Skip to main navigation Skip to search Skip to main content

Causal interpretations can be based on mechanistic knowledge

  • James B. Grace
  • , Glenn R. Guntenspergen
  • , Kevin J. Buffington
  • , Justine A. Neville
  • , Karen M. Thorne
  • , Michael J. Osland
  • , Melinda Martinez
  • , Joel A. Carr
  • , Debra A. Willard

Research output: Contribution to journal › Review article › peer-review

Abstract

There exists a long-standing disconnect between statistical and mechanistic approaches to the development of causal understanding. Statistical approaches, which have dominated the literature, have focused on the need to obtain perfectly unbiased estimates of causal effects often using either experimental, quasi-experimental or other methods. Mechanistic approaches have instead focused on investigating how systems work by elucidating the structures and processes whereby variations in one system property can propagate to other system properties. Explicit references to ‘causal effects’ have tended to require adherence to statistical methods and standards, inadvertently downplaying the suitability of mechanistic knowledge for that purpose. It has been recently demonstrated that both mechanistic and statistical approaches can contribute to the long-term goal of developing causal knowledge and understanding. Proponents of statistical causal inference have seldom recommended that mechanistic evidence be relied upon to support causal interpretations. This paper provides a clear and thorough example where a causal interpretation can be supported based on mechanistic knowledge. Arguing for a causal interpretation based on knowledge of mechanisms has typically been an informal process and one that has thus far infrequently led to explicit declarations of causal knowledge by scientists. To overcome this problem, we illustrate a recently described procedure referred to as ‘causal knowledge analysis’ to summarize explicit support for causal interpretations. In this paper, we first clarify the basis of the long-standing disagreement by describing the crux of the problem as viewed from a statistical perspective and by describing how it can be overcome when there is sufficient mechanistic knowledge. We then offer a proof-of-concept example based on robust documentation and description of the mechanisms whereby plants causally regulate the responses of coastal marsh elevation to changes in sea level. Synthesis—The evidential requirements for declaring a relationship to be causal have been obscured until very recently, leading to a long neglect of this issue by scientists. Meanwhile, subject matter experts have accumulated a vast body of undeclared causal knowledge that we now need to recognize in order to position scientists as essential players in defending causal interpretations.

Original languageEnglish (US)
Pages (from-to)3084-3098
Number of pages15
JournalJournal of Ecology
Volume113
Issue number11
DOIs
StatePublished - Nov 2025
Externally publishedYes

Bibliographical note

Publisher Copyright:
Published 2025. This article is a U.S. Government work and is in the public domain in the USA.

Keywords

  • causal inference
  • causal knowledge
  • coastal marshes
  • mechanistic causal knowledge

Fingerprint

Dive into the research topics of 'Causal interpretations can be based on mechanistic knowledge'. Together they form a unique fingerprint.

Cite this