Northwestern University Law Review, Vol. 99, No. 3 (Spring 2005), pp. 1053-1099


This Article proceeds in four parts. Part II is a brief overview of harmless-error doctrine in the context of habeas challenges to state criminal convictions, focusing on the nature of the inquiry and the doctrinal deadlock described above. Part III is an empirical analysis of the post-Brecht cases in the federal courts of appeals. To search for a way out of the doctrinal deadlock, I started with a relatively straightforward question: what has happened to harmless-error analysis since Brecht? To answer this question, I reviewed and, with the help of a research assistant, coded all of the 315 harmless-error analysis on habeas review in published opinions over the last decade. Despite the different approaches, a common threat emerges: the language and logic of causation is everywhere, as courts struggle to assess the causal impact of an error on the jury's verdict. Contrary to the conventional view, judges using different approaches to harmless-error are actually trying to answer the same fundamental question: did the error cause the conviction?

Part IV proposes reconceptualizing harmless-error analysis as a determination of causation in a constitutional tort claim and using this reconception to provide a way out of the doctrinal morass. By turning to tort-law debates about what it means to cause harm, I grapple with the question: what does it mean for an error to cause a conviction? Indeed, it appears that different conceptions of causation -- described in tort law as the "but for" versus the "substantial factor" tests -- account for many of the differences in harmless-error outcomes in the federal courts. I argue that a hybrid approach can better serve the normative ideal of determining the factual causation at criminal trials while avoiding appellate fact-finding that violates the Sixth Amendment guarantee of a trial by jury.

In Part V, I turn to prescription: how should judges implement this proposed reconception of harmless-error? First, I observe the difference between the harmless-error cases and the social science literature with respect to the dominant model of how jurors use evidence to reach a verdict. I argue that in the absence of direct evidence about the impact of errors on particular jurors, drawing inferences from empirical research on juries as part of the harmless-error analysis is both appropriate and desirable. And in focusing on the issue of causation, judges and lawyers should turn their attention to evidence of influence on the jury, as opposed to simply weighing the evidence of guilt.