Three Days in a Classroom: the HGN Test, Part 3 of 5

Day three, and there's a paper on the desk in front of me.

Twenty five questions, four points each, and it passes at eighty percent. That's the written exam at the end of the practitioner course, the one that qualifies an officer to administer HGN and the rest of the standardized battery. I paid to sit for it like everybody else in that room.

They're fair questions. Eight of the twenty five are about HGN, and they ask how many inches the pen goes from the nose, how many seconds the eye gets held out at the corner, and how many clues are possible in each eye.

I went down that list twice, looking for one particular question.

Last time I gave you a question to ask at your suppression hearing. Ask the officer what the eighty eight percent was measured against. And I told you that many of them can't answer it, and that the reason isn't that they're hiding it from you.

So here's the reason. That question isn't on the exam. Nobody in that room ever gets asked it, so nobody in that room ever has to know it.

I'm Rory Safir, and this is Reasonably Safir. Part three of five on horizontal gaze nystagmus.

Here's the plan for the next twenty minutes, so you always know where you are. First, the room itself. Three days, and what those days are actually spent on. Then the two exams at the end of it, and the one thing neither of them ever measures. Then four sentences that used to sit in these manuals and don't anymore, and a whole section that disappeared with them. And at the end, a stopwatch, and the question I'd bring to your next hearing.

I paid to sit the practitioner course, the one that teaches an officer how to administer these exercises. Then I went back for the instructor course, which is a different course with a different job: it trains you to train officers. And I've sat ARIDE on top of both, which goes back through the original battery and the older tests besides. That's the only reason I can tell you what's in the room. What's in there is not what you'd expect. Nobody should mistake it for a weak curriculum. It's careful, and it has one thing missing from the middle of it.

You can't see the result. You can see everything else. This is the everything else that nobody ever looks at.

So let me take you in there, starting with the size of it.

The whole practitioner course is three days. About twenty four hours. That is the entire foundation of the expertise the officer is going to claim in front of your jury.

Inside those three days there are supposed to be at least two alcohol workshops. Volunteers get dosed and breath tested to a measured level, and the students run the battery on people who have actually been drinking. That's the only part of the course where anybody finds out whether they can do this at all.

Mine were live. Both of them, and I've done live workshops several times since. Handheld breath testers, and a room full of volunteers who had been drinking for real, different ages and different tolerances in the same room.

And this is the thing I wish every juror could see one time. Some of them, you could tell immediately. And some of them you would have had absolutely no idea about. People blowing numbers that would end their driving privilege, sitting there looking like anybody else in the room.

The mood in there is light, by the way. It's a funny afternoon, which is its own kind of strange when you think about what everybody in that room is learning to do.

Now here's the first thing I want you to carry out of this. Those workshops can be replaced with video. The standards allow substituting approved video of drinking subjects for the first workshop, or for both of them, and the current instructor guide has a session for exactly that. It's titled, in the guide, Dry Lab.

Which means that under the program's own rules, an officer can be qualified on HGN having never once run it on an intoxicated human being.

Now the middle of the curriculum.

Part one, I gave you the motion. The eye drifts off the target and gets snapped back out to it, and the jerk is what the officer is scoring. Here's what nobody teaches the officer.

Your brain doesn't hold your eye off to the side with a muscle that just stays clenched. It holds it out there with a running calculation, and that calculation leaks. The cerebellum, the part at the back of your head that handles coordination, is what plugs the leak. When the plug gets weaker, the eye slides back toward the middle and the nerves snap it out again. That's the jerk.

And this is real, and I want to be clear that I'm not disputing it. Researchers dosed healthy volunteers to point zero six and to point one zero, measured how their eyes held a target off to the side, and watched the drift get worse. Alcohol makes the cerebellum worse at plugging that leak. The eye sign is genuine physiology, and I'd say so on the stand.

So I went looking for that explanation in the officer's own book.

Those manuals aren't secret. NHTSA publishes them, and I've been through every edition back to the year two thousand. The word cerebellum does not appear. Not once, and not in any edition. The officer's manual says alcohol causes this. It never says how.

