PSC 100 Topic 2 Perception
Gestalt principles: Pragnanz
"pregnant" "good figure" we interpret the sensory input as corresponding to the simplest possible set of well formed objects
Photoreceptor cones
3 kinds of cones, each sensitive to a different range of wavelengths. red, green, blue cones every color is coded by the retina in terms of specific amounts of the red, green, blue some men only have two cones causing their perception of color to be different.
Our percepts reflect an interaction with what factors
External factors- influence the energy that is transduced by receptors internal factors- things inside our brain that influence how we construct a representation of the world based on the energy that hits our receptors. i.e lights, reflectance.
Gestalt Psychology
Gestalt is a german word that means figure in the sense of a unified or meaningful whole.
recognition by components.
Irving Biederman theory common objects can be broken down into a set of simple 3D parts called "geons" the geons and their relations remain the same if the viewpoint is changed. different objects may consist of the same geons but with different spatial relationships. Goal: figure out the parts and relationships to form a structural description Geons are the components
Fluorescent vs Incandescent
Most of the energy in a fluorescent light bulb is in the visible wavelengths making it more efficient than the incandescent.
Conditional Probability
P(A|B) Probability that A is true given that B is true
Four parts of perception
Perception as an active process lightness perception and perception of motion object recognition face perception
Bayes' Theorem meaning
Probability that the hypothesis is true given the evidence = (the prob. that the hypothesis is true w/ no evidence * probability that the evidence would occur if the hypothesis is true) / (probability that the evidence would occur no matter what
Rotation and Scaling
Problem: an object casts a different image on the retina depending on the distance and viewpoint of the observer --Solution: we mentally adjust the size and rotation of the image so that it can be matched against the template less mental rotation means faster reaction time (linear relationship)
Two types of object properties
Some are stable and diagnostic (color of an orange, the parts of a blender) Others are temporary and not diagnostic (the size of the orange's image on your retina, the outline of a blender)
Bayes' Theorem
The probability that some hypothesis is true given some evidence states the probability of the hypothesis being true given the evidence is equal to the probability that it would be true even without that evidence multiplied by the probability that this pattern of evidence would occur if the hypothesis is true (probability we'd get the slightly odd shaped "H "if it really is an "H"), and then divided by the probability that we would get this pattern even if it wasn't an "H."!
Ponzo illusion
This illusion happens because the railroad tracks are getting closer and closer as we go higher up in the picture, and the stones are getting smaller and smaller. the visual system uses these factors to guess that the upper part of the scene is farther away than the lower part
voice recognition example
Voicemail beginning in elmo PRIOR probability of the first word on the phone is hello is much higher than elmo. The voice recognition system did not take that into account.
Is recognition by components (structural description theory) or template theory correct?
We do not know one possibility: brain uses both
Color Constancy
We have color constancy mechanisms that partially cancel the effects of different light sources the light that reaches our eyes will depend on the objects AND the wavelengths present in light source. the color constancy adjust our perception of color. factor out the lighting and allow us to see the scene almost the say even though the actual distribution of light is very different.
sampling error will sometimes lead to
a higher sample mean for one group and sometimes a higher one for the other group. Not Systematic
When trying to perceive one object
a key property of the environment is the other objects that are present in the environment. i.e one object may occlude another (partly block one another) an orange looks bigger than a pineapple due to it being closer to you.
algorithm
a sequence of operations that is guaranteed to reach a correct solution
sampling error and individual differences are often the reason
a study fails to find a significant effect
Biederman: variable, non-diagnostic features are
accidental properties
within participants design
all participants receive all conditions reduce sampling error and probability of a false negative uses PAIRED t test
why can be see the different shades in sunlight and candlelight
although the number of photons hitting your eyes under sunlight is much larger, you don't perceive this as a change in the lightness of the A your perception is that it is still white but the intensity of the source of illumination has changed.
Receptor properties
another factor that influences our perception is the nature of the receptors that transduce the energy from the environment, turning it into electrical signals
heuristics
any approach to problem solving or self-discovery that employs a practical method, not guaranteed to be optimal, perfect, logical, or rational, but instead sufficient mental shortcuts faster than algorithm almost right all the time
photoreceptors
are in the retina at the back of the eye the light has to go through the eye and through these other cells to get there. This causes distortion and scattering of the light, which impairs our vision slightly.
law of optics in physics
are used to create computer generated images when an object blocks a source of light, it creates a shadow according to the geometry due to the way light scatters the edge of the shadow is slightly blurry. Thats how we tell the difference.
