The Confrontation Clause of the Sixth Amendment to the U.S. Constitution limits the use of testimonial hearsay statements by an unavailable witness at a criminal trial, as does its state counterpart in Article I, Sec. 23 of the North Carolina Constitution. A hearsay statement is an out-of-court statement offered for its truth. A statement is testimonial if the primary purpose of the statement was to establish past facts for use at a later prosecution. Ohio v. Clark, 576 U.S. 237, 245 (2015). The Confrontation Clause does not protect against the admission of nontestimonial statements (although hearsay statements still must meet an exception or exemption). In a recent decision, the North Carolina Supreme Court analyzed a challenge to the admission of the defendant’s phone records offered by the State at trial. Overruling the Court of Appeals on the point, the North Carolina Supreme Court found that the phone records were nontestimonial as purely machine-generated data.” The case is a good reminder of the distinctions between testimonial and nontestimonial statements and may have implications for future confrontation issues. Read on for the details.
Facts. In State v. Lester, 293PA23-2, ___ N.C. ___; 910 S.E.2d 642 (Jan. 31, 2025), the defendant was charged with various sex crimes with a minor child. During its investigation, the State obtained the defendant’s phone records from Verizon. The phone records were generated automatically as part of Verizon’s recordkeeping process and showed the defendant’s communications for the roughly three-month period relevant to the investigation. This included numbers of calls made and received, the lengths of the calls, and their times and dates. Once the police received the phone records, an officer used a computer software program to isolate the communications between the defendant and the victim. The pared-down version of the phone records showed about 100 calls back and forth between the two and became important evidence at the defendant’s trial. The victim testified, but there was no physical evidence in the case, and the phone records were used to substantiate the victim’s account. The defendant objected to the use of the records, arguing they were testimonial statements and that their admission violated his confrontation rights (the defendant also challenged the records as impermissible hearsay, but the focus of this post is on the constitutional issue). The trial court overruled the objection and admitted the phone records. The jury ultimately convicted the defendant of all charges
The Court of Appeals Decision. The defendant appealed. The Court of Appeals found that the records were testimonial hearsay and should not have been admitted without the defendant having an opportunity to cross-examine the Verizon records custodian. In a sparsely worded opinion, the court seemed to reason that the summary of the phone records prepared by the law enforcement officer was primarily made for use at the defendant’s trial and the summary was therefore testimonial. The court found that this error was not harmless and ordered a new trial. The hearsay issue raised by the defendant on appeal was left unresolved in light of the court’s holding. The State sought and obtained discretionary review of the Court of Appeals decision at the North Carolina Supreme Court.
The North Carolina Supreme Court Decision. In an opinion authored by Justice Earls, a unanimous North Carolina Supreme Court reversed. It held that computer-generated statements like the information found in the Verizon phone records were nontestimonial, because they are created by an automated process, without any human involvement whatsoever. In the words of the court:
[C]omputer generated data represent the self-generated record of a computer’s operations resulting from its programming. This evidence is unique because it is created entirely by a machine, without any help from humans. When triggered, the computer mechanically processes inputs, extracts information, and generates results. The response is encoded in the machine’s programming—it is the product of 1s and 0s rather than independent choice. Id. at 13 (cleaned up).
The court compared the phone records at issue in Lester to other records generated without human input, such as seismograph machines that automatically record seismic activity, or the so-called “black boxes” that automatically record airplane flight activity. These machines operate independently of human involvement, recording activity in real time. “Because computer-generated data are the fruit of a self-sufficient and automated processes, they are the machine’s work alone.” Id. at 14. The records of such an automated process are not testimonial for purposes of the Confrontation Clause, because the machine did not (and could not) intend the records to be used at a later prosecution. Because the Court of Appeals erred in finding the phone records were testimonial, the court reversed and remanded the case back to intermediate appellate court for consideration of the defendant’s other challenges.
Thoughts and Takeaways. The Court of Appeals decision in Lester raised an interesting question: Do nontestimonial statements transform to testimonial ones when they are summarized or condensed for trial? The North Carolina Supreme Court answered that question firmly in the negative. When considering the primary purpose of a statement, a court must consider all the circumstances at the time the statement was made. The fact that Verizon produced the phone records to law enforcement in response to a court order did not transform the otherwise nontestimonial machine-generated statements into testimonial ones. “[T]he Court of Appeals conflated the timing of the records’ production with the timing of their creation.” Id. at 21. While not specifically addressed by the court’s opinion, it seems the same logic would apply to the law enforcement officer’s act of summarizing the records down to the communications relevant to the case—if the statements were nontestimonial at the time of their creation, they do not transform into testimonial statements by virtue of their packaging for trial.
