Can Many Weak Clues Add Up to a Conviction? The Logic and Limits of Circumstantial Evidence

Spanish case law has long accepted that a conviction may rest on circumstantial evidence without direct proof; we examined the requirements in our commentary on Supreme Court judgment 355/2026 of 21 May. This article addresses a different and logically prior question: can a collection of clues which, examined one by one, are inconclusive, be transformed into conclusive proof by the mere fact of being piled together? The theory of inference answers no, and it is worth explaining why: in forensic practice, the “joint assessment of the evidence” sometimes works as a verbal alchemy that turns many suspicions into one certainty.

The problem: the temptation to add

The reasoning criticised here is easy to recognise: the accused was near the scene; he had a financial motive; his account contains inaccuracies; there is no proven alibi; a hearsay witness attributes an ambiguous remark to him. None of those facts, on its own, establishes authorship. Yet the ruling declares that, “assessed as a whole”, the indicia “reinforce one another” and lead “univocally” to guilt. The operation has an evident psychological appeal: each added clue looks like one more brick in a wall. The question is whether that image survives logical analysis.

Let us state the thesis up front: accumulating indicia increases probative force only under strict conditions — each item must have probative value of its own, the items must be independent of one another, and they must not all hang from the same source. Where those conditions are absent, adding inconclusive clues does not produce a conclusion: it produces a bulkier suspicion. And a bulky suspicion is still not proof.

What logic says: chains, cables and leaky buckets

Anglo-American legal epistemology distinguishes two structures of inference, elaborated on the basis of John Henry Wigmore's classic treatise on the law of evidence. In a chain inference, each intermediate conclusion serves as a premise for the next: the whole is, like any chain, only as strong as its weakest link, and the reliability of the result decreases with every uncertain link added. In a cable inference — the rope metaphor comes from nineteenth-century English case law — several independent strands converge on the same conclusion, and the cable may hold even if one strand gives way.

The distinction matters because the calculus of probability is unforgiving. If a conclusion requires several uncertain, mutually independent premises to be true at the same time, the probability of the conjunction is the product of their probabilities: two premises reliable at eighty per cent yield sixty-four per cent together; four, barely forty-one. And even without independence, the probability of the conjunction never exceeds that of the weakest premise. This is the well-known conjunction paradox that L. Jonathan Cohen set out in The Probable and the Provable (Clarendon Press, Oxford, 1977): aggregating uncertain elements that must hold together does not raise the probability of the whole, it lowers it. Chaining doubtful inferences impoverishes the conclusion instead of enriching it.

And there is an even more elementary limit: a fact that individually establishes nothing — one equally compatible with guilt and with innocence — has zero probative value, and zero multiplied by ten is still zero. The philosopher Antony Flew coined the canonical image against merely cumulative arguments in 1966: if one leaky bucket will not hold water, there is no reason to think that ten can. From medical literature, Petr Skrabanek and James McCormick named that same illusion the “faggot fallacy” — from the old word for a bundle of sticks — in Follies and Fallacies in Medicine (1989): the belief that multiple pieces of weak evidence, bundled together, amount to strong evidence, when a bundle of insecure evidence remains insecure. Quantity does not replace quality: a heap of suspicions does not manufacture a single proven fact.

The honest caveat: when indicia genuinely reinforce one another

It would be dishonest to present only half of the argument. Probability theory also teaches that several independent items of evidence, each with genuine probative value of its own — each more likely if the accused is guilty than if he is innocent — do reinforce one another when they converge: that is the cable structure, and it is why properly built circumstantial evidence can legitimately support a conviction. Susan Haack made the point precise in Evidence Matters (Cambridge University Press, 2014): combined evidence can warrant a conclusion to a higher degree than any of its components, but only if each piece is secure independently of the conclusion it is meant to prove; interlocking conjectures that merely lean on one another produce no knowledge.

