Let’s distinguish three kinds of scientific theories: predictive, origin, and comprehensivetheories.
Predictivetheoriesdeclare universal laws or rules that always apply. These typically apply to measurable or verifiable quantities, so they can either be measured against a common scale or their instances can be counted and then measured by the probabilities of an occurrence. Newton’s 3 laws of motion are an example.
Origintheoriesare chronological accounts that explain how the particular configurations we observe came about. Our current understanding of the 8-planet solar system is an example of this; as these theories build from incomplete evidence and records, by connecting the dots, they are always somewhat speculative, and are routinely revised as further evidence comes in.
Comprehensive theories are much more expansive, providing a conceptual framework for the topic as a whole. They typically combine multiple predictive theories and origin theories along with some overarching principles or axiomatic assumptions that provide coherence and context for the other two, but are not themselves directly testable. Newton’s and Einstein’s theories of gravity are examples of this; both made claims that were not immediately testable, while containing predictive theories that were testable.
Predictivetheoriespredict what will happen in limited or controlled circumstances where the universal laws they declare can be put to decisive tests. The ideal experiment or observation controls all variables except for two, the independent variable which the experiment or observation controls and the dependent variable, changes in which can be directly correlated to changes in the observer-controlled independent variable. Universal laws said to determine the correlation between these two variables can thus be put to a test controlled by the observer.
Predictive theories are governed by unchanging laws that apply throughout the whole universe and in each of its parts. To test these theories empirically, one need only create a microcosm of the universe, setting up a carefully limited, controlled, and measured (and, thus, artificial) original configuration. In this microcosm, isolated variables can be measured and correlated as one systematically varies only the independent variable, providing measurable, quantified data. The controlled experiment creates an isolated module in which the forces, laws, and causal correlations operative in every other module in the universe can be isolated, made thereby observable, and put to decisive tests.
Origin theoriescannot be so tested. Remember, they try to explain the creation of a particular configuration — a result of universal laws and original configurations, but vastly complicated by the unplanned interaction of all events in the ongoing sequence of events leading up to the particular configuration. There can be few if any observer-controlled variables here.
Origin theories, unlike predictive theories, cannot be directly tested. 1) They are not isolated, controlled modules, but uncontrolled event free-for-alls, in which multitudinous independent chains of causality interact in ways too complex and multivariate to be controlled and made predictable by the observer. 2) They are not representative modules, in principle, repeated everywhere else in the universe; they are unique chronologically accreted results of the random, undesigned interactions of causal events blind to everything outside the forces acting immediately upon them. 3) Because they generate one-of-a-kind situations and entities, they are often mistaken for history (i.e., what happened in the human past to give us the present that we humans find ourselves in), but they are not history, because the generative events occur without intentions or purposes, being strictly causal and deterministic, in the strong sense. Such origin theories include cosmological accounts of the origin of our present day planet, solar system, galaxy, and universe, as well as geological accounts of the origins of geographical features, rock strata, and continents, as well as evolutionary biological accounts of the origin of all species, and of the extinction of lost species.
Origin theories are random from the perspective of an overarching design or purpose (even in history, large-scale events have no overarching intelligent design, just the clashing designs of competing agents and interests), yet not arbitrary, for, if one could observe, measure, catalog, and inventory all the constituent events, the outcome is necessary in accordance with universal laws. For the current configuration resulted from universal laws acting on a long-ago original configuration; but that configuration has been so altered in the meantime, and impacted by so many independent variables, that we can neither observe nor measure its constituent events. No controlled module can be constructed from such a menagerie.
Instead, we must piece together a narrative, roughly analogous to the investigation of a homicide, especially if we focus on the forensic evidence, and the testimony, not from human observers (except in the case of history), but from mechanical recording devices. Short of directly observing the final homicidal event itself (rarely possible for the jury and judge), we can only examine the current configuration of evidence, piece together a sporadic chronological and forensic record from the evidence, and match it against a list of suspects, each suspect’s status as a suspect being supported by an origin theory claiming that the suspect’s doings sufficiently match the evidence at hand to account for the origin of, in this case, a murder victim. And if the comparative matching of each origin theory makes one such theory the likeliest, the suspect named by that theory becomes our prime suspect.