Say there are formula whose terms correspond to individual causal properties of a system, whereby action on one of these terms necessarily brings about a determinate response in another, eg., f=kQq/r^2... if I increase one of the charges, necessarily, the force increases. There is a sense in which a highly specific explanation of "electrostatic force"
includes "charge", so that deabstracting this formula would yield a logically necessary structure.
Then there are formula on the other end, "teen mazagine regressions", where the terms are superstitious aggregates (eg., star signs, number of times you've kissed, etc.).
And formula/models/etc. inbetween, where terms like 'temperature' correspond to a property that no part has, but there is a "nearly determinsitic" relationship between them, such that necessarily for any given set of measures (x, y,t) you have a temperature.. but that you cannot, in practice, actually set {(x, y, t)} so you have a "stochastic-causal relationship".
My claim is that almost all pscyho-social science is of the teen magazine variety: running linear regressions on surveys and then falsely attributing reality the apparent property given by the correlation coefficient. This is just outright pseudoscience and why much of this stuff is unreporducible nonesense.
There is nothing even like "temperature" to our mental faculties, which if explained with a scientific theory, would comprise a vast number of specific causal properties. To say we "forget something" is to describe a combination of 100s of processes each of which have different ecological expressions.
Likewise, "IQ" is exactly the same sort of psuedoscience: a linear regression through survey responses of an entire population and then a fraudulent pseudoscientific project of taking this correlation coefficient and turning it into a singular mental faculty. 110% BS