Human Nature

, Volume 23, Issue 2, pp 208–217

Estimating the Prevalence of Nonpaternity in Germany


    • Institute of Experimental PsychologyHeinrich-Heine-University of Duesseldorf
  • Jochen Musch
    • Institute of Experimental PsychologyHeinrich-Heine-University of Duesseldorf
  • Juergen Enczmann
    • Institute for Transplantation Diagnostics and Cell TherapeuticsHeinrich-Heine-University of Duesseldorf Medical School
  • Johannes Fischer
    • Institute for Transplantation Diagnostics and Cell TherapeuticsHeinrich-Heine-University of Duesseldorf Medical School

DOI: 10.1007/s12110-012-9143-y

Cite this article as:
Wolf, M., Musch, J., Enczmann, J. et al. Hum Nat (2012) 23: 208. doi:10.1007/s12110-012-9143-y


The prevalence of nonpaternity in human societies is difficult to establish. To obtain a current and fairly unbiased estimate of the nonpaternity rate in Germany, we analysed a dataset consisting of 971 children and their parents in whom human leukocyte antigen (HLA) typing had been carried out in the context of bone marrow transplantation. In this sample, nine exclusions (0.93%) could be identified on the basis of more than 300 HLA-haplotypes defined by four HLA genes. Given this number of exclusions, a maximum likelihood estimate of the nonpaternity rate in the population of 0.94% was obtained with asymptotic 95% confidence limits of 0.33% and 1.55%, respectively. This result is in accordance with recent surveys as well as findings from Switzerland for a comparable sample, and it suggests that earlier estimates of the nonpaternity rate which were often in excess of 10% may have been largely exaggerated.


NonpaternityCuckoldryHuman leukocyte antigenBone marrow transplantationMaximum likelihood

Since ancient times, an asymmetry between human males and females was noted regarding the certainty of their parenthood: Unlike males, females can always be sure of their maternity (“Mama’s baby; papa’s, maybe”). For several reasons, it is of considerable interest to accurately assess the paternity rate in human societies. Cultural and social differences in nonpaternity rates cannot be investigated without determining the nonpaternity rate at the population level (Cerda-Flores et al. 1999). Moreover, the nonpaternity rate is a central variable in evolutionary theory. Many evolutionary models are based on the assumption that the nonpaternity rate is larger than zero, and that it was the constant evolutionary pressure of female infidelity in combination with paternal uncertainty that led to the evolution of an entire arsenal of anticuckoldry tactics among human males (Platek and Shackelford 2006). For example, it has been argued that male sexual jealousy, mate-guarding behavior, and differential grandparental solicitude may be psychological adaptations to the problem of paternity uncertainty (Buss et al. 1992; Buss and Shackeford 1997; Euler and Weitzel 1996). Determining the nonpaternity rate is also of importance for the analysis of the transmission of genetic disorders (Aguilar-Martinez et al. 2007; Bonaiti-Pellié et al. 1992) and for the estimation of the genetic influence on quantitative traits (Lathrop et al. 1983). This is because cases of nonpaternity represent forms of error in pedigree analysis that may have significant implications for heritability estimates; knowing the nonpaternity rate of the general population therefore allows researchers to adjust for this source of variance (Procopio 2005).

The actual rate of nonpaternity, however, is difficult to determine. Rates of about 10% and more have frequently been cited (e.g., Cohen 1977; Philipp 1973; Stewart 1989). However, MacIntyre and Sooman (1991) criticized that most of the empirical support for such figures seems to be based on hearsay, anecdotes, or unverifiable sources. Several recent reviews have therefore tried to separate myth from evidence (Anderson 2006; Bellis et al. 2005; Voracek et al. 2008). In these reviews more than 70 studies of the nonpaternity rate in different countries were collected and analyzed. Among these studies, two major groups can be distinguished. On the one hand, samples from paternity-testing laboratories have often been reported. Estimates of the nonpaternity rate based on such samples, however, are likely to be gross overestimates because these samples were drawn from a population in which fathers were already uncertain regarding their paternity. It is therefore not surprising that Bellis et al. (2005) as well as Anderson (2006) arrived at very high estimates of nonpaternity of 26.9% and 29.8%, respectively, for this type of data.

