Feature

For years, a cadre of psychologists has believed that animals can understand the concept of numbers. But many remained unconvinced until Columbia University's Herb Terrace, PhD, and his graduate student Elizabeth Brannon did what some are calling the "definitive" study.

In it they find that rhesus monkeys easily learn to order pictures in terms of the number of items they contain. And, unlike any other study, the researchers find that the monkeys successfully transfer their ability to order pictures containing one to four items to correctly order pictures containing five to nine items.

The researchers published their initial round of data in Science in 1998 (Vol. 282, p. 746-749) and have just published the full article, including several new twists, in this month's Journal of Experimental Psychology (JEP): Animal Behavior Processes (Vol. 26, No. 1).

Even those who were already won over by earlier studies of animal counting are impressed with the experiments.

"What's novel about this article is that they used higher numbers than others have traditionally used and they showed that if you train animals on one through four they can respond to five through nine," says psychologist John Capaldi, PhD, of Purdue University, who has studied number ability in mice.

A well-defined category

Brannon, who joins the center for cognitive neuroscience at Duke University in March, conducted the studies to discover whether monkeys could learn rules for putting objects into categories, then apply those rules to a new set of objects. She and Terrace designed the experiments using a technique Terrace developed called "simultaneous training paradigm," which rewards animals for correctly touching objects in a specific order.

In their study, the researchers created displays with one, two, three or four abstract elements, such as circles, ellipses, squares and diamonds of various sizes and colors. In each training session, they showed the monkeys a computer screen that displays each of the four numbers. The monkeys' job was to touch each display in numerical order. Two of three monkeys, Rosencrantz and Macduff, learned to touch the displays in ascending order--the display with one element first, the one with two elements second, up to four. Another monkey, Benedict, learned to touch the displays in descending order.

Critics of past animal number studies often complain that researchers fail to control for mechanisms other than number that the animals might use to order stimuli. For example, animals might use total surface area to recognize that a display with four items has an overall larger surface area than a display with three items. Brannon and Terrace carefully designed their displays to rule out everything but number.

"They used some very clever manipulations to control for everything but number," confirms Capaldi.

Brannon and Terrace trained the monkeys on 35 displays until the animals were accurate 50 percent of the time at touching them in ascending or descending order. The researchers then tested the monkeys on 150 new displays, and their performance didn't falter, says Brannon.

Confirming the findings

But this finding wasn't enough to convince Brannon and Terrace that the monkeys truly understood the ordinal nature of numbers. They certainly recognized the difference between one, two, three and four items. But did they understand that two always lands between one and three, and three between two and four?

Instead, the monkeys may have learned to order the number images as they learn to order arbitrary sequences. For example, they might assign displays with three elements to a category "A," displays with four elements to category "B," displays with two elements to category "C" and displays with one element to category "D." Then they might learn to order those categories in the arbitrary order: D, C, A, B.

To rule out that option, the researchers tested the monkeys again, this time using pairs of displays that represented numbers one through nine. Some pairs represented only numbers with which the monkeys were familiar. Others represented one familiar number and one novel number--three and six, for example. And others represented two novel numbers--such as five and eight.

Rosencrantz and Macduff were tested to see if they could touch the pairs in ascending order, and Benedict was tested to see if he could touch them in descending order.

In the first round of testing, Rosencrantz and Macduff correctly ordered 75 percent of the novel number pairs without receiving any reinforcement for correct answers. Benedict ordered no more than chance correctly. In a second round of testing, which included reinforcement for correct answers, however, all three did well: Benedict raised his scores above chance on all the novel number pairs.

"This is one of the first instances where we're seeing expertise developing in monkeys," says Terrace. "We trained them for only six months on these numbers--imagine how they'll do in two to three years. Remember, it takes children learning numbers thousands of repetitions to get it right."

Brannon is working with researchers at New York University to compare the monkeys' performance with that of children who are starting to learn number concepts. That, she says, may help them understand how people and animals might think about numbers in the absence of language. For example, do monkeys and very young children count or perhaps use some kind of innate neural "accumulator" that keeps track of the concept "how many?"

Future implications

In the JEP study, the researchers have already found several similarities between monkeys and people on similar tasks. For example, the monkeys are more accurate and quicker to order pairs of numbers the further apart the numbers are--ordering three and nine is easier than ordering three and four. Researchers have found the same trend in people for years.

Terrace, Brannon and others are conducting more studies to better understand what these findings mean.

Meanwhile, "the results of these experiments provide compelling evidence that number is a meaningful dimension for rhesus monkeys," write Terrace and Brannon.

This article is part of the Monitor's "Science Watch" series, which reports news from APA's journals.