Over the first several months of 2014, the Hillsborough County Sheriff’s Office arrested 15 people on drug charges only to have the Florida state crime lab determine the substances thought to have been drugs in fact were not.

The arrests had been made by deputies on Florida’s Gulf coast using what are known as chemical field tests — inexpensive kits designed to detect the presence of illegal drugs. A suspected substance is dropped into a pouch of chemicals. The liquid turns a certain color if the substance is, say, methamphetamine or cocaine.

Amid the rash of mistakes, Christopher Baumann, a lieutenant who’d worked narcotics and homicide at the agency, decided to conduct an experiment. He placed material listed as methamphetamine into a kit specifically meant to indicate methamphetamine and watched as the liquid in the pouch turned a purple-ish color. He understood that to mean the test was positive for methamphetamine. Then he activated another kit — cracking open ampules of the chemical reagent — but placed nothing inside the pouch. The liquid turned the same color.

Baumann, concerned, notified his superiors of his results in a memo. They took swift action: Officers in the department were ordered to turn in their remaining field test kits. Prosecutors dropped charges against a number of people arrested for possession of methamphetamine, according to the local public defender’s office.

But sheriff’s officials subsequently made an embarrassing discovery. It was Baumann who had erred, not the tests. He’d misinterpreted the color changes. The department had to notify the local prosecutor again, though little could be done about the methamphetamine arrests that might have been dismissed prematurely.

Beyond scuttling a handful of cases, Hillsborough County’s confusion reflects a broader issue with field tests, which are used by thousands of law enforcement agencies across the country to bring drug charges and obtain guilty pleas from those charged.

Many officers appear to receive little or no formal training in the proper use and interpretation of these tests and there are no requirements for them to do so.

It’s hard to say exactly what proportion of police agencies have formal training regimens for field tests. Some do: A spokesman for the Phoenix Police Department said its controlled-substance officers were the only ones allowed to administer field tests. They have to take a two-day class to get certified and another four hours of training annually to be recertified. To be certified or recertified, they have to pass a practical exam with a grade of 100 percent.

But a 2011 federal survey of 10 jurisdictions that used field tests found just two provided any kind of formal training. Most, like Hillsborough County, seemingly leave officers to figure out the science on their own.

ProPublica tried repeatedly to interview Hillsborough County Sheriff David Gee about his department’s use of the kits and what training is provided to officers. Gee declined, as did all other officials in his department. In response to a request for any information or material pertaining to field test training made more than a year ago, Sissy Shepherd, the department’s training specialist, wrote, “I have researched and found that the Training Division does not have any material pertaining to Field Drug Test Kits.” Subsequent inquiries concerning any updates to training requirements went unanswered.

No government agency tracks how commonly police use field tests, but what little research exists suggests they are ubiquitous. Officers arrest roughly 1.2 million people a year nationwide on low-level drug crimes, many involving evidence derived from the chemical kits.

Much of that evidence is never confirmed by a certified lab. More than 90 percent of drug convictions in the U.S. come by guilty plea, typically at preliminary hearings. A ProPublica analysis of arrest and criminal court data indicates at least 100,000 convictions a year rely on field-test results to identify unknown powders and crystals as illegal drugs.

In Hillsborough County alone, there were almost 14,000 drug convictions between 2010 and 2014, nearly 99 percent of them via plea deals.

Some material initially suspected of being a drug such as methamphetamine does make it to the state’s crime lab for confirmation. State records show that in 2014, when the Hillsborough sheriff’s office submitted material described as methamphetamine to the state lab, one in four submissions were not meth. It is unclear what exact percentage of the material submitted involved field tests. The statewide data on such lab confirmation submissions for suspected meth, however, are similar. Of all material suspected of being meth in 2015, 21 percent proved not to be.

Jon Shane, a retired captain with the Newark police and now a professor at John Jay College of Criminal Justice focusing on police policies and practices, said departments using field tests should have formal training and a written policy to go with it.

“To deploy any kind of equipment like that without training people is completely inappropriate,” Shane said. “You don’t know if the people in the field actually know what they are doing with it.”

---

Field tests were originally conceived and embraced as a quick way for police officers to figure out if suspicious substances might be illegal drugs. Over the past 30 years, the kits also became a convenient answer for crime labs staggering under the burden of ever-greater numbers of drug cases as American law enforcement ramped up its war on drugs. Officers in the streets could do the tests and prosecutors could use them to bring defendants before a judge.

But the science was always prone to error, and the administration of the tests can be tricky. Depending on the test, officers have to follow several steps and interpret a range of changes in color. Some of the color-coded charts provided by manufacturers resemble complicated high school chemistry projects. Often, officers conduct the tests in something less than laboratory conditions — in a police station or on the side of the road.

In Houston, more than 300 people were wrongly convicted in recent years based largely on problems with field tests, including the inability of officers to interpret them accurately. Charles McClelland, who served as Houston’s police chief until earlier this year, said he has come to think officers shouldn’t be using the tests at all.

“Police officers aren’t chemists,” McClelland said. “We shouldn’t be doing field tests on the hood of patrol cars.”

Field test results remain inadmissible at trial in nearly every jurisdiction; instead, prosecutors must present lab tests that use more reliable methods. However, most drug convictions come by plea deal shortly after prosecutors file charges, with field tests alone identifying the evidence as a drug.

Training is not a cure-all for these risks, experts agree, a truth that has played out in Las Vegas.

The Las Vegas Metropolitan Police Department requires its officers to attend a field test certification class before performing tests to make drug arrests. Department policies say officers learn to weigh the suspected drugs, ensure the area where the tests are being done is well-lit and clean, and then fill out a form to document each test. Officers must go through the training again if the crime lab finds they have in the course of any 12 months made paperwork mistakes or had two instances in which tests purportedly indicating the presence of illegal drugs later proved wrong.

Yet, even with this level of instruction, mistakes happen, the Las Vegas department’s own officials concede. Officers misread the color changes in the pouches. Mishandling can lead to contamination. In 2014, the Las Vegas crime lab officials lost enough faith in the reliability of the tests that they formally advocated abandoning them for use in making arrests for certain kinds of drugs, including methamphetamine.

Still, experts said training can help standardize best practices and demonstrates that some degree of care is being taken by police agencies.

“The field test should be governed by policy,” Shane said. “And the policy should be governed by very thorough research. And the research should include positions from the forensics lab and the company who developed the test in addition to contemporary guidelines on field testing narcotics.”

---

In the year after the episode involving Baumann, the Hillsborough County Sheriff’s Office did not change its approach to training. Instead, the department switched to a different kit for testing possible methamphetamine made by Sirchie, the manufacturer of its old kits.

North Carolina-based Sirchie sells an array of evidence collection products to police agencies and has sold chemical field tests under various names for nearly four decades.

Sirchie suggested Hillsborough switch methamphetamine tests after Baumann’s error, department emails show. “Thank you for your understanding in this issue,” Jack Thorndike, Sirchie’s training representative, wrote to agency officials in August of 2014. “I believe this exchange will service your needs very well.”

The alternate kits contain Marquis reagent, a mix of sulfuric acid and formaldehyde that produces an array of colors when mixed with more than 100 different compounds. Methamphetamine is supposed to produce orange that shifts to brown. Sugar can cause the same color change.

Sirchie routinely offers training to law enforcement agencies that buy its tests. ProPublica asked Thorndike to describe the training — whether it was a one-time offer, how many hours it involved, how many officers typically availed themselves of it — but he refused.

ProPublica also contacted half a dozen police agencies whose officers said they’d received training from Sirchie in online biographies and résumés. None of them would describe the instruction they had received.

Detective William Majeskie with the New Berlin Police Department outside Milwaukee lists himself as a “Master Instructor” of NARK II, a Sirchie brand, on LinkedIn. When contacted, Majeskie directed us to talk with Thorndike. Told that Thorndike had declined to detail the training Sirchie provides to police agencies, Majeksie said he would not discuss it either.

Sirchie spells out some of what might be drilled in through training in its 23-page “operator’s manual,” which can be found online. It instructs officers how much of the suspicious substance to place inside the pouches, what colors should develop for the test to be considered positive, and safety precautions for handling acids. Sirchie’s manual mentions inaccurate results only in passing, without explaining how they happen. “There is no drug identification system presently in use which completely eliminates the occurrence of false positives and false negatives,” the manual states.

Another Hillsborough County case makes clear that officers may need more guidance than is provided by these instructions alone.

On Jan. 3, 2015, sheriff’s deputies said they pulled over Gil Ramirez Martinez for having windows that were tinted too darkly. Martinez acknowledged to the deputies that he was driving without a license. The officers handcuffed Martinez, put him in the back of a squad car, and searched his car.

Martinez, who doesn’t speak much English, raises chickens in the Tampa area to support his family. Some of them had fallen ill and he was on his way home with a vial of chicken medicine —­ penicillin — and hypodermic needles to administer it when flashing lights appeared in his rear view mirror. When the deputies conducted their search, they found syringes containing clear liquid in the car’s trunk.

Best practices established by the National Forensic Science Technology Center explicitly say police should not field test liquids. Yet Sirchie and other leading kit manufacturers instruct officers on ways to test liquids in their pouches. And in Hillsborough, that’s exactly what officers did. Arrest records show they tested the clear liquid found in Martinez’s trunk using a NARK II kit, determined it was methamphetamine, and took Martinez to jail.

Martinez said he was never shown the field test and didn’t understand why he was being arrested for possession of antibiotics.

“I was just worried about why I was being pulled over,” Martinez said through his wife Danila Ramirez Martinez, who translated for ProPublica. “Was it a crime having chicken medicine?”

Martinez spent seven hours in jail charged with possession of a controlled substance. His family had to find bail money. He had to hire a lawyer. He had to fight the charges.

A month after Martinez’s arrest Florida’s state lab determined the clear liquid was not meth.

It was chicken medicine, said Amy Ruiz, Martinez’s lawyer.