dissenting:
I respectfully dissent.
I am convinced that no rational trier of fact could have found Talbert guilty beyond a reasonable doubt on the evidence adduced at his trial. The only significant evidence produced against Talbert was his thumbprint on the bowling pin. The principle is universally established, however, that convictions obtained through fingerprint evidence will be upheld only if the Government produced either evidence that the print could have been impressed only during the commission of the offense,1 or separate *532evidence of guilt.2 Contrary to the majority’s assertions, neither of these corroborative circumstances are apparent in the record before us.
First, the Government failed to establish that Talbert’s print was impressed during the commission of the crime. There was no evidence tending to show that Talbert lacked prior access to the bowling pin. Compare Borum, 380 F.2d at 597, with Stevenson v. United States, 380 F.2d 590, 594 (D.C.Cir.1967). To the contrary, there was undisputed evidence that Talbert had prior access. Moreover, there was no affirmative evidence tending to show that Talbert actually touched the pin at the time of the offense: The Government failed to place Talbert at or near the veterans’ facility during the relevant time period. Rather, the evidence tended to place Talbert at a Los Angeles bus station approximately twenty miles from the hospital.
The fingerprint expert who testified at Talbert’s trial was unable to lend support to the speculative inference that Talbert grasped the pin at the time of the murder. The expert could not determine how long Talbert’s print had been implanted on the pin. The expert also asserted that a second print, placed on top of Talbert’s, would not have obscured Talbert’s print and that it was possible to touch the pin without leaving a print at all. Although the expert detected no latent prints over Talbert’s, neither did he detect identifiable prints on Nornes’ personal effects.
Finally, the apparent unlikelihood of both Talbert and the killer grasping the same pin *533fails to supply the necessary corroboration to establish that Talbert touched the pin at the time of the murder. The only inquiry required by either precedent or logic is whether Talbert could have touched the bowling pin on a prior occasion. Hiet v. United States, 365 F.2d 504, 505 (D.C.Cir.1966). The likelihood of the killer grasping the same pin is a matter of statistical probability and we simply do not possess sufficient information on the basis of the record before us to evaluate the odds. Perhaps the bowling pin in question lay apart from the others, perhaps it obstructed a pathway. The likelihood of the killer touching the same pin would be much greater if Talbert had previously touched many of the pins. Yet, the record is clear that only a few of the pins were tested for prints, although the pins laying in the gutters had little or no dust on them.3 Moreover, the pin may already have been in Nornes’ room when Talbert or the killer touched it. The witness who testified at trial that he did not see a bowling pin in Nornes’ room the day before the murder also did not see any bowling balls in the room. Nornes’ room was used, however, to store bowling balls and it contained several racks of balls when the witness peered through the window. Logic demands that we abstain from reliance on unestablished, unsupported, statistical “evidence” when reviewing the strength of the Government’s case.
Second, the Government failed to establish separate evidence of guilt to corroborate the defective fingerprint evidence. The majority disagrees, citing Talbert’s marijuana transactions with Nornes, Tal-bert’s knowledge that Nornes carried large sums of money with him, Talbert’s presence in Los Angeles during the time period in which Nornes was killed, Talbert’s “threatening” comments, and the fact that Tal-bert’s thumbprint was on the neck of the bowling pin. Upon scrutiny, the weakness of this so called corroborative evidence is appalling.
Nothing in Talbert’s and Nornes’ relationship suggested a murderous inclination on Talbert’s part. The only conceivable significance of their marijuana transactions was that from Nornes’ public transfer of cash to Talbert for the purchase of marijuana, accomplished in the presence of approximately sixteen psychiatric patients, Talbert and others learned that Nornes carried large sums of money with him in his money belt. Yet, there was no evidence that Tal-bert had any particular need for money.4 During Talbert’s two-day visit to Santa Barbara, beginning on Saturday, August 2nd, he paid cash in advance for a motel room with a one hundred dollar bill. On a previous trip to Santa Barbara, beginning on July 28th, it appeared to the motel manager that Talbert was carrying “a couple of hundred dollars” with him. On Talbert’s August 2nd visit, the manager offered Tal-bert a job which Talbert declined explaining that he was expecting a disability check soon. Finally, Talbert had been looking for Nornes on August 1st to repay the approximately $650 that he had taken from Nornes to make a marijuana purchase that was never completed.
