Testing

Field Sobriety Tests: Refusing Them, and What 9 Courts Held

Three roadside tests make up what officers are trained to administer as a standardized battery, and courts do not treat all three the same way. Two of them are behavior a witness can describe. The third is treated, in state after state, as scientific evidence — which means it carries a foundation requirement, and a ceiling on what it can be used to prove.

This page is built on what courts have actually held, quoted from the official reporters and read on the date shown against each opinion in the sources. It is deliberately not built on the federal training manual, and the last section explains why.

The three tests, and the one that is treated as science

The opinions below name the same battery. Blake lists it as “walk and turn, the one-leg stand, and the HGN”; Campbell names the same three and notes the officer was trained and certified to administer them.

Horizontal gaze nystagmus

Nystagmus is involuntary jerking of the eye. The officer moves a stimulus horizontally and watches how the eye tracks it. Three of the opinions read here state that six clues are available, three per eye — and that is the only clue count any of them gives.

Walk-and-turn

A divided-attention task: hold a position while listening to instructions, then walk a line heel-to-toe, turn as instructed, and return. Courts treat this as behavior that a witness observed.

One-leg stand

Balance held together with counting aloud. Same category as the walk-and-turn.

The split matters because it decides what the prosecution has to do before the evidence comes in. Florida’s Fourth District put the distinction in a single sentence:

We answer these questions by distinguishing between: (1) psychomotor field sobriety tests, in which the defendants are requested to perform certain tasks; and (2) the horizontal gaze nystagmus (HGN) test, which is scientific evidence of a physiological phenomenon associated with intoxication.

— State v. Meador, 674 So. 2d 826 (Fla. 4th DCA 1996)

The psychomotor tests are admissible on that view without a scientific predicate, subject to a limit the court attached: they may not be “characterized in a manner which unduly emphasizes the significance of the test results.” The eye test is a different animal.

The line the courts that reach it drew

Read the nine states’ opinions together and something that no single opinion announces becomes obvious. They apply different evidentiary standards — Frye, Kelly, Harper, the Rule 702 family, Florida’s traditional scientific predicate, and in South Carolina no scientific predicate at all — and the courts that reach the question end up in the same place:

HGN can support a conclusion that a driver was impaired. It cannot simply be turned into a number — barred outright in Texas, Arizona and South Carolina, and in Georgia not without a scientific foundation.

Texas said so while holding the test reliable:

We are unable to conclude, however, that the HGN technique is a sufficiently reliable indicator of precise BAC.

— Emerson v. State, 880 S.W.2d 759 (Tex. Crim. App. 1994)

Arizona said so as a limit written into the holding itself:

the horizontal gaze nystagmus test satisfies the Frye test for reliability and may be admitted in evidence to corroborate or attack, but not to quantify, the chemical analysis of the accused’s blood alcohol content.

— State v. Superior Court, 149 Ariz. 269, 718 P.2d 171 (1986)

And Georgia reversed a conviction over exactly that step — an officer testifying that four of six clues indicated a blood alcohol level at or above 0.08:

We conclude that the evidence presented by the State in this case was insufficient to establish the scientific validity or reliability of any correlation between a particular number of clues on an HGN test and a numeric blood alcohol content, whether a specific percentage or “equal to or greater than” a specific percentage.

— Spencer v. State, 302 Ga. 133, 805 S.E.2d 886 (2017)

South Carolina drew the same line while treating HGN as an ordinary field test rather than a scientific one:

We hold that evidence arising from HGN tests is not conclusive proof of DUI. A positive HGN test result is to be regarded as merely circumstantial evidence of DUI. Furthermore, HGN tests shall not constitute evidence to establish a specific degree of blood alcohol content.

— State v. Sullivan, 310 S.C. 311, 426 S.E.2d 766 (1993)

And Washington drew it for drugs, in the drug-recognition setting:

Moreover, an officer may not testify in a fashion that casts an aura of scientific certainty to the testimony. The officer also may not predict the specific level of drugs present in a suspect.

— State v. Baity, 140 Wash. 2d 1 (2000)

Five courts across three decades — Arizona in 1986, South Carolina in 1993, Texas in 1994, Washington in 2000, Georgia in 2017 — on different standards. That is a stronger signal than any one of them alone, and it is the single most useful thing to know about this evidence.

What each state’s courts actually held

Texas — reliable, and still not a number

Emerson ran HGN through the Texas novel-scientific-evidence framework and cleared it twice over: “the theory underlying the HGN test is sufficiently reliable pursuant to Texas Rule of Criminal Evidence 702”, and “the technique employed in the HGN test, as designed and promoted by NHTSA, is reliable pursuant to Rule 702.” Then came the sentence that limits it. The full Texas penalty framework is on the Texas DWI laws page.

