2. Psychophysics
Learning objectives include understanding the
following concepts;
īŧ Importance of psychophysics;
īŧ Absolute & differential threshold;
īŧ Psychophysical methods;
īŧ Theory of signal detection
3. Psychophysics
īŽ Definition:
īŽ involves the determination of the psychological
reaction to events that lie along a physical
dimension.
īŽ Psychophysics = Psycho + Physics
īŽ Psycho or psychology is the science of
behavior
īŽ Physics studies the matter and energy of the
stimulus
īŽ Psychophysics ī Psychological events +
Physical events
4. ContâĻ
īŽ Edwin G. Boring (1950), âEminent historian of
experimental psychologyâ
īŽ Introduction of techniques to measure the relation
between internal impressions (the psycho of
psychophysics) and the external world (physics)
marked the onset of scientific psychology
5. Psychophysical Methods
īŽ Gustav Fechner formalized the psychophysical
methods, which measure attributes of the world in
terms of their psychological values.
īŽ Methods ī psychological judgments varied in
particular ways according to the
īŽ intensity of the stimulus;
īŽ the particular sensory modality of the stimulus
(i.e., judgments of visual stimuli differed from
judgments of auditory stimuli, which differed from
judgments of taste stimuli, and so on)
6. ContâĻ
Some relationships between Physical Stimuli
and Psychological Judgements
īŽ Physical Visual Intensity Psychological
Brightness
īŽ Physical Auditory Intensity Psychological
Loudness
īŽ Physical Measure of Weight Psychological
Heaviness
īŽ Physical Electrical Intensity Psychological
Pain
7. Scientific Topics
īŽ Operational Definitions
īŽ procedures used to produce a concept and
allow us to communicate successfully about
the concepts we are studying
īŽ Ensure that scientists use technical terms
īŽ Measurement Scales
īŽ Assignment of numbers or names to objects
and their attributes
īŽ Small n Design
īŽ Based on small numbers of subjects
8. Operational Definition
īŽ Provide technical meaning to the concept
īŽ Formula for building a construct
īŽ Other scientists can duplicate it
īŽ Specifying the ways used to produce and measure it
īŽ Clear and can be copied
īŽ E.g. Operationally define a construct called centigrams as the
product of your height in centimeters and your weight in grams.
Since any scientist can easily determine the centigram score,
this is a valid operational definition.
īŽ Tied to theory or body of research literature make sense and are
valid.
9. Introducing the Variables
īŽ Dependent Variables
īŽ Psychophysical studies ī one or two kinds of
judgements about stimuli
īŽ One stimulus ī an absolute judgement is
required
īŽ Absolute judgements ī simple statements for
presence or absence
īŽ Two stimuli ī a relative judgement is required
īŽ Relative judgement ī simple statements
about comparison
10. Independent Variables
īŽ Major IVs ī Magnitude and quality
īŽ Magnitude ī Changing the intensityâthe
physical correlate of loudnessâof a tone would
be a manipulation of stimulus magnitude, as
would be changing the weight of an object or the
concentration of an odor.
īŽ The frequencyâthe physical correlate of the
pitchâof a tone would be manipulated to produce
a qualitative change in the stimulus
īŽ Other qualitative judgments ī Various foods or
styles of different singers
11. Control Variables
īŽ Oberserverâs willingness to make particular responses
īŽ Attitude remain constant from trial to trial
īŽ E.g. An observer who is very willing to make a positive
judgment (âYes, I saw itâ) should maintain this same
willingness over the course of the experiment.
īŽ Classical or traditional psychophysics ī Once an
observer was trained, the attitude was supposedly
controlled.
īŽ Modern psychophysical theories, such as the theory of
signal detection do not accept this assumption.
īŽ Observer makes a response ī decision that depends
both on the stimulus and on the psychological factors
involved
12. Thresholds: Classical Psychophysics
īŽ Threshold ī Common language ī part of
doorway one step through or over to enter a
room
īŽ Classical psychophysicists ī Stimuli had to
cross such a hypothetical barrier to enter the
brain or mind
īŽ Strong stimuli ī easily jump over the
threshold
īŽ Feeble stimuli ī will not jump
īŽ Questionī how strong a stimulus must be if
a signal is to cross the threshold
13. ContâĻ
īŽ Slowly increase the intensity of the stimulus
e.g., tone or light, until the observers respond
âyes there it isâ
īŽ Problem ī Repetition of the process ī the
point at which an observer suddenly detects
the stimulus changes from trial to trial.
