimination but proception more reliable for discriminating along depth than horizontal
suggsting an MLE curve between 2 visually and proprioceptively specified location
what are issues with optimal integration (do we always integrate info optimally)
need to know uncertainty which is hard to estimate (easier in sensory perception >cog reasoning)
and calculation takes long times (can use heuristic but are suboptimal)
are probabilities encoded in brain?
little direct electrophysiological evidence
what is correspondence problem
if can hear and see something does it mean there are 2 things, or is it the same
in correspondence problem, when are things integrated
if sound like from same place
but if what you see/hear seems different (see dog and hear cat), then this is more complex
in correspondence problem, when are signals most likely to be integrated
if occur simultaneously with no spatial discrepancy and not likely to be integrated when spatial discrep large, or temporal sequence of event not appropriate
how can optimal integration behaviours be studied
forced-choice discrimination paradigm, ppt compare perceived sizes
