In World Hypotheses: A Study in Evidence (1942), Chapter III: Evidence and Corroboration, the third section de-emphasizes _Uncriticized Evidence_ to focus on _Refined Evidence_.
Pepper introduces an adjective-noun phrase _multiplicative corroboration_ that is expressed in the sketch as verb-adverb _corroborating multiplicatively_; _structural corroboration_ is expressed as _corroborating structurally_. Those processes result in the _products_ of _data_ and _danda_.
digraph PepperCh03 { // Global setup rankdir = BT // Node list {rank=same cc re} cc [shape=oval color="blue" label="Cognizing\ncritically"] re [shape=box color="green" label="Refined\nEvidence"] cl [shape=oval color="blue" label="Criticizing\nlegitimately"] co [shape=box color="green" label="Corroboration"] dt [shape=box color="green" label="Data"] dn [shape=box color="green" label="Danda"] cmu [shape=oval color="blue" label="Corroborating\nmultiplicatively"] cst [shape=oval color="blue" label="Corroborating\nstructurally"] mm [shape=oval color="blue" label="Corroborating\nman with man"] ff [shape=oval color="blue" label="Corroborating\nfact with fact"] // Adding edges cc -> re [arrowhead="onormal" arrowsize="0.8" label="yields"] cl -> cc [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] co -> re [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] cl -> co [arrowhead="onormal" arrowsize="0.8" label="yields"] dt -> co [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] dn -> co [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] cmu -> dt [arrowhead="onormal" arrowsize="0.8" label="yields"] cst -> dn [arrowhead="onormal" arrowsize="0.8" label="yields"] mm -> cmu [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] ff -> cst [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] }
> ยง3.Types of corroboration in refined knowledge > [....] The work of legitimate criticism in cognition, then, is corroboration. And the value of the evidence is in proportion to its expected corroboration. [p. 47] > There are two types of corroboration and accordingly two types of critical evidence. > * There is corroboration of man with man, and corroboration of fact with fact. > * Let us call the first "multiplicative corroboration" and the second "structural corroboration." > * And let us call the products of multiplicative corroboration "data," and the products of structural corroboration "danda." > We shall justify these names later. [pp. 47-48, editorial paragraphing added]
Data is the product of corroborating multiplicatively, or corroborating man with man. Danda is the product of corroborating structurally, or corroborating fact with fact.
_Rough Data_ is contrasted with _Highly Criticized Data_, (and the _Wholly Uncriticized Dubitanda_ shown in the larger sketch in Evidence and Corroboration. In parallel is the _Rough Danda_ contrasted with _Highly Criticized Danda_.
digraph PepperCh03 { // Global setup rankdir = BT // Node list {rank=same cc re} cc [shape=oval color="blue" label="Cognizing\ncritically"] re [shape=box color="green" label="Refined\nEvidence"] cl [shape=oval color="blue" label="Criticizing\nlegitimately"] co [shape=box color="green" label="Corroboration"] ob [shape=oval color="blue" label="Observing"] hy [shape=oval color="blue" label="Hypothesizing"] dt [shape=box color="green" label="Data"] dn [shape=box color="green" label="Danda"] cmu [shape=oval color="blue" label="Corroborating\nmultiplicatively"] cst [shape=oval color="blue" label="Corroborating\nstructurally"] mm [shape=oval color="blue" label="Corroborating\nman with man"] ff [shape=oval color="blue" label="Corroborating\nfact with fact"] rt [shape=box color="green" label="Rough\nData"] hct [shape=box color="green" label="Highly\nCriticized\nData"] rn [shape=box color="green" label="Rough\nDanda"] hcn [shape=box color="green" label="Highly\nCriticized\nDanda"] // Adding edges cc -> re [arrowhead="onormal" arrowsize="0.8" label="yields"] cl -> cc [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] co -> re [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] cl -> co [arrowhead="onormal" arrowsize="0.8" label="yields"] dt -> co [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] dn -> co [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] cmu -> dt [arrowhead="onormal" arrowsize="0.8" label="yields"] cst -> dn [arrowhead="onormal" arrowsize="0.8" label="yields"] mm -> cmu [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] ff -> cst [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] rt -> dt [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] rn -> dn [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] hct -> dt [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] hcn -> dn [arrowhead="onormal" arrowsize="1.5" label="is part of\n(consists)"] rt -> ob [arrowhead="odot" label="is\nrequired\nby"] ob -> hct [arrowhead="onormal" arrowsize="0.8" label="yields"] rn -> hy [arrowhead="odot" label="is\nrequired\nby"] hy -> hcn [arrowhead="onormal" arrowsize="0.8" label="yields"] }
> Now, these two types of corroboration can be found in common sense, or very close to common sense. For there is no sudden leap from uncriticized to criticized fact. > * Common sense has the germs of criticism in it and performs some degree of criticism by itself. Or rather, viewed from the perspective of highly criticized fact, rough criticism appears closer to common sense than to science, mathematics, or metaphysics. > * To refer to this factual material which lies between highly criticized data and danda at one extreme and wholly uncriticized dubitanda at the other, we may use the terms "rough data" and "rough danda." > We shall find this distinction between data and danda proper, and rough data and rough danda, very useful in avoiding certain cognitive pitfalls. [p. 68, editorial paragraphing added] The sketches show that _Rough Data_ through the process of _Observing_, becomes _Highly Criticized Data_. Also, _Rough Danda_, through the process of _Hypothesizing_, becomes _Highly Criticized Danda_.
Pepper use an example of wanting to know if a chair is sufficiently strong to bear a man's weight. * Corroborating multiplicatively (man with man) would have several friends trying by sitting in the chair. * Corroborating structurallly (fact with fact) would put together the evidence of (i) the kind of wood, (ii) nails and glue employed, (iii) the fact of the firm having turned out serviceable furniture for many years, .... etc.
> Whichever I do, my belief is clearly based on a cumulative corroboration of evidence. But the nature of the corroboration differs with the two methods employed. > * In the first trial, it consists in what may be roughly called a repetition of the same fact. I agree with myself in many repeated observations, and my friends agree with me that the chair was strong. > * In the second, the corroboration comes from an agreement of many different facts in the determination of the nature of one central fact. > * In the first, the persuasive force of the corroboration comes from the number of observations and even more from the number of men who agree about them. It is a social force. > * In the second, the persuasive force comes from the massiveness of convergent evidence upon the same point of fact. It is the structural force of the evidence itself and is not peculiarly social. [p. 49, editorial paragraphing added]
Pepper describes a method of _observing_ associated with data, and a method of _hypothesizing_ associated with danda. > The first method seems to be predominantly one of observation; the second, one of hypothesis. This is roughly correct, though the further criticism is carried the less does this distinction count, and at the very end the situation appears almost reversed. The highly refined data are observations sharpened to so fine an edge that the highly refined danda seem to contain much more observation. We shall presently have many occasions to notice this. [pp. 49-50]