Page:Sm all cc.pdf/155

 can trust its subconscious or casual use. For those who are mastering evidence evaluation, and even occasionally for experienced scientists, evaluation aids are useful. Here we describe three such techniques: model/observation tables, outlines, and concept maps.

A model/observation table succinctly compares several competing hypotheses. Usually various observations are relevant, with some favoring one idea while others favor another. Ideally, the scientist would examine each hypothesis systematically, reject those refuted by one or more evidence sets, and conclude that only a single hypothesis survives unscathed. In practice, we generally must weigh many inconclusive and partially contradictory data. The challenge to the scientist is to consider simultaneously this variety of evidence; a model/observation table is one way.

The model/observation table is a specialized and somewhat qualitative version of a truth table: list the models (or hypotheses) horizontally, list the relevant observations vertically, and then symbolically summarize the consistency of each observation with each model. Select symbols that are readily translatable into position along the continuum from strong confirmation to strong refutation:

+: strong confirmation +: weak or ambiguous confirmation 0: not relevant, or no data available(alternatively, use a blank if ‘0’ implies ‘no’ to you) -: weak or ambiguous refutation --: strong refutation.

Table 11. Example of a model/observation table.

For example, Table 11 summarizes the consistency of four published models with a group of five experimental findings. A quick scan of this table permits us to see that the leading hypotheses are those of D [1980] and of D & E [1981]; the latter is somewhat more successful but not decisively so. The hypothesis of A [1975], though consistent with many observations, is refuted by the observation of no y/z correlation. The hypothesis of B&C [1976] has mixed and unimpressive consistency with the observations. This quick overview allows identification of which observations are the most useful and consequently warrant the most careful attention. For example, the observation that x<w obviously is of no help in distinguishing among the possibilities.

The model/observation table is an easy way to focus one’s attention onto the most diagnostic relationships among observations and hypotheses. It counteracts the universal tendency toward letting one relationship dominate one’s thoughts. It encourages systematic evaluation of all relevant types of evidence. The table is not meant to be a simple tabulation of consistency scores, resulting