Rohlfs et al. analyze the performance of familial searching in structured populations. They state: “We find that, using current U.S. forensic loci, a familial searching method can be designed that has power greater than 80% to detect first-degree relatives in forensic DNA databases while yielding only a very small number of false positives (on the order of 10−5 per database entry).” They also caution that misidentification of more distant relatives is more common: “We also find that, while unrelated individuals are rarely misidentified as first-degree relatives, distant relatives (e.g., cousins) have a substantial probability of being misidentified as first-degree relatives using currently proposed methods.”

The study implements the Myers et al. familial identification procedure used in California to estimate power and false positive rate, and to estimate the rates of misidentification of distant relatives as first-degree relatives. It reports: "Our results agree with previous work, showing that with the prescribed methodology, false positive rates of parent-offspring and sibling identification are low, on the order of 10−5 to 10−9." However, it also concludes that "for more distant Y-haplotype sharing relatives (half-siblings, first cousins, half-first cousins or second cousins) there is a substantial probability that the more distant relative will be incorrectly identified as a first-degree relative" and that a first cousin could be misidentified as a full sibling with probabilities that can be substantial depending on population background.

The authors note that "the chance of a false positive—a non-relative achieving this less-stringent partial match threshold—greatly exceeds the probability that the same non-relative is a false exact match." They add that "owing to nontrivial false positive rates, close relatives of database entrants can be exposed to inappropriate forensic investigation when they have not in fact contributed to query profiles." The study finds that using ancestry inference can reduce error: "false positive rates for familial search with use of ancestry inference to specify the allele frequencies are similar to those seen when allele frequencies align with the population of origin" but concludes that even then such reductions "may continue to produce rates that are found to be intolerably high."

This U.S. Department of Justice primer explains: “In general, the more stringent a familial search program is in excluding individuals who are unlikely to be relatives of the alleged perpetrator, the greater the likelihood that the program will miss some potential relatives but the lower the probability that it will incorrectly identify an unrelated individual as a potential relative.” It notes that carefully designed policies can keep error rates low: “Properly implemented familial searching protocols can minimize the risk that an unrelated individual will be identified as a possible relative, but cannot eliminate the possibility that more distant relatives may be returned as potential first-degree relatives.” The document emphasizes that empirical estimates of error “depend heavily on the search parameters, database composition, and follow-up confirmation testing.”

The authors note that when only 15 STR loci are available, "candidate lists inflate, and the ranks of both the UHR and putative relative are unstable, limiting investigative utility," and that this can result in "false positive familial candidate matches." They emphasize that "a high prevalence of homozygous genotypes in an individual's DNA profile can result in false positive familial matches" and recommend using expanded kits with at least 22 STR markers to "reduce the number of candidate matches generated" and maximize discriminatory power. The paper further proposes mitigation protocols (masking tri-allelic loci, dual-pass strategy, replicate testing) explicitly aimed at minimizing erroneous familial hits.

The OSAC (Organization of Scientific Area Committees) standard cites key validation studies and summarizes: “Rohlfs, R.V. et al. ‘Familial Identification: Population Structure and Relationship Distinguishability.’ PLoS Genetics, vol. 8(2), 2012, e1002469” and “Rohlfs, R.V. et al. ‘The Influence of Relatives on the Efficiency and Error Rate of Familial Searching.’ PLoS ONE, vol. 8(8), 2013, e70495.” The standard notes that these studies “provide estimates of the power and false positive rates of familial DNA searching methods under various population genetic scenarios.” OSAC stresses that laboratories should “understand and document the expected false positive and false negative rates associated with their familial search protocols” and consider the increased probability of misclassifying more distant relatives as first-degree relatives when interpreting results.

This PLOS ONE paper (often cited in later work) models familial searching in offender databases. It reports that for correctly specified first-degree relationships, the procedures can achieve high power with "very low false positive rates" for unrelated individuals under certain allele frequency assumptions and database sizes. However, the simulations also show that misidentification risk increases when more distant relatives or population substructure are not properly accounted for; in such cases, distant relatives can be misclassified as first-degree, leading to investigative focus on the wrong nuclear family despite a low probability of completely unrelated individuals being identified.

In an earlier simulation study of familial searching, Rohlfs, Fullerton & Weir (2012) reported that the false positive rate for identifying an unrelated individual as a first-degree relative can be substantial when allele frequencies are misspecified. They showed that misspecification of allele frequencies away from the true population can "substantially inflate false positive rates" and that the increase in false positives grows with the degree of misspecification. The work emphasizes that partial-match familial searching has an inherent risk of misidentifying non-relatives as relatives, and that this risk is sensitive to statistical assumptions rather than being uniformly very low.

