The defendant, Andrew P. Lorton, appeals his conviction after a bench trial in Claremont District Court (Robbins, J.) of driving while under the influence of intoxicating liquor (DWI). See RSA 265:82 (Supp. 2002). We reverse.
The record supports the following facts. On March 23, 2002, at approximately 11:30 p.m., the defendant was pulled over in Charlestown after Trooper Drew Keith of the New Hampshire State Police observed him make a right-hand turn without using a turn signal. Upon approaching the driver’s door, the trooper detected an odor of alcohol within the vehicle and a “modérate” odor of alcohol coming from the defendant’s breath. He also noted that the defendant’s eyes were “red and glassy” and that his face was “red and flushed.” The defendant produced his license and registration without incident and explained that he had been at home that *733evening when an intoxicated friend, who was in the passenger’s seat, called to ask for a ride home from a local restaurant. Upon inquiry, the defendant admitted to having one beer at the restaurant.
At Trooper Keith’s request, the defendant exited his vehicle and agreed to perform a set of field sobriety tests. The trooper gave instructions and demonstrated the walk-and-turn test and the one-leg stand test. With regard to the walk-and-turn test, Trooper Keith testified that the defendant did not walk heel to toe as instructed and that he held his hands slightly more than six inches from his side. The trooper testified that on the one-leg stand test, the defendant raised his arms higher than six inches and put his leg down after counting to ten rather than to thirty as instructed. The tests were conducted on the edge of the roadway, which, according to the trooper, was wet but not slippery. Based upon the defendant’s performance on these tests and his admission of having consumed alcohol, Trooper Keith concluded that he was under the influence and placed him under arrest. After he was taken to the police station, the defendant elected not to take a breath test. At the end of the booking process, the trooper drove the defendant home and returned his car keys. Following a bench trial in June 2002, the defendant was found guilty. This appeal followed.
The defendant argues that there was insufficient evidence presented at trial to support his conviction. Specifically, he contends that the State failed to establish impairment beyond a reasonable doubt. When the evidence presented is circumstantial, it must exclude all rational conclusions except guilt in order to be sufficient to convict. State v. Richardson, 141 N.H. 139, 141 (1996). In applying this standard, we view the evidence in the light most favorable to the State and examine each piece of evidence “in the context of all the evidence, not in isolation.” Id. (quotation omitted).
RSA 265:82 provides, in pertinent part, that: “No person shall drive or attempt to drive a vehicle upon any way: (a) While such person is under the influence of intoxicating liquor____” RSA 265:82,1(a). We have stated that in order to convict a person under this provision, the State need only prove “impairment to any degree.” State v. Taylor, 132 N.H. 314, 316 (1989).
Viewing all the evidence in the light most favorable to the State, we cannot say beyond a reasonable doubt that the circumstantial evidence in this case excluded all rational conclusions except that the defendant was guilty. The evidence against the defendant included Trooper Keith’s observations of his red, glassy eyes, his red, flushed face and his performance on two field sobriety tests. In addition, the trooper smelled *734an odor of alcohol on the defendant’s breath. The defendant admitted to having had a drink and refused to submit to a breath test.
On the contrary, Trooper Keith testified that the defendant was not operating his vehicle erratically and that he stopped in a safe and prudent manner. He produced both his license and registration without incident, which Trooper Keith acknowledged was a form of a divided-attention test. Furthermore, the defendant had no trouble exiting the vehicle and walked with a normal gait without stumbling or faltering in any way. Moreover, he stood as directed during the instructional stage of the field sobriety tests, and during the walk-and-tum test, he walked in a straight line without an actual line on the pavement to follow, did not lose his balance, counted the appropriate number of steps and properly executed the turning phase of the test.
Field sobriety tests were developed on behalf of the National Highway Traffic Safety Administration (NHTSA) and are described in the NHTSA student manual for DWI detection and standardized field sobriety testing. The manual outlines the test condition standards for each test, stating that the walk-and-turn test “requires a designated straight line, and should be conducted on a dry, hard, level, nonslippery surface.” National Highway Traffic Safety Administration, Student Manual: DWI Detection and Standardized Field Sobriety Testing, VIII-21. The manual also lists standardized “clues” for police officers to look for during a suspect’s performance on the tests and indicates that there is a strong likelihood of impairment when a suspect exhibits two or more “distinct clues” on both the walk-and-turn and the one-leg stand tests. Id. at VIII-21, VIII-24. “[I]t is the officer’s subjective evaluation of the suspect that results in the determination of whether a ‘clue’ is present or not.” United States v. Horn, 185 F. Supp. 2d 530, 538 (D. Md. 2002). Although the walk-and-tum and the one-leg stand tests “enhance the basis and reliability of [the officer’s] opinion as to whether the driver’s performance has been adversely affected by intoxicating liquor,” their results “do not possess the scientific reliability or the same degree of certitude attributed to the chemical analysis of the alcoholic blood content by a breathalyzer or other such methods.” State v. Arsenault, 115 N.H. 109, 111 (1975).
Here, Trooper Keith deemed the defendant’s performance on the walk- and-tum test a “failure” based upon the minimum of two assignments of error — his observation that the defendant’s steps were one to two inches apart and his “rough estimate” that the defendant held his hands out “slightly” more than six inches from his side. The trooper testified, however, that although he instructed the defendant to walk heel to toe and *735to keep his arms by his sides, test subjects are allowed a half inch between their feet and can raise their arms up to six inches from their sides without any assignment of error. Thus, the trooper’s “rough estimate” that the defendant’s arms were “slightly” higher than six inches from his body does not provide a “distinct clue,” as directed by the NHTSA manual.
Because the field sobriety tests were conducted under less than ideal conditions and the results of the defendant’s performance on the walk-and-tum test were essentially unremarkable, it is just as rational to conclude that his performance was attributable to innocent factors as it is to conclude that his performance was due to impairment.
Perhaps the most damaging evidence against the defendant was his refusal to take the breath test. The defendant explained that he refused because he had learned on television that breath tests produce inaccurate results. Viewed in the light most favorable to the State, however, the refusal supports an inference that the defendant sought to suppress evidence of his guilt. See RSA 265:88-a (1993); State v. Cormier, 127 N.H. 253,256-58 (1985). Yet the probative value of this inference is not the same in all cases. Cf Cormier, 127 N.H. at 260 (Batchelder, J., concurring specially) (trial court has discretion to exclude evidence of refusal). The inference is much stronger, for example, when coupled with an admission. See, e.g., State v. Parmenter, 149 N.H. 40, 43 (2002) (defendant refused test, stating, “I know Pm over and you know I’m over”); South Dakota v. Neville, 459 U.S. 553, 555 (1983) (defendant refused test, stating, “I’m too drunk, I won’t pass the test”). The absence of any significant corroborating evidence in this case — i.e., erratic driving or outright failure of a sobriety test — further weakens the inference. Accordingly, under the facts of this case, there was insufficient evidence to establish beyond a reasonable doubt that the defendant operated his vehicle while impaired.
Reversed.
Brock, C.J., and Nadeau, Dalianis and Duggan, JJ., concurred.