Juan Antonio Matiz Monroy — Lawyer, Universidad de los Andes · Specialist in Criminal Law, Universidad Sergio Arboleda · Professional licence 294.746 (C.S.J.)
Two judges can receive the same file and hand down opposite judgments without it being possible to say which of them got it wrong. Not because one assessed the evidence badly, but because they do not agree on how much evidence is needed to convict. That point —the threshold— is what the law should have fixed precisely, and did not.
In criminal proceedings that limit is written down. Article 381 of the Criminal Procedure Code requires «knowledge beyond all doubt as to the offence and the criminal responsibility of the accused, founded on the evidence argued at trial», and article 7 conditions it on the judge’s «conviction». Neither provision, however, states how one recognises that the point has been reached.
What is missing is well known, and it is not peculiar to criminal proceedings. Every decision about facts needs a rule saying when the evidence suffices. That rule looks at the evidence, not at the conviction of the person judging, and it fixes a level.
In criminal matters that level is stated as a motto: knowledge beyond all doubt. Under that same motto fit different delimitations, more or less demanding for the prosecution. Choosing between them allocates the risk of error.
In Colombia the picture is problematic. The formula in force is too vague to mark a threshold, and the choice between the possible formulations has not been argued with sufficient rigour or conceptual clarity. The threshold is therefore written by each judicial decision.
From that come the costs, and from that come two measures. The first is for the legislature to fix the threshold. The second, while it does not, is for each judgment to fill it with the reading most favourable to the accused and to justify that choice.
1. Evidence to convict in Colombia: the law named the limit but did not measure it
1.1. What a standard of proof requires
Procedural guarantees do not operate on their own: they need an explicit threshold. Giving reasons for a judgment, challenging it for insufficiency of evidence and asserting the presumption of innocence all presuppose knowing how much evidence suffices.
That is why the absence of legally provided standards «entails a failure to respect both the right to due process […] and the right to evidence» (Ferrer Beltrán, Prueba sin convicción, p. 65). The scholarship on the subject agrees on two points: a subjective criterion is not a standard, and the level is a choice once the formal conditions are met.
A rule deserves to be called a standard if it meets three methodological requirements. It must appeal to the justificatory capacity of the evidence, not to the conviction of the person judging. It must fix an intelligible threshold, with the least possible vagueness. And it must be formulated in criteria of logical probability, not percentages. To those three Ferrer adds a fourth, on the relationship the standards governing different decisions in the same proceeding should bear to each other (Ferrer, Prueba sin convicción, pp. 29-30).
Those criteria are the degree of corroboration of the hypothesis, the weight of the available evidence and the refutation of alternative explanations. The second requirement is the one most often confused here. Reasoned assessment teaches how to weigh evidence, but does not say when it suffices.
From that follows something usually overlooked. The motto «beyond all doubt» is not by itself a standard: it fails the second requirement, because its vagueness allows no threshold to be recognised. Ferrer offers seven formulations that do meet it, ordered from most to least demanding (Ferrer, Prueba sin convicción, p. 209).
They neither exhaust the possibilities nor ask to be adopted. They show that the requirements are formal, not about level, and that meeting them does not yet say how much should be required.
Choosing between them means deciding which of the two errors is more bearable: convicting someone who does not deserve it or acquitting someone who does. Here the first three suffice: they share the first condition —that the prosecution hypothesis explain the data and that its predictions have been confirmed by evidence— and they differ in how much refutation they require.
| Formulation | What the prosecution must establish in order to convict |
|---|---|
| Standard 1 | That every plausible hypothesis compatible with innocence has been refuted, whether raised or not. |
| Standard 2 | That the alternative hypothesis the defence formulated has been refuted, if it is plausible and compatible with innocence. |
| Standard 3 | That that hypothesis has been refuted, but only if the defence produced some evidence confirming it. |
Satisfying the first implies satisfying the second or the third, but not the reverse (Ferrer, Prueba sin convicción, p. 215).
On the same file the result changes: under the first, an explanation compatible with innocence prevents conviction even if no one raised it; under the third, that same explanation is ignored if the defence produced no evidence to support it. It is not a drafting nuance, it is how much the State has to prove in order to convict.
1.2. The law did not fix a threshold that can be reviewed
The Code wrote the limit twice. Article 7 conditions conviction on the judge’s «conviction»; article 381 requires «knowledge beyond all doubt». Neither fixes a threshold: one refers to a mental state and the other names the motto without saying what has to be proved to reach it.
