OPINION OF THE COURT
SCHNEPP, J.Defendant appeals from a judgment convicting him of manslaughter in the second degree, vehicular manslaughter, criminally negligent homicide, driving while under the influence of alcohol and other charges in connection with a series of motor vehicle accidents which occurred on October 6, 1984 on and near the Ford Street Bridge in the City of Rochester. Defendant’s speeding car sideswiped two vehicles as it crossed the bridge from east to west and finally broadsided a car at the western approach to the bridge, killing the driver. When he emerged from his car, defendant told an off-duty police officer who witnessed the accident that he could not slow down because his car’s accelerator stuck, causing him to lose control of the vehicle. Defendant has consistently maintained this claim throughout these proceedings and produced expert evidence at trial that his vehicle had mechanical and design defects which could have caused it to accelerate uncontrollably.
Police officers called to the scene of the accident observed that defendant’s eyes were bloodshot, that he was unsteady on his feet and that there was a strong odor of alcohol on his breath. The police also administered a preliminary breath test on an Alco-Sensor device which apparently indicated that defendant had consumed alcohol (see, Vehicle and Traffic Law § 1193-a). The trial testimony reflected that defendant was then arrested "based on the results of the test”, although there was testimony that he had been earlier arrested based on the observations by the police. Within one hour following the Alco-Sensor screening and his arrest, defendant submitted to a breathalyzer test which indicated that his blood alcohol level was .14%. At trial, the People introduced the results of the breathalyzer test and also adduced testimony concerning *75the preliminary breath test on the Alco-Sensor. Defendant objected to any reference to the Alco-Sensor test; however, the trial court allowed the testimony as relevant to the question of whether the police had reasonable grounds to proceed with a breathalyzer test (see, Vehicle and Traffic Law § 1194 [1] [1], [2]) and later instructed the jury that this proof could be considered "merely on the issue of the police officer’s reasonable grounds to believe that the defendant was driving while intoxicated.”
The prosecution adduced proof that the breathalyzer test had been administered within two hours of defendant’s arrest, that the breathalyzer was in proper working condition when the test was given and that the chemicals used in the test were of the proper kind and in the proper proportion (see, People v Gower, 42 NY2d 117). Defendant challenged only the reliability of the test results, however, and did this through the testimony of the breathalyzer operator, developed on cross-examination, that defendant had a cut on his lip when tested, that the operator found blood on the outside of the disposable mouthpiece of the breathalyzer after the test, although he was not sure if there was any blood in defendant’s mouth, and that, although the mouthpiece has a stop valve to prevent contamination of the machine, the presence of blood in the mouth can produce an inaccurate test result. On summation, after defense counsel theorized that blood from defendant’s mouth or lip may have contaminated the breath sample, the prosecutor erroneously stated that there was no evidence of the effect of such blood on the breathalyzer test.
At trial defendant also produced a witness, whom the prosecution conceded to be an expert mechanic, to testify concerning the alleged defects in his vehicle. This witness testified that a cable controlling the throttle had partially unraveled and could have caused unintended acceleration and that the placement of the cruise control unit on defendant’s model car was a design defect which could also result in such acceleration. The witness was apparently prepared to testify that the manufacturer had changed the design and relocated the cruise control units on later model cars; however, this testimony was excluded by the court as irrelevant.
In our view, the receipt of evidence concerning the AlcoSensor test, the misstatement by the prosecutor on summation concerning the critical testimony of the breathalyzer operator and the exclusion of defendant’s expert evidence on design *76changes in later model cars were errors which cumulatively denied defendant a fair trial and require a reversal.
The People argue that the Alco-Sensor test was not admitted for the purpose of showing that defendant was actually intoxicated but only to establish that the police had grounds to request him to submit to a breathalyzer test. Defendant contends that the evidence concerning the Alco-Sensor test was used improperly by the jury as additional evidence of intoxication. We agree that this proof, from which the only logical inference is that defendant failed the Alco-Sensor test, may have appeared to the jury to be additional reliable scientific evidence of intoxication, particularly since the officer testified that he had been "trained in the operation of this device” and had used it for five years and the court failed to caution the jury immediately that the testimony had been received for the limited purpose of establishing a basis to proceed with a breathalyzer test.