It is written down, though. It's in the prosecutor's reference, a National Traffic Law Center guide on this exact test, called The Science and the Law, and it names the eye movement control centers in the cerebellum and the brainstem, out loud, in a sentence.

So the explanation exists, in the government's own reference shelf, and the officer giving the test was never handed it.

Which brings me to the exams. First, how hard this test is to give, because it bears on what comes next.

This test is very hard to do correctly, and the difficulty is physical rather than intellectual. You hold the stimulus in the right place and keep it there. You remember every pass and how many of each. You say a specific set of words, in a specific order, out loud, from memory. And you actually see the thing, which is genuinely difficult, because a small involuntary jerk of somebody's eye is not an obvious event.

I was in a warm room with good light, with an instructor standing next to me, with nobody to arrest, and I found it hard. Now picture doing it for real. Roadside, at one in the morning, with traffic going past, and on somebody who might run.

There are two exams at the end of that course, and the gap between them tells you a lot.

One is the written test I opened with.

The other is a proficiency demonstration. You stand up in front of an instructor and run the whole battery on somebody while the instructor marks off the steps. And check the edition, because that sheet changed. On the current twenty twenty three form there are twenty nine steps, ten of them on HGN, and the candidate has to explain and complete every one. On the two thousand four, two thousand six, and twenty thirteen forms there are twenty six, and only some are marked as the ones that can't be skipped.

So the program that watched the candidate move a pen correctly in front of a witness was content to qualify the candidate on the book while they missed one question in five. And it never tells you which five they missed.

And here's the thing it took the instructor course to see.

The proficiency exam has no accuracy standard in it, anywhere.

Look at what it scores. Was the pen twelve to fifteen inches from the nose. Was it slightly above eye level. Was there a check for equal pupils and resting nystagmus. A check for equal tracking. Was the pass about two seconds. Was the eye held out at maximum deviation a minimum of four seconds, with no white showing. Did the pen take about four seconds moving out to the angle, and did each eye get checked twice.

Every one of those is procedure. Not one of them asks whether the officer was right.

And I have to be fair, because accuracy does get compared somewhere. In the dry lab the recording sheet has two columns side by side, one for what the student observed and one for what was actually there, and the log officers are told to keep pairs their own call against the driver's measured breath result. So I won't tell you nobody ever mentions it. My point is narrower and it still stands. Nothing that's graded, and nothing that decides whether an officer walks out of that room qualified, turns on whether their reading of a pair of eyes matched anything.

The grade is for the hands.

There's one more thing on that sheet worth going and getting.

Pull the version of the checklist your officer was scored on. On the two thousand four, two thousand six, and twenty thirteen forms, some steps are marked as the ones that can't be skipped and some aren't, and three of the ones that aren't are the pre-checks: equal pupil size, resting nystagmus, and equal tracking. In the twenty fifteen edition they became mandatory, and the candidate has to complete each of the steps listed. Part one, I spent a while on those pre-checks, and on the manual's own word for one of them, which is a medical assessment.

Now be careful with that, because it's easy to oversell, and the sheet cuts both ways. This is a fact about how the form was marked up. It isn't proof that your officer skipped anything, and they may well have done all three. What it tells you is that for about a decade the program didn't make them mandatory in order to pass. So the useful question here is about the calendar. Ask what year the officer trained.

And that takes me back to the written exam. I told you which question isn't on it. Here's the rest of what isn't on it. It never asks what a false positive rate is. It never asks what those accuracy figures were computed against. And it never asks what a correct decision meant in the research that produced them.

There's one more absence I find harder to explain. Eight of those twenty five questions are about HGN, and not one of them asks how long the hold at the corner has to be. Four seconds is the clue. It isn't on the test.

I want to say this next part clearly, because it changes the tone of the cross. None of that is being hidden from you. The examination never raised it. The officer was taught a number, was tested on the number, and nobody in that room ever said what the number was made of.

Which is why I don't cross these officers as liars. I cross them as people handed a tool and never told what was wrong with it. Juries can smell the difference, and the second one is more damaging anyway, because the officer will agree with you.

