Rich Prather <rprather@mit.edu>
If pigeons will use features to recognise objects, how specific a feature will they pay attention to. As in the charlie brown example, would the pigeons be affected if his eyes and nose were moved?
Assuming pigeons can pick out paintings by different artist, what would the difference between how a pigeon does it and how a human would. do they both just recognise some feature that is common to a certain artist paintings? If so then a pigeon would not be able to recognise a noncubist picaso painting.
-------------------------------------------------------------------------------------------
"Amrys O. Williams" <amrys@mit.edu>
Research Question:
In the colored corn experiments, are the pigeons forming the concept "corn" or simply "food"?
Short Answer Question:
Describe two experiments that support local templates as a mechanism for recognition in pigeons.
Multiple Choice Question:
Polymorphic concepts are
a) complex
b) a disjunction of conjunctions
c) difficult to grasp
d) none of the above
e) all of the above
-------------------------------------------------------------------------------------------
Charisse Massay <charisse@mit.edu>
1. Open research question: What if you were to build little 3D scenes out of little lego trees and little people. Make several of these 3D scenes to scale and present pictures of different angles to the pigeon or you can present the actual scene in a little glass aquarium of sorts. The tests would be the same as the ones with just pictures but the experimenter could use actual scenes or the scenery separate from everything else. this setup would be more like what the pigeon sees in real life and features can be isoated on an independent and 3D basis.
2. Short answer question: What are the two major hypotheses for pigeon recognition? Which seems more likely? What evidence is there for this hypothesis instead of the other? [global/local]
3. Multiple choice: Which of the following is *not* an action performed by gull chicks in response to hunger?
a. chirping
b. pecking at mother's beak
c. pecking at mother's legs
d. pecking at regurgitated material
-------------------------------------------------------------------------------------------
Liina Pylkkanen <liina@mit.edu>
1. Open research question
Is pidgins' pattern recognition truly categorical? Imagine a situation
where the pidgin is trained on a certain category boundary for
stimuli that visually form a continuum (say, a continuum between
being a square or a circle). In the test session, does the
pidgin show categorical bahavior (s-curve) or does response rate
increase linearly?
2. A short answer question
What's the evidence that pidgins learn categories rather than
map new stimuli to innate categories?
3. A multiple choice question
There is evidence that the gull's natural concept of a beak is
(a) learned
(b) innate
(c) species-specific
-------------------------------------------------------------------------------------------
Richard Russell <rrussell@mit.edu>
1) open research question/project idea: train birds with slides of different kinds of trees at a given distance and then test them on recognizing other trees at the same distance that have different shapes than those that were in the initial training set.
Positive results would indicate that the birds generalized from the various templatees to some concept of 'tree'.
2) Short answer:
What is some evidence that pigeons use local and not global templates in object recognition?
3) Multiple choice:
Studies have generally found that in object recognition pigeons use
A) global templates
B) local templates
C) mental encyclopedias
D) polymorphic concepts
-------------------------------------------------------------------------------------------
Matt Cain <mcain@mit.edu>
Open Research:
If gulls of one species are raised by parents of another species, with which species do they try to mate?
Short Answer:
Q) What features do pigeons use to discriminate the works of Monet from those of Picasso?
A) Features such as color, line, and presence of natural objects all seem to be employed to some degree, but none are necessary or sufficient.
Multiple Choice:
What is the most important species determiner between gulls for mating purposes?
a) wing tip markings
b) color contrast in the eye area
c) scent
d) physical interaction
-------------------------------------------------------------------------------------------
Janice Chen <kanile@mit.edu>
1. A gull chick's ability to discriminate between its parent's beak and
other objects in the environment becomes more refined after the
first few days of life. Is this increase in ability purely
visual? Is the chick using other cues (sound, smell, motion) to
enhance its parent-recognizing capability?
2. Why do chicks prefer to peck at vertical rods projecting from
above eye level, rather than at those below eye level?
3. A newly hatched chick responds best to
a) very specific aspects of its parents appearance
b) very simple stimuli characteristic of its parents bill
c) an image matching its first view upon hatching
-------------------------------------------------------------------------------------------
"Jennifer C. Shieh" <jcshieh@mit.edu>
1. Determine what feature allows pigeons to discriminate between paintings done in the Impressionist style and the Cubist style, and learn whether the pigeon demonstrates a preference for a style.
2. Explain how gulls demonstrate knowledge of their own image.
3. What experimental evidence supports the theory that birds may use absolute discrimination as a mechanism of recognition?
a. Ability to classify an image geometrically similar to a training image.
b. Inability to identify scrambled images as the same as a trained image.
c. Inability to demonstrate a knowledge of the abstract concept of insideness.
d. Ability to discriminate an oak leaf from other tree's leaves.
-------------------------------------------------------------------------------------------
Kennet Belenky <kbelenky@mit.edu>
1)
How does age affect a pigeon's ability to learn various discrimination tasks.
2)
Q) How did the oak-leaf experiment demonstrate the pigeon's ability to encode 2-d relations of local features?
A) Pidgeons strongly generalized features from one species of oak to all oaks. The various oak leaves have different physical outlines but share a common feature template of lobes. The generalization showed that pidgeons encoded the relationship of various parts, rather than the form of the whole leaf.
3)
The gulls disciriminate against other gulls that don't look like themselves. How does a gull know what it looks like?
a) Reflective surfaces. Gulls with painted eye-rings adopted a new identity among the target species.
b) Imprinting. Gulls look for a mate that looks like dear old mom.
c) Social interaction. A third party gull determines if a couple matches.
d) No theory has been conclusively proven.
-------------------------------------------------------------------------------------------
Andrew Yip <ayip@mit.edu>
Open research question/project idea:
Pigeons have been tested on their ability to recognize natural objects such as trees and water. However, these studies used 2D images containing backgrounds. Using stereoscopic images of natural objects without backgrounds could provide insights into how pigeons recognize and categorize natural objects.
Short answer question:
What evidence suggests that pigeons can use combinations of local templates for object recognition?
Multiple choice question:
What visual characteristics help gulls select a mate of the same species?
a) eye color
b) eye ring color
c) wingtip color
d) all of the above
-------------------------------------------------------------------------------------------
Keith Thoresz <thorek@ai.mit.edu>
1. Pigeons are adept at navigating long distances from the air, where exposure to 3D structure from parallax is limited. How would the ability to extract 3D structure from scenes benefit pigeon navigation?
2. If pigeons can retain in memory thousands of images over several years, this would seem strong evidence for their having strong brute force memory structures. Does the finding that simple transformations of the images impair recognition undermine this conclusion?
3. Which of the following was not a result of the pigeon research discussed in class.
a. Pigeons are incapable of seeing pictures as real objects
b. Pigeons use local template matching
c. Pigeons are able to learn complex concepts
-------------------------------------------------------------------------------------------
Jodi Davenport <jodi@mit.edu>
Research or Project suggestion:
Develop a full model of pigeon performance in recognizing scenes. Compare this with actual recognition performance.
Short answer ?:
What is a polymorphic recognition scheme?
Mult. Choice ?:
Which of the following can pigeons NOT discriminate?
a. Letter "A"s and number "2"s.
b. Different types of oak leaves.
c. Scenes with trees from scenes without trees.
d. Paintings by Picasso from paintings by Monet.