Peirce and the Problem of Unobservable Causes

After invoking Hume, Gould turns briefly to the American philosopher C. S. Peirce, and the article takes another interesting turn. Peirce denied the need for an invariant assumption by arguing that inductive conclusions are self-correcting. As observations accumulate, recurrent patterns become clearer, errors are refined away, and knowledge moves toward closer approximation. This is an attractive idea. Science does often improve through repetition, correction, and expanded sampling. But Gould argues that geology faces a problem Peirce’s account does not fully solve: the causes of many past events are “in principle, unobservable.”

This phrase is crucial. Gould is not saying merely that we have not yet observed ancient causes. He is saying that, for historical reasons, some causes cannot be directly observed because they occurred in the past and are gone. The ancient glacier that made a striation cannot be watched. The vanished organism that left a trace fossil cannot be filmed. The storm that deposited an ancient bed cannot be instrumented. The eruption, impact, flood, reef, swamp, or extinction event is not simply hidden behind a curtain. It has passed out of direct access.

Peirce’s self-correcting induction works best when we can observe a recurring sequence many times. If we repeatedly see event A followed by event B, we can refine our expectation. But historical geology often asks a different question: what unobserved past cause produced this observed present trace? We can study modern glaciers producing striations and then infer that ancient striations were produced similarly. But the ancient causal sequence itself cannot be observed again.

Gould’s glacial example is elegant. We can learn from “modern glaciers” and then infer that striations in ancient rocks were “similarly caused.” The logic depends on more than repeated observation. It depends on the invariance of laws governing causal sequences. Without that assumption, the modern relation between glacier and striation would not authorize an inference about ancient striations.

This is not a weakness unique to geology. Many historical sciences share it. Evolutionary biology cannot directly observe most speciation events in deep time. Archaeology cannot watch ancient rituals or decisions unfold. Cosmology cannot rerun the early universe. Paleoclimatology cannot place instruments in a vanished atmosphere. These sciences rely on traces, models, analogues, and lawful inference. Their subjects are absent, but their evidence remains.

Gould’s point helps defend historical science against a common misunderstanding. Some people imagine that only repeatable laboratory experiment counts as strong science. Historical sciences show otherwise. They demonstrate that explanation can be powerful when it uses traces, constraints, independent evidence, and stable causal principles. A detective does not need to witness the crime if the evidence is rich enough and the causal reasoning is sound. Geology is detective work with billion-year evidence lockers.

But Gould’s point also imposes discipline. Because past causes cannot be directly observed, the scientist must be careful about inference. The ancient trace may resemble a modern product, but resemblance alone is not enough. One must ask whether the proposed cause is physically plausible, whether alternative causes can be excluded, whether the regional context supports the interpretation, and whether multiple lines of evidence converge. Methodological uniformitarianism makes inference possible; it does not make every analogy safe.

This is where the limits of modern analogues become important. A present process can illuminate a past trace, but ancient conditions may have differed. A glacial striation example is comparatively straightforward because the mechanical relation is robust. Other cases are harder. Extinct organisms may have ecological roles without exact living equivalents. Ancient microbial worlds may have operated under atmospheric and oceanic conditions unlike today’s. Sedimentary structures may form under several different processes. Chemical signatures can be altered after deposition. The past can be lawful and still difficult.

Peirce’s emphasis on self-correction remains valuable here. Historical sciences do correct themselves. New evidence, better dating, improved instruments, broader comparisons, and more refined models can overturn earlier interpretations. The inability to observe the original event does not freeze knowledge. But Gould is right that self-correction depends on a background assumption of lawful continuity. We can revise our interpretation of a trace because we believe causes and effects maintain intelligible relations.

The post should explore the emotional drama of this problem. Historical scientists live with ghosts. Their objects are present, but their causes are absent. A fossil is present. The animal is absent. A layer is present. The depositional environment is absent. A crater is present. The impact is absent. This gives geology a strange temporal intimacy. The past is gone, yet it touches the hand through stone.

Gould’s distinction between observable and unobservable causes also clarifies why methodological uniformitarianism should not be confused with substantive uniformitarianism. The assumption needed for inference is not that ancient glaciers moved at the same speed as modern glaciers, or that all glacial episodes were materially identical. The needed assumption is that the causal laws linking ice, pressure, debris, motion, and rock abrasion are stable. Uniform law, not uniform condition, does the work.

This distinction can reshape how we think about evidence. Evidence is not simply a thing. It is a thing interpreted through a causal framework. A scratch becomes a glacial striation only within a network of observations, laws, comparisons, and geological context. A fossil becomes evidence of evolutionary history only within principles linking form, descent, variation, preservation, and time. The past does not speak in raw data. It speaks through disciplined interpretation.

The limitation in Gould’s discussion is that he treats invariant law as necessary, but the practical work of historical inference often depends on middle-level regularities, not only abstract laws. Between physics and interpretation lie process models: how rivers braid, how reefs grow, how bones fossilize, how ash layers disperse, how ecosystems collapse. These regularities may be conditional. They travel across time, but not automatically. Scientists must know when an analogue is robust and when it is fragile.

Still, Gould’s core insight stands. Peirce’s self-correcting induction is not enough by itself for geology because geology asks us to infer causes we cannot directly observe. That inference requires confidence in the continuity of causal law. Methodological uniformitarianism names that confidence, though Gould thinks the name has outlived its usefulness.

The ancient causes are gone. The traces remain. Between them stands inference, delicate but powerful, held together by the assumption that nature did not change its rules while no one was looking.