There's a section in the participant manual about how to testify, called Trial Tips and Techniques, and two of its instructions are worth putting to the witness. It tells the officer to make eye contact with whoever they're trying to convince and to maintain it even if it feels unnatural. And it tells them to give definitive, positive, and sure answers, and never to preface a response with I think or I believe.

I'll say the other half myself, because a good prosecutor will say it on redirect and it's true. That same list opens with tell the truth, and the definitive answers item starts by telling them not to guess and says in terms that it's okay to say I don't know, or I can't remember. So that half is fair. This isn't a manual teaching anyone to bluff. What it teaches is how to sound certain, and it sits next to a curriculum that never showed them the reasons to be uncertain.

And one more thing about the training, and it's the short one. There's no required refresher. The standards recommend one every three years or so, and then leave the methods and the frequency up to the individual states. The certificate doesn't expire on its face.

And Florida is one of those states, so it's worth knowing what Florida did with that recommendation. It didn't take it. The retraining Florida actually requires of a certified officer is forty hours every four years, and the subjects are named: human diversity, use of force, firearms qualification, domestic violence, juvenile sexual offender investigations, discriminatory profiling and professional traffic stops, misuse of electronic databases, and officer health and wellness.

Field sobriety testing isn't on that list.

Neither is impaired driving detection, and neither is horizontal gaze nystagmus.

I want to be careful with that, because an officer's own agency can require more, and some do. So the sentence that holds up is that Florida doesn't require it, and whether the department did is a question for the training file.

So the officer testifying about the science of this test may be working off twenty four hours in a classroom a decade ago, with nothing since, and no obligation to have done anything since.

Just ask the officer for the date.

Now I want to show you four sentences that used to sit in these manuals, and where each of them went. They're the reason this part exists.

And before I read you the list, here's what it's worth. Every one of these is a fact the program used to hand officers in print and doesn't anymore. Which means every one of them is a question the officer will be answering cold, in front of your jury, with no page to stand on.

Through the August two thousand six manual, the definition of clue two read like this. People exhibit slight jerking of the eye at maximum deviation, even when unimpaired, but this will not be evident or sustained for more than a few seconds.

Even when unimpaired.

By the twenty thirteen edition that clause is gone, and it hasn't come back. The admission that ordinary sober eyes jerk at the corner, which is the entire reason the four second hold exists, hasn't appeared in a student manual since two thousand six.

That's the first sentence.

The second. That same August two thousand six manual carried a warning printed in full capital letters. If any one of the standardized field sobriety test elements is changed, the validity is compromised.

In the twenty thirteen manual that warning is absent.

In twenty fifteen it comes back, in lower case, and the sentence now reads that the validity may be compromised. That's where it still sits today, as one sentence.

Is became may be.

And flag the trap before you walk into it. Do not stand up and tell a jury the warning was deleted, because a prepared officer will know the current edition cold, and a good prosecutor may have the manual on the table. Either way it comes out on redirect and you get asked to read the sentence aloud. It's right there.

What actually happened cuts deeper anyway.

If your officer trained between May of twenty thirteen and October of twenty fifteen, they trained on a manual that handed them the percentages and not the caveat. And that gap belongs to the program before it belongs to the officer.

The third. The Colorado validation study, which the manual quotes for an eighty six percent figure. In the twenty thirteen, twenty fifteen, and twenty eighteen editions, that sentence said correct arrest and release decisions at a point zero five BAC were eighty six percent accurate.

In the twenty twenty three edition, the words at a point zero five BAC are gone.

Same figure, with the threshold it was measured against no longer printed. So an officer trained on the current manual might quote your jury eighty six percent, or might not remember the number at all. Either way the manual never handed them the threshold, and the threshold sat well below Florida's. If the number matters in your case, the impeachment is the page itself.

And be careful with that one, because the study has an answer, and a good prosecutor will have read it. Colorado's own authors wrote that if their state had had an point zero eight statute, the differences would be relatively minor. So the use of this is narrow. The officer can quote the figure and can't tell you what it was measured against, which is the same hole you found in part two.

