Readers Elaborate on Fashion and Truth, Fact and Theory

Highlighted in Michael Riordan’s Opinion piece is the danger of relaxing the criteria for what constitutes scientific fact. He is, however, in danger of blunting a valuable new tool of science when he identifies computer experiments as part of the problem rather than part of the solution.

Fifty years ago, Enrico Fermi, John Pasta, and Stanislaw Ulam invented the computer experiment and predicted recurrence in nonlinear systems. They programmed the early MANIAC computer at Los Alamos Laboratory to simulate an array of 64 weakly coupled nonlinear oscillators. The researchers expected the array to relax into a random equipartition of energies. Instead, it periodically returned to the starting condition. Fermi affectionately referred to that phenomenon as a “little discovery.” ¹ Since then, Fermi-Pasta-Ulam recurrence has been experimentally confirmed and has become a key concept in understanding the behavior of complex nonlinear systems.

A few years ago, NSF Director Rita Colwell gave a talk in which she referred to simulation as “the third branch of science.” ² She based that statement on the use of computer simulation in fields such as astrophysics and Earth sciences, where system complexity prevents evaluation of theoretical predictions by any means other than computer simulation. In those fields, computer simulations bridge the gap between theory and experiment for complex nonlinear systems so that the theoretical predictions can be compared far more precisely to experimental results. Without simulations, approximations must be used, which limit accuracy and introduce unknown errors into predictions.

Thus, computer modeling and simulation are primarily theoretical tools. A powerful adjunct for the theorist, they provide additional predictions but never replace experiment. For example, the numerical predictions of gravity-wave emission from merging black holes are beyond analytical check and will only be confirmed with data from the Laser Interferometer Gravitational-Wave Observatory (LIGO) and other experiments. Modeling and simulation can point to new directions for both experimental and theoretical investigation, so they truly merit being called the third branch.