is almost always a false economy, unless the new apparatus or experiment is only a slight refinement of a familiar one.
The experimental prototype is a routine step in applied science, where it provides a bridge between theory and routine practical application. Applied science usually has two scales of prototype: laboratory prototype and then pilot plant. Only after both have been tried does a company decide whether commercial production is viable.
The prototype is a less common step in basic research, although some of the same factors that encourage its use in applied science apply to basic science. The prototype allows feasibility study, detection of practical problems, and improvement of design parameters. It also permits identification and isolation of unanticipated factors that could affect the success or accuracy of the experiments. Consequently, far different standards and objectives apply to the prototype than to the final apparatus:
- The prototype is much less expensive and time-consuming to build than the final apparatus. Often it is smaller, less robust, and less streamlined.
- The prototype is much more versatile than the final apparatus. Modification is easier, parts can be exchanged, and ‘quick and dirty’ fixes are acceptable.
- Depending on the type of measurement, the prototype may have a smaller or larger range of measurement values than the final apparatus will need to have.
- Precision and accuracy may be lower on the prototype, and both are improved as problem variables are isolated. The prototype is not necessarily well calibrated, because we are probably more interested in sensitivity analysis than in accuracy of these results.
- Measurements may be more cumbersome and slower on the prototype than on the final apparatus.
A prototype is likely to be needed whenever equipment design is substantially modified. It can even be a worthwhile time saver when one is building equipment or an experimental setup similar to published ones and depending on rather scanty published details. It is better to discover that the author left out a ‘slight complication’ when trying out a prototype than in the midst of an experiment.
The pilot study is the procedural analogue to an equipment prototype, and many of the considerations above apply equally well to pilot studies. Different standards [Beveridge, 1955] concerning variables and their control apply to pilot studies than to the formal experimental series:
- One can use extreme values for a variable in the pilot study to see if they have any effect. If they do seem to have an effect, then the formal experiment can focus on the range of most interest for this variable. At that stage, higher accuracy and precision probably will be required. Indeed, statistical analysis of the pilot data can indicate how many measurements will be needed to detect the effect (Chapter 2).
- In some situations, many variables could have a significant effect, and it is not obvious which needs to be incorporated into design of the formal experiment. One can lump many variables in the pilot study. Of course, some caution is needed to prevent cancellation of the effect of one variable by the opposing effect of another. This approach is most effective if one knows the probable direction of the potential influence of each variable. If a composite effect is found, then formal experiments can be designed that will systematically isolate the effects of each variable.
