Washington: You make a decision based on a "lucky guess" which feels right, but that may not be guesswork at all, says the latest research.

A Northwestern University study offers precise electrophysiological evidence that such decisions may be due to the surprising accuracy of memories that can't be consciously accessed.

During a special recognition test, guesses turned out to be as accurate or more accurate than when study participants thought they consciously remembered.

"We may actually know more than we think we know in everyday situations, too," said Ken Paller, professor of psychology at Northwestern.

"Unconscious memory may come into play, for example, in recognising the face of a perpetrator of a crime or the correct answer on a test. Or the choice from a horde of consumer products may be driven by memories that are quite alive on an unconscious level," he said.

The study links lucky guesses to valid memories and suggests that people need to be more receptive to multiple types of knowledge, Paller said.

Paller and Joel L. Voss, who received his Ph.D. at Northwestern and is now at the Beckman Institute, are co-investigators of the study.

During the first part of the memory test, study participants were shown a series of colourful kaleidoscope images that flashed on a computer screen. Half of the images were viewed with full attention as participants tried to memorize them.

While viewing each of the other images, they heard a spoken number, such as 3, 8 or 4, which they had to keep in mind until the next trial, when they indicated whether it was odd or even. On every trial they had to listen to a new number and press a button to complete the number task.

In other words, they could focus on memorising half of the images but were greatly distracted from memorising the others. A short time later, they viewed pairs of similar kaleidoscope images in a recognition test, said a Northwestern release.

"Remarkably, people were more accurate in selecting the old image when they had been distracted than when they had paid full attention," Paller said. "They also were more accurate when they claimed to be guessing than when they registered some familiarity for the image."

These findings were published online in Sunday edition of Nature Neuroscience.