The majority’s reliance on the fact that Talbert was in Los Angeles during a portion of the time period in which Nornes was killed is difficult to comprehend. That Tal-bert was in Los Angeles and, therefore, may have had an opportunity to kill Nornes lends no support to the conclusion that Tal-bert actually committed the act of killing Nornes. United States v. Hoke, 610 F.2d 678, 679 (9th Cir.1980) (A defendant’s failure to establish an alibi does not constitute evidence of guilt). In fact, the evidence concerning Talbert’s whereabouts on Sun*534day, August 3rd, is more exculpatory than neutral. In order for Talbert to have killed Nornes, Talbert would have had to travel from a downtown Los Angeles bus station twenty miles to the veterans’ hospital and then return to the bus station within approximately three hours. In so doing, however, Talbert would have had no assurance that he would find Nornes if he made the trip. Nornes was a vagrant of whose immediate whereabouts Talbert could have no advance knowledge.
Talbert’s so called “threatening” comments, viewed in the contexts in which they were spoken, fail to evince the hostile intent with which the majority seeks to infuse them. Talbert’s intentions when he stated that he wanted to return Nornes’ money “as soon as [he] could get [his] hands on him” were obviously benign. Although a jury may draw reasonable inferences from the facts presented, I find it inconceivable that a jury could reasonably draw an inference that Talbert intended a threat with this language. Talbert’s stated purpose in looking for Nornes was completely innocent and his colloquial wording extremely commonplace. Moreover, Talbert’s statement that he wouldn’t mind bumping Nornes on the head for his marijuana, uttered approximately one month before Nornes was killed, was made facetiously in the context of a friendly discussion about going downtown, “getting high” and seeing a movie. Talbert’s words, spoken in jest, are hardly sinister enough to corroborate the otherwise inadequate evidence of Tal-bert’s culpability.
Finally, the fact that Talbert’s thumbprint was found on the neck of the bowling pin is virtually irrelevant: For nearly any conceivable purpose, the neck of a bowling pin is the obvious place by which to grasp and move it.
In view of the unconscionable weakness of the Government’s case against Talbert, I think it indisputable that the jury inferred that Talbert committed the act of killing Nornes from the circumstantial evidence of Talbert’s thumbprint on the bowling pin. This court has held that “[w]hile inferences from facts that have been established by circumstantial evidence may be sufficient to sustain a verdict of guilt, mere suspicion or speculation cannot be the basis for the creation of logical inferences.” United States v. Thomas, 453 F.2d 141, 143 (9th Cir.1971), cert. denied, 405 U.S. 1041, 92 S.Ct. 1312, 31 L.Ed.2d 581 (1972). In order to infer from the thumbprint that Talbert killed Nornes, the jury must also have concluded that Talbert placed his thumbprint on the bowling pin during the commission of the offense. This conclusion lacks foundation in any direct or circumstantial evidence produced at trial and must, therefore, be based on mere speculation.
The logic of the Borum opinion, in rejecting a conviction based on such an unsupported inference, is sound. Prior to the majority’s opinion in the case before us, no federal court has rendered a decision inconsistent with Borum5 In Borum, a reasonable doubt as to the defendant’s guilt rested on the hypothesis that sometime, somewhere, the defendant touched the jars that contained or that were to contain the victim’s private coin collection. There was no evidence in Borum that the victim knew the defendant or had ever invited the defendant into his home. In contrast, the record below establishes that Talbert had been invited into Nornes’ shelter and might innocently have touched any of the objects therein.
To sustain Talbert’s conviction is to adopt the outrageous and unprecedented position that someone may be convicted of a crime simply because his fingerprint was found at the scene, and despite evidence that he could have impressed the fingerprint at a time other than during the commission of the crime. This outcome exposes the innocent to a tremendous threat of false accusation. Fingerprint evidence ought not be subject to the intolerable abuses possible, and actual, under the majority’s view. I am hopeful that, as precedent, the majori*535ty’s opinion will be confined to the facts of this case and will not be used in abuse of the logical weight to be accorded fingerprint evidence in future cases.