The first defect has a name in the scholarship. «Any formulation of the standard of proof that refers to mental or psychological states of the decision-maker fails this requirement», writes Ferrer, and article 7 appears on his list of examples, alongside the Spanish, Brazilian and Mexican statutes (Ferrer, Prueba sin convicción, pp. 29-30).
It might be objected that article 7 is cured by reading it with article 381. The objection does not hold: 381 changes the word, but a motto without a chosen formulation still indicates «only the imprecise degree of firmness the conviction should have» (Ferrer, Prueba sin convicción, p. 31). Neither article, moreover, says «reasonable»: that adjective was added by practice.
A rule like that cannot operate as a form proper to the trial. There is no threshold known before the case, no measure two judges can share, and no way to verify whether it was reached. It is the failure to respect «both the right to due process […] and the right to evidence» the scholarship anticipated (Ferrer, Prueba sin convicción, p. 65).
A vague formula could still become a standard if the Chamber chose a level and held to it. That is what has to be checked.
1.3. The case law did not close what the statute left open
The check does not change the result. The gap left by the statute —a motto without a measure and a criterion tied to conviction— remains intact in the practice of the courts. The level is fixed judgment by judgment, and each judge fixes it in the case before them.
Four examples illustrate it. There are uses tied to the decision-maker’s conviction; the same motto demanded in two quantities of evidence; a ruling that requires refutation without delimiting the set; and different names for the hypothesis capable of acquitting, none of them defined.
The threshold is formulated as the conviction of the person judging. There are decisions requiring the evidence to allow one «to reach a state of conviction beyond all reasonable doubt» (CSJ, Cas. Penal, Sent. SP3583-2021, ago. 18/2021, Rad. 57196. M.P. Gerson Chaverra Castro).
A standard looks at what the evidence justifies; this use looks at what the judge believed. That state does not exist before the ruling and cannot be tested on appeal.
The same motto demands two quantities of evidence. In one ruling it cannot be ruled out that the hypothesis is «implicit in the indictment and/or detected by the judge during the oral trial, even if the parties make no express reference to it» (CSJ, Cas. Penal, Sent. SP3168-2017, mar. 8/2017, Rad. 44599. M.P. Patricia Salazar Cuéllar). In another, reasonable doubt arises «where the defence presents an alternative hypothesis» which «must find reasonable support in the evidence» (CSJ, Cas. Penal, Sent. SP5295-2019, dic. 4/2019, Rad. 55651. M.P. Patricia Salazar Cuéllar).
Take the same explanation compatible with innocence, raised by no one. Under the first ruling there is no conviction; under the second there may be. They are Standard 1 and Standard 3 of the table, and the accused does not know which will apply to them.
One ruling requires refutation and does not say what is included. It recognises that the prosecution must exclude the alternative hypotheses, but does not specify whether they are all of them or only those the defence raises: it asked that the thesis be «capable of withstanding critical scrutiny» (CSJ, Cas. Penal, Sent. oct. 26/2011, Rad. 36357. M.P. Julio Enrique Socha Salamanca).
Ferrer noted it about that same ruling: «it is not clear which alternative hypotheses must be refuted» (Ferrer, Prueba sin convicción, p. 217, n. 19). The previous one chose twice, and not the same way; this one did not choose, and no one enters a trial knowing what they must knock down.
The hypothesis capable of acquitting changes its name. In one ruling it is enough for the defence «to show that the alternative hypothesis is genuinely plausible» (CSJ, Cas. Penal, Sent. SP282-2017, ene. 18/2017, Rad. 40120. M.P. Patricia Salazar Cuéllar); in another it must be «a plausible, concrete, coherent hypothesis compatible with the facts established in the proceeding» (CSJ, Cas. Penal, Sent. SP682-2026, jul. 8/2026, Rad. 68437. M.P. Diego Eugenio Corredor Beltrán).
Take a single hypothesis. Under the first name it is enough that it be plausible; under the second it can fall for not being treated as concrete. It falls on an adjective no one defines, not because the evidence was measured.
The four examples show that the case law has not established a uniform standard, and what that absence entails. Nor did the statute write one, so whoever comes to trial does not know the threshold in advance. That puts pre-existing law and the forms proper to each trial under strain.
2. Without a threshold there is no review of the decision
The problems that follow are not abstract objections: they come out of those two findings. If the threshold is set by each decision and not by the statute, guarantees stop operating.
