Identifying the Mother of William Franklin: A Proposal for a Genetic-Genealogical and Documentary Research Program

Abstract

The maternity of William Franklin (c. 1730–1813), illegitimate son of Benjamin Franklin and royal governor of New Jersey, has never been resolved. No contemporary document names the woman; no family tradition survives naming her; the historiography has produced candidates without producing evidence. This paper argues that the question is now tractable in a way it has not previously been, because a documented and genetically tested descent from William Franklin exists, and because autosomal segment analysis can, under favorable conditions, distinguish maternal from paternal ancestry in the fifth to eighth generation. The paper proposes a three-track program: reconstruction of a candidate field from Philadelphia’s documentary record on stated criteria; assembly of a Franklin-descendant genetic reference panel; and subtraction of the known paternal contribution to isolate maternal-side segments for matching against candidate lineages. It sets out the design in detail, states the statistical limits honestly, and specifies in advance what results would and would not license. The realistic best outcome is not a name but a substantially narrowed field with one or two lineages ranked above the rest — and the paper argues that this outcome is worth pursuing, both for its own sake and because the woman in question has been the object of three centuries of remarkably incurious speculation.

1. The Problem

1.1 What is known

William Franklin was born in Philadelphia, most probably between late 1728 and 1731, with the balance of scholarly opinion favoring 1730 or early 1731 (Skemp, 1990). Benjamin Franklin acknowledged him without qualification, raised him in his household, and secured his education, his military commission, his clerkship of the Pennsylvania Assembly, and eventually the royal governorship of New Jersey. Deborah Read Franklin, who entered a common-law union with Franklin on 1 September 1730, raised William as her own; the household’s other children, Francis Folger (1732–1736) and Sarah (1743–1808), were hers.

That is the extent of secure knowledge. Franklin never named the mother in any surviving letter, memorandum, or account entry, and the Autobiography — expansive on his own errata, including the “hard-to-be-governed passion of youth” that drove him toward “intrigues with low women” — passes over her in silence (Franklin, 1791/1986; Lemay & Zall, 1986).

1.2 What has been alleged

Three lines of speculation recur.

Deborah Read as mother. Advanced on the ground that the marriage’s timing and William’s silent absorption into the household would be neatly explained. Against it: Deborah’s marked coldness toward William, attested across decades of family correspondence, is difficult to reconcile with maternity; the arrangement offered no advantage over simply acknowledging a pre-nuptial child, which was socially manageable; and the hypothesis explains the concealment less well than concealment of a different woman does (Skemp, 1990; Lopez & Herbert, 1975).

Barbara, the household servant. Originating with a 1763 pamphlet by Franklin’s political enemies alleging that he had kept a servant named Barbara as William’s mother, worked her without wages, and buried her without a stone. The pamphlet is a campaign document in the Pennsylvania proprietary fight, is unsupported by any independent record, and is exactly the sort of claim such literature manufactured (Lopez & Herbert, 1975; Skemp, 1990).

An unnamed woman of low station. The consensus position, which is not a hypothesis but a restatement of the silence — grounded in Franklin’s own account of his conduct in this period and in the absence of any indication that the mother had family able to compel acknowledgment.

1.3 Why the silence is itself evidence

Franklin was a compulsive record-keeper. He preserved account books, letterbooks, and memoranda across seven decades, and his papers constitute one of the fullest personal archives of the eighteenth century (Labaree et al., 1959–). The absence of the mother from that archive is therefore not a random gap. It is either deliberate suppression, or destruction by Franklin or his heirs, or — least likely — a woman with whom he had no financial or correspondent relationship at any point.

Deliberate suppression narrows the field in a useful direction. A woman of standing would have generated pressure, negotiation, and probably a settlement, all of which leave traces. A woman of no standing generates nothing to suppress. The silence is most consistent with a servant, an indentured woman, a laboring woman, or a widow without kin — that is, with the least documented stratum of colonial Philadelphia, which is precisely the stratum where a genetic approach has the greatest comparative advantage over an archival one.

2. The Enabling Condition

2.1 A tested descendant line exists

The 1998 Jefferson–Hemings study included, as part of its comparison set, Y-chromosome data from a documented male-line descendant of William Franklin (Foster et al., 1998). This establishes two things: that the descent is documented to the standard required for a published study, and that the line has already accepted participation in genetic-historical research.

