The Volokh Conspiracy

Mostly law professors | Sometimes contrarian | Often libertarian | Always independent


Automated License Plate Readers, the Mosaic Theory, and the Fourth Amendment

The Massachusetts Supreme Judicial Court weighs in.


One of the fascinating questions raised by the United States Supreme Court's 2018 decision in Carpenter v. United States is how the Fourth Amendment applies to government use of automated license-plate readers (ALPRs) and the querying of ALPR databases.  Last week, the Massachusetts Supreme Judicial Court handed down what I believe is the first appellate decision on the question, Commonwealth v. McCarthyMcCarthy is likely the first of many appellate decisions on this question, so I thought it deserves a close look.

Applying the controversial mosaic theory of the Fourth Amendment, McCarthy ruled that use of ALPRs can violate the Fourth Amendment, but that it depends on how much the ALPR query happens to reveal.  Because the surveillance in this particular case was relatively limited, it did not tell the police enough to create a mosaic and was therefore not a Fourth Amendment search.

McCarthy is a fascinating case, although I think it's ultimately a good example of why the mosaic theory is unworkable and should be rejected (as I have argued here and here).  I'll start with the facts, turn to the majority's reasoning, explain Chief Justice Gants's concurrence, and then offer some thoughts of my own.

I.  The Facts

This is a drug case.  Investigators suspected McCarthy of distributing heroin from his home to a co-conspirator's home on Cape Cod.  There are two bridges on to Cape Cod, the Sagamore bridge and the Bourne bridge, and the state has ALPRs set up on those two bridges.  To see if their suspicions about McCarthy's travels were right,  the police set up an alert on the state ALPR system to tell the police when McCarthy's car crossed one of the two bridges.  The alert they set up operated in "real time," but only for the cameras fixed on those particular bridges.

For two and a half months, the police learned the precise dates, times, directions, and specific lanes that McCarthy's car traveled on the two bridges to or from Cape Cod.  The police then used a later ALPR hit to know when McCarthy's car had crossed one of the bridges and was likely meeting with his co-conspirator.  Upon seeing that meeting, the police made their arrest.

The legal issue raised in the case was whether the government violated McCarthy's constitutional rights by conducting this monitoring of his car's location.

II.  The Reasoning

The Supreme Judicial Court ruled that use of ALPRs can violate the Fourth Amendment, but that they didn't in this particular case.

The court began by endorsing the mosaic theory of the Fourth Amendment.  The SJC had previously suggested a mosaic theory was viable under its state constitution.  But here the court explicitly endorsed the theory for both the state constitution and (more importantly) the federal Fourth Amendment.

Looking at its state constitutional cases together with several federal constitutional cases, the court reasoned that

these cases articulate an aggregation principle for the technological surveillance of public conduct, sometimes referred to as the mosaic theory. When collected for a long enough period, the cumulative nature of the information collected implicates a privacy interest on the part of the individual who is the target of the tracking. . . . As the analogy goes, the color of a single stone depicts little, but by stepping back one can see a complete mosaic.

A detailed account of a person's movements, drawn from electronic surveillance, encroaches upon a person's reasonable expectation of privacy because the whole reveals far more than the sum of the parts. The difference is not one of degree but of kind.  Prolonged surveillance reveals types of information not revealed by short-term surveillance, such as what a person does repeatedly, what he does not do, and what he does ensemble. Aggregated location data reveals a highly detailed profile, not simply of where we go, but by easy inference, of our associations—political, religious, amicable and amorous, to name only a few—and of the pattern of our professional and avocational pursuits.

(internal quotations and citations omitted, both here and in the passages below)

The court next concluded that use of ALPRs is a Fourth Amendment search if enough ALPRs are used in particular ways:

In determining whether a reasonable expectation of privacy has been invaded, it is not the amount of data that the Commonwealth seeks to admit in evidence that counts, but, rather, the amount of data that the government collects or to which it gains access. . . . For this reason, our constitutional analysis ideally would consider every ALPR record of a defendant's vehicle that had been stored and collected by the government up to the time of the defendant's arrest. That information, however, is not in the record before us.

With enough cameras in enough locations, the historic location data from an ALPR system in Massachusetts would invade a reasonable expectation of privacy and would constitute a search for constitutional purposes. The one-year retention period indicated in the EOPSS retention policy certainly is long enough to warrant constitutional protection. Like CSLI data, ALPRs allow the police to reconstruct people's past movements without knowing in advance who police are looking for, thus granting police access to a category of information otherwise [and previously] unknowable. Like both CSLI and GPS data, ALPRs circumvent traditional constraints on police surveillance power by being cheap (relative to human surveillance) and surreptitious.

So what's the test for when use of ALPRs is sufficient to trigger a Fourth Amendment search?

