Of course, most officers (even the instructors who teach the FST's to other officers) don't administer them properly, and don't really bother to score them at all, but instead insist that the tests are merely for the officer's benefit in forming an opinion as to whether the driver was under the influence.  In fact the tests are to be scored according to a standardized procedure and the driver either passes or fails the test based on the number of "clues" observed by the officer.  The officer is also supposed to write down the scoring, indicating each clue, right as the test is being performed... but they don't do that either.

The Standardized Tests 

The three standardized tests are the Horizontal Gaze Nystagmus (HGN) and the Walk & Turn.

The HGN test is sometimes called the "follow the finger test" because that's pretty much what the officer has you do as he slowly passes his finger or a pen or some other object back and forth about 12-15 inches in front of your eyes.  Many people think they do well on this without knowing what is being tested.  The officer is looking for involuntary jerking of the eye, something you never even notice as your eye begins to twitch.  Based on where and how much the eye twitches, the officers are supposedly able to determine with about a 70% accuracy rate whether your blood alcohol level is above or below .10.  Fortunately Tennessee courts have ruled that this evidence is inadmissible unless the prosecution brings in an expert witness to testify as to what the results actually mean.  Since the prosecution almost never does this, the results of the HGN test don't matter much in Tennessee.

The other two tests supposedly involve both testing your ability to follow directions and also to perform the tests.  The names of the Walk & Turn test and the One-Leg Stand test are both rather descriptive of what the officers requires.  The Walk & Turn test involves walking nine steps heel-to-toe (with the heel of the front foot touching the heel to the back foot each time) in a straight line without raising the arms for balance and turning in a sort of pivot-step on reaching the ninth step and then walking back, all the time counting each step out loud.  The One-Leg Stand involves standing on one leg, with your hands to your side and not raising them for balance, with the foot in the air raised about 6-12 inches off the ground right in front of the foot you are standing on, and slowly counting "One-Thousand, Two-Thousand, Three-Thousand" and so on to thirty.  Starting too soon, losing your balance, starting over or failing to understand or follow the directions all result in points added to your score even when it is scored properly.

A great many sober people can never pass the tests.  Women wearing high heels, anyone with balance problems, leg or back injuries, more than 100 pounds overweight or more than 60 years old are not even supposed to take the tests because the results are not valid.

"Reliability" of the FST's 

But there is also research indicating that the "reliability" of the FST's, even when properly administered is much lower than the police want us to believe.  The police claim that when all three of the FST's are used, the controlled tests show they are accurate 81% of the time.  This is bullshit.  Let's look at the actually study leading to the finding of this high accuracy rate.

In a controlled setting, inside, in a well lighted room, with healthy and fit "drivers" performing the tests, with no wind blowing, no one exhausted after a long drive, and no fear or anxiety such as drivers feel when pulled over at midnight, officers who had just been instructed on how to administer and score the tests were given several other officers to perform the tests on.  While one group of officers was administering the tests, another group of officers were posing as "drivers" who had just been pulled over and they were taking the tests.

Each of the officers posing as drivers and being tested had been given either no alcohol to drink or a controlled dose of alcohol which according to the officer's body weight was to be enough to bring the officer's blood alcohol level to either .05 (half way to being under the influence), .10 (under the influence) or .15 ( which most charts will tell you is the point at which there is obvious impairment, with speech being slurred and clear balance problems).  The officers being tested were equally divided into the four groups, meaning there were the same number of officers at each level.

For the sake of illustration let's say there were 25 who had nothing to drink, 25 at .05, 25 at .10, and 25 at .15.

Now all of those who had nothing to drink should have been "gimmes", anyone should have been able to get them right, since they would have had no odor of alcohol.  All of those at the .15 level also should have been gimmes because of the obvious level of impairment.  Of the remaining 50, let's say you decided whether you thought the person was impaired, with a blood alcohol level of .10 or above simply by the toss of a coin -- you would get 25 right and 25 wrong.  This would give you a total of 75 right, meaning you were right 75% of the time.... and the FST's, in the best of settings are only supposedly right 81% of the time.... a whopping 6% better than nothing.

Non-Standardized FST's 

There are also a number of other commonly used non-standardized FST's, including the following: Finger-to-Nose; Coin Pick-Up; Recite A-B-C's (generally the request is that you recite them backwards, or from one point in the alphabet to another backward); Counting (generally counting backwards, from perhaps 87 to 41); Finger Count (touching the tips of the fingers to the tip of the thumb, and usually doubling on the little finger and the index finger); Palm Pat; Tracing Mazes; Romberg Body Sway; Letter Cancellation (reading a passage and counting the number of times a particular letter appears in the passage); Math Skills (adding & subtracting).

There's a great Steve Martin movie from a few years back in which Martin's character is pulled over for DUI and the officer has him do a list of progressively more difficult FST's, each of which the driver does with ease, until the officer finally has him juggle flaming knives and read a German newspaper.  While I have never heard of an officer getting that bizarre, the bottom line is that they want you to fail the tests... and they are the only ones scoring the tests... and they will sometimes keep giving tests until they feel you have failed one.  This is why I tell people not to take the Field Sobriety Tests.  If the officer is wanting to arrest you, he's going to do so regardless how you do on the FST's, and once charged the officer will use the FST results to say you were under the influence.  By refusing to take the FST's you don't give the officer the opportunity to offer your performance as evidence you were drunk.

There are a number of other things that can indicate that a person is not impaired, including all of the following: Not Weaving, Lights Turned On, Driving At Proper Speed, Stopping At Light, Proper Turning, Responding Promptly to Lights/Siren, Pulling Over Appropriately, Understanding Officer's Directions, Ability To Respond Intelligently, Ability to Follow Directions, Not Fumbling for Wallet, Ability to Open Door, Not Leaning On Car Door, Not Stumbling While Walking, Polite, Cooperative, Respectful, Understanding Miranda warning (right to remain silent), Knowing and Voluntary and Intelligent Waiver of the Miranda Warning, Remembering Phone Numbers, Good Clean Signature, this list could be endless.  Police officers and prosecutors often like to ignore these indications a driver is not impaired, but they are real indications nonetheless.

Bottom line -- don't do the Field Sobriety Tests, not any of them... but do remain polite and courteous in your refusal.