Psc experiments are not usually designed to
ask whether an effect is present in every individual draw conclusions about average person
The visual system makes
assumptions about what is and is not likely to happen in the world
what are good and bad criticisms in study
bad: sampling error, doesn't apply to everyone good: sampling bias confound alternative explanations drawing conclusions from null effects causation from correlation
JJ gibson wanted to figure out
better ways to teach these pilots how to land their jets
experience and knowledge in perception
both sensory inputs (bottom up processing) and our previous experience and knowledge of the world (top-down processing) influence our perception i.e TAE and CHT. The CAT our brain peices it together.
motion helps to
break camouflage
parsing
breaking the complex scene up into individual objects find edges filling in behind objects grouping together elements
faces have the same structural description but
differ in metric properties (exact sizes, shapes, and relative distance of the parts) the exact sizes and shapes of the parts and the relative distances among them are what we use to tell the difference between two people
we take in energy through
different types or receptors and construct a percept (mental representation that are in the world.)
Lesions in area MT (important for motion processing) of one brain hemisphere
do not have a large impact because the other hemisphere can be used
figure ground segmentation
each pixel on the screen has a red, green, and blue value the visual system must determine what is the foreground, the background, and the boundaries between them pulling the foreground and background apart because they are moving differently from each other
early stages of the visual cortex have
edge enhancement algorithms
Visible light
electromagnetic energy in a particular range of wavelengths
confounds are a property of
experimental design. Not a property of the pattern of the results
Things we use motion for in vision
figure ground segmentation (what the background is) locomotion distance perception 3D structure avoiding collisions
surface interpretation
figuring out the surface properties of each object (color, texture, 3D shape)
counterbalancing
half the subjects do rote memorization first and then elaborate encoding and the other half do the reverse
people with MT lesions in both hemispheres
have issues with walking around the room and finding a door or picking up a coffee cup
face are perceived and stored
holistically, not as individual parts we also store the overall pattern of relationships among the parts
The visual system fills
in the missing information i.e apple if the ink is not there to tell our visual system that it needs to fill in the gaps then we do not fill in the missing info very well
classic idea about info
information that comes into our sensory receptors is impoverished
Looming
is a pattern of motion that occurs when something is heading straight for your face the size of the RETINAL image starts expanding rapidly as the object gets very close to your face how motion tells us something is heading right towards us and we need to take quick action
when multiple theories can explain a given result
it is usually necessary to design new experiments where the results are consistent with only one theory
quasi-experiment
like a true experiment but without random assignment
recognition
linking lower-level perceptual properties with objects stored in memory process of linking a sensory input with a representation in memory
experiment
manipulate 1 variable and measure effect on variable 2 in group of subjects involves random assignment to different levels of variable 1
observational study
measure variable 1 and variable 2 in a group of subjects
what is the advantage of observational studies over experimental studies
more naturalistic because it is hard to do in controlled natural environments much cheaper
confound
more than a single factor systematically differs across conditions systematic
JJ Gibson figured out
motion in particular provides a lot of information for visual perception
JJ gibson realized
motion provides a cue to tell you where you are headed the point you are heading towards stays stable; the rest of the world dilates (or moves outward) around this point
we mostly use motion for
navigation and figuring out the space around us
Why are photoreceptors in the back
need to be next to the pigment epithelium, an opaque substance in the back of the eyes. the pigments in photo-receptors capture photons of light to release neurotransmitters. it can not work again until it sends a molecule called retinal to the pigment epithelium to be regenerated
Biederman: stable, diagnostic features are
non accidental properties. Brick: 3 parallel edges, innery Y vertex, 3 outer arrow vertices most angles
The fact that recognition by components has trouble with unusual viewpoints is
not a problem with the recognition by components theory because it matches this typical characteristic of human perception
Structural Description Theories
objects are represented abstractly as parts and relations between parts the visual system would form a structural description of the sensory input of that letter which would be compared with all of our structural descriptions in memory
Stages of Form Perception
parsing surface interpretation recognition
what happens when you look at an object
photons of light bounce off the pipe and enter your eyes this energy is interpreted by your brain to construct a mental representation of a pipe.