The fourth is from the drug side. There's an eye check called lack of convergence, which is just crossing your eyes. In the twenty thirteen drug recognition manual, in the same paragraph that taught it, the manual called it the least reliable of any of the eye tests, because a significant portion of the population may have an inability to cross their eyes.

By twenty fifteen it may not be as reliable as the other eye tests, because some people may have that inability.

And by the twenty twenty three edition, the caution is gone entirely.

Those are the four sentences. Now here's the one that isn't a sentence.

Open the two thousand six manual to the contents page for the session that teaches these three tests, and read down the list of segments. The validation studies. Horizontal gaze nystagmus. Walk and turn. One leg stand. And down at segment H, sitting between One Leg Stand and Taking Field Notes: Limitations of the Three Tests.

A named segment, sitting in the curriculum, whose whole job the instructor's text spells out. In it you explain the circumstances under which each of these tests might not be reliable indicators of alcohol impairment.

It's in the two thousand two manual. It's in two thousand four. It's in both two thousand six editions. In twenty thirteen it's gone, and it has not come back in twenty fifteen, in twenty eighteen, or in twenty twenty three.

The other four were sentences. This was a section.

Now go to the learning objectives at the front of that same session in the current manual, the list of things the program says an officer must be able to do on walking out of the room. One of them is identify the limitations of the three tests.

That objective is still there. The segment that taught it has been gone for over a decade.

So the program still requires officers to know the limits. It just stopped telling them what the limits are.

Nothing about human eyes changed between twenty thirteen and twenty twenty three. What changed is what the program was willing to put in print. And what that means for your case is narrower and more useful than any theory about why. What a student got handed depended on the year they happened to sit in that room.

And that walks me to the idea this whole part has been heading for.

Standardized and validated are two different words.

Standardized means the procedure and the scoring have been fixed, so the same steps get done the same way in Pinellas and in Pensacola. That's procedural uniformity, and it's genuinely useful.

Validated is a separate question. It asks whether the thing measures what it says it measures.

A procedure can be perfectly standardized and still measure nothing.

And on the road side, somebody went and measured how this test actually gets administered once the classroom is over.

A researcher named Booker sat down with fifty two arrest tapes from car mounted cameras, covering forty nine different officers, all of them cases where the officer testified the driver was impaired. He put a stopwatch on the HGN test in every one.

The times ran from nine seconds to eighty three. The average was twenty one seconds.

Now hold that twenty one in your head, because the question is twenty one seconds against what.

Run this test at the fastest pace the manual allows, at every single step, and the movement alone takes forty eight seconds. The two second passes, the four second holds, the four second sweeps out to the angle, each eye, each one twice. Forty eight is not a good test. Forty eight is the floor, and it doesn't include the instructions, or taking the glasses off, or checking equal pupil size, resting nystagmus, and equal tracking, which are the three checks that are supposed to catch a head injury or a neurological problem before anything the officer sees gets counted.

The average officer on those tapes spent twenty one.

In only eight of the fifty two tests did the officer stop the pen even once for the four seconds the manual requires. Fifteen percent. And that study is from two thousand one, and it's one study. Here is its last number.

One officer out of fifty two ran the test the way the manual teaches it.

So that's the room, and that's what walks out of it. An officer taught carefully, tested on procedure, graded on the hands, and never once asked the question you're about to ask.

Here's the question, and it isn't a question at all. It's a stopwatch. Pull the video of your stop and put a clock on the HGN test, from the first pass to the last hold. Run at the fastest pace the manual allows, the movement alone takes forty eight seconds. If the tape shows twenty, the test on that video is not the test in the manual, and you can prove it with the officer's own training materials and a second hand.

Next time we go to the law. And I'm going to tell you that the case everybody in this state cites for letting this evidence in is actually the case for keeping it out.

Every manual edition and every study from this episode is free. NHTSA publishes the manuals, and if you'd rather skip the hunting, email me at Rory at the Safir lawyer dot com and I'll send you all of it.

I'm Rory Safir. Take care of yourself, and take care of your clients.