A second type of data which has also frequently been used to estimate the nonpaternity rate was mainly collected in genetic studies. However, as pointed out by Anderson (2006) and Bellis et al. (2005), the samples in these latter studies can hardly be considered to be random—and thus unbiased—either. This is because the participants in genetic and lineage studies are usually aware or informed of the fact that cases of nonpaternity may be revealed in the course of the investigation. If the participants are able to drop out, nonpaternity rates are underestimated to the extent to which families in which the social father is not a child’s biological father fail to participate. It is therefore hardly surprising that for this second type of data, Anderson (2006) arrived at very low estimates of nonpaternity of only 1.7% (excluding studies of unknown methodology) and 3.3% (including such studies). Bellis et al. (2005) as well as Voracek et al. (2008) arrived at quite similar figures of 3.7% and 3.1%, respectively, in their review of the nonpaternity rate in samples investigated for reasons other than disputed paternity (including such of unknown methodology).

While most of the above-mentioned studies are thus biased toward either low or high nonpaternity rates, they can probably still be regarded as providing upper and lower bounds for the actual prevalence of nonpaternity. There is some reason to expect, however, that the true nonpaternity rate might be closer to the lower bound provided by samples collected for genetic studies. This is because in these studies, a possible downward bias due to the selective dropping out of men who do not believe that they fathered their putative children may have been compensated for by a possible upward bias due to the unintentional inclusion of misidentified stepchildren and covert adoptions (Anderson 2006). Moreover, some of the samples collected in genetic studies seem to meet more closely the central criteria that have to be met in order to arrive at an unbiased estimate, namely, random sampling and a low dropout rate. One such study investigated virtually all newborns for a 13-month period in a hospital in London (Johnstone 1954). Among the 2,596 screened children, only 17 (or 0.7%) exclusions were found. Adjusting this figure for the low probability of exclusion (PE) resulting from the use of a rather inefficient marker system (namely, ABO blood groups) still yields a relatively low nonpaternity rate estimate of about 4% (Edwards 1957).

Another sample that met rather strict criteria of unbiasedness was investigated by Sasse et al. (1994). This study investigated a dataset consisting of 1,264 individuals from Switzerland in whom human leukocyte antigen (HLA) tissue typing had been carried out to search for bone marrow donors.

Samples drawn from a population of bone marrow transplant patients have several benefits over samples collected in genetic or lineage studies. First, there is no evidence that the nonpaternity rate is associated with diseases such as leukemia or lymphoma that are prompting screenings for appropriate bone marrow donors. Moreover, findings from several studies also suggest that the incidence of childhood cancer is not related to environmental or social factors such as place, ethnicity, socioeconomic status, and date of study (Dockerty et al. 2001; Doll 1989; IARC 1997). None of these variables can therefore bias estimates of the nonpaternity rate. The same is also true for birth order, another variable that is potentially related to nonpaternity (Schacht and Gershowitz 1963) but known to be unrelated to the risk of lymphoma or leukemia (Hemminki and Mutanen 2001).

As a second advantage of investigating such screening data, typing routines stipulate that unintentionally identified cases of nonpaternity have to remain secret and must not be communicated. Mothers and potential nonfathers are therefore not forewarned and not encouraged to drop out from the screening. Moreover, the thorough testing of both parents is important to cross-validate the HLA pattern of the patient. This guarantees that every effort is made to investigate both the mother and father of a sick individual. Additionally, the urgent need to find an appropriate donor usually leads both parents to comply with the request to undergo the typing procedure. For these reasons, among all samples that may possibly be collected to estimate nonpaternity rates, samples collected in the course of a bone marrow donor search for the purpose of HLA typing of families perhaps most closely meet the requirements of a truly random sample of the population. This conclusion, however, rests on the premise that the sample under investigation is drawn from a population in which each individual has equal access to medical care in case of a bone marrow disease. Given that there is a universal, state-supported health care system both in Germany and in Switzerland, the present sample and that of Sasse et al. (1994) meet this requirement. It is therefore remarkable that Sasse et al. (1994) observed only 8 exclusions in the 1,264 families they investigated, resulting in a maximum likelihood estimate of the nonpaternity rate in Switzerland of only 0.64%.

