approach effective also expose such weaknesses as the potential for a bias that depends on the citizenship of the panelists who gather data for analysis. This increase the importance of including substantial proportions of non-US participants in all panels.
Multidisciplinary fields like materials science and engineering and immunology pose special challenges. For example, the immunology panel had to extract data from collaborative and international research; had to compare large enterprises with multiple smaller ones; and had to extract information on the specific field of immunology from related research fields in large, aggregate databases.
The method by which the most important fields and subfields are identified is critical. For example, immunology is not considered a "discipline" in the traditional sense and does not have departmental status in most universities. The selection of subfields is a somewhat subjective process that might differ between one benchmarking exercise. Rather than being a drawback, however, such differences will reflect the continual shifting of the borders of modern fields. A field should be considered by the array of related domains between which investigators can move without leaving the realm of their expertise.
It is likely that benchmarking could be effective on a 3- or 5-year cycle because large fields of research change relatively slowly. Annual benchmarking probably would not be sensitive enough to reveal changes.
Our series of experiments has revealed that no benchmarking technique is sufficient by itself and that the utility of particular techniques varies by field. Therefore, each panel should use a variety of comparable qualitative and quantitative methods to afford cross-verification of results. The methods should be kept as independent as possible.
Because the accuracy of benchmarking depends heavily on panel members' personal knowledge of fields, panel members were more closely involved with the writing of the report than is frequently the case with committee-written reports.
Use of indicators that provide information on degree of uncertainty and reliability might enhance the presentation of the panel assessments of leadership status.
The extensive use of benchmarking would be enhanced by reliable, up-to-date information. The US field-specific data that are collected do not provide sufficient or timely information; non-US data are even more problematic.
A finding that the United States is the world leader in a research field might lead some to conclude that additional resources for that field are not warranted. This might or might not be the case. For example,