A simple physics experiment has shed light on slithering - the most common type of motion used by snakes.
Rather than pushing off nearby obstacles, snakes exploit the fact that their scales have different "grip" in different directions.
They also boost speed by lifting curved parts of their bodies off the ground, redistributing their weight.
The research is published in the Proceedings of the National Academy of Sciences.
Since the first theories of snake locomotion arose 70 years ago, the assumption has been that they "push off" objects or irregularities in their paths such as plants or outcrops of rock.
Snakes on a plane
Researchers at New York University and the Georgia Institute of Technology in the US have challenged that theory using an experiment lifted straight out of a physics textbook.
The experiment measures a snake's "coefficient of static friction" - that is, how much its body grips a surface when at rest.
Anaesthetised snakes were pointed in different directions along a flat board held at an angle relative to the ground. By lifting the board until the snakes began to move, the reptiles' friction coefficient could be determined.
A snake facing down the board towards the ground moved easily when held at a shallow angle. A snake lying sideways gripped more, requiring a higher angle before moving, and a snake facing uphill required the highest angle.
The researchers say this clearly points to a mechanism by which the snakes use scales on their bellies as pushing off points.
"What people thought was going on in the macroscopic scale with the sides of their bodies is actually going on in the microscopic scale, with their belly scales," said David Hu, an applied mathematician at the Georgia Institute of Technology.
"If the friction were equal in all directions, a snake would just slither in place, as if it were on a treadmill," he added - a possibility that the team tested by putting snakes on a smooth surface on which they could gain no traction.
As the snakes came out of their anaesthetic daze, the researchers noted that individual scales twitch, confirming the idea that each scale can be independently controlled for maximum traction.
The team then developed a simple mathematical model to describe how the snakes would push forward using just the friction provided by their scales.
However, the model came up with maximum speeds lower than those observed in live snakes.
The team then began to look at how the snakes distribute their weight as they slither.
"Most people think that when snakes slither they are completely pressed flat against the ground, but actually they lift their bodies," Dr Hu explained.
"Sometimes it's perceptible and sometimes they just unload parts of their body so their weight distribution is only towards the centre."
The team went on to use a visualisation scheme involving polarised light and gelatine, which changes its optical properties under pressure.
The gelatine showed that the snakes indeed concentrate their weight where their bodies were least curved.
Fitting this behaviour into their model, the team found that not only did it predict the speeds that real snakes achieved, but also that the movement was 50% more efficient.
The weight re-distribution eliminates a lot of wasted slithering effort because they lifted the parts not contributing to forward motion.
"It's analogous to the way we walk or run," Dr Hu said.
"You shift your weight to the left or the right leg, but you don't drag the other leg. The snakes are only touching the ground where the friction force is going to help them move."
The work is particularly relevant to efforts in engineering snake-like robots, such as the search-and-rescue robots made in the Hirose Fukushima laboratory at the Tokyo Institute of Technology.