In this review, Bieber, Brenner and Lazer discuss empirical and simulated performance of familial searching in offender databases. They state that carefully designed methods can keep false positives low: “Computer simulations using CODIS loci suggest that, with appropriate statistical thresholds, the probability that an unrelated individual in a large database will be misidentified as a first-degree relative of the true source of the crime-scene DNA can be made extremely small.” At the same time, they acknowledge: “There is a substantially greater risk that more distant relatives (e.g., cousins) will be identified by familial searching and then misclassified as first-degree relatives, necessitating careful follow-up investigation and confirmatory testing.” The article frames familial searching as powerful but in need of safeguards and empirical validation.

Reporting on operational UK casework, the authors state that of the completed familial search investigations, "17% resulted in the identification of a relative of the true offender." They describe that when a relative is in the National DNA Database and a likelihood ratio or allele-counting methodology is combined with age, ethnic appearance and geographic filters, "the relative would be expected to occur in the candidate list" at a manageable rank. Simulation experiments cited in the paper show that there is an "approximately 78% chance that a sibling of a true offender will be in the top 100 profiles" using a likelihood-ratio based approach, but they also stress that partial profiles and degraded samples are excluded because these would produce large candidate lists and undermine reliability.

Summarising multiple empirical and simulation studies, the paper explains that familial searching is designed so that "the DNA profiles of close genetic relatives will exhibit a greater degree of genetic similarity" than unrelated individuals, which is exploited using likelihood ratios and ranked candidate lists. It reports that in real UK casework, relatives of the true offender were present in the candidate lists in a substantial proportion of cases, but that only 17% of completed investigations actually identified a relative, implying that most candidates investigated were not true relatives. The authors note that partial profiles, mixtures, or degraded samples are excluded from the service because these conditions increase the risk of generating large numbers of spurious partial matches.

In a simulation study of relative identification using forensic STR markers, the authors show that when the database contains millions of profiles, "many unrelated individuals will have genotypes that are similar enough to be considered candidate relatives" under typical likelihood-ratio thresholds. They report that while first-degree relatives can usually be distinguished from unrelated individuals with high power, "some more distant relatives and some unrelated individuals may be misclassified as closer relatives" depending on the threshold chosen, and stress that the trade-off between sensitivity and false-positive rate must be explicitly managed in familial-search policies.

This review discusses error in forensic DNA more broadly, including database matches and familial searching. The authors note that reported random match probabilities can be extremely small, but other sources of error (laboratory mistakes, sample mix-ups, database issues) can dominate: “Forensic practitioners, however, have struggled to provide realistic overall error rates, since laboratory error, contamination, and cognitive bias can increase the chance that an incorrect DNA association is reported.” The paper emphasizes that while the statistical chance of coincidental DNA matches is tiny, “the overall error rate of a DNA identification must also consider process errors, which in some case studies are orders of magnitude higher than the random match probability.” This context is relevant to assessing the ‘true’ misidentification risk in partial or familial searches.

NYU summarizes the PLOS ONE study of California’s familial DNA search method: "The study found that familial DNA searches do a good job of locating a relative if one is in the database, and conversely that a search is also unlikely to return a match that appeared to be related to the crime scene source, but in fact was not." At the same time, "the results also showed that a more distant relative in the existing DNA database could have up to a 42% chance of being incorrectly labeled as a first-degree relative of the person who left the crime scene DNA." This highlights low misidentification risk for unrelated individuals but a substantial misclassification risk among actual relatives of different degree.

This NIJ report notes that familial DNA searching (FDS) uses moderate and low stringency levels to identify partial matches. It states: "Studies examining the efficacy of FDS (including lineage testing) with statistical simulations of data generally find that the technique reliably removes non-familial matches for certain family relationship types (Bieber, Brenner, & Lazer, 2006; Ge et al., 2011; Hicks et al., 2010; Myers et al., 2011; Rohlfs, Fullerton, & Wier, 2012; Slooten & Meester, 2012)." However, it also acknowledges that "some studies have also identified the potential for false positives that exist despite the advanced abilities of the statistical FDS software (Pu & Linacre, 2008; Mueller, 2008; Reid, Lee & Lee, 2008)." Thus, misidentification risk for unrelated persons is generally reduced by confirmatory lineage testing, but false positives remain a recognized issue.