The line runs through William Temple Franklin (1760–1823), William’s own illegitimate son, whose descendants are traceable. Additional descent may exist through other channels; establishing the full descendant population is the first task of Track B below.

2.2 Why the Y chromosome is the wrong tool here

The 1998 study used Y-STR haplotypes because the question was paternal. For maternity, the Y chromosome is useless: it descends father to son and carries no maternal information whatever.

Mitochondrial DNA would be ideal — it descends mother to child unbroken — but it requires an unbroken female line from the woman in question to a living person. William Franklin’s only known child was male. Unless an undocumented daughter is discovered with surviving matrilineal descent, mtDNA is unavailable. This is the central structural obstacle, and it is what pushes the problem onto autosomal ground.

2.3 The autosomal opportunity and its cost

Autosomal DNA recombines each generation and descends from all ancestral lines. A living descendant of William Franklin therefore carries some genetic contribution from William’s mother — but diluted by roughly a factor of two per generation and, critically, subject to stochastic loss.

At eight generations removed, expected contribution from any one ancestor is about 0.39 percent of the genome, roughly 27 centimorgans. But the expectation understates the risk: beyond about six generations, an appreciable fraction of descendants inherit nothing from a given ancestor. Simulation work indicates that at seven generations roughly a tenth of descendants carry no detectable segment from a specific ancestor, and the proportion climbs steeply thereafter (Speed & Balding, 2015; Ball et al., 2016). By ten generations the majority of descendants carry nothing from most of their ancestors at that depth.

This has a direct design consequence: the program cannot rest on any single descendant. It requires a panel, because segments lost in one line survive in another, and it requires the panel to be as genealogically shallow as the surviving descent permits.

3. Track A: Constructing the Documentary Candidate Field

Genetic matching requires candidates to match against. A match to an unnamed segment tells us nothing; a match to a segment shared with the documented descendants of a named Philadelphia woman tells us a great deal. Track A builds the target list.

3.1 Inclusion criteria

A woman enters the field if she satisfies all of:

Presence. Residence in or near Philadelphia between mid-1728 and mid-1731. Franklin returned from London in October 1726 and, apart from short absences, was continuously in the city.
Age. Plausibly between fifteen and thirty-five at the relevant date.
Status. Free, indentured, or servant. Enslaved women are retained in the field but flagged separately (see 3.4).
Opportunity. Any documented connection to the printing trade, to Franklin’s residence and shop on Market Street, to the Junto’s membership households, to the boarding house of the Read family, or to the taverns and coffeehouses of the High Street corridor.
Absence of disqualifier. No documented residence elsewhere, marriage generating a competing paternity, or death predating conception.

3.2 Source base

The Philadelphia record for this period is thinner than for later decades but not empty:

Philadelphia County Court of Quarter Sessions records, including bastardy and maintenance proceedings — the single richest source, since the Poor Law framework required naming the reputed father, and a failure to appear where one would expect an appearance is itself informative.
Indenture and servant registers, and runaway servant advertisements in the Pennsylvania Gazette (which Franklin himself printed from 1729).
Christ Church and Gloria Dei parish registers; Philadelphia Monthly Meeting records.
Tax and constables’ returns, and the 1734 and later assessments.
Overseers of the Poor accounts.
Franklin’s own shop ledgers and account books, held in the Franklin Papers, for names appearing and then vanishing (Labaree et al., 1959–).

3.3 What the bastardy records may show

A woman who bore a child out of wedlock and named no father, or named a father who then quietly settled, is the profile of interest. Pennsylvania’s procedure required the mother to name the reputed father, and settlements were sometimes recorded in abbreviated form. A search targeted on 1729–1731 filings, cross-referenced against subsequent disappearance from the record, will produce a manageable list. Franklin’s name appearing in such a filing would of course end the inquiry; its absence is expected and consistent with a private settlement.

3.4 Handling the possibility of enslaved or Black maternity

The field should not be constructed on an assumption of white maternity. Franklin owned enslaved persons at various points and Philadelphia’s Black population, both enslaved and free, was appreciable. Contemporary allegation on this point exists but is embedded in hostile political literature and cannot be taken at face value in either direction.

