1 - psychophysics

why measure perception?

• physical stimuli are “real” and can be measured

• perception is a private experience

• some say can measure perception bc may offer clues abt nature of the brain, the way it processes info and ultimately bio processes that lead to sensation + perception

  • in clinical setting, allow for proper diagnostic

what are methods used to study sensation + perception?

1. threshold: faintest sound you can hear

2. scaling - measuring provite experience: rate on scale

3. signal detector theory - measuring difficult decisions: takes cog processes into consideration

   • e.g., did you see light or was that imagination?

4. sensory neurosci: how do sensory receptors and nerves underlie our perceptual experience?

5. neuro imaging: what parts of brain active during crt tasks?

6. computational models: can we create models of sensory sys that adapt and learn, like humans?

what is classical psychphysics?

• pioneered by Gustav Fechner

  • true father of experimental psych

• psychophysics → study of quantitative relationships bet. physical stimuli and psych experiences

  • mind meets matter

how can we describe the relationship between mind and matter?

• relate physical stimuli to perceptual experiences using mathematical models bc useful to compare data

• func: mathematical description of how one variable is related to another; generally formula

• exponential: at lower intensities, no change pain sensation but w/ higher intensities you get increased perception pain

• log: small increases at lower intensities = high sensation

  • high slope at low intensity

what is absol thresh, subthresh & suprathresh?

• absol thresh → min stimulus lvl required to be registered by brain as sensory event; where func begins

• subthresh: below lvl detection

• suprathresh: above lvl detection

how can we measure thresholds?

1. method of adjustment

2. method of limits

3. method of constant stimuli

what is method of adjustment?

• simplest

• light shine computer and turn knob until can detect it OR turn it down until can just barely see it

what is method of limits?

• series stimuli

• start w/ subthresh and increase intensity OR w/ suprathresh n decrease intensities

• avg of thresh = absol

what is method of constant stimuli?

• random order

• more accurate bc no bias in expectation

why do we need to repeat measures over and over when measuring thresholds?

• variability in NS + computer screen, not same thresh each time

what type of data does a psychophysics experiment generate?

• ideal detector:

  • always sense suprathresh stimuli

    • NOT TYPICAL

• psychometric func: uncertainty arund stimulus intensities near abs. thresh

• select a 50% response lvl as abs. thresh bc bio, cog + physical variability

• normally, S-shaped funcs → ogive

what determines the shape of the psychometric func?

• absol thresh. only gives starting pnt

• for rest of func, need to have idea of:

  1. what slope at suprathrsh lvl and

  2. how slope changes w/ increasing intensities

• difference thresh (or just noticeable difference, JND), ΔI (Ernst Weber): how does stimulus need to change in order to produce a detectable change in perception?

how to find diff thresh?

experiment: present subject w/ 2 stimulis and ask which is heavier (i.e., greater stimulus “intensity“) → repeat

• one is standard and other is target

• calc % of times subj said target heavier than standard

• 2 alternatives → forced choice

• when said target is heavier 50% of the time → same so expect say light ½ time and heavier ½ time = perceptual equivalence pnt

• until get to 75% of the time target called heavier

  • target weight - standard weight = JND

if change the weight of standard → need a larger change in weight to detect a change at all when stimulus intensity is increased

describing Weber’s observations mathematically

• Weber’s law: ΔI = K’

• ΔI is a const. proportion (k) of stimulus intensity (I)

• Weber’s fraction, k, must be experimentally determined

• in graph → notice w/ increasing standard light intensity func get flatter

  • there is also increase in thresh w/ intensity (need a lil light to notice diff at low intensities)

sensory dimension and weber fraction (k)

• e.g., for smell → means need to add 7% of stimulus order to detect diff in intensity

• pitch is the most sensitive (need to change v lil)

how is the JND related to changes in perception?

• assumption: JND is like unit sensory perception = unit change in sensory perception

• difference = 1 unit change

• only func work as increase JND and keep sensation same → Fechner’s law

  • S = K ^log (I)

    • S = psych sensation/perception

    • k = constant (NOT same k)

    • I = stimulus lvl/intensity

• law assumes that all JNDs are perceptually equivalent, i.e., ΔI is a “unit of the mind”

• need to ad more to keep ΔS

• psych experinence increases less quickly than actual physical stimulus increases → slope decreases as intensity increases

  • NOT TRUE for ALL sensory modality

what is modern psychophysics?