Arizona — corroborate or attack, never quantify

Blake is the oldest of the opinions read here and the most explicitly drafted. HGN may be admitted “to corroborate or attack, but not to quantify, the chemical analysis” — and the court added that it “may not be used to establish the accused’s level of blood alcohol in the absence of a chemical analysis”. Arizona’s own sentencing floors turn on that reading, which is why the distinction has teeth there: see Arizona DUI law.

The same opinion also settled something broader — that roadside sobriety tests are searches under the Fourth Amendment, justified by reasonable suspicion based on specific articulable facts rather than requiring probable cause.

Georgia — a count of clues is not a BAC

Spencer is the most concrete. The officer testified that four of six clues “generally indicates a blood alcohol level equal to or greater than a .08”; no studies were identified, none were admitted, and no scientific or medical testimony was presented. The conviction was reversed. Georgia DUI law covers the rest of the ladder.

California — a new scientific technique

Leahy answered a narrower question and answered it structurally: HGN falls within Kelly, so it is not something an officer can simply describe from the stand without the foundation being laid. See California DUI law.

Florida — two different kinds of evidence

Meador is the cleanest statement of the split, and it is a District Court of Appeal decision rather than a Florida Supreme Court one — worth knowing if you are reading it as authority. Florida DUI law covers the statutory side.

Colorado — that is expert testimony

Campbell approaches it from the witness rather than the technique: an officer testifying about HGN results is giving expert testimony under CRE 702 and has to be qualified accordingly. See Colorado DUI law.

South Carolina — a field test like the others

Sullivan went the other way on method. The circuit court had ruled the HGN testimony inadmissible, doubting a police officer’s ability to conduct and interpret what it saw as a medical test; the Supreme Court reversed that ruling, holding HGN evidence admissible, like the other field sobriety tests, when it is used to elicit objective signs of sobriety or insobriety. It then fenced it in: circumstantial evidence only, never conclusive, and never a specific blood alcohol content. See South Carolina DUI law.

Washington — no aura of scientific certainty

Baity is a drug case. It held the twelve-step drug-recognition protocol, including HGN applied to drug intoxication, generally accepted under Frye, confined the holding to evaluations where all twelve steps were done, and limited the officer: no aura of scientific certainty, and no prediction of the specific level of drugs. See Washington DUI law.

Minnesota — foundation, and an instruction on request

Klawitter is also a drug-recognition case. The court allowed an officer’s opinion based on the protocol and on nystagmus testing, with conditions:

After careful consideration we conclude that opinion testimony based on nystagmus testing is admissible if a sufficient foundation has been laid for the opinion expressed and provided that the trial court, when requested, gives an appropriate cautionary instruction.

— State v. Klawitter, 518 N.W.2d 577 (Minn. 1994)

It added that the state may not exaggerate the officer’s credentials by referring to the officer as a “Drug Recognition Expert”. See Minnesota DWI law.

Can you refuse a field sobriety test?

The tests need the driver’s cooperation, and some drivers decline. The legal question is what a refusal then costs, and among the opinions read for this page one decides it: Washington’s Mecham.

Mark Mecham was stopped on an outstanding warrant, arrested and handcuffed, and then asked to do the three tests. He refused, and the refusal was used against him at trial. The Supreme Court reversed his conviction, on a split that is worth reading closely:

Accordingly, five justices hold that an FST is a seizure but not a search so long as the suspect has not already been arrested for an unrelated offense and the seizure is justified under Terry.

— State v. Mecham, 186 Wash. 2d 128 (2016)

Four justices — the lead opinion — would have gone further and held that he had no constitutional right to refuse, so the refusal was admissible as evidence of consciousness of guilt. The fifth, Justice Fairhurst, agreed that the tests are ordinarily a seizure justified by reasonable suspicion, but said that using them on a man already under arrest for an unrelated offense was a search needing a warrant or a warrant exception, so the refusal should not have come in. Four dissenting justices also voted to reverse, on two different grounds: two because a field sobriety test is a search, and two because a refusal of tests the officer described as voluntary should carry no evidentiary weight at all.

So the Washington answer has two parts, and only one of them is settled by a majority. Before an arrest for something else, five justices treat the tests as a seizure that reasonable suspicion can justify; only the four-justice lead opinion says a refusal in that setting may be used as evidence. After an arrest on an unrelated warrant, five justices voted to reverse a conviction in which the refusal was used. Arizona’s Blake, above, treats roadside tests that involve no long delay or unreasonable intrusion as searches that reasonable suspicion can justify.

