The primary aim of a scientific investigation is to find the most like
ly model for a situation out of a host of alternative explanations. Th
e strength of evidence provided by anecdote, logical argument, mathema
tical modelling, observation, and designed studies (manipulative and o
bservational) are discussed and the effectiveness of randomisation and
orthogonal designs in separating hypotheses compared. Pseudoreplicati
on is shown to be often misunderstood. It consists of two concepts: th
e importance of adequate replication and the independence of the sampl
ing units. While replication is necessary to separate out the effects
of different factors and to provide an error term for inference, contr
ary to popular belief independence of sampling units is not necessary.
Finally the interpretation of evidence is discussed and the distincti
on made between formal and informal generalisation to a population.