Law of physics in perception
physics play a big role in our perception, as they govern how energy is influenced by the environment. i.e inverse square law
Template Theory
picture like representation of objects stored in memory we compare incoming sensory input with our templates to see which template is best matched with the sensory input.
3D structure becomes easier to see when
put in motion
Prior probability P(H)
reflects our top-down knowledge and experience with the world
difference between sampling error and sampling bias
sampling error is RANDOM sampling bias is SYSTEMATIC in sampling error it will only make some groups better some of the time in sampling bias one group would have a consistent advantage.
Gestalt principles: similarity
similar shapes are grouped together, causing perception of vertical groups
sampling bias
sometimes an experiment will have this researcher picks 8 people in the front row and 8 in the back. (students in front better)
when recognizing individual faces, we are doing
subordinate-level categorization
face inversion effect
the difficulty of recognizing faces when they are inverted or upside down is specific to face perception
sampling error
the fact that when we sample from a small subset of a population the sample wont be perfectly representative of the population
3d structure from motion
the motion effects that underlay figure ground segregation, contraction, dilation, and motion parallax combine together to allow the perception of the 3D shape based on motion cues
if you are headed for the end of the landing strip (Locomotion Dilation)
the motion will dilate out away from that spot lets us know we are heading to the point
Shape constancy
the outline changes from a rectangle to trapezoid but we do not see the shape of the door as changing. The ability of the visual system to see an object as having the same 3D shape in the face of changes in your viewpoint that influence the 2D image that hits the eyes.
Gestalt psychologists believed that
the perception of an object is more than just the perception of the individual parts
The appearance of an object also depends on
the properties of the object the color of an object depends on which wavelengths of light it absorbs and which wavelengths it reflects
In template theory, the brain compares
the sensory input with every single template we have in memory and we recognize the object as the template that has the highest correlation if all correlations are poor, we simply do not recognize the sensory input
we can figure out the spatial relations among the geons in the object to form
the structural description of the object we then match the structural description with any of those stored in memory
when is there sampling error
there is ALWAYS sampling error in any area of science where samples are used. if the difference between the groups is statistically significant, the sampling error is only 5%
color constancy and shape constancy
they are comparable the visual system is combining multiple sources of information to make a best guess about what is really going on in this world
Templates can still work even if
they do not match perfectly. One template simply has to match well enough with the sensory input we can have multiple templates for a given object
motion parallax
things that are closer to you appear to move faster than things that are farther away how we determine which things are close to us and which are far away
Being critical
thinking carefully about whether somebody has made a convincing case
the simultaneous contrast illusion
this illusion is the result of a heuristic in which the visual system uses the region surrounding an object to make a judgement about the lightness of the object. (how well the object reflects light) and the overall amount of light coming from the light source.
Recognition by components is designed
to do basic level categorization, figuring out what basic category an object belongs to it is NOT designed to explain the difference between two members of the same category (2 guitars)
Perception involves
unconscious hypothesis testing
How does color constancy work
using the distribution of wavelengths across the whole scene to make a guess about the wavelengths present in the source of light. then they factor out the lighting in your perception of the objects in the scene
When perceiving an object
we actively construct a representation of every object we perceive in the world
How do we detect infrared light
we can not, as we do not have the receptors. Its hard ti imagine what it is like to percieve infared energy with them snakes have receptors on their heads that do.
what guess do we make about occluders
we must make a guess about what lies behind them, filling in the missing information. we unconsciously infer what must be behind the occluders
lightness constancy
we see the lightness of an object as being the same regardless of the intensity of the light source. different from color constancy where we see the COLOR being the same when the pattern of wavelengths of light changes.
Gestalt principles: good continuation
we tend to group things together if they follow a straight line or smooth curve
subordinate level categorization
when identifying specific members of a category
Locomotion Contraction
when we move away from something the visual scene appears to contract inwards towards the point that you are moving away from helps us know we are heading away from a point
sensory input can be consistent
with multiple different interpretations so we must use our prior knowledge to determine the most likely interpretation of the world.
Property of the environment for auditory
your auditory system has mechanisms of cancelling out echoes, if the delay between the original sound and the echo is short.