WILTON ALLEN DEDGE, APPELLANT,

v.

STATE OF FLORIDA, APPELLEE

SHARP, Judge.

Dedge appeals from judgments adjudicating him guilty of burglary with assault,1 two counts of sexual battery with a weapon,2 and aggravated battery.3 Because the eyewitness testimony identifying Dedge as the perpetrator of these crimes was equivocal, the errors at trial relating to his identification by a dog scent lineup cannot be considered harmless. Therefore, we reverse the judgments and remand this case for a new trial.

The state offered three means of identifying Dedge as the perpetrator: the victim’s testimony, a hair analysis, and a dog scent lineup. Shortly after her attack, the victim described her assailant as standing six feet to six feet two inches tall, and weighing one hundred sixty to two hundred pounds. Dedge’s height is approximately five feet five, and he weighs one hundred twenty-five pounds. A microanalyst for the Florida Department of Law Enforcement examined and compared hair samples found on the victim’s pillowcase and bed sheets with samples from Dedge. Although similarities were found, the expert witness could not say that the hair found on the sheets came from Dedge, but merely that Dedge “cannot be eliminated” as a suspect.

The primary identification of Dedge as the perpetrator of the crimes came through the activities of Harrass II, a highly trained and experienced police dog. Preston, an expert dog handler who has extensively participated in manhunts and lineups, and Harrass II, his purebred German shepherd, conducted a scent lineup in this case. This is apparently the first time a dog scent lineup has been discussed in an appeal in Florida. For this scent lineup, the victim’s sheets and four dirty sheets from the jail that Dedge did not use or touch, were placed in a lineup outside the presence of Preston and Harrass II. Har-rass II smelled the contents of a paper bag in which were paper towels used by Dedge. Preston then commanded Harrass II to search, and walked the dog down the lineup of sheets. During the second time Preston walked Harrass II down the lineup, after again telling the dog to search, Harrass II stopped and went over one of the sheets with his nose, which indicated to Preston that the scent Harrass II was looking for was on that sheet. That sheet was the victim’s sheet that Dedge had allegedly been on. Harrass II ignored the other sheets and again stopped at the same sheet in the lineup. Preston and Harrass II subsequently went to the victim’s home, and Harrass II also indicated that Dedge had been there by touching his nose to areas where he smelled Dedge.

Although the scent lineup was persuasive, there were two major errors committed at trial that relate to it. The first involved the disqualification of Sonny Bran-non, the defense’s only expert witness on human scent discrimination. The ability of a dog to identify human scents on objects at the crime scene more than three months after a crime is committed was a key point in this case.

Although Brannon was unavailable as a live witness, his testimony had been videotaped for the jury and judge to see and hear. The prosecutor challenged Bran-non’s qualifications as an expert in human scent discrimination. Both the prosecutor and the defense counsel argued as to what those qualifications were. Solely from listening to the arguments of counsel and without viewing the videotape proffer, the trial judge determined that Brannon did not qualify as an expert witness in human scent discrimination. However, “the question regarding an expert witness’ qualifications is a question of a fact to be decided by the trial court determined by the testimony adduced.” Upchurch v. Barnes, 197 So. 2d 26, 28 (Fla. 4th DCA 1967) (emphasis added). The trial court’s failure to hear any testimony and to conduct a proper voir dire by reviewing the videotape of Brannon cannot be deemed harmless error in this case.

The second error was the improper admission of hearsay. On redirect examination, the state’s expert witness, Preston, was asked if L. Wilson Davis, an established author and expert in the field of human scent discrimination, had ever testified as to the reliability of Harrass II. Preston answered affirmatively. This testimony was only relevant to show that an out-of-court witness had vouched for the reliability of Harrass II to identify scents. See Postell v. State, 398 So. 2d 851 (Fla. 3d DCA 1981). The defense also objected on the grounds that the question Went beyond the scope of the cross-examination. Both were valid objections. Since Harrass II’s abilities were the key to Dedge’s identification, the error is not harmless.

We reverse and remand this case for a new trial.

REVERSED AND REMANDED.

DAUKSCH and COBB, JJ., concur. . § 810.02(2)(a), Fla.Stat. (1981).

. §§ 794.011(2) & (3), Fla.Stat. (1981).

.§ 784.045(l)(b), Fla.Stat. (1981).