īŽ Classical psychophysicists ī to deal with
variability ī developed statistical methods to
estimate the best value for the threshold
14. Method of Limits
īŽ Fechner developed the method of limits
īŽ A psychophysical procedure for determining the
sensory threshold by gradually increasing or
decreasing the magnitude of the stimulus presented
in discrete steps
īŽ Experiment: Using the method of limits to determine
the threshold for a tone
īŽ Results would look like those shown in Table 6.1
īŽ Each column represents data from one block of trials.
15.
16. ContâĻ
īŽ First block ī clearly audible tone, to which the
observer responds âyes.â The tone intensity is
lowered in successive steps until the observer reports
âno,â thus ending that trial block.
īŽ Next block of trials starts with an intensity so low that
the observer cannot hear the tone and responds âno.â
On successive trials ī intensity increased until the
observer reports yes.
īŽ Process of alternating trial blocks continues until
Table 6.1 is complete.
īŽ Each block ī different intensity to avoid extra cues
that might mislead the observer
17. ContâĻ
īŽ If ī observer perfect stimulus detector ī the point at
which responses switched from âyesâ to ânoâ (or vice
versa) would always be the same
īŽ Ideal point ī threshold
īŽ Stimuli less intense ī value would never be
detected, and stimuli greater than or equal to this
ideal threshold would always be detected
īŽ Unfortunately, real data from real people do not have
this ideal characteristic; instead, they look like the
data in Table.
18. ContâĻ
īŽ Observers ī influenced by their expectations about
when they think it is time to change their response
from âyesâ to ânoâ or vice versa
īŽ E.g. if a series requires several âyesâ responses
before the threshold is reached, some observers ī
giving too many âyesâ responses and prematurely
respond âno.â Other observers ī cautious about
changing their responses ī delay too long.
īŽ Indeed, the same observer at different times may
commit both of these kinds of errors
īŽ Operational Definition of Threshold: mean (average)
of the points in each trial block at which the observer
switches from âyesâ to ânoâ (or ânoâ to âyesâ)
19. ContâĻ
īŽ Operational definition ī statistical
īŽ A threshold defined this way, based on an observerâs
ability to detect a signal, is called an absolute
threshold since the yes-no judgments are not based
on a comparison of two stimuli but are absolute
judgments about a single stimulus
īŽ Difference threshold: Based on relative judgments,
in which a constant unchanging comparison stimulus
is judged relative to a series of changing stimuli
20. ContâĻ
īŽ Example: observer ī lift pairs of weightsâone
weight always remaining the sameâand to judge if
the new weight is heavier, lighter, or equal to the
standard weight.
īŽ Several series of ascending and descending trials are
given.
īŽ The upper threshold is the average point at which the
observer changes from âheavierâ responses to
âequalâ responses. The lower threshold is the point at
which âequalâ responses give way to âlighterâ
responses. The difference between these two values
is called the interval of uncertainty
22. ContâĻ
īŽ Operational Definition of Difference Threshold
īŽ half the interval of uncertainty
īŽ In Table 6.2, this equals 10 grams.
īŽ The mean of the upper and lower thresholds is called
the point of subjective equality (300 grams in Table
6.2)
īŽ Properties of Difference Threshold
īŽ Ernst Heinrich Weber discovered important properties
1) The difference threshold increases with increases in the
magnitude of the standard stimulus. E.g., 10 grams is the
difference threshold when 300 grams is the standard, and
the corresponding value for a 600-gram standard stimulus
is a difference threshold of 20
Example of Candle in a Room
23. ContâĻ
2) Weberī famousī determining a second property of the difference
threshold: For a particular sensory modality, the size of the
difference threshold relative to the standard stimulus is
constant.
īą the ratio of 10 grams to 300 grams is the same as the ratio of 20
grams to 600 grams, 1/30 in this case. According to Weberâs
discovery, this means that the difference threshold for a 900-gram
standard stimulus should be 30 grams, and it should be 40 grams
for a 1,200-gram standard.
īą Fechner called relative constancy of the difference
threshold Weberâs law.
īą Formula: âI/I = K
I= Magnitude of the standard stimulus
âI= Difference threshold
K= Symbol of constancy
24. ContâĻ
īŽ Weberâs also known as ī Weber fraction
īŽ Varies in size for different senses
īŽ Example: Larger for brightness than it is for heaviness
īŽ Method of limits --- quite inefficient
īŽ Each column contains many successive responses either
yes or no that do not change
īŽ Staircase method ī Newer version of the method of
limits (Cornsweet, 1962)
īŽ concentrates responses around the threshold
īŽ For the first trial, it is similar to the method of limits.