Review of the decision. Giving reasons means showing why the evidence reaches the required level. If that level is not defined, there is no measure against which to show it and the reasoning is reduced to describing the evidence. The extreme case is the criterion tied to conviction: holding a belief is not a voluntary act, so «it makes no sense to require a justification of it, which is conceptually impossible to give» (Ferrer, Prueba sin convicción, p. 192).
Appeal. What cannot be shown cannot be argued either. Challenging for insufficiency of evidence requires a measure against which to test the ruling.
Without it, «the indeterminacy of that threshold of sufficiency empties the parties’ own right of appeal of content […] as regards factual aspects» (Ferrer, Prueba sin convicción, p. 114). All that is left is arguing whether the judge weighed each item of evidence well, which is another matter.
The presumption of innocence. A guarantee that cannot be argued on appeal is left to whoever ruled. It only protects «if it is possible to know the threshold of evidentiary sufficiency from which the presumption is defeated» (Ferrer, Prueba sin convicción, p. 114).
Predictability. The chain does not stop at the file. The same evidence, another judge, another threshold. Whoever assesses a plea agreement calculates blind, and with that the preventive function suffers, because the threat of punishment only motivates if it can be anticipated.
What is lost is not severity: it is calculability.
3. Two ways out: the statute and, meanwhile, the most demanding threshold
None of those losses is cured by weighing the evidence better: they all come from the rule being missing. That is why there are two measures, in this order. The first is the one that properly applies; the second is what is available until the first arrives.
3.1. The substantive way out: the legislature writes the threshold
The level of demand allocates the risk of error between the State and the accused. That is a political decision, not an epistemological one. That is why standards «must be expressed in general rules […] and cannot depend on individual decisions» (Ferrer, Prueba sin convicción, p. 150).
The Court has said it in so many words: «it rests with the legislature, not the judge, to determine the degree or level of corroboration or sufficient probability required» (CSJ, Cas. Penal, Sent. SP3274-2020, sep. 2/2020, Rad. 50587. M.P. Patricia Salazar Cuéllar).
Article 29 of the Constitution requires trial under pre-existing laws and with observance of the forms proper to each trial. Article 93 incorporates the presumption of innocence of the American Convention and of the International Covenant, which requires guilt proved «according to law».
These are two successive decisions: how much corroboration is required and how that requirement is formulated in objective criteria. The first is for the legislature; the second cannot be left to each ruling, because then there is no rule to review.
Those conditions answer the losses in the previous section. The rule would have to be prior and statutory, known before the trial. Objective, referring to the relationship between the evidence and the hypotheses. And precise, capable of marking a recognisable threshold.
It would also have to be qualitative, in criteria of inductive probability and not percentages. Reviewable, verifiable by someone other than the person who decided. And graduated, with thresholds rising according to the seriousness of the decision, not that of the offence or the status of the victim.
3.2. Meanwhile: the reading most favourable to the accused
The gap does not permit waiting for reform, because the judge has to rule anyway. And where the law is silent, the choice is not left to each person’s own criterion: it is governed by the same parameters the legislature should follow and by the rules of interpretation of criminal law.
In criminal matters that criterion is fixed. Article 7 of the Code orders doubt to be resolved in favour of the accused and prohibits reversing the burden of proof.
Dealing with the interpretation of criminal law, Velásquez puts it thus: «in no circumstance may the principle of favor rei be disregarded, that is, doubts must always be resolved in favour of the accused and not against them» (Velásquez Velásquez, Fundamentos de derecho penal. Parte general, p. 184).
Hence the position maintained here. While the legislature does not write the threshold, whoever must fill it should apply the reading most favourable to the accused. That is the most demanding of those the Court already uses. It is the first: refuting every plausible hypothesis compatible with innocence, whether raised or not.
That reading is not foreign to the Court’s own line. The ruling that opened it distinguishes by the direction of the hypothesis the judge draws from the evidence: if it points to acquittal, it «shall make the presumption of innocence prevail over any result of the argument between the parties» (Rad. 36357, cit.); if it points to conviction, it counts only where the defence failed in its burdens.
There is a practical reason for requiring it. In several of the decisions that fix the threshold, what decides is not the requirement imposed on the defence but that the prosecution did not prove its own hypothesis; the adjective about the opposing hypothesis comes afterwards, as added reasoning.
That no formula eliminates all imprecision does not permit choosing the threshold in silence. That is why the minimum requirement is that the judgment justify, in its own words, the threshold it applied: which hypotheses it considered, which it discarded and why.
