However, one last caution. What has happened in Japan for 12 long years warns that an affluent society like America’s might not be immune to a self d… - Paul Samuelson

In the preface to the reissue of Risk, Uncertainty and Profit, Frank Knight makes the penetrating observation that under the conditions envisaged above the velocity of circulation would become infinite and so would the price level. This is perhaps an over-dramatic way of saying that nobody would hold money, and it would become a free good to go into the category of shell and other things which once served as money. We should expect too that it would not only pass out of circulation, but it would cease to be used as a conventional numeraire in terms of which prices are expressed. Interest bearing money would emerge. Of course, the above does not happen in real life, precisely because uncertainty, contingency needs, non-synchronization of revenues and outlay, transaction frictions, etc., etc., all are with us. But the abstract special case analyzed above should warn us against the facile assumption that the average levels of the structure of interest rates are determined solely or primarily by these differential factors. At times they are primary, and at other times, such as the twenties in this country, they may not be. As a generalization I should hazard the hypothesis that they are likely to be of great importance in an economy in which there is a “quasi-zero" rate of interest. I think by this hypothesis one can explain many of the anomalies of the United States money market in the thirties.

As a theorist I have great advantages. All I need is a pencil (now a ball pen) and an empty pad of paper. There are analysts who sit and look vacantly out the window, but after the age of 20 I was not one of them. I ought to envy the new generation who have grown up with the computer, but I don’t. None of them known to me sits idly at the console, improvising and experimenting in the way that a composer does at the piano. That ought to become increasingly possible. But up to now, in my observation, the computer is largely a black box into which researchers feed raw input and out of from which they draw various summarizing measures and simulations. Not having access to look around in the box, the investigator has less intuitive familiarity with the data than used to be the case in the bad old days.

I think the acceptance of "mathematical expectation of utility" or its "arithmetic mean" was an unthinking carryover from the mathematical theory of the law of large numbers as applied to asymptotic processes.