Defendant Dale Leon Jonas appeals from a conviction of aggravated robbery, a first degree felony. He argues that the trial court abused its discretion in refusing to submit to the jury his requested cautionary instruction on the reliability of eyewitness identification. Jonas also contends that the evidence is insufficient to establish guilt beyond a reasonable doubt. We reverse on the first ground and remand for retrial.
On March 18, 1984, just after midnight, David E. Larson was driving south on Wasatch Boulevard, near the mouth of Little Cottonwood Canyon, when the headlights of his truck illuminated a man standing on the shoulder of the road and apparently attempting to flag down someone to assist him. Larson pulled over to the side of the road and got out of his truck. Although the evidence is conflicting, Larson either walked to the front of his truck, where the headlights were on, or to the rear of the truck, where the only light came from the red taillights and emergency blinkers. As Larson got close to the man, he felt a sharp pain in his ribs. He looked down and saw the barrel of a handgun. Larson did not recall anything after that until a couple of hours later when he awoke, groggy and with a bump on his head.
A Salt Lake County deputy sheriff found Larson semiconscious in a snowbank near the right side of his truck about two hours later. Larson’s wallet, containing $250, a driver’s license, and a fitness club membership card, was missing. He then gave the deputy a description of his assailant as a clean-shaven man, six feet tall, with long, blond hair, wearing an army jacket and brown Levi-type pants with a tear in the right knee. Footprints and tire tracks, which were apparently not those of Larson or his vehicle, were observed at the scene of the crime. Although pictures of the tire tracks and footprints were taken, nothing was included in the pictures to enable a viewer to determine their dimensions. No castings of the footprints or tire tracks were made.
Three days later, Mr. Larson attempted to identify his assailant by reviewing a series of photographs provided by a detective. At that time, he mentioned that a prominent identifying feature of his assailant was “a crooked nose, as if it may have been broken at some time.” Mr. Larson then selected a photograph of an individual having a crooked nose, although that person did not have blond hair. Then Mr. Larson went through a second set of photographs and picked out another picture, identifying it as “a close possibility ... except for the nose.” This “close possibility” was the defendant, Dale Leon Jonas. The detective who was assisting Mr. Larson with the photo array then took the picture of Jonas and returned with six photos, including a more recent one of Jonas. From this new group of six pictures, Larson selected the picture of Jonas and identified him as his assailant.
*1380Jonas was arrested on March 22, 1984. No search was ever conducted of his home, nor was there any attempt to locate or search a vehicle for an army jacket and brown Levi-type pants with a tear in the right knee. The record is silent as to any comparisons made of the tire track and footprint photos with any tires or shoes associated with Jonas.
At trial, the only evidence linking Jonas to the crime was Larson’s eyewitness testimony. At the close of trial, defendant’s counsel requested a cautionary instruction regarding the inherent weakness of eyewitness identification. The trial court declined to give a cautionary instruction and instead ga ve a standard burden-of-proof instruction which did not address the special problems associated with eyewitness identification.
Until our recent decision in State v. Long, 721 P.2d 483 (Utah 1986), this Court had consistently held that the decision to give a jury cautionary instructions regarding the reliability of eyewitness identification was a matter strictly within the trial court’s discretion and to be decided by considering “the totality of the circumstances.” State v. Reedy, 681 P.2d 1251, 1254 (Utah 1984); see also State v. Schaffer, 638 P.2d 1185 (Utah 1981); State v. Malmrose, 649 P.2d 56 (Utah 1982); State v. Newton, 681 P.2d 833 (Utah 1984); State v. Watson, 684 P.2d 39 (Utah 1984); State v. Bingham, 684 P.2d 43 (Utah 1984); State v. Tucker, 709 P.2d 313 (Utah 1985); State v. Booker, 709 P.2d 342 (Utah 1985).
In Long, we abandoned the discretionary approach to cautionary instructions “whenever eyewitness identification is a central issue in a case and such an instruction is requested by the defense.” 721 P.2d at 492. However, the Long decision was specifically limited in its application to cases tried after its date of issuance. Id. Trial of the present case preceded Long, and therefore defendant’s claim must be evaluated under the prior case law to determine whether the trial court abused its discretion in refusing to give the requested cautionary instruction.
In each of our pre-Long cases involving the refusal of a trial court to give a cautionary instruction, we have held that the trial court did not abuse its discretion under the particular circumstances presented, although we did expressly recognize “that under suitable circumstances a cautionary instruction of the type requested would be required.” State v. Tucker, 709 P.2d at 316.
However, in all of the pre-Long cases, unlike the one before us, it is highly likely that the result would have been exactly the same even if a cautionary instruction had been given. In Reedy, the eyewitness was an attendant at a gas station who had two excellent opportunities to view his assailants in broad daylight, both prior to and during the actual robbery. In Schaffer, the witness had ample opportunity to observe the defendant as he and his accomplice entered the pizza parlor where the witness was working, ordered food, and waited for other customers to clear out of the pizza parlor before robbing money from two separate cash registers, pushing the witness and her manager through a well-lighted kitchen, and locking them into the restaurant’s refrigerator. In Malmrose, the victim who identified the suspect was raped in broad daylight and observed her assailant for seven or eight minutes during the attack. In Newton and Watson, the identifying witnesses had opportunities to observe the suspects in well-lighted stores over a period of several minutes both before and during the robberies. In Bingham, two eyewitnesses observed the defendant at noon on a sunny day under circumstances of alerted sensitivity caused by the defendant’s suspicious behavior and heightened by the witnesses’ membership in the Neighborhood Watch program. In Booker, the victim was completely conscious while her assailant was present for up to twenty minutes in her darkened apartment and during which time he talked *1381to her extensively.1 In Malmrose, Watson, and Booker, there was also corroborating evidence to support the eyewitness identification. And in Tucker, the circumstantial corroborating evidence was sufficiently strong to render the eyewitness identification almost redundant. Moreover, in all of those cases involving photo array identification, there was no evidence of suggestive investigatory procedures.
In contrast, the present case presents a classic example of circumstances under which a cautionary instruction on reliability of eyewitnesses is necessary. The victim had only a few seconds, at most, to observe his assailant; the observation occurred at night; and the evidence is conflicting as to whether it occurred in the illumination of the headlights or of the taillights of the victim’s truck. During this time, the victim was struck in the ribs with a handgun and knocked unconscious. The witness initially identified his assailant as clean shaven, but then testified at trial that the assailant had a moustache. (Jonas had a full moustache when he appeared at the trial four months after the crime.) During the photo array, the witness initially searched for a man with a crooked nose, but ultimately identified defendant, who has no irregularities in the shape of his nose or in his facial features. The victim’s trial testimony demonstrated that there was a great deal of uncertainty and hesitancy in the manner in which he arrived at his identification of defendant during the photo array. In addition, there were no other witnesses to confirm the victim’s identification of defendant, and there was no corroborating evidence of any kind.
Under these circumstances, the trial court abused its discretion in not giving a cautionary instruction about the weaknesses inherent in eyewitness identification. The conviction is reversed, and the case is remanded for a new trial on that ground.
HALL, C.J., and DURHAM, J., concur.. State v. Booker was essentially a voice identification case in which the veracity of eyewitness identification was of secondary significance.