There lies ground for litigation: asking that the ruling justify it and complaining on appeal about the one taken for granted. None of that guarantees a result, but it turns into a matter of argument what today passes without discussion.
4. Conclusions
Articles 7 and 381 of the Criminal Procedure Code name the limit for conviction but fix no threshold. One refers to the judge’s conviction; the other states the motto «beyond all doubt» without saying what has to be proved. Ferrer shows that the motto admits different formulations, which allocate the risk of error differently. The Supreme Court’s case law has not chosen one either: it has used several, with different results on the same file.
Without a prior threshold there is no reviewable reasoning, no workable appeal and no operative presumption of innocence. The substantive way out is for the legislature to write a prior, objective, precise and reviewable rule. Meanwhile, article 7 orders doubt to be resolved in favour of the accused. Hence the position maintained here: the most demanding reading, standard 1, and a duty on each judgment to justify the threshold it applied.
That argument is not improvised at judgment: it is prepared from the hypothesis the defence takes to trial. In the initial consultation we review the file and the standard against which the evidence in your case is being weighed. On that assessment, the firm can take on the criminal defence. At trial and on appeal, it can ask that the ruling justify the threshold applied.
Frequently asked questions
What people usually ask us
What does Colombian law require in order to convict?
Article 381 of the Criminal Procedure Code requires «knowledge beyond all doubt as to the offence and the criminal responsibility of the accused, founded on the evidence argued at trial», and article 7 speaks of the judge’s «conviction» beyond all doubt. Neither provision states how much evidence satisfies that level, so the concrete threshold is determined by each judicial decision.
Are «beyond all doubt» and «beyond reasonable doubt» the same thing?
The Colombian Code says «beyond all doubt», without the adjective. «Reasonable» was added by judicial practice, under Anglo-American influence. The difference matters because with that adjective comes an argument —which doubt counts as reasonable— that the statutory text did not regulate.
Does the defence have to prove its alternative hypothesis?
The Supreme Court has said it both ways. In 2017 it held that in Ley 906 proceedings the party advancing the factual hypothesis bears the burden of showing it; in 2022 it accepted that the hypothesis generating the doubt may be implicit in the indictment or detected by the judge, even if the parties do not mention it. Article 7 of the Code, for its part, places the burden of proof on the prosecuting body and prohibits reversing it.
Can the standard the judge applied be argued on appeal?
It can be argued that the judgment did not make explicit the threshold it applied, or that it applied a less demanding one than it should have. The difficulty is precisely the one this article describes: with no predetermined threshold, the complaint lacks a clear measure against which to test the ruling, and tends to be confused with the argument about the assessment of each item of evidence.
Legal and bibliographic references
Legislation. C.N., arts. 29 y 93. · C.P., art. 6.º (L. 599/2000). · CPP, arts. 7.º, 380 y 381 (L. 906/2004). · Convención Americana sobre Derechos Humanos, art. 8.2 (L. 16/1972). · Pacto Internacional de Derechos Civiles y Políticos, art. 14.2 (L. 74/1968).
