There are a number of webpages where one can read about "improving the validity" of a scientific process or measurement. Indeed, the entire Wikipedia page for Validity in statistics (which is the sense that I'm interested in) refers to validity as being a matter of degrees of correctness. However, it seems to me that validity, the noun meaning "the quality of being correctly deduced from a premis" as one definition has it, is an either-or thing - either your measurement is valid, or it is not. Is this second interpretation of validity too narrow for use in scientific contexts?
Validity is a multivariate concept, and abstract to the point of being beyond absolute proof. As such, you can discuss the degree of validity since it is not a simple dichotomous absolute.
You can do various things to improve (or decrease) validity.
Here is a Google Ngram chart for three comparative verbs + the validity (improve the validity [blue line], increase the validity [red line], and strengthen the validity [green line]) and for two either/or verbs (establish the validity [yellow line] and prove the validity [dark teal line]), for 1800–2005:
The chart indicates that each of the two either/or combinations has consistently been more common than any of the three comparative combinations. On the other hand, the chart also indicates that the comparative combinations have become significantly more common from the late 1920s onward than they were prior to that. This suggests to me that more people—both in and outside scientific settings—are seeing validity as something that may be more or less firmly established, and not merely as something that either has been established or has not.
Nineteenth-century instances of comparative validity: law and statescraft
A Google Books search doesn’t find many instances of such usage before the second half of the twentieth century, but it does find some, including some that go far back indeed. The earliest of these arise in a legal context—specifically a Scottish legal context. From Sir James Innes & Brigadier General Walter Ker against John Ker, Esq., (January 12, 1808), in Decisions of the Court of Sessions, from November 1807 to November 1808 (1809):
The stipulations of the contract were the price of the acquisition ; and even by the pursuer's argument, whatever excludes the notion of their being gratuitous must strengthen their validity in law. If, instead of these various stipulations, the defender has contracted to pay a sum of money, the constitution of the feu [according to Black’s Law Dictionary, ”In Scotch law, a holding or tenure where the vassal, in place of military service, makes his return in grain or money.”] rights would have been still less lucrative, but not less legal.
From Patrick Irvine, Considerations on the Inexpediency of the Law of Entail in Scotland (1826):
A deed of conveyance is generally revised with great care and attention by the granter of it, if it is only to take effect upon his death ; and both by the granter and receiver of it, if it arises from an onerous transaction ; and likewise by their professional advisers, The instrument of sasine [according to Black’s Law Dictionary: “In Scotch law. The symbolical delivery of land, answering to the livery of seisin of the old English law.”] following upon it, cannot increase the validity of its warrant, if defective in legal formality. But any error (even clerical error) in the sasine may prove fatal to the security which it is intended to complete.
From Smith v. Nash (July 8, 1850), in A Selection of Supreme Court Cases in New South Wales, from 1825 to 1862 (1893):
For the only malice confessed by the plea in that case was the malice in law, which was necessarily implied from the words complained of, and which need not have been stated in the declaration at all, to strengthen its validity in point of substance.
From Graham and others, Pursuers, v. Stewart and others, Lord Lynedoch’s Trustees, Defenders (March 15, 1853), in The Scottish Jurist: Containing Reports of Cases Decided in the Supreme Courts of Scotland (1853):
It has been shewn, on grounds which need not be repeated, that neither the truster, nor the conveyancers who acted for him at the date of the trust, had any idea that any alteration was necessary, in the entail of 1726, to strengthen its validity. They believed it perfect, and that it required no correction ; for when the new entail was executed in terms of the entail of 1726, it was confidently supposed the lands would stand on a valid and impregnable title of entail, according to the law of Scotland, which might have been endangered, but could not be improved, by alteration.
From Zacharias Topelius, Times of Gustaf Adolf, volume (1883):
”I am content with you, my friend,” said the Jesuit, with apparent satisfaction. "These documents, which bear the stamp of truth, will be sufficient to prove the conversion of King Gustaf Wasa and King Charles; and this introduction, signed by you, will further strengthen their validity. I will now return to Germany through Sweden, in order to have these documents printed, either through our adherents in Stockholm, or, if found, possible, in Lübeck or Leyden.”
