The government’s cross-appeal presents a far more difficult question. Following the jury verdict of guilt on the second count (carrying a firearm during a crime of violence), the court granted a judgment of acquittal finding “no evidence that the defendant actually carried a firearm, as opposed to a toy gun.” On appeal, we examine the evidence in the light most favorable to the government. If a rational trier of fact could have concluded that every essential element of the crime charged was proved beyond a reasonable doubt, then the issue should have been left to the jury. United States v. Medina-Garcia, 918 F.2d 4, 6-7 (1st Cir.1990).
The firearm statute, 18 U.S.C. § 924(c), provides in relevant part that whoever carries a firearm during the commission of a crime of violence shall be sentenced to an additional five years’ imprisonment (or more if the weapon is of a type here not involved). It is common ground that the gun need not be proved to be loaded or operable in order to convict, United States v. Gonzalez, 800 F.2d 895, 899 (9th Cir.1986), but that a toy or replica will not do. United States v. Westerdahl, 945 F.2d 1083, 1088 (9th Cir.1991). The district court summed up the evidence and held it inadequate to permit a reasonable jury to find, beyond a reasonable doubt, that Kirvan carried a real gun as opposed to a toy.
The government’s riposte is to point to a square holding by then Circuit Judge Scalia in Parker v. United States, 801 F.2d 1382, 1385 (D.C.Cir.1986), cert. denied, 479 U.S. 1070, 107 S.Ct. 964, 93 L.Ed.2d 1011 (1987), that non-expert testimony affirming that a robber used a gun is enough. The holding was followed without much discussion by the Fourth Circuit in a case where five witnesses had described the object as a gun. United States v. Jones, 907 F.2d 456 (4th Cir.1990), cert. denied, 498 U.S. 1029, 111 S.Ct. 683, 112 L.Ed.2d 675 (1991). Kirvan in turn points us to cases which, in upholding convictions under this or similar statutes, recounted or relied upon testimony from a firearms expert or at least a witness who saw the robbery and claimed to know about weapons. E.g., United States v. Buggs, 904 F.2d 1070 (7th Cir.1990); Westerdahl, 945 F.2d at 1088.
If fake guns were extraordinarily rare in bank robberies, it might be fairly easy, absent affirmative proof, to dismiss the possibility that the gun was a toy. The jury had no actual data, which might be inadmissible in any event,2 although it does have considerable latitude in making intuitive judgments about how the world works. United States v. Guerrero-Guerrero, 776 F.2d 1071, 1075 (1st Cir.1985), cert. denied, 475 U.S. 1029, 106 S.Ct. 1233, 89 L.Ed.2d 342 (1986). Nor does policy tilt the balance, as it might if we faced an issue where the government had ready access to direct evidence {e.g., whether a bank is federally insured) and no excuse for leaving the matter in doubt. Of course, Kir-van has such access but we will not decide the matter by relying upon his failure to produce the gun for inspection.3
We need not decide here whether the government’s burden could be met merely by unembellished lay testimony that “the robber carried a gun.” In this instance, the object was identified by two witnesses as a gun; one said that it was black and had a five inch barrel and the other, who was closer, supplied more detail: he said that it appeared “shiny, silver” in color; that it was “[ljarge, very large for a handgun”; and that when it fell to the floor, it made “[a] very loud noise. Heavy object hitting the floor.”
Without deciding whether less would do, we think that this detail permitted a rational jury to conclude that this was a “real” gun: it was a plausible size, colored like a real gun, and quite heavy. One witness could easily describe gun metal as black and another as
2
See 1 McCormack, Evidence § 210, at 949-50 (4th ed. 1992) (cases discourage mathematical proof and probability data in criminal cases). Such data may exist. E.g., Washington Post, October 2, 1986, p. Cl ("Neil Hurley, chief of the grand jury section of D.C. Superior Court, said that at least 10 percent of the armed robbery cases that he sees involve fake guns.").
3
The problem with an adverse inference is the Fifth Amendment’s bar against compelled self-incrimination. See Griffin v. California, 380 U.S. 609, 85 S.Ct. 1229, 14 L.Ed.2d 106 (1965). But cf. Barnes v. United States, 412 U.S. 837, 93 S.Ct. 2357, 37 L.Ed.2d 380 (1973) (upholding an instruction that unexplained possession of stolen property permits an inference of knowledge).