CONCEPT
Experimental Neurosis
Pavlov's discovery that a conditioned system pushed past the boundary of its discriminative capacity does not return appropriate uncertainty but collapses into confident disorder—the original laboratory demonstration of what AI now calls confabulation.
The most disturbing episode in
Ivan Pavlov's laboratory occurred when he pushed his discrimination experiments past the point where they could succeed. He had trained a dog to salivate at a circle (rewarded) and not at an ellipse (unrewarded), and then progressively rounded the ellipse to make it more circle-like. For a while the dog kept pace, discriminating finer and finer distinctions. But at the boundary of what the animal's discriminative machinery could resolve, something broke. The dog's behavior—previously calm and orderly—collapsed into agitation: squealing, struggling against the apparatus, tearing at the equipment, and losing not only the fine discrimination but discriminations it had long mastered. Pavlov called the condition experimental neurosis. The conditioned system, asked to draw a distinction finer than it could draw, did not fail gracefully. It fell apart. What makes this discovery more than a curiosity is what it reveals about the structure of competence in any trained system. A system that genuinely understood the task would recognize