The conditions of genuine corroboration are three. First, each item must have individual probative value; the guidelines of the European network of forensic science institutes put it as a simple rule: a finding whose likelihood ratio equals one provides no assistance in deciding between the hypotheses, and accumulating independent findings of that kind provides none either. Second, the items must be independent of one another; five echoes of the same rumour are not five clues but one clue counted five times. Third, they must not hang from a common source whose collapse drags them all down, as happens when everything stems from the same police report, the same witness or the same disputed expert opinion. The Royal Statistical Society warned about this in its guide on statistical evidence in criminal proceedings (2010): two or more independent items of evidence can be mutually corroborative, but applying the product rule to events that are not independent is a serious error. A notorious British miscarriage of justice illustrates it: the probabilities of two sudden infant deaths in the same family were multiplied as if they were independent — they were not — until the jury was told the odds were one in seventy-three million; the calculation was statistically invalid and the conviction was eventually quashed. When a judgment does not verify these three conditions and merely lists facts and proclaims their “mutual corroboration”, it is not reasoning: it is counting leaky buckets.

The doctrine of the Constitutional Court and the Supreme Court

The striking thing is that Spanish case law, in its most demanding formulation, says exactly this. The Constitutional Court built the canon in judgments 174/1985 and 175/1985, both of 17 December: the indicia must be “fully proven”, because certainties cannot be built “on the basis of mere probabilities”, and the judgment must set out both the indicia and the reasoning that leads from them to authorship. Judgment 229/1988 of 1 December hammered home that indicia cannot be “mere suspicions”. The canon was completed with the review of the sufficiency of the inference: the presumption of innocence is violated when the reasoning is “excessively open, weak or indeterminate” (judgment 135/2003 of 30 June), that is, when the inference is so open that it admits such a plurality of alternative conclusions that none of them can be taken as proven (judgments 189/1998 of 28 September and 133/2014 of 22 July).

The Second Chamber of the Supreme Court treats the classic requirements as settled, as recalled, among the most recent rulings, in judgment 304/2025 of 2 April: indicia fully established, unequivocally incriminating in nature, plural or — if single — of singular probative force, concomitant with the fact to be proven and, where several, interrelated so that they reinforce one another. Judgment 532/2019 of 4 November, which systematised twenty criteria for assessing circumstantial evidence, warns that indicia must not be confused with suspicions and that a conviction cannot be a “judgment of suspicions”, requiring the prosecution hypothesis to attain a “prevailing probability” over the alternatives. Judgment 593/2017 of 21 July had already specified that the force of circumstantial evidence stems “from the interrelation and combination” of the indicia, which converge and reinforce one another when they all point rationally in the same direction. And judgment 830/2024 of 3 October puts it as clearly as one could wish: the value of circumstantial evidence “is not measured by a simple aggregation of items of evidence, but by the logical interaction between them”. Finally, judgment 1000/2025 of 9 December recalls that the alternative hypothesis capable of defeating the inference must be a reasonable one, not a mere conjecture, and judgment 355/2026 of 21 May that the inference suffices only when it blocks any other reasonably possible hypothesis.

Properly read, this doctrine adopts the cable model and rejects the pile-of-suspicions model: it demands individual probative value (indicia “fully established” and “unequivocally incriminating”), genuine convergence (interrelation, not juxtaposition) and rational closure (exclusion of reasonable alternatives). The problem, then, is not the doctrine: it is its degraded application, when the “joint assessment” is invoked as a ritual formula that dispenses with checking, item by item, that those conditions are met. As an old evidentiary aphorism has it, indicia are weighed, not counted.