Two things that are often confused with the three tests are not the same:

• The roadside breath device. Colorado’s statute has the officer tell the driver that the preliminary screening test may be refused, and keeps both its result and any refusal out of the trial, except at a hearing on probable cause held outside the jury’s presence (C.R.S. § 42-4-1301(6)(i).)
• The evidential breath or blood test after arrest. That is implied consent, and a refusal there is a different matter: in Colorado it is admissible at a DUI or DWAI trial under § 42-4-1301(6)(d); what it costs elsewhere is in refusing a breathalyzer.

None of the other opinions read for this page decides the refusal question, so this page does not state a rule for those states.

Why the clue list is the evidence

Because these tests are scored against a fixed list rather than on an overall impression, the record of which clues were observed is what the evidence actually consists of. A report that states a conclusion without the underlying clues has skipped the step that makes the test a test, and Spencer is what that looks like when it reaches a supreme court: an assertion about what the clues meant, with nothing behind it.

This is also why the battery is administered as a battery. Each test is scored separately and recorded separately, and the questions a defense asks are about administration — was the stimulus moved at the right speed and held long enough, was the surface level, were the instructions given before the performance was scored.

None of that is a question about whether the driver was drunk. It is a question about whether the thing in the report is the test the officer was trained to give.

What this page does not say, and why

An earlier version of this page carried a table of validation statistics — accuracy percentages for the battery and for each test alone, and correlation coefficients between HGN and blood alcohol — attributed to a San Diego study run with seven officers. That table has been removed.

The figures could not be traced to a document that could be opened. The natural source is NHTSA’s own manual and the validation studies behind it, and on September 21, 2026 every federal route to them refused this site’s requests: nhtsa.gov, one.nhtsa.gov and the Department of Transportation’s National Transportation Library all returned 403. Publishing accuracy percentages for a forensic test from memory is precisely the thing this site exists not to do.

Two smaller removals for the same reason. The earlier version gave clue counts of eight for the walk-and-turn and four for the one-leg stand. Those numbers appear in none of the opinions read for this page, so the table above says so rather than repeating them. And the claim that HGN is the most predictive of the three tests rested on the same unsourced table; it is gone with it.

What replaced all of it is ten opinions from nine states, quoted, each linked to the volume it was printed in. If the federal sources become reachable, the validation data will be added with its citation — not before.

Related

• What makes a DUI checkpoint lawful — the other roadside procedure with its own rulebook, with nine states’ high courts quoted
• Refusing a breathalyzer — what implied consent trades, and why refusing does not leave the state without a case
• DUI lawyer by state — what the charge is called where you were arrested
• DUI law by state — limits, deadlines and prior-offense windows side by side

Sources

All ten opinions were read in full from the Caselaw Access Project mirror of the official reporters, on the date shown against each. The citation given is the bound reporter; the link is the document that was opened.

• Emerson v. State, 880 S.W.2d 759 (Tex. Crim. App. 1994), Texas Court of Criminal Appeals, decided 1994-04-13 — full text as read on September 21, 2026
• State v. Superior Court, 149 Ariz. 269, 718 P.2d 171 (1986), Arizona Supreme Court, decided 1986-04-07 — full text as read on September 21, 2026
• Spencer v. State, 302 Ga. 133, 805 S.E.2d 886 (2017), Supreme Court of Georgia, decided 2017-10-02 — full text as read on September 21, 2026
• People v. Leahy, 8 Cal. 4th 587 (1994), Supreme Court of California, decided 1994-10-27 — full text as read on September 21, 2026
• State v. Meador, 674 So. 2d 826 (Fla. 4th DCA 1996), Florida District Court of Appeal, decided 1996-05-15 — full text as read on September 21, 2026
• Campbell v. People, 443 P.3d 72 (Colo. 2019), Colorado Supreme Court, decided 2019-06-24 — full text as read on September 21, 2026
• State v. Sullivan, 310 S.C. 311, 426 S.E.2d 766 (1993), Supreme Court of South Carolina, decided 1993-01-11 — full text as read on October 27, 2026
• State v. Baity, 140 Wash. 2d 1 (2000), Washington Supreme Court, decided 2000-02-03 — full text as read on October 27, 2026
• State v. Klawitter, 518 N.W.2d 577 (Minn. 1994), Minnesota Supreme Court, decided 1994-06-30 — full text as read on October 27, 2026
• State v. Mecham, 186 Wash. 2d 128 (2016), Washington Supreme Court, decided 2016-06-16 — full text as read on October 27, 2026

All posts · Guides · State laws