… is from page 219 of Thomas Kuhn’s 1977 collection, The Essential Tension ; specifically, it’s from Kuhn’s 1961 paper “The Function of Measurement in Modern Physical Science ” (original emphasis):
The road from scientific law to scientific measurement can rarely be traveled in the reverse direction. To discover quantitative regularity one must normally know what regularity one is seeking and one’s instruments must be designed accordingly; even then nature may not yield consistent or generalizable results without a struggle.
And in the social sciences – which study phenomena vastly more complex than those studied in the physical sciences  – that struggle is far more difficult and less likely, in any instance, to succeed than it is in the physical sciences.
Doing serious scientific work is difficult and demanding. Armchair theorizing alone is never adequate and typically malignant. Similarly, ‘letting the data speak for themselves’ is futile. Data alone never speak. Only human’s speak – and it is only we who interpret the data about which we speak to our professional colleagues and to the general public. And (I think especially in the social sciences), data are never handed to us in ‘natural’ forms, ones determined by the gods. We imperfect still-seeking-understanding human being must collect data; organize and sort data; and (of course) interpret data. Each of these stages involves not only an inevitable theory to guide the scientist but also judgment.
I suspect that most sensible people wish that the above weren’t so. Sometimes I do. It would be far easier for us (although much less intellectually exciting) if nature – physical and social – dumped into our brains pre-formed data that speak for themselves. Or even if nature pre-formed all the data with clear answers blazoned across them, requiring of us only the fortitude and curiosity to find these answer-filled data, much like young children find hidden dyed eggs on Easter morning.
The interplay between theory, observation, measurement, and scientific conclusion, however, remains enormously complex. It is an intellectual venture for which good judgment is indispensable yet also, I believe, the rarest of all the good qualities of intellectuals. The reason judgment is so rare is not only that there is no recipe for it that can be taught in graduate school. The deeper reason good judgment is so rare is that it seldom dazzles and it often counsels caution – not prohibition, but caution – before leaping away from established theories onto the newest and shiniest theories. Good judgment, at least in the social sciences, is rare also because to have it one must, in addition to equipping oneself with at least adequate ability in technique, spend a good deal of time and thought studying history, philosophy, economics, political science, literature, and the intellectual history of whatever discipline is your specialty.
The economics profession (which is the only one that I’m competent to judge) is filled with scores of brilliant people, many of whom lack good judgment because, it is evident, the range of their knowledge is too narrow. (And, to add an epilogue, too many modern economists are under the misapprehension that improved techniques for gathering and measuring and processing data are the chief keys to a better understanding of social and economic reality.)