The Alco-Sensor testimony was clearly not admissible to show intoxication. It is well settled that "[t]here must be a sufficient showing of reliability of the test results before scientific evidence may be introduced” (People v Spaight, 92 AD2d 734, 735; compare, People v Donaldson, 36 AD2d 37, 40). "[Scientific evidence will only be admitted at trial if the procedure and results are generally accepted as reliable in the scientific community” (People v Hughes, 59 NY2d 523, 537). Thus, the Alco-Sensor evidence should have been excluded because as it was presented to the jury it served as proof of intoxication and the People failed to lay a proper foundation showing its reliability for this purpose. No expert testimony was submitted as to the accuracy of this device and the scientific principles on which it is based. The record is completely barren of scientific evidence which would establish the reliability of the test. Moreover, cases from other jurisdictions hold that the Alco-Sensor test is not reliable evidence of intoxication (see, Boyd v City of Montgomery, 472 So 2d 694, 697 [Ala Crim App]; State v Thompson, 357 NW2d 591, 593-594 [Iowa]; State v Smith, 218 Neb 201, 352 NW2d 620, 624; State v Orvis, 143 Vt 388, 465 A2d 1361, 1362-1363; cf. State v Albright, 98 Wis 2d 663, 298 NW2d 196, 203 [Ct App]).
Although an Alco-Sensor test is not admissible as evidence of intoxication, breath screening devices have won acceptance as being sufficiently reliable to establish probable cause for an arrest (see, Matter of Smith v Commissioner of Motor Vehicles, 103 AD2d 865, 866; see also, Boyd v City of Montgomery, 472 *77So 2d 694, 697, supra; State v Thompson, 357 NW2d 591, 593, supra; State v Orvis, 143 Vt 388, 465 A2d 1361, 1362-1363, supra) and may be used by the police to establish a basis to request a breathalyzer test (Vehicle and Traffic Law §§ 1193-a, 1194 [1] [2]). The People claim that the evidence was properly received for this limited purpose; however, as we have previously stated, this limitation was not sufficiently expressed to the jury. Moreover, the issue as to whether grounds existed to request a breathalyzer test was not raised in this case.
The People contend that they are required to prove reasonable grounds to require a breathalyzer test in every driving while intoxicated (DWI) case in order to lay a proper foundation for proof of the result of a breathalyzer test. Their view is supported by the "pattern” Criminal Jury Instructions which state that the jury must find "that the presence of [a percentage of alcohol] in the defendant’s blood was determined in accordance with the law” (3 CJI [NY] V&TL 1192 [2] p 2277). According to the CJI, Vehicle & Traffic Law §§ 1192 and 1194 require the jury to find that five steps have been taken in order to conclude that the results of the chemical test are valid. Our concern is with the first of such steps which requires the jury to find either: "[T]hat [the arresting officer] had reasonable grounds to believe that the defendant was operating a motor vehicle” while he was intoxicated (at 2277; see, Vehicle and Traffic Law § 1194 [1] [1]) or, that the arresting officer gave defendant a breath test to determine if defendant had consumed any alcohol, and the test administered showed that defendant had consumed alcohol (at 2278-2279; see, Vehicle and Traffic Law § 1194 [1] [2]).
We disagree with the implication of these instructions that either reasonable cause to arrest or the results of a preliminary breath screening test must be shown in every DWI case regardless of whether the defendant has raised lack of reasonable grounds to require a breathalyzer test as an issue. Case law holds that there must be evidence in the record from which the trier of fact can determine whether a breathalyzer test was performed within two hours of arrest (see, People v Mertz, 68 NY2d 136, 146; People v Ambrozik, 104 AD2d 693, 694). The People argue that proof of probable cause to arrest or failure of a screening test is also required to establish a foundation for the breathalyzer test; that is simply not the case. It is only when the legality of such a test or an arrest is challenged, e.g., by a pretrial suppression motion, that the People have the burden "to come forward with evidence” of *78probable cause and a hearing Judge is faced with deciding whether the People have met their burden of proof. The analysis required is "largely the same as that used by a magistrate in passing on an application for an arrest or search warrant” (People v Dodt, 61 NY2d 408, 415). Moreover, with relation to the admission of breathalyzer test results the only foundational requirement is that "evidence must be introduced both that the breathalyzer was in proper working condition when the test was given to defendant, and that the chemicals used in the test were of the proper kind and in the proper proportion” (People v Garneau, 120 AD2d 112,115). Section 1194 (1) provides that a test conducted more than two hours after arrest is stale, thus the time element also must be proved, but we can discern no requirement in the statute that reasonable grounds to require a breathalyzer test must be presented to the jury as an issue of fact in every DWI case.
