Approaches to the Study of Perception

1
Approaches to the Study of Perception
Introduction to Psychophysics
2
Approaches to the study of perception

• Phenomenology Description
• Behaviour and Psychophysics
  Measurement
• Anatomy and Physiology Inference
• One goal of sensory neuroscience is to link
  perceptual behaviour (and experience) to
  physiological mechanisms

3
The Phenomenological Approach

• Often the starting point - sparks curiosity
• Provides a description of phenomena
• Provides the raw material for research
  
  

4
Hermann Grid
What produces those grey spots?
5
The Psychophysical Approach

• Examines the relationship between physical
  stimuli and sensory experience
• May provide the basis for inferring activity at
  the neural level

6
The Anatomical/Physiological Approach

• A direct examination of the underlying structures
  and mechanisms
• Can be done independently or in parallel with
  psychophysics
• May provide the basis for explaining
  psychophysical data

7
Introduction to Visual Psychophysics

• Psychophysics is the study of the relationship
  between physical stimuli and the perceptual
  behaviour
• Historically, psychophysics was directed at
  answering philosophical questions about the
  relationship between mind and body.
• Modern psychophysics is mainly a set of tools for
  investigating the nervous system - information
  processing capacities

8
Gustav FechnerThe founder of psychophysics
The term Psychophysik was coined by Fechner as
a result of a mystical vision that he
experienced on October 22, 1850. Fechner Day
(October 22) is now celebrated by
psychophysicists in memory of Fechner and his
contributions.
9
The contributions of Fechner

• Was interested in solving the mind-body problem
• Wanted to establish the relationship between
  changes in the physical domain and changes in
  subjective sensation
• He developed the classical psychophysical
  methods as tools to investigate this
  relationship
• He formulated the first psychophysical law

10

• Psychophysics is concerned with functional
  relationships between stimulus characteristics
  and perceptual behaviour
• Modern psychophysics uses these data to try to
  draw inferences about underlying physiological
  mechanisms
• The data are gathered using different
  psychophysical methods

11
The Domains of Psychophysics

• Detection
• absolute threshold
• Discrimination
• just noticeable difference (jnd)
• point of subjective equality
• Scaling
• rating subjective magnitude