However, once an estimate of the threshold is obtained,
the staircase method never presents stimuli that are far
from this estimate.
26. ContâĻ
īŽ Table 6.3. As soon as the threshold estimate is
crossed, the direction of stimulus intensity
reverses.
īŽ Improves the efficiency of the method by keeping
the stimuli much closer to the threshold than is the
case for the method of limits.
īŽ Operational Definition of Threshold: the mean
value of all stimuli presented, starting with the
second trial (column 2 in Table 6.3)
īŽ Parr, heatherbell, & White, 2002 Example of
Wine
27. No Thresholds: The Theory of Signal
Detection
īŽ Our perception in general is controlled by evidence and
īŽ signal or stimulus creates evidence
īŽ intensity of the signal and
īŽ the acuity of the observer, which partly determines a âyesâ
response.
īŽ Other determiners of a decision to say âyes, there is a
stimulus present,â including factors that influence the
willingness of the observer to say a signal is present.
īŽ Figure 6.3 shows the decision process is influenced by
both the evidence and response biases
īŽ Decision depends on costs and benefits associated with
it.
Depends on
28.
29. ContâĻ
īŽ Example of Blind Date and Marriage proposal
īŽ According to decision theory:
īŽ Conservative decision-makers ī marriage
īŽ Liberal decision-makers ī blind date
īŽ This response bias does not depend on the stimulusâindeed, the
same person could be involved in both instancesâbut only on the
costs and benefits of the decision.
īŽ Sensory End of Signal Detection
īŽ Sensory process transmits a value to the decision process
īŽ Value high ī decision process is more likely to yield a âyesâ
response once costs and benefits have been considered.
īŽ Value low ī decision process is more likely to yield a ânoâ
response, even if costs and benefi ts favor a âyesâ decision.
īŽ What determines the value sent by the sensory process?
30. ContâĻ
īŽ Signal-detection theory assumes that noise, a
disturbance that can be confused with signals, is always
present when a human attempts to detect signals.
īŽ Background disturbance is owing to such things as
environmental changes, equipment changes,
spontaneous neural activity, and direct experimental
manipulations.
īŽ Just to make sure that the assumption that noise is
present during attempts at detection, a typical signal-
detection experiment will present white noiseâa hissing
sound such as that heard when you tune your television
to an unoccupied channelâalong with the signal.
31. ContâĻ
īŽ Noise ī auditory or visual or can occur in
any modality; consider ī auditory system for
now
īŽ Experiment:
32.
33. ContâĻ
īŽ Hit: correctly detecting a signal when it is
presented
īŽ False Alarm: Incorrectly responding âyesâ
when only noise is presented
īŽ With a liberal decision strategyâcriterion set
to the leftâthe number of hits will be high; but
since there are numerous âyesâ responses,
the number of false alarms will also be high.
īŽ With a conservative decision strategy, false
alarms will be lowâbut so will hits.
34. ContâĻ
īŽ If we plot hits as a function of false alarms, as the
criterion moves from conservative to liberal, we get
the representation depicted in Figure.
35. ContâĻ
īŽ Figure ī receiver-operating characteristic (or
ROC) function.
īŽ Both hits and false alarms are infrequent
(conservative criterion) at the lower left of the curve.
īŽ As the criterion becomes more liberal, both hits and
false alarms become more likely, and the ROC curve
moves upward to the right.
īŽ The slope of the ROC function tells us the criterion.
īŽ Flat slopes reveal a liberal decision criterion
(generally, the upper right of the curve)
īŽ Steep slopes a conservative criterion (usually, the
lower left of the curve)
36. ContâĻ
īŽ There is no operational definition of a threshold.
īŽ Two quantities are operationally defined d and beta
īŽ The sensitivity of the observer is called d' and is
defined as the distance between signal and noise
distributions in Figure 6.4 or as the maximum
distance between the ROC curve and the diagonal in
Figure 6.6.
īŽ The criterion of the decision processes is called beta
(β) and is the slope of the ROC function at the point
of interestâfor example, a hit rate of 55 percent.
37. ContâĻ
īŽ Notion of an absolute threshold as determined by a
stimulus of a particular intensity has been denied by signal-
detection theory
īŽ DâAmato (1970) ī response or decision threshold. Only
when a stimulus yields evidence that exceeds the decision
threshold, what we have been calling or the criterion, do
we have correct detection of the signal. Of course, dâ
determine the detectability of the signal but not necessarily
what the subject reports.