. See, e.g., United States v. Lonsdale, 577 F.2d 923, 924-25 (5th Cir.1978); United States v. Corso, 439 F.2d 956, 957 (4th Cir.1971); United States v. Collon, 426 F.2d 939, 941 (6th Cir. *5321970). See generally Annot. 28 A.L.R. 140-52 (Later Case Service 1981).
. See, e.g., United States v. Durham, 512 F.2d 1281, 1289 (5th Cir.), cert. denied, 425 U.S. 871, 96 S.Ct. 137, 46 L.Ed.2d 102 (1975); United States v. Roustio, 455 F.2d 366, 370 (7th Cir.1972); United States v. Scarpellino, 431 F.2d 475, 478 (8th Cir.1970). See generally Annot., 28 A.L.R. 140-52 (Later Case Service 1981).
The majority misleadingly states that “[tjhis court has held that fingerprint evidence, if sufficiently probative, may by itself support a conviction,” citing United States v. Crenshaw, 698 F.2d 1060 (9th Cir.1983), and United States v. Scott, 452 F.2d 660 (9th Cir.1971). These are the only two cases in which this court has previously addressed the adequacy of fingerprint evidence to sustain a conviction. In each case, there was both separate evidence of guilt and evidence that the prints had been impressed during the commission of the crime. Crenshaw, 698 F.2d at 1064; Scott, 452 F.2d at 662. Perhaps in using the qualification “if sufficiently probative,” the majority intends to preserve the universally established rule requiring corroboration of fingerprint evidence. Nevertheless, the majority’s language is ambiguous and may be subject to erroneous application in future cases. I wish, therefore, to set the record straight: In no prior case has this court held that fingerprint evidence, standing alone, may be sufficient to sustain a conviction.
In Crenshaw, three co-defendants appealed robbery convictions. One of the defendants, Gordon, argued that the District Court erred in admitting certain identification evidence against him and that some of the prosecutor’s statements made at trial were prejudicial to him. The evidence produced against Gordon at trial was substantial: (1) He left three fingerprints, known to have been left by one of the robbers, at the scene; (2) two bank employees, each twice, selected his picture from a photo line-up; (3) he carried a concealed weapon matching the description of one of the weapons used in the robbery; and (4) he occupied a hotel room with the other two robbers after the theft. Although this court found no error in connection with Gordon’s prosecution, the opinion erroneously and unnecessarily adds that “[tjhis court has held that fingerprint evidence alone is sufficient to convict,” 698 F.2d at 1064, citing Scott.
In Scott, the defendant was convicted of theft from a federal savings and loan association. The evidence produced against him at trial consisted of his fingerprint on a packet of checks identified as having been stolen on the relevant night, and his fingerprints found on a battery inside a flashlight left at the scene of the crime. This court affirmed the defendant’s conviction. In so doing, this court distinguished the “fingerprint only” cases on the grounds that
[h]ere, the fíashlight was identified as having been left at the Association during the crime. The identification of the defendant’s fingerprints upon one battery inside the flashlight might well have been insufficient circumstantial evidence to survive a motion for acquittal. But when that evidence is combined with positive fingerprint evidence upon the Association’s stolen check and its stolen traveler’s checks, the ring of circumstantial evidence tightens around the defendant. 452 F.2d at 662 (emphasis added).
Clearly, neither Crenshaw nor Scott held that fingerprint evidence, standing alone, may be sufficient to convict. Any contrary language in either opinion is pure dictum.
. The majority cites testimony that the bowling pins were dusty. The witness actually testified, more specifically, that the numerous pins in the gutters were essentially free of dust while the pins in the racks were dusty and appeared undisturbed.
. The majority’s statement that Talbert had a particular need for money is virtually unfounded. The only basis in the record for this assertion is from the fact that he telephoned his parents asking whether his disability check had arrived. The supposition that Talbert had a special need for money is clearly outweighed by evidence to the contrary.
. My research disclosed that only the jurisdiction of Kentucky has adopted a contrary rule. Mason v. Commonwealth, 357 S.W.2d 667, 668 (Ky.1962).