• debate: whether stimulus-sensation relationship could be measured continued after Fechner and Weber’s work

• Stevens proposed new set method studying perception:

  • beleived could directly measure sensation

  • beegan era of modern psychophysics

magnitude estimation & power law

• scaling: a general psychophysical procedure to estimate the amount of sum related to perception

• magnitude perception: a scaling approach in which subject provide direct ratings of thir sensations

• experiment: subj assign num to the standard stimulus (modulus) and provide a relative numerical rating for other stimulus of varying intensities

  • discovered relationship bet. stimulus and sesation can be direction measured bu subj

    • subj ratings were consistent w/ power law

      • S = a x I^b

        • S = sensory experience

        • a - scaling const

        • B = exponent that determines if increases lot/lil at diff intensities

what is the val of power exponent (b)?

• each sensory experience is related to stimulus intensity by a specific exponent

• nature of the relationship, or “power func” or decreasing, depending on exponent

• NOTE: for exponents >1, Fechner’s law does NOT hold, use Steven’s power law

• graph:

• for blue and purple: at low lvl notice diff but not at high intensities

• for green n red: at low lvls, if increase shock a lil bit, sensation doesn’t change much but tiny lil increase in intensity can than lead to big changes in sensory experience

• shows Fechner’s law does NOT hold

what determines exponent in power func?

• sensory transducer theory: idea that transduction of the physical stimulus into the biological stimulus is basis power law

  • the neural output of sensory sys must follow power law relationship w/ the stimulus

    • e.g., elec shock

      • neural response gets much bigger w/ intensity

what are some other scaling techniques?

• cross-modality matching: compare stimulus from one sensory modality to stimulus of another sensory modality

  • relationships seem to be similar across individuals

• graph → equal sensation func

  • e.g., shock w/ sound of shock

why is JND important?

• relevant to many living things, not just humans

  • e.g., peacock feather diff might not be diff to females so can evolve to have less

    • good for them bc less heavy

what is signal detection theory?

• recall: sig detection can vary → abs. thresh depends on likelyhood that signal>noise to produce a perceptible event

• SDT: theory that takes into account nonsensory factors that can affect sig detection

  • uses stat concepts that take into account cog factors (i.e., sources of variability) that may influence a subj’s decision - making process

  • assumes the decision depends on sensitivity of sensory sys + udgment by subj (cog factor)

example 1 - detecting signal above noise

• scenario: you’re in the shower, waiting for a v important phone call

  • water making conts sound = noise

  • sound phone ringing = sig

  • noise n signal + noise r represented by distribution

• graph A:

  • magnitude perception varies

  • on avg, percpetion of loudness of shower noise = “x”

  • plot of all sensations would have a normal distribution

• graph B:

  • when ringtone plays, signal is added to noise

• graph C:

  • must decide on criterion (β) lvl of response, i.e., an int. thresh set by observer → min response lvl

    • β is automatic → unconscious

example 2 - detecting signal above noise

• radiologists use mammograms to screen women for breast cancer

• on a mammogram, cancer appears as a fuzzy white region (=signal)

• there are other fuzzy regions of the mammogram not due to cancer (=noise)

• decision for cancer diagnosis needs to be made

what are the 4 possible outcomes in an SDT experiment?

• in STD experiment, can test effect of noise alone by giving subj a num of trials in which no signals present

  • can show how sensitive participants r to noise

how sensitive are you to ringtone?

• sensitivity to a stimulus is illustrated by the separation bet. the distribution of your response to noise alone and sig+noise

  • by knowing relalationship of hits to false alarms, can calc sensitivity measure, d’, if vol phone is higher → will have a high d’

can β change according to expectation and motivation?

• blue arrow → more liberal criterion

  • influence of waiting/expecting news

• green → conservative lvl

  • e.g., if know it’s your mom calling, don’t go, can call back later

• for stimulus likelihood

  • expect stimulus 30% time → conservative

  • expect it 70% time → liberal

  • increases likelihood hit and false

what is the reiciver operating characteristic (ROC) curve?

• plot hit vs. false alarms for a sig of fixed intensity

• illustrates the effet of diff criterion effects in SDT experiment

• simply changing nonsensory factors can affect detection

  • response bias

• so there is no abs. thresh bc it depends on many factors

inof in ROC curve

• NOTE: higher d’ = distributions are far apart

1. provides estimate of relative sensitivities of diff individuals (d’)

   • can figure out by matching hit & false alarm rates to appropriate ROC curve

   • increasing d’ increases separation bet. distributions and the likelihood of a hit (less likelihood false)

     • as d’ increases → sensitivity increases

   • straight line on the graph = guessing (probability hit and false same)

2. provides measure of how nonsensory factors may influence judgement (β)

   • can see how probability change when signal and us stay same but there are expectancies & motivation involved

NOTE: ROC will never bend beneath chance performance (i.e., more false alarms than hits), bc the sig distribution is always to R of noise distribution → area under blue curve is always greater than area under red curve since blue curve = sig + noise