R v Villaroman, 2016 SCC 33 (Cromwell):

[37] When assessing circumstantial evidence, the trier of fact should consider “other plausible theor[ies]” and “other reasonable possibilities” which are inconsistent with guilt: R. v. Comba, [1938] O.R. 200 (C.A.), at pp. 205 and 211, per Middleton J.A., aff’d [1938] S.C.R. 396; R. v. Baigent, 2013 BCCA 28, 335 B.C.A.C. 11, at para. 20; R. v. Mitchell, [2008] QCA 394 (AustLII), at para. 35. I agree with the appellant that the Crown thus may need to negative these reasonable possibilities, but certainly does not need to “negative every possible conjecture, no matter how irrational or fanciful, which might be consistent with the innocence of the accused”: R. v. Bagshaw, [1972] S.C.R. 2, at p. 8. “Other plausible theories” or “other reasonable possibilities” must be based on logic and experience applied to the evidence or the absence of evidence, not on speculation.

[38] Of course, the line between a “plausible theory” and “speculation” is not always easy to draw. But the basic question is whether the circumstantial evidence, viewed logically and in light of human experience, is reasonably capable of supporting an inference other than that the accused is guilty.

[39] I have found two particularly useful statements of this principle.

[40] The first is from an old Australian case, Martin v. Osborne, 55 C.L.R. 367, at p. 375:

In the inculpation of an accused person the evidentiary circumstances must bear no other reasonable explanation. This means that, according to the common course of human affairs, the degree of probability that the occurrence of the facts proved would be accompanied by the occurrence of the fact to be proved is so high that the contrary cannot reasonably be supposed. [Emphasis added.]