The Innocence Project’s annotated bibliography summarizes research on familial DNA searching, noting that "partial matching methods presently have a significant rate of false positives (i.e., supposed genetic relatives who, upon analysis, turn out not to be related)." It further explains that "partial matches may be uncovered either fortuitously (by chance) or deliberately" and stresses the need to model the likelihood of chance matches "in order to reduce false-positive results." The discussion highlights that false positive risk is a recognized concern in the scientific literature on partial DNA matching.

Studying California’s familial search protocol, the authors report that the method "does a good job of locating a relative if one is in the database" and is "unlikely to return a match that appears to be related to the crime-scene source, but in fact is not." However, they also find that more distant relatives can be severely misclassified: "a more distant relative in the database could have up to a 42% chance of being incorrectly labeled as a first-degree relative of the person who left the crime scene DNA." This illustrates that while some error rates may be low under specific criteria, misidentification risks in partial familial matching can be substantial for certain relationship classes.

This review article summarizes existing studies of familial DNA searching. It notes: “Empirical and simulation-based studies uniformly suggest that familial searching strategies can be tuned to achieve high power for detecting first-degree relatives with very low false positive rates among unrelated individuals, particularly when likelihood ratio methods and additional markers are used.” However, it cautions that “the likelihood of incorrectly inferring the degree of relationship, especially misclassifying distant relatives as first-degree relatives, remains non-trivial in many population scenarios,” and that “the practical impact on misidentification rates depends on database size, population structure, and the extent of confirmatory testing and traditional investigation that follow an initial partial match.”

This review explains that familial searches extend database use to identify potential relatives via partial matches. It notes that statistical frameworks and thresholds are used to control false positives, stating that simulations show relatively low false positive rates for correctly specified first-degree relationships under certain protocols, while also highlighting concerns about coincidental matches and over-interpretation of partial profiles. The authors emphasize that misidentification risk depends on database size, allele frequency assumptions, and relationship hypotheses, and they discuss both reported low error rates and critiques warning about potential misclassification and investigative burden from spurious leads.

Murphy (2010) reviews familial searching and partial match policies and notes that "partial matches" identified in routine database searches can generate investigative leads but also raise concerns about erroneous implication. The article explains that partial matching is less discriminating than full-profile matching and thus increases the number of individuals who may appear related by chance, which in turn heightens the risk of false positive investigative targets. The review emphasizes that these risks have been a central topic in policy debates over familial searching.

Discussing partial and familial DNA matches, the author notes that there are "numerous cases of lab techs who make mistakes or argue that there was a DNA match when there was none." Citing Erin Murphy’s work, the piece emphasizes that the belief that DNA is uniquely reliable can obscure the fact that "the results of a partial or familial match could have led investigators to the wrong sources — a common occurrence." The article frames familial and partial matching as particularly vulnerable to interpretive error and bias compared with straightforward full-profile matches.

A forensic-technology review in Profiles in DNA states that familial searching can increase cold hit rates and discusses error control: "Given the discrimination power afforded by current Y STR profiling kits is around 0.999, most false indirect associations can be eliminated." It concludes that "when using a Y STR match threshold, rarely would an investigator be incorrect in following up the lead provided by the DNA association." At the same time, the article stresses that software and statistical practices must be designed with known false-positive and false-negative rates and that indirect associations (partial matches) require careful evaluation to minimize misidentifications.

The NYCLU, critiquing New York’s proposed familial searching policy, argues that these methods introduce additional error: “Familial DNA searching and partial-match DNA analysis use fewer genetic markers than are used in standard testing protocols to determine a match between a forensic DNA sample and a DNA profile held in a databank… By definition, then, these techniques introduce imprecision, and the potential for error, in the analysis of forensic DNA and in the criminal investigation based upon this analysis.” It highlights the lack of empirical data: “The use of partial-match DNA techniques is not well understood; nor is there sufficient documentation to evaluate the use of familial DNA searching… A RAND Corporation report concluded that evaluating the efficacy of CODIS is difficult because ‘data are seriously lacking.’” The testimony notes that false leads can subject innocent relatives to investigation even when they are unrelated to the crime.

An Oxford University Press blog post discusses how DNA database searches have evolved to include familial and partial matches: "These searches have recently been expanded to include 'partial matches,' potentially implicating relatives of the individuals in the database." It notes that familial searching has been critiqued for turning relatives into suspects, but reframes partial matches as identifying "persons of interest" that provide leads for investigation, not definitive identifications. The author underscores that partial matches are investigative tools subject to confirmation and that they raise ethical and legal questions partly because of the possibility of false leads.