Early twentieth-century instances of comparative validity: social sciences
Scientific instances—usually instances from the social sciences—begin to appear in the early twentieth century. From Johns Hopkins University, Studies in Education, issue 10 (1928):
The statement has frequently been made that it is better to use two tests and average the results than to use only one test, for the sake of the greater validity of the measurement of the children. The above evidence indicates that this is true if the right tests are chosen for the combination, but that not any two tests will serve advantageously. It is apparent that if used indiscriminately one may harm the results rather than increase their validity.
From Bureau of Public Personnel Administration, Institute for Government Research, Public Personnel Studies, volumes 6–7 (1928[?]) [combined snippets]:
The six items not valid as they originally stood are being modified somewhat in an attempt to improve their validity. It might be of interest to note that for some of the items the number of good testees answering correctly was as high as 19 in excess of the number of poor testees, while for about three-fourths of the items the difference was between five and fifteen.
From Giles Ruch, The Objective or New-Type Examination: An Introduction to Educational Measurement (1929):
Kelley gives a formula which is at times of the greatest value in discovering whether it is worth while to attempt to improve the validity of a test through increasing its reliability.
From University of Iowa, Studies in Education, volume 6 (1931):
To the extent that it was possible to secure a measure of the language ability of the children tested, the validity of the tests was secured. In some tests, additional exercises might improve the validity. In others, the elimination of certain items would no doubt raise the relative degree of validity.
An outlier here is an early-twentieth-century instance of relative validity in earth science. From F.E. Wright & E.S. Larsen, “Quartz as a Geologic Thermometer” (1909), quoted in Edson Bastin, “Geology of the Pegmatites and Associated Rocks of Maine” (1911):
Briefly stated, the four criteria which can be used to distinguish, at ordinary temperatures, quartz which was formed above 575 from quartz which has never been heated to that temperature are [details of four distinctions follow]. Into all these criteria an element of probability enters, and in testing quartz plates, with this end in view, a number of plates should be examined to strengthen the validity of the inferences drawn.
Recent instances of relative validity: physical sciences
By the 1980s the relativity of validity had become an accepted point in some physical sciences as well. From Asad Amr, Energy Systems in the United States (1981):
Friedman's method of ranks is used to confirm further this hypothesis and to strengthen the validity of the analysis of variance. Next, the nature of differences among fuels and criteria is investigated by means of the percentile rank analysis.
From E.I. Galperin & Peter Kennett, Vertical Seismic Profiling and Its Exploration Potential (1985):
By combining the vertical and the fixed geophone level profiling in VSP reflection observations, it is possible to obtain phase correlation of the observed data and to improve the validity of structural mapping.
From Roger Mead, The Design of Experiments: Statistical Principles for Practical Applications (1988):
Enough of this philosophy. In the remainder of the chapter, we examine some methods of examining or testing assumptions; we consider transformations of scale to try to improve the validity of assumptions and a more general approach to fitting models; and we discuss methods to deal with missing observations and how to detect others.
And from Chenfeng Wang & Marina Gavrilova, "A Novel Topology-Based Matching Algorithm for Fingerprint Recognition in the Presence of Elastic Distortions" in Computational Science And Its Applications - ICCSA 2005 (May 2005):
Finally, we develop an efficient global matching scheme based on the comparisons of minutiae sets and singular points to increase the validity of the matching results. Using the same amount of information on minutiae set as traditional minutiae based algorithms we achieve a faster performance and higher accuracy rates using our proposed algorithm.
A hundred years ago, Google Books searches suggest, few authors writing about scientific experimentation and analysis treated validity as a thing that can be improved, increased, or strengthened. However, authors writing on law and statecraft had already been treating it so since at least 1808. The 1900s saw a considerable rise in instances where writers in sciences—especially social sciences—and (later) statistics used comparative verbs with validity.
At this point, it seems to me, the cows are out of the barn and wandering all over the hillside. In the context of social sciences, medicine, earth sciences, and statistics, at least, many practitioners in the field view validity as something that can be improved, increased, or strengthened by adjusting one's methodology. The older, narrower either/or notion of validity does not comprehend the gray world of improvable validity that these practitioners are talking about.