Selective attention and Pavlovian conditioning.

The present results show that the common practice of using self-indexing conditioned stimuli (CSs) in research on Pavlovian conditioning is a major source of experimental bias. The typical stimulus used is either a light flash or a sound pulse in a light/sound-shielded chamber. Under these conditions the onset characteristics of the CS signal totally predominate over the durational characteristic, i.e. the pattern information. Thus a visual pattern presented as a CS in a dark chamber is confounded with a brightness change from darkness to light. In the first experiment, animals were conditioned with a brightness CS using a self-indexing signal paradigm. When tested for specificity of the conditioning, they showed complete transfer of learning to either a visual pattern or even an auditory CS. These findings indicated that the traditional conditioning paradigm is biased towards non-specific sensory learning. The second experiment showed that specific sensory conditioning is critically dependent on selective attention mechanisms. When the onset characteristics of the CS signal were de-emphasized by the use of equal energy background illumination in the intertribal interval (ITI) during conditioning, the animals were not able to feature extract either the onset or the durational component of the CS signal from the ITI background despite prolonged training. It was only by starting with conditioning that was initially anchored to the CS onset characteristics that a perceptual fade-in procedure would bias attention to feature extract the durational characteristics of the CS. Thus conditioning occurred only when the rabbit's attention was directed to detection of the gratings display without any associated changes in visual albedo. Perhaps the most important finding of the present experiments is that the use of self-indexing CS signals in Pavlovian conditioning inevitably introduces non-specific sensory processing involving multiple sensory input pathways in the conditioning. This inherent uncertainty of the sensory input pathways presents a problem for clarifying the role of sensory pathways in the neural mechanisms of NM conditioning. In addition, the use of self-indexing CSs inevitably leads to an underestimation of the role of forebrain mechanisms in Pavlovian conditioning.