īŽ This means that detecting and reporting the presence of a
signal are determined by d' and; together, these two
quantities determine what a classical psychophysicist
would call a threshold.
īŽ Calculating d': the sum of the two z values yields d'
38. Advantage of signal-detection methods
īŽ The ability to measure both sensitivity and
response bias
īŽ In many areas of applied psychology, the
ability to distinguish between these two
processes is very important
39. Measurement Scales
īŽ Measurement ī Systematic way of assigning
number or names to objects and their attributes.
īŽ Assign names or numbers to objects and their
attributes ī need measurement scale
īŽ E.g. When we measure temperature, for example,
we usually use either the Fahrenheit scale or the
centigrade scale. These two temperature scales
are inappropriate for measuring weight, which can
be measured in pounds or kilograms.
40. Properties of Measurement Scales
īŽ Four Properties ī combination of these properties
determines what is measured
īŽ Difference ī fundamental property --measurement
scales have instances that are different from each
other
īŽ Some temperatures are colder (or warmer) than
others, some people are male and some female,
and so on.
īŽ Magnitude ī Not universal
īŽ Determine the magnitude of attributes
īŽ scale can show that one attribute is greater than,
less than, or equal to another instance of that
attribute
41. ContâĻ
īŽ Equal Intervals
īŽ some scales can determine whether there are
equal intervals between magnitudes
īŽ 1-pound difference between two weights is the
same when considering both 1 versus 2 pounds
and 70 versus 71 pounds.
īŽ True Zero
īŽ true zero point on the scale
īŽ zero on the scale indicates that nothing of the
attribute being measured exists
īŽ cannot have less than zero weightâit has a true
zero point of no weightâbut you can have less
than zero degrees centigrade
42. Types of Measurement Scales
īŽ Nominal Scales (nominal is from the Latin nomalis, which means âpertaining to
namesâ)
īŽ measure just the property of difference and nothing
else.
īŽ Ordinal Scales (means in order. Includes âFirst,â âsecondâ and âninety ninth.â)
īŽ measure differences and magnitudes.
īŽ Interval Scales (as values of equal intervals that mean something.)
īŽ possess the properties of difference, magnitude, and
equal intervals.
īŽ Ratio Scales
īŽ all four properties of measurement scales (difference,
magnitude, equal interval, and a true zero).
43. Nominal Scale Examples
âĸ Gender (Male, Female, Transgender).
âĸ Eye color (Blue, Green, Brown, Hazel).
âĸ Type of house (Bungalow, Duplex, Ranch).
âĸ Type of pet (Dog, Cat, Rodent, Fish, Bird).
âĸ Genotype ( AA, Aa, or aa).
44. Ordinal Scale Examples
âĸ High school class ranking: 1st, 9th, 87thâĻ
âĸ Socioeconomic status: poor, middle class, rich.
âĸ The Likert Scale: strongly disagree, disagree,
neutral, agree, strongly agree.
âĸ Level of Agreement: yes, maybe, no.
âĸ Time of Day: dawn, morning, noon, afternoon,
evening, night.
âĸ Political Orientation: left, center, right.
45. Interval Scale Examples
âĸ On the other hand, temperature (with the exception of
Kelvin) is not a ratio scale, because zero exists (i.e.,
zero on the Celsius scale is just the freezing point; it
doesnât mean that water ceases to exist).
âĸ Celsius Temperature.
âĸ Fahrenheit Temperature.
âĸ IQ (intelligence scale).
âĸ SAT scores.
âĸ Time on a clock with hands.
46. Ratio Scale Examples
īŽ Exactly the same as the interval scale except that the
zero on the scale means: does not exist. For example, a
weight of zero doesnât exist; an age of zero doesnât exist.
âĸ Age.
âĸ Weight.
âĸ Height.
âĸ Ruler measurements.
âĸ Income earned in a week.
âĸ Years of education.
âĸ Number of children
47. Fechnerâs Law
īŽ Fechner ī psychophysical research done by Weber to try
to develop a measurement scale for sensations.
īŽ According to Weberâs law, the difference threshold bears a
constant relation to the standard stimulus: âI/I = K.
īŽ Fechner assumed that Weberâs law was correct and, with
two additional assumptions, developed his own law of
sensation measurement.
īŽ Fechner first assumed that the absolute threshold indicates
the point of zero sensation.
īŽ He then assumed that the just-noticeable difference (JND),
which is the internal sensation evoked by two stimuli that
differ by one difference threshold, is the unit defining the
intervals of an internal psychological scale.