The methodological point is that this hypothesis is testable by the proposed design and testable comparatively cheaply, since admixture analysis on the descendant panel would register African ancestry at the expected fraction if present. A negative result on that analysis eliminates a live possibility; a positive result would redirect the documentary search entirely. Either outcome is a gain. The hypothesis should be entered into the design explicitly rather than left as innuendo, which is how it has circulated for two and a half centuries.

3.5 Expected field size

A realistic expectation is a primary field of perhaps thirty to eighty named women meeting all criteria, with a secondary field of several hundred meeting presence and age but lacking any opportunity connection. Only the primary field is genetically tractable, and only that subset of it whose own descendants can be traced to living tested individuals — likely a third or fewer.

This is the program’s binding constraint, and it should be stated at the outset: most of the candidate field will not be testable, because most eighteenth-century laboring women left no traceable descent. The program can therefore confirm a candidate but can rarely exclude one.

4. Track B: The Franklin Descendant Panel

4.1 Composition

Three groups are needed:

Group 1 — William Franklin descendants. All traceable living descendants through William Temple Franklin and any other line. These carry the target maternal segments. Genealogical depth from William is likely eight to ten generations, placing the recovery probability per individual in the range where panel size matters critically.

Group 2 — Benjamin Franklin descendants not through William. Descendants through Sarah Franklin Bache, whose issue is numerous and well documented. These carry Benjamin’s contribution but not William’s mother’s. Group 2 is the subtraction key and is the design’s essential element.

Group 3 — Josiah Franklin and Abiah Folger collateral descendants. Benjamin’s siblings’ lines, which carry the Franklin and Folger ancestral segments independent of Benjamin. These strengthen paternal-side attribution by identifying which segments belong to the Franklin and Folger stock generally rather than to Benjamin specifically.

4.2 Panel size

Given per-individual recovery probability at eight to ten generations, a panel of fewer than twenty Group 1 members is unlikely to be adequate. Thirty to fifty is the realistic target, with priority given to the genealogically shallowest available descents and to representation from as many distinct branch points as possible — segments lost in one branch may survive in a cousin branch, so branch diversity matters more than raw count.

4.3 Platform and consent

Whole-genome sequencing or dense SNP genotyping (600k+) is preferable to consumer arrays, though consumer-database matching will be indispensable for Track C and the panel should be represented in the major matching databases where participants consent.

Consent must be explicit, informed, and specific to historical research, with participants advised that results may reveal unexpected relationships within their own recent ancestry — a routine and consequential hazard of genetic genealogy. Ethical frameworks developed for genetic-genealogical historical research and for forensic investigative genetic genealogy provide applicable models (Kennett, 2019; Erlich et al., 2018; Greytak et al., 2019).

5. Track C: Subtraction and Matching

5.1 The logic

Every autosomal segment in a Group 1 member descends either from Benjamin Franklin’s ancestry, from William’s mother, or from the many ancestors entering the line after William. The analytical task is to isolate the second category.

Step 1 — Build the paternal reference. From Groups 2 and 3, construct the set of segments attributable to Benjamin Franklin’s ancestry (Franklin and Folger stock). Any segment appearing in Group 1 and also in Group 2 or 3 at the appropriate identity-by-descent threshold is paternal and is set aside.

Step 2 — Phase and subtract post-William ancestry. Each Group 1 member’s genealogy after William Franklin is documented. Where that documented ancestry is itself represented in matching databases, its contribution is identified and removed. This is laborious and imperfect; it is also the step that determines whether the residual is interpretable.

Step 3 — Identify the residual. What remains after Steps 1 and 2 in any Group 1 member is a candidate maternal segment. Segments appearing in the residual of multiple Group 1 members from distinct branches, at the same chromosomal location, are the strongest candidates, since independent branch survival is far more consistent with descent from William’s mother than with coincidence.

Step 4 — Match outward. Query the surviving residual segments against the major matching databases and against Track A candidate lineages specifically. A residual segment shared with documented descendants of a Track A candidate constitutes the target finding.

Step 5 — Triangulate. A single match is suggestive only. Confirmation requires the same segment shared among multiple Group 1 members and multiple candidate-lineage descendants, at a location and length consistent with the genealogical distance.

5.2 Statistical thresholds

Segment length thresholds must be set conservatively. Short segments below about 7 cM are heavily contaminated by false positives and by population-level background sharing; below 5 cM the signal is largely noise (Durand et al., 2014; Ball et al., 2016). At the genealogical distance in question, true shared segments will often fall in exactly this problematic range, which is the central technical difficulty of the whole design.