The statutory requirement that the police have reasonable grounds, in the form of observation of behavior or failure of a screening test, to require that a person submit to a breathalyzer test is analogous to the requirement that an arrest must be founded on probable cause and creates similar issues for the court. "The question of probable cause is a mixed question of law and fact: the truth and existence of the facts and circumstances bearing on the issue being a question of fact, and the determination of whether the facts and circumstances found to exist and to be true constitute probable cause being a question of law” (People v Oden, 36 NY2d 382, 384; see, People v Morales, 42 NY2d 129, 134, cert denied 434 US 1018). "If the facts and circumstances adduced as proof of probable cause are controverted so that conflicting evidence is to be weighed, if different persons might reasonably draw opposing inferences therefrom, or if the credibility of witnesses is to be passed upon, issues as to the existence or truth of the facts and circumstances are to be passed upon as a question of fact; however, when the facts and circumstances are undisputed * * * the issue as to whether they amount to probable cause is a question of law” (People v Oden, supra, p 384). Here, defendant did not controvert at trial the existence of grounds to administer a breathalyzer test. Under these circumstances, the question whether the police had grounds to administer such a test presented a pure question of law and should not have been submitted to the jury. Thus, proof that defendant failed the Alco-Sensor test, which was ostensibly admitted *79solely for the purpose of establishing the existence of reasonable grounds to administer a breathalyzer test under Vehicle and Traffic Law § 1192 (1) (1) or (2), should have been excluded as irrelevant to the issues before the jury. In our view, evidence regarding the Alco-Sensor test had no place in the trial and the objection to its admission should have been sustained. The jury should not have been given the opportunity "to use the screening test result to corroborate the evidential test result” (Brent & Stiller, Handling Drunk Driving Cases, ch 13, at 229; see, State v Thompson, 357 NW2d 591, 593, supra; State v Smith, 218 Neb 201, 352 NW2d 620, 624, supra).
Although the receipt of evidence relating to the Alco-Sensor test was error, reversal is not required if the evidence was merely cumulative of other proof of intoxication and can be deemed harmless. Where the reliability of the results of a breathalyzer test showing defendant’s blood alcohol content to be greater than .10% are not credibly challenged, improper references to preliminary screening tests may be disregarded as harmless error (see, Boyd v City of Montgomery, 472 So 2d 694, 697-698, supra; State v Smith, 218 Neb 201, 352 NW2d 620, 624, supra). Here, however, defendant raised a credible challenge to the accuracy of the breathalyzer test. The breathalyzer operator testified that the presence of blood in the subject’s mouth may result in a false reading on the breathalyzer and that he observed a cut in defendant’s lip and found blood on the mouthpiece of the breathalyzer after he tested defendant. Under these circumstances, evidence that defendant failed the Alco-Sensor test tended to corroborate the challenged questionable breathalyzer results and may not be deemed harmless (see, State v Thompson, 357 NW2d 591, 594, supra). We conclude "that there is a significant probability, rather than only a rational possibility * * * that the jury would have acquitted the defendant had it not been for the error * * * which occurred.” (People v Crimmins, 36 NY2d 230, 242.)
The prosecutor’s misstatement on summation concerning the critical testimony of the breathalyzer operator also tended to bolster the credibility of the breathalyzer results. While the jury was cautioned that its recollection of the evidence was controlling, the prosecutor’s comments were improper and easily might have confused the jury.
Finally, the trial court improperly precluded defendant’s expert from testifying as to a design modification concerning *80the placement of cruise control units on later model cars and as to the reasons for the change. The prosecutor stipulated that the witness was an expert and under these circumstances there was no basis to exclude the testimony. The evidence was certainly relevant to defendant’s claim that his car malfunctioned and accelerated out of control and should have been allowed. Subsequent design modifications, although inadmissible in some civil cases (see, Caprara v Chrysler Corp., 52 NY2d 114, rearg denied 52 NY2d 1073), may be admitted in criminal cases if relevant; the rationale underlying excluding such evidence in civil cases where the defendant is the accused culpable designer or manufacturer is ill-applied in criminal cases. Public policy does not justify the exclusion of this evidence where the manufacturer of the product cannot be prejudiced by its admission and the defendant has a constitutional right to present evidence. There is no reason to reject evidence of this character in criminal cases if relevant.
Accordingly, the judgment should be reversed and a new trial granted.
Denman, J. P., Boomer, Pine and Lawton, JJ., concur.
Judgment unanimously reversed, on the law, and new trial granted.