The Deductive-Nomological Model of Explanation

Core Idea

The deductive-nomological (D-N) model, developed by Hempel and Oppenheim, provides a formal account of scientific explanation: to explain an event is to show that it follows deductively from premises consisting of true universal laws and specific initial conditions. The explanans must contain laws; the explanatum must be logically entailed. This model emphasizes laws, logical structure, and makes explanation parallel to prediction. However, it faces challenges: some valid D-N arguments feel intuitively like poor explanations, asymmetries between explanation and prediction emerge, and not all scientific explanations fit the D-N form.

Explainer

From your introduction to philosophy of science, you know that science aims not just to describe phenomena but to *explain* them. The D-N model is an attempt to cash out exactly what explanation means — to give it the same logical precision that deductive reasoning gives to proof. The central idea is elegant: you explain an event by showing it was *to be expected* given the laws of nature and the circumstances. Explanation becomes a deductive argument: from true universal laws plus true initial conditions, the event we want to explain follows as a logical consequence.

Consider a simple example. Why did the metal rod expand when heated? Because (law) all metals expand when heated, and (initial condition) this rod is metal and was heated. The explanandum — rod expanded — follows deductively. Or more ambitiously: why did the planet reach that position at that time? Because (Newtonian gravitational law) every mass attracts every other mass with force GMm/r², plus the initial positions and velocities — and from those premises, the position follows mathematically. The explanans (the explaining premises) must contain at least one genuine universal law; without the law, you have description, not explanation.

One of the model's most striking features is the symmetry of explanation and prediction. On the D-N account, every explanation is a prediction that could have been made in advance, and every successful prediction (from laws plus conditions) is potentially an explanation. If you knew the laws and initial conditions beforehand, you could have predicted the event; explaining it after the fact uses the same logical structure. This symmetry seems like a virtue — it ties explanation to predictive power — but it generates serious counterexamples. You can "explain" flagpole shadow length by deriving it from the flagpole height, sun angle, and laws of optics. But reversing the argument — "explaining" flagpole height from shadow length — produces a valid D-N argument that intuitively explains nothing.

This asymmetry problem reveals that the D-N model captures something real about explanation while missing something important. It captures the role of laws and logical entailment. It misses the role of *causes* and *direction* — we explain effects from causes, not causes from effects, even when the logic runs both ways. Real scientific explanation, it turns out, has structure beyond formal deducibility, which motivates causal and unification models you will encounter next.