The Bad News about Discovery Learning:
A Nugget from the Annals of Research
Michael Marshall raises chickens, but still can’t tell their sex at a day old.
From Common Knowledge, Volume 11, Number 3, 1998.
© 1998 Core Knowledge Foundation. Not to be copied or reproduced without permission from the Core Knowledge Foundation, 801 E. High Street, Charlottesville, VA 22902.
What’s the Fastest Way to Learn to Tell the Sex of a Chick?
In an experiment designed to test perceptual skills, the use of an expert’s crib sheet describing the key features that identify the sex of a chick, a notoriously mysterious art, rapidly turned novices into experts, demonstrating that a skill is most effectively learned when a proper content base is first established.
The 1987 experiment,* conducted by Irving Biederman, now a professor of cognitive psychology at the University of Southern California, and Margaret Shiffrar, rebuts the central assumption of discovery learning, the notion that children retain information best when they must acquire it on their own, through persistent effort rather than by having it presented to them verbally.
Professional chick-sexers can sort 1,000 day-old chicks per hour, examining each for less than a second, with 98 percent accuracy. Under bright light and magnification, chicks are grasped in the left hand, vents concealing their genital area are forced open briefly and a judgment is made of the curvature of an eminence called the “bead ”that is about the size of the head of a pin. If the bead appears convex, the chick is male, and if concave or seemingly flat, the chick is female. Years are spent to learn this skill, which is commercially vital to the poultry industry. Male chicks have no economic value and are sacrificed before any expense is incurred from them. Secondary sex characteristics such as the rooster’s comb only appear a month after hatching. The experiment sought an answer for the question of how a perceptual skill so difficult to acquire could be performed with such a high degree of speed and accuracy.
A chance to test expert versus novice skill was suggested to Biederman in 1984 when he saw a newspaper story about the retirement of Heinz Carlson of Petaluma, California, a professional chicken sexer who had sorted 55 million chicks over his 50 years on the job.
The study revolved around a rare set of 18 photographs of the genital regions of pullets and cockerels. Taken by the Poultry Science Association in 1941,the photographs are probably the only ones to exist that show ambiguous specimens of the bead that challenge the discriminating powers of even expert chicken sexers. Periodically over two decades, Biederman and other cognitive scientists studied the photographs, attempting to identify perceptual markers that could be used to distinguish the sexes, and finally admitted failure.
The perceptual basis for telling male chicks from females was discovered in the late 1920s by the Japanese. Since it inspects for a concave or convex outline, the method has the useful characteristic of being perceptually consistent from any angle of view. In 1934 a Japanese delegation to British Columbia demonstrated the method to American and Canadian hatcheries, and schools for teaching the technique were soon set up in Washington and California. Carlson was one of the first graduates of the California school.
Because of the similarities of the male and female bead, it was obvious that novice sexers would not know where to look in the pictures or what to look for. Carlson taught the researchers the method using the photographs.
To compare their performance before and after instruction, untrained subjects were shown the photographs and asked to identify the chick’s sex. Their average success rate was 60.5 percent, or correct on nearly 11 out of 18.The slightly better than 50/50 chance score was attributed to the subjects’ tendency to judge any chick whose bead they considered to be prominent as likely to be male. A group of professional sexers shown the photographs averaged a score of 72 percent correct, a rate below their usual proficiency which two factors were thought to influence: first, that they could not inspect a live bird and, second, that the photos were of rare and ambiguous cases in the first place. One particular chick was wrongly judged by every professional, for example.
Half the naive subjects were then shown a short instructional sheet and given one minute of explanation of where and what to look for. They were then asked to sort a new random arrangement of the photographs again. The other half of subjects made a second sort with no instruction or feedback from their first effort. Those receiving instruction were now 89.9 percent accurate. The average score of those not receiving instruction actually fell to 54 percent.
“More important than the improvement in the performance of the instructed subjects was the fact that their item analysis more closely resembled that of the professionals as a consequence of instruction,” said Biederman. “In other words, after training, the pictures the nonprofessionals missed tended to be the same ones missed by the professional sexers.
“We demonstrated that a considerable portion of the visual learning in the classification of the pictures could be achieved by a brief instruction that merely described a nonaccidental contrast in shape.”
By extension, he said, more rapid learning of rare items could be expected from a deliberate study of unusual cases. “It seems plausible that an album of such [extremely rare] instances could lead to much more rapid mastery of these configurations by overcoming problems of rarity and allowing immediate comparisons with similar types. A disproportionate allocation of effort to rare events is characteristic of learning a number of skills that require extremely high accuracy levels, such as diagnostic medicine or piloting planes,” he said.
As for the implications his experiment holds for the tenets of discovery learning, Biederman said, “There is no empirical evidence to support the idea of discovery learning. My own view of it is that it is grossly inefficient. Look, it took 2,000 years to figure out chicken sexing.
“Of course there are some things you can learn for yourself; for instance, it is important to work through a mathematical proof for yourself. But knowledge requires so much supporting structure that it can’t be acquired readily without first building the content base. In the case of learning how to sex chickens that means being given an idea of where to look and what to look for. Just having someone try to discover something means you seldom get any discoveries.”
The notion that discovery learning promotes critical thinking is likewise flawed, he said. Thinking invariably come back to matters of content, generally how much one has at his or her command. “Imagine that a fourth-grader working a math problem on the blackboard is having trouble and is told to ‘think’about it. When you ask people what happened when they were told to think, they answer that their mind went blank. There is nothing a person can do when told to ‘think.’ What that amounts to is a command to ‘try harder’and trying harder means getting deeper into the content itself to get more information to apply to the problem.”
*Journal Of Experimental Psychology: Learning Memory and Cognition.1987, Vol.13, No.4, 640-645
 |
 |
 |
|
||
 |
|
 |
|
||
Last updated: Fri, March 14 2008