12
Thresholds
Much of psychophysics is concerned with
thresholds
Threshold - means the 'beginning point'. In
early psychology it referred to the boundary
point which separated that which was not
conscious from that which was conscious.
13
Method of adjustment
To illustrate the method of adjustment, lets
imagine that we conduct an experiment on the
Müller-Lyer illusion. In this illusion the line
in the top figure with the arrows pointing out
looks shorter than the line in the bottom figure
with the arrows pointing in even though the two
lines are identical in length. Suppose we want
to measure the extent of this illusion.
14
Suppose we developed a computer program in which
we could vary the length of lines and the
computer stored the values of these line
lengths. We could then have observers adjust
the length of one of the lines until both lines
looked identical in length. For example, suppose
Observers adjusted the left line and the right
line was the standard (180 units (e.g. pixels)
in length). We would use both 'ascending' and
'descending' trials in our experiment.
15
A
D
The standard (ST) 180 pixels Point of
Subjective Equality (PSE) 143.5 pixels (Mean
setting for both ascending and descending
trials) The PSE indicates the setting in which
the left line looks on average the same as the
right line.
Constant Error (CE) PSE-ST 143.5-180 -36.5
pixels The CE indicates the average amount by
which the observer underestimated the standard
line length and in this case is 36.5 pixels.
Variable Error (VE) 4 pixels (mean standard
deviation (SD)) The difference between the PSE
and the observers setting on any trial is
called the variable error (VE) as it varies in
magnitude from the PSE over trials. VE is
measured by the standard deviation.
16
Method of Adjustment
Advantages and Disadvantages Advantages -
relatively high test-retest reliability -
more interesting for observer - quite
efficient Disadvantages - method is not
suitable for use with variables that do not
vary continuously - errors of
anticipation - subject to observer bias
17
The method of adjustment (5.2.4)
18
Method of limits
To illustrate this technique, suppose we want to
measure when a figure looks perfectly square.
This question might be motivated by a
well-known visual illusion called the
horizontal-vertical illusion. In this illusion,
a vertical line bisects a horizontal line. Even
though both lines are the same length, the
vertical line looks longer. This experiment
could tell us whether such an illusion operates
in simple closed figures like a square.
19
In our experiment we present rectangles in which
we vary one dimension (e.g. the Y dimension) by
a fixed amount- this is our step size. Here are
some samples below. The figure in the middle is
a perfect square while the rectangle on the far
left has a Y axis 10 longer than the X axis,
and that at the far right has a Y axis 10
shorter than the X axis.
20
We could keep track of our data like this
Threshold value -1.8
21
Method of Limits
Advantages and Disadvantages Advantages - it
is simple - reduces bias Disadvantages -
tends not to be precise - it is inefficient
since many trials far above and below
threshold are often performed.
22
The method of limits (5.2.1 )
23
Method of Constant Stimuli
How much light is required for an observer to
detect it? Suppose we choose 9 levels of light
intensity from subthreshold to
suprathreshold levels. We will present the
stimuli briefly in a light flash. To find these
levels we would first have to make some
preliminary observations and this is an important
part of such an experiment. We might, for
example, use another method such as the method
of adjustment to find such levels. Once we
have found the appropriate light levels we can
now do our experiment.
24
As in all psychophysical research, the task
involves presenting a stimulus to an observer and
measuring a response a 'simple' response. In
this case, the task on each trial is to tell
whether a flash was presented or not. A trial is
a designated interval in which a stimulus is
presented (or not in some cases as we will see
later). Since the experimenter knows whether or
not a flash was presented, the observer's
performance can be measured. The amount of light
necessary for a certain level of performance (the
threshold) can therefore be found.
25
A hypothetical threshold function with perfect
discrimination
26
Results of constant stimuli experiment
Psychometric Function
27
After obtaining such results, Fechner defined the
threshold as that point at which the observer
detects the stimulus 50 of the time.
Note that this definition maintains the concept
of a threshold, but assumes that the threshold
varies from to trial.
Later we will question the assumption of a
threshold.
28
Method of Constant Stimuli
Advantages and Disadvantages
Advantages - very precise and reliable -
reduces bias Disadvantages - inefficient -req
uires some prior knowledge of the threshold
value of interest in order to be able to
select the range of values to be tested.
29
Method of constant stimuli (5.2.5)
30
Adaptive Methods
To cope with some of the problems in the
classical psychophysical methods, so-called
adaptive methods have been developed. Many of
these adaptive methods are variants of the
classical method of limits. One of the main
problems with the method of limits is that it is
inefficient, because many trials that are well
below or well above threshold are
presented. Also,it is open to several sources of
bias because observers can guess what is going
on. In the adaptive methods, the stimuli that
are presented depend on how the observer has
already responded such that the next stimulus in
a series depends upon the response the observer
just made.
31
One adaptive method is a staircase method
Basically the procedure involves starting at some
stimulus value above threshold, for example, and
decreasing this value until the observer says
'no or not seen. On the next presentation
the value is increased and if the observer says
'no', it is increased until she says yes.
When The observer says 'yes' or seen the value
of the stimulus is decreased until he/she says
'no' and so on.
32
Staircase methods (5.2.3)
33
Double staircase-method
It is possible, however, that an observer may
become aware of the simple stimulus presentation
scheme used in a simple staircase - this
knowledge could influence the observers
responses. To deal with this and other problems
researchers often run two staircases
simultaneously, randomly switching from one to
the other.
34
Forced-Choice Procedures
While the Method of Constant Stimuli eliminates
some of the problems of the previous methods, it
still has some problems of response criteria
the point at which "No" responses become"Yes"
responses is determined not only by the stimulus
threshold, but also by other subjective
factors such as the response criterion. A
method developed to circumvent the problem of
response criteria is the Method of Forced-Choice.

35
Method of Forced Choice
Here subjects are presented with two or more
alternatives, and must select one on each trial
even if the stimulus was not clearly seen.
36
Square on Left or Right?
(2 AFC)
The observer must identify some characteristic of
the stimulus (i.e. square) other than its
intensity, for example. In this case, the
observer must indicate its location. This proves
that the observer can detect the stimulus.
37
Threshold is 75 correct Chance is 50 correct
38
Forced Choice Methods
- factor out criterion differences among
observers - produces lower thresholds - shows
that nervous systems register more information
than one is usually aware of
39
The determination of thresholds still plays a
major role in contemporary psychophysical research
Of course, researchers are aware of the
problems with the different techniques and design
their experiments using the best techniques