Case law. CSJ, Cas. Penal, Sent. oct. 26/2011, Rad. 36357. M.P. Julio Enrique Socha Salamanca. · Sent. SP4316-2015, abr. 16/2015, Rad. 43262. M.P. María del Rosario González Muñoz. · Sent. SP1467-2016, oct. 12/2016, Rad. 37175. M.P. Patricia Salazar Cuéllar. · Sent. SP282-2017, ene. 18/2017, Rad. 40120. M.P. Patricia Salazar Cuéllar. · Sent. SP3168-2017, mar. 8/2017, Rad. 44599. M.P. Patricia Salazar Cuéllar. · Sent. SP9916-2017, jul. 11/2017, Rad. 44997. M.P. Patricia Salazar Cuéllar. · Sent. SP1750-2018, may. 23/2018, Rad. 49009. M.P. Fernando Alberto Castro Caballero. · Sent. SP5295-2019, dic. 4/2019, Rad. 55651. M.P. Patricia Salazar Cuéllar. · Sent. SP358-2020, feb. 12/2020, Rad. 53127. M.P. Luis Antonio Hernández Barbosa. · Sent. may. 6/2020, Rad. 49906. M.P. Luis Antonio Hernández Barbosa. · Sent. SP3583-2021, ago. 18/2021, Rad. 57196. M.P. Gerson Chaverra Castro. · Sent. SP162-2023, abr. 19/2023, Rad. 58235. M.P. Gerson Chaverra Castro. · Sent. SP3274-2020, sep. 2/2020, Rad. 50587. M.P. Patricia Salazar Cuéllar. · Sent. SP5451-2021, dic. 1/2021, Rad. 51920. M.P. Hugo Quintero Bernate. · Trib. Sup. Bogotá, Sala Penal, Sent. feb. 5/2025, Rad. 110016000096201880033-01. M.P. Efraín Adolfo Bermúdez Mora. · Trib. Sup. Medellín, Sala Penal, Sent. feb. 1/2022, Rad. 05-001-60-00206-2017-34027. M.P. Nelson Saray Botero. · Sent. SP1743-2022, may. 25/2022, Rad. 59213. M.P. Diego Eugenio Corredor Beltrán. · Sent. SP170-2023, may. 10/2023, Rad. 62852. M.P. Luis Antonio Hernández Barbosa. · Sent. SP349-2023, ago. 23/2023, Rad. 56154. M.P. Diego Eugenio Corredor Beltrán. · Sent. SP1860-2024, jul. 17/2024, Rad. 62051. M.P. Myriam Ávila Roldán. · Sent. SP1870-2024, jul. 17/2024, Rad. 57193. M.P. Hugo Quintero Bernate. · Sent. SP2024-2024, jul. 31/2024, Rad. 59068. M.P. Carlos Roberto Solórzano Garavito. · Sent. SP2491-2024, sep. 11/2024, Rad. 62354. M.P. Myriam Ávila Roldán. · Sent. SP124-2026, mar. 4/2026, Rad. 63263. M.P. Myriam Ávila Roldán. · Sent. SP127-2026, mar. 11/2026, Rad. 62025. M.P. Diego Eugenio Corredor Beltrán. · Sent. SP189-2026, abr. 15/2026, Rad. 60455. M.P. Carlos Roberto Solórzano Garavito. · Sent. SP682-2026, jul. 8/2026, Rad. 68437. M.P. Diego Eugenio Corredor Beltrán.
Scholarship. Anderson, Terence; Schum, David y Twining, William. Analysis of Evidence. 2nd ed. Cambridge University Press, 2005. · Beccaria, Cesare. De los delitos y de las penas. · Dei Vecchi, Diego. «Prueba sin convicción en su justa medida». Doxa, n.º 45, 2022. · Ferrajoli, Luigi. Derecho y razón. 6.ª ed. Madrid: Trotta, 2004. · Ferrer Beltrán, Jordi. Prueba sin convicción. Estándares de prueba y debido proceso. Madrid: Marcial Pons, 2021. · Gascón Abellán, Marina. «Sobre la posibilidad de formular estándares de prueba objetivos». Doxa, n.º 28, 2005. · González Lagier, Daniel. «Distinciones, estipulaciones y sospechas…», 2018. · Marín Vásquez, Ramiro Alonso. «El estándar de prueba de conocimiento más allá de duda razonable». Derecho Penal y Criminología, vol. 35, n.º 99, 2014, pp. 113-138. · López Medina, Diego Eduardo. El derecho de los jueces. 2.ª ed. Bogotá: Legis, 2006. · Laudan, Larry. «Por qué un estándar de prueba subjetivo y ambiguo no es un estándar». Doxa, n.º 28, 2005. · Mir Puig, Santiago. Derecho penal. Parte general. 7.ª ed. Barcelona: Reppertor, 2004. · Muñoz García, Miguel Ángel. El conocimiento más allá de toda duda razonable del dolo en la intervención delictiva. Tesis doctoral. Bogotá: Universidad Externado de Colombia, 2022. · Roxin, Claus. Derecho penal. Parte general, t. I. Madrid: Civitas, 1997. · Velásquez Velásquez, Fernando. Fundamentos de derecho penal. Parte general. 3.ª ed. Bogotá: Tirant lo Blanch, 2020.
About the author
Juan Antonio Matiz Monroy — A graduate of the Universidad de los Andes (2017) with a specialisation in Criminal Law from the Universidad Sergio Arboleda (2021), he built a solid command of oral litigation at the Universidad del Rosario and at the Universidad de Salamanca, Spain. He sharpened his rigour in evidential reasoning with CESJUL and the Universidad Externado de Colombia. In parallel he went deeper into asset forfeiture —a field that requires understanding the State’s logic on property with the same rigour with which liberty is defended—. He is currently a candidate for the Master’s in Evidential Reasoning at the Universitat de Girona (Spain) and the Università degli Studi di Genova (Italy). Professional licence 294.746 (C.S.J.).