Theory, Model, Method
Data and Theories
The ways linguists may build their theories on the basis of data and use correspondent methods may be different.
The first view of a linguistic way of approaching data and building theories was held by many workers and is unfortunately propagated by teaching scientific topics in schools and in popular accounts of science in the media. According to this view, a scientist must begin by collecting observations or produce data by experiments. After he has made a large and sufficient number of such observations or experiments, he proceeds to a generalization (see above) about these data. This generalization is expected to be supported by the original (given) data. After several attempts at generalizing he may proceed to a new (modified) hypothesis by looking at new data. The modified hypothesis should cover the old original data and the new data. This way of arriving at hypotheses is called inductive. The most typical inductive approach in linguistics is found in American Structuralism with its Discovery Procedures. It is based on the assumption that one has to start without any pre-knowledge or pre-conception about the linguistic object under examination.
The other, more accurate account of scientific method is the following. The scientist has some ideas, some knowledge, or may be interested in a problem (including some knowledge) as the input to his theory–construction. How he comes to that knowledge is of no theoretical consequence or importance. The scientist then formulates a first working hypothesis as a tentative answer to his problem. A good hypothesis is based on common Scientific standards. It is then tested against a collection of observations or experimental data and might be modified on this data basis. This way is called deductive insofar as it assumes that the hypothesis is derived (deduced) from already existent knowledge and then tested by empirical data.
Exercises on Data and Theories