Another interesting aspect of the case is that the North Carolina Supreme Court seems to take an expansive view of what constitutes “purely machine-generated data.” The examples of seismograph machines and airplane “black boxes” seem analogous to the kind of fully automated data produced by a phone company documenting the activity on an account. In its discussion of computer-generated data, though, the opinion compares these types of data to the output produced by gas chromatography machines, commonly used for testing controlled substances. The Lester opinion cites State v. Ortiz-Zape, 367 N.C. 1 (2013), for the proposition that the results of a gas chromatography analysis are likewise raw data. While the opinion notes that a human will need to interpret such data and opine on the meaning of the results (which may be testimonial depending on the circumstances), the data itself is not testimonial. “The key point is that no human judgment contributes to producing this information—the machine simply records and reports what it measures.” Lester Slip op. at 19.
This part of the opinion gave me pause. Unlike seismographs, black boxes, or phone records, a human being is involved in gas chromatography testing. Like most other forms of forensic testing, a person must handle the sample to be analyzed, prepare, and evaluate the sample prior to analysis, input the prepared sample into the machine, ensure that proper protocols are followed (such as calibration of the machine and preparation and testing of quality control samples), and avoid contaminating the sample and instrument, at least. This level of human involvement in forensics testing like drugs and DNA analysis arguably distinguishes it from “purely machine-generated data.” While the comparison to gas chromatography machines is dicta and not necessary to the holding in the case, it may be an indication of how the court will view challenges to substitute analyst testimony, where one witness testifies about the test results generated by another, nontestifying analyst (discussed at length here). If the data produced by a lab machine is always nontestimonial machine-generated data (despite the machine requiring the assistance of a human analyst to function), then it may still be possible for a substitute analyst testify to the lab results of another, absent analyst, despite the defendant never having an opportunity to question the person who performed the testing. If that is correct, then the justification for substitute analyst testimony will simply have shifted from the former “nonhearsay basis of opinion” logic to “nontestimonial machine-generated data” to get to the same place.
Such a result seems questionable under Smith v. Arizona, 602 U.S. 779 (2024), and other state and federal cases categorizing forensic test results created for use at trial as testimonial. See, e.g. Melendez-Diaz v. Massachusetts, 557 U.S. 305 (2009); State v. Craven, 367 N.C. 51 (2013) (so holding). The U.S. Supreme Court rejected a similar argument about gas chromatography results in Bullcoming v. New Mexico, 564 U.S. 647 (2011). While New Mexico categorized the test results in the case as nontestimonial machine-generated data, requiring no human interpretation or exercise of judgment, the Court seemed to reject that approach:
[The analyst] certified that he received Bullcoming’s blood sample intact with the seal unbroken, that he checked to make sure that the forensic report number and the sample number corresponded, and that he performed on Bullcoming’s sample a particular test, adhering to a precise protocol. He further represented . . . that no circumstance or condition affected the integrity of the sample of the validity of the analysis. These representations, relating to past events and human actions not revealed in raw, machine-produced data, are meet for cross-examination. Id. at 660 (cleaned up).
Without such preliminary information about how the lab data was generated, it may be that the substitute analyst’s opinion interpreting the data generated by the testing of another analyst only avoids being testimonial in nature when it is offered as a hypothetical. Indeed, a substitute analyst opining about hypotheticals was expressly contemplated by the Court in Smith. See Smith at 799. Further, it may be that an interpretation of raw test result data by the substitute analyst is subject to challenge under N.C. Evid. R. 702. When the witness cannot establish the preliminary steps needed to ensure that the machine generated an accurate, trustworthy test by reliable application of the scientific methods to the facts of the case, the testimony might not meet the requirements for admission for expert testimony. Assuming Rule 702 does not bar the hypothetical testimony, the scope of the substitute expert’s opinion in this situation could presumably be narrowly circumscribed under the rule to ensure it does not exceed the bounds of a permissible hypothetical. How much weight such hypothetical opinion testimony would carry with a factfinder remains an open question.
As my favorite law professor used to say, it may be that the devil is in the details, and we will have to wait and see how North Carolina courts approach the question of substitute analyst testimony in light of Smith v. Arizona. In the first North Carolina decision to apply Smith, the Court of Appeals seemed to take a bright-line approach to substitute analyst testimony, ordering a new trial where a substitute analyst was allowed to present forensic results identifying a substance as methamphetamine. State v. Clark, ___ N.C. App. ____; 909 S.E.2d 566 (Dec. 3, 2024); temp. stay allowed, ___ N.C. ___; 909 S.E.2d 323 (Dec. 20, 2024) (discussed here). If the state high court decides to weigh in, you can be sure to read about it here. Meanwhile, defenders should continue objecting to substitute analyst testimony as a Confrontation Clause violation, as well as challenging the substitute analyst’s opinion as unreliable expert testimony.
I can be reached as usual at dixon@sog.unc.edu for any questions or feedback.