In this policy resolution, NACDL states that there is "presently insufficient data demonstrating the effectiveness of using familial DNA searching as a law enforcement tool" and warns that the practice can "target innocent individuals who have never been arrested or convicted of crimes." The document calls for empirical research to determine kinship-index thresholds and expected candidate-list sizes that "minimize the danger that innocent individuals will be subject to investigation" and recommends collecting data on "what percentage of candidates are ruled out" after additional DNA testing, reflecting concern about the rate at which initial partial matches prove to be false leads.

This book chapter argues that familial forensic identification rests on probabilistic inferences about relatedness and therefore cannot be treated as equivalent to direct DNA identification. The author notes that as databases grow, "the probability that an unrelated person will share enough alleles to be considered a candidate relative increases," which raises concerns about false leads and misidentifications. The chapter critiques policies that treat familial hits as strong evidence of guilt and emphasizes the need for cautious interpretation and corroboration to avoid wrongful implication of distant relatives or unrelated individuals.

This 2025 review on familial DNA searching notes that partial matches at moderate or low stringency can produce many coincidental matches in databases of 13–15 STR profiles, especially for close relatives. It observes that "while most true parent-child matches show strong support for their relatedness, there have been instances where unrelated individuals had test results that mistakenly suggested a parent child relationship." The paper highlights that family DNA analysis "can be problematic due to potential false matches" and warns that the method "can sometimes mistakenly link individuals to crimes based on partial DNA similarities," emphasizing the need for updated statistical techniques to reduce misinterpretation and unjust use of familial DNA evidence.

Discussing forensic practice, the article notes that "partial profiles will match up with many more people than a full profile" and cautions that "partial matches are more likely to lead to false positive identification of suspects who are already in the DNA database." It emphasizes that problems such as mixture interpretation and database over-representation of certain groups can combine with partial DNA matches to generate erroneous leads and miscarriages of justice, underlining that misidentification risks are a real concern when relying on partial DNA matches.

This educational overview defines a partial match in CODIS as occurring "when a Combined DNA Index System (CODIS) search is conducted and the results clearly show that the offender profile is not the source of the crime scene profile... but the possibility does exist that a close biological relative of the offender might be the source." It notes that partial matches are generated under moderate or low stringency search settings and that policies (referencing SWGDAM recommendations) treat them as investigative leads rather than identifications, recommending additional testing such as lineage analysis to reduce false positives from partial DNA similarity.

This 2026 article explains that familial DNA searching is used "when no direct database match is obtained" and that it "seeks to identify potential biological relatives through partial matches, most commonly first-degree relatives." The authors discuss how likelihood ratios are employed to quantify the strength of putative relationships and state that appropriate thresholds can reduce but not eliminate coincidental matches. They note that familial searches can generate false leads and that careful statistical interpretation and confirmatory analyses are essential to manage the risk of misidentification arising from partial DNA matching.

An empirical evaluation of familial searching in an operational database reports that, under the tested protocol, the method generated relatively few non-familial candidates among top-ranked matches for close relatives, suggesting low false-positive rates for first-degree relationships under controlled conditions. Nonetheless, the authors stress that performance depends on database composition, search thresholds, and follow-up testing, and they caution that their findings "should not be generalized to all familial searching implementations" or more distant relationship categories.

Forensic genetics literature generally distinguishes between two types of misidentification risk in familial searching: (1) the probability that an unrelated individual in a database will appear to be a first-degree relative of the crime scene contributor, and (2) the probability that a true but more distant relative (e.g., cousin) will be misclassified as a first-degree relative. Published simulation studies of protocols such as the California Myers et al. method report that type (1) error can be extremely low (often quoted in the range of 10−5 or lower per search under specified conditions), while type (2) error can be much higher, with some studies citing misclassification probabilities for certain distant relatives that approach or exceed tens of percent in specific populations.

Forensic scientist Dan Krane discusses observed error rates in a large DNA database when comparing profiles pairwise. He recounts an analysis of approximately 15,000 profiles: “When you did all possible pairwise comparisons of those 15,000 people… there were a whole bunch that matched at 17 out of 18 [loci]… and then a whole bunch that matched at 18 out of 18.” Krane explains that 18-of-18 matches were duplicate entries, but that 17-of-18 matches indicated mistakes in profiling: “This translated into an error rate of one in 300.” He argues that such laboratory or database error rates can overshadow the extremely small random match probabilities often cited, suggesting that real-world misidentification risk may be driven more by process errors than by the theoretical statistics alone.