To summarize, the existing empirical evidence regarding the nonpaternity rate in human populations is rather mixed. Some of the better-controlled studies seem to suggest nonpaternity rates well below the often-cited rate of about 10%. To obtain rather unbiased estimates of the nonpaternity rate, it is however essential to examine families that are chosen at random. For this reason, we retrospectively examined individuals and their parents for whom HLA typing had been carried out in the context of a bone marrow donor search. Since we examined data originating from a German transplant database, our study also provided what—to the best of our knowledge—is the first fairly unbiased estimate of the nonpaternity rate in Germany, a country in which so far only three published studies have been conducted on the basis of blood group or DNA markers to estimate the nonpaternity rate. In two of these studies, especially high estimates of 50.2% and 16.8%, respectively, have been reported. The validity of these estimates is questionable, however, given that they were both based on data collected to examine cases of disputed paternity (Henke et al. 1999; Krawczak et al. 1993). The third study is only a report of a statement made during a discussion rather than a full study; in this statement, the nonpaternity rate was reported to be “at least 10%” (Ritz 1985:90). Since Ritz (1985) does not provide any information concerning the procedure and the sample that was being referred to in this statement, this third study cannot be regarded as providing a reliable estimate of the nonpaternity rate either.


Family and Marker Data

We analyzed the data that were collected for a bone marrow transplant database at a German university hospital over a span of 15 years (1993–2008). The data represented 971 individuals, their biological mother, and their presumptive father, all of whom were typed for four different HLA markers (A, B, DR, DQ). The 971 individuals were members of 454 different families. Prior to conducting our analysis, all personal data were removed from the dataset to guarantee anonymity in accordance with the medical-ethical guidelines for genetic testing in humans (WMA 2008; ZEKO 1999).

Data Treatment and Exclusion Rate

The HLA system—the major histocompatibility complex in humans—contains a large number of genes situated on chromosome 6 that are related to immune system function. As with other genetic markers, a child inherits alleles from this genetic region (HLA haplotypes) in equal parts from both parents. Mendelian inconsistencies occur, however, if the presumed father is not the biological father.

Observable inconsistencies in the HLA-typing of a family are usually scrutinized closely and followed up by a retest of the presumed father to verify the result. It is therefore highly probable that the present data set contains very few, if any, typing errors. Importantly, during the entire data collection phase, not a single case occurred where the mother or presumed father of an individual refused to be typed.

Since every known marker system is imperfect in that it can detect cases of genetic inconsistency with a certain probability only, the observable number of excluded individuals is an underestimate of the true number of children of cuckoldry in a population. To arrive at a more exact estimate, it is necessary to take the inefficiency of a marker system in detecting inconsistencies into account. The probability of exclusion—defined as the probability of excluding a random male from the population as the biological father of a given child—is a function of the marker system and the frequency distribution of the relevant alleles in the respective population. The individuals in our sample were tested for four different, highly polymorphic HLA genes resulting in more than 300 different haplotypes. A nearly perfect probability of detecting cases of nonpaternity was thus guaranteed (Chakravarti and Li 1983). Assuming that each HLA haplotype has equal frequency, Sasse et al. (1994) computed a total probability of exclusion for the HLA marker system of 0.99. Rather than relying on this very high probability of exclusion, we performed a sensitivity analysis to adjust for any potential bias resulting from possible nonrandom associations between HLA markers, which may lead to an overestimate of the probability of exclusion (Bodmer and Bodmer 1978). To this end, we also computed nonpaternity rate estimates assuming probabilities of exclusions of only 0.95 and 0.90. Using these very low and therefore conservative alternative estimates of the probability of exclusion—in addition to the probability of exclusion of 0.99 that was used by Sasse et al. (1994)—allowed us to determine a reliable upper bound for the nonpaternity rate in Germany.