48. ContâĻ
īŽ Because Weberâs law was assumed to be accurate,
Fechner believed that all JNDs produce equal
increments in sensation, as shown in Figure 6.9.
īŽ Each JND step on the psychological scale
corresponds to the physical stimulus that is one
difference threshold greater than the preceding
stimulus.
īŽ The first unit beyond the zero point corresponds to
the physical stimulus which is one JND above the
absolute threshold.
īŽ The next point will be one JND above that or two
JNDs above the absolute threshold
49.
50. ContâĻ
īŽ This process can be continued to build a
psychological scale.
īŽ Once this is done, there is a fixed mathematical
relationship between the value of the physical scale
corresponding to some point on the psychological
scale and the physical value corresponding to the
preceding point on the internal psychological scale.
īŽ To find the physical scale value that corresponds to a
particular psychological value, first take the physical
value of the previous step on the external scale (e.g.,
X in Figure 6.9) and multiply it by the Weber fraction.
51. ContâĻ
īŽ We then add this product to our original value, so that
Y = X + the product of X times the Weber fraction in
Figure 6.9 (likewise, Z = Y + the product of Y times
the Weber fraction).
īŽ Summing in this fashion yields successive physical
values that correspond to successive JNDs on the
internal psychological scale.
īŽ When this relationship is expanded and solved
mathematically, find that the psychological scale
value (á´Ē) is proportional to the logarithm of the
physical-stimulus value. This equation (á´Ē = K log
Stimulus) is called Fechnerâs law
52. ContâĻ
īŽ According to Fechnerâs law, all JNDs produce
equivalent increments in sensation; therefore, it
appears that we have a ratio scale (DâAmato, 1970).
īŽ The sensation corresponding to six JNDs should be
twice the sensation of three JNDs.
īŽ Question: Fechner actually devised a ratio scale of
sensation or not??
53. ContâĻ
īŽ First, Fechnerâs zero point is arbitrary rather than
absolute. The absolute threshold is defined
statistically and includes many sensations that do not
exceed the decision criterion
īŽ Second, we know that Weberâs law is only
approximately true; this could result in psychological
and physical units of varying sizes. There is an
additional difficulty with Fechnerâs formulation.
Fechner assumed that each JND was psychologically
equal, but if you ask people about the magnitude of
the sensory effects produced by stimuli of varying
JNDs above threshold, there is poor correspondence
between the two (DâAmato, 1970).
54. ContâĻ
īŽ Thus, Fechnerâs work is neither a ratio scale
nor an interval scale. At best, it is an ordinal
scale indicating that sensations are ordered
in a particular way with regard to the physical
stimuli that produce them
55. Stevenâs Power Law
īŽ S. S. Stevens (1961) attempted to develop an
internal scale of sensation more directly
īŽ Fechner used an indirect scaling method, in which
the psychological scale was built up by putting
successive JNDs in a row
īŽ The observers did not judge the magnitudes of the
JNDs directly, so the psychological scale values are
derived from measures of discrimination; therefore,
they are indirect.
īŽ Stevens used several direct scaling techniques, in
which the observer responded in psychological scale
units in the first place
56. ContâĻ
īŽ The primary direct scaling procedure used by
Stevens was the method of magnitude estimation,
which requires the observer to state a number that
represents his or her sensation of the stimulus
intensity.
īŽ The first stimulus that the experimenter presents is
arbitrarily assigned some convenient number, say,
100.
īŽ Then other stimuli are assigned numbers, depending
on how close the perceived intensity is to the first
stimulus.
57. ContâĻ
īŽ For example, the experimenter could present a tone
of moderate intensity and tell you it has a value of
100. Then a weaker tone might be presented, so you
would give it a lower number, say, 87.
īŽ These numbers reported by the observer represent
perceived psychological values directly. When data
are gathered in this way, the equation relating
psychological value to physical value differs from the
logarithmic relationship of Fechnerâs law. Instead, the
equation obtained by Stevens (1961) is á´Ē = K
(Stimulus)n , where n is an exponent. This equation
is called Stevensâ law
58. Small n Design
īŽ Large number of tightly controlled observations are made
on a small number of observers
īŽ Why Use Small n Designs
īŽ Many experiments require special participants, such as
specialists in radiology (interpreting X-rays), who are
scarce relative to the large numbers of undergraduates
that typically are used in experiments. Thus, a
psychophysical experiment on what data are used by
experts to detect breast cancer might include six
mammography specialists
īŽ Individual Differences
īŽ Personality and IQ differences
īŽ Poorly controlled conditions of large groups