The mitigation is triangulation and branch independence rather than threshold relaxation. A 6 cM segment appearing in four Group 1 members from three separate branches and in two descendants of a single Track A candidate is meaningful; a 12 cM segment appearing once is not.

Endogamy is a further complication. Colonial Philadelphia was not severely endogamous, but colonial New England — from which the Folger line derives — was appreciably so, which inflates background sharing on the paternal side and makes Step 1 more important rather than less (Ball et al., 2016).

6. Honest Statement of Limits

Any proposal of this kind should state its failure modes before its promises.

The identification may be impossible in principle. If William’s mother’s own line died out, or produced no traceable descent, no amount of segment analysis will find her, because there is nothing to match against. Given the demographic profile of the likely candidate — a laboring or servant woman who may well have had no other children — this outcome is entirely plausible and may be the most probable single outcome.

Recovery may fail stochastically. At ten generations, even a panel of forty may retain very little of the target ancestry. The program should model expected recovery from the actual genealogical depths before recruitment, and should be prepared to conclude that the panel is inadequate.

Subtraction may not resolve. Step 2 requires documenting and genetically characterizing eight to ten generations of intervening ancestry for every panel member. Incomplete documentation in any line leaves unattributable residual that cannot be distinguished from the target.

Positive findings will be probabilistic, not certain. The realistic ceiling is a ranked field: one or two lineages substantially more consistent with the data than the rest. This is a genuine advance over the present state, in which the field is unbounded, but it is not a name established beyond dispute, and it should never be reported as one.

Negative findings will be weakly informative. Failure to match a candidate lineage is consistent with that candidate being wrong, with segment loss, with that lineage’s descendants being untested, or with threshold effects. Exclusion is much harder than inclusion here, and the program should not claim exclusions it cannot support.

Non-paternity events in either the Franklin descent or candidate descents will corrupt the analysis at rates of roughly one to two percent per generation, compounding across eight to ten generations to a non-trivial probability in any given line. Multiple independent branches are the only defense.

7. Sequencing and Resources

Phase 1 (12–18 months). Genealogical reconstruction of the full William Franklin descendant population; documentary construction of the Track A candidate field; simulation modeling of expected recovery given actual descent depths. Phase 1 is inexpensive and determines whether Phases 2 and 3 are justified. The program should be prepared to stop here if modeling indicates recovery below a pre-specified threshold.

Phase 2 (18–24 months). Recruitment and consent; sequencing of Groups 1, 2, and 3; construction of the paternal reference; admixture analysis addressing §3.4.

Phase 3 (24–36 months). Subtraction, residual identification, database matching, triangulation, and targeted testing of candidate-lineage descendants identified in Phase 1.

Institutional partners with existing holdings and standing are the American Philosophical Society, Yale’s Franklin Papers project, the Historical Society of Pennsylvania, and the Philadelphia City Archives. Any of these could host Phase 1 at modest cost.

7.1 Pre-registration

The analytical thresholds, triangulation requirements, and interpretive criteria should be fixed and published before the residual segments are examined. The alternative — setting thresholds after seeing which candidate a permissive threshold would favor — is how genetic-genealogical claims about historical figures have most often gone wrong, and the field’s credibility has already been strained by confident announcements resting on post hoc criteria.

8. Why This Is Worth Doing

Three arguments, in ascending order of weight.

Methodological. The design generalizes. The same subtraction logic applies to any historical maternity or paternity question with a documented descendant population and a constructible candidate field. Establishing whether it works, and characterizing where it fails, has value independent of this particular result.

Historiographical. William Franklin’s estrangement from his father — the loyalist governor and the revolutionary printer, permanently sundered by 1776 — is among the period’s most examined family ruptures, and the maternal half of his ancestry has been entirely absent from the analysis. Whether that absence mattered to William, and whether the mother’s identity or status bore on Deborah Franklin’s documented coldness, are questions that cannot even be posed without a candidate.