Among the 971 individuals in the data set, nine (or 0.93%) exclusions were found. Based on this figure and in accordance with the procedure detailed in Sasse et al. (1994), we calculated two different maximum likelihood (ML) estimates of the nonpaternity rate. The first estimate was based on the assumption that cases of nonpaternity occur independently and with the same probability for each child, whereas the second estimate was based on the more relaxed—and possibly more realistic—assumption that the chance for a case of nonpaternity is higher in some families than in others. The formulas for both estimates are given in Sasse et al. (1994).

Table 1 shows both ML estimates (assuming constant or varying nonpaternity rates) separately for three different probabilities of exclusion (0.90, 0.95, and 0.99). As can be seen from the table, assuming a varying rather than constant nonpaternity rate led to only minimally larger 95% confidence intervals (CI). In a similar vein, ML estimates of the nonpaternity rate varied only very little, between 0.94% and 1.03%, as a function of the presumed probability of exclusion. Thus, taken together, regardless of the underlying assumptions the nonpaternity rate in the present sample is estimated to be only marginally higher than the figure reported by Sasse et al. (1994) for a comparable sample from Switzerland (0.64%; CI [95%] = 0.19–1.09).
Table 1

Estimates of the nonpaternity rate for three different probabilities of exclusion


Probability of Exclusion (%)




Constant nonpaternity rate

MLa estimate of nonpaternity rate




CI (95%)b




Varying nonpaternity rate

MLa estimate of nonpaternity rate




CI (95%)b




aML = maximum likelihood

bCI = confidence intervals; confidence intervals are based on standard errors calculated as the square roots of the asymptotic variances


Based on a sample of father-mother-child triplets collected in a search for bone marrow donors, we were able to estimate the prevalence of nonpaternity in Germany at roughly 1%. This is a somewhat lower estimate than those provided in recent surveys of Anderson (2006), Bellis et al. (2005), and Voracek et al. (2008). It agrees well with the nonpaternity estimate for a comparable Swiss sample (Sasse et al. 1994), however. Although two previous studies of the nonpaternity rate in Germany arrived at considerably higher estimates of 50.2% (Henke et al. 1999) and 16.8% (Krawczak et al. 1993), those studies analyzed mothers and fathers trying to resolve paternity disputes. They were therefore far from being representative of the German population at large. The present data therefore fill a gap by providing what seems to be the first fairly unbiased estimate of the nonpaternity rate in Germany. The large discrepancy between the present and the two diverging earlier estimates emphasizes the importance of using unbiased samples when trying to obtain valid estimates of the nonpaternity rate.

In spite of its virtues, the present sample is not a truly random sample of the population. Children whose fathers died before the data collection took place, or who were conceived through anonymous sperm donors, were of course not eligible for sampling. Our sample might therefore perhaps best be described as an unbiased sample of bone marrow transplant patients seeking treatment at a university hospital. However, given that the catchment area of the hospital included not only the urban center of a large German city but also the surrounding rural communities, and provided that there is universal access to the state-supported health system for every German regardless of his or her socioeconomic status, the present sample may still be considered quite representative of the general population.

One might wonder whether there are alternative explanations for even the small number of genetic inconsistencies we observed in our data set. It is indeed possible that abnormal inheritance patterns such as uniparental disomy—a person receiving two copies of a chromosome from one parent and no copies from the other parent—led to apparent inconsistencies, resulting in an overestimate of the nonpaternity rate. Such conditions are exceedingly rare, however, and were not detected in any of the cases in the present study.

Inconsistencies may also be the result of mutations in the HLA genes. Although we cannot rule out the possibility that mutations led to false paternity exclusions in our sample, we consider this to be highly unlikely for two reasons. First, compared with other polymorphic markers the mutation rate of HLA genes is generally quite low (Jobling et al. 2004). Second, in the present dataset there were no hints indicating that inconsistencies resulted from a mutation. If there was an inconsistency, it was not limited to a single gene and therefore a possible result of mutation; rather, in these cases the haplotype combination of the excluded child differed substantially from that of his or her presumed father.

