paper, many researchers and laymen believed that working memory's capacity was about seven items," Cowan says."Other researchers believed the smaller figure of about three or four items.Still other researchers were perplexed by the discrepancy and concluded that the capacity ‘just depends' on many factors."
The 7-item memory model originated with psychologist George Miller in 1956, says Rouder. "Miller noted that in some simple tasks, people's performance would decline if more than seven items were presented." But that model began to falter when researchers embraced "chunking," the idea that the memory of seven items is actually a memory of several chunks (such as chunks of three, two and two items).
Chunking can remove the strict limits from working memory. "We remember more than seven items when we can ‘chunk' two items into one new item," says Rouder.
People who memorize pi to hundreds of places are nothing more than good chunkers, Cowan adds. Most forms of mnemonics, the common memory aids beloved of elementary school teachers everywhere, are nothing more than good chunking.
As Gibson sees it, Cowan and his colleagues "have been arguing for some time that the estimate of 7-item retention was too high and was probably achieved in previous studies with the help of various strategies like chunking. One could argue that when strategies like chunking or rehearsing an answer that might inflate the capacity of working memory are prevented, the limit is lower -- more like three or four objects."
Confirmation of the four-item memory model came about 10 years ago, when psychologists Steve Luck and Ed Vogel "published a very influential paper in which they asked if there was a basic-capacity limit in immediate consciousness."
Their approach, which Rouder calls "the gold standard for assessing the number of basic items in consciousness," was to flash five differently colored squares for about 100 milliseconds to study participants in a series of changing displays -- plenty of time for working memory to kick in, but not enough time to formulate an educated guess.
"People had to indicate whether the display was the same or had changed," Rouder explains. "Performance was perfect if the number of squares was small." But with a larger number of squares, he adds, performance declined dramatically.
Using a variant of the same approach, Rouder and Cowan confirmed not only the limit of three or four items, but also its all-or-nothing nature.