Historical justice. This is the argument that should be stated plainly. Benjamin Franklin’s illegitimate paternity has been accepted, discussed, and psychologized for two and a half centuries on nothing more than his own acknowledgment. The woman who bore the child has attracted, in the same period, a handful of paragraphs and one hostile pamphlet. The asymmetry is not accidental: it reflects a durable convention by which the acknowledged father becomes a subject of biography and the unacknowledged mother becomes a gap in the record. Comparative work on founding-era paternity questions shows the pattern repeating wherever the woman lacked standing.

The convention is not, at this point, a necessity. The tools exist to test it. Whether they succeed in this case is uncertain; that the attempt is owed is not.

References

Ball, C. A., Barber, M. J., Byrnes, J., Carbonetto, P., Chahine, K. G., Curtis, R. E., Granka, J. M., Han, E., Hong, E. L., Kermany, A. R., Noto, K., Ohla, J., Rañola, J. M., Sotoodeh, M., Wang, Y., & Sturk-Andreaggi, K. (2016). AncestryDNA matching white paper: Discovering genetic matches across a massive, expanding genetic database. Ancestry.com DNA.

Bettinger, B. T. (2019). The family tree guide to DNA testing and genetic genealogy (2nd ed.). Family Tree Books.

Bettinger, B. T., & Wayne, D. P. (2016). Genetic genealogy in practice. National Genealogical Society.

Browning, S. R., & Browning, B. L. (2012). Identity by descent between distant relatives: Detection and applications. Annual Review of Genetics, 46, 617–633. https://doi.org/10.1146/annurev-genet-110711-155534

Durand, E. Y., Eriksson, N., & McLean, C. Y. (2014). Reducing pervasive false-positive identical-by-descent segments detected by large-scale pedigree analysis. Molecular Biology and Evolution, 31(8), 2212–2222. https://doi.org/10.1093/molbev/msu151

Erlich, Y., Shor, T., Pe’er, I., & Carmi, S. (2018). Identity inference of genomic data using long-range familial searches. Science, 362(6415), 690–694. https://doi.org/10.1126/science.aau4832

Foster, E. A., Jobling, M. A., Taylor, P. G., Donnelly, P., de Knijff, P., Mieremet, R., Zerjal, T., & Tyler-Smith, C. (1998). Jefferson fathered slave’s last child. Nature, 396(6706), 27–28. https://doi.org/10.1038/23835

Franklin, B. (1986). The autobiography of Benjamin Franklin (J. A. L. Lemay & P. M. Zall, Eds.). W. W. Norton. (Original work composed 1771–1790)

Greytak, E. M., Moore, C., & Armentrout, S. L. (2019). Genetic genealogy for cold case and active investigations. Forensic Science International, 299, 103–113. https://doi.org/10.1016/j.forsciint.2019.03.039

Henry, N. S. (1935). Franklin’s illegitimate son. Pennsylvania Magazine of History and Biography, 59(1), 35–48.

Kennett, D. (2019). Using genetic genealogy databases in missing persons cases and to develop suspect leads in violent crimes. Forensic Science International, 301, 107–117. https://doi.org/10.1016/j.forsciint.2019.05.016

Labaree, L. W., Willcox, W. B., Oberg, B. B., et al. (Eds.). (1959–). The papers of Benjamin Franklin. Yale University Press.

Lemay, J. A. L. (2006). The life of Benjamin Franklin, Volume 1: Journalist, 1706–1730. University of Pennsylvania Press.

Lopez, C.-A., & Herbert, E. W. (1975). The private Franklin: The man and his family. W. W. Norton.

Nash, G. B. (1988). Forging freedom: The formation of Philadelphia’s Black community, 1720–1840. Harvard University Press.

Salinger, S. V. (1987). “To serve well and faithfully”: Labor and indentured servants in Pennsylvania, 1682–1800. Cambridge University Press.

Skemp, S. L. (1990). William Franklin: Son of a patriot, servant of a king. Oxford University Press.

Skemp, S. L. (1994). Benjamin and William Franklin: Father and son, patriot and loyalist. Bedford Books.

Smith, B. G. (1990). The “lower sort”: Philadelphia’s laboring people, 1750–1800. Cornell University Press.

Speed, D., & Balding, D. J. (2015). Relatedness in the post-genomic era: Is it still useful? Nature Reviews Genetics, 16(1), 33–44. https://doi.org/10.1038/nrg3821

Wulf, K. (2000). Not all wives: Women of colonial Philadelphia. Cornell University Press.