Press perturbations, in which one or more species densities are experi
mentally altered and held at higher or lower levels, are common field
approaches used to understand community dynamics. The outcomes of such
experiments are often difficult to anticipate solely on the basis of
intuition. This is because the effects of a perturbation may pass thro
ugh a complex network of direct and indirect pathways in a food web, a
nd the outcome may be highly sensitive to the strength of interactions
among species. One solution to understanding outcomes of press experi
ments is to quantify first the community matrix, the matrix of measure
d direct interactions between all species in a food web, and then obta
in the inverse of this matrix. The inverse of the community matrix pre
dicts the effect of all species presses on all other species. I evalua
ted the utility of the inverse community matrix in predicting the outc
omes of press experiments in an old-field food web. I used data from f
ield and laboratory experiments to quantify the interaction strengths
between grasshoppers, four old-field plants, and nitrogen supply. Thes
e values were used to parameterize the community matrix and obtain its
inverse in a Monte Carlo simulation. The simulation was used to predi
ct the mean and standard error in the outcome of a simultaneous nitrog
en and herbivore press on food-web structure and dynamics. The predict
ions were compared with data from an enclosure experiment in the field
in which I manipulated nitrogen supply and herbivore abundance. There
was a high degree of uncertainty predicted and observed in the study
system. Despite this, I show that the degree and the sources of uncert
ainty were predictable for each species. This suggests that the invers
e community matrix offers a useful theoretical benchmark for understan
ding the outcome of field press experiments.