Instead of factors leading to an overestimate of the nonpaternity rate, some potential reasons for underestimating the number of cuckolds exist in the present study. Although the highly polymorphous HLA markers usually allow detection of inconsistencies with a very high probability, inconsistencies are more likely to go undetected when the presumed and the biological father are related to each other (e.g., brothers or cousins). It could be argued that if a large proportion of all cases of infidelity occur in such close kinship, our assumed probability of exclusion levels of 0.99, 0.95, or 0.90 are overly optimistic and may have led to an underestimate of the nonpaternity rate. However, we are not aware of any evidence supporting the notion that brothers or cousins are more likely to become the biological father of a child of cuckoldry than a person outside the family. Nevertheless, to address this concern, we conducted an additional sensitivity analysis by hypothetically assuming an extremely low probability of exclusion of only 0.50. Even under this most conservative assumption, the proportion of children of cuckoldry was estimated at only 1.85%, with 95% confidence limits of 0.64% and 3.06%. Thus, even under most unrealistic worst-case assumptions, it seems safe to conclude that the nonpaternity rate in Germany does not surpass 3%.

From an anthropological point of view, it would be interesting to determine whether there are any geographic differences with regard to the nonpaternity rate. The present estimate falls close to other presumably unbiased estimates of the nonpaternity rate, ranging between 0.64% and 1.35%, that were obtained in studies from Switzerland (Sasse et al. 1994), England (Brock and Shrimpton 1991; Sykes and Irven 2000), and the United States (Broman 1999). Thus, at least in current Western industrialized societies, the nonpaternity rate seems to be rather low. However, socioeconomic variables may be influencing this rate. Both Cerda-Flores et al. (1999) and Schacht and Gershowitz (1963) found nonpaternity rates to be a function of socioeconomic status: families with a high socioeconomic background tended to have fewer such children.

One might speculate whether the nonpaternity rates in ancestral human populations were more close to the present estimate or to the meta-analytic estimates for current populations provided by Anderson (2006), Bellis et al. (2005), and Voracek et al. (2008). To the extent that human sexual behavior was shaped in the past, behavioral data from today may provide an alternative means of arriving at a plausible estimate of the nonpaternity rate. One class of studies tried to estimate the number of cuckolded men by assessing sexual behavior patterns via questionnaire and arrived at considerably higher estimates than the present study. In a survey of 2,708 British women by Bellis and Baker (1990), 6% reported extrapair copulations, and 1.8% even had double matings within the previous 28 days. Based on the probability of conception by day of menstrual cycle, the authors calculated an estimate of the nonpaternity rate of about 7–14%, with about half being the result of double mating. In a more representative study of sexual behavior in America, Laumann et al. (1994) also found that about 16% of 3,500 randomly selected Americans reported more than one concurrent sexual relationship in the preceding year, and 24% of men and 15% of women reported extramarital affairs over the duration of their marriage. Similar figures have also been reported by Wiederman (1997).

Gaulin et al. (1997; cf. Hoier et al. 2001) proposed yet another approach to estimate nonpaternity rates based on a behavioral measure. They asked participants to rate the amount of solicitude they had received throughout their childhood from their matri- and their patrilateral aunts and uncles, respectively. Assuming that relatives will invest in offspring in direct proportion to their genetic relatedness, which in turn is a function of their certainty of biological parenthood, Gaulin et al. (1997) tried to estimate the nonpaternity rate via kin investment data. Based on the extent to which more solicitude was provided by matrilineal aunts and uncles, who can be more sure of being genetically related to their nephews and nieces than their patrilineal counterparts, Gaulin et al. (1997) arrived at an estimate of the nonpaternity rate of between 13% and 20%. Using the same approach in a German sample, Hoier et al. (2001) estimated the nonpaternity rate at about 8–32%. Based on these data resulting from competing approaches, and provided that ancestral populations did not have access to contraceptives, one might conjecture that nonpaternity rates may have been considerably higher in the human past. In present-day Germany, however, a child of cuckoldry seems to be a rather rare event.

Copyright information

© Springer Science + Business Media, LLC 2012