Hands, not eyes, hold clue to illusion Seeing is believing, but acting is more

accurate. That’s what two Canadian psy- chologists conclude after investigating the size-weight illusion, an error that aris- es when people try to estimate the weights of two objects of different sizes but the same mass.

J. Randall Flanagan and Michael A. kltzner of Queen’s University in Kingston, Ontario, presented 40 people with two square boxes topped by handles, each unit weighing 0.39 kilograms. One box measured 5.2 centimeters on each edge; the other, 10.9 cm. As expected, all volunteers guessed

that the larger box was heavier before they touched the cubes. After lifting each box in turn 20 times, using the tips of their thumb and index finger, all reported that the smaller box was heavier. Anoth- er group fell prey to the illusion even when told before lifting the boxes that they were the same weight.

The usual explanation for the illusion, which has appeared in psychology jour- nals since the 1890s, is that it’s due to a mismatch between expectations and sensory feedback. “You might expect a large object to be heavy,” says David Mil- ner, a psychologist at the University of Durham in England, “and when it isn’t, it seems startlingly light.” It’s similar, he suggests, to lifting a halfempty suitcase.

Flanagan found this mismatch hypothe- sis unsatisfying, however, because it doesn’t specify how a discrepancy in feedback could become a perception. He used sensors on the boxes’ handles to measure the lifting force and grip strength that the volunteers applied-learning not just what they saw but what they did.

Initially, the participants used more power to lift the larger box. After hoisting the boxes 5 to 10 times, however, they applied equal force to the boxes. Yet de- spite this correct sensorimotor adapta- tion, the cognitive side of the illusion per- sisted. The volunteers still said that the smaller box was heavier, the researchers report in the July NATURE NEUROSCIENCE.

“I’m happy that this happens,” says Melvyn A. Goodale of the University of Western Ontario in London. In the early 199Os, he and Milner advanced some of the theories of visual processing underly- ing Flanagan and Beltzner’s study.

Milner and Goodale reinterpreted work on visual processing that identified a “what” pathway concerned with labeling objects and a “where” pathway con- cerned with locating them. Instead, Mil- ner and Goodale suggested a division be- tween “vision for cognition” and “vision for action.”

Adjustments to grip and force, says Milner, are examples of vision for action. They deal with information conveyed by the immediate scene. Judgments about object weight, in contrast, draw more

heavily on an integration of the visual scene with cognitive comparisons among weights of familiar objects, he says.

From the primary visual cortex, these information streams diverge. Vision for action uses a dedicated path to the brain’s posterior parietal cortex, where sensory information is processed. Vision for cognition takes a route to the infer- otemporal region, which supports more- abstract cognitive functions.

Even though there’s interaction be- tween the two streams, says Flanagan, information in one isn’t necessarily avail- able to the other. This prevents the cog- nitive system from accessing sensorimo- tor learning. Researchers have reported similar disjunctures in other optical illu- sions (SN: 2/14/98, p. 108).

Outside of experimenter-contrived worlds, Goodale says, the assumptions un- derlying size-weight and optical illusions “are correct 99.9 percent of the time.”

Moreover, these judgments’ enduring utility has given them staying power.

Two boxes of the same weight, one shown with sensors to measure litling force.

Only temporal lobe damage has, so far, made the cognitive side of the illusion disappear.

“I’ve sort of fantasized about experi- ments,” says Flanagan, hastening to add that he hasn’t actually done them, “where you would put people in a room where everything was inverted. So, the chester- field would be the lightest thing, and your cutlery would be the heaviest.” Maybe after a week in there, he says, you could make the illusion go away. -R. Bennett

Silencing a gene slows breast-tumor fighter Just as every actor needs a director to

shout “Action,” each gene in the body needs a promoter-nearby DNA that switches on the gene at appropriate times. The promoter’s cue can take on life-or-death importance when the gene is p53, a tumor suppressor.

Researchers now report that a protein that activates the p53 promoter is fre- quently missing in breast cancer cells. Of tumor samples taken from 30 women with breast cancer, 20 were devoid of the protein, called HOXA5, says study coau- thor Saraswati Sukumar, a molecular bi- ologist at the Johns Hopkins Medical In- stitutions in Baltimore.

The p53 gene itself encodes a protein that signals cells to commit programmed suicide, or apoptosis, when their growth patterns run amok, as in cancer. Try as it may, however, p53 can’t always play the hero. In at least onefifth of breast cancer cases, for example, p53 has been mutat- ed and its protein rendered ineffective. Sukumar also suspects that p53 is si- lenced in some breast tumors.

To ascertain whether HOXA5 plays a vital role in activating p53, Sukumar and her colleagues infused either functional or defective HOXA5 genes into breast- tumor cells in laboratory dishes. The working genes induced cell death, where- as flawed HOXA5 genes permitted the tumor cells to keep thriving, the re- searchers report in the June 22 NAW.

HOXA5’s cancer-fighting capabilities may not be limited to activation of the p53 promoter, Sukumar says. The protein also spurred production of another tu- mor-suppressing protein, encoded by the

p21 gene. “We have a feeling that more genes are [switching] on,” she says.

Further experimentation suggested a mechanism by which the HOXA.5 gene might come to be disabled in cancer patients. The researchers took breast cancer cells lacking p53 protein from 20 patients. In 16 cases, they found that hy- drocarbon fragments, or methyl groups, had latched onto the promoter region for the HOX45 gene.

While methylation is a common occur- rence in DNA throughout the body, the promoter DNA next to the HOX45 gene “is normally not a methylated region,” says Sukumar. If methylation has si- lenced HOX45, the gene couldn’t direct a cell to produce the protein needed to ac- tivate p53’s promoter, she concludes.

Meanwhile, none of the breast cells from 10 healthy women showed a meth- ylated promoter for HOX45, she says.

“This . . . is potentially a very interest- ing finding,” says Louise C. Strong, a can- cer geneticist at the University of Texas M.D. Anderson Cancer Center in Houston. Scientists are now examining the effects of removing methyl groups from portions of DNA. By demethylating HOX45 pro- moter DNA, scientists might be able to fight breast cancer by activating p53 in patients, she says.

HOAX5 belongs to the Hox family of genes, which guide formation of certain body parts. Sukumar notes that the new study is the first to link a Hox protein to apoptosis. Future research might delve into ways of delivering functional HOX45 genes to tumors in patients lacking the protein. -N. Seppa

JUNE 24,2000 SCIENCE NEWS, VOL. 157 407