What legal scholarship demands

Academic scholarship converges on the same point: the judge's conviction is not a criterion of truth, and the number of indicia is no substitute for their quality. In the Spanish rationalist tradition, Jordi Ferrer Beltrán (La valoración racional de la prueba, Marcial Pons, 2007) argues that a hypothesis can be considered proven only when corroborated under a controllable methodology, and in Prueba sin convicción (Marcial Pons, 2021) states the thesis of the “justificatory irrelevance of judicial conviction”: that the court is inwardly convinced justifies nothing; objective standards of proof are needed. Marina Gascón Abellán (Los hechos en el Derecho, 3rd ed., Marcial Pons, 2010) recalls that judicial knowledge of facts is inductive and its results merely probable, which demands every possible safeguard and rigorous reasoning. Perfecto Andrés Ibáñez (“Acerca de la motivación de los hechos en la sentencia penal”, Doxa, no. 12, 1992) dismantled the reading of free assessment of evidence as an ineffable inner conviction; Juan Igartua Salaverría (Valoración de la prueba, motivación y control en el proceso penal, Tirant lo Blanch, 1995) stressed that free assessment is not the judge's subjective sovereignty but an operation subject to reasoning and review; and Manuel Miranda Estrampes (La mínima actividad probatoria en el proceso penal, J.M. Bosch, 1997) specified what minimum incriminating evidence can defeat the presumption of innocence.

In the international literature, Michele Taruffo (La prova dei fatti giuridici, Giuffrè, 1992; Spanish translation, Trotta, 2002) conceived of evidence as an epistemic instrument aimed at the true determination of facts, not at mere persuasion; and Larry Laudan (Truth, Error, and Criminal Law, Cambridge University Press, 2006) showed that the criminal standard of proof is a mechanism for distributing the risk of error between false convictions and false acquittals: a standard that softens precisely when the evidence is poor stops protecting those it was meant to protect. Together with Cohen, Wigmore, Flew and Haack, cited above, the most authoritative voices agree: there is no law of logic by which many near-proofs amount to one proof.

The critique: when “joint assessment” masks the need to decide

The uncomfortable question remains: if the logic is so clear and the judicial doctrine so demanding, why does the rhetorical aggregation of clues keep reappearing? The honest answer is not epistemological but institutional. The judge is under an inexcusable duty to decide (Article 1.7 of the Civil Code): once the trial is over, someone must sign a ruling, and the system does not allow the scientifically most honest answer to an equivocal body of evidence, which would be “we do not know”. To that pressure is added a complacent reading of Article 741 of the Criminal Procedure Act, which instructs the court to assess the evidence “according to its conscience”: if conscience is understood as an uncontrollable inner conviction, the “joint assessment” becomes a black box into which many suspicions enter and out of which one certainty comes, with no one able to audit the transformation.

Yet the legal system does provide an outcome for an insufficient body of circumstantial evidence, and it is not conviction: it is acquittal. That is what the presumption of innocence (Article 24.2 of the Constitution) and the principle in dubio pro reo are for. When a ruling turns a constellation of equivocal facts into certainty by merely invoking their “joint appraisal”, that sum of non-decisive indicia stems from the need to hand down a judgment, not from a scientific deduction. And that is precisely the kind of decision that the constitutional standard of rationality of the inference — the prohibition of excessively open, weak or indeterminate inferences — exists to prevent.

Practical consequences for the defence

Concrete guidelines for the criminal defence against a circumstantial prosecution follow from the above. First: check that each basic fact is established by direct evidence, and identify the neutral facts, without probative value of their own, that add nothing to the whole. Second: audit the independence of the indicia, spotting redundancies (the same fact counted several times), items derived from one another and items hanging from the same disputed source. Third: demand individualised reasoning on the value of each item and on the inferential link, not a generic appeal to joint assessment; bear in mind, however, that the Supreme Court rejects a purely atomised analysis that ignores the interaction between the indicia — judgment 830/2024 calls it the “deconstructive method” — so an effective challenge does not merely fragment the picture: it shows that the alleged interaction is not genuine, because the items are redundant, dependent or neutral. Fourth: build and, where possible, substantiate a reasonable alternative hypothesis, because if the inference remains open, the conviction violates the presumption of innocence.

This article is for general information only and does not constitute legal advice. Every case requires an individualised study of the evidence, of its inferential structure and of the available avenues of challenge.