(FORTUNE Magazine) – When is it irresponsible to spend too little on research?

Most corporate executives will give you a blank stare if you ask them this question. That's because they probably think it is irresponsible to spend anything on research. Ask similar questions about other business investments, however, and you'll get better answers, with references to such things as ROI and EVA--metrics that help CEOs and CFOs get a quantitative handle on their companies.

Researchers don't help the matter. They market themselves very poorly. If you ask a researcher what the benefit of his work will be, you'll likely get an answer couched in technical jargon and wiggle words. The thing is, he's right. The more fundamental and important the research, the more unpredictable the results. Great work often stems from inauspicious origins.

No wonder many companies throw up their hands when it comes to research. Some just don't do it. Others treat research as a kind of philanthropy. Even the government is downsizing basic science, or at least trying to predict the market by funding applied, rather than basic, research.

Perhaps this is the best approach--but I wouldn't bet on it. The technological age in which we live is built on decades of basic research.The tail end of the 20th century is a period of enormous technological turmoil for many industries. If ever there was a time to invest in innovation, this would seem to be it.

I believe research can pay for itself, like any other business investment. Done well, it is an investment that will create an enormously high return--the foundation for long-term growth.

Basic research isn't just for high tech. Consider Procter & Gamble's leadership in detergents and disposable diapers, or Gillette's edge in shaving. Each rose to the top on the strength of R&D investments that rival those of any Silicon Valley firm.

Even academic research pays, albeit to society rather than shareholders. Perhaps the most dramatic example is the ultimate ivory tower--the Institute for Advanced Study in Princeton, N.J., where I've spent time, first as a graduate student and these days as a trustee. The IAS was founded in 1930 by the family behind Macy's and Bamberger's department stores. Its mission is the pursuit of knowledge for its own sake, and its staff has included the world's deepest thinkers, like Albert Einstein.

Fifteen years after IAS opened its doors, computer architecture as we know it--as featured in every Pentium, Cray, and digital watch--was invented there. So was game theory, a powerful tool in the economics of markets. The IAS didn't pursue these ideas commercially and instead let the world run with them. We're collectively richer for these generous gifts of intellectual property, but make no mistake: If it were a business, the institute's ROI would be very good indeed.

But who could have predicted that return? I believe that with careful study, the risks in research can be understood.

The output of research has a wide distribution. At one end of the scale are major breakthroughs; at the other, good ideas that yield small improvements. In the middle are many minor to medium breakthroughs. A collection of brilliant researchers can't help but produce results at all scales. This distribution of research represents what is called a scale invariant curve, a form of "self-organizing criticality," which describes a system first discovered in the context of the physics of sand dunes.

If you pour sand into a pile, it mounds up till it reaches an uneasy equilibrium, known to physicists as a critical state. Add a little more sand--even one grain--and an avalanche starts. Avalanches of all sizes occur, from the movement of a single grain to a chain reaction that shifts the entire pile.

The study of these avalanches shows they are inversely related: If you know how often an avalanche of a certain size will occur, then avalanches twice this size will occur half as often.

Cutting-edge research results also appear to constitute a self-organizing system. New results cause flurries of activity, not unlike the avalanches in the sand pile. It may even be possible to predict the size vs. frequency relationship, and thus relative return on major breakthroughs vs. short-term work. In fact, Microsoft now has a research project aimed at quantifying this.

Even without precise results, it's easy to see that research programs should be designed to take advantage of discoveries of all sizes. Surprisingly few are set up to do so.

Some research labs are isolated from the rest of corporate engineering, sent off to think deep thoughts in the piny woods. Only a major breakthrough could bridge this isolation gap--getting small ideas from research to production is impossible. And when the Big One does occur, production usually isn't ready because there is no ongoing connection to research.

Other labs insist on timid research programs focused on short-term incremental work. I think it's clear that you get more small ideas by thinking big than by thinking small.

Most executives will tell you that a company living one quarter to the next can't afford to invest in long-range research. That seems like common sense, but I think the converse is true: Any company that can afford long-term investments is foolish not to invest in innovation and take an active hand at inventing its own future.