As I understand the court's opinion, the test is applied ex post.  After the surveillance has occurred, you look back and consider if the surveillance ended up revealing enough about a person's life to have been a search.  This should be evaluated based on the totality of the picture that emerged based on all of the circumstances:

[T]he constitutional question is not merely an exercise in counting cameras; the analysis should focus, ultimately, on the extent to which a substantial picture of the defendant's public movements are revealed by the surveillance.

For that purpose, where the ALPRs are placed matters too. ALPRs near constitutionally sensitive locations—the home, a place of worship, etc.—reveal more of an individual's life and associations than does an ALPR trained on an interstate highway. A network of ALPRs that surveils every residential side street paints a much more nuanced and invasive picture of a driver's life and public movements than one limited to major highways that open into innumerable possible destinations. For while no ALPR network is likely to be as detailed in its surveillance as GPS or CSLI data, one well may be able to make many of the same inferences from ALPR data that implicate expressive and associative rights.

So where is the line? Well, the court could not say exactly where the line is:

On this record, however, we need not, and indeed cannot, determine how pervasive a system of ALPRs would have to be to invade a reasonable expectation of privacy. While a testifying expert alluded to cameras "all over the State," the record is silent as to how many of these cameras currently exist, where they are located, and how many of them detected the defendant.

However, the court did conclude that the surveillance here was not enough:

[W]e consider the constitutional import of four cameras placed at two fixed locations on the ends of the Bourne and Sagamore bridges. . . .

There is no real question that the government, without securing a warrant, may use electronic devices to monitor an individual's movements in public to the extent that the same result could be achieved through visual surveillance. It is an entirely ordinary experience to drive past a police officer in a cruiser observing traffic on the side of the road, and, of course, an officer may read or write down a publicly displayed license plate number. In this way, a single license plate reader is similar to traditional surveillance techniques.

On the other hand, four factors distinguish ALPRs from an officer parked on the side of the road: (1) the policy of retaining the information for, at a minimum, one year; (2) the ability to record the license plate number of nearly every passing vehicle; (3) the continuous, twenty-four hour nature of the surveillance; and (4) the fact that the recorded license plate number is linked to the location of the observation. These are enhancements of what reasonably might be expected from the police.

The limited number of cameras and their specific placements, however, also are relevant in determining whether they reveal a mosaic of location information that is sufficiently detailed to invade a reasonable expectation of privacy. The cameras in question here gave police only the ability to determine whether the defendant was passing onto or off of the Cape at a particular moment, and when he had done so previously.

This limited surveillance does not allow the Commonwealth to monitor the whole of the defendant's public movements, or even his progress on a single journey. These particular cameras make this case perhaps more analogous to CSLI, if there were only two cellular telephone towers collecting data. Such a limited picture does not divulge the whole of the defendant's physical movements,  or track enough of his comings and goings so as to reveal the privacies of life.

That brought the court to its conclusion:

While we cannot say precisely how detailed a picture of the defendant's movements must be revealed to invoke constitutional protections, it is not that produced by four cameras at fixed locations on the ends of two bridges. Therefore, we conclude that the limited use of ALPRs in this case does not constitute a search within the meaning of either art. 14 or the Fourth Amendment.

(emphasis added)

III.  Chief Justice Gants's Concurrence

In a concurrence, Chief Justice Gants suggested that perhaps the mosaic theory should encompass multiple tiers of surveillance, with different degrees of surveillance possible with different court order requirements.  If surveillance is sufficient to constitute a search, he wrote, "we have two options":

Our first option is to determine based on the facts of a particular case when the locational mosaic of a targeted individual's movements crosses the threshold of the reasonable expectation of privacy. A mosaic above that threshold would require a search warrant based on probable cause, but a mosaic below that threshold would not require any court authorization.

Alternatively, we could strike a balance analogous to that struck by the United States Supreme Court in Terry v. Ohio, 392 U.S. 1, 21 (1968), and decide that there are two locational mosaic thresholds: a lesser threshold that may be permissibly crossed with a court order supported by an affidavit showing reasonable suspicion and a greater threshold that is permissibly crossed only with a search warrant supported by probable cause.

The reasonable suspicion standard would require "specific and articulable facts" demonstrating reasonable suspicion that the targeted individual has committed, is committing, or will commit a crime, and that there are reasonable grounds to believe that the data obtained from the query are relevant and material to an investigation of the crime.

Chief Justice Gants goes on to explain how this would work.  As I understand the proposal,  the government could apply for court orders for various levels of surveillance, based on its prediction of how much surveillance it will conduct and how much cause was needed for that level.  After the surveillance occurs, the reviewing court would assess whether it feels the government collected only a reasonable-suspicion amount of surveillance or a probable-cause amount of surveillance.  The court would then ask if the affidavit viewed ex post established ex ante enough cause to satisfy the mosaic demands of the surveillance that ended up occurring:

This second alternative would mean that law enforcement agencies would need to obtain court authorization more often before retrieving targeted individual historical locational information in their possession because queries that would not require a showing of probable cause might still require a showing of reasonable suspicion. But the benefit to law enforcement would be that, if the police sought a court order based on reasonable suspicion and a reviewing court determined that the query sought locational data that could yield a mosaic of movement requiring a showing of probable cause, the search would not be found unconstitutional (and the information collected would not be suppressed) if the reviewing court found that the affidavit supported a finding of probable cause.

In contrast, where no court order was obtained and a reviewing court determined that probable cause or reasonable suspicion was required to support the retrieval of historical locational information, the data retrieved from the query would have to be suppressed even if law enforcement could have met the applicable standard.

Under Chief Justice Gants's view, the tiered Fourth Amendment mosaic regime would also imply a government record-keeping requirement:

Regardless of which alternative the court ultimately chooses, a reviewing court will need to know the extent of the mosaic that was possible from the retrieval of historical locational information regarding the movements of a targeted individual, because only then can the court accurately determine whether the threshold had been crossed.

Therefore, unless the law enforcement agency has sought prior court approval to search for particularized locational data in its possession, the agency will have to preserve each and every search query for the retrieval of historical locational information regarding a targeted individual.

For instance, if the State police maintain 1,000 ALPRs at different locations throughout the Commonwealth, it matters whether they searched for a suspect's vehicle from the data yielded by all 1,000 cameras or only by four cameras, and it matters whether they gathered this data for one day or one hundred days. And regardless of whether a court authorized the search, the agency must preserve the historical locational data regarding a particular individual that the agency retrieved as a result of such queries from the data in its possession, even when that exceeds the amount of data that the agency uses in an investigation or at a trial.

Further, this Fourth Amendment theory also implies certain discovery obligations:

And the agency must make this preserved data and search request available in discovery when sought by the defendant. Only then will a court have the information it needs to determine whether the retrieval of locational information regarding a targeted individual crossed a constitutional threshold that requires court authorization and either reasonable suspicion or probable cause.

IV.  My Own Views

This is fascinating majority opinion and a fascinating concurrence.  The court offers a developed and engaged take on how the Fourth Amendment might apply to public surveillance.  I appreciate the Justices' engagement with some really deep questions.  In my view, though, McCarthy ultimately offers a demonstration of why the mosaic theory is unworkable.

The court's approach is to look ex post to see if the government learned enough about a person for the surveillance to need a warrant ex ante.  But the court can't say where the line is.  And that makes it hard both for the government to know what it should do and for reviewing courts to say when the line has been crossed.  Here, four cameras on two ends of two bridges for two and half months wasn't enough. In another case, the court says, a single camera "may well be" enough.  Who knows where the line may be?

This is the Fourth Amendment equivalent of Justice Stewart's comment about then-existing obscenity standards in Jacobellis v. Ohio.  "[P]erhaps I could never succeed in intelligibly" saying when a movie is unprotected obscenity, Justice Stewart famously commented. "But I know it when I see it, and the motion picture involved in this case is not that."

The McCarthy court's reasoning seems similar. Perhaps you can't succeed in intelligibly identifying when a search has occurred, but the court knows it when it sees it.  And surveillance "produced by four cameras at fixed locations on the ends of two bridges" is not that.

This is a particularly challenge approach given the likely need for a warrant if a search is later found.  After the data is collected, the appellate courts will then look at all the data, figure out what it revealed, and then say whether a warrant was needed before the government tried to collect it.

Think of how it would work with a query into an existing database instead of a real-time query like the one in McCarthy.  For example, imagine the government wanted to know where McCarthy's car had been for the last 10 days.  Maybe that query would reveal 10 hits from two cameras.  Or maybe that query would reveal 1,000 hits from 300 cameras.  If you're an officer, you can't know whether your query will reveal a "substantial picture of the defendant's public movements" until you know the results, at which time you can find out if you needed a warrant before you ran the query you just finished.  Or maybe more accurately, you'll then have the information needed to ask a judge whether the judge finds the picture "substantial" and therefore a search.  Odd.

To be clear, I don't have a particular view of how the Fourth Amendment should apply to ALPRs.  Maybe using them should be a search, maybe it shouldn't be.  But as I argued in this book chapter, Implementing Carpenter, I think the answer shouldn't rely on the mosaic theory.  The mosaic theory raises too many unanswerable questions to implement; it would keep courts and investigators guessing for years about what the law requires.

A bright-line answer is needed, I think. Maybe the bright-line rule should be that accessing ALPR data is always a search.  Maybe it should be that accessing ALPR data is never a search.  But I don't think it works to saying that it is sometimes a search, based on the totality of the circumstances depending on how things ended up going.

Other ALPR cases are pending, and it will be fascinating to see if they follow or diverge from the approach in McCarthy.  As always, stay tuned.