Why shoul a science experiment be replicable?

Why should a science experiment be replicable?

The basis of science knowledge is in its consistency, and that means reproducibility. An experiment performed the exact same way under the exact same conditions should always produce similar results. An experiment that can only performed once is essentially an observation of what happened that one time. Conclusions and understanding drawn from an unreproducible experiment, no matter how well it is assumed to be understood, can't refute the possibility that the results are totally different performed by someone else.

Take gravity as an example. A simple experiment is dropping something. Everyone on the planet earth can drop an object and confirm the behavior - it falls. This establishes knowledge - things appear to fall towards planets. However, this simple experiment is not well designed. The acceleration measured will differ from place to place on the planet and without scientists defining their altitude, the density of the crust and mantle below them, the phase of the moon, the amount of air resistance etc., they will have all differing measurements and be unable to explain why their results are differing by small fractions.

Also, there is little understanding behind gravity and thus the variables are not all accounted for. Who would know from their everyday experience would propose the following theories: space is curved around particles with mass; certain atomic particles are responsible for gravitational fields; gravity was different during the beginnings of the universe. Gravity, something felt by everyone, is a theory.

The scientific process is the method by which a scientist discovers and describes the universe around him. An experiment is the process used to confirm or disprove a hypothesis. An ideal experiment takes into account all variables that would effect the outcome, leaving only one variable the question is concerned with in doubt. The comparison of an isolated variable with experimental results is the foundation of consistency, which assists in understanding. Unfortunately, there is no such thing as an ideal experiment, and only by constant observation and confirmation